Rigid Body Dynamics, A Lagrangian Approach

Advances in Mechanics and Mathematics
45
Hamad M. Yehia
Rigid Body
Dynamics
A Lagrangian Approach
@seismicisolation
@seismicisolation
Volume 45
Series Editors
David Gao, Federation University Australia
Tudor Ratiu, Shanghai Jiao Tong University
Advisory Board
Anthony Bloch, University of Michigan
John Gough, Aberystwyth University
Darryl D. Holm, Imperial College London
Peter Olver, University of Minnesota
Juan-Pablo Ortega, University of St. Gallen
Genevieve Raugel, CNRS and University Paris-Sud
Jan Philip Solovej, University of Copenhagen
Michael Zgurovsky, Igor Sikorsky Kyiv Polytechnic Institute
Jun Zhang, University of Michigan
Enrique Zuazua, Universidad Autónoma de Madrid and DeustoTech
@seismicisolation
@seismicisolation
More information about this series at https://link.springer.com/bookseries/5613
@seismicisolation
@seismicisolation
Hamad M. Yehia
Rigid Body Dynamics

A Lagrangian Approach
@seismicisolation
@seismicisolation
Hamad M. Yehia (Deceased)
Department of Mathematics
Faculty of Science
Mansoura University
Mansoura, Egypt
ISSN 1571-8689 ISSN 1876-9896 (electronic)

ISBN 978-3-030-96335-4 ISBN 978-3-030-96336-1 (eBook)
https://doi.org/10.1007/978-3-030-96336-1
Mathematics Subject Classification: 70-XX, 70Exx, 70Hxx
© Springer Nature Switzerland AG 2022

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilms or in any other physical way, and transmission or information storage and
retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known
or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
This book is published under the imprint Birkhäuser, www.birkhauser-science.com by the registered
company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
@seismicisolation
@seismicisolation
This book is dedicated to my family:
Nagwa, Hani and Magid
and to the memory of
my teacher and friend Vladimir G. Dumin
@seismicisolation
@seismicisolation
Introduction
Rigid body dynamics is one of the oldest and most challenging subjects in
classical mechanics. It was initiated by Leonhard Euler, who formulated the equations
of motion for a general (asymmetric) torque-free body and obtained the first inte-
grable, in quadratures, case known after his name. With the efforts of D’Alembert,
Poinsot, Lagrange and Poisson, the equations of motion of a body about a fixed
point under the action of forces were written in their present form known as the
Euler–Poisson equations: a system of six first-order differential equations for which
three integrals are known. Lagrange found the second integrable case, the case of a
heavy axi-symmetric body, usually named as Lagrange’s top. In both cases of Euler
and Lagrange, an integral of motion followed from general principles of mechanics,
constancy of the angular momentum in the first and due to the cyclic angle of rotation
about the axis of symmetry in the second. An important moment was that in both
cases, the equations of motion were solved to the end and the solution expressed
through elliptic functions, invented by Jacobi, and certain integrals involving them.
The search for integrable cases continued, but, although the problem attracted the
attention of several eminent mathematicians, the search did not lead to any other
cases. A whole century later, Sofia Kowalevski found a new integrable case of the
heavy rigid body. That was not in virtue of a physical conservation principle, but
using a purely mathematical condition: all solutions of the equations of motion
should have only poles as their singularities as functions of time in the complex
t-plane. This property is satisfied by the solutions in the two known integrable cases
of Euler and Lagrange, being expressible in terms of elliptic functions of time. Having
isolated three cases of this type, two cases of Euler and Lagrange and a new third
one, Kowalevski tried and found the complementary integral in the third case. That
integral turned out to be the first instance ever of a polynomial integral of degree
four in the dynamical variables in a dynamical problem. Kowalevski also reduced the
problem to quadratures and expressed all the dynamical variables in terms of hyper-
elliptic functions of time, which are far more complicated than elliptic functions, but
share with them the property of having only poles as singular points in the complex
plane.
vii
@seismicisolation
@seismicisolation
viii Introduction
From this point on, instead of searching for new integrable cases, researchers
directed their efforts to find methods to prove non-integrability, non-existence of a
fourth integral of motion, in addition to the elementary three general ones, which
is algebraic (Liouville, Husson and Burgatti) or even single-valued (Poincaré). This
trend reached its perfection after the KAM (Kolmogorov–Arnold–Mozer) theory
for integrable Hamiltonian systems was established. In the last few decades, new
concepts were applied to rigid body dynamics in the works of several authors, among
which those of Kozlov and Ziglin play a distinguished role.
The last quarter of the nineteenth century was a golden era for rigid body
dynamics. More complicated problems including certain generalizations of the clas-
sical problem were investigated. The integrable case of the motion of a body carrying
a symmetric rotor (the gyrostat) was found by Joukovsky and Volterra. Brun found
integrals of motion for the motion of a body acted upon by asymmetric Newtonian
force of attraction. The equations of motion of a body in a liquid were constructed
by Kirchhoff and integrable cases of that problem are associated with the names of
Clebsch, Steklov and Lyapunov. The same period is also characterized by the appear-
ance of several cases of exact particular solutions of the classical problem, which
were obtained by Staude, Hess, Goryachev, Chaplygin, Kowalewsky, Bobylev and
Steklov. Those cases constitute more than half of the exact solutions of the classical
problem, known to us to date.
The next half-century or so has elapsed without significant advancements in prob-
lems of rigid body dynamics as concerns integrable cases. The second half of the
twentieth century, on the other hand, witnessed a renaissance of the subject. Interest
has grown in integrable problems in general and, in particular, in those of rigid body
dynamics. Several important results were obtained, including new exact solutions of
the classical problem and the problem of motion of a heavy gyrostat. New problems
emerged and underwent intensive investigation. One of them was that of motion of a
body acted upon by more general conservative potential and gyroscopic forces. The
last four chapters of this book present mostly innovations brought into the subject in
the past few decades, in which the author had some substantial contributions.
The idea of writing this book emerged more than a decade ago. It was delayed so
long due to the social and political upheaval that arose in Egypt at the time and had a
direct impact on every aspect of life. The original motivation was two-fold, first and
foremost, there was a need for a new survey on the subject of rigid body dynamics.
The core of such a survey should be classification and a complete up-to-date account
of all the known but scattered in the literature integrable cases and particular solutions
of the diverse problems treated within this subject.
Only in integrable cases can one study the motion in the whole phase space and
draw conclusions about the behaviour of the mechanical system over an infinite
interval of time. It will be evident as we go through the book that integrable cases are
a rarity in problems of rigid body dynamics, and in some problems, there are even
proofs that no more integrable cases are there to be found in the future. This situation
ensures the high importance of every integrable case and justifies that each one is
recorded under the name of its discoverer. Also of great importance are particular
exact solutions, obtained only under certain conditions on the initial state of motion
@seismicisolation
@seismicisolation
Introduction ix
of the body. Those constitute the only window through which one can visualize or
study a little fragment of the solution in non-integrable problems over an infinite
time interval.
Direct methods already used in the construction of new integrable problems since
the late nineteenth century have been exhausted and are no longer capable of iden-
tifying integrable cases in more complicated problems. Completely new ways of
thinking were required to establish new methods and to make substantial advance-
ments in the subject. One of those ways was writing equations of motion in the form
of the Lax pairs and using advanced tools of algebraic geometry. In this way, many
integrable systems were constructed in the dynamics of particles and some in rigid
body dynamics. Later, those methods helped to construct the formal exact solution
of some problems. An example is the integrable case of motion of a body acted upon
by two skew uniform fields, which generalizes Kowalevski’s case of a single field
due to Reyman and Semenov-Tian-Shansky. Two integrable versions of this problem
were known when the problem admits a linear integral, but using the above method
made it possible to obtain the case, with a quadratic integral, unifying those versions.
The solution of the complexified equations was also pointed out in terms of Theta
functions.
Another method of finding integrable problems in rigid body dynamics was an
inverse one. The author of the present book developed a method for the construction of
integrable generalized natural systems of two degrees of freedom, which admit inte-
grals of motion in the form of polynomial in the velocities with coefficients depending
on the position. This method led to the appearance of vast families of such systems,
living on two-dimensional Riemannian manifolds. Designating special values for the
parameters, it was possible to construct integrable cases on some known manifolds.
In this way, a comparatively large collection of general and conditional integrable
problems in rigid body dynamics was constructed over the past decades. Examples
are the case of a gyrostat, which generalizes the classical case of Kowalevski, and a
large set of conditional cases. Every one of the integrable cases poses new mathe-
matical challenges: to investigate qualitative properties using integrals of motion, to
achieve separation of variables and study topological properties of integral manifolds
in the phase space. Another possible task is the construction of explicit time solution
of the equations of motion, usually by inverting quadratures in case of separation of
variables or by using Lax pair representation of a certain type.
Moreover, the application of certain transformations to known integrable cases
with cyclic coordinates has led to the construction of much more general inte-
grable cases. This produced new general integrable cases depending on several extra
parameters and added physical effects to all the integrable cases known earlier.
Thus, the state of the subject has radically changed since the time of the well-
known monographs of Leimanis and Magnus. In 2005, Borisov and Mamaev
published their book “Rigid Body Dynamics” and an English translation appeared
in 2017. This marvellous book lays emphasis on mathematical structures and gener-
alizations of integrable cases to higher dimensions. Several important topics are
not covered in it. Moreover, reading this book requires professional mathematical
knowledge.
@seismicisolation
@seismicisolation
x Introduction
The second motivation for writing the book was to showcase some novel methods
developed by the author that have led to substantial results and completely new collec-
tion of integrable cases in rigid body dynamics. Although almost all these cases have
definite physical interpretation, they have not been given due attention. In the mean
time, as we essentially use the Lagrangian approach, against a currently prevailing
trend that the Hamiltonian (or Poisson bracket) approach is the one that should be
followed in all problems of mechanics whenever it is applicable. We had to explain
why we intentionally use the Lagrangian formalism to present some features of
mechanical systems that are vividly seen in that formalism, while disguised in the
Hamiltonian formalism. Our standpoint is that the use of this or that approach is not
an arguable issue. None of them can be said to be absolutely better than the other.
Each approach must be applied, with no prejudice, to problems for which it is most
suited. The problem of classifying and tabulating integrable cases of motion of a
rigid body subject to potential and gyroscopic forces, which occupies most of the
book, is an example. Those cases are time-irreversible, i.e. equations of motion are
not invariant under the change of sign of the time variable. As will be established
later in the book, every such case is completely determined by two (scalar and vector)
functions V and μ. The scalar is the potential and the vector uniquely determines
the moment exerted by gyroscopic forces. The pair (V, μ) uniquely characterizes
the physics of the problem and its equations of motion and thus can be used as a
basis for the classification and tabulation of integrable cases. On the other hand, in
inverse methods used to construct integrable systems, gauge terms that arise as a part
of the solution of partial and ordinary differential equations enter in the definition
of momenta. The Hamiltonian function and the Hamiltonian equations of motion
depend on those terms and obscure the necessary terms that determine potential
and gyroscopic forces acting on the body together with the physics of the problem.
Hamiltonian equations of motion of an irreversible mechanical system can be written
in an infinite number of equivalent forms. In fact, to determine whether two Hamil-
tonians are equivalent, one has to do some steps that are equivalent to finding the
equations of motion in the Lagrangian form. Concrete examples are given in the last
chapters of the book, beginning with Chap. 10. Nevertheless, in Chaps. 10 and 12,
after classification of integrable cases on the Lagrangian basis, to conform with the
reference character of the book, we also give Hamiltonians and complementary inte-
grals in terms of momenta for all integrable cases. These cases involve gyroscopic
moments, which depend on the position in a complicated manner. For such systems,
the Lagrangian approach is not just an awkward presentation, but it faithfully and
uniquely presents the physics of the problem under consideration. On the other hand,
a system of Hamiltonian equations of motion can represent a whole class of physi-
cally different mechanical systems on that level of complication. Nevertheless, the
whole integrability theory is most easily and clearly in terms of the Hamiltonian
approach.
In the plan of this book, it soon became clear that the original motivation to include
all new changes in the field of rigid body dynamics to produce something similar to
Routh’s tractate of the late nineteenth century is too ambitious and rather impractical.
The changes in the subject in the preceding half-century are far more extensive, to
@seismicisolation
@seismicisolation
Introduction xi
be included in one volume. A narrower line had to be set as an aim for the book. We
have, thus, made the decision to make a survey of known integrable cases in various
problems in the dynamics of a rigid body moving about a fixed point under given
forces, including certain problems of motion of a body with no fixed point, but which
reduce, after some transformation or reduction, to the first type of problems (with a
fixed point). Examples of such problems are the problem of motion of a body in a
liquid and that of motion of a satellite in a circular orbit about a spherical planet.
Thus, no place was left for some important problems. The problem of motion of
a rigid body subject to a non-holonomic constraint, rolling on a plane or a surface,
is an example. Other examples are the motion of a gyroscope in gimbals, problems
involving motion of a system of connected bodies in a general state of motion and
the motion of a body with a cavity, completely filled with a liquid in a state of
vortex flow. The last problem is described by Poincaré–Joukovsky’s equations of
motion. We have also excluded a large set of existing solutions, which are not in
finite exact form, like asymptotic solutions and series solutions or perturbations of
exact solutions in power series of a small parameter. On the other hand, problems of
rigid body dynamics in which integrability is not a principal issue are not considered.
An example is the controlled rotational motion of the rigid body.
Even in the main course of the problem considered in the book, we had to make
some definite selections of the material to be included. In the first place, we intended
to make a complete up-to-date account of all known integrable cases in the subject. A
considerable part of such content is scattered in the literature and would be presented
in the form of a book for the first time. The information about integrable cases should
contain conditions for their existence, full historical context of their discovery or
development from former cases and sufficiently detailed forms of the first integrals
in each case. This covers all general integrable cases, i.e. cases integrable in the
whole phase space (for arbitrary initial conditions) as well as conditional integrable
cases, i.e. integrable on a fixed level of the relevant linear integral of the motion
(the areas integral). In most elementary cases, we tried to illuminate as much as
possible the process of obtaining the explicit solution of the equations of motion.
By this, we mean the expression of all the physical phase variables in terms of
time. As will be seen in most of the integrable problems considered in this book,
the separation of variables, inverting quadratures and constructing explicit solutions
have turned into a separate art and in their majority still represent open mathematical
challenges. Even in solved cases, a frequently met drawback is that some explicit
solutions are expressed using complex functions of time, a situation that obstructs
their use in numerical calculations or simulation. For this reason, greater importance
is devoted to the construction of some particular solutions expressible in terms of
elliptic, trigonometric or simpler functions of time. Apart from the integrable cases,
complete, and somewhat detailed, account of all the twelve known exact particular
solutions of the classical problem of motion of a heavy rigid body is given. But this
could not be pursued in other higher problems of the hierarchy. That could simply
double the size of the book and also the time to compile the existing information.
@seismicisolation
@seismicisolation
xii Introduction
The book is divided into two parts:

The first part, the elementary part, grew mainly as a course on rigid body dynamics
delivered over years to undergraduate mathematics students of the Faculty of Science
and it can be used for this purpose. This part includes Chaps. 1–7, covering the mate-
rial necessary for a mathematics or physics student to get acquainted with the subject
of rigid body dynamics, its main problems, techniques and historical development. A
few sections of Chap. 8 can be selected to augment that content with some examples
of a particular solution in rigid body dynamics.
We begin in Chap. 1 with a study of the characteristics of mass distributions: the
centre of mass and the inertia matrix. Even in this classical material, some innovative
element was introduced. A new theorem is given, determining natural bounds on the
location of the centre of mass of a body with given moments of inertia. Chapter 2 is
devoted to different ways of the description of finite rotations, infinitesimal rotations
and the angular velocity vector. Introduced here are Euler’s angles, the rotation matrix
and quaternions for describing the orientation of the body.
Chapter 3 includes a brief study of the classical problem of motion of a rigid body
about a fixed point under the action of its own weight. Different forms of the equations
of motion and their integrals are derived in different reference frames fixed in the
body and moving with it. Equations of motion are obtained in Lagrangian, Routhian
and Hamiltonian forms.
In Chap. 4, the three general integrable cases of the classical problem known after
the names of Euler, Lagrange and Kowalevski are presented in some detail. Explicit
time solution of the equations of motion is given in terms of elliptic functions of time
for Euler’s case of a torque-free body. The solution of Lagrange’s case is reduced to
an elliptic quadrature, which may be used to express it in elliptic functions as well.
However, we relied, following Poisson, on the use of integrals of motion to establish
certain qualitative aspects of the motion, without referring to explicit time solution.
Kowalevski’s case is formally reduced to hyper-elliptic quadratures. Some degenerate
cases are solved in elliptic or simpler functions in Appendix B. The conditional case
of integrability bearing the names of Goryachev and Chaplygin is also presented with
its separation of variables belonging to Chaplygin. Degenerations of hyper-elliptic
quadratures are presented in some detail in Appendix C.
Chapter 5 is devoted to the study of the problem of motion of a heavy gyrostat. In
its simplest form, the gyrostat is a rigid body in which a symmetric rotor is placed
with its axis fixed in the carrier body by cylindrical smooth joint(s) and given a
constant angular speed with respect to the main body. The gyrostatic effects are
in wide use in several problems of science and technology. Equations of motion
were formulated in the last decades of the nineteenth century. An integrable case is
readily recognized, which is a trivial generalization of Lagrange’s case, when the
main body is axially symmetric and the rotor is aligned along its axis of symmetry.
The second case generalizes Euler’s case in the classical problem by adding a rotor
in an arbitrary direction fixed in the body. This case was found by Joukovsky and
shortly later by Volterra. The third general integrable case, Yehia’s case, was found as
a generalization of Kowalevski’s case in the classical problem by adding a gyrostatic
momentum. The conditional case of Goryachev and Chaplygin that was generalized
@seismicisolation
@seismicisolation
Introduction xiii
to the gyrostat by Sretensky is also presented. The chapter concludes with some
applications of the gyrostat dynamics to stabilize certain motions.
In Chap. 6, the problem of motion of a gyrostat about a fixed point while acted upon
by the force of a Newtonian centre of attraction is presented. Especially interesting
is the case when the attraction centre is far from the fixed point. This case is treated
in some detail, for it has several applications in certain problems in astronomy and
physics. Two integrable cases are known. They generalize Euler’s and Lagrange’s
cases. In this chapter, we also present what is called Brun’s problem, which is equiva-
lent to a special version of the former problem. A quite interesting property is proved
that Brun’s potential is the only one that admits an integral of motion, quadratic in
the velocities.
Chapter 7 contains a brief account of the problem of motion of a body having no
fixed point. Equations of motion will be useful in certain topics later to be exposed in
this book. A quite interesting example, the motion of a top on a smooth plane (called
Poisson’s top), is considered at least for its educational importance.
The second part of the book contains mostly new research material that was not
compiled before in book form.
The original course included a few examples of particular solutions of the equa-
tions of motion. In the final plan of the book, I found it necessary to make separate
Chap. 8 collecting the basic information and results about all the twelve known exact
particular solutions of the equations of motion in the classical problem. This infor-
mation, which accumulated in the period from the 1890s to 1970s, to the date the last
case was found, has never been presented in a source in the English language. Just a
few cases are pointed out in Leimanis’ book [256], not all of them are correct. The
same situation applies to Magnus’ monograph [270]. Borisov and Mamaev mention
only half of these cases, with somewhat detailed analysis of the earlier results of
Staude, Hess, Bobylev, Steklov and Grioli. In our presentation, some new features
were added. In most cases of a particular solution, we show the curve drawn during
the motion by the apex of the vertical unit vector γ on the unit sphere fixed in the
body. This graph gives a full idea on how the body moves relative to the vertical
through the fixed point, i.e. up to a rotation about it.
Exact particular solutions of the gyrostat problem are considered only in one brief
section of Chap. 8. Those solutions were intensively studied almost exclusively by
the school of Mechanics in Donetsk. All those cases are listed with the essential
information on each case. Those are nine cases generalizing their counterparts of the
classical problem and four new cases with no classical analogs. For each case, we
provided necessary information and references that would help the interested reader
to track every case in original works.
In Chap. 9, we consider the analogy between the motion of a rigid body about
a fixed point and the problem of motion of a particle on a smooth ellipsoid. This
analogy, noted first by Minkowski, furnishes several easy ways for the reduction of
the order of equations of motion, using the known integrals of motion. The climax in
this direction is the maximal reduction to a single differential equation of the second
order named as the “orbital equation”. This equation settles once for all the question
@seismicisolation
@seismicisolation
xiv Introduction
of maximal reduction of order raised in rigid body dynamics the since 1890s and
discussed by a long list of authors.
Chapter 10 is devoted to the problem of motion of a body by inertia in an ideal
incompressible fluid extending to infinity in all directions. This problem originally
belongs to the field of hydrodynamics, which is described by boundary-value prob-
lems on partial differential equations. Nevertheless, the efforts of several authors led
to the result that the pressure of the fluid on the body can be completely avoided,
leaving us with a mechanical system of six degrees of freedom. Equations of motion
of that system were given by Kirchhoff and Clebsch for a simply connected body
and by Lamb for a multi-connected (perforated) body.
Also in this chapter, a new form of the equations of motion and a transformation
have changed the way of presentation so deeply, gave a new insight into the problem
and revealed its inherent relation to other physical problems that were treated before
as completely separate from each other. An analogy has been established between
this problem and a special form of the problem of motion about a fixed point of a
heavy and magnetized body, which carries immovable in it electric charges under
the action of an axi-symmetric combination of gravitational, electric, magnetic and
Lorentz forces. This opened the way to study systematically, for the first time, the
motion of a heavy, magnetized and electrically charged body or gyrostat. The full
list of known integrable cases, seven general and two conditional, valid for the two
equivalent problems, is given in a unified form. For each case, we give relevant
historical information and essential contributions to its study. For completeness, we
also provide the Hamiltonian and the complementary integral for every integrable
case in the tables, beginning from Chap. 10, where gyroscopic forces will play a more
prominent role and Hamiltonians found by inverse methods are usually obscured by
gauge terms.
The analogy just described above of the problem of motion of a body in a liquid
and the alternative problem has placed the last problem on the top of a hierarchy
of the problems considered in all previous chapters and paved the way to create a
higher and richer level of that hierarchy that was never treated before. It may have
been considered as hopelessly complicated to yield significant results.
In Chap. 11, the use of the Lagrangian approach and certain peculiarities of the
equations of motion has pushed the whole subject beyond its common limits. Equa-
tions of motion are given in their historical context. They formally generalize the
new alternative form of equations of motion of a body in a liquid and go much further
from the physical point of view. Transformations are given, which generate new inte-
grable cases of the most complicated nature from the ones in the lower hierarchies,
by adding more parameters into their structures.
In Chap. 12, unprecedented and quite complicated integrable cases involving
large numbers of parameters were constructed in an exotic, but effortless, way. In
fact, we have used certain tricky properties of the Lagrangian formalism to add extra
parameters of physical significance to the structures of the known general integrable
cases of a body in a liquid. The number of those additional parameters depends on
the structure of the potential part of the Lagrangian of the integrable problem. The
new cases are the only known examples in our days of integrable cases of motion of
@seismicisolation
@seismicisolation
Introduction xv
a rigid body acted upon by potential and gyroscopic forces of the most complicated
structure. The variety of those cases may shed some light on some of the most
intractable problems of mechanics concerning the motion of natural and artificial
bodies in extreme conditions when the fields applied to the body have comparable
effects and none of them can be treated as negligible.
In this chapter, we also introduce a new type of generalization of each general
case of integrability of motion of a body in a liquid a conditional integrable case
involving an arbitrary function. This type of generalization is valid on a fixed level
of the cyclic (areas) integral.
It was argued by some authors that those two types of generalization are trivial
and lose their value if the problem under consideration is written in Hamiltonian
formalism. We use the Hamiltonian formalism to show that this view of the subject
is not factual.
In Chap. 13, we give a full list of the known up-to-date conditional integrable cases
of the problem of motion of a rigid body. The tables for those cases are given here in
their last and most general form, with no regard to their physical interpretation. Cases
in those tables are ordered according to the degree of the complementary integral as a
function of the components of the angular velocity of the body. Some of them acquire
a physical meaning only for certain values of the parameters present in them. Other
ones do not seem presently to have physical meaning at all, mostly because their
potentials involve singular terms of certain types, not usually attainable by natural
fields. The 22 cases known at present of this class were obtained mainly in the works
of the author and some with his coworkers. Most of those cases have resulted as
special cases of certain generalized natural multi-parameter integrable systems that
were constructed by the author over the last few decades. Some of those cases have
even stimulated research to find new ways for the separation of variables and other
mathematical topics.
Chapter 14 is devoted to a systematic presentation of the present status of the
problem of motion of a rigid body about a fixed point under the action of an asym-
metric combination of potential and gyroscopic forces (crossed fields). Equations
of motion are derived in the Euler–Poisson variables. Known integrable cases are
collected and classified. First presented are integrable cases of a body acted upon
by two and three skew uniform fields, then cases with a potential that is quadratic
in the direction cosines. Apart from some special cases, in both types of problems,
the mechanical system has strictly three degrees of freedom, i.e. does not admit
a cyclic integral. In the last two sections of this chapter, we present two classes of
problems admitting a cyclic coordinate. The problem of motion of a (physically) axi-
symmetric body under the action of asymmetric forces admits the Eulerian proper
rotation angle as a cyclic coordinate. The symmetry leading to a cyclic integral in
the second problem is not about a fixed axis, neither in space nor in the body. It may
be interpreted as axial symmetry in the quaternion space. The cyclic coordinate is
the sum (or difference) of the two Eulerian angles of precession and proper rotation.
For the last two classes, the method described in Chap. 11 and applied in Chap. 12
gives some exotic generalizations of the well-known cases.
@seismicisolation
@seismicisolation
xvi Introduction
Exercises constitute an essential component of the book. A large number of them

provide some supplemental information to the main text or introduce in a brief way
additional topics of special interest that could not be presented in more detail.
The elementary part of the present book should be easily readable by anyone who
has completed courses on calculus, differential equations and analytical dynamics.
Some parts require some knowledge of elliptic integrals and Jacobi’s elliptic func-
tions. The advanced part provides mostly new material but it is written in the most
elementary way. It will be tractable for readers of different mathematical back-
grounds: students, Mathematicians, Physicists and Engineers. I hope that it will stim-
ulate research in the field. Many of the new cases may be investigated for explicit
solution either by separation of variables or by the use of Lax pairs. Topological
classification and qualitative properties of motion can be studied for every case, for
which separation of variables is achieved. Many cases are waiting for appropriate
concrete physical interpretation.
This book is intended to be a reference book for integrable cases and exact solu-
tions. I have taken possible care of checking the large number of formulas involved,
mostly by using computer packages of symbolic computation.
Throughout the book, I used the usual notations for mathematical terms and
operations. Vectors and matrices are denoted by bold symbols, and scalar and vector
products by dot and cross, respectively. We have found it much easier and more
consistent to use for multiplication of a vector v by a matrix M the usual matrix form
vM, instead of the mostly used operator form Mv. The first produces vectors in the
usual row form and brings some advantage in avoiding the need in many sources to
switch between row and column forms of vectors.
During work on this book, I enjoyed generous help from many friends, to whom
I express my sincere gratitude. A conversation with David Gao at a conference
in Poland was encouraging and inspiring. Michael and Irina Kharlamov and Pavel
Ryabov provided me with some old papers in Russian. Gennady Gorr secured for
me issues of MTT and longtime chats discussing many details of the subject. Our
long-years friendship was in no way affected by our differences on some of the
content of books coauthored by him. Alexey Borisov made some publications of RCD
available to me, in addition to his books coauthored by Ivan Mamaev, including their
marvellous book on rigid body dynamics (2005) and its recent English translation
(2017). On the other hand, Ahmed Ghaleb has read extensive parts of the manuscript
and made many suggestions to improve the English text. Adel Elmandouh and Ashraf
Hussein helped me by checking some mathematical calculations and resolving some
issues concerning LaTeX editing of the manuscript. Hani Yehia and Ashraf Hussein
helped me to improve the quality of some graphics.
Mansoura, Egypt Hamad Yehia

March 2021 [email protected]
@seismicisolation
@seismicisolation
Contents
Part I The Elementary Part

1 Distribution of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 The Moment of Mass—The Centre of Mass . . . . . . . . . . . . . . . . . 3
1.1.1 Moments of a Mass Distribution . . . . . . . . . . . . . . . . . . 3
1.1.2 Centre of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Second Moments and Inertia Matrix of a Mass Distribution . . . 4
1.2.1 Second Moments Matrix of Mass Distribution . . . . . . 5
1.2.2 Inertia Matrix of Mass Distribution . . . . . . . . . . . . . . . . 5
1.3 Properties of the Inertia Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 The Triangle Inequalities . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.2 Theorem of Parallel Axes . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.3 Ellipsoid of Inertia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.4 The Gyration Ellipsoid . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.5 Representation of Principal Moments of Inertia . . . . . 11
1.4 Relations Between the Centre of Mass and the Inertia
Ellipsoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Solved Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2 Description of Rotation of a Rigid Body About a Fixed Point . . . . . . 21
2.1 The Position of a Rigid Body. Euler’s Angles . . . . . . . . . . . . . . . . 21
2.2 The Rotation Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.1 The Angle of Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Description of Finite Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4 Representation of Finite Rotation by Means of a Vector . . . . . . . 28
2.4.1 The Rotation Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5 Hamilton–Rodrigues’ Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.6 The Angular Velocity Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
xvii
@seismicisolation
@seismicisolation
xviii Contents
2.7 Space and Relative Time Rates of Change of a Vector . . . . . . . . 34

2.7.1 Components of the Angular Velocity
in the Body Axes and Space Axes . . . . . . . . . . . . . . . . . 35
2.7.2 The Use of the Euler–Rodrigues Parameters . . . . . . . . 35
2.8 Quaternions and Representation of Finite Rotation . . . . . . . . . . . 37
2.9 Composition of Two Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3 The Classical Problem: The Motion of a Heavy Rigid Body
About a Fixed Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.1 Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2 The Heavy Rigid Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 The Angular Momentum of a Rigid Body . . . . . . . . . . . . . . . . . . . 43
3.4 Kinetic Energy of a Moving Body . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Equations of Motion in the Moving Coordinate System . . . . . . . 45
3.5.1 The Use of the Variables ω, γ . Special Axes
Related to the Inertia Matrix . . . . . . . . . . . . . . . . . . . . . 45
3.6 Integrals of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.6.1 The Energy Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.6.2 The Area’s Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.6.3 The Geometric Integral . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.6.4 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.7 Special Axes Associated with the Inertia Matrix . . . . . . . . . . . . . 47
3.8 The Use of Principal Axes of Inertia of the Body . . . . . . . . . . . . . 48
3.9 Determination of Euler’s Angles . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.10 The Movable and Immovable Hodographs . . . . . . . . . . . . . . . . . . 50
3.11 The Use of the Variables G, γ . Special Axes Associated
with the Gyration Ellipsoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.12 Equations of Motion in Generalized Coordinates . . . . . . . . . . . . . 52
3.13 Canonical Equations of Motion in Euler’s Angles . . . . . . . . . . . . 53
3.14 The Routhian Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.15 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4 General and Conditional Integrable Cases of the Classical
Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1 Euler’s Case (1758). The Torque-Free Rigid Body . . . . . . . . . . . 60
4.1.1 Explicit Time-Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.1.2 Permanent Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.3 The Degeneracy of Elliptic Function . . . . . . . . . . . . . . 64
4.1.4 The Case of Dynamically Axi-Symmetric Body . . . . . 64
4.1.5 Euler’s Angles in Terms of Time . . . . . . . . . . . . . . . . . . 65
4.1.6 Geometrical Interpretation of the Motion
(Poinsot 1851) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2 Lagrange’s Case (1788). The Top with a Fixed Point . . . . . . . . . 69
4.2.1 The Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.2.2 The Study of the Motion . . . . . . . . . . . . . . . . . . . . . . . . . 72
@seismicisolation
@seismicisolation
Contents xix
4.3 Kowalevski’s Case (1888) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.3.1 Integration of the Equations of Motion . . . . . . . . . . . . . 76
4.4 The Goryachev–Chaplygin Case: A Conditional
Integrable Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4.1 The Fourth Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4.2 Separation of Variables. Solution
of the Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . 81
4.5 Integrability and Nonintegrability Issues . . . . . . . . . . . . . . . . . . . . 83
5 The Motion of a Heavy Gyrostat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1 Models of the Gyrostat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1.1 The Classical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1.2 The Free Rotor Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.1.3 Joukovsky’s Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.2 Equations of Motion in Hamiltonian Form . . . . . . . . . . . . . . . . . . 89
5.3 Tables of Integrable Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.4 The Case of Joukovsky and Volterra . . . . . . . . . . . . . . . . . . . . . . . 91
5.5 The Case of Axially Symmetric Gyrostat . . . . . . . . . . . . . . . . . . . 91
5.6 Yehia’s Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6.1 Separation of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.7 The Conditional Case of Sretensky . . . . . . . . . . . . . . . . . . . . . . . . 96
5.8 Some Applications of the Gyrostat Motion . . . . . . . . . . . . . . . . . . 96
5.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6 Motion of a Rigid Body About a Fixed Point in the Field
of a Distant Newtonian Centre and Brun’s Problem . . . . . . . . . . . . . . 99
6.1 Approximate Form of the Potential . . . . . . . . . . . . . . . . . . . . . . . . 100
6.2 Brun’s Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.3 Equations of Motion and Integrals of Motion . . . . . . . . . . . . . . . . 102
6.4 Integrable Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.4.1 Brun’s Case [47] (Analog of Euler’s Case) . . . . . . . . . 102
6.4.2 The Generalization of Lagrange’s Case . . . . . . . . . . . . 104
6.4.3 The Place of Brun’s Potential . . . . . . . . . . . . . . . . . . . . . 104
6.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
7 The Motion of a Body with No Fixed Point . . . . . . . . . . . . . . . . . . . . . . . 109
7.1 General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
7.2 Poisson’s Top. A Top on a Smooth Horizontal Plane . . . . . . . . . . 111
7.2.1 Regular Precession of Poisson’s Top . . . . . . . . . . . . . . . 114
7.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Part II The Advanced Part

8 Particular Solutions of the Classical Problem and Its
Generalizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
8.1 The Notion of a Particular Solution . . . . . . . . . . . . . . . . . . . . . . . . 119
@seismicisolation
@seismicisolation
xx Contents
8.2 Planar Motion (Motion of the Body as a Physical

Pendulum) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.2.1 Rotational Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.2.2 Vibrational Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.2.3 The Limiting Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
8.2.4 Orbital Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
8.3 Permanent Rotation of a Heavy Rigid Body About
a Fixed Point [343] (1894) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
8.3.1 Possible Axes of Permanent Rotation . . . . . . . . . . . . . . 125
8.3.2 Description of the Motion . . . . . . . . . . . . . . . . . . . . . . . . 126
8.3.3 Further Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.3.4 Excercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.4 Hess’ Case (1890) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
8.4.1 Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
8.4.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
8.4.3 The Use of Special Axes . . . . . . . . . . . . . . . . . . . . . . . . . 132
8.4.4 Solution of the Case f = 0 . . . . . . . . . . . . . . . . . . . . . . 134
8.5 The Case of Bobylev and Steklov (1896) . . . . . . . . . . . . . . . . . . . 136
8.5.1 Region I : The First Class of Motions . . . . . . . . . . . . . . 138
8.5.2 Region I I . The Second Class of Motions . . . . . . . . . . 140
8.6 Steklov’s Case (1899) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8.6.1 Conditions and Solution . . . . . . . . . . . . . . . . . . . . . . . . . 141
8.6.2 The First Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.6.3 The Second Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.6.4 Some Properties of the Motion . . . . . . . . . . . . . . . . . . . 145
8.6.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.7 Goryachev’s Case [114] (1899) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.7.1 Conditions and Solution . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.7.2 Properties of the Motion . . . . . . . . . . . . . . . . . . . . . . . . . 152
8.7.3 The History of Goryachev’s Case . . . . . . . . . . . . . . . . . 156
8.8 Chaplygin’s Case (1904) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8.8.1 Properties of the Motion . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.9 Kowalewski’s Case [239] (1908) . . . . . . . . . . . . . . . . . . . . . . . . . . 162
8.10 Grioli’s Case (1947): The Regular Precession About
a Tilted Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
8.10.1 Motion of the Centre of Mass . . . . . . . . . . . . . . . . . . . . 173
8.10.2 Orbits of Motion on the Poisson Sphere . . . . . . . . . . . . 174
8.11 Dokshevich’s First Case [63] (1966) . . . . . . . . . . . . . . . . . . . . . . . 175
8.11.1 Use of Special System of Axes . . . . . . . . . . . . . . . . . . . 175
8.11.2 Orbits on the Poisson Sphere . . . . . . . . . . . . . . . . . . . . . 178
8.12 The Case of Konosevich and Pozdnyakovich [230, 231]
(1968) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
8.12.1 The First Subcase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
8.12.2 The Second Subcase . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
@seismicisolation
@seismicisolation
Contents xxi
8.13 Dokshevich’s Second Case [64] (1970) . . . . . . . . . . . . . . . . . . . . . 182

8.13.1 Periodicity of the Motion . . . . . . . . . . . . . . . . . . . . . . . . 186
8.13.2 Orbits on the Poisson Sphere . . . . . . . . . . . . . . . . . . . . . 187
8.14 Unsuccessful Cases and Incorrect Claims . . . . . . . . . . . . . . . . . . . 188
8.14.1 Shiff’s Work [331] (1903) . . . . . . . . . . . . . . . . . . . . . . . 189
8.14.2 Field’s Works [85, 86] (1934) . . . . . . . . . . . . . . . . . . . . 189
8.14.3 Corliss’ Works [56, 57] (1932–1934) . . . . . . . . . . . . . . 189
8.14.4 Fabbri’s Works [80, 81] (1934) . . . . . . . . . . . . . . . . . . . 189
8.14.5 Concerning Mertsalov’s Work [283] (1946) . . . . . . . . 190
8.14.6 Gao’s Work [98] (2003) . . . . . . . . . . . . . . . . . . . . . . . . . 190
8.14.7 The Work of Yanxia and Keying [371] (2005) . . . . . . . 191
8.14.8 Ershkov’s Work [76] (2014) . . . . . . . . . . . . . . . . . . . . . . 191
8.15 Particular Solutions in the Problem of Motion of a Heavy
Gyrostat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
8.15.1 Generalizations of Particular Cases Known
in the Classical Problem . . . . . . . . . . . . . . . . . . . . . . . . . 192
8.15.2 Solvable Cases of the Gyrostat, Having No
Classical Analog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
8.15.3 Known Cases of the Classical Problem, Which
Are not Presently Generalized to the Gyrostat
Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
9 The Rigid Body in a Potential Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
9.1 The Routhian in γ1 , γ2 , γ3 as Redundant Variables . . . . . . . . . . . 199
9.1.1 Expression of ω in Terms of γ and γ̇ . . . . . . . . . . . . . . 200
9.1.2 The Case of Complete Dynamical Symmetry
of the Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
9.2 Maximal Reduction of Order of the Differential Equations
of Motion of a Rigid Body About a Fixed Point . . . . . . . . . . . . . . 202
9.2.1 Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
9.2.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
9.3 Reduction to the Motion of a Particle on an Ellipsoid . . . . . . . . . 208
9.4 The Use of Elliptic Coordinates on the Inertia Ellipsoid
[224] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
9.5 The Use of Isometric Coordinates on the Inertia Ellipsoid . . . . . 211
9.6 Reduction to the Simplest Form of Orbital Differential
Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
9.6.1 Sphero-Conic Coordinates on the Poisson
Sphere [378] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
9.6.2 Reduction to a Single Differential
Equation [378] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
9.6.3 Special Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
@seismicisolation
@seismicisolation
xxii Contents
9.7 Separable Potentials in Rigid Body Dynamics

(Conditional Integrable Problems) . . . . . . . . . . . . . . . . . . . . . . . . . 218
9.7.1 Potentials Separable for Axi-Symmetric Body . . . . . . 219
9.7.2 Potentials Separable for an Asymmetric Body . . . . . . 220
9.7.3 Potentials Separable for a Body of Spherical
Dynamical Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
9.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
10 The Problem of Motion of a Body in a Liquid . . . . . . . . . . . . . . . . . . . . 229
10.1.1 Kirchhoff’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
10.1.2 Example: Permanent Translational Motions . . . . . . . . 234
10.1.3 Clebsch’s Form of Kirchhoff’s Equations . . . . . . . . . . 234
10.2 Thomson-Lamb’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
10.3 On Different Forms of the Equations of Motion . . . . . . . . . . . . . . 238
10.4 A New Form of the Equations of Motion . . . . . . . . . . . . . . . . . . . 239
10.5 Steklov and Kharlamov Analogies and Their
Generalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
10.5.1 The Equivalent Problem of Motion About
a Fixed Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
10.5.2 Steklov’s Analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
10.5.3 Kharlamov’s Analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
10.6 Completing the Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
10.6.1 Solution of the Equivalent Problem . . . . . . . . . . . . . . . . 246
10.6.2 Solution of the Original Problem . . . . . . . . . . . . . . . . . . 247
10.7 Uniform Translational-Rotational Motion of a Body
in a Liquid (Permanent Rotations of a Body with a Fixed
Point About a Vertical Axis) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
10.8 Stationary Motions About an Axis Inclined to the Vertical . . . . . 251
10.9 A Several-Parameter Particular Solution . . . . . . . . . . . . . . . . . . . . 252
10.10 Alternative Hamiltonian Formulation . . . . . . . . . . . . . . . . . . . . . . 253
10.11 The Uniform Precession Transformation [383] . . . . . . . . . . . . . . 257
10.11.1 Direct Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
10.11.2 Lagrangian Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
10.11.3 Physical and Mechanical Significance
of the Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
10.11.4 Uniform Precession Transformation
in Hamiltonian Formalism . . . . . . . . . . . . . . . . . . . . . . . 262
10.12 Generalization of General Integrable Cases . . . . . . . . . . . . . . . . . 264
10.12.1 Generalization of the Integrable Case Found
by Sokolov . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
10.12.2 Steklov’s Case and Its Generalizations . . . . . . . . . . . . . 266
10.13 Generalization of Conditional Integrable Cases . . . . . . . . . . . . . . 268
10.13.1 Generalization of Goryachev–Chaplygin’s,
Sretensky’s and Sokolov–Tsiganov Cases . . . . . . . . . . 269
@seismicisolation
@seismicisolation
Contents xxiii
10.14 Generalizations of Particular Solvable Cases . . . . . . . . . . . . . . . . 270

10.14.1 Example 1. Equilibria and Permanent Rotations
About a Vertical Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
10.14.2 Example 2. Permanent Rotations About
a Tilted Axis and Precessional Motions About
the Vertical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
10.14.3 Example 3. generalization of grioli’s precession
[402, 405] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
10.14.4 Example 4. Regularly Precessing Pendulum . . . . . . . . 275
10.15 Tables of Integrable Cases of Motion of a Rigid Body
in a Liquid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
10.15.1 General Integrable Cases . . . . . . . . . . . . . . . . . . . . . . . . 286
10.15.2 Conditional Integrable Cases on the Level f = 0 . . . . 291
10.16 Further Studies on Integrable Cases . . . . . . . . . . . . . . . . . . . . . . . . 292
10.16.1 Separation of Variables, Explicit Solutions and . . . . . . 292
10.16.2 Topological Classification of Integrable Cases . . . . . . 294
10.17 Chaplygin’s Case of Integrability . . . . . . . . . . . . . . . . . . . . . . . . . . 295
10.17.1 Separation of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 296
10.17.2 Forms of Motion on the Poisson Sphere . . . . . . . . . . . . 299
10.17.3 Explicit Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
10.18 Integrability Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
10.18.1 Results Concerning Kirchhoff’s Equations . . . . . . . . . 305
10.19 Remark Concerning Particular Solutions of the Problem . . . . . . 307
10.20 The Donetsk School of Mechanics and Its Attitude
to Competing Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
10.20.1 The Attitude to the Uniform Precession
Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
10.20.2 The Attitude to the Equations of Motion
in the Form (10.45) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
10.21 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
11 The General Problem of Motion of a Rigid Body Acted
upon by a Coaxial Combination of Potential and Gyroscopic
Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
11.3 Relation to Grioli’s and Kharlamov’s Equations . . . . . . . . . . . . . 322
11.3.1 Grioli’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
11.3.2 M. Kharlamov’s Equations . . . . . . . . . . . . . . . . . . . . . . . 323
11.4 Potential of, and Torques on, a Heavy, Magnetized
and Electrically Charged Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
11.5 On General and Conditional Integrable Cases in Rigid
Body Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
11.6 Transformation of the Equations of Motion . . . . . . . . . . . . . . . . . 329
@seismicisolation
@seismicisolation
xxiv Contents
11.7 Maximal Reduction of the Order of the Equations

of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
11.7.1 The Case of Complete Dynamical Symmetry . . . . . . . 333
11.8 Extensions of Integrable Problems . . . . . . . . . . . . . . . . . . . . . . . . . 334
11.9 Transformations of Cyclic Variables . . . . . . . . . . . . . . . . . . . . . . . 337
12 The Most General Integrable Cases in Rigid Body Dynamics . . . . . . 343
12.1 General Integrable Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
12.1.1 Table of General Integrable Extensions
of General Integrable Cases . . . . . . . . . . . . . . . . . . . . . . 344
12.1.2 About the Hamiltonian Formulation . . . . . . . . . . . . . . . 352
12.2 Conditional Integrable Deformations of General
Integrable Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
12.2.1 Table of Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
12.2.2 Example of Physical Application . . . . . . . . . . . . . . . . . 359
13 Miscellaneous Cases Integrable on a Single Level of the Areas
Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
13.1 Cases with a Quadratic Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
13.1.1 Separable Integrable Potentials . . . . . . . . . . . . . . . . . . . 363
13.1.2 Non-separable Cases with a Quadratic Integral . . . . . . 364
13.2 Cases with a Cubic Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
13.3 Cases with a Quartic Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
13.3.1 Cases Stemming from Kowalevski’s Case . . . . . . . . . . 367
13.3.2 Cases Stemming from Chaplygin’s Case . . . . . . . . . . . 368
13.3.3 Cases Combining Kowalevski’s
and Chaplygin’s Cases . . . . . . . . . . . . . . . . . . . . . . . . . . 370
13.3.4 A Case with a Quartic Integral Outside
the Above Classification . . . . . . . . . . . . . . . . . . . . . . . . . 372
13.3.5 Two Conditional Cases Valid on a Single,
Not Necessary Zero, Level of the Linear
Integral [421] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
13.4 Integrable Extensions of Conditional Integrable Cases . . . . . . . . 373
14 The Rigid Body Acted upon by a Skew Combination of Fields . . . . . 375
14.1.1 Interpretation of Forces . . . . . . . . . . . . . . . . . . . . . . . . . . 377
14.1.2 The Motion of a Magnetizable Rigid Body
in an Ideal Fluid and In a Uniform Magnetic
Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
14.1.3 Example: The Motion of a Satellite in a Circular
Orbit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
@seismicisolation
@seismicisolation
Contents xxv
14.2 The Rigid Body (Gyrostat) Acted upon by More than One
Uniform Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
14.2.1 The Motion of a Body Acted upon by Two
Uniform Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
14.2.2 The Motion of a Body Acted upon by Three
Irreducible Uniform Fields . . . . . . . . . . . . . . . . . . . . . . . 388
14.3 Integrable Cases of a Body with a Homogeneous
Quadratic Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 390
14.3.1 Brun’s Case of the Asymmetric Body
in an Asymmetric Gravitational Field . . . . . . . . . . . . . . 390
14.3.2 Case of Dynamically Spherical Body . . . . . . . . . . . . . . 392
14.4 The Motion of an Axi-Symmetric Body Under the Action
of Asymmetric Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
14.4.1 Description of the Problem . . . . . . . . . . . . . . . . . . . . . . . 393
14.4.2 General Integrable Cases . . . . . . . . . . . . . . . . . . . . . . . . 395
14.4.3 Conditional Integrable Cases . . . . . . . . . . . . . . . . . . . . . 399
14.5 Motion of a Body with Combined (Quaternion) Symmetry . . . . 401
14.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
14.5.2 Routhian Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
14.5.3 Equivalence of Two Problems . . . . . . . . . . . . . . . . . . . . 403
14.5.4 Basic Equivalent Integrable Problems . . . . . . . . . . . . . . 407
14.5.5 Generalization Through Transformation . . . . . . . . . . . 413
Appendix A: Some Useful Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415

Appendix B: Kowalevski’s Case: Appelrot’s Four Classes of Simple
Motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
Appendix C: Particularly Simple Classes of Motions in Goryachev–
Chaplygin’s Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
Appendix D: Gyrostatic Generalization of the Appelrot Classes . . . . . . . . 433
Appendix E: The Conditional Case of Sretensky . . . . . . . . . . . . . . . . . . . . . . 439
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
@seismicisolation
@seismicisolation
Part I
The Elementary Part
@seismicisolation
@seismicisolation
Chapter 1
Distribution of Mass
In the study of the dynamics of a rigid body, we deal in a natural way with certain
quantities which are determined by the distribution of mass in that body. In this
chapter, we introduce those quantities and study their properties and the relations
between them.
1.1 The Moment of Mass—The Centre of Mass
A rigid body is defined as a finite distribution of mass in which the relative positions
of all its mass elements do not change with time, regardless of the position of the
body in space and the external forces exerted on it. We do not assume any conditions
on the structure or shape of the body, which may be composed of invariable rigidly
connected parts that can comprise in any way discrete point masses and continuous
line, surface or volume distributions of mass. The term “element of mass” we use
below should be interpreted in each case accordingly.
1.1.1 Moments of a Mass Distribution
The moment of mass of a given rigid body is a vector defined by the integral

σ= rdm, (1.1)
where dm is an infinitesimal mass element, r is the position vector of that element

and the integral is taken over all mass elements of the body. In a given system of
axes O x yz, the components are
© Springer Nature Switzerland AG 2022 3
H. M. Yehia, Rigid Body Dynamics, Advances in Mechanics and Mathematics 45,
https://doi.org/10.1007/978-3-030-96336-1_1
@seismicisolation
@seismicisolation
4 1 Distribution of Mass

( xdm, ydm, zdm), (1.2)
Those are the first moments of the mass distribution with respect to the given coor-
dinate system.
Generally, moments of arbitrary degree n for a mass distribution are also defined

σ(n 1 , n 2 , n 3 ) = x n 1 y n 2 z n 3 dm, (1.3)
where n 1 , n 2 , n 3 are non-negative integers and n = n 1 + n 2 + n 3 . In dynamics of

rigid bodies, zeroth-, first- and second-degree moments (n = 0, 1, 2) appear naturally
in the equations of motion. Higher moments are also met when the potential of a
rigid body is calculated in certain models of gravitational potential in the field of
attraction of other bodies. The simplest case is that of approximating the potential of
the body in the Newtonian field of a far centre of attraction. We shall return to this
point later with more detail.
1.1.2 Centre of Mass
Let M be the total mass of the body

M= dm. (1.4)
Obviously, M is positive and finite. The vector

1 1 1
r0 = σ/M = ( xdm, ydm, zdm) (1.5)
M M M
defines a unique point in the body, called the centre of mass. This point has the
fundamental property that the resultant of forces exerted on the body by an arbitrary
uniform gravity field always passes through it. However, this property can be lost for
any non-uniform gravitational field.
1.2 Second Moments and Inertia Matrix of a Mass

Distribution
In the course of our study of rigid body dynamics, we deal with two related matrices
(in fact, tensors):
@seismicisolation
@seismicisolation
1.2 Second Moments and Inertia Matrix of a Mass Distribution 5
1.2.1 Second Moments Matrix of Mass Distribution
Define the symmetric matrix Ī = ( I¯i, j )i,3 j=1

I¯i, j = ri r j dm, i, j = 1...3, (1.6)
where ri stands for the i-th component of the position vector r =(x, y, z) of the mass
element dm of the body, i.e.
⎛ 2 ⎞
x dm x ydm x zdm
Ī = ⎝ x ydm y 2 dm yzdm ⎠ . (1.7)
x zdm yzdm z 2 dm

Diagonal elements x 2 dm, y 2 dm and z 2 dm are called moments of inertia of
the body with respect to the planes yz, zx and x y, respectively.
1.2.2 Inertia Matrix of Mass Distribution
In most dynamical considerations, we more frequently meet the inertia matrix

defined as
I = tr (Ī)δ − Ī, (1.8)
where δ is the unit matrix. This makes

⎛ 2 ⎞
(y + z 2 )dm − x ydm − x zdm
I =(Ii, j )i,3 j=1 = ⎝ − x ydm (z 2+ x 2 )dm − yzdm ⎠ . (1.9)
− x zdm − yzdm (x 2 + y 2 )dm
The diagonal elements of the inertia matrix I11 , I22 and I33 are called moments of
inertia of the mass distribution with respect to the axes x, y, z, respectively, while
the off-diagonal ones are termed the products of inertia with respect to the three
coordinate planes. As the mass element is always positive, the moments of inertia
are non-negative. Moreover, a moment of inertia of a body about an axis vanishes
only if the body mass is distributed (continuously or discretely) on that axis.
Note that
tr (I) = 2tr (Ī), (1.10)
so that the inverse of the relation (1.8) can be written as
1
Ī = tr (I)δ − I. (1.11)
2
@seismicisolation
@seismicisolation
1.3 Properties of the Inertia Matrix
1.3.1 The Triangle Inequalities
Although the moments of inertia are always positive, not every three positive quan-
tities can represent moments of inertia of some body about three perpendicular axes.
In fact, moments of inertia of a body satisfy the inequalities
I11 + I22 − I33 0, I22 + I33 − I11 0, I11 + I33 − I22 0. (1.12)
The equality holds only for plane mass distributions. Those inequalities suggest that
the three moments of inertia about three perpendicular axes can be represented by
lengths of three sides of a triangle. We shall return to this point with more detail later
in this chapter.
To prove those inequalities, we notice from (1.7) that the diagonal elements of Ī
are non-negative, for being quadratic moments of mass with respect to the coordinate
planes. The three inequalities follow from the relation (1.11). Equality holds only
for bodies whose mass is distributed in one of the coordinate planes.
1.3.2 Theorem of Parallel Axes
Let I be the inertia matrix of a given body of total mass M with respect to some
Cartesian frame O x yz with origin O at the centre of mass of the body. We shall
calculate the inertia matrix I with respect to another Cartesian frame O x y z parallel
to the first, so that O has relative to O the position vector r1 = (x1 , y2 , z 1 ). The first
of those elements is

I11 = (y 2 + z 2 )dm

= [(y − y1 )2 + (z − z 1 )2 ]dm

= (y 2 + z 2 )dm + (y12 + z 12 ) dm − 2y1 ydm − 2z 1 zdm.

Since ydm = zdm = 0, we get

I11 = I11 + M(y12 + z 12 ). (1.13)
@seismicisolation
@seismicisolation
1.3 Properties of the Inertia Matrix 7
In a similar way, we can show that

⎛ ⎞
M(y12 + z 12 ) −M x1 y1 −M x1 z 1
I = I + ⎝ −M x1 y1 M(z 12 + x12 ) −M y1 z 1 ⎠ , (1.14)
−M x1 z 1 −M y1 z 1 M(x12 + y12 )
which can be stated as follows:

Theorem 1.1 The inertia matrix of a body about a system of axes is equal to its
inertia matrix relative to a parallel system with the origin at the centre of mass of
the body plus inertia matrix of a point mass equal to the mass of the body placed at
the centre of mass of the body relative to the first system.
From (1.13), we deduce that among all parallel axes, the axis around which the
moment of inertia of a given body is minimal is the one passing through centre of
mass of the body.
1.3.3 Ellipsoid of Inertia
Let I be the inertia matrix of a given body with respect to a Cartesian reference
frame O x yz. Denote by r the position vector of the current mass element dm. The
moment of inertia of that body about an arbitrary straight line O P, passing through
O, whose direction cosines are given by the vector t = (α, β, γ) can be written in
the form (Fig. 1.1)
Fig. 1.1 Body-element

reference frame
@seismicisolation
@seismicisolation

IO P = |r × t|dm

= [|r|2 |t|2 −(r · t)2 ]dm

= [(x 2 + y 2 + z 2 )(α2 + β 2 + γ 2 ) − (αx + β y + γz)2 ]dm
= I11 α2 + I22 β 2 + I33 γ 2 + 2I12 αβ + 2I23 βγ + 2I13 αγ. (1.15)

2
We now take a point Q on O P such that I O P is inversely proportional to O Q
with a proportionality constant of dimensions M L 4 and magnitude equal to 1, i.e.
I O P = 1. 2 . Combining this with (1.15), we get
OQ
2
O Q (I11 α2 + I22 β 2 + I33 γ 2 − 2I12 αβ − 2I23 βγ − 2I13 αγ) = 1,
and denoting by (X, Y, Z ) the coordinates of Q, we can put the last equation in the
form
I11 X 2 + I22 Y 2 + I33 Z 2 + 2I12 X Y + 2I23 Y Z + 2I13 X Z = 1. (1.16)
This means that the point Q lies on a quadratic surface with centre at the origin.
Since the moment of inertia of the body around any line is always positive and finite,
the surface (1.16) is an ellipsoid, usually termed “inertia ellipsoid” or “Cauchy’s
ellipsoid”. Equation (1.16) can be written in short form as
rI · r = 1, (1.17)
where r = (X, Y, Z ).
With the inertia ellipsoid, we gain a clear geometric image of the distribution of
moments of inertia about axes meeting at one point. We know from geometry that
the axes can be rotated to a position Oξηζ, say, so as to eliminate the products of
inertia and hence all mixed terms in (1.16) and write the ellipsoid equation in the
standard form
Aξ 2 + Bη 2 + Cζ 2 = 1, (1.18)
where A, B, C are called the principal moments of inertia. They are the eigenvalues
of the inertia matrix I, i.e. roots of the equation

I11 − λ I12 I13

I12 I22 − λ I23 = 0 (1.19)

I13 I23 I33 − λ
or, in expanded form,
@seismicisolation
@seismicisolation
λ3 − (I11 + I22 + I33 )λ2

+(I11 I22 + I22 I33 + I33 I11 − I122
− I23
2
− I13
2
)λ
−I11 I22 I33 + 2I12 I23 I13 + I11 I23 + I22 I13 + I33 I12
2 2 2
= 0. (1.20)
One should notice here that the coefficients of the last equation, which defines
a unique set of principal moments of inertia, must be invariant with respect to the
rotation of the axes, i.e. they must be independent of the current set of axes at O.
This implies the three relations between the elements of the current inertia matrix at
O and the principal moments of inertia at the same point:
I11 + I22 + I33 = A + B + C,

I11 I22 + I22 I33 + I33 I11 − I12
2
− I23
2
− I132
= AB + BC + AC,
I11 I22 I33 − 2I12 I23 I13 − I11 I23 − I22 I13 − I33 I12 = ABC.
2 2 2
(1.21)
The axes ξ, η, ζ of the ellipsoid (in which I is diagonal) are called the principal
axes of inertia and they are in the directions of the eigenvectors of I , i.e. the solutions
of the linear system
⎛ ⎞⎛ ⎞
I11 − λ I12 I13 ξ
⎝ I12 I22 − λ I23 ⎠ ⎝ η ⎠ = 0 (1.22)
I13 I23 I33 − λ ζ
corresponding to the three eigenvalues λ = A, B, C. The planes ξ = 0, η = 0, ζ = 0

are called principal planes.
Remark 1 As clear from (1.18), the semi-axes of the ellipsoid of inertia are
1 1 1
a = √ ,b = √ ,c = √ ,
A B C
and thus the semi-minor axis of the ellipsoid of inertia is the axis of maximal moment
of inertia and the semi-major axis of the ellipsoid of inertia is the axis of minimal
moment of inertia.
Remark 2 An arbitrary ellipsoid with semi-axes a, b, c cannot be the ellipsoid of
inertia of a real body unless those semi-axes satisfy the three inequalities a12 + b12
, + c12 a12 , c12 + a12 b12 .
1 1
c2 b2
Remark 3 When two of the principal moments of inertia are equal, A = B, say, the
ellipsoid of inertia becomes an ellipsoid of revolution with axis of symmetry ζ. Then
all axes passing through O in the equatorial plane ξη of the ellipsoid are principal
axes.
Remark 4 When A = B = C, the point O is called a spherical point. All axes
through it are principal axes.
@seismicisolation
@seismicisolation
Remark 5 The matrices I¯, I have the same principal axes and in those axes, in
virtue of (1.11),

B +C − A C + A− B A+ B −C
I¯ = diag( Ā, B̄, C̄) = diag , , , (1.23)
2 2 2
and regarding the inverse relation (1.8), we also have
I = diag( B̄ + C̄, C̄ + Ā, Ā + B̄). (1.24)
As we shall see in several occasions in the sequel, it is advantageous to use the second
moments matrix instead of the inertia matrix. Several relations take a simpler form
when expressed in terms of Ī. Moreover,
Remark 6 Contrary to moments of inertia, any given three positive quantities can
serve as the principal second moments of a mass distribution with respect to three
orthogonal axes.
Remark 7 If the axes are arranged so that A B C, then Ā ≤ B̄ ≤ C̄.
Remark 8 The condition of axial dynamical symmetry A = B is equivalent to

Ā = B̄.
Remark 9 The diagonal elements of the inertia matrix defined in (1.8) (expressed
in terms of Ī) automatically satisfy the triangle inequalities
B + C − A = 2 Ā 0, C + A − B = 2 B̄ 0, A + B − C = 2C̄ 0. (1.25)
1.3.4 The Gyration Ellipsoid
The inverse I −1 of the inertia matrix is called the gyration matrix. The gyration
ellipsoid can be defined, in a way analogous to inertia ellipsoid, by the equation
rI−1 ·r = 1, (1.26)
where r = (x, y, z) is a current point on that ellipsoid. Thus, inertia and gyration
ellipsoids have the same principal axes and their semi-axes are reciprocal to each
other. In the principal axes of inertia, the gyration ellipsoid is
x2 y2 z2
+ + = 1. (1.27)
A B C
@seismicisolation
@seismicisolation
1.3.5 Representation of Principal Moments of Inertia
In many problems involving rigid body dynamics, the results depend on the values of
the moments of inertia. This can be represented in several ways. Here, we describe
two of the most frequently used methods:
1.3.5.1 The First Method
In most problems, only the ratios between moments of inertia and not the moments
themselves are significant to the results. The simplest method is to represent moments
of inertia in the plane of their ratios (ξ, η) = ( BA , CA ). The set of admissible points,
i.e. points which correspond to real bodies are those inside or on the border of the
semi-infinite strip shown in Fig. 1.2. This strip is divided into zones with different
relative order of the moments by lines representing cases of two equal moments and
cases when the body is a plane disc. We refer to those zones by Roman numerals.
I) A > B > C,
II) B > A > C,
III) B > C > A,
IV) A > C > B,
V) C > A > B,
VI) C > B > A.
1.3.5.2 The Second Method
Ratios of moments of inertia are represented by points in the plane of the two quan-
tities
Fig. 1.2 Representing

moments of inertia in the
plane of their ratios
@seismicisolation
@seismicisolation
Fig. 1.3 Representing ratios

of moments in the plane
A1 B1
B −C C−A
A1 = , B1 = . (1.28)
A B
Recalling that the largest value of B is A + C, we note that the maximum value of
A1 is 1. Also, the largest value of C is A + B and thus A1 has a minimum of −1.
Similarly, we can show that −1 ≤ B1 ≤ 1. The admissible area in the given plane is
the square depicted in Fig. 1.3.
Comparing the three methods, we note that the circular boundaries between zones
are not comfortable in use. Two of the zones extend to infinity, in which the moment
A approaches zero. In the second method, the boundaries are straight lines, but we
also have two infinite zones. In the third method, all zones are finite and boundaries
are straight, but the quantities defining the plane are not as simple as in the second
method. Thus, we shall prefer to use one of the last two methods when we need to
illustrate the dependence of certain results on moments of inertia of the body.
1.4 Relations Between the Centre of Mass and the Inertia

Ellipsoid
In several circumstances, we deal with a rigid body and results of the study depend
on the parameters: moments of inertia of that body, its mass and the position of its
centre of mass. In the analysis of such results, one has to know the relative order of
parameters. Also, in numerical simulations of motion, one needs to choose definite
numerical values of all the parameters, corresponding to a realistic body.
We know that three moments of inertia A, B and C are subject only to triangle
inequalities. When one of the triangle inequalities for moments of inertia approaches
@seismicisolation
@seismicisolation
1.4 Relations Between the Centre of Mass and the Inertia Ellipsoid 13
equality, say, C = A + B, the body turns into a distribution of mass in the plane
z = 0. Obviously, some condition (or conditions) must hold in the general case and
turn to z = 0 in the case of equality.
This question is answered by the following:
Theorem 1.2 [389] Let O x yz be the Cartesian coordinate system coinciding with
the principal axes of inertia at O of a body of mass M and given principal moments
of inertia A, B and C, respectively. The centre of mass of the body lies inside or on
the ellipsoid
x2 y2 z2
2
+ 2 + 2 = 1, (1.29)
a b c
where
Ā B +C − A 2 B̄ C+ A−B 2 C̄ A+ B −C
a2 = = , b == = ,c = = .
M 2M M 2M M 2M
(1.30)
Remark 10 Note that if C → A + B, then c → 0 and the centre of mass lies in the
ellipse
x2 y2
+ =1 (1.31)
B/M A/M

of the plane z = 0, where A = y 2 dm, B = x 2 dm.
This theorem can be proved1 by applying the conditions that the quadratic form

F(α0 , α1 , α2 , α3 ) = (α0 + x1 α1 + y1 α2 + z 1 α3 )2 dm
is non-negative for all real α0 , α1 , α2 , α3 where dm is the mass element at (x1 , y1 , z 1 )

and the integral is taken over the whole mass of the body. This gives

F(α0 , α1 , α2 , α3 ) = α02 dm

+2α0 (α1 x1 dm + α2 y1 dm + α3 z 1 dm)

+α12 x12 dm + α22 y12 dm + α32 z 12 dm
= M[α02 + 2α0 (α1 x + α2 y + α3 z)]

+ Āα12 + B̄α22 + C̄α32 .
1 For detailed proof, generalization to arbitrary dimension and degenerate cases, see [408].
@seismicisolation
@seismicisolation
For a mass distribution M > 0, and the required condition is that all top left deter-
minants of the matrix ⎛ ⎞
1 x y z
⎜ x Ā 0 0 ⎟
⎜ M ⎟
⎜ ⎟
⎝ y 0 B̄
M
0 ⎠
C̄
z 0 0 M
should be non-negative. It turns out that, unless the mass distribution lies in a plane,
those conditions reduce to three ones of the form
x 2 ≤ a 2 , y 2 ≤ b2 , z 2 ≤ c2 , (1.32)
where
Ā 2 B̄ 2 C̄
a2 = ,b = ,c = (1.33)
M M M
and the last one, resulting from the whole determinant, which we put in the form
x2 y2 z2
+ + ≤ 1. (1.34)
a2 b2 c2
Conditions (1.32) restrict the centre of mass of the mass distribution to lie inside or
on the cuboid centred at the origin and with sides 2a, 2b, 2c. The condition (1.34) is
stronger than the three previous conditions. It tells that the centre of mass lies inside
or on the ellipsoid with semi-axes a, b, c inscribed in this cuboid.
The semi-axes of the ellipsoid in the condition (1.34) have the simplest form
(1.33) in terms of the matrix of second moments. Written in terms of moments of
inertia, they become
B +C − A 2 C+ A−B 2 A+ B −C
a2 = ,b = ,c = . (1.35)
2M 2M 2M
Remark 11 Theorem 1.2 was firstly used in [388], where it helped to establish
that the number of zones of stability and instability of pendulum-like motion cannot
exceed a given finite number. Disregarding such result led to assuming countable
number of zones in some works, e.g. [17]. In some engineering works in order to
simulate the motion of a rigid body, authors have to give the parameters of the
problem certain numerical values. Ignoring the above conditions can lead to the use
of impossible combination of parameters like the location of the centre of mass that
does not belong to any physical body, e.g. [160]. Recently, this result was applied in
[315].
@seismicisolation
@seismicisolation
1.5 Solved Examples 15
1.5 Solved Examples
Example 1
Find the loci of the axes at O about which the moments of inertia are equal.
Solution
Let (ξ, η, ζ) be a point on one of those axes, about which the moment of inertia
is I. The point of intersection of this axis with the inertia ellipsoid
Aξ 2 + Bη 2 + Cζ 2 = 1
lies on the sphere

I (ξ 2 + η 2 + ζ 2 ) = 1.
Subtracting the two equations, we obtain
(A − I )ξ 2 + (B − I )η 2 + (C − I )ζ 2 = 0. (1.36)
This homogeneous quadratic equation determines the required locus, which is a cone
with vertex at O. This cone is called equimomental cone (Fig. 1.4).
For determinacy, we assume the moments of inertia to be organized so that A >
B > C. Equation (1.36) is meaningful only for values of I satisfying C ≤ I ≤ A
(Fig. 1.5).
The shape of the cone depends on the value of I as follows:
(1) For I = C, the equimomental cone degenerates into the line ξ = η = 0, i.e. into
the z−axis.
(2) For values of I on the interval C < I < B, Eq. (1.36) takes the form
Fig. 1.4 Iso-momental lines

on the ellipsoid of inertia
@seismicisolation
@seismicisolation
Fig. 1.5 Projections of

iso-momental lines on the
x z-plane
(A − I )ξ 2 + (B − I )η 2 = (I − C)ζ 2 .
The axis of the cone is the z−axis.

(3) When I = B, the middle moment of inertia, the cone degenerates into the two
planes
(A − B)ξ = ± (B − C)ζ
which intersect at the middle axis of the inertia ellipsoid.

(4) For values of I on the interval B < I < A, Eq. (1.36) can be written in the form
(A − I )ξ 2 = (I − B)η 2 + (I − C)ζ 2 .
The axis of the cone is the x−axis.

(5) Finally, when I = A, the cone degenerates into x−axis.
Example 2
A0 , B0 and C0 are the central principal moments of inertia of a given body (the
principal moments of inertia at its centre of mass). Show that no point exists, at which
the three principal moments of inertia A, B, C are equal (spherical point), unless the
smaller two of the central moments are equal.
Solution
Let the point P(x, y, z) be a spherical point referred to axes at the centre of mass
O of the body. All axes passing through P are principal axes. Choose the axis O P
to be the z−axis. The inertia matrix in parallel axes at P is
⎛ ⎞
A + M z2 0 0
⎝ 0 B + M z2 0 ⎠ ,
0 0 C
@seismicisolation
@seismicisolation
1.5 Solved Examples 17
and its diagonal elements are the principal moments at P. Equating those, we get
A + M z 2 = B + M z 2 = C.
This is possible only when A = B < C, i.e. the z-axis is the axis of dynamical
symmetry and the moment aroundit is the maximal moment at O. In that case, there
are two spherical points (0, 0, ± C−A
M
), both lying on that axis at equal distances
from the centre of mass.
Example 3
(a) Find the cone formed by all axes passing through some point O of a rigid
body, so that each of them is a principal axis of inertia at one of its points.
Solution
(a) Let x yz be a system of axes at the given point O and r = (x, y, z) be the
position vector of the current mass element in that system. Take an axis z passing
through another point O at a distance s from O and two orthogonal axes to complete
a new Cartesian system x y z at O . Let α, β, γ be three orthogonal unit vectors in
the directions of the new axes. We have
x = r · α, y = r · β, z = r · γ − s.
The condition that Oz be a principal axis of inertia of the given body at O gives
two equations
x z dm = 0, y z dm = 0,
which can be written in the form
α · (γ Ī − Msr0 ) = 0, β · (γ Ī − Msr0 ) = 0
in which r0 is the centre of mass and Ī is the second moments matrix. The vector in
brackets is thus orthogonal to both α and β, i.e. parallel to γ. That is
γ × (γ Ī − Msr0 ) = 0. (1.37)
Eliminating s by multiplying scalarly by r0 , we get the equation of the required

locus as
r0 · (γ × γ Ī) = 0. (1.38)
In this form, this equation can be expressed in terms of components with respect to
any axes at O. In particular, if we take x yz as the principal axes of Ī (and also of the
inertia matrix I) at O, then (1.38) takes the form
@seismicisolation
@seismicisolation

x0 y0 z 0

γ1 γ2 γ3 = 0 (1.39)

Āγ1 B̄γ2 C̄γ3
or, equivalently,
x0 (C̄ − B̄)γ2 γ3 + y0 ( Ā − C̄)γ3 γ1 + z 0 ( B̄ − Ā)γ2 γ1 = 0. (1.40)
This homogeneous quadratic equation represents a cone with vertex at O, called

Ampère’s cone. Recalling the relation (1.11), one can write the relations (1.38–1.40)
in terms of the inertia matrix by simply dropping bars from Ī( Ā, B̄, C̄). In particular,
the last equation becomes
x0 (C − B)γ2 γ3 + y0 (A − C)γ3 γ1 + z 0 (B − A)γ2 γ1 = 0. (1.41)
The three principal axes are generators of that cone. Other generators are passing
through the points r0 , r0 Ī−1 , r0 I−1 . The general view of Ampère’s cone is shown
in Fig. 1.6. We shall meet this cone in the sequel, in connection with the stationary
rotations of a rigid body known as Staude’s rotations (see Chap. 8 Sect. 8.3).
Fig. 1.6 The intersection of

Ampere’s cone with the unit
sphere, for
A : B : C :: 2 : 3 : 4 and
x0 : y0 : z 0 :: 2 : 3 : 7
@seismicisolation
@seismicisolation
1.6 Exercises 19
1.6 Exercises
1. Show that for an axis to be a principal axis of inertia of a given rigid body at two
points on it, this axis should pass through the centre of mass of the body, and in
that case, it is a principal axis for all its points.
2. Given a rigid body and a point O on an axis Oz fixed in it, show that one can
always find two orthogonal pairs of axes x, y and x1 , y1 at O orthogonal also to
z, such that
(a) the product of inertia with respect to x, y vanishes;
(b) the moments of inertia about x1 , y1 are equal, and find the angle between
the two pairs.
3. Given a body with central principal moments of inertia A, B, C, show that its
matrix of inertia can be obtained at an arbitrary point of space by replacing
the
body withan ellipsoidof the same mass M, uniform density and semi-axes
( 5(B+C−A)
2M
, 5(C+A−B)
2M
, 5(A+B−C)
2M
), respectively, in the directions of the cen-
tral principal axes.
4. Show that the two circular cross-sections of the gyration ellipsoid (1.27) relative
to principal axes at O lie in the two planes

x C(A − B) ± z A(B − C) = 0.
In Hess’ case, the centre of mass (x0 , y0 , z 0 ) lies on the line drawn from O
perpendicular to one of those cross-sections, show that the following conditions
are satisfied:
y0 = 0, x02 A(B − C) = z 02 C(A − B).
5. Show that the equation of the gyration ellipsoid (1.26), in arbitrary Cartesian axes
at the same origin, can be put in the form

1 x y z

x I11 I12 I13
= 0.
y I12 I22 I23

z I13 I23 I33
6. Show that the equation of the ellipsoid (1.29), inside or on which lies the centre
of mass (x, y, z) of a body in an arbitrary system of axes O x yz, can be written
in the form
M Mx My Mz

M x I¯11 I¯12 I¯13

M y I¯12 I¯22 I¯23 = 0,

M z I¯13 I¯23 I¯33
Ī being the second moments matrix of the body in those axes.
@seismicisolation
@seismicisolation
Chapter 2
Description of Rotation of a Rigid Body
About a Fixed Point
Although in motion of a rigid body about a fixed point, we deal with continuous
change of position, i.e. with a sequence of infinitesimal rotations, it turns out that
the study of finite rotations is essential in understanding several concepts concerning
infinitesimal ones. In this chapter, we present a brief theory of finite rotations to
elucidate their properties and relation to infinitesimal rotations. We shall concentrate
on the use of Euler’s angles and the Euler–Rodrigues or quaternion coordinates as
the most relevant to the Lagrangian approach. For space considerations, some other
alternative descriptions, like Cayley–Klein, are not considered. We also use the most
common notation. For a useful historical survey, the reader may consult [332] and
several references therein.
2.1 The Position of a Rigid Body. Euler’s Angles
The position of a rigid body moving about a fixed point O is completely determined
by the position of a Cartesian coordinate system O x yz fixed in the body and moving
with it with respect to the system O X Y Z fixed in space. The number of parameters
necessary for the description must be three, the number of degrees of freedom of
the rotational motion of the body. Several types of angles are used to this end, and
their choice depends mainly on the suitability for the concrete problem of motion
under consideration. For example, some angles are most suitable to use in the study
of motion of a ship and others are suitable for describing the flight of a plane.
Although, of course, different sets of angles must be equivalent for arbitrary position
of the body, each of them can have specific privileges or drawbacks in regard to
certain application.
One of the most frequently used sets of angles in many applications and especially
for theoretical purposes are Euler’s angles ψ, θ and ϕ. They are defined as follows
(see Fig. 2.1):
https://doi.org/10.1007/978-3-030-96336-1_2
@seismicisolation
@seismicisolation
22 2 Description of Rotation of a Rigid Body About a Fixed Point
Fig. 2.1 Euler’s angles
We begin by the body axes O x yz coinciding with the space-fixed axes O X Y Z ,

and from that position, we give the body system a rotation by an angle ψ (the pre-
cession angle) about the Z -axis, so that the body system takes the position O K Y1 Z .
Then, the last system is given a rotation by an angle θ (the angle of nutation) about
O K . This brings the body system to the position O K Y2 z. We now fix the z-axis in the
body and give the body system a rotation by an angle ϕ (the angle of proper rotation)
about the z-axis to reach its final position O x yz, fixed in the body. In this way, ψ
is the angle of rotation of the body about the space axis Z , θ is the angle between z
and Z and ϕ is the angle of rotation about z. The line O K is the intersection of the
two planes O x y and O X Y. It is called the line of nodes.
Now, let α, β, γ; i, j, k be the unit vectors along the space axes X Y Z and the
body axes x yz, respectively. Let also n be a unit vector along the nodal line O K
and j1 , j2 be unit vectors along OY1 and OY2 , respectively. One can express the
components of the fixed unit vectors with respect to the moving axes. For example,
α = cos ψn − sin ψj1

= (cos ψ cos ϕ − cos θ sin ψ sin ϕ, − cos ψ sin ϕ − cos θ sin ψ cos ϕ, sin θ sin ψ),
β = (sin ψ cos ϕ + cos θ cos ψ sin ϕ, − sin ψ sin ϕ + cos θ cos ψ cos ϕ, − sin θ cos ψ),
γ = (sin θ sin ϕ, sin θ cos ϕ, cos θ). (2.1)
We can also write the unit vector n in the body and space axes, respectively, as
@seismicisolation
@seismicisolation
2.1 The Position of a Rigid Body. Euler’s Angles 23
n = cos ϕi − sin ϕj = cos ψα + sin ψβ. (2.2)
Conversely, we can express the moving unit vectors i, j, k in the fixed (space)
basis. They have the form
i = (cos ψ cos ϕ − cos θ sin ψ sin ϕ, sin ψ cos ϕ + cos θ cos ψ sin ϕ, sin θ sin ϕ),
j = (− cos ψ sin ϕ − cos θ sin ψ cos ϕ, − sin ψ sin ϕ + cos θ cos ψ cos ϕ, sin θ cos ϕ),
k = (sin θ sin ψ, − sin θ cos ψ, cos θ). (2.3)
2.2 The Rotation Matrix
Let (X, Y, Z ) be the coordinates of a point P in the system of axes X Y Z . When the
system X Y Z is rotated by an angle around Z -axis, the point P is displaced to the
new point P (X , Y , Z ). One can easily write
X = X cos + Y sin ,
Y = −X sin + Y cos ,
Z = Z,
which can be put in the matrix form

⎛ ⎞ ⎛ ⎞
X X
⎝ Y ⎠ = R ⎝ Y ⎠,
Z Z
where ⎛ ⎞
cos sin 0
R = ⎝ − sin cos 0 ⎠ .
0 0 1
R is called the rotation matrix. Note that R is an orthogonal matrix (all rows are
orthogonal and also columns), its inverse is its transpose (R−1 = R T ) and its deter-
minant is 1.
Now we consider the characteristic equation of R. That is

cos − λ sin 0

− sin cos − λ 0 = 0.

0 0 1 − λ
@seismicisolation
@seismicisolation
In the final form, this becomes
(1 − λ)(λ2 − 2λ cos + 1) = 0. (2.4)
As we see, one of the characteristic roots equals 1. It corresponds to the eigenvector

(0, 0, 1), which coincides with the axis of the rotation. The other two roots are
complex
cos ± cos2 − 1 = cos ± i sin = e±i .
Thus, the eigenvector vector corresponding to the unit eigenvalue of the rotation
matrix coincides with the axis of the rotation and the argument of the complex pair
of eigenvalues directly expresses the angle of rotation.
Now we apply the same conception to an arbitrary rotation. Let r be a vector
whose components are (X, Y, Z ) in the space-fixed axes and (x, y, z) in the body-
fixed axes. We can find the relations between the components of the vector in the
two systems as follows:
The components of r in the system O K Y1 Z after a rotation by an angle ψ around
the Z -axis ⎛ ⎞ ⎛ ⎞ ⎛ ⎞⎛ ⎞
x1 X cos ψ sin ψ 0 X
⎝ y1 ⎠ = Rψ ⎝ Y ⎠ = ⎝ − sin ψ cos ψ 0 ⎠ ⎝ Y ⎠ (2.5)
z1 Z 0 0 1 Z
and in the system O K Y2 Z after a rotation by an angle θ around O K

⎛ ⎞ ⎛ ⎞ ⎛ ⎞⎛ ⎞
x2 x1 1 0 0 x1
⎝ y2 ⎠ = Rθ ⎝ y1 ⎠ = ⎝ 0 cos θ sin θ ⎠ ⎝ y1 ⎠ . (2.6)
z2 z1 0 − sin θ cos θ z1
Finally, after a rotation by an angle ϕ around the z-axis, we find the components of
r in the body system
⎛ ⎞ ⎛ ⎞ ⎛ ⎞⎛ ⎞
x x2 cos ϕ sin ϕ 0 x2
⎝ y ⎠ = Rϕ ⎝ y2 ⎠ = ⎝ − sin ϕ cos ϕ 0 ⎠ ⎝ y2 ⎠
z z2 0 0 1 z2
⎛ ⎞
X
= R⎝Y ⎠, (2.7)
Z
where
R = Rϕ Rθ Rψ
⎛ ⎞⎛ ⎞⎛ ⎞
cos ϕ sin ϕ 0 1 0 0 cos ψ sin ψ 0
= ⎝ − sin ϕ cos ϕ 0 ⎠ ⎝ 0 cos θ sin θ ⎠ ⎝ − sin ψ cos ψ 0 ⎠
0 0 1 0 − sin θ cos θ 0 0 1
@seismicisolation
@seismicisolation
2.2 The Rotation Matrix 25
⎛ ⎞
cos ψ cos ϕ − cos θ sin ψ sin ϕ sin ψ cos ϕ + cos θ cos ψ sin ϕ sin θ sin ϕ
= ⎝ − cos ψ sin ϕ − cos θ sin ψ cos ϕ − sin ψ sin ϕ + cos θ cos ψ cos ϕ sin θ cos ϕ ⎠ .
sin θ sin ψ − sin θ cos ψ cos θ
(2.8)
Comparing with (2.1), we conclude that

⎛ ⎞
α1 β1 γ1
R = ⎝ α2 β2 γ2 ⎠ ,
α3 β3 γ3
and again we note that R has all rows and columns orthogonal unit vectors. Its
inverse is its transpose R−1 = R T (check that RR T = R T R = δ) and its determinant
equals 1. Note also that the columns of are the components of the fixed unit vectors
referred to the moving (body) axes while its rows are the components of the unit
vectors i, j, k, in the directions of the movable axes x yz, referred to the fixed axes
XY Z.
2.2.1 The Angle of Rotation
We now proceed to form the characteristic equation of the rotation matrix (2.8). After
some manipulations and factorization, we get
|R − λδ| = (1 − λ){λ2 + [(1 − cos θ) − (1 + cos θ) cos(ψ + ϕ)]λ + 1}

θ ψ+ϕ
= (1 − λ){λ2 + 2[1 − 2 cos2 cos2 ]λ + 1} = 0. (2.9)
2 2
From here, we see that one of the characteristic roots of the rotation matrix is 1. The
eigenvector v corresponding to that root satisfies the equation
Ra = a,
i.e. the rotation represented by the matrix R leaves that vector unchanged. This vector
coincides with the axis of the rotation. On the other hand, to obtain the rotation angle
of the rotation , we compare the quadratic factors in (2.9) and (2.4). We get
θ ψ+ϕ
cos = 2 cos2 cos2 − 1. (2.10)
2 2
This can be also written in the form
θ ψ+ϕ
cos = cos cos (2.11)
2 2 2
in which we have chosen positive sign.
@seismicisolation
@seismicisolation
Remark: In spite of their simplicity, Euler’s angles suffer from the defect that the
two angles ψ and ϕ lose their independence when the third angle θ takes one of the
values 0 or π. In those cases, the rotation matrix takes the form
⎛ ⎞
cos(ψ ± ϕ) sin(ψ ± ϕ) 0
⎝ − sin(ψ ± ϕ) cos(ψ ± ϕ) 0 ⎠
0 0 1
which involves not the two angles, but their sum when θ = 0 and difference when
θ = π.
2.3 Description of Finite Rotation
We have just seen that the finite rotation can be always and completely represented by
a rotation matrix. On the other hand, it is evident that such rotation can be completely
determined by giving the axis of rotation and the angle of rotation about that axis,
i.e. one can say the rotation is determined by a scalar quantity and a direction.
Nevertheless, it cannot be represented in the full sense by a vector, since an essential
rule of vector algebra, the commutation rule, is not followed by matrices. This means
that performing two consequent rotations R1 and then R2 gives different resultant
from that of reverse order R2 and then R1 . This can be clearly illustrated by the
following
Example: Let us perform to the body in Fig. 2.2a two consecutive rotations, each
by a right angle,1 the first about the x-axis and the second about the y-axis. Figure
2.2b shows what we get in this case, but Fig. 2.2c shows the completely different
result of performing the rotations in the reverse order.
Analytically, R1 and R2 can be represented by the matrices
⎛ ⎞ ⎛ ⎞
1 0 0 0 0 −1
R1 = ⎝ 0 0 1 ⎠ , R2 = ⎝ 0 1 0 ⎠ .
0 −1 0 10 0
Resultant of the first sequence is

⎛ ⎞
010
S = R2 R1 = ⎝ 0 0 1 ⎠
100
and the second sequence
1We mean rotation by an angle described in the positive sense about an axis, i.e. counterclockwise
as viewed from the positive end of that axis.
@seismicisolation
@seismicisolation
2.3 Description of Finite Rotation 27
Fig. 2.2 Finite rotations are not commutative
⎛ ⎞
0 0 −1
S = R1 R2 = ⎝ 1 0 0 ⎠ .
0 −1 0
The two matrices S and S are obviously different. We can go a little further and
construct the characteristic equation for S. That is
1 − λ3 = (1 − λ)(λ2 + λ + 1) = 0,
and its roots are 1, e±i 3 . This means that the angle of the rotation S is equal to
2π 2π
3
.
To find the axis of the rotation, we solve the equations
Sv = v ,
which gives the column vector ⎛ ⎞

1
v = ⎝ 1 ⎠ .
1
Thus, S is a rotation by angle 2π

3
around the axis in the direction parallel to the vector
v =(1, 1, 1).
Similarly, one can show that S is a rotation by angle 2π 3
around the vector
v =(1, 1, −1).
@seismicisolation
@seismicisolation
Fig. 2.3 Geometry of a rotation vector
2.4 Representation of Finite Rotation by Means of a Vector
The above example shows that finite rotation cannot be completely represented by a
vector as we represent the position vector or the velocity of a particle. Thus, we shall
try now to find the formula that expresses the finite rotation as a vector quantity and
to find a suitable rule for the resultant of two rotations, a rule that must account for
the non-commutation of rotations.
2.4.1 The Rotation Vector

−→
Let us begin with some vector r = O P and an axis O Q with a unit vector e in its
direction. The rotation of r by an angle around O Q in the positive direction carries
−−→
r to its new position r = O P and the point P along the circular arc
P P to P (4.1).

Let also O be the centre of P P . Our aim now is to express r in terms of r and the
angle and direction of the rotation (see Fig. 2.3).
The plane Fig. 2.4 shows the circle O P P . O R is orthogonal to P P and RS is

orthogonal to O P. Note that e is the outward unit vector normal to the plane of the
figure and S R = O P sin 2 cos 2 , so that S R = e × O P sin 2 cos 2 .
From simple geometry, we find that

P P = 2P R = 2(P S + S R) = 2[−O P sin2 + e × O P sin cos ].
2 2 2
But since
O P = r − (r · e)e = (e · e)r − (e · r)e = −e × (e × r),
@seismicisolation
@seismicisolation
2.4 Representation of Finite Rotation by Means of a Vector 29
Fig. 2.4 Rotation angle
then we obtain

P P = 2[sin2 e × (e × r)+ sin cos e × r],
2 2 2
so that we can finally write
r = r + P P

= r + 2[sin cos e × r+ sin2 e × (e × r)]. (2.12)
2 2 2
This is the Rodrigues formula, which expresses the rotated vector r in terms of the
initial vector r, the angle of rotation and the direction e of the rotation axis. It can
be written also in the form
r = r + (sin )e × r+(1 − cos )e × (e × r). (2.13)
It is valid for arbitrary angle and arbitrary direction of the rotation. As expected, for
a point on the axis of rotation, r = e and r = r. Also, a rotation with an angle 2π
brings all points of space to their initial positions.
To push forward the concept of a vector representing a finite rotation, we assume
that the rotation angle = π, i.e. cos 2 = 0. Then we can write (2.12) in the form

r = r + 2 cos2 [tan e × r+ tan2 e × (e × r)]
2 2 2
2
=r+ [tan e × r+ tan2 e × (e × r)].
1 + tan2 2 2 2
Introducing the notation

ρ = tan e, (2.14)
2
@seismicisolation
@seismicisolation
then we rewrite the last formula as

2
r = r + ρ × [r + ρ × r]. (2.15)
1 + ρ2
Using (2.14), one can verify that
(1) ρ( + 2π, e) = ρ(, e).

(2) ρ(−, −e) = ρ(, e).
Those properties are geometrically obvious.
(3) Formula (2.14) is not suitable for expressing any rotation with an angle π about
any axis. That is the singular point of the function tan . This is not related to
the rotation itself, but due to the way of representing the rotation as a vector in
(2.14). The previous formulas (2.12), (2.13) are still valid for the angle = π.
(4) Finite rotation is thus represented by a vector, which may be written in terms of
its components as ρ = (ρ1 , ρ2 , ρ3 ) or ρ = ρ1 i + ρ2 j + ρ3 k, but this determines
the magnitude and direction of the rotation and does not mean at all that the
rotation is the resultant of its parts, or equivalent to any sequence of those parts.
(5) It is evident that rotation vectors do not commute and cannot be summed accord-
ing to rules of vector algebra. However, it can be easily shown that infinitesimally
small rotations do commute and obey the rule of summation of vectors.
Let ρ1 be a rotation by a small angle 1 the rotation vector ρ1 = tan 21 e1 = 21 e1 .
After neglecting nonlinear terms in the rotation vector, formula (2.15) takes the
form
r = r + 2ρ1 × r. (2.16)
If ρ1 is followed by another small rotation ρ2 = 22 e2 , the vector r is transformed

to
r = r + 2ρ2 × r
= r + 2ρ1 × r + 2ρ2 × (r + 2ρ1 × r). (2.17)
Neglecting the nonlinear term, we get
r = r + 2(ρ1 +ρ2 ) × r. (2.18)
Small rotations are summed according to vector addition rule, and their sum does
not depend on the order of the rotations.
Now we return to formula (2.14) to see how a rotation ρ = tan 2 e acts on the unit
vectors α, β, γ fixed in the directions of X Y Z and bring them to be coincident with
the unit vectors i, j, k, respectively. According to (2.15), we have
@seismicisolation
@seismicisolation
2.4 Representation of Finite Rotation by Means of a Vector 31
2
i = α+ ρ × (α + ρ × α),
1 + ρ2
2
j=β+ ρ × (β + ρ × β),
1 + ρ2
2
k=γ+ ρ × (γ + ρ × γ). (2.19)
1 + ρ2
One immediately notices that
ρ · i = ρ · α, ρ · j = ρ · β, ρ · k = ρ · γ,
i.e. the components of the rotation vector are the same in the directions of the initial
and final axes. We shall denote those components by ρ = (ρ1 , ρ2 , ρ3 ).
We now express the rotation matrix in terms of the components of the rotation
vector
⎛ ⎞
i·α i·β i·γ
R = ⎝ j·α j·β j·γ ⎠
k·α k·β k·γ
⎛ ⎞
ρ22 +ρ23
1 − 2 1+ρ 2 ρ11+ρ
ρ2 +ρ3
2 ρ11+ρ
ρ3 −ρ2
⎜ ρ ρ −ρ ⎟
2 2 2
=⎜ ρ21 +ρ23
⎝ 2 1+ρ2 1 − 2 1+ρ2 2 1+ρ22 2 ⎠ .
1 2 3 ρ2 ρ3 +ρ1 ⎟
(2.20)
ρ1 ρ3 +ρ2 ρ2 ρ3 −ρ1 ρ1 +ρ2
2 1+ρ2 2 1+ρ2 1 − 2 1+ρ2
2.5 Hamilton–Rodrigues’ Parameters
We now introduce the four quantities λ0 , λ1 , λ2 , λ3 to express the three components

of the rotation vector, such that
(λ1 , λ2 , λ3 )
(ρ1 , ρ2 , ρ3 ) = . (2.21)
λ0
As we have one redundant parameter, we assume that the new parameters satisfy the
condition
λ20 + λ21 + λ22 + λ23 = 1, (2.22)
so that the end of the four-dimensional vector = (λ0 , λ1 , λ2 , λ3 ) lies on a three-

dimensional sphere of unit radius. This implies the relation
λ0 λ̇0 + λ1 λ̇1 + λ2 λ̇2 + λ3 λ̇3 = 0. (2.23)
From (2.21), we calculate
@seismicisolation
@seismicisolation
λ21 + λ22 + λ23 1

1 + ρ2 = 1 + = 2, (2.24)
λ20 λ0
and substituting (2.21) in (2.20) and using the last relation, we obtain the expression
of the rotation matrix
⎛ 2 ⎞
λ0 + λ21 − λ22 − λ23 2(λ0 λ3 + λ1 λ2 ) 2(−λ0 λ2 + λ1 λ3 )
R = ⎝ 2(−λ0 λ3 + λ1 λ2 ) λ20 + λ22 − λ21 − λ23 2(λ0 λ1 + λ2 λ3 ) ⎠ . (2.25)
2(λ0 λ2 + λ1 λ3 ) 2(−λ0 λ1 + λ2 λ3 ) λ20 + λ23 − λ21 − λ22
This form of the rotation matrix is more symmetric than that in terms of the
rotation vector or in terms of Euler’s angles. Moreover, it does not have the problem
of degeneration of Euler’s angles at θ = 0 or π, nor the singularity of the rotation
vector corresponding to a rotation by an angle π. This makes Euler–Rodrigues’
parameters in certain problems appropriate for use as variables describing motion
and finite rotations of the rigid body.
On the other hand, one can readily notice that the two sets of the Hamilton–
Rodrigues parameters ± correspond to the same rotation matrix. Thus, any expres-
sion designating a quantity of physical meaning should contain only even terms in
, otherwise it will be double-valued on the group of rotations S O3. This remark
will have some implications in later chapters.
Remark 12 The expression (2.25) for the rotation matrix can be decomposed into
three parts of simpler structure (two symmetric and one antisymmetric):
⎛ ⎞ ⎛ 2 ⎞
100 λ1 λ 1 λ 2 λ 1 λ 3
R = (2λ20 − 1) ⎝ 0 1 0 ⎠ + 2 ⎝ λ1 λ2 λ22 λ2 λ3 ⎠
001 λ1 λ3 λ2 λ3 λ23
⎛ ⎞
0 λ3 −λ2
+2λ0 ⎝ −λ3 0 λ1 ⎠ (2.26)
λ2 −λ1 0
or in the shorter tensor form
Ri j = (2λ20 − 1)δi j + 2λi λ j + 2λ0 i jk λk , (2.27)
where δ is the Kronecker delta and is the Levi-Civita tensor.

Remark 13 It is clear from (2.25) that the points (λ0 , λ1 , λ2 , λ3 ) and (−λ0 , −λ1 ,
−λ2 , −λ3 ) represent the same rotation matrix. The sphere (2.22) covers the con-
figuration space of the rotating body twice. The configuration space can, thus, be
represented by one half of that sphere, say, the half on which λ0 0.
Remark 14 From (2.24) and (2.14), we have
1
= 1 + ρ2 = 1 + tan2 = sec2 ,
λ20 2 2
@seismicisolation
@seismicisolation
2.5 Hamilton–Rodrigues’ Parameters 33
so that

λ0 = cos , (2.28)
2
and hence
(λ1 , λ2 , λ3 ) = λ0 ρ

= cos tan e
2 2

= sin e.
2
Thus, we can write the following expression for the Euler–Rodriguez parameters

(λ0 , λ1 , λ2 , λ3 ) = (cos , sin e). (2.29)
2 2
2.6 The Angular Velocity Vector
During motion of the body, Euler’s angles change with time t, and also the rotation
matrix. To express the time derivatives of an arbitrary quantity, one may first check
the following relations using expressions (2.1)
∂α ∂β ∂γ
∂ψ
= α × γ, ∂ψ
= β × γ, ∂ψ
= 0,
∂α ∂β ∂γ
∂θ
= α × n, ∂θ
= β × n, ∂θ
= γ × n, (2.30)
∂α ∂β ∂γ
∂ϕ
= α × k, ∂ϕ
= β × k, ∂ϕ
= γ × k.
From those, we get
∂α ∂α ∂α
α̇ = ψ̇ + θ̇ + ϕ̇ (2.31)
∂ψ ∂θ ∂ϕ
= α × γ ψ̇ + α × nθ̇ + α × kϕ̇
= α × (ψ̇γ+θ̇n+ϕ̇k). (2.32)
Let us now introduce the notation
ω = ψ̇γ+θ̇n+ϕ̇k. (2.33)
The vector ω is called the angular velocity of the body and it is, in fact, the usual
vector sum of the three vectors ψ̇γ,θ̇n,ϕ̇k, which represent the angular velocities
ψ̇, θ̇ and ϕ̇ about the axes Z , K and z, respectively. In a similar way, we can get two
expressions for the derivatives of β and γ, so that we can write
@seismicisolation
@seismicisolation
α̇ = α × ω, β̇ = β × ω, γ̇ = γ × ω. (2.34)
Those equations satisfied by α, β, γ are called Poisson’s equations and they

express the constancy of those vectors in space, as we shall see soon. They play an
important role in the dynamics of rigid body as will be seen in due course.
2.7 Space and Relative Time Rates of Change of a Vector
During the motion of the body, the three unit vectors i, j, k fixed in the body along
the axes x, y, z change with time. The rate of change of one of them, i, say, is the
di
velocity dt of its end point in space. Hence, we have
di dj dk
= ω × i, = ω × j, = ω × k, (2.35)
dt dt dt
where ω is the instantaneous angular velocity of the body.
Now, let u be a vector given by its components (u 1 , u 2 , u 3 ) in the body system of
axes, so that we write
u = u 1 i+u 2 j+u 3 k. (2.36)
The time derivative of this vector is

du du 1 du 2 du 3 di dj dk
= i+ j+ k+u 1 +u 2 +u 3 .
dt dt dt dt dt dt dt
Using (2.35) in the last three terms, we get
du du 1 du 2 du 3
= i+ j+ k + u 1 ω × i+u 2 ω × j + u 3 ω × k
dt dt dt dt
du 1 du 2 du 3
= i+ j+ k + ω × (u 1 i + u 2 j + u 3 k).
dt dt dt
Now we introduce the notation u̇ = du dt

1
i + du
dt
2
j + du
dt
3
k, i.e. u̇ is the time derivative
of the vector u as if the unit vectors i, j, k were constant vectors, or as seen by an
observer fixed in the body and moving with it. This derivative will be called the
relative derivative or the relative rate of change of u. The last relation becomes
du
= u̇+ω×u. (2.37)
dt
Thus, we have split the space derivative into two terms: the relative derivative u̇ in
the body system and the term ω × u resulting from the rotation of the body system.
When ω = 0 the two derivatives coincide.
@seismicisolation
@seismicisolation
2.7 Space and Relative Time Rates of Change of a Vector 35
As an example, we apply this rule to the three vectors α, β and γ fixed in space.
As dα
dt
= dβ
dt
= dγ
dt
= 0, we have
α̇ + ω × α = 0, β̇ + ω × β = 0, γ̇ + ω × γ = 0, (2.38)
so that we again obtain Poisson’s equations (2.34).
2.7.1 Components of the Angular Velocity in the Body Axes

and Space Axes
The direction of the angular velocity at the fixed point determines a line called the
instantaneous axis of rotation. Points of the body lying on that line at any moment of
time are instantaneously at rest. The magnitude of ω is a measure of the angular speed
of rotation of the body. In case of rotation about a fixed axis, the angular velocity is
the time rate of change of the angle of rotation about that axis, but it is not possible
in general to write a rotation angle such that the angular velocity is represented as
its rate of change.
We shall denote by p, q, r and p , q , r the components of the angular velocity
ω in the moving and in the fixed axes, respectively. Using (2.2) and (2.3) together
with (2.33), we have
p = ψ̇ sin θ sin ϕ + θ̇ cos ϕ,

q = ψ̇ sin θ cos ϕ − θ̇ sin ϕ,
r = ψ̇ cos θ + ϕ̇, (2.39)
and similarly, we write
p = ϕ̇ sin θ sin ψ + θ̇ cos ψ,

q = −ϕ̇ sin θ cos ψ + θ̇ sin ψ,
r = ϕ̇ cos θ + ψ̇. (2.40)
2.7.2 The Use of the Euler–Rodrigues Parameters
From (2.11) and (2.28), we have λ0 = cos 2θ cos ψ+ϕ 2

. One can easily obtain expres-
sions for the other three parameters by comparing corresponding elements of the
rotation matrix R in formulas (2.8) and (2.25). It is even easier to compare the anti-
symmetric parts, e.g. one can see from (2.26) and (2.8) that
@seismicisolation
@seismicisolation
R23 − R32 sin θ(cos ϕ + cos ψ)

λ1 = =
4λ0 4 cos 2θ cos ψ+ϕ
2
θ ψ−ϕ
= sin cos
2 2
and so on, so that we get the expressions
θ ψ−ϕ θ ψ−ϕ
λ1 = sin cos , λ2 = sin sin ,
2 2 2 2
θ ψ+ϕ θ ψ+ϕ
λ3 = cos sin , λ0 = cos cos . (2.41)
2 2 2 2
The angular velocity has the expression
p = 2(λ0 λ̇1 − λ1 λ̇0 + λ3 λ̇2 − λ2 λ̇3 ),

q = 2(λ0 λ̇2 − λ2 λ̇0 + λ1 λ̇3 − λ3 λ̇1 ),
r = 2(λ0 λ̇3 − λ3 λ̇0 + λ2 λ̇1 − λ1 λ̇2 ). (2.42)
This may be written in the vector form
ω = 2[λ0 λ̇ − λ̇0 λ − λ × λ̇], (2.43)
where λ denotes the three-dimensional vector (λ1 , λ2 , λ3 ). In this notation, (2.22)

and (2.23) take the form
λ20 + λ2 = 1, (2.44)
λ0 λ̇0 + λ · λ̇ = 0. (2.45)
The formula (2.43) together with (2.44), (2.45) can be used to obtain a remarkable
expression for the square of the angular velocity. Squaring both sides of (2.43) and
noting that the third term on the right-hand side is orthogonal to the other two, we
write
2
ω 2 = 4[λ20 λ̇ − 2λ0 λ̇0 λ · λ̇λ̇ + λ̇20 λ2 + |λ × λ̇|2 ]
2 2
= 4[λ20 λ̇ − 2λ0 λ̇0 λ · λ̇ + λ̇20 λ2 + λ2 λ̇ − (λ · λ̇)2 ]
2
= 4[(λ20 + λ2 )λ̇ + 2λ20 λ̇20 +λ̇20 λ2 − (λ0 λ̇0 )2
2
= 4[(λ20 + λ2 )λ̇ + (λ20 + λ2 )λ̇20 ]
2
= 4(λ20 + λ2 )(λ̇20 + λ̇ )
2
= 4(λ̇20 + λ̇ ),
so that, finally, we have
@seismicisolation
@seismicisolation
2.7 Space and Relative Time Rates of Change of a Vector 37
˙ 2.
ω 2 = p 2 + q 2 + r 2 = 4(λ̇21 + λ̇22 + λ̇23 + λ̇20 ) = 4 (2.46)
That is four times the square of the speed of the point = (λ0 , λ1 , λ2 , λ3 ) moving
on the unit sphere (2.22).
2.8 Quaternions and Representation of Finite Rotation
Quaternions, or hypercomplex numbers, discovered by Hamilton, are a generalization

of the ordinary complex number system. A quaternion is composed of one real
component and three imaginary ones. A general quaternion can be written in the
form
Q = (a, A1 , A2 , A3 ) = a + A1 i + A2 j + A3 k, (2.47)
where i, j, k are imaginary units satisfying the multiplication rules
i 2 = j 2 = k 2 = 1,
i j = − ji = k, jk = −k j = i, ki = −ik = j. (2.48)
In (2.47), the first part a is an ordinary real part and the remaining parts can be viewed
as a vector A = A1 i + A2 j + A3 k, and thus, we write
Q = a + A. (2.49)
Now it is easy to check that the product of the quaternions Q and Q = a + A

according to the rules (2.48) can be put in the usual form using scalar and vector
products of vectors as
Q Q = aa − A · A + aA + a A + A × A , (2.50)
and if we define the conjugate quaternion Q̄ = a − A, we easily note that the quantity
Q Q̄ = Q̄ Q = a 2 + A · A (2.51)
is a positive real number which we adopt as the squared magnitude of the quaternion

|Q| = Q Q̄ = a 2 + A21 + A22 + A23 . (2.52)
From the last, we get that for a non-zero quaternion Q
Q̄
Q = 1. (2.53)
|Q|2
@seismicisolation
@seismicisolation
That is the multiplicative inverse of Q (which satisfies Q Q −1 = Q −1 Q = 1) is

Q −1 = |Q|Q̄
2.
Consider now the quaternion formed by Euler–Rodrigues’ parameters Q = λ0 +

λ1 i + λ2 j + λ3 k. It can be put in the form

Q = λ0 + λ0 ρ = cos (1 + ρ). (2.54)
2
Note that

|Q| = 1, Q −1 = cos (1 − ρ). (2.55)
2
Let r be a vector, i.e. a quaternion with zero real part. Applying (2.50), we calculate
the product

QrQ −1 = cos2 (1 + ρ)[(0 + r)(1 − ρ)]
2

= cos2 (1 + ρ)(ρ · r + r + ρ × r)
2
2
= cos [ρ · r + r + ρ × r − ρ · (r + ρ × r)
2
+(ρ·r)ρ + ρ × (r+ρ × r)]

= cos2 [r + 2ρ × r + (ρ · r)ρ + ρ × (ρ × r)]
2

= cos2 [r + 2ρ × r + 2ρ × (ρ × r) + ρ2 r]
2

= cos2 [(1 + ρ2 )r + 2ρ × r + 2ρ × (ρ × r)]
2
2
= cos (1 + ρ2 )[],
2
and using (2.14), we finally get
2ρ
QrQ −1 = r + ×(r + ρ × r). (2.56)
1 + ρ2
Comparing this formula with (2.15), we note that the rotation ρ transfers the vector
r to
r = QrQ −1 , (2.57)
so that the rotation ρ is completely determined by the quaternion Q of unit magnitude.

Also, we have
@seismicisolation
@seismicisolation
2.8 Quaternions and Representation of Finite Rotation 39
Q −1 r Q = Q −1 (QrQ −1 )Q
= (Q −1 Q)r(Q −1 Q)
= r,
so that the inverse of the rotation is given by the quaternion Q −1 = Q̄.
2.9 Composition of Two Rotations
Formula (2.57) is due to Cayley. Although equivalent to (2.15), it is much simpler in

dealing with finite rotations. We use it now to obtain a formula for the composition
of two rotations.
Consider a rotation through an angle 1 around the axis in the direction e1 . This
rotation is completely described either by the rotation vector ρ1 = tan 21 e1 or by the
quaternion q1 = λ0 + λ1 i + λ2 j + λ3 k. Let the rotation vector ρ2 = tan 22 e2 and
the quaternion q2 = μ0 + μ1 i + μ2 j + μ3 k correspond to another rotation through
an angle 2 around e2 . We have
1 2
q1 = cos (1 + ρ1 ), q2 = cos (1 + ρ2 ). (2.58)
2 2
The vector r is transformed by the first rotation to
r = q1 rq1−1 (2.59)
and then by the second rotation to
r = q2 r q2−1 = q2 q1 rq1−1 q2−1 = (q2 q1 )r(q2 q1 )−1 . (2.60)
Thus, the resultant rotation corresponds to the quaternion
Q = q2 q1
= (Q 0 , Q 1 , Q 2 , Q 3 ), (2.61)
where
Q 0 = λ0 μ0 − λ1 μ1 − λ2 μ2 − λ3 μ3 ,
Q 1 = λ1 μ0 + λ0 μ1 + λ3 μ2 − λ2 μ3 ,
Q 2 = λ2 μ0 + λ0 μ2 + λ1 μ3 − λ3 μ1 ,
Q 3 = λ3 μ0 + λ0 μ3 + λ2 μ1 − λ1 μ2 . (2.62)
If we like to express the resultant rotation in terms of the rotation vectors, we use
(2.58) to write
@seismicisolation
@seismicisolation
1 2
Q = cos cos (1 + ρ2 )(1 + ρ1 )
2 2
1 2
= cos cos [(1 − ρ1 · ρ2 ) + ρ1 + ρ2 + ρ2 × ρ1 ]
2 2
1 2 ρ + ρ2 + ρ2 × ρ1
= cos cos (1 − ρ1 · ρ2 )[1 + 1 ]. (2.63)
2 2 (1 − ρ1 · ρ2 )
Comparing this with (2.58), we can write the resultant rotation quaternion in the
form

Q = cos (1 + ρ). (2.64)
2
We find
ρ1 + ρ2 + ρ2 × ρ1
ρ= (2.65)
(1 − ρ1 · ρ2 )
and
1 2
cos = cos cos (1 − ρ1 · ρ2 )
2 2 2
1 2 1 2
= cos cos − sin sin e1 · e2
2 2 2 2
1 2 1 2
= cos cos − sin sin cos χ. (2.66)
2 2 2 2
Rodrigues’s formula (2.65) gives the resultant rotation vector and (2.66) gives the
angle of the resultant rotation in terms of the two rotation angles and the angle χ
between the two axes of the rotations.
Note that Rodrigues’s formula (2.65) is not valid when ρ1 · ρ2 = 1. In that case,
from (2.66), we see that the rotation angle = π. This is expected whenever we
deal with vectors of rotation.
2.10 Exercises
1- Show that the resultant of two half turns around different axes intersecting at an
angle θ is equivalent to a rotation at an angle 2θ around the axis orthogonal to the
two axes.
2- Put the formula (2.42) for the angular velocity in the quaternion form
ω = 2q −1 q̇, (2.67)
and hence prove the formula (2.46).
@seismicisolation
@seismicisolation
Chapter 3
The Classical Problem: The Motion of a
Heavy Rigid Body About a Fixed Point
In the present chapter, we present detailed analysis of the classical problem of motion
of a rigid body about a fixed point under the action of its own weight. This problem
has a long history that began with the work of Euler and continued to the present
day. Various powerful methods belonging to eminent specialists in mechanics and
mathematics were applied to this problem without stopping, sometimes successfully
and sometimes with less success. The list of basic contributors to this problem from
our perspective, the construction of integrable cases, will be clearly presented as we
proceed through this preliminary chapter and the following two chapters, after we
make clear the meanings of general integrable, conditional integrable and particular
solvable cases of the classical problem. This chapter is mainly concerned with basic
concepts and various forms of the equations of motion, each of which would be more
suited for use in certain investigations of the classical problem.
3.1 Equations of Motion
In this section, we derive the equation of rotational motion of the rigid body about a
fixed point, under the action of arbitrary forces, which are not necessarily conservative
or even having a potential. For such general setting, the Lagrangian approach is not
a suitable choice and it is preferable to use ordinary vector mechanics. Denote by
r the position vector of a mass element dm and by v = dr dt
its velocity. The angular
momentum of the body, denoted here by G, is the sum of moments of momenta of
the elements about the origin O, the fixed point of the body,

G= r × (vdm), (3.1)

https://doi.org/10.1007/978-3-030-96336-1_3
@seismicisolation
@seismicisolation
42 3 The Classical Problem: The Motion of a Heavy Rigid Body About a Fixed Point
where the integral is taken over the whole mass of the body. Differentiating the last
relation, we get
dG dv
= v × vdm + r × dm. (3.2)
dt dt
The first integral vanishes, and from the equation of motion of the mass element dm,
we have
dv
dm = dF + dF ,
dt
in which dF, dF are, respectively, the resultant external and internal forces exerted
on that element. Inserting this into (3.2), we write

dG
= r × (dF + dF ). (3.3)
dt
Since the internal

forces appear only in equal and opposite pairs, their overall moment
vanishes, i.e. r × dF = 0. Thus, we finally have
dG
= L, (3.4)
dt

where L = r × dF is the resultant moment of all the external forces acting on the
body about the fixed point. This is the equation of rotational motion of the rigid
body about a fixed point, under the action of arbitrary forces with moment L. It is
curious that this equation is similar to the equation of motion of a particle dP
dt
= F,
but replacing the Linear momentum P and the force F by the angular momentum
and the moment of forces about the fixed point.
3.2 The Heavy Rigid Body
Equation (3.4) is quite general. It is valid for an arbitrary rigid body subject to arbitrary
system of forces. In this chapter, we are concerned with the simplest case of motion
of a body subject only to its own weight. For such a body, let g be the intensity of
the gravity field directed vertically downwards. We have

L= r × (gdm)

= rdm×g.
Recalling the definition of the centre of gravity (the centre of mass) of the body in
Chap. 1, we write the last relation in the form
@seismicisolation
@seismicisolation
3.2 The Heavy Rigid Body 43
L = Mr0 × g. (3.5)
Without loss of generality, for most applications, one can take the Z -axis in the
vertical direction upwards, so that the gravity field becomes
g = −gγ. (3.6)
Thus, in the case of a heavy rigid body, the equation of motion takes the form
dG
= −Mgr0 × γ. (3.7)
dt
3.3 The Angular Momentum of a Rigid Body
The mass element at the point r has velocity
v = ω×r. (3.8)
Recalling the definition (3.1), we write

G= r × (ω×r)dm

= [r2 ω−(ω · r)r]dm.
The i-th component is

3
G i = ωi r dm −
2
ω j r j ri dm
j=1

3
= ωj [r2 δi j − ri r j ]dm
j=1

3
= ω j Ii j ,
j=1
where I = (Ii j )i,3 j=1 is the inertia matrix in the system of axes at the fixed point O.
Making use of the symmetry of the inertia matrix, we write the last expressions in
the form
G = ωI. (3.9)
@seismicisolation
@seismicisolation
Remark 15 In most textbooks, the last relation is usually written as
G = Iω,
where G is a column vector. We use the notation (3.9) to express G as a normal

row vector. This is especially convenient in applying rules of vector algebra to that
vector.
The relation (3.9) can be written in the expanded form as

⎛ ⎞
I11 I12 I13
G = ( p, q, r ) ⎝ I12 I22 I23 ⎠
I13 I23 I33
= (I11 p + I12 q + I13r, I12 p + I22 q + I23 r, I13 p + I23 q + I33 r ). (3.10)
In the inertial system of axes at O, the body moves and its orientation changes with
time. Equation (3.7) then involves 12 variable quantities: six moments and products
of inertia, three coordinates of the mass centre and three components of the angular
velocity. This makes the equations of motion quite complicated and impractical to
use.
The system of axes fixed in the body with origin at the fixed point O enjoys
the advantage that the inertia matrix is constant and also the position vector of the
centre of mass. This makes it plausible to use a coordinate system fixed in the body
to express Eq. (3.7) in it. A question arises, how to express that equation which is
derived in the inertial system of axes in the body system? The answer will be given
soon.
3.4 Kinetic Energy of a Moving Body
Summing the kinetic energy of mass elements and making use of the formulas of the
last subsection, we get

1 2
T = v dm
2

1
= v · (ω×r)dm
2

1
= ω · [r × v]dm
2
1
≡ ωI · ω. (3.11)
2
In expanded form, this means
@seismicisolation
@seismicisolation
3.4 Kinetic Energy of a Moving Body 45
1
T = (I11 p 2 + I22 q 2 + I33 r 2 + 2I12 pq + 2I23 qr + 2I13 pr ). (3.12)
2
3.5 Equations of Motion in the Moving Coordinate System
Let us now write Eq. (3.7) in the body system. It takes the form, called Euler’s
equation,
Ġ + ω × G = Mgγ × r0 . (3.13)
In addition to the vector ω, this equation involves the vertical unit vector γ, which
has variable components in the body system O x yz. Being constant in space, the
vector γ satisfies dγ
dt
= 0 in the inertial frame. In the body system, this is equivalent
to
γ̇ + ω × γ = 0, (3.14)
which bears the name of Poisson’s equation.

The pair of vector Eqs. (3.13) and (3.14), known as the Euler–Poisson equations,
constitute a closed system of six scalar first-order differential equations in six vari-
ables, which can be chosen as either ω, γ or G, γ.
3.5.1 The Use of the Variables ω, γ. Special Axes Related to

the Inertia Matrix
In that case, using (3.9), we write the Euler–Poisson equation as
ω̇I + ω × ωI = Mgγ × r0 , γ̇ + ω × γ = 0, (3.15)
which is the most commonly used form of those equations. For arbitrary choice of
the body axes, they have the expanded form
I11 ṗ + I12 q̇ + I13 ṙ + (I33 − I22 )qr + I23 (q 2 − r 2 ) + p (I13 q − I12 r ) = Mg(z 0 γ2 − y0 γ3 ),

I12 ṗ + I22 q̇ + I23 ṙ + (I11 − I33 ) pr + I13 r 2 − p 2 + q (I12 r − I23 p) = Mg(x0 γ3 − z 0 γ1 ),

I13 ṗ + I23 q̇ + I33 ṙ + (I22 − I11 ) pq + I12 p 2 − q 2 + r (I23 p − I13 q) = Mg(y0 γ1 − x0 γ2 ),

(3.16)
γ̇1 + qγ3 − r γ2 = 0, γ̇2 + r γ1 − pγ3 = 0, γ̇3 + pγ2 − qγ1 = 0. (3.17)
In this form of equations, the inertia matrix has six elements. The system is not solved
for the derivatives, a situation that is not in favour of a process of solution.
@seismicisolation
@seismicisolation
3.6 Integrals of Motion
Equations of motion (3.13), (3.14) or in expanded form (3.16), (3.17) are essentially
nonlinear. For their solution, in the sense of reduction to quadratures, the application
of Jacobi’s theorem about the last integrating multiplier (See, e.g. [305]) requires the
knowledge of four integrals of motion.
The first step is to see how much general integrals the above system admits in its
most general form.
3.6.1 The Energy Integral
The rigid body is assumed to be smoothly fixed at O and moving in the uniform field
of gravity, whose potential is
V = Mgr0 · γ. (3.18)
Regarding expressions (3.11), (3.12) and (3.18), one can immediately write the
energy integral as
I1 ≡ T + V
1
≡ ω · G + Mgr0 · γ
2
1
≡ ωI · ω + Mgr0 · γ = h, (3.19)
2
h being the arbitrary constant of conserved total energy of the motion. In the expanded
form, we can write
1
I1 ≡ (I11 p 2 + I22 q 2 + I33 r 2 + 2I12 pq + 2I23 qr + 2I13 pr )
2
+Mg(x0 γ1 + y0 γ2 + z 0 γ3 )
= h. (3.20)
3.6.2 The Area’s Integral
Now we rewrite the equation of rotational motion (3.7)
dG
= −Mgr0 × γ,
dt
and note that on multiplying scalarly by the vector γ on both sides, we get
@seismicisolation
@seismicisolation
3.6 Integrals of Motion 47
dG
γ· = 0,
dt
which may be now written as
d
(G · γ) = 0,
dt
so that we obtain the second general integral of motion
I2 ≡ G · γ = f, (3.21)
f being an arbitrary parameter. In the moving axes, it has the form
I2 ≡ ωI · γ = f. (3.22)
In a general body system, it may be written as
I2 = (I11 p + I12 q + I13 r )γ1 + (I12 p + I22 q + I23 r )γ2 + (I13 p + I23 q + I33 r )γ3 = f.
The integral of motion (3.21) or (3.22) is linear in the components of the angular
velocity. In accordance with the tradition prevailing in celestial mechanics, it is called
the areas integral.
3.6.3 The Geometric Integral
The vector γ is defined as the unit vector directed vertically upwards. From this
definition, it directly follows that its square is a constant of motion, with its constant
value normalized to 1:
I3 ≡ γ 2 ≡ γ12 + γ22 + γ32 = 1. (3.23)
3.6.4 Exercise
Use Euler–Poisson’s equations of motion in the vector form (3.15) to directly obtain
the three general integrals of motion.
3.7 Special Axes Associated with the Inertia Matrix
Equation (3.16) can be somewhat simplified, by a suitable choice of the body axes.
For example, we can take the z-axis as the one joining the fixed point with the centre
@seismicisolation
@seismicisolation
of mass, so that r0 can be written as
r0 = (0, 0, z 0 ). (3.24)
Moreover, we still have the freedom to rotate the x, y-axes in their plane to a position
in which
I12 = 0. (3.25)
We shall call the final set of axes the special axes associated to the inertia matrix.
Those axes are most convenient in describing some particular solutions of the clas-
sical problem and other problems in rigid body dynamics, such as the regular pre-
cessions. This will be made clear later on.
The Euler equations now take the form
I11 ṗ + I13 ṙ + (I33 − I22 )qr + I23 (q 2 − r 2 ) + I13 pq = Mgz 0 γ2 ,

I22 q̇ + I23 ṙ + (I11 − I33 ) pr + I13 r 2 − p 2 − I23 pq = −Mgz 0 γ1 ,
I13 ṗ + I23 q̇ + I33 ṙ + (I22 − I11 ) pq + r (I23 p − I13 q) = 0, (3.26)
while Poisson’s equations still have the form (3.17), i.e.
γ̇1 + qγ3 − r γ2 = 0, γ̇2 + r γ1 − pγ3 = 0, γ̇3 + pγ2 − qγ1 = 0.
The general integrals of motion can be written in the given coordinate system in the
form
1
I1 ≡ (I11 p 2 + I22 q 2 + I33 r 2 + 2I23 qr + 2I13 pr )
2
+Mg(x0 γ1 + y0 γ2 + z 0 γ3 )
= h,
I2 ≡ (I11 p + I13 r )γ1 + (I22 q + I23 r )γ2 + (I13 p + I23 q + I33r )γ3 = f,
I3 ≡ γ12 + γ22 + γ32 = 1. (3.27)
3.8 The Use of Principal Axes of Inertia of the Body
In the special case, when the body axes are chosen to be the principal axes of the
body at O, we have
I = diag(A, B, C), G = (Ap, Bq, Cr ), r0 = (x0 , y0 , z 0 ). (3.28)
The equations of motion take the form
@seismicisolation
@seismicisolation
3.8 The Use of Principal Axes of Inertia of the Body 49
A ṗ + (C − B)qr = Mg(z 0 γ2 − y0 γ3 ),
B q̇ + (A − C) pr = Mg(x0 γ3 − z 0 γ1 ),
C ṙ + (B − A) pq = Mg(y0 γ1 − x0 γ2 ),
γ̇1 + qγ3 − r γ2 = 0, γ̇2 + r γ1 − pγ3 = 0, γ̇3 + pγ2 − qγ1 = 0, (3.29)
and the integrals of motion become
1
I1 ≡ (Ap 2 + Bq 2 + Cr 2 ) + Mg(x0 γ1 + y0 γ2 + z 0 γ3 ) = h,
2
I2 ≡ Apγ1 + Bqγ2 + Cr γ3 = f,
I3 ≡ γ12 + γ22 + γ32 = 1. (3.30)
The equations of motion acquire in (3.29) their simplest and most symmetric
form. The most favoured form is in scientific and technical literature. Those equations
involve six parameters: three principal moments of inertia and three quantities formed
by multiplying three coordinates of the centre of mass by the body weight. Unlike
Eq. (3.16), Eq. (3.29) is readily solved in the derivatives ṗ, q̇, ṙ , which is quite an
advantage.
In the sequel, we shall mostly adhere to this form of the equations of motion. Only
in exceptional occasions, we find other forms more appropriate or easier to use.
3.9 Determination of Euler’s Angles
Solving the system of six equations of motion (3.29), we determine the vectors ω(t)
and γ(t)as functions of the time t and only five arbitrary constants of integration,
since the initial values of γ satisfy the geometric integral without arbitrary constant.
This determines the Eulerian angles of nutation and proper rotation θ and ϕ as
γ1
θ = cos−1 γ3 , ϕ = tan−1 . (3.31)
γ2
To complete the solution of the dynamical problem, i.e. to determine the orientation
of the body in space, we should also determine the precession angle ψ. To this end,
we use (2.39) of Chap. 2 to write
pγ1 + qγ2
ψ̇= , (3.32)
γ12 + γ22
so that we finally obtain
@seismicisolation
@seismicisolation
t
pγ1 + qγ2
ψ = ψ0 + dt, (3.33)
0 γ12 + γ22
ψ0 is the sixth integration constant of the solution. This completes the solution of the
problem of motion about a fixed point.
3.10 The Movable and Immovable Hodographs
Applied to the angular velocity vector the relation (2.37) gives
dω
= ω̇, (3.34)
dt
i.e. the angular velocity has the same rate of change in space as in the body. This
formula, noted by Poisson, means that the infinitesimal change in the angular velocity
at any moment of time is the same in both space and body reference frames. This has
a very useful interpretation. Let the spatial curves , named as the movable angular
velocity hodograph, and 0 , the immovable angular velocity hodograph, be the loci
of the angular velocity vector in the body and space system of axes, respectively. The
two curves have the same tangent at every moment of time. The motion of the body
in space can be represented as rolling the movable hodograph without slipping on
the immovable hodograph 0 (fixed in space). The hodograph motion was studied
as a way of geometric visualization of the motion in solvable cases. A voluminous
literature exists on this topic. Interested readers may see, e.g. [108, 121] for several
concrete examples.
3.11 The Use of the Variables G, γ. Special Axes

Associated with the Gyration Ellipsoid
Let G = (P, Q, R) denote the angular momentum of the body and its components
referred to the body axes. In that case, inverting the relation (3.9), we write
ω = GA, A = I−1 , (3.35)
so that (3.13) and (3.17) take the form
Ġ + GA × G = Mgγ × r0 , γ̇ + GA × γ = 0, (3.36)
and the integrals of motion become
@seismicisolation
@seismicisolation
3.11 The Use of the Variables G, γ. Special Axes Associated with the Gyration Ellipsoid 51
1
GA · G+Mgr0 · γ = h,
2
G · γ = f,
γ 2 = 1. (3.37)
The main advantage of Eq. (3.36) is that they are solved for the derivatives, in the
sense that each of the six equations involves only one derivative of one component of
G or γ. In this form, also the areas integral takes its simplest form. The situation can
be made more advantageous by using the so-called “Special axes associated with
the gyration ellipsoid”, introduced and extensively used by Kharlamov [191]. They
are formed in the following way: Choose the z-axis as the one joining the fixed point
with the centre of mass, so that r0 can be written as
r0 = (0, 0, z 0 ), (3.38)
and then rotate the x, y-axes in their plane to a position in which
A12 = 0. (3.39)
In those special axes, the angular velocity
ω = (A11 P + A13 R, A22 Q + A23 R, A33 R + A13 P + A23 Q), (3.40)
and Euler–Poisson’s Eq. (3.36) become
Ṗ + (A22 − A33 )Q R − A13 P Q + A23 (R 2 − Q 2 ) = Mgz 0 γ2 ,

Q̇ + (A33 − A11 )P R + A23 P Q + A13 (P 2 − R 2 ) = −Mgz 0 γ1 ,
Ṙ + (A11 − A22 )P Q + (A13 Q − A23 P)R = 0,
γ̇1 + (A22 Q + A23 R)γ3 − (A33 R + A13 P + A23 Q)γ2 = 0,
γ̇2 + (A33 R + A13 P + A23 Q)γ1 − (A11 P + A13 R)γ3 = 0,
γ̇3 + (A11 P + A13 R)γ2 − (A22 Q + A23 R)γ1 = 0. (3.41)
As to the integrals of motion in the special axes, we note that the areas and
geometric integrals still have the form as in (3.37), but the energy integral takes the
form
1
I1 = (A11 P 2 + A22 Q 2 + A33 R 2 + 2 A23 Q R + 2 A13 P R) + Mgz 0 γ3 = h.
2
(3.42)
@seismicisolation
@seismicisolation
3.12 Equations of Motion in Generalized Coordinates
The Euler–Poisson form is mostly preferred in the study of rigid body motion. Nev-
ertheless, in certain situations, it is advantageous to write the Lagrangian form of
the equations of motion, sometimes using the Eulerian angles as generalized coordi-
nates and other times using different coordinates or some redundant coordinates, for
example, the components of the vector γ or the quaternions. This formalism turns
out to be most useful in the case of a dynamically symmetric body, but we shall not
impose this condition for the time being.
The Lagrangian can be written in arbitrary coordinate system fixed in the body,
but to obtain a more tractable form, we use the principal axes of inertia of the body
at the fixed point as body axes. We write
1
L= (Ap 2 + Bq 2 + Cr 2 ) − Mgr0 · γ. (3.43)
2
Using Eqs. (2.39) and (14.1), the Lagrangian takes the form
1
L= [A(ψ̇ sin θ sin ϕ + θ̇ cos ϕ)2
2
+B(ψ̇ sin θ cos ϕ − θ̇ sin ϕ)2 + C(ψ̇ cos θ + ϕ̇)2 ]
−Mg(x0 sin θ sin ϕ + y0 sin θ cos ϕ + z 0 cos θ). (3.44)
We note at once two properties of the Lagrangian leading to two integrals:

(1) The system is conservative and hence admits the energy integral
1
I1 ≡ [A(ψ̇ sin θ sin ϕ + θ̇ cos ϕ)2
2
+B(ψ̇ sin θ cos ϕ − θ̇ sin ϕ)2 + C(ψ̇ cos θ + ϕ̇)2 ]
+Mg(x0 sin θ sin ϕ + y0 sin θ cos ϕ + z 0 cos θ) = h. (3.45)
(2) The angle of precession ψ is a cyclic coordinate and this leads to the cyclic
integral
∂L
I2 ≡
∂ ψ̇
= A sin θ sin ϕ(ψ̇ sin θ sin ϕ + θ̇ cos ϕ)
+B sin θ cos ϕ(ψ̇ sin θ cos ϕ − θ̇ sin ϕ) + C cos θ(ψ̇ cos θ + ϕ̇)
= (A sin2 θ sin2 ϕ + B sin2 θ cos2 ϕ + C cos2 θ)ψ̇
+(A − B) sin θ sin ϕ cos ϕθ̇
= f. (3.46)
@seismicisolation
@seismicisolation
3.12 Equations of Motion in Generalized Coordinates 53
It is evident that those integrals are the same as the two in (3.30). Note that the
geometric integral in (3.30) turns into an identity in the Euler angles as coordinates.
In fact, γ1 , γ2 , γ3 are redundant coordinates, i.e. they are dependent coordinates
subject to the geometric integral as a constraint.
3.13 Canonical Equations of Motion in Euler’s Angles
For certain important applications, such as different perturbation procedures, it may

be advantageous to use the Hamiltonian formalism. We shall give now the Hamilto-
nian function and canonical equations of motion in Euler’s angles and their conjugate
momenta pψ , pθ , pϕ . From (3.44), we get
∂L
pψ = = D ψ̇ + (A − B) sin θ sin ϕ cos ϕθ̇ + C cos θϕ̇,
∂ ψ̇
∂L
pθ = = (A − B) sin θ cos θ sin ϕψ̇ + (A cos2 ϕ + B sin2 ϕ)θ̇,
∂ θ̇
∂L
pϕ = = C(ψ̇ cos θ + ϕ̇), (3.47)
∂ ϕ̇
where
D = A sin2 θ sin2 ϕ + B sin2 θ cos2 ϕ + C cos2 θ. (3.48)
Then, after solving (3.47) for ψ̇, θ̇, ϕ̇, we calculate the Hamiltonian
H = ψ̇ pψ + θ̇ pθ + ϕ̇ pϕ − L
(A cos2 ϕ + B sin2 ϕ) (A sin2 ϕ + B cos2 ϕ) 2
= ( p ψ − p ϕ cos θ) 2
+ pθ
2 AB sin2 θ 2 AB
(A − B) sin ϕ cos ϕ pϕ2
− ( pψ − pϕ cos θ) pθ +
AB sin θ 2C
+Mg(x0 sin θ sin ϕ + y0 sin θ cos ϕ + z 0 cos θ). (3.49)
The equations of motion can be written in the form
∂H ∂H
ṗψ = − , ψ̇ = ,
∂ψ ∂ pψ
∂H ∂H
ṗθ = − , θ̇ = ,
∂θ ∂ pθ
∂H ∂H
ṗϕ = − , ϕ̇ = . (3.50)
∂ϕ ∂ pϕ
@seismicisolation
@seismicisolation
Since H does not depend on ψ, we have ṗψ = 0, i.e. pψ = const. In conformity

with (3.46), we take this constant to be f, so that
pψ = f. (3.51)
The second and third pairs of equations give
(B − A) sin ϕ cos ϕ
ṗθ = ( f cos θ − pϕ ) pθ
AB sin2 θ
(A cos2 ϕ + B sin2 ϕ)
+ ( f − pϕ cos θ)( f cos θ − pϕ )
AB sin2 θ
−Mg (x0 cos θ sin ϕ + y0 cos θ cos ϕ − z 0 sin θ) ,
(B − A) sin ϕ cos ϕ (A sin2 ϕ + B cos2 ϕ)
θ̇ = ( f − pϕ cos θ) + pθ ,
AB sin θ AB
(A − B) sin ϕ
ṗϕ = [ ( f − pϕ cos θ) + pθ cos ϕ]
AB sin θ
cos ϕ
×[ ( f − pϕ cos θ) − pθ sin ϕ]
sin θ
−Mg sin θ (x0 cos ϕ − y0 sin ϕ) ,
pϕ (A − B) sin ϕ cos ϕ cos θ
ϕ̇ = + pθ
C AB sin θ
(A cos2 ϕ + B sin2 ϕ) cos θ
− ( f − pϕ cos θ). (3.52)
AB sin2 θ
If a solution is obtained for the last system giving θ, ϕ, θ̇, ϕ̇ as functions of time,
the precession angle ψ can be then determined by integrating the second equation in
(3.50), which is now written as
(A cos2 ϕ + B sin2 ϕ) (A − B) sin ϕ cos ϕ

ψ̇ = ( f − pϕ cos θ) − pθ . (3.53)
AB sin2 θ AB sin θ
3.14 The Routhian Reduction
From (3.46), we find
f − (A − B) sin θ sin ϕ cos ϕθ̇ − C cos θϕ̇

ψ̇ = . (3.54)
(A sin2 θ sin2 ϕ + B sin2 θ cos2 ϕ + C cos2 θ)
One can now use Routh’s procedure to ignore the cyclic coordinate ψ and reduce
the problem of motion to a system of two degrees of freedom. The Routhian of the
system is
@seismicisolation
@seismicisolation
3.14 The Routhian Reduction 55
R = L − f ψ̇
= R2 + R1 − V1 , (3.55)
where
1
R2 = {C sin2 θ(A sin2 ϕ + B cos2 ϕ)ϕ̇2 −
2D
1
− C(A − B) sin 2θ sin 2ϕθ̇ϕ̇
2
+[D(A cos2 ϕ + B sin2 ϕ) − (A − B)2 sin2 θ sin2 ϕ cos2 ϕ]θ̇2 },
f
R1 = [C cos θϕ̇ + (A − B) sin θ sin ϕ cos ϕθ̇],
D
f2
V1 = V + . (3.56)
2D
The function V1 is called the reduced potential, while V is the original potential
of the problem. The equations of motion are
d ∂R ∂R d ∂R ∂R
− = 0, − = 0. (3.57)
dt ∂ θ̇ ∂θ dt ∂ ϕ̇ ∂ϕ
We shall not write them down in the expanded form because they lack symmetry and
they are not easy to use in general. However, they can be used much easily in case
of a dynamically symmetric body. Such concrete applications are not in the focus
of the present book and can be found in several books on perturbation problems.
Those are two second-order equations in the two variables θ and ϕ. After solving
those equations and expressing the two angles in terms of time, one can determine
the ignored angle ψ by integrating (3.54) with respect to time.
Remark 1: The Lagrangian (3.44) (and equations of motion derived from it in
any generalized coordinates) is time-reversible, i.e. the Lagrangian and equations
remain invariant if the sign of time t is changed. On the contrary, the Routhian and
Routhian equations of motion are not time-reversible. They are invariant only on the
simultaneous change of signs of t and f.
When f = 0, the Routhian becomes
R = R2 − V,
and the Routhian equations of motion are time-reversible.

Remark 2: In the case of axial dynamical symmetry B = A, a significant sim-
plification occurs in the Routhian (3.55). It renders to the form
A f
R= (C sin2 θϕ̇2 + D θ̇2 ) + C cos θϕ̇ − V1 ,
2D D
f2
V1 = V + . (3.58)
2D
@seismicisolation
@seismicisolation
Finally, when A = C, i.e. in the case of complete dynamical symmetry, D = A

and we have
A
R = (sin2 θϕ̇2 + θ̇2 ) + f cos θϕ̇ − V (θ, ϕ). (3.59)
2
This is the Lagrangian of a particle moving on a smooth sphere with θ and ϕ as
polar coordinates on that sphere. The particle is subject to forces with potential V
and gyroscopic forces represented by the term linear in ϕ̇. Note that the last term is
proportional to f and thus vanishes when f = 0.
3.15 Exercises
(1) A heavy rigid body is moving about a fixed point, which is not coincident with the
centre of mass of the body. Use the energy integral and Euler’s equations to express
the vector γ in terms of the angular velocity ω and its time derivative ω̇ in the form
1 1
γ= [r0 × (ω̇I + ω × ωI) + (h − ωI · ω)r0 ], (3.60)
Mg|r0 |2 2
where h is the energy constant and r0 = 0 is the position vector of the centre of mass
of the body with respect to the fixed point.
(2) Use the last result to reduce the equations of motion of the classical problem
to the form of three autonomous first-order differential equations in the components
of the angular velocity with respect to an arbitrary system of axes fixed in the body
in time as independent variable to the form1
(ω̇I + ω × ωI) · r0 = 0,
1
(h − ωI · ω)2 + (ω̇I + ω × ωI)2 = M 2 g 2 |r0 |2 ,
2
1
r0 · [(h − ωI · ω)ωI+ω̇I×ωI − |ωI|2 ω] = Mg f |r0 |2 , (3.61)
2
where f is the areas constant and other parameters as defined above.
Hint: Use the following equations:
(ω̇I + ω × ωI) · r0 = 0,
|γ|2 = 1,
ωI · γ = f.
1For this form, or (3.63), to be equivalent to the original Euler–Poisson system, a condition on the
motion must be satisfied (See the two theorems in Sect. 8.1).
@seismicisolation
@seismicisolation
3.15 Exercises 57
(3) Show that in terms of G, the formula (3.60) and reduced Eq. (3.61) take the
following form:
1 1
γ= [r0 × (Ġ + GA × G) + (h − GA · G)r0 ], (3.62)
Mg|r0 |2 2
and
(Ġ + GA × G) · r0 = 0,
1
(h − GA · G)2 + (Ġ + GA × G)2 = M 2 g 2 |r0 |2 ,
2
1
r0 · [(h − GA · G)G+Ġ×G − |G|2 GA] = Mg f |r0 |2 , (3.63)
2
where A = I−1 .
@seismicisolation
@seismicisolation
Chapter 4
General and Conditional Integrable
Cases of the Classical Problem
In the classical problem of motion of a rigid body about a fixed point, there are only
three general and one conditional integrable cases. By a general integrable case,
we mean that case in which the problem of motion is integrable for arbitrary initial
conditions, while the term “conditionally integrable” is reserved for cases which
are integrable only on a fixed level of the areas integral. In this chapter, we give a
brief account of integrable cases of the classical problem. We first rewrite the Euler–
Poisson equations of motion in their most used form in the system of principal axes
of inertia of the body at the fixed point:
A ṗ + (C − B)qr = Mg(z 0 γ2 − y0 γ3 ),
B q̇ + (A − C) pr = Mg(x0 γ3 − z 0 γ1 ),
C ṙ + (B − A) pq = Mg(y0 γ1 − x0 γ2 ),
γ̇1 + qγ3 − r γ2 = 0, γ̇2 + r γ1 − pγ3 = 0, γ̇3 + pγ2 − qγ1 = 0,
(4.1)
and also their integrals of motion
1
I1 ≡ (Ap 2 + Bq 2 + Cr 2 ) + Mg(x0 γ1 + y0 γ2 + z 0 γ3 ) = h,
2
I2 ≡ Apγ1 + Bqγ2 + Cr γ3 = f,
I3 ≡ γ12 + γ22 + γ32 = 1. (4.2)
The list of integrable cases is provided in the following Tables 4.1 and 4.2:
In both Tables 4.1 and 4.2 a1 = Mgx0 /C, a2 = Mgy0 /C. Although in the cases
of Kowalevski and Goryachev–Chaplygin, one can always rotate the coordinate axes
x y in their plane in order that the x-axis passes through the centre of mass, i.e.
y0 = 0, we here keep both coordinates in the general position, keeping in mind some
future occasions, when the extra-parameter makes it possible to build integrable cases
https://doi.org/10.1007/978-3-030-96336-1_4
@seismicisolation
@seismicisolation
60 4 General and Conditional Integrable Cases of the Classical Problem
Table 4.1 General (Unconditional) cases

Author Conditions
1 Euler [78] (1758). gr0 = 0.
I4 = A 2 p 2 + B 2 q 2 + C 2 r 2 .
2 Lagrange’s top [251] (1788). B = A,
(Axially symmetric case)
I4 = Cr. x0 = y0 = 0,
3 Kowalevski (1889) [238] A = B = 2C,
I4 = z 0 = 0.
( p 2 − q 2 − a1 γ1 + a2 γ2 )2 +
(2 pq − a1 γ2 − a2 γ1 )2 .
Table 4.2 Conditional case f = 0

1 Goryachev–Chaplygin [115] A = B = 4C, z 0 = 0
(1900) [52] (1901)
I4 = r ( p 2 + q 2 ) − γ3 (a1 p + a2 q)
containing additional parameters in a nonlinear way (See, e.g. case 6 of Table 12.1
in Chap. 12).
4.1 Euler’s Case (1758). The Torque-Free Rigid Body
4.1.1 Explicit Time-Solution
In the same work [78], where he published the final form of the equations of motion
of a rigid body about a fixed point, Euler noted that the case of a free rigid body
(when the moment of forces applied to the body about the fixed point vanishes)
can be solved and reduced that case to a quadrature expressing the relation of the
components of angular velocity to time. In the present section, we shall solve the
equations of motion in Euler’s case explicitly using the elliptic functions invented
by Jacobi after almost a century, in the middle of the nineteenth century.
The equations of motion are obtained from Eq. (3.29) of the previous chapter
by setting either g = 0 (the case of absence of gravity force) or the equivalent case
r0 = 0 (the case of a heavy body fixed from its centre of mass). Euler’s equation can
be written as
dG
≡ Ġ + ω × G = 0, (4.3)
dt
or, in components,
A ṗ − (B − C)qr = 0,
@seismicisolation
@seismicisolation
4.1 Euler’s Case (1758). The Torque-Free Rigid Body 61
B q̇ − (C − A) pr = 0,
C ṙ − (A − B) pq = 0. (4.4)
Apart from Poisson’s equations, (4.4) form a closed system of three equations in
three variables, and can be solved independently. One can also notice that the energy
integral (3.30) for this case takes the form
1
I1 ≡ (Ap 2 + Bq 2 + Cr 2 ) = h, (4.5)
2
so that it gives an integral of the system (4.4).
We can easily construct another integral of that system. In fact, Eq. (4.3) expresses
the constancy of the angular momentum vector G in the inertial space. The modulus
of this vector will be constant in all frames and in particular in the body frame. We
can write the integral
G2 = const. = G 20 ,
Alternatively, we can multiply (4.3) scalarly by G to obtain
˙ = 0,
G·G
which leads to the same result. Thus, the new integral can be written as
A2 p 2 + B 2 q 2 + C 2 r 2 = G 20 . (4.6)
Now, we construct the general solution for the equations of motion. First, we
introduce the two constants D and μ through the relations
2h = μ2 D, G 0 = Dμ, (4.7)
provided none of the constants G 0 , h is zero, which can happen only on an equilibrium
( p = q = r = 0). The two Eqs. (4.5, 4.6) become
Ap 2 + Bq 2 + Cr 2 = Dμ2 , (4.8)
A2 p 2 + B 2 q 2 + C 2 r 2 = D 2 μ2 . (4.9)
Note that D has the dimension of a moment of inertia, while μ has that of an angular
velocity.
Eliminating r from the two Eqs. (4.8, 4.9), we have
1
p2 = [μ2 D(D − C) − B(B − C)q 2 ]. (4.10)
A(A − C)
Also, eliminating p from the same equations
@seismicisolation
@seismicisolation
1
r2 = [μ2 D(A − D) − (A − B)q 2 ]. (4.11)
A(A − C)
Without loss of generality, we assume A > B > C, it is easy to show that the variables
p and r have real values only if the parameter D satisfies the condition
C D A. (4.12)
Inserting the two expressions (4.10, 4.11) into the middle equation in (4.4), we obtain
±1 2
B q̇ = √ μ D(A − D) − B(A − B)q 2 μ2 D(D − C) − B(B − C)q 2 .
AC
(4.13)
Noting that the right-hand side of the last equation is the square root of a 4th degree
polynomial, we conclude that the separation of variables leads to an elliptic integral
of the first kind. Since q 2 admits the values from zero to the minimum value of
μ2 D(A − D) μ2 D(D − C)
q12 = , q22 = .
B(A − B) B(B − C)
Thus, we have two cases:
4.1.1.1 The First Case
If q22 < q12 (this is hold only if D−C

B−C
< A−D
A−B
, i.e. C < D < B) then q 2 is changed
from 0 to q22 . Setting
q = q2 x,
where x is a new variable that varies from −1 to 1. Inserting the last expression in
Eq. (4.13), we get
ẋ = λ 1 − x 2 1 − k 2 x 2 , (4.14)
where
D(A − D)(B − C) (A − B)(D − C)
λ = ±μ , k2 = . (4.15)
ABC (A − D)(B − C)
Separating the variables in (4.14) and integrating, we get

x
du
λ(t − t0 ) = , (4.16)
0 (1 − u )(1 − k 2 u 2 )
2
where t0 is an arbitrary integration constant. Equation 4.16 can be solved to give
x = snλ(t − t0 ). (4.17)
@seismicisolation
@seismicisolation
Taking all obtained expressions into account, we can write

D(D − C)
p = ±μ cnλ(t − t0 ),
A(A − C)

D(D − C)
q=μ snλ(t − t0 ),
B(B − C)

D(A − D)
r = ±μ dnλ(t − t0 ). (4.18)
C(A − C)
The last expressions show that the components of angular velocity are periodic func-
tions in the time t with period λ4 K (k) for p, q and λ2 K (k) for r . Furthermore, the sign
of p and q change during the motion while r does not change its sign. Therefore, the
rotation about the third axis that has the minimum value of inertia is always in one
and the same direction.
4.1.1.2 The Second Case
In this case, B < D < A. The reader can easily check that the last formulas are
replaced by

D(D − C)
p=μ dnλ(t − t0 ),
A(A − C)

D(A − D)
q=μ snλ(t − t0 ),
B(A − B)

D(A − D)
r =μ cnλ(t − t0 ), (4.19)
C(A − C)
where
D(D − C)(A − B) (A − D)(B − C)
λ = ±μ ,k = .
ABC (A − B)(D − C)
Note that this time the rotation about the first axis (x-axis, with the maximum moment
of inertia) remains in one direction.
4.1.2 Permanent Rotations
Permanent rotations are motions with constant angular velocity in the body, i.e.
motions satisfying ω̇ = 0. Since dω
dt
= ω̇ + ω × ω = 0, those motions are are also
@seismicisolation
@seismicisolation
time independent in space. For them Eqs. (4.4) become
(C − A)qr = (A − C)r p = (B − A) pq = 0.
They have three solutions: { p = , q = r = 0}, {q = , p = r = 0} and {r =

, p = q = 0}, each of which represent a permanent rotation with an arbitrary angu-
lar velocity about one of the principal axes of inertia of the body which takes an
arbitrary fixed position in the inertial space.
4.1.3 The Degeneracy of Elliptic Function
When D = C, we have k = 0 and Eq. (4.18) gives p = q = 0 and r = ±μ. This

refers to the permanent rotation of the body around the third axis that has the minimum
principal inertia. The second degeneration occurs when D = B. Then, we have k = 1
and the expressions (4.18) or (4.19) become

B(B − C) 1
p = ±μ ,
A(A − C) cosh λ (t − t0 )
q = μ tanh λ (t − t0 ),

B(A − B) 1
r = ±μ ,
C(A − C) cosh λ (t − t0 )

where λ = μ (A−B)(B−C)
AC
. This is an asymptotic motion. As the time t → ±∞, this
motion becomes a permanent rotation with angular velocity ±μ around the middle
axis.
4.1.4 The Case of Dynamically Axi-Symmetric Body
In this case, when B = A(> C) (the case of a body with a prolate inertia spheroid),
we can obtain relevant formulas of the solution by substituting B = A in (4.18). We
obtain

D(D − C)
p = ±μ cos λ(t − t0 ),
A(A − C)

D(D − C)
q=μ sin λ(t − t0 ),
A(A − C)
@seismicisolation
@seismicisolation

D(A − D)
r = ±μ (4.20)
C(A − C)

where λ = ±μ D(A−D)(A−C)
A2 C
. It is easy to identify the two results.
On the other hand, when B = C < A, i.e. in the case of a body with an oblate
ellipsoid of inertia, from (4.19) we get

D(D − C)
p=μ ,
A(A − C)

D(A − D)
q=μ sin λ(t − t0 ),
C(A − C)

D(A − D)
r =μ cos λ(t − t0 ), (4.21)
C(A − C)
where
D(D − C)(A − C)
λ=μ .
AC 2
4.1.5 Euler’s Angles in Terms of Time
In Euler’s case, there is no distinct direction in space to be taken as the direction of

gravity. The orthonormal basis vectors α, β, γ can be chosen arbitrarily in space.
There also exist three areas integrals
G·γ = G 0 , G·α = G 0 , G·β = G 0 (say). (4.22)
For determinacy, one can use the direction of the constant angular momentum to be
that of γ, i.e. one can take
G (Ap, Bq, Cr )
γ= = . (4.23)
G0 G0
Corresponding to this choice, one must take G 0 = G 0 = 0 in (4.22). Using the
expressions (4.18), we can write the Eulerian angles in terms of time. Thus, we
have

Cr C D(A − D)
cos θ = = dnλ(t − t0 ),
G0 D C(A − C)
@seismicisolation
@seismicisolation

Ap A(B − C) cnλ(t − t0 )
tan ϕ = = , (4.24)
Bq B(A − C) snλ(t − t0 )
as elliptic functions in the time t. The angle of precession ψ can be determined from

A−C
pγ1 + qγ2 Ap 2 + Bq 2 D
ψ̇ = =G 2 2 =μ 1− A
,
γ12 + γ22 A p + B 2q 2 C 1+ C(A−B) 2
A(B−C)
s
where s = snλ(t − t0 ). Thus, ψ is obtained as an elliptic integral of third kind in time

D A−C t dt
ψ = ψ0 + μ t − t0 − . (4.25)
t0 1 + sn2 λ(t − t0 )
C A C(A−B)
A(B−C)
Further analytical evaluation of the last integral may be found in several treatizes,
e.g. [9] and will not be pursued here.
In the case of dynamical symmetry A = B, we get
Cr0
cos θ = ,
G
G0
ψ̇ = , (4.26)
A
C
ϕ̇ = r0 − ψ̇ cos θ = (1 − )r0 .
A
It is evident that this motion describes a uniform rotation around the axis of symmetry
while this axis rotates around the vertical with a constant angular velocity. Euler’s
angles are then given by
Cr0
θ = cos−1 ,
G
G0
ψ = ψ0 + (t − t0 ), (4.27)
A
C
ϕ = ϕ0 + (1 − )r0 (t − t0 ).
A
This motion is named regular precession. The axis of symmetry of the body is called
the figure axis and the axis fixed in space in the direction of G and γ is called the
precession axis. This type of motion will be met in several other situations later in
this book.
@seismicisolation
@seismicisolation
4.1.6 Geometrical Interpretation of the Motion (Poinsot

1851)
The explicit analytical solution of Euler’s case provided in Sect. 4.1.1 is not of much
help in giving a clear geometric idea about how the body moves in space. An elegant
geometric description of the motion constructed by Poinsot [308] has become popular
in textbooks on the subject. We give a quick presentation of this construction here.
In the system of principal axes of inertia of the body O x yz at the fixed point O,
the inertia ellipsoid of the body has the equation
Ax 2 + By 2 + C z 2 = 1. (4.28)
Let the body be in motion at certain moment of time with instantaneous angular
velocity ω = ( p, q, r ) referred to that system. Denote by r =(x, y, z) the position
vector of the pole P (the point of intersection of the vector ω with the inertia ellipsoid).
Note that since P lies on the instantaneous axis of rotation of the body, it is always
in instantaneous rest.
It is easy to deduce the following properties of the motion of the inertia ellipsoid.
Property 1: The angular velocity of the body is proportional to the line segment
O P, cut of the rotation axis by the inertia ellipsoid.
In fact, one can write
ω = λr, (4.29)
where λ is a multiplier to be determined. Substituting this expression into (4.28), we

obtain
λ−2 (Ap 2 + Bq 2 + Cr 2 ) = 1,
and in virtue of the energy integral this gives
λ2 = 2h.
It follows that
1
λ= √ .
2h
so that the proportionality factor λ is a constant. We have here fixed the positive sign
for λ, in order to have r in the same direction of ω, and thus from (4.29) we get
√
ω= 2hr. (4.30)
Property 2: The tangent plane to the inertia ellipsoid at the pole P keeps a fixed
direction in space for all the time of motion.
The unit vector n in the direction orthogonal to the tangent plane at P can be
written as
@seismicisolation
@seismicisolation
rI ωI G
n= = = , (4.31)
|rI| 2
|ωI|2 G
where we have used (4.30). As G is a constant vector in space, property 2 is proved.

Property 3: The tangent plane to the inertia ellipsoid at the pole P remains fixed
in space for all the time of motion.
To prove this property it is easy, using properties 1 and 2, to deduce that the
orthogonal distance of the centre O of the ellipsoid from the tangent plane at P
√
rI · r 1 2h
δ = r · n = =√ = (4.32)
|rI| 2
rI · r
2 G
is really constant during motion.

Now we are ready to make the final conclusion about the motion of the body based
on the motion of its ellipsoid of inertia (fixed in that body):
The motion of a rigid body by inertia about the fixed point O is realized by rolling
the inertia ellipsoid, with centre at O, without slipping on the immovable plane
(See Fig. 4.1).
At any time moment, the point of contact P is momentarily at rest. The curve
traced out on the ellipsoid of inertia by the pole P is called polhode, and it is
obviously a closed curve. The locus of P on the plane has the name herpolhode.
It is not closed in general, but it may close under some conditions on the parameters
of the body and of the motion. In the case of a dynamical axial symmetry, both the
polhode and herpolhode become circular.
Fig. 4.1 Poinsot’s

construction
@seismicisolation
@seismicisolation
It should be noted that Poinsot’s description is purely geometrical. It shows the

consecutive positions occupied at different moments by the inertia ellipsoid of the
body, but has nothing to do with the time sequence in which those positions occur.
Exercise 1 Show that in Euler’s case the angular momentum of the body describes
a cone, whose equation referred to the principal axes of inertia at the fixed point is
D 2 D D
(1 − )x + (1 − )y 2 + (1 − )z 2 = 0.
A B C
4.2 Lagrange’s Case (1788). The Top with a Fixed Point
Lagrange’s top is characterized by axial symmetry, when two of the principal

moments of inertia at O are equal and the centre of the mass of the body lies on
the axis of dynamical symmetry, so that one can write
A = B, x0 = y0 = 0. (4.33)
This case was first described as a solvable problem by Lagrange in his historical book
on analytical mechanics [251]. A significant contribution of Poisson [309] made a
clear and elegant picture of the motion of the apex of the top. This picture will be
briefly described below in this section.
The study of motion of Lagrange’s top is most interesting as for theoretical under-
standing of gyroscopic effects of rotating bodies and from the point of view of scien-
tific and technical applications. In fact, the approximate theory of gyroscopes relies
on the fundamental properties of motion of Lagrange’s heavy top and the axisym-
metric version of the free top (in absence of gravity). That is why that top is a
favourite subject in all books on analytical dynamics in general and, particularly, in
books on gyroscopes. (See, e.g. [112] for precession of equinoxes and [132, 270] for
gyroscopic effects of rotors). One of the most detailed descriptions of the dynamical
behaviour of Lagrange’s top can be found in [222], where expressions for the angular
velocities and the geometric variables are also given in terms of theta functions (See
also [269, 368]). Interest in the subject never faded away over years. Various aspects
of Lagrange’s top were studied extensively in the literature. The reader can see [34,
61, 62, 104, 106, 143, 262, 310, 362], just as examples. In this section, we give a
brief account of the analytical solution and the most important qualitative properties
of the motion.
@seismicisolation
@seismicisolation
4.2.1 The Solution
In the case of Lagrange, both coordinates ψ and ϕ are cyclic and one can use
Routh’s procedure to ignore them and reduce the problem of motion in the clas-
sical Lagrangian form to one degree of freedom. However, we shall follow here a
slightly different approach, more consistent with that used in other cases throughout
this book. In fact, under conditions (4.33) Euler–Poisson’s equations reduce to
A ṗ − (A − C)qr = Mgz 0 γ2 ,
Aq̇ − (A − C) pr = −Mgz 0 γ1 ,
C ṙ = 0, (4.34)
γ˙1 + qγ3 − r γ2 = 0,
γ˙2 + r γ1 − pγ3 = 0,
γ˙3 + pγ2 − qγ1 = 0. (4.35)
The last equation of (4.34) at once gives the fourth integral
r = r0 (4.36)
where r0 is an arbitrary parameter. We also write the three general integrals:

The energy integral
A( p 2 + q 2 ) + Cr 2 + 2Mgz 0 γ3 = 2h, (4.37)
the areas integral

A( pγ1 + qγ2 ) + Cr γ3 = f, (4.38)
and the geometric integral

γ12 + γ22 + γ32 = 1 (4.39)
where h, f are arbitrary parameters. According to Jacobi’s theorem, the equations

of motion can be solved completely and the solution is expressed in terms of elliptic
functions as in Euler’s case. In fact, using Eq. 4.35 and the geometric integral, we
obtain
γ˙3 2 = ( pγ2 − qγ1 )2 = p 2 (1 − γ32 − γ12 ) + q 2 (1 − γ32 − γ22 ) − 2 pqγ1 γ2

= (1 − γ32 )( p 2 + q 2 ) − ( pγ1 + qγ2 )2 . (4.40)
One can also use the four integrals of motion in the last equation to obtain the
following equation for γ3 :
@seismicisolation
@seismicisolation
4.2 Lagrange’s Case (1788). The Top with a Fixed Point 71
f − Cr0 γ3 2
γ˙3 2 = (1 − γ32 )(E − aγ3 ) − ( ) , (4.41)
A
where E = A1 (2h − Cr02 ), a = 2Mgz

A
0
and, without loss of generality, we assume
a > 0. In this equation, one can separate the variables and integrate to obtain the
relation
dγ3
t= √ , (4.42)
F(γ3 )
where
1
F(γ3 ) = (1 − γ32 )(E − aγ3 ) − ( f − Cr0 γ3 )2 . (4.43)
A2
The function F(γ3 ) is a cubic polynomial in the variable γ3 and thus, the integral in
(4.42) is an elliptic integral of the first kind and its inverse γ3 (t) is an elliptic function.
The three roots of F, which we denote by u 1 , u 2 , u 3 (u 1 ≤ u 2 ≤ u 3 ), play a decisive
role in determining the character of the motion. We shall now closely investigate the
function F on the real γ3 line. First we note that
1
F(∓1) = − ( f ± Cr0 )2 ,
A2
F(+∞) = +∞. (4.44)
That is, apart from the very special cases f ± Cr0 = 0, F is negative at the terminal
points of the interval [−1, 1]. One of the roots, namely u 3 , lies to the right of γ3 = 1
and for physically meaningful motion the other two roots must be real and must
lie in the interval [−1, 1]. The point representing γ3 moves on the interval [u 1 , u 2 ],
where F is non-negative, (See Fig. 4.2), traversing this interval in one direction
and then in the other with its velocity vanishing at the ends. In the generic case of
u
three different roots of F, γ3 is periodic in t with period T = 2 u 12 √ dγ 3
F(γ3 )
. Explicit
solution was given in [134] in terms of Jacobi’s elliptic functions of time. The most
detailed solution is given in the treatise of Klein and Sommerfeld [222] in terms
of Theta functions. Alternative form in Weierstrass’ functions, the most popular in
Fig. 4.2 The path of the apex on the fixed sphere
@seismicisolation
@seismicisolation
application at that time, can be found in Whittaker’s book [368]. We shall not give
explicit analytical formulas here, but we can draw, following Poisson, a very useful
qualitative picture of the motion of the body.
Now, we turn to determine Euler’s angles. The nutation angle θ is readily expressed
as θ = cos−1 γ3 . The two relations
pγ1 + qγ2 = ψ̇ sin2 θ, r = ψ̇ cos θ + ϕ̇
can be used, together with the areas integral, to obtain expressions
f − Cr0 γ3
ψ̇ = , (4.45)
A(1 − γ32 )
γ3 ( f − Cr0 γ3 )
ϕ̇ = r0 − . (4.46)
A(1 − γ32 )
One can obtain ψ, ϕ by integrating the last two expressions with respect to time. The
general solution of the problem is thus constructed. It contains six arbitrary constants:
h, f, r0 and the integration constants that appear in the last three integration processes.
4.2.2 The Study of the Motion
Lagrange’s top is assumed to have only dynamical symmetry about the z−axis. It is
possible to think of it as an ordinary top having complete axial symmetry, geometric
and physical. From this perspective, the way in which the proper rotation angle
ϕ changes does not make difference to the observer. The significant part of the
motion is that of the figure axis. This is determined by the two spherical coordinates
θ = cos−1 γ3 and ψ, governed by Eqs. (4.42) and (4.45), respectively.
4.2.2.1 Motion of the Apex
During the motion of the body, the apex of the figure axis traces a curve on the
unit sphere fixed in space and with centre at the fixed point. The figure axis nutates,
ascending from the angle θ1 = cos−1 u 1 to θ2 = cos−1 u 2 and then descending to θ1 ,
in a periodic manner, while precessing about the vertical. The precession velocity ψ̇,
as given by (4.45), takes its minimum value at the upper-most position θ2 . This value
may be positive, zero or negative, depending on the quantity f − Cr0 u 2 . Accordingly,
the spherical curve traced by the apex touches the upper circle u = u 2 in the same
direction as it makes with the lower one, has a cusp on that circle or touches it in the
opposite direction. The trace of the apex takes one of the forms a, b or c, depicted in
Fig. 4.2.
@seismicisolation
@seismicisolation
4.2 Lagrange’s Case (1788). The Top with a Fixed Point 73
4.2.2.2 Regular Precession of Lagrange’s Top
Generic motion of Lagrange’s top corresponds to motion of the figure axis between
the two circles u 1 and u 2 . In Fig. 4.3a, u varies on the interval [u 1 , u 2 ]. Regular
precession corresponds to the nutation angle θ taking a constant value θ∗ (say), and
then from (4.45, 4.46) we find that the other two Eulerian angles ψ, ϕ change with
time in constant rates. This occurs in two qualitatively different ways:
1- When u 1 = u 2 = u ∗ , −1 ≤ u ∗ ≤ 1. This happens at inclination u ∗ , provided
two conditions F(u ∗ ) = F (u ∗ ) = 0 are satisfied. These conditions give, respec-
tively,
(Cr0 u ∗ − f )2
E = au ∗ + ,
A(1 − u ∗2 )

1 ∗2 ∗2
f = [Cr 0 (1 + u ) ± (1 − u ) C 2 r02 − 2 Aau ∗ ]. (4.47)
2u ∗
u1=u2=u* u3
1 u
-1 0 1 u -1
a-Generic motion. b-Regular processions.
u1=u2=1
-1 0 u3 u u
c-Stable standing position. d-Unstable standing position.
Fig. 4.3 Possible cases of motion of Lagrange’s top
@seismicisolation
@seismicisolation
When the figure axis makes with the vertical upwards an angle θ∗ = cos−1 u ∗ ( π2 <
θ∗ ≤ π), there are two different real values of f corresponding to two precessional
motions that may be named as fast and slow precessions. On the other hand, when
0 ≤ θ∗ < π2 , the two precessions are possible only when |r0 | exceeds a minimal
√
∗
value 2 CAau (Fig. 4.3b). This family of precessional motions include the hanging
gyroscope position u ∗ = −1 and the standing gyroscope position u ∗ = 1, termed
also as the sleeping top position (Fig. 4.3c). Regular precessions of this family are
all stable, since a slight perturbation of the motion causes splitting of the two roots
in a small neighbourhood of u ∗ . This leads to a small periodic change in the nutation
angle θ and consequently small wobbling in the rates ψ̇ and ϕ̇.
2- When u 1 < 1, u 3 = u 2 = 1. This gives a different standing position, corre-
sponding to Fig. 4.3d. On perturbation, the equal roots split into u 3 > 1, u 2 < 1. The
figure axis begins a finite periodic motion, in which it goes near to position u 1 before
it returns near to u 2 . This standing position is unstable.
Comparing Fig. 4.3c, d, one concludes that the standing gyroscope position is
stable when F has a maximum at its double root u = 1 and unstable when it has a
minimum at that position. Analytically, the condition for a stable upright spinning
position is
F(1) = F (1) = 0, F (1) < 0. (4.48)
This finally gives the condition
A2 4Mg Az 0
r02 > 2a = . (4.49)
C2 C2
√
There is a critical value r0∗ = 2 MgC
Az 0
of the angular speed |r0 | of the sleeping top,
under which the upward position is unstable and above which that position becomes
stable. The same stabilizing effect can be achieved even when the body stands at rest
in its upper equilibrium position, with the help of a symmetric rotor whose axis is
fixed in the body in alignment with the body axis. We shall return to this point in the
next chapter.
4.3 Kowalevski’s Case (1888)
In both cases of Euler and Lagrange, an integral of motion followed from general
principles of mechanics, constancy of the angular momentum in the first and cyclicity
of the angle of rotation about the axis of symmetry in the second. An important
moment was that in both cases the equations of motion were solved to the end
and the solution expressed through Jacobi’s elliptic functions and certain integrals
involving them.
The search for integrable cases continued, but, although the problem attracted
attention of many eminent mathematicians, the search didn’t lead to any other cases.
@seismicisolation
@seismicisolation
4.3 Kowalevski’s Case (1888) 75
A whole century later, a new integrable case of the heavy rigid body was found by
Sofia Kowalevski. That was not in virtue of a physical conservation principle, but
using a purely mathematical condition: all solutions of the equations of motion (4.1)
should have only poles as their singularities as functions of time in the complex
t-plane. This property is satisfied by the solutions in the two known integrable cases
of Euler and Lagrange, being expressible in terms of elliptic functions of time. It
implies that the Euler–Poisson variables ω, γ can be represented in the vicinity of a
pole at t0 (say) by a Laurent series of the form

∞
∞
1 (k) 1
ωi = ω (t − t 0 ) k
, γi = γ (k) (t − t0 )k , (4.50)
(t − t0 )αi k=0 i (t − t0 )βi k=0 i
in which αi , βi are positive integers and the coefficients ωi(k) , γi(k) are (in general
complex) constants to be determined. In order for the series (4.50) to represent the
general solution of the six-ordered system (4.1), five constants must remain arbitrary
in it. The sixth one, as explained in Sect. 3.9, is associated with the integration process
that determines the angle of precession.
Kowalevski used the values αi = 1, βi = 2 and analyzed the successive systems
of equations in the coefficients resulting from substituting the expansion in the Euler–
Poisson equations of motion and their integrals. As a result, she isolated three cases
of the required type, the two cases of Euler and Lagrange and a new third one,
characterized by the conditions A = B = 2C, z 0 = 0. Thus, the centre of mass of
the body lies in the plane of equal moments of inertia, i.e. z 0 = 0. By virtue of the
condition A = B, we can assume that the centre of the mass lies on x-axis, so that
we finally write Kowalevski’s conditions for the new third case1
A = B = 2C, y0 = z 0 = 0. (4.51)
Note that those conditions guarantee only that property of the solution in the complex
plane, but with no definite conclusion about integrability of the third case, i.e. about
the existence of a complementary fourth integral, necessary for the integration of the
equations of motion. Kowalevski tried and found the complementary integral in the
third case, thus proving the integrability of that case. She also reduced problem to
quadratures and expressed the solution in terms of hyper-elliptic functions of time,
which are more complicated than elliptic functions, but they share with them the
property of having only poles as singular points in the complex plane.
In this section, we throw some light on the Kowalevski case. In our presentation,
more attention is paid to available principal results. Detailed citations to relevant
sources covering the abridged material are given.
1Some authors argue that the condition y0 = 0 is a restriction on the physical parameters of the
problem and call for correcting this error [98, 371]. For a discussion see Chap. 8 Sect. 8.14.
@seismicisolation
@seismicisolation
4.3.1 Integration of the Equations of Motion
Under Kowalevski’s conditions (4.51), the Euler–Poisson equations take the form
2 ṗ − qr = 0,
2q̇ + pr = aγ3 ,
ṙ + aγ2 = 0, (4.52)
γ˙1 + qγ3 − r γ2 = 0, γ˙2 + r γ1 − pγ3 = 0, γ˙3 + pγ2 − qγ1 = 0, (4.53)
in which a = Mgx0
C
. The three general integrals of motion may be written as
I1 = 2( p 2 + q 2 ) + r 2 + 2aγ3 = 2h,
I2 = 2( pγ1 + qγ2 ) + r γ3 = f,
I3 = γ12 + γ22 + γ32 = 1. (4.54)
From the first two equations of (4.52), we get
d
2 ( p + iq) = −ir ( p + iq) + iaγ3 , (4.55)
dt
and from the first two equations of (4.53)
d
(γ1 + iγ2 ) = −ir (γ1 + iγ2 ) + iγ3 ( p + iq). (4.56)
dt
Multiplying (4.55) by p + iq and (4.56) by a and subtracting, we obtain
d
( p + iq)2 − a(γ1 + iγ2 ) = −ir ( p + iq)2 − a(γ1 + iγ2 ) .
dt
The last equation can be written in the equivalent form
d
ln ( p + iq)2 − a(γ1 + iγ2 ) = −ir.
dt
Taking the conjugate of the last equation, noting that all the variables are real, we get
d
ln ( p + iq)2 − a(γ1 + iγ2 ) = ir,
dt
and adding the last two equations gives
d
ln |( p + iq)2 − a(γ1 + iγ2 )|2 = 0.
dt
@seismicisolation
@seismicisolation
Thus, we obtain the fourth first integral in the form
|( p + iq)2 − a(γ1 + iγ2 )|2 = k 2 ,
where k is an arbitrary integration constant. Its final form takes the form
I4 = ( p 2 − q 2 − aγ1 )2 + (2 pq − aγ2 )2 = k 2 . (4.57)
According to Jacobi’s theorem, the general solution of Euler–Poisson equations in

this case can be constructed. To find it, Kowalevski introduced two new variables
s1 , s2 and showed that they satisfy the following two ordinary differential equations
ds1 ds2
√ +√ = 0,
(s1 ) (s2 )
s1 ds1 s2 ds2 dt
√ +√ = , (4.58)
(s1 ) (s2 ) 2
where
a2 f 2
(s) = −(s − h − k)(s − h + k)F(s), F(s) = s[(s − h)2 + a 2 − k 2 ] − ,
2
(4.59)
i.e. (s) is a polynomial of degree five in the variable s. Those equations can be put
in the equivalent form
ṡ1 1
√ = ,
(s1 ) 2(s1 − s2 )
ṡ2 −1
√ = . (4.60)
(s2 ) 2(s1 − s2 )
Equations (4.58) are termed Abel–Jacobi equations. On integration, they give

s1 s2
ds1 ds2
√ + √ = 0,
s0 (s1 ) 0
s2 (s2 )
1s1 s2
s1 ds1 s2 ds2 1
√ + √ = (t − t0 ), (4.61)
s10 (s1 ) 0
s2 (s2 ) 2
where s10 , s20 are the initial values at t = t0 . The inversion of those ultra-elliptic
quadrature is a classical solved problem. It gives the two variables s1 , s2 in terms of
ultra-elliptic functions of time, depending on the parameters h, f, k and two more
integration constants. Alternatively, they can be expressed in terms of Theta functions
in two variables. A detailed account of the inversion problem can be found in [113].
Kowalevski was also able to express all the dynamical variables in terms of s1 , s2
[238, 256]. To this end she widely used the properties of elliptic functions, which were
@seismicisolation
@seismicisolation
so popular at the time. Kötter [234] simplified and systematized Kowalevski’s method
and made it somewhat more transparent (See also [259] and references therein).
Several later trials to simplify Kowalevski’s and Kötter’s derivations by using Lax
representation of the equations of motion and methods of algebraic geometry have
led to more complicated solutions in terms of Theta functions of three variables, e.g.
[1, 26] (See also comments in [307]).
The motion of the body in Kowalevski’s case takes its simplest form in the plane
of the Kowalevski separation variables s1,2 . Let ei , i = 1, ..., 3, be the roots of the
cubic polynomial F(s). Two options are possible, either e1 is real and e2 , e3 are
complex conjugate, or the three roots are real and e1 ≥ e2 ≥ e3 . We shall introduce
also the notation e4 = h − k, e5 = h + k. Note that, for determinacy, k can always
be considered a non-negative constant, while a is made positive by a suitable choice
of the coordinate x axis.
Each of the variables s1 , s2 takes its values on one of the admissible intervals
on which (s) takes non-negative values. But, in view of Eq. (4.60), both variables
cannot vary simultaneously in one and the same interval. After the works of Appelrot
[9] (See also [108]), it turned out that in the generic case, when has no equal roots,
then real motions in Kowalevski’s case correspond to the intervals of variation of
Kowalevski’s variables shown in the following table (Table 4.3).
Kowalevski’s expressions for the phase variables in terms of the auxilliary vari-
ables s1 , s2 can be written as
β1 P1 + β2 P2 + β3 P3
p=− ,
α1 P1 + α2 P2 + α3 P3
1
q= ,
α1 P1 + α2 P2 + α3 P3
√ α1 P23 + α2 P31 + α3 P12
r = 2 ,
α1 P1 + α2 P2 + α3 P3
√ β1 P23 + β2 P31 + β3 P12
γ3 = − 2 ,
α1 P1 + α2 P2 + α3 P3
3 3
(en − e4 )αn Pn4 ∓ (en − e5 )αn Pn5
n=1 n=1
γ1 ± iγ2 = − , (4.62)
3 3
αn Pn4 ∓ αn Pn5
n=1 n=1
Table 4.3 Admissible intervals of Kowalevski’s variables s1 , s2 when has all roots distinct
1 −∞ < s2 ≤ e4 < e1 ≤ s1 ≤ e5 F has one real root e1
2 −∞ < s2 ≤ e1 < e4 ≤ s1 ≤ e5
3 −∞ < s2 ≤ e3 < e2 < e1 ≤ s1 ≤ e5 e3 < e4 < e1
4 −∞ < s2 ≤ e4 < e3 < e2 < e1 ≤ s1 ≤ e5
5 −∞ < s2 ≤ e3 < e2 < e1 < e4 ≤ s1 ≤ e5
@seismicisolation
@seismicisolation
where

Pn = (s1 − en )(s2 − en ),
√ √
Pn Pm (s1 ) (s2 )
Pnm = − ,
(s1 − s2 ) (s1 − en )(s1 − em ) (s2 − en )(s2 − em )
and the constant coefficients are given by

√
2en
αn = ,
F (en )
√ √ √
e2 e3 e3 e1 e1 e2
β1 = , β2 = , β3 = .
F (e1 ) F (e2 ) F (e3 )
However, those formulas serve to show how complicated the solution is. The generic
analytical solution is almost useless in the study of qualitative properties of motion.
Even in some cases, when the solution takes simpler form for certain values of the
parameters, it turns out to be more practical to return to direct solution of the equations
of motion under given conditions, instead of substituting in the generic solution.
In real motion, the point (s1 , s2 ) moves either in a rectangular area or in a semi-
infinite strip. In this area, the s1 , s2 are related to time through the ultra-elliptic inte-
grals. For such motions, the general behaviour of the angles of precession and proper
rotation was briefly investigated by Kozlov [240]. However, as the expressions of
the dynamical variables given by Kowalevski and Kötter are too complicated to help
drawing more qualitative conclusions about the motion, it acquires great importance
to point out some special classes of motions for which the ultra-elliptic quadratures
degenerate into elliptic or even into simpler ones. Those classes correspond to cases
when the polynomial (s) has multiple roots. They were first studied by Appelrot
[11] and investigated later in [67, 158] (See also [183] or [108]). An account of those
cases will be presented in Appendix B.
4.4 The Goryachev–Chaplygin Case: A Conditional

Integrable Case
Here, we are going to deal with a new type of integrable cases in rigid body dynamics
that we shall meet frequently later in this book, a conditional case. That is a case
when the complementary fourth integral exists only under one condition: the initial
motion (ω 0 , γ 0 ) satisfies the restriction
ω 0 I·γ 0 = 0,
and hence, from the integral of areas, we shall have at all subsequent times
@seismicisolation
@seismicisolation
ωI·γ = f = 0.
The angular momentum vector is horizontal at the initial moment and will stay so
all the time.
4.4.1 The Fourth Integral
When the distribution of mass in the body satisfies the conditions
A = B = 4C, y0 = z 0 = 0, (4.63)
the equations of motion (3.29) take the form
4 ṗ = 3qr,
4q̇ = −3 pr + aγ3 ,
ṙ = −aγ2 ,
γ̇1 + qγ3 − r γ2 = 0, γ̇2 + r γ1 − pγ3 = 0, γ̇3 + pγ2 − qγ1 = 0
(4.64)
where a = Mgx0 . As was shown first by Goryachev, it can be easily verified that
d aq
[r ( p 2 + q 2 ) − apγ3 ] = f, (4.65)
dt 4C
where f is the constant of areas. Now, we note, as was first done by Chaplygin, that
when
f = G·γ = 0, (4.66)
i.e. when the vertical component f of the kinetic moment of the body vanishes, the
quantity in the square bracket becomes an integral of motion
r ( p 2 + q 2 ) − apγ3 = G (4.67)
where G is an arbitrary constant. Note that, as in the three general cases of inte-
grability, this integral is a polynomial, cubic in the Euler–Poisson variables. The
integral (4.67) was obtained by Chaplygin and will be named Chaplygin’s integral.
The particular case found slightly earlier by Goryachev is subject to the additional
√
restriction G = 0 and an invariant relation 3 pγ3 = br, b is an arbitrary constant.
In the three general cases of integrability, the problem of motion was integrable
for all initial conditions, which is not the case here. The integral (4.67) is valid only
for motions satisfying the condition (4.66), i.e.
@seismicisolation
@seismicisolation
4.4 The Goryachev–Chaplygin Case: A Conditional Integrable Case 81
4 pγ1 + 4qγ2 + r γ3 = 0. (4.68)
Of course, it suffices to have this condition satisfied at the beginning of motion, and
it will be satisfied at all times since the quantity in hand is an integral of motion. This
condition means that the kinetic moment of the body should lie at the initial moment
of motion in a horizontal plane passing through the fixed point, and it will continue
to be in the same plane at all times.
4.4.2 Separation of Variables. Solution of the Equations of

Motion
As we have mentioned earlier, the mere presence of the four integrals of motion
does not always mean that a known procedure can be utilized to reduce equations
of motion to a separation of variables and, consequently, to complete the process
of explicit expression of all the phase variables of the problem in terms of certain
functions of time. In most cases, one has to design for a new case a new method,
almost independent of methods applied in the solution of other cases.
Here, we present for Goryachev–Chaplygin case the method which Chaplygin was
able to devise for separation of variables. We first write all the integrals of motion
1 h
[4( p 2 + q 2 ) + r 2 ] + aγ1 = = E,
2 C
4 pγ1 + 4qγ2 + r γ3 = f /C = 0,
r ( p 2 + q 2 ) − apγ3 = G,
γ12 + γ22 + γ32 = 1. (4.69)
Following Chaplygin, we introduce new variables u and v by the relations
r = u − v,
4( p 2 + q 2 ) = uv. (4.70)
Substituting in (4.69) and solving the first three equations, which are linear in
γ1 , γ2 , γ3 , and inserting the solution in the fourth equation, we determine p and
q. After some manipulations we write down the solution of the algebraic system
(4.69)–(4.70) is
U1 V2 − U2 V1
p= ,
8a
U1 V1 + U2 V2
q= ,
8a
r = u − v,
@seismicisolation
@seismicisolation
U +V
γ1 = − ,
2a(u + v)
U1 U2 − V1 V2
γ2 = − ,
2a(u + v)
U1 V2 + V1 U2
γ3 = − , (4.71)
2a(u + v)
where
U = u 3 − 2Eu − 4G, V = v 3 − 2Ev + 4G,

U12 = U − 2au, V12 = V − 2av,
U22 = −U − 2au, V22 = −V − 2av. (4.72)
Thus, we have expressed the six Euler–Poisson variables in terms of the two
variables u and v. It remains to determine the dependence of u, v on time. Using the
first equation of (4.70), the third of (4.64) and (4.71), we get
du dv U1 U2 − V1 V2
− = −aγ2 = . (4.73)
dt dt 2(u + v)
Also, using the second equation of (4.70), the first and second of (4.64) and (4.71),
we find
du dv
v +u = 8( p ṗ + q q̇)
dt dt
= 2aqγ3
U1 V1 + U2 V2 U1 V2 + V1 U2
=− , (4.74)
8a (u + v)
so that we finally obtain two differential equations in u and v

du dv
2(u + v) = U1 U2 , 2(u + v) = V1 V2 . (4.75)
dt dt
The last equations can be put in the form
du dv udu vdv dt
− = 0, + = . (4.76)
U1 U2 V1 V2 U1 U2 V1 V2 2
Integrating from the initial position u 0 , v0 at t0 , we can now write

u v
du dv
− = 0,
u0 4a u − (u − 2Eu − 4G)
2 2 3 2 v0 4a v − (v − 2Ev + 4G)2
2 2 3
u v
udu vdv 1
+ = (t − t0 ).
u0 4a 2 u 2 − (u 3 − 2Eu − 4G)2 v0 4a 2 v 2 − (v 3 − 2Ev + 4G)2 2
(4.77)
@seismicisolation
@seismicisolation
4.4 The Goryachev–Chaplygin Case: A Conditional Integrable Case 83
With this step, we have completed the process of separation of variables as was
proposed by Chaplygin. Integrals in (4.77) are in the standard form of hyper-elliptic
integrals, where the denominators of terms are square roots of polynomials of the
sixth degree. The inverse functions, i.e. the solution of the system (4.77) in the
variables u, v as functions of time can be expressed in terms of Theta functions of
two variables, each of which is a linear function of time. This step was completed
for the present problem by Marcolongo and Olsson and the reader may be referred to
their original works [272, 298]. The behaviour of the variables u, v depending on the
number and multiplicity of real roots was studied by Dokshevich [67]. The analysis
is much simpler than in Kowalevski’s case, since here we deal with a system with
only two parameters.
It is noteworthy that the separation of variables in Goryachev–Chaplygin’s case
was attained in [240] by an elegant approach, using canonical Andoyer–Deprit vari-
ables. Also, certain qualitative properties of motion were studied, e.g. the long time
behaviour of the angles of precession and proper rotations. Particularly simple solu-
tions of Goryachev–Chaplygin case are discussed in Appendix C.
4.5 Integrability and Nonintegrability Issues
As mentioned above, Kowalevski used the analytical theory of differential equations

to isolate all possible combinations of the physical parameters A, B, C, x0 , y0 and z 0 ,
for which the general solution of the Euler-Poisson system of equations of motion,
for arbitrary initial conditions, can be expressed as single-valued functions that have
only poles as their singularities in the complex plane of time t. Her approach led to
the third integrable case in the classical problem of motion of a rigid body about a
fixed point. Kowalevski’s result was continued and supplemented by Appelrot [10]
and Lyapunov [266]. Lyapunov’s method is characterized by the introduction of a
small parameter and involves the analysis of the monodromy of variational equations
for some particular solutions of Euler–Poisson’s equations. It showed that, except for
the cases of Euler, Lagrange and Kowalevski, the general solution of those equations
branches in the complex time plane in all cases.
The subject of integrability was a favourite field for several mathematicians over
a large period of time.
As we have noticed, the fourth integral in the three general integrable cases has
polynomial form in the Euler–Poisson variables. Husson investigated the possibility
that there exists an additional general integral algebraic in those variables [154, 155].
It turned out that this happens only in the three known integrable cases. Ziglin [435]
has also shown that a meromorphic complementary first integral of the Euler–Poisson
equations exists only in the cases of Euler, Lagrange and Kowalevski and also in the
conditional ( f = 0) Goryachev–Chaplygin case. We note also the recent proof of
the non-existence of meromorphic integrals [268], obtained by means of a technique
based on differential Galois theory.
@seismicisolation
@seismicisolation
One has to mention also the contributions and different approaches of Poincaré,
Kozlov [240, 245] and Ziglin [434, 435] and others. For a more detailed presen-
tation see, e.g. [41]. Detailed analysis of the complete algebraic integrability and
the anaytical structure of the solution in the complex plane for the integrable cases
can be found in [2]. However, we note also that although methods of proving inte-
grability or non-integrability had succeeded in the study of the classical problem,
they become much less effective in more general problems involving the effects of
additional potential and gyroscopic forces on the body.
@seismicisolation
@seismicisolation
Chapter 5
The Motion of a Heavy Gyrostat
Originally, the gyrostat, as the terminology was coined by Lord Kelvin, is a heavy
rigid body with a rotor or a fly-wheel spinning with a constant angular speed about
its axis of symmetry. The subject gained a great interest at the first two decades of
the twentieth century. Examples are the two books by Crabtree [59] (1909) and Gray
[133] (1918), devoted exclusively to describing gyroscopic phenomena, specially the
stabilizing effects of rotors, and the ways to make use of them in warfare of World
War I. Today, gyroscopic apparatuses are indispensable in cell phones, in so many
applications in terrestrial and cosmic navigation and in technology. Most useful is
the stabilizing effect of fast rotors on normally unstable motions and equilibria.
In this chapter, different types of mechanical systems having the same equations
of motion as the gyrostat are presented. General and conditional integrable cases
of motion are presented. In fact, these are generalizations of the relevant integrable
cases in the classical problem, and reduce to them when the gyrostatic momentum
vanishes.
At present, several particular solutions to the problem of motion of a gyrostat are
known, namely, 13 solutions. Some of them are generalizations of classical counter-
parts by adding a gyrostatic momentum. Other cases lose their meaning when the
gyrostatic moment vanishes.
5.1 Models of the Gyrostat
5.1.1 The Classical Model
Consider a system S, composed of two joint rigid bodies. The first, S0 , which we
shall call the carrier (or the main) body, is fixed in the inertial space from its point

https://doi.org/10.1007/978-3-030-96336-1_5
@seismicisolation
@seismicisolation
86 5 The Motion of a Heavy Gyrostat
O. The second body, the rotor S1 , is an axially symmetric body fixed from its axis
of symmetry in the main body. Its centre of mass, O1 , lies on its axis of symmetry.
Usually, such a symmetric body is called gyroscope. Because of the symmetry of the
rotor, its rotation does not change the distribution of mass in the system. Let I and
r0 be the inertia matrix and the position vector of the centre of mass of the system,
referred to the system of axes O x yz fixed in the main body. Let also J be the inertia
matrix of the rotor with respect to a system of axes O1 x1 y1 z 1 fixed in it with z 1
along its axis of symmetry. From symmetry, it is clear that O1 x1 y1 z 1 is a system of
principal axes of the rotor and hence we can write J = diag(J1 , J1 , J ).
Let the rotor be set and kept in motion about its axis with a constant angular
−−→
velocity , by means of some device. Let r1 = O O1 and denote by r the position
vector of a mass element dm of the system with respect to O. The velocity of that
element is ω × r if it belongs to S0 and ω×r+ e × r for elements of S1 , where r
is the position vector of the mass element of the rotor with respect to O1 . The angular
momentum of the system can be written as

G= r × (ω×r)dm + r × (ω×r+e × r )dm
S0 S1

= r × (ω×r)dm + r × (e × r )dm
S0 S1

= ωI+ (r1 + r )×(e × r )dm
S1

= ωI + r1 ×(e × r dm) + r ×(e × r )dm
S1 S1
= ωI + 0+eJ
= ωI+J e. (5.1)

Here we have used S1 r dm = 0. The last expression will be written as
G = ωI+κ, (5.2)
where κ = J e is the gyrostatic momentum, the angular momentum of the rotor

relative to the carrier body. It is directed along the axis of symmetry of the rotor.
Now we write down the equation of motion of the system. The mutual forces
between the main body and the rotor are internal forces in the system and do not
appear in this equation. One has
Ġ + ω × G = Mgγ×r0 .
Since κ is kept constant in the body, κ̇ = 0, and the last equation reduces to
ω̇I + ω × (ωI + κ) = Mgγ×r0 . (5.3)
@seismicisolation
@seismicisolation
5.1 Models of the Gyrostat 87
This is the final form of the dynamical equation of motion of the gyrostat. Together
with Poisson’s equation
γ̇ + ω × γ = 0, (5.4)
one obtains a closed system which we now write in the following scalar form of six
first-order differential equations:
A ṗ + (C − B)qr + κ3 q − κ2 r = Mg(z 0 γ2 − y0 γ3 ),
B q̇ + (A − C) pr + κ1r − κ3 p = Mg(x0 γ3 − z 0 γ1 ),
C ṙ + (B − A) pq + κ2 p − κ1 q = Mg(y0 γ1 − x0 γ2 ), (5.5)
This system admits the general integrals:
I1 ≡ Ap 2 + Bq 2 + Cr 2 + Mg(x0 γ1 + y0 γ2 + z 0 γ3 ) = h,
I2 = γ12 + γ22 + γ32 = 1,
I3 = (Ap + κ1 )γ1 + (Bq + κ2 )γ2 + (Cr + κ3 )γ3 = f. (5.7)
The first integral is usually termed Jacobi’s integral for the system, since it is different
from the total energy of the system, which contains terms linear in the components
of ω.
When the angular speed of the rotor vanishes, gyrostatic momentum κ = 0,
and equations (5.5) and the integrals (5.7) reduce to their counterparts of the classical
problem.
5.1.2 The Free Rotor Model
In the previous model, the angular velocity of the rotor was kept constant relative to
the carrier body. In an interesting alternative, due to Levi-Civita [261], the rotor is
left to move freely around its axis of symmetry fixed in the body, so that the system
will have an additional rotational degree of freedom. Let χ be the angle of rotation of
the rotor relative to the body. Using the same symbols as in the previous subsection,
the kinetic energy of the system is expressed as the sum of two parts

1 1
T = (ω×r)2 dm + (ω×r+χ̇e × r )2 dm
2 S0 2 S1

1 1 2
= (ω×r) dm + χ̇
2
(ω×r) · (e × r )dm + χ̇ (e × r )2 dm
2 S0 S1 2 S1
@seismicisolation
@seismicisolation

1 1
= ωI·ω + χ̇ (r1 +r )×(e × r )dm·ω + χ̇2 r × (e × r )dm·e.
2 S1 2 S1
Noting that

r1 ×(e × r dm) = 0, r × (e × r )dm = eJ =J e,
S1 S1
we obtain
1 1
T = ωI·ω + J χ̇e·ω+ J χ̇2
2 2
and hence the Lagrangian of the system may be written as
1 1
L= ωI·ω + J χ̇e·ω+ J χ̇2 − a·γ (5.8)
2 2
where a = Mgr0 . Obviously, the angle χ is a cyclic variable. The corresponding
cyclic integral is
∂L
= J (e·ω + χ̇) = κ, (5.9)
∂ χ̇
κ is an integration constant. Note that this integral means that the component of the
total angular velocity of the rotor along its axis of symmetry remains constant during
motion, i.e.
κ
e·ω + χ̇ = .
J
Now, ignoring the cyclic coordinate, we obtain the Routhian
R = L − κχ̇
1 1 κ2
= ωI·ω − J (ω·e)2 +κe·ω − a·γ −
2 2 2J
1
= ω Ĩ · ω + κe·ω − a·γ (5.10)
2
where
I˜i j =Ii j − J ei e j (5.11)
κ 2
and a constant 2J has been ignored. In the way described in Chap. 3, the Euler
dynamical equation derived from this Routhian are
ω̇ Ĩ + ω × (ω Ĩ + κ) = γ × a, (5.12)
where κ = κe.
@seismicisolation
@seismicisolation
5.1 Models of the Gyrostat 89
This equation has the same structure as (5.3), but the matrix Ĩ is not simply the
matrix of inertia of the system, but depends on the axial moment of inertia of the rotor
and on the orientation of its axis relative to the main body. The meaning of the vector
κ, the gyrostatic momentum is different in the two equations. Moreover, it should
be noted that, in view of (5.11), the matrix Ĩ may not satisfy conditions, normal
to ordinary inertia matrix of the simple body, like positivity of diagonal elements,
triangle inequalities, etc.
5.1.3 Joukovsky’s Model
Generalizing previous particular cases considered by Stokes and Neumann,

Joukovsky established that “a fluid mass with an initial velocity in a multiply-
connected cavity in the rigid body performs an action that is similar to the action of
some rotor attached to the rigid body” [163] (see also [41, 286]).
The gyrostatic moment can also be due to internal cyclic degrees of freedom such
as circulation of fluid in tubes inside the body or to forced stationary motions as
motors, whose axes are fixed in the body.
As will be seen in a later chapter, terms in the equations of motion similar to
gyrostatic momentum appear in problems of motion of a perforated rigid body (a
body bounded by a multi-connected surface) in a liquid, as a result of the presence
of circulations through perforations.
5.2 Equations of Motion in Hamiltonian Form
As in the classical problem (Chap. 3), one may use some generalized coordinates
like Euler’s angles, to construct the Hamiltonian function and the canonical equations
of motion, involving those coordinates and momenta conjugate to them. This form
of the equations of motion is rarely used in applications and is left as an exercise.
Non-canonical equations
∂R
M= = ωI+κ, (5.13)
∂ω
so that
ω= (M−κ)I−1 . (5.14)
Also, the Hamiltonian corresponding to the same Routhian as a function in M and

γ is
1
H = ωI·ω + a·γ
2
@seismicisolation
@seismicisolation
1
= (M−κ)I−1 · (M−κ) + a·γ
2
1
= MI−1 · M−κI−1 · M + a·γ, (5.15)
2
so that the equations of motion can be written as
∂H ∂H
Ṁ = M× +γ× ,
∂M ∂γ
∂H
γ̇ = γ× . (5.16)
∂M
or, in the expanded form,
Ṁ = M × (M − ˇ)I−1 + γ × a,
γ̇ = γ×(M−κ)I−1 . (5.17)
5.3 Tables of Integrable Cases
Equations (5.5 and 5.6) have three known general and one conditional integrable
cases, which generalize the four cases of a simple heavy body. Those are listed in
the following Tables 5.1 and 5.2.
Table 5.1 Unconditional cases

Author Conditions
1 Joukovsky 1885 [163] and Volterra 1899 [366]
Euler κ1 = κ2 = κ3 = 0. gr0 = 0.
I4 = (Ap + κ1 )2 + (Bq + κ2 )2 + (Cr + κ3 )2 .
2 Axially symmetric case B = A,
(Generalization of Lagrange’s top) x0 = y0 =
0,
I4 = Cr + κ3 . κ 1 = κ2 =
0.
3 Yehia 1986 [380], [383]∗ A=B=
2C,
Kowalevski 1889 (κ = 0 ) [238] z 0 = 0,
I4 = ( p 2 − q 2 − a1 γ1 + a2 γ2 )2 + (2 pq − a1 γ2 − a2 γ1 )2 κ 1 = κ2 =
+2κ (r − κ) ( p 2 + q 2 ) − 4κγ3 (a1 p + a2 q), 0.
where κ = κ3 /C, a1 = Mgx0 /C, a2 = Mgy0 /C
∗ The case (3) was rediscovered in 1987 by Komarov [225] and Gavrilov [109]. In the monograph
[41], it is attributed to Yehia, Komarov and Gavrilov, but in the Russian literature it is mostly called
Kowalevski–Yehia’s case
@seismicisolation
@seismicisolation
5.4 The Case of Joukovsky and Volterra 91
Table 5.2 Conditional cases f = 0

1 Sretensky (1963) [341]. A = B = 4C, z 0 = 0,
Goryachev–Chaplygin κ1 = κ2 = 0, κ3 = Cκ.
1900–1901 (κ = 0).
I4 = (r − κ)( p 2 + q 2 ) − γ3 (a1 p + a2 q).
where κ = κ3 /C, a1 = Mgx0 /C, a2 = Mgy0 /C
5.4 The Case of Joukovsky and Volterra
The first integrable case, which generalizes Euler’s case, i.e. a balanced gyrostat or
a gyrostat under no external torques, was noted in 1885 by Joukovsky in his study of
the motion by inertia of a body containing liquid-filled cavities [163]. He also devised
a geometric-mechanical interpretation of the motion in that case. Independently, in
a trial to explain the displacement of Earth’s poles by adding a rotor to the model of
rigid Earth, Volterra gave in 1899 the full solution of the equations of motion in terms
of Weierstrass’ elliptic functions σi of time [366]. Those functions are complex in
general. An alternative but real solution in terms of Jacobi’s elliptic functions was
constructed by Wittenburg [369]. Volterra’s solution and stability analysis of the
permanent rotations were reconsidered in [18].
5.5 The Case of Axially Symmetric Gyrostat
The axi-symmetric gyrostat is a trivial generalization of Lagrange’s top and the

solution of the equations of motion for it is practically the same as that of Lagrange’s
case. In a later Chap. 12, we will show a much richer generalization of this case.
We now prove the following
Theorem 5.1 Any integrable case of an axi-symmetric body in a potential field, in

which both φ and ψ are cyclic variables, can be generalized by the addition of a
rotor aligned with the axis of symmetry.
Theorem 5.2 Consider the motion of an axi-symmetric gyrostat, with a gyrostatic

momentum κ aligned along the axis of symmetry of the carrier body. The motion
of the axis of the gyrostat is identical with the motion of a simple body with the
same moments of inertia of the carrier body and moving in the same potential. The
gyrostatic momentum is compensated by an additional angular speed κ/C given to
the body about its axis, C being the axial moment of inertia of the body.
Conversely, the motion of the axis of a simple axi-symmetric body, which given
an additional angular speed around that axis is identical to the motion the axis of
a similar body, which carries a rotor with gyrostatic momentum κ = C.
@seismicisolation
@seismicisolation
Those two theorems can be proved by writing the Lagrangian of the simple body
in a field with potential V (θ). Let the moments of inertia of the body be C about its
z-axis of symmetry and A about any axis orthogonal to it. For such body, we have
from (3.44)
1
L= [A( p 2 + q 2 ) + Cr 2 ] − V (θ)
2
1
= [A(θ̇2 + sin2 θψ̇ 2 ) + C(ψ̇ cos θ + ϕ̇2 ) − V (θ). (5.18)
2
We now study the motion in another reference frame, which is rotating abut the z-axis
with a constant angular rate . This can be achieved by a substitution
ϕ = ϕ + t,
which preserves the holonomicity of the system. The Lagrangian transforms to

1
L= {A(θ̇2 + sin2 θψ̇ 2 ) + C[ψ̇ cos θ + (ϕ̇ + )2 )]} − V (θ)
2
1
= {A(θ̇2 + sin2 θψ̇ 2 ) + C(ψ̇ cos θ + ϕ̇ )2 }
2
1
+C(ψ̇ cos θ + ϕ̇ ) − V (θ) + C2 .
2
Ignoring the last constant term, this can be regarded as describing the motion of the
body referred to fixed axes in it, but with the coordinate ϕ instead of ϕ and with the
same potential V and additional gyroscopic term C(ψ̇ cos θ + ϕ̇ ) = Cr . The
last term is the contribution of a gyrostatic momentum κ directed along the z-axis
κ= (0, 0,C).
This proves Theorem 5.1 and the first part of Theorem 5.2. The second part of
Theorem 5.2 follows naturally.
Those results may be used to express the quadrature resulting from separation of
variables in the case 2 of Table 5.1, i.e. the gyrostatic generalization of Lagrange’s
case from the quadrature (4.42) by replacing r0 by r0 + κC3 , so that it becomes

dγ3
t= .
(1 − γ32 )(E − aγ3 ) − A12 [ f − (Cr0 + κ3 )γ3 ]2
5.6 Yehia’s Case
The history of the third case in Table 5.1 has experienced some confusion and misun-
derstandings. It was direct and easy, guided by the same principle of conservation of
angular momentum, to obtain the integrable case of Joukovsky, as a generalization of
@seismicisolation
@seismicisolation
5.6 Yehia’s Case 93
Euler’s case in the classical problem, by adding a constant gyrostatic momentum. The
generalization of Lagrange’s case of a symmetric body was even easier. Nevertheless,
the search of a gyrostatic generalization of Kowalevski’s case was so futile, that it
was generally believed that, unlike Euler’s and Lagrange’s cases, Kowalevski’s case
does not admit generalization by the addition of a gyrostatic momentum. This trend
may have been augmented by three contradicting results published in the mid-sixties
by Keis:
(1) In the first of those works [167] (1963), Keis claimed having obtained general-
izations to the gyrostat problem for four known integrable and particular cases
of the classical problem, namely, Lagrange’s, Hess’, Bobylev–Steklov’s and
Delone’s cases. The first case, the generalization of Lagrange’s top, is trivially
simple and the next two cases will be commented on in the relevant section on
particular solutions. The last case (Delone’s) is a special case of Kowalevski’s,
when Kowalevski’s integral takes a zero value and splits into two invariant rela-
tions (See Sect. 4.3). Keis added to the body a constant gyrostatic momentum,
aligned with the centre of mass in the equatorial plane of the inertia spheroid
and claimed that the resulting system admits two invariant relations generalizing
those of Delone’s case. This claim was cited as being true in the review book
[256].
(2) In the second paper [168] (1964), may be after realizing the flaw in his 1963
paper (cited in [168]), Keis used the method of Husson [154, 155] to give
another theorem asserting that the equations of motion of the heavy gyrostat
with Kowalevski’s configuration A = B = 2C admit an algebraic complemen-
tary integral only when the gyrostatic momentum vanishes (κ1 = κ2 = κ3 = 0).
He then formulated it as1 “If x02 + y02 + z 02 = 0 and κ21 + κ22 + κ23 = 0 a fourth
algebraic integral is possible only when A = B, x0 = y0 = 0, κ1 = κ2 = 0”,
i.e. only in the case of Lagrange when both the centre of mass and the gyrostatic
momentum are directed along the axis of dynamical symmetry of the body. This
meant that Kowalevski’s case has no extension to the gyrostat problem.
(3) In the third paper [170] (1965), Keis used Golubev’s method [113] (In fact,
Poincaré’s method of small parameter) to establish a new result. The search for
all cases, when all the solutions of the equations of motion of a heavy gyrostat
are single-valued, reduces to investigation of the solution in three cases:
(a) The torque-free gyrostat, or gyrostat fixed from its centre of mass, (x0 =
y0 = z 0 = 0).
(b) The axi-symmetric gyrostat ( A = B, x0 = y0 = 0, κ1 = κ2 = 0).
(c) The “Kowalevski gyrostat” (A = B = 2C, y0 = z 0 = 0, κ1 = κ2 = 0).
In this paper, the presence of the third component of the gyrostatic momentum
does not give rise to multi-valued solutions in any of the three cases, up to
the second degree of a small parameter. However, the conditions for case c
are considered as necessary. The only way to ensure integrability is to find the
1 Here we use the notation adopted in the present book.
@seismicisolation
@seismicisolation
complementary fourth integral of motion, a step which was not considered by

Keis. Strangely, in the third paper the author does not refer to any of the other
two papers, each of which announces a conflicting result.
Probably, influenced by the result in the second paper of Keis, published in the
most influential Russian mechanics journal PMM, Kharlamov and coworkers con-
centrated on the search for particular solutions of the equations of motion. In this
respect, they have succeeded in constructing the most part of the cases of that type
known up to date. To this end, they used equations of motion in the form of Euler–
Poisson and various modified forms. Kharlamov [198] obtained a particular solution
of the heavy gyrostat with the Kowalevski configuration A = B = 2C, y0 = z 0 = 0,
involving a gyrostatic moment along the axis of dynamical symmetry of the body
under certain restrictions on the initial motion. His case characterized by the exis-
tence of an invariant relation quadratic in the angular velocities fits as a special case
of the third general integrable case in Table 5.1.
The full generalization of Kowalevski’s case by the addition of a rotor to the body
came out, in our work [380], in almost a century (exactly 98 years) after the publica-
tion of Kowalevski’s case (See also [383]). Actually, it was not found as a solution of
Euler–Poisson equations or any of their modifications. It was one of the first results
obtained by the completely new method devised by the author of the present book
to construct integrable 2D conservative mechanical systems, which admit a comple-
mentary integral polynomial in the velocities. After constructing a several-parameter
integrable time-irreversible system of the above type, the parameters of the system
are given certain values, such that the metric of the system could be identified with
that of the Routhian reduction of the rigid body dynamics and then potential and
gyroscopic forces could be identified and only then the appropriate Euler–Poisson
equations are verified and the presentation of the new case in [380] was made in the
last context.2 The details of the method will not be presented here for space consid-
erations, but the reader can get some acquaintance with it from the early papers [381,
419]. This method has proved fruitful and still gives new integrable cases of much
more complicated problems in particle and rigid body dynamics (see, e.g. [411, 413,
422, 423]). All cases pertaining to rigid body dynamics obtained in this way will
be described later in this book. They form the most part of the list of conditional
integrable cases in Chap. 13.
The question of integrability of Eqs. (5.5 and 5.6) did not attract as much interest
as the problem of a simple heavy body. Only one result on this aspect is known.
It generalizes the above-mentioned theorem of Husson to the present problem of
motion of a gyrostat.
Theorem 5.3 (Gavrilov [110, 111]) The equations of motion of heavy gyrostat (5.5
and 5.6) possess an additional algebraic first integral only in the three cases of
Joukowsky, Lagrange and Yehia.
2This result was announced at the International Conference on Mechanics held in Moscow Uni-
versity, January 1986.
@seismicisolation
@seismicisolation
5.6 Yehia’s Case 95
5.6.1 Separation of Variables
For greater clearness we first write down the equations and integrals of motion in the
present case, after adding a simplifying condition y0 = 0, which can be attained by
a coordinate rotation. The equations have the form
2 ṗ − q(r − κ) = 0,
2q̇ + p(r − κ) = aγ3 ,
ṙ + aγ2 = 0,
γ˙1 + qγ3 − r γ2 = 0, γ˙2 + r γ1 − pγ3 = 0, γ˙3 + pγ2 − qγ1 = 0, (5.19)

κ3
in which a = Mgx0
C
,κ = C
. The three general integrals of motion may be written as
I1 = 2( p 2 + q 2 ) + r 2 + 2aγ3 = 2h,
I2 = 2( pγ1 + qγ2 ) + (r + κ)γ3 = f,
I3 = γ12 + γ22 + γ32 = 1,
I4 = ( p 2 − q 2 − a1 γ1 + a2 γ2 )2 + (2 pq − a1 γ2 − a2 γ1 )2
+2κ (r − κ) ( p 2 + q 2 ) − 4κγ3 (a1 p + a2 q)
=K (5.20)
and here we retained the same names for h and f after dividing by C and K is an
arbitrary constant.
Unlike the case of Kowalevski’s top (with κ = 0), the explicit solution of the
equations of motion in terms of time in Yehia’s case is still unsuccessful. Separation
coordinates analogous to Kowalevski’s s1 , s2 (Chap. 4 Sect. 4.3) were not found, even
on the zero level of the areas integral. However, there is an indirect indication about
the class of functions needed to describe this solution. An idea of special interest
was presented in [145], which relates Kowalevski’s case to a special version ( f = 0)
of an integrable case of the problem of motion of a rigid body in a liquid, known as
Clebsch’s first case (Case 2 of Table 10.1. Chap. 10) A bi-rational complex trans-
formation was found relating the two sets of variables describing the two integrable
problems, so that explicit solution for one of the cases can be obtained from that of
the other. In the meantime, Clebsch’s case is known to be solvable in terms of Theta
functions with two arguments.
As was established by Gavrilov [110]: “The gyrostat of Yehia can be realized in
a similar way as (the full f = 0) Clebsch’s geodesic motion on E 3 . This leads, in
particular, to formulas for its explicit solution in terms of genus-two hyper-elliptic
Theta functions [233]”. Of course, this construction is not practical as a method of
solution and there must be another direct way to obtain the solution. This way has
not been found yet.
Komarov and Tsiganov [229] (See also [228]) considered the trajectory iso-
morphism of what they call “the Kowalevski gyrostat” and the Clebsch problem.
@seismicisolation
@seismicisolation
Although appeared fifteen years later, the last works do not contain any reference to
Gavrilov’s work.
In a recent work [326], Ryabov shows that the separated equations of the Yehia
(Kowalevski–Yehia) case, on its zero level of area’s integral, can be formally written
in the Abel–Jacobi form analogous to (4.58) with (s) as a polynomial of degree
five in the variable s. However, the relation of the original variables of the problem
to separated ones are not obtained, so that the problem of separation of variables
cannot be considered complete yet, even on the level f = 0.
An earlier work relying on the Lax pair representation of the equations of motion
constructs separation variables that are simultaneously suitable, on the zero level
of the areas integral, for what the author names as gyrostatic generalizations of
Kowalevski’s and Goryachev–Chaplygin’s cases [249]. It is claimed there that the
given separation is “much simpler than the Kowalevski separation”. However, the
impact of that situation on the solution of the equations of motion was not considered.
Further (unpublished) results are announced by Fedorov et al. concerning the
case of a gyrostat in two constant uniform fields, which includes the present case as
a special version. This will be commented later in Chap. 14 Sect. 14.2.1.1.
The gyrostatic generalizations of Appelrot’s classes of motions are discussed in
Appendix D.
5.7 The Conditional Case of Sretensky
In [341], Sretensky found the modification of the complementary integral in the gen-
eralization of Goryachev–Chaplygin’s case. He also generalized the procedure due
to Chaplygin for explicit solution (See Chap. 4 Sect. 4.4) by changing the definitions
of the three quantities r, U and V in (4.70)–(4.72) to be
r = u − v − κ,
U = u(u − κ)2 − 2Eu − 4G,
V = v(v + κ)2 − 2Ev + 4G. (5.21)
The ultra-elliptic quadratures remain the same as in (4.77). The investigation of the
critical sets and bifurcation diagrams in Sretensky’s case is performed in [172, 174]
(See also [183]). The results generalize relevant ones for Goryachev–Chaplygin’s
case, but they are much more complicated in view of the presence of three significant
parameters. For more detail about Sretensky’s case see Appendix E.
5.8 Some Applications of the Gyrostat Motion
We have seen that the presence of the gyrostatic momentum leads to appearance of
a gyroscopic moment κ × ω in the equations of motion.
@seismicisolation
@seismicisolation
5.9 Exercises 97
5.9 Exercises
1- Show that the Lagrangian
1
L= ωI·ω + κ · ω − V (γ),
2
describes the motion of a gyrostat with gyrostatic momentum κ about a fixed point,
while acted upon by axially symmetric forces with potential V (γ) (γ is the unit
vector along the axis of symmetry of the forces). Deduce the equations of motion in
the form
∂V
ω̇I + ω × (ωI+κ) = γ × ,
∂γ
γ̇ + ω × γ = 0.
2- In the previous problem, show that all possible axes of stationary motions lie
on the cone, with vertex at the origin and generators passing through the points of
the spherical curve
∂V 2 ∂V
[γ · (γI× )] − [κ·(γ × γI)][κ · (γ × )] = 0,
∂γ ∂γ
γ 2 = 1,
and the angular velocity of the gyrostat about the axis in the direction of γ is given
by any of the expressions
∂V ∂V
γ · (γI × ∂γ
) κ·(γ × ∂γ
)
ω= = ∂V
.
κ·(γ × γI) γ·(γI× ∂γ )
3- An axially symmetric gyrostat, moving about a fixed point under its own weight,
has its centre of mass on its axis of symmetry and the gyrostatic momentum is
collinear with that axis. Show that the upper equilibrium position of the gyrostat,
which is unstable in the absence of gyrostatic moment, can always be stabilized and
find the minimum angular velocity necessary for that effect.
4- Write down the Hamiltonian and Hamiltonian equations of motion of a heavy
gyrostat moving about a fixed point, using Euler’s angles as generalized coordinates.
5- A system composed of a main body S0 fixed from one point O, while carrying
another body S1 whose axis O P is fixed in S0 by means of a smooth cylindrical
hinge and freely rotates about this axis. Let O O = r1 and e is the unit vector in the
direction of O P. Show that the kinetic energy of the system is
1 1
T = ωI·ω + χ̇eJ·ω + χ̇2 eJ · e+M1 χ̇ω·[r1 ×(e × r0 )],
2 2
@seismicisolation
@seismicisolation
where I is the inertia matrix of the system at the fixed point, J is the inertia matrix
of the second body at O , M1 is the mass of S1 and r0 is the position vector of its
centre of mass.
@seismicisolation
@seismicisolation
Chapter 6
Motion of a Rigid Body About a Fixed
Point in the Field of a Distant Newtonian
Centre and Brun’s Problem
In the preceding chapters, we concentrated on the study of motion of a rigid body

about a fixed point in the uniform gravity field. This field serves also as a good
approximation for the forces acting on the body in most purposes. However, for
certain applications, like precise surveying instruments, the slight variations in the
Earth’s gravitational field from point to another must be taken into consideration.
These variations play a decisive role in the determination of rotational motion of
artificial satellites of Earth, especially those whose tasks demand high precision of
orientation, for terrestrial, cosmic or astronomical purposes.
In the present chapter, we study in detail another approximate model for one rigid
body moving in the gravitational field of a fixed body. In this model, we make two
assumptions:
(a) The fixed body is spherically symmetric. This enables us to treat it as a point
mass concentrated at the centre of the body.
(b) The diameter of the moving body is very small compared to the distance
between the fixed point of the moving body and the centre of the fixed body.
Let a body of mass M be in motion about its fixed point O at a distance R from
a fixed Newtonian attraction centre of mass M at O . Denote by γ the unit vector
−−→
in the direction O O and by V the potential of the body in the field of the centre. To
calculate the moment of gravitational forces about O, we take a mass element dm
of the body whose position vector referred to O is r :
μdm
The potential of this element will be d V = − |Rγ+r| , the force exerted on it by the
attraction centre is dF = − μ(Rγ+r)dm
|Rγ+r|3
and its moment about O is dL = − μRγ×rdm
|Rγ+r|3
,
where μ is Gauss’ constant of the centre (μ = M ×Newton’s gravitational constant).
Integrating over the whole mass of the body, we get

dm
V = −μ , (6.1)
|Rγ+r|

https://doi.org/10.1007/978-3-030-96336-1_6
@seismicisolation
@seismicisolation
100 6 Motion of a Rigid Body About a Fixed Point in the Field …

rdm
L = −μRγ× (6.2)
|Rγ+r|3

and recalling that |Rγ+r| = R 2 + 2Rγ·r+r 2 , one can write (2) in the form
∂V
L = γ× . (6.3)
∂γ
6.1 Approximate Form of the Potential
It is evident from (6.3) that the knowledge of the potential V is sufficient for complete
determination of the moment L. However, in most cases it is difficult to calculate V
from (6.1) due to the complexity of the shape of the body, its mass density or because
the mass distribution in the body is unknown. In those cases, an efficient solution is
to expand the potential in powers of R1 and keeping terms up to the third degree in
this parameter. First we write the expansion of the function
1 1 2γ·r r2 1
= [1 + + 2 ]− 2
R 2 + 2Rγ·r+r 2 R R R
1 1 2γ·r r2 3 2γ·r r2 1
= [1 − ( + 2)+ ( + 2 )2 + o( 3 )]
R 2 R R 8 R R R
1 γ·r r2 3 1
= − 2 − 3
+ 3
(γ·r)2 + o( 4 ).
R R 2R 2R R
Inserting this expression into (6.1), and neglecting terms of degrees higher than the
third, we obtain

1 1 1 3
V = −μ[ dm − 2 γ· rdm − r dm +
2
(γ·r)2 dm]
R R 2R 3 2R 3

M M 1 1
= −μ[ − 2 γ·r0 + (3 γ γ
i j xi x j dm − tr (I))]
R R 2R 3 2
M M 1 1
= −μ[ − 2 γ·r0 + 3
(3γ Ī·γ − tr (I))]
R R 2R 2
M M 1 1 1
= −μ[ − 2 γ·r0 + 3
(3γ( tr (I)δ − I)·γ − tr (I))]
R R 2R 2 2
M M 1
= −μ[ − 2 γ·r0 + (tr (I) − 3γI·γ)]. (6.4)
R R 2R 3
In expanded form that is
μM μM
V =− + 2 (x0 γ1 + y0 γ2 + z 0 γ3 )
R R
@seismicisolation
@seismicisolation
6.1 Approximate Form of the Potential 101
μ 3μ
− 3
tr (I) + (Aγ12 + Bγ22 + Cγ32 ). (6.5)
2R 2R 3
As from (6.3) only derivatives of V with respect to components of γ, the constant
terms may be discarded and thus the potential can be written as
1
V = Mg(x0 γ1 + y0 γ2 + z 0 γ3 ) + λ(Aγ12 + Bγ22 + Cγ32 )
2
1
= Mgr0 · γ + λγI · γ, (6.6)
2
where g is the gravity field intensity of the centre at the fixed point O of the body and
λ = 3gR
. Formula (6.6) gives the approximate form of the potential of the body by the
knowledge of its centre of mass and moments of inertia, without need to completely
specify the distribution of mass in the body.
Remark: Note that for a Newtonian centre of attraction both parameters g and
λ are positive. In certain physical problems, the Newtonian attraction is replaced by
Coulomb’s electric interaction (e.g. [58]). In that case g and λ can be either positive
or negative and I,Mgr0 are replaced, respectively, by the inertia matrix of the electric
charges and the moment of those charges multiplied by the intensity of the electric
field of the central charge at O.
6.2 Brun’s Problem
Brun considered the motion of the following model [47]:

Let a rigid body be in motion about a fixed point O, while each of its mass elements
is influenced by a force proportional to its mass and its distance from a fixed plane at
O in the direction perpendicular to that plane. In the usual notation, the axes O x yz
are taken as the system of principal axes of the body at the fixed point and O X Y Z as
the inertial frame, with the Z -axis orthogonal to the fixed plane and the unit vector
γ along the Z direction.
The potential of the body

1 1
V = N Z dm = N (r·γ)2 dm
2
2 2
1 1
= N γ Ī·γ = − N γI·γ + const (6.7)
2 2
where N is proportionality constant. This potential is just the quadratic part of the
potential (6.6), with λ replaced by −N .
Thus, Brun’s problem is a special version of the general problem of motion of a
rigid body in an approximate Newtonian field, namely, the case when the body is
fixed from its centre of mass.
@seismicisolation
@seismicisolation
6.3 Equations of Motion and Integrals of Motion
Taking (6.6) into account, one can write the equations of motion of the body in the
approximate field of a Newtonian centre of attraction in the form
ω̇I+ω×ωI = γ × (Mgr0 + λγI),

γ̇+ω×γ= 0. (6.8)
Or in expanded form
A ṗ + (C − B)(qr − λγ2 γ3 ) = Mg(z 0 γ2 − y0 γ3 ),

B q̇ + (A − C) ( pr − λγ1 γ3 ) = Mg(x0 γ3 − z 0 γ1 ),
C ṙ + (B − A)( pq − λγ1 γ2 ) = Mg(y0 γ1 − x0 γ2 ),
This is a closed system of six first-order differential equations. For this system,
we have three obvious general integrals of motion. They are the energy, areas and
geometric integrals
1 1
I1 = ωI·ω+Mgr0 · γ + λγI · γ= h,
2 2
I2 = ωI·γ = f,
I3 = γ·γ= 1. (6.10)
6.4 Integrable Cases
As will be seen later in this book, the problem under consideration is in fact a special
version of the much more general problem considered in Chap. 10, dealing with the
motion of a body in a liquid or a magnetized and electrically charged body about
a fixed point. However, in virtue of the importance of this version in applications,
we present here its integrable cases. For the system (6.9) to be integrable, just as in
the problem of motion in the uniform gravity field, we have to find a fourth integral
independent of those three. It turned out that this can be achieved in two cases.
6.4.1 Brun’s Case [47] (Analog of Euler’s Case)
In this case r0 = 0, i.e. the body is fixed from its centre of mass. To obtain the
complementary integral, we write the Euler–Poisson equation
@seismicisolation
@seismicisolation
6.4 Integrable Cases 103
ω̇I+ω × ωI = λγ × γI
multiplying scalarly by ωI to get
ωI·ω̇I+ωI · (ω×ωI) = λωI · (γ × γI).
That is
d 1
|ωI|2 = −λγ·(ωI×γI)
dt 2
= −λABCγI−1 ·(ω×γ). (6.11)
The last relation can be verified very easily by writing the triple scalar product in
the form of a determinant and using the determinant’s properties. Using Poisson’s
equation in this relation, we write
d 1
|ωI|2 = λ ABCγI−1 · γ̇
dt 2
d 1
= λABCγI−1 · γ
dt 2
and hence finally we get
|ωI|2 − λ ABCγI−1 · γ = c, (6.12)
or in expanded form
γ12 γ2 γ2
I4 = A2 p 2 + B 2 q 2 + C 2 r 2 − λ ABC( + 2 + 3 ) = c. (6.13)
A B C
This integral is an obvious generalization of the fourth integral in Euler’s case (the
square of the modulus of the angular momentum). Nevertheless, complete solution
of the equations of motion (6.9) for Brun’s case with the integrals (6.10) and (6.13) is
much more complicated than in Euler’s case. Kobb [223] expressed the Hamiltonian
of the problem using Euler’s angles as generalized coordinates and the momenta
conjugate to them. By writing Hamilton–Jacobi equation and constructing a com-
plete solution for it, he reduced the solution to certain quadratures, but he has not
tried to solve those quadratures explicitly for the coordinates in terms of time. A
similar approach was adopted by Kharlamova in [204] using different coordinates,
the sphero-conic coordinates on the Poisson sphere (See Chap. 9).
@seismicisolation
@seismicisolation
6.4.2 The Generalization of Lagrange’s Case
Let A = B and x0 = y0 = 0, i.e. the body admits axial dynamical symmetry about
the z-axis passing through the fixed point and the centre of mass of the body lies on
the axis of dynamical symmetry. Under those conditions, it is easy to write the third
equation of (6.9) in the form
C ṙ = 0
and thus the integral is the same as in Lagrange’s case
I4 = r = r 0 . (6.14)
To obtain the solution of the equations of motion, one can proceed exactly as in
Lagrange’s case in Chap. 4 (Sect. 4.2). The relation between γ3 and time t has the
same form as in (4.42)

dγ3
t= √ , (6.15)
F(γ3 )
but with
F(γ3 ) = (1 − γ32 )(E − aγ3 − a1 γ32 ) − (b − cr0 γ3 )2 , (6.16)
i.e. F(γ3 ) is here a polynomial of the fourth degree. Thus, γ3 is expressible in terms
of elliptic functions of time. This procedure was pointed out in [250], where the
Eulerian angles are given expressions in terms of Weierstrass’ elliptic functions.
Arkhangelsky [14] proved that the equations of motion (6.8) do not admit a com-
plementary single-valued integral in any more cases than the above two. Note also
that the case of axially symmetric body can be readily generalized by adding a gyro-
static moment, i.e. a rotor along the axis of symmetry of the body. The integral I4
becomes Cr + k3 , i.e. I4 = r is still a constant of the motion.
6.4.3 The Place of Brun’s Potential
At present we know very little about the answer to an important question: for which
potentials are the Euler–Poisson equations
∂V
ω̇I + ω×ωI = γ× , γ̇ + ω×γ = 0 (6.17)
∂γ
integrable? Even for the simplest case of a quadratic integral we know a single
partial result, obtained in [381] using reduced equations in isometric coordinates
on the inertia ellipsoid. We deduce it here in a direct way from Euler–Poisson’s
equations. We formulate it as
@seismicisolation
@seismicisolation
6.4 Integrable Cases 105
Theorem 6.1 Equation (6.17) is integrable with a quadratic complementary integral

for arbitrary A, B, C, and for all initial conditions only for the potential
1
V = λ(Aγ12 + Bγ22 + Cγ32 ), λ arbitrary constant (6.18)
2
and their integral is
I4 = A2 p 2 + B 2 q 2 + C 2 r 2 − λ(BCγ12 + C Aγ22 + ABγ32 ). (6.19)
Proof Let for a general potential V (γ) and μ = 0 the equations of motion have an
integral quadratic in the angular velocities. This integral must have the form
I4 = A2 p 2 + B 2 q 2 + C 2 r 2 + F(γ1 , γ2 , γ3 ), (6.20)
so that F ≡ 0 when V ≡ 0. Differentiating this integral we get
∂F
2ωI·ω̇I+ · γ̇ = 0,
∂γ
and using (6.17) this becomes
∂V ∂F
2ωI·(γ× − ω×ωI)+ · (−ω×γ) = 0,
∂γ ∂γ
and finally we obtain the relation
∂V ∂F
2ωI·(γ× )−ω · (γ× ) = 0, (6.21)
∂γ ∂γ
which is satisfied for arbitrary vector ω.

Now we replace ω by γ in Eq. (6.21), to obtain an equation for the potential V
∂V
2γI·(γ× ) = 0,
∂γ
whose solution is readily

V = V (γI·γ, γ 2 ).
Since we have γ 2 = 1, the last expression can be written as
V = V (γI·γ). (6.22)
In a similar way, we replace ω by γI−1 in Eq. (6.21), to obtain the equation for F
∂F
γI−1 · (γ× )= 0.
∂γ
@seismicisolation
@seismicisolation
It follows that
F = F(γI−1 ·γ). (6.23)
Now, inserting (6.22) and (6.23) into (6.21), one gets
4V (γI·γ)ωI·(γ × γI) − 2F (γI−1 ·γ)ω · (γ × γI−1 ) = 0. (6.24)
Comparing the structures of the two terms in the last equation and noting that γI · γ
and γI−1 · γ are independent functions for a tri-axial body (A = B = C), The two
derivatives V and F must be constants. Neglecting an insignificant constant in each
of V and F, we write
1
V = λ(Aγ12 + Bγ22 + Cγ32 ), F = N1 (γ12 /A + γ22 /B + γ32 /C). (6.25)
2
Again, inserting the last expressions into (6.24), we obtain
λγ · (ωI × γI) + N1 γI−1 · (ω × γ) = 0.
Applying the identity (A.2) in Appendix A, we rewrite the last relation as
(λABC + N1 )γI−1 · (ω × γ) = 0
and thus we obtain

N1 = −λABC.
This completes the determination of the integral as in (6.13).
6.5 Exercises
1- In the classical problem of motion of a rigid body about a fixed point in the
constant uniform gravity field two equilibrium positions are possible. In the problem
of motion in approximate Newtonian field, show that in any equilibrium position one
of the generators of Ampère’s cone (Staude’s cone) must be vertical (passes through
the fixed point and the centre of attraction).
2- Noting that λ is positive in (6.8) for a centre of attraction, show that the vertical
generator of Staude’s cone in an equilibrium position must be one of the axes inad-
missible for Staude’s rotation in the uniform gravity field. [Compare the equations
of equilibrium to equations of Staude’s rotation and note that ω 2 in Staude’s rotation
is replaced by −λ]
3- For a body fixed from its centre of mass in a uniform gravity field, any position
is a possible equilibrium position. Show that for a body fixed from its centre of
mass in the approximate field of a Newtonian centre there are only six equilibrium
positions and find them.
@seismicisolation
@seismicisolation
6.5 Exercises 107
4- For a body fixed from its centre of mass, show that a uniform rotation is possible
only in two cases:
(a) The rotation about a principal axis with
√ arbitrary angular velocity.
(b) The rotation with angular velocity ± λ about an arbitrary axis.
5- Show that all the axes of uniform rotation of the body in approximate field of a
Newtonian centre are generators in Ampère’s cone and that possible axes of Staude’s
rotation constitute a subset of possible axes of uniform rotation in the Newtonian
field.
6- Using the terminology of the present chapter, F is the exact resultant force
exerted on the body by the centre of attraction. Show that
(Rγ+r)
(a) F = −μ |Rγ+r| 3 dm,
(b) The component of this force in the direction of γ is F·γ = − ∂V , V being the
dm ∂R
exact potential (V = −μ |Rγ+r| ).
(c) The resultant force F can be written in terms of the potential in the form
∂V 1 ∂V
F=− γ + γ × (γ × ).
∂R R ∂γ
Note that F · L = 0, which agrees with the fact that the resultant attraction must be
a single force passing through the centre of attraction whatever be the position of the
body.
(d) The magnitude of the resultant force F is

∂V 2 1 ∂V 2
|F| = ( ) + 2 |γ × | .
∂R R ∂γ
(e) The reaction of the fixed point on the body at any equilibrium position must
be vertical.
7- A rigid body is fixed from a point and its centre of mass lies on a principal plane
for that point, say, the x y-plane. The body is acted upon by the gravitational force due
to a distant Newtonian attraction. Show that the body can perform a plane pendulum-
like motion about a horizontal axis, described by the conditions p = q = γ3 = 0,
and determine the variables r, γ1 , γ2 as functions of time.
8- Brun’s model problem is modified, so that each mass element of the body is
influenced by a force proportional to its mass and its distance from an arbitrary fixed
plane and in the direction perpendicular to that plane. Show that the potential is
equivalent to the general potential (6.6).
9- Investigate the stability of the uniform rotation of a body fixed from its centre
of mass O in the approximate Newtonian field of a far centre at O , determined in
Exercise 4-a. Show that the rotation with angular speed r0 about the z-axis, which is
directed to pass through the centre of attraction and the fixed point of the body, is
(a) unstable when the z-axis is the middle principal axis of inertia of the body at
O,
@seismicisolation
@seismicisolation
(b) stable for all values of r0 , when C is the least principal moment of inertia, and
(c) when C is the largest principal moment of inertia, the rotation is unstable if
√ √
3g a + b
r0 < √ ,
R 1 + ab
and stable if √ √
3g a + b
r0 > √ ,
R 1 + ab
−−→
where a = C−B
A
, b = C−A
B
, g is the acceleration of gravity at O and R = | O O|.
[Beletsky [20]. See also [256], Sect. 18.4]
@seismicisolation
@seismicisolation
Chapter 7
The Motion of a Body with No Fixed
Point
7.1 General Considerations
In previous chapters, we studied problems of motion of a rigid body with one point
fixed in space, i.e. in an inertial reference frame. In the present chapter, we study
certain problems of motion when the body is not fixed from any point. For the
moment, we shall not begin with constructing a Lagrangian for the motion. To keep
the applicability of the equations of motion as wide as possible, we assume that the
body is subject to a set of forces, which are not necessarily time independent and
which may not have a potential.
As in Chap. 3, we write the equation of motion of an element dm of the body mass
d 2r
dm = dF + dF (7.1)
dt 2
summing over the mass of the body one gets
d 2 r0
M = F. (7.2)
dt 2
The centre of mass of the body moves as a particle acted upon by a force equal to
the resultant of all external forces acting on the body.
On the other hand, let ρ be the position vector of the mass element dm with respect
to the centre of mass, so that
r = r0 + ρ. (7.3)
Multiplying (7.1) vectorially by ρ and integrating over the body, we obtain the
equation for the rotational motion in the form
dGc
= Lc , (7.4)
dt

https://doi.org/10.1007/978-3-030-96336-1_7
@seismicisolation
@seismicisolation
110 7 The Motion of a Body with No Fixed Point

where Gc = ρ× dρ dt
dm and Lc = ρ × dF are, respectively, the angular momen-
tum and the resultant moment of external forces about the centre of mass. The law
of rotational motion about the centre of mass of the body resembles that of motion
relative to the inertial frame.
Example 1 A rigid body is free to move in vacuo under the action of a uniform
gravity field. Describe the motion.
The centre of mass of the body moves as a projectile, while the body performs a
torque-free motion about the centre of mass as in Euler’s case.
In particular, if the body begins with zero angular velocity, it will continue trans-
lational motion of its centre of mass without change in its orientation. If the body
begins with a rotation about one of its principal axes of inertia, it will continue rotation
with the same angular velocity about the same axis, which keeps fixed orientation in
space.
Kinetic energy of the rigid body: To construct the Lagrangian of the motion,
one needs to have an expression for the kinetic energy. That is

1 dρ 2
T = (vc + ) dm
2 dt

1 1 dρ
= Mvc2 + ( )2 dm. (7.5)
2 2 dt
The kinetic energy of the body in a general translational and rotational motion is the
sum of two terms: the kinetic energy of a mass M, equal to the mass of the body and
moving with its centre of mass, and the kinetic energy of the rotation of the body
about its centre of mass. The last formula gives a simple expression of the kinetic
energy, very useful in application to many problems of rigid body dynamics. In the
next section, we study an example of such application.
Example 2 A heavy magnetized rigid body is free to move in vacuo under the action
of two skew uniform gravity and magnetic fields. Describe the motion.
Let (X, Y, Z ) be the coordinates of the centre of mass P of the body relative to
an inertial frame O X Y Z with axis Z directed vertically upwards and let H = H e
be the magnetic field of fixed magnitude and fixed direction in space determined by
the unit vector e. The Lagrangian of the problem can be written as
1
L= M( Ẋ 2 + Ẏ 2 + Ż 2 ) − Mg Z
2
1
+ (Ac p 2 + Bc q 2 + Cc r 2 ) − m · H,
2
where M, m are the mass of the body and its magnetic moment, respectively, and
Ac , Bc , Cc are the central principal moments of inertia. One can see at once that the
translational motion of the centre of mass and the rotational motion are completely
@seismicisolation
@seismicisolation
7.1 General Considerations 111
independent. The centre of mass P moves as a projectile on a parabola with its axis
vertical and vertex upwards. The equations of the rotational motion are obtained
from the Lagrangian
1
L = (Ac p 2 + Bc q 2 + Cc r 2 ) − H m · e. (7.6)
2
This Lagrangian is form-identical with (3.43) of the classical problem of motion
of a heavy body about a fixed point and the problem can be put in Lagrangian,
Hamiltonian or in the Euler–Poisson form. It follows immediately that the present
problem is generally integrable only in the following three cases:
(1) The analog of Euler’s case. The case of no magnetic effect H = 0 (no magnetic
field) or m = 0 (no magnetization in the body)
(2) The analog of Lagrange’s case. The central inertia ellipsoid is a spheroid and the
magnetic moment is parallel to the symmetry axis.
(3) The analog of Kowalevski’s case. Ac = Bc = 2Cc and the magnetic moment lies
in the equatorial plane (m 3 = 0).
The conditional integrable case of Goryachev and Chaplygin has its analog valid
on the level f = 0, f being the component of the angular momentum in the direction
of the magnetic field (ωIc · e = f ). The same applies to all particular solutions of
the classical problem, a complete list of which will be provided in Chap. 8.
7.2 Poisson’s Top. A Top on a Smooth Horizontal Plane
In our previous study of Lagrange’s top (the axi-symmetric top), the assumption was
made that the pin of the top serves as a fixed point in the inertial space. A related
model was first considered by Poisson [309], but much rarely mentioned in textbooks
on the subject. For a somewhat detailed treatment see [221, 222].
In this model, a symmetrical top moves with its apex Q constrained to move
without friction on a horizontal plane, so that the reaction R of the plane on the
body remains in the vertical direction. Denote by P the centre of mass of the top.
From symmetry, P lies at a point on its axis of the top at a distance a (say) from Q.
Let (x, y, z) be the coordinates of P relative to an inertial frame O x yz with axis z
directed vertically upwards and the x y-plane is the plane of motion of the apex. In
this system, the apex Q has coordinates (X, Y, 0). The rotational motion of the top
will be described by the Eulerian angles: ψ between the vertical plane containing
Q P and the x z-plane, θ between Q P and the vertical upwards and finally ϕ the
angle of proper rotation of the top about its axis (See Fig. 7.1).
To derive equations of motion, one may use (7.5) to write down the Lagrangian
of the motion as
@seismicisolation
@seismicisolation
Fig. 7.1 Poisson’s top
1
L= M(ẋ 2 + ẏ 2 + ż 2 )
2
1
+ [A(θ̇ 2 + sin2 θ ψ̇ 2 ) + C(ψ̇ cos θ + ϕ̇)2 ]
2
−Mgz
where M is the mass of the top, C is its axial moment of inertia, A is the moment
of inertia about any axis passing through P and orthogonal to Q P and g is the
acceleration of gravity. Recalling that z = a cos θ, we can rewrite the Lagrangian in
the form
1
L= M(ẋ 2 + ẏ 2 + a 2 sin2 θ θ̇ 2 )
2
1
+ [A(θ̇ 2 + sin2 θ ψ̇ 2 ) + C(ψ̇ cos θ + ϕ̇)2 ]
2
−Mgz. (7.7)
The mechanical system under consideration has five degrees of freedom. It also
admits five integrals of motion: the energy integral and four integrals corresponding
to the four cyclic coordinates x, y, ψ and ϕ, so that there is no need to write any of
the Lagrange’s equations of motion, but only to write the integrals
ẋ = U,
ẏ = V,
ψ̇ cos θ + ϕ̇ = r0 ,
A sin2 θ ψ̇ + Cr0 cos θ = n,
1 1
M(U 2 + V 2 + a 2 sin2 θ θ̇ 2 ) + [A(θ̇ 2 + sin2 θ ψ̇ 2 ) + Cr02 ] + Mgz = h, (7.8)
2 2
@seismicisolation
@seismicisolation
7.2 Poisson’s Top. A Top on a Smooth Horizontal Plane 113
where U, V, r0 , n, h are arbitrary constants of motion.

From the first two integrals, it turns out that the centre of mass of the top moves
with uniform horizontal velocity. This agrees with the fact that the body moves under
the action of only two vertical forces: its weight and the reaction of the plane. From
the last three integrals, one gets
n − Cr0 cos θ
ψ̇ = ,
A sin2 θ
cos θ (n − Cr0 cos θ )
ϕ̇ = r0 − (7.9)
A sin2 θ
and the final equation for the angle θ
(n − Cr0 cos θ )2
(A + Ma 2 sin2 θ )θ̇ 2 = 2(E − Mga cos θ ) − , (7.10)
A sin2 θ
where E = h − 21 M(U 2 + V 2 ) − 21 Cr02 .

From the analytical point of view, setting cos θ = u, we transform (7.10) to the
form
1
[A + Ma 2 (1 − u 2 )]u̇ 2 = 2(E − Mgau)(1 − u 2 ) − (n − Cr0 u)2 , (7.11)
A
which leads through separation of variables to the relation

u
A + Ma 2 (1 − u 2 )
t= du
2(E − Mgau)(1 − u 2 ) − A1 (n − Cr0 u)2
u
a du
= √ , (7.12)
2g F(u)
where
(u − u 1 )(u 2 − u)(u 3 − u)
F(u) = ,
u 24 − u 2

and 0 ≤ u 1 ≤ u 2 ≤ 1 ≤ u 3 , u 4 = 1 + Ma A
2 > 1.
Now, we note that Eq. (7.9) is identical with the ones considered for Lagrange’s top
in Chap. 3. Also, (7.11) is similar to its corresponding Eq. (4.41) in Lagrange’s case,
and differs from it only by presence of the term Ma 2 (1 − u 2 ) in the coefficient of u̇ 2
on its left-hand side. However, this term does not change the sign of the coefficient.
Thus, the general qualitative character of the motion of the Poisson top is almost the
same as in Lagrange’s top and it will not be repeated here.
In general, the last integral is hyper-elliptic, compared to the elliptic integral (4.42)
for Lagrange’s top. Inverting this relation we express u = cos θ in terms of time, and
then integrating (7.9) with respect to time we obtain ψ and ϕ. In Klein’s work [221],
@seismicisolation
@seismicisolation
where the top under consideration is termed the “toy top”, explicit expressions of
the Cayley–Klein parameters describing the motion as hyper-elliptic integrals in u
were given, so that together with (7.12) this gives a parametric representation of the
solution. Those results were detailed and refined in [339], where also degenerate
cases when hyper-elliptic integral reduce to elliptic were singled out.
To find the equation of the trajectory of the tip of the top on the plane, we write
X = x − a sin θ cos ψ = x0 + U t − a sin θ cos ψ,

Y = y − a sin θ sin ψ = y0 + V t − a sin θ sin ψ. (7.13)
The reaction of the plane on the apex can be found from the equation of motion of
the centre of mass in the z direction
M z̈ = R − Mg,
so that
R = M(g + z̈)
= M(g + a ü). (7.14)
Remark: Without any effect on the rotational motion of the top, it can be assumed
that U = V = x0 = y0 = 0, so that x = y = 0. This fixes the choice of the inertial
frame as the one whose Z -axis passes through the initial position of the centre of
mass P of the top. In this frame, P moves only vertically up and down the Z -axis.
7.2.1 Regular Precession of Poisson’s Top
As it was in the case of Lagrange’s top, regular precession corresponds to the nutation
angle θ taking a constant value θ ∗ (say), and then from (7.9) we find that the other
two Eulerian angles ϕ, ψ change with time in constant rates. This occurs in two
qualitatively different ways:
1- When u 1 = u 2 = u ∗ , 0 < u ∗ ≤ 1. This can happen at arbitrary inclination of
the body axis to the vertical, including the standing gyroscope positions (u = 1).
Regular precessions of this type correspond to Fig. 7.2a for the function F. They are
all stable, since a slight perturbation of the motion causes splitting of the two roots
in a small neighbourhood of u ∗ . This leads to a small periodic change in the nutation
angle θ and consequently small wobbling in the rates ψ̇ and ϕ̇.
2- When u 1 < 1, u 3 = u 2 = 1. This gives a different standing position, corre-
sponding to Fig. 7.2b. On perturbation, the equal roots split into u 3 > 1, u 2 < 1. The
figure axis begins a finite periodic motion, in which it goes to position u 1 before it
returns to u 2 . This standing position is unstable.
@seismicisolation
@seismicisolation
7.3 Exercises 115
Fig. 7.2 Stability of regular precession for Poisson’s top
7.3 Exercises
1. Show that the resulting problem of example 2 in Sect. 7.1 above can be generalized
to the case of a body-gyrostat.
2. The Poisson top is given the upper equilibrium position. Show that this position,
which is unstable when r0 = 0, can be stabilized by giving the top an angular speed
r0 about its axis of symmetry provided
√
Mga A
|r0 | > 2 .
C
3. A rigid body moves freely in a gravitational field with homogeneous quadratic
potential
1
Vg = (aξ 2 + bη2 + cζ 2 )
2
in the inertial frame. Show that the equations of translational motion of the centre
of mass and the rotational motion about the centre of mass are completely separate.
For more details see Chap. 14 Sect. 14.3.1.
@seismicisolation
@seismicisolation
Part II
The Advanced Part
@seismicisolation
@seismicisolation
Chapter 8
Particular Solutions of the Classical
Problem and Its Generalizations
8.1 The Notion of a Particular Solution
By particular solutions, we mean cases that are solvable under conditions on the
initial state of motion, which cannot be stated as a condition only on the value f
of the areas integral. After the great success of Sofia Kowalevski (Kovalevskaya)
in isolating a new general integrable case, several Russian authors, including such
celebrated names as Joukovsky, Lyapunov, Steklov and Chaplygin, concentrated on
research on various aspects of rigid body dynamics. Their contributions have built
up a large part of the information available today about various problems in the
dynamics of rigid body.
In this chapter, we present known particular solutions of the Euler–Poisson equa-
tions describing motion in the classical problem. These are twelve solutions, valid
under various conditions on the initial motion and on the distribution of mass in
the body. This number does not include other particular solutions, which are special
cases of general and conditional cases. For example, four particular solutions of the
classical problem considered by Appelrot, who coined for them the name “Classes
of the simplest motions”, fit under the integrable case of Kowalevski [11].
In our presentation, we adhere to chronological order and also to referring each
case to the author who was first to publish it, in case of repetition. After providing
conditions and detailed solution in each particular case, we draw the orbit (or family
of orbits) of motion on the Poisson sphere fixed in the body. That is the trajectory
traced by the apex of the vertical unit vector γ(t) on that sphere as the body moves.
This simulates the motion of the body relative to the vertical and in fact determines
the motion of the body in the inertial space up to the angle of precession.
Section 8.14 of the present chapter is devoted to a brief discussion of certain
particular solutions, which were announced and turned out later to be either in error
or repeating known solutions. At the end of this chapter, we have also inserted a brief
survey of all the known up-to-date particular solutions of the problem of motion of
a heavy gyrostat. In most cases, we aimed at giving minimal information that would
help the interested reader to track any of the cases in original works.

https://doi.org/10.1007/978-3-030-96336-1_8
@seismicisolation
@seismicisolation
120 8 Particular Solutions of the Classical Problem and Its Generalizations
In constructing particular solutions of the classical problem, we deal with six

first-order differential equations with three first integrals of motion, namely energy,
areas and geometric integrals. A question arises, to what extent one can rely on using
integrals of motion and ignore some of the six differential equations? This question
is first answered in the following theorem:
Theorem 8.1 Let ω(t) and γ(t) be two vectors satisfying Euler’s dynamical equa-
tion
Ġ + ω × G = Mgγ × r0 , (8.1)
and the integrals of motion
1
I1 ≡ G · ω + Mgr0 · γ = h,
2
I2 ≡ G · γ = f,
I3 ≡ γ 2 = 1, (8.2)
where G = ωI. Then, these vectors satisfy Poisson’s equation
γ̇ + ω × γ = 0, (8.3)
provided the equation

G · (r0 × γ) = 0 (8.4)
is not satisfied identically for all t.

Proof Differentiating I1 with respect to time and using (8.1), we obtain
r0 · (γ̇ + ω × γ) = 0. (8.5)
In a similar way, differentiating I2 , we get
G · (γ̇ + ω × γ) = 0, (8.6)
while differentiating I3 yields

γ · γ̇ = 0. (8.7)
The general solution of (8.7) is
γ̇ = − × γ, (8.8)
where is a vector to be determined. Inserting this into (8.5) gives
( − ω) · (r0 × γ) = 0,
so that one can write, since a component of parallel to γ is insignificant,
@seismicisolation
@seismicisolation
8.1 The Notion of a Particular Solution 121
= ω + λr0 ,
where λ is a still undetermined multiplier.
Now, only Eq. (8.6) remains to be satisfied. Inserting the last expression into (8.8)
and then γ̇ into (8.6), we obtain
G · [−(ω + λr0 ) × γ + ω × γ] = 0,
and this finally becomes

λ[G · (r0 × γ)] = 0.
If the triple scalar product is not identically zero during motion, then λ = 0, = ω
and (8.8) becomes identical with Poisson’s equation.
Thus, in constructing particular solutions of the Euler–Poisson system, it is suf-
ficient in most cases to take into account Euler’s dynamical equation, along with
the three first integrals of motion. The resulting solution will also satisfy Poisson’s
equation, unless it comes out that the three vectors G, r0 ,γ remain coplanar all the
time.
We now note the following: Multiplying the dynamical Eq. (8.1) scalarly by G,
we get
G · Ġ = MgG · (γ × r0 ),
so that if condition (8.4) is satisfied, then
G2 = G 20 (const .).
The converse is also true. Thus, the last theorem can be reformulated as follows:
Theorem 8.2 If ω(t) and γ(t) satisfy Euler’s dynamical equation and the three
relations I1 = h, I2 = f and I3 = 1 for some pair {h, f }, and if the magnitude of
the angular momentum of the body is not identically a constant, then ω(t) and γ(t)
satisfy also Poisson’s equation and constitute a solution of the classical problem.
This result will be used frequently in the present chapter. In most cases, it is
comfortable to use the ansatz in the three integrals and to satisfy Euler’s dynamical
equations, with no need to verify Poisson’s equations.
It should be noted that a solution satisfying (8.1)–(8.2) can be erroneous as a
solution of the Euler–Poisson equations. An example is the recently published result
of Ershkov [76] claiming a new particular solution in which the centre of mass lies
either on a principal axis of inertia or in a principal plane. To this end, the author uses
only the dynamical equation and the three integrals. As in these solutions G = C0 γ,
i.e. G2 = const., three Poisson’s equations should have been checked, which was
not done by the author of [76]. For more details, see Sect. 8.14 of the present chapter.
@seismicisolation
@seismicisolation
8.2 Planar Motion (Motion of the Body as a Physical

Pendulum)
This is the motion of the body about the fixed point as a physical pendulum, i.e.
motion about a horizontal axis which should be fixed in the body and space. This
motion is also called planar motion, as all body elements move parallel to a fixed
vertical plane. Let us take the axis of rotation to be the z-axis. The conditions for
such a motion are
p = q = γ3 = 0. (8.9)
Substituting into the equations of motion (3.29), one obtains
0 = Mgz 0 γ2 ,
0 = −Mgz 0 γ1 ,
C ṙ = Mg(y0 γ1 − x0 γ2 ),
γ̇1 − r γ2 = 0, γ̇2 + r γ1 = 0. (8.10)
This means z 0 = 0, i.e. the centre of mass of the body lies in the principal x y-plane.
Measuring Euler’s angle ϕ from the downward vertical, one sees that conditions (8.9)
reduce to the conditions
π
θ= , ψ̇ = 0, r = −ϕ̇, γ1 = − cos ϕ, γ2 = − sin ϕ, (8.11)
2
so that only one equation remains. Instead of integrating this equation, we use the
energy integral to write
1 2
Cr + Mg(x0 γ1 + y0 γ2 ) = h, (8.12)
2
where h is the total energy of the motion. Denoting by α the angle between the
position vector of the centre of mass and the y-axis and by s the distance from the
fixed point to the centre of mass, so that
x0 = s cos α, y0 = s sin α, (8.13)
and completing substitution from (8.11), we get
1
C ϕ̇2 − Mgs(cos α cos ϕ + sin α sin ϕ) = h, (8.14)
2
or, equivalently,
1
C ϕ̇2 = h + a cos(ϕ − α), (8.15)
2
@seismicisolation
@seismicisolation
8.2 Planar Motion (Motion of the Body as a Physical Pendulum) 123
where a = Mgs. Now, separating variables in the last equation, we find

2 dϕ
± dt = √ .
C h + a cos(ϕ − α)
Introducing a new angle by the relation ϕ = α + 2 and integrating the last

relation yield

1 d
± (t − t0 ) = , (8.16)
2C 0 h + a − 2a sin2
where t0 is the time at which = 0, i.e. when the centre of mass passes at the vertical
position below the fixed point (the stable equilibrium position). The minimum energy
for possible motions is h = −a occurs at this position. Three qualitatively different
types of motions are possible:
8.2.1 Rotational Motion
When h > a we can write (8.16) in the form

h+a d
± (t − t0 ) = , (8.17)
2C 0 1 − k 2 sin2
where
2a
k2 = < 1.
h+a
The angle is readily expressed as

h+a
= am(u, k), u = ± (t − t0 ),
2C
sin = sn u, cos = cn u, (8.18)
and hence
γ1 = − cos ϕ = − cos(α + 2)

= − cos α(1 − 2 sin2 ) + 2 sin α sin cos
= − cos α(1 − 2 sn2 u) + 2 sin α sn u cn u,
γ2 = − sin(α + 2)
= − sin α(1 − 2 sn2 u) − 2 cos α sn u cn u, (8.19)
@seismicisolation
@seismicisolation
and the angular velocity is given by

d du h+a
r = −2 = ±2 dn u. (8.20)
du dt 2C
The body rotates with periodic angular velocity and makes complete revolution in
one and the same direction all the time. The three variables γ1 , γ2 , r are periodic
functions in t with period
2C
T = 2K (k) . (8.21)
h+a
As the energy of rotation h increases to infinity, k → 0, K (k) → π2 , dn u → 1 and

r → 2h C
and the periodic time for fast rotations becomes 2πr
.
8.2.2 Vibrational Motion
When −a ≤ h < a, the energy of motion is not sufficient to let the centre of mass
of the body go all the way to the highest point vertically above the fixed point. The
motion is a vibration such that ϕ − α varies between ±δ, δ = cos−1 −ha
. The modulus
k of the elliptic functions becomes >1 and using elliptic functions with modulus

1 δ h+a
ν = = sin = , (8.22)
k 2 2a
the solution (8.19), (8.20) is replaced by
γ1 = − cos α(1 − 2ν 2 sn2 w) + 2ν sin α sn w dn w,

γ2 = − sin α(1 − 2ν 2 sn2 w) − 2ν cos α sn w dn w,

h+a
r =2 cn w, (8.23)
2C
where
a
w=± (t − t0 ). (8.24)
C
The periodic time of motion is

C
T = 4K (ν) . (8.25)
a

For small vibrations about the lower equilibrium position, T = 2π C
a
.
@seismicisolation
@seismicisolation
8.2 Planar Motion (Motion of the Body as a Physical Pendulum) 125
8.2.3 The Limiting Motion
Now we turn to the critical case between the above two classes of motions. When
h → a, both k and ν → 1 and the period becomes infinitely large. The solutions
(8.19), (8.20) and (8.23) both reduce to
γ1 = − cos α(1 − 2 tanh 2 w) + 2ν sin α tanh w sech w,

γ2 = − sin α(1 − 2 tanh2 w) − 2ν cos α tanh w sech w,

a
r =2 sech w. (8.26)
C
As w → ±∞, (γ1 , γ2 ) → (cos α, sin α) and r → 0. The motion becomes asymp-

totic to or from the unstable (upper) equilibrium position.
8.2.4 Orbital Stability
The trajectory traced by the plane motion on the Poisson sphere is a great circle
γ3 = 0 for the case of rotations and an arc of such circle in the case of vibration. By
the orbital stability of the motion, we mean the stability of the solution γ3 = 0 with
respect to lateral perturbations preserving the values of the total energy h and the
areas integral f = 0. The study of orbital stability of pendulum motions in the linear
setting was initiated in [379], continued in [375, 376, 388] and generalized for the
gyrostat in [388, 420].
8.3 Permanent Rotation of a Heavy Rigid Body About a

Fixed Point [343] (1894)
8.3.1 Possible Axes of Permanent Rotation
Following Staude [343], we now look for possible rotations of the body with a
uniform angular velocity about an axis, which remains immovable both in space and
in the body. Setting ω̇ = 0, γ̇ = 0, in (3.15), we get
ω × ωI = Mgγ × r0 , ω × γ = 0. (8.27)
The second equation has the solution ω = ω0 γ, ω0 is a constant, so that the motion
is possible only about a vertical axis. Substituting ω in the dynamical equation gives
γ × (ω02 γI − Mgr0 ) = 0. (8.28)
@seismicisolation
@seismicisolation
Multiplying this equation scalarly by r0 , we obtain the equation of the locus of the
axes of permanent rotations
r0 · (γ × γI) = 0. (8.29)
In the system of principal axes of the body, this equation can be written in the
determinantal form
x0 y0 z 0

γ1 γ2 γ3 = 0 (8.30)

Aγ1 Bγ2 Cγ3
or, in the expanded form,
x0 (B − C)γ2 γ3 + y0 (C − A)γ3 γ1 + z 0 (A − B)γ1 γ2 = 0. (8.31)
This is a homogeneous quadratic equation. It represents a cone, fixed in the body and
whose vertex is at the fixed point. This cone is called Staude’s cone. It is identical
with Ampére’s cone mentioned in Chap. 1 in a completely different context as the
cone formed by lines passing through the fixed point, each of which is a principal
axis with respect to one of its points. The last property of Staude’s cone seems to be
first proved by Lecornu [255], but is usually ascribed to van der Woude [370].
There are three cases when Staude’s cone becomes undetermined. Those are when
(1) the body is fixed from its mass centre (x0 = y0 = z 0 = 0). That is Euler’s case;
(2) the body has spherical dynamical symmetry at the fixed point ( A = B = C);
(3) the body is dynamically axi-symmetric and the centre of mass lies on the sym-
metry axis. This is Lagrange’s case.
Staude’s cone degenerates into two planes in three cases:
(1) When the body is dynamically axi-symmetric.

(2) When the centre of mass lies in a principal plane or on a principal axis of inertia.
(3) To those cases one has to add the case when a triangle inequality holds (B + C =
A, C + A = B, A + B = C), in which Staude’s cone formally persists to be
non-degenerate, but loses its meaning because the body renders to a plane disc
with centre of mass lying outside the plane. According to the theorem of Sect. 1.4,
we have z 0 = 0.
8.3.2 Description of the Motion
We now try a more detailed description of the permanent rotations. Let a given vector
γ determine a direction in the main body of an axis of permanent rotations, i.e. a
generator of Staude’s cone. This generator is directed vertically upwards and the
body moves about it with a constant angular speed ω0 . During this motion the centre
of mass (and also any other mass element) of the body uniformly traces a circular
@seismicisolation
@seismicisolation
8.3 Permanent Rotation of a Heavy Rigid Body About a Fixed Point [343] (1894) 127
path in a horizontal plane. In the given motion not only the total energy of motion is
conserved, but also both kinetic and potential energies remain constant.
The square of the angular velocity about this axis may be determined from any
of the equations
ω02 (B − C)γ2 γ3 = Mg(y0 γ3 − z 0 γ2 ),

ω02 (C − A)γ3 γ1 = Mg(z 0 γ1 − x0 γ3 ),
ω02 (A − B)γ1 γ2 = Mg(x0 γ2 − y0 γ1 ), (8.32)
obtained by writing (8.28) in components. Note that those equations are compatible
in virtue of (8.29). This means that if a permanent rotation is possible (ω02 positive)
about an axis with one of its ends directed vertically upwards, then this motion
can have one of the two angular velocities ± ω02 , i.e. equal and opposite angular
velocities. From the last equations, we notice that ω02 changes sign when the direction
of γ is reversed, so that the other end of the axis of permanent rotation cannot be
itself directed upwards for another possible permanent rotation. In other words, every
generator of Staude’s cone is a possible axis of permanent rotation with one and only
one end of it directed vertically upwards. Obvious exception is the generator that
passes through the centre of mass. This generator can be directed vertically up or
down, rendering the body to its two equilibrium positions (since ω02 = 0) with the
centre of mass above or below the fixed point.
We now proceed to clarify the picture of Staude’s cone and the distribution of the
axes of permanent rotations on that cone. We assume that the centre of mass lies at a
point in the positive octant. This can be always achieved, excluding for the moment
some degenerate cases, when the centre of mass of the body lies on one of its principal
axes of inertia. From the determinant Eq. (8.30), we notice that Staude’s cone passes
through the three principal axes of inertia of the body and also through the centre of
mass r0 and its diametrically opposite point with respect to the origin (−r0 ). Denote
by P and Q, the points of intersection of the generator passing through r0 and (−r0 ),
respectively, with the Poisson sphere.
Staude’s cone has vertex at the origin, and it intersects the Poisson sphere fixed in
the body in two closed spherical curves, one of which passes through P, the positive
ends of x, z-axes and the negative end of y. The other spherical curve is the reverse
of the first with respect to the origin, i.e. passes through Q, −x, −z, y (See Fig. 8.1).
For further determinacy, we shall assume that A > B > C. From Eqs. (8.32), it
follows that ω0 is real only on the axes intersecting the unit sphere in the four arcs
(P, z), (x, −y), (Q, −x) and (y, −z), shown as thick lines in Fig. 8.2. Each axis
should be directed vertically upwards and the body must be given the appropriate
(positive or negative) angular velocity about that axis. Note that the angular velocity
corresponding to the permanent rotation about O P is ω0 = 0, i.e. this is the upper
equilibrium position of the body. Similarly, the axis O Q corresponds to the lower
equilibrium position. Note also that ω0 tends to infinity whenever the axis of rotation
approaches one of the six ends of the principal axes of inertia.
@seismicisolation
@seismicisolation
Fig. 8.1 Staude’s cone and

its intersection with the
sphere
Fig. 8.2 Admissible arcs for

Staude’s permanent rotations
are represented by thick (red)
lines (Hidden lines are
dashed)
Fig. 8.3 The centre of mass

lies in a principal plane
(x y-plane)
Figure 8.3 depicts the admissible arcs on the sphere, corresponding to a centre of
mass lying in the principal plane x y. For more detailed presentation of degenerate
cases, see [321] (and also [256]).
@seismicisolation
@seismicisolation
8.3 Permanent Rotation of a Heavy Rigid Body About a Fixed Point [343] (1894) 129
8.3.3 Further Studies
Permanent rotations have attracted researchers for a long time and from different
points of view. Linear stability analysis was performed for some special cases in
[131], (See also [132]), [43] and others. More systematic results were obtained by
Rumiantsev [321], who relied on wide use of linearized equations in the variations
and construction of Lyapunov functions.
Staude’s rotation is characterized by the constancy of the vectors γ and ω in
the body system (as well as in space). In such motion, only the Eulerian angle of
precession ψ varies with time, while θ and ϕ take constant values. Noting that ψ is
a cyclic coordinate, it becomes evident that Staude’s rotations can be characterized
as the equilibrium positions of the reduced system. This means permanent rotations
are located at the extremal points of the reduced potential V1 of the problem on the
Poisson sphere (See (3.56)):
f2
V1 = Mgr0 · γ + . (8.33)
2(Aγ12 + Bγ22 + Cγ32 )
Let P be one of those points. The permanent rotation about O P is stable if V1 attains
a minimum at P, unstable if it has a saddle point. If V1 has a maximum at P, then
one cannot draw any conclusion about the stability of the permanent rotation based
on V1 alone. In that case gyroscopic stabilization is possible and the equations of
perturbed motion should be considered in the nonlinear setting. For more details
and results in this direction, see [105, 351] and references therein. Detailed stability
analysis of permanent rotations with special emphasis on those corresponding to the
three integrable cases can be found in [330].
8.3.4 Excercises
(1) Show that the reaction of the fixed point on the body in the state of permanent
rotation is given by the formula
R = Mgγ − Mω02 ρ,
where ω0 is determined from (8.32) and ρ is the current radius vector of the centre
of mass of the body relative to the axis of rotation.
(2) Determine all possible uniform rotations of a rigid body about a fixed point
O of it, when acted upon by a single constant force applied to a point distinct from
both the fixed point and the centre of mass [Alfieri [6]].
@seismicisolation
@seismicisolation
8.4 Hess’ Case (1890)
One of the earliest particular solutions in the dynamics of a rigid body was discovered
by Hess in 1890 [150] and actually rediscovered by some other authors, e.g. Appelrot
[10] and Shiff [331]. It differs from all other particular cases in an essential aspect,
its solution is not completely reduced to quadratures. One may say it is a case of
partial solution of the Euler–Poisson equations.
8.4.1 Equations of Motion
Let the principal axes of inertia of the body be arranged so that A > B > C, and
assume that the centre of mass C g lies in the x z-plane containing the major and minor
axes of the inertia ellipsoid y0 = 0. Equations (3.29) take the form
A ṗ + (C − B)qr = Mgz 0 γ2 ,
B q̇ + (A − C) pr = Mg(x0 γ3 − z 0 γ1 ),
C ṙ + (B − A) pq = −Mgx0 γ2 ,
γ̇1 + qγ3 − r γ2 = 0, γ̇2 + r γ1 − pγ3 = 0, γ̇3 + pγ2 − qγ1 = 0.
(8.34)
Eliminating γ2 between the first and third equations, we get
Ax0 ṗ + C z 0 ṙ = q[(B − C)x0 r + (A − B)z 0 p]. (8.35)
Thus, if the distribution of mass in the body satisfies the condition
(A − B)z 0 (B − C)x0
= = s (say),
Ax0 C z0
i.e. if
x0 A(B − C) = ±z 0 C(A − B), (8.36)
then Eq. (8.35) can be written as
d
(Ax0 p + C z 0 r ) = sq(Ax0 p + C z 0 r ). (8.37)
dt
If the motion at a certain time moment satisfies the condition that
Ax0 p + C z 0 r = 0, (8.38)
@seismicisolation
@seismicisolation
8.4 Hess’ Case (1890) 131
the time derivative of this expression will also vanish, and hence it will be a constant
of motion that vanishes at all times. This gives what we call an invariant relation.
Note that the invariant relation (8.38) has a dynamical meaning. It signifies that the
component of the angular momentum of the body vanishes in the direction from the
fixed point to the centre of mass of the body.
Remark: Condition (8.36) together with y0 = 0 mean that the centre of mass
C g lies on the line drawn from the origin perpendicular to one of the two circular
cross-sections of the gyration ellipsoid (See Exercise 4 of Chap. 1).
Taking this remark into account, one can see that the angular momentum of the
body in Hess’ case lies always in one of the circular cross-sections of the gyration
ellipsoid.
8.4.2 Solution
Let δ be the angle between OC g and the z-axis. Regarding (8.36), we have

A(B − C) C(A − B)
cos δ = , sin δ = ± . (8.39)
B(A − C) B(A − C)
Denote by 1 and 3 the projections of the vector γ on OC g and the orthogonal to

it. Then one can write
3 = cos δγ1 + sin δγ3 ,

1 = − sin δγ1 + cos δγ3 . (8.40)
Those variables satisfy the relation
12 + 32 = γ12 + γ32 = 1 − γ22 . (8.41)
We now show that the variable 3 can be determined by a quadrature. In fact,

using the integrals of motion
Ap 2 + Bq 2 + Cr 2 + 2Mgs3 = 2h,
Apγ1 + Bqγ2 + Cr γ3 = f,
together with (8.38), one can find

√
A(1 − 32 ) p = − sin δ( Sγ2 + f 1 ),
√
B(1 − 32 )q = − S1 + f γ2 ,
√
C(1 − 32 )r = cos δ( Sγ2 + f 1 ), (8.42)
@seismicisolation
@seismicisolation
where S = S(3 ) is the cubic polynomial
S = 2B(1 − 32 )(h − s3 ) − f 2 . (8.43)
From Poisson’s equations and using the last expressions (8.42), we readily obtain
for 3 the equation
B ˙ 3 = S(3 ). (8.44)
Thus, 3 can be expressed as an elliptic function of time by inverting the relation

3
du
t= . (8.45)
2B(1 − u 2 )(h − su) − f 2
The last formula can be compared to formulas (4.42) of Sect. 4.2, concerning
Lagrange’s top. Equation (8.45) has the same structure as in Sect. 4.2 provided in
the last we put c = C/A = 0, i.e. the Lagrange top has all its mass concentrated on
the 3 -axis. This means that the centre of mass of the body moves with respect to
the vertical as a spherical pendulum. This property was noted by Joukovsky [165].
8.4.3 The Use of Special Axes
Hess’ case takes a simpler form if we choose the body axes to be associated with the
gyration ellipsoid. The equations of motion take the form (3.41). Let the x, y-axes
be in the plane of one of the circular cross-sections of that ellipsoid and the z-axis
orthogonal to it. As the middle principal axis of inertia of the body is the intersection
of the two circular cross-sections, this axis can always be chosen as the y-axis, so
that A23 = 0. The gyration matrix becomes
⎛ ⎞
A11 0 A13
A = ⎝ 0 A11 0 ⎠ . (8.46)
A13 0 A33
In those axes, the centre of mass lies on the z-axis at the point (0, 0, z 0 ) (say), and
Hess’ integral becomes R = 0. The equations of motion (3.41) become
Ṗ − A13 P Q = cγ2 ,
Q̇ + A13 P 2 = −cγ1 , (8.47)
@seismicisolation
@seismicisolation
8.4 Hess’ Case (1890) 133
γ̇1 − Pγ2 A13 + A11 Qγ3 = 0,

γ̇2 + Pγ1 A13 − A11 Pγ3 = 0,
γ̇3 + A11 (Pγ2 − Qγ1 ) = 0, (8.48)
where c = Mgz 0 , and the integrals of motion
1
A11 (P 2 + Q 2 ) + cγ3 = h,
2
Pγ1 + Qγ2 = f,
γ12 + γ22 + γ32 = 1. (8.49)
It will be more convenient to introduce new variables by the relations
P = ρ cos χ, Q = ρ sin χ. (8.50)
Those variables are related to the Eulerian angles θ, ϕ. Recalling the expression of
γ, one can use the integrals (8.49) to obtain the relations
1 1
γ3 = cos θ = (h − A11 ρ2 ),
c 2
ρ sin θ sin(χ + ϕ) = f. (8.51)
In the same special axes, the angular velocity of the body can be expressed, using
(8.46), as
ω ≡ ( p, q, r ) = (A11 P, A11 Q, A13 P)

= (A11 ρ cos χ, A11 ρ sin χ, A13 ρ cos χ). (8.52)
Now, comparing this expression with the formula (2.39) and using the areas integral
in (8.49), one can find the rates of change of the Eulerian angles
A11 f
ψ̇ = ,
sin2 θ
θ̇ = A11 ρ cos(χ + ϕ),
A11 f cos θ
ϕ̇ = A13 ρ cos χ − . (8.53)
sin2 θ
Using (8.50) and (8.47), we find for ρ and χ the equations
dρ √
ρ = −c F, (8.54)
dt
where
1
F = A11 ρ2 [1 − (h − A11 ρ2 )2 /c2 ] − f 2 ,
2
@seismicisolation
@seismicisolation
and
cf
χ̇ = −[A13 ρ cos χ + ]. (8.55)
ρ2
The first equation determines ρ2 as an elliptic function of time t (the equivalent of

(8.45)). Inserting this function into (8.55) we obtain for χ a nonlinear first-order dif-
ferential equation with elliptic coefficients. Substituting tan χ2 = η, the last equation
can be put in the form of Riccati equation and if further one uses the substitution
η = ζ̇/ζ, it can be reduced to a linear second-order equation with elliptic coefficients
[293]. An equivalent form was obtained by Golubev using a somewhat different
method in [113]. A detailed analysis of Hess’ case was performed by Kovalev in
[236, 237], aiming mainly at constructing the angular velocity hodograph for Hess’
case. A linear equation for ζ(ρ), equivalent to Nekrassov’s equation, was derived
[237] (See also [108]). However, none of those forms helped to complete the solu-
tion in the general case.
8.4.4 Solution of the Case f = 0
The simplest case and the only one when the solution of the equations of motion is
completed is that when the areas constant f vanishes and then Eq. (8.55) becomes
separable. In that case, the second equation in (8.51) determines the angle χ in the
form
χ = −ϕ, (8.56)
and, hence, Eqs. (8.53) become
ψ̇ = 0,
θ̇ = A11 ρ,
ϕ̇ = A13 ρ cos ϕ. (8.57)
The first equation means that the Eulerian angle ψ during motion preserves a constant
value. Without loss of generality, this value can be taken as ψ = 0. The motion of
the body can be interpreted as follows:
The body rotates with angular velocity ϕ̇ about its barycentric axis (the z-axis,
carrying the centre of mass), while that axis performs a periodic pendulum-like
motion in a fixed vertical plane passing through the fixed point. That is a type of
precession with a variable angular speed θ̇ about a horizontal axis, namely the type
pointed out by Bressan [46] (1957) for the rigid body in the case of Hess.
Explicit expressions of the variables in terms of time differ for the three cases
of pendulum motions: rotational, vibrational and asymptotic. We write down here
the full expressions for the case of complete pendulum rotations, characterized by
@seismicisolation
@seismicisolation
8.4 Hess’ Case (1890) 135
the condition h > c. Detailed qualitative analysis of possible types of motion can be
found in [108].
In that case from the above formulas we can find
cos θ = −1 + 2 sn2 u, (8.58)

2(h + c)
ρ= dn u, (8.59)
A11
where
h+c
u = A11 t,
2
and the modulus of Jacobi’s elliptic functions

2c
k= .
h+c
On the other hand, dividing the last two equations in (8.57) and separating vari-
ables, we get
dϕ A13
= dθ.
cos ϕ A11
And on integration, this gives
A13
ϕ = tan−1 (sinh( θ)), (8.60)
A11
where an insignificant integration constant is taken zero. The trajectory (8.60) on the
Poisson sphere is the locus of the end of the vertical unit vector γ
A13 A13
γ = (sin θ tanh( θ), sin θ sech( θ), cos θ)
A11 A11
on that sphere. As θ increases indefinitely with the pendulum rotation of the z-axis
in the fixed vertical plane, the angle of proper rotation ϕ about that axis tends to the
value π/2, as depicted in Fig. 8.4. The larger the quantity AA13
11
, the faster ϕ approaches
π/2. We conclude that the motion of the body asymptotically approaches the motion
as a physical pendulum about the medium principal axis of inertia, which takes a
permanent horizontal position.
@seismicisolation
@seismicisolation
Fig. 8.4 Orbit of apex of γ

on the Poisson sphere
8.5 The Case of Bobylev and Steklov (1896)
Almost simultaneously, in 1896, two communications of Bobylev [27] and Steklov

[346] appeared announcing one and the same solvable case of the equations of motion
of a heavy rigid body fixed from one point.
The key condition in this case is that one of the components of the angular velocity
permanently vanishes, say, q = 0, i.e. the angular velocity lies permanently in the
principal x z-plane. A great simplification occurs in the equations of motion (3.29)
if one adds the assumption that the centre of mass of the body lies on the z-axis, i.e.
x0 = y0 = 0. Then, under the compatibility condition C = 2 A, we will have
A ṗ = aγ2 ,
−Apr = −aγ1 ,
C ṙ = 0,
γ̇1 − r γ2 = 0, γ̇2 + r γ1 − pγ3 = 0, γ̇3 + pγ2 = 0, (8.61)
where a = Mgz 0 . From those equations, we immediately get
r = r0 = const,
a
p= γ1 , r0 = 0,
Ar0
γ̇1
γ2 = . (8.62)
r0
Now, using the areas integral, we obtain

1 a 2
γ3 = f − γ1 , (8.63)
2 Ar0 r0
@seismicisolation
@seismicisolation
8.5 The Case of Bobylev and Steklov (1896) 137
so that all variables are expressed in term of γ1 and it remains to determine γ1 using
the geometric integral, which leads to the first-order separable equation

1 a 2 2
γ̇1 = ±r0 1 − γ12 − f − γ . (8.64)
4 A2 r02 r0 1
Note that one can verify that all the Euler–Poisson equations (8.61) are all satisfied,
in virtue of (8.62)–(8.64).
To simplify the classification of the possible motions, we introduce the new dimen-
sionless parameters ρ, j by the relations

a √
r0 = ρ , f = a A j. (8.65)
A
Thus, the vectors ω and γ will be expressed as follows:

a γ1
ω= ( , 0, ρ),
A ρ

A γ̇1 j γ2
γ = (γ1 , , − 12 ). (8.66)
a ρ 2ρ 2ρ
Note that a simultaneous change of the signs of parameters r0 , f (ρ, j) results in a

change of signs of p, γ̇1 and leaves unchanged the vector γ as function of time. On
the other hand, formulas (8.66) are not valid for ρ = 0. Thus, it suffices to consider
motions with parameters ρ, j in the open right half-plane ρ > 0. Motions in the other
half-plane are the same but traversed in the reverse direction.
Now, integrating Eq. (8.64) gives

a (t − t0 ) dγ1
± = √ , (8.67)
A 2ρ F(γ1 )

where F(γ1 ) = 4ρ4 − j 2 ρ2 + 2 jρ − 4ρ4 γ12 − γ14 and t0 is an arbitrary con-
stant. The integral in the right-hand side is an elliptic integral of the first kind. Invert-
ing this integral, we express γ1 as an elliptic function in the time t. Depending on the
two parameters ρ and j, the fourth-degree polynomial F(γ1 ) may have either two or
four real roots, so that in the ρ j-plane we have two different regions corresponding
to two qualitatively distinct classes of motions.
A clear geometrical picture of the two classes is obtained by examining the tra-
jectories drawn during the motion of the body by the tip of the vector γ on the
Poisson sphere. That is the curve of intersection of the sphere γ12 + γ22 + γ32 = 1 and
γ2
the parabolic cylinder γ3 = 2ρj − 2ρ12 . It can be readily seen that when −1 ≤ 2ρj < 1
the intersection is a single simple closed curve symmetric with respect to the yz-
plane, beginning at j = −2ρ with the single point (0, 0, −1). Example of this type
@seismicisolation
@seismicisolation
Fig. 8.5 Path of the vector γ on the unit sphere
is illustrated in Fig. 8.5a. For j > 2ρ if j ≤ ρ3 + ρ1 , the intersection consists of two

similar closed curves symmetrically situated with respect to the yz-plane (Fig. 8.5b).
On the boundary j = ρ3 + ρ1 , the two curves shrink to two points. We now describe
the analytical solution of the equations of motion in the two regions.
8.5.1 Region I: The First Class of Motions
In the region I, −2ρ ≤ j ≤ 2ρ, (see Fig. 5.1) the polynomial F(γ1 ) has only two
real roots. Equation (8.67) will have a cn solution and the Euler–Poisson variables
take the form
M√ j M2
γ = (M cn(u, k) , − d sn(u, k) dn(u, k), − 2 cn2 (u, k)),
ρ 2ρ 2ρ

a M
ω= ( cn(u, k), 0, ρ), (8.68)
A ρ
where

M= ρ[ j − 2ρ3 + 2ρd], d = ρ4 − jρ + 1,
1 j − 2ρ3
k 2 = (1 + ),
2 2ρd

ad
u=± (t − t0 ) . (8.69)
A
@seismicisolation
@seismicisolation
Fig. 8.6 Contour lines of

k, ν

4K (k)
The expressions (8.68) are periodic functions in time with period √
4
A
a
.
4 ρ − jρ+1
Contours of k in the ρ j-plane are illustrated in Fig. 8.6. The motion is composed of
a uniform rotation about the z-axis and a vibration about the x-axis.
On the two rays j = −2ρ and j = 2ρ > 2, the modulus k = 0 (see Fig. 8.7)and
the motion reduces to Staude’s rotation with the uniform angular velocity r0 = aA ρ
around the z-axis, whose positive half bearing the centre of mass is directed vertically
downwards on the first ray and upwards on the second.
On the line j = 2ρ , 0 < ρ ≤ 1, (8.68) reduces to

±2ρ 1 − ρ2 2 1 − ρ2 sinh(w) 2 1 − ρ2
γ=( , ,1 − ),
cosh(w) cosh2 (w) cosh2 (w)

a ±2ρ 1 − ρ2
ω= ( , 0, ρ), (8.70)
A cosh w
@seismicisolation
@seismicisolation
Fig. 8.7 Limiting cases

where w = a
A
(1
− ρ2 )(t − t0 ). Those solutions describe asymptotic motions and

tend to Staude’s rotation with velocity ω = aA ρ around z-axis directed upwards.
8.5.2 Region I I. The Second Class of Motions
In the region I I (see Fig. 1)
1
j ≤ ρ3 + , j ≥ 2ρ, (8.71)
ρ
the polynomial F(γ1 ) has four real roots and Eq. (8.67) will take the form
γ1 = ± N dn(w, ν), (8.72)
where

a j − 2ρ3 + 2ρd
w= (t − t0 ),
Aρ 2
4ρd
N = ρ( j − 2ρ3 + 2ρd), ν 2 = . (8.73)
j − 2ρ3 + 2ρd
Contours of ν are shown in Fig. 2. The vectors γ, ω will have the form
2d j N2
γ = (±N dn(w, ν), sn(w, ν) cn(w, ν), − 2 dn2 (w, ν)),
N 2ρ 2ρ

a N
ω= (± dn(w, ν), 0, ρ). (8.74)
A ρ
@seismicisolation
@seismicisolation
On the boundary curve j = ρ3 + ρ1 , ρ ≤ 1, the motion degenerates into Staude’s

regular rotation with angular velocity ω = aA 1ρ around one of the directions γ =

(± 1 − ρ2 , 0, ρ) (see Fig. 3). On the line j = 2ρ , 0 < ρ ≤ 1, we get the pair of
solutions (8.70) describing motions asymptotic to Staude’s rotations around z-axis.
Remark: The, so-called, orbital stability of the Bobylev–Steklov motion was
recently investigated in [426]. Some results will be presented in due course.
8.6 Steklov’s Case (1899)
8.6.1 Conditions and Solution
Suppose that the centre of mass lies on the first principal axis, i.e. y0 = z 0 = 0.
Equations of motion (3.29) take the form
A ṗ + (C − B)qr = 0,
B q̇ + (A − C) pr = aγ3 ,
C ṙ + (B − A) pq = −aγ2 , (8.75)
γ̇1 + qγ3 − r γ2 = 0, γ̇2 + r γ1 − pγ3 = 0, γ̇3 + pγ2 − qγ1 = 0, (8.76)
where we have put Mgx0 = a. Without loss of generality we assume a > 0. The
general first integrals of motion become
1
I1 ≡ (Ap 2 + Bq 2 + Cr 2 ) + aγ1 = h,
2
I2 ≡ Apγ1 + Bqγ2 + Cr γ3 = f,
I3 ≡ γ12 + γ22 + γ32 = 1, (8.77)
where h, f are certain parameters, determined by initial state of motion.

When B = C, the equations describe Lagrange’s integrable case discussed in
Chap. 2. Thus, without loss of generality, one can choose the y-axis to be the one
of greater moment of inertia, i.e. assume that B > C. Steklov assumes a solution of
(8.75), (8.76), such that
γ2 = n 2 pq, γ3 = n 3 pr, (8.78)
without indicating any line of thought leading to that assumption. It is not hard to
reveal that under this assumption (8.75) can be cast in the form of equations of motion
of a heavy body about its centre of mass, i.e. can be rendered to the integrable case of
Euler and Poinsot, see Sect. 4.1, thus guaranteeing integrability and solution in terms
of elliptic functions of time. Equations (8.76) will serve as compatibility conditions,
may be, leading to some restrictions on the parameters of the problem.
@seismicisolation
@seismicisolation
Now, substituting (8.78) into (8.75)–(8.77) and using again the resulting dynam-
ical equations, we determine the values of the parameters:
f = 0,
A2
h = a[1 − ],
2(A − B)(A − C)
(A − B) (A − C)
n2 = ,
a (A − 2 C)
(A − B) (A − C)
n3 = , (8.79)
a (A − 2 B)
where = ±1 and the relations between the variables

A A−C 2 a (A − 2 C)
q2 = [ p − ],
B −C A−2B ( A − B) (A − C)
A A−B 2 a (A − 2 B)
r2 = [− p + ]. (8.80)
B −C A − 2C (A − B) (A − C)
It remains to find expression for p in terms of time. To this end, we use (8.80) in
the first of Eqs. (8.75). We get

A−C 2 a (A − 2 C) A−B 2 a (A − 2 B)
ṗ = ± [ p − ][− p + ],
A−2B (A − B) (A − C) A − 2C (A − B) (A − C)
(8.81)
so that p and hence q, r are elliptic functions of time. However, we deal here with
real motion, where all variables are real, and this leads only to two classes of motion,
characterized by the two different choices = ±1. For those classes we have some
conditions on the moments of inertia satisfied in addition to the triangle inequalities:
B > A > 2C in the first case ( = 1),

A > B > C, A > 2C in the second case ( = −1).
In his original 1899 paper [347], Steklov considered only the first class of motions,
but formulas derived there cover also the second class, which was noted and consid-
ered in detail by De Angeli [8] in 1934. The latter class was also rediscovered by
Kuzmin in 1952 [248]). It is also noteworthy that the method devised by Steklov to
obtain his solution was used and generalized in later works of other authors as Gory-
achev [114], Chaplygin [54], Kowalewski [239] and Kharlamov [196]. Solutions
obtained in this way will be presented below.
The regions where the conditions on moments of inertia are satisfied are shown in
Fig. 8.8 in the plane of (B/A, C/A). They are represented in Fig. 8.8, respectively,
@seismicisolation
@seismicisolation
8.6 Steklov’s Case (1899) 143
Fig. 8.8 Regions of reality of Steklov’s solution
Fig. 8.9 a Contours of h. b Contours of k1 , k2
by the two triangular regions I (P Q R) and I I (Q RS). Note that on the line P R in
region I I we have B + C = A and the body becomes a distribution of mass in the
yz-plane, while its centre of mass lies on the x-axis outside that plane. The line P R
is thus excluded from region II.
The energy constant h in the first region takes a value ranging from 3a at S to
infinity on Q R. In the second region it ranges from a at P to infinity on Q
R. Contour
lines of h are shown in Fig. 8.9a. Figure 8.9b shows contours of k1 = B−C B−A
and

k2 = A−B , the moduli of elliptic functions in regions I and I I, respectively. Note
A−C
√
that k1 changes from 0 on Q R to 1/ 2 at S, while k2 changes from 0 on Q R to 1
at P.
We now write the solution of the equations of motion in the final explicit form.
@seismicisolation
@seismicisolation
8.6.2 The First Class
In this case the centre of mass lies on the middle axis of inertia. The components of
the angular velocity are expressed as

(2B − A)(A − 2C)a
p= cn(u 1 ),
(B − A)(A − C)2

A(A − 2C)a
q= sn(u 1 ),
(B − A)(B − C)(A − C)

(2B − A)Aa
r = dn(u 1 ), (8.82)
(B − A)(A − C)2

(B−C)a
where u 1 = 1 (t − t0 ), t0 is an arbitrary constant and 1 = (B−A)(A−C)
. The vector
γ is given by
A sn2 (u 1 ) − C A
γ1 = =1− cn2 (u 1 ),
A−C A−C

A(2B − A)
γ2 = sn(u 1 ) cn(u 1 ),
(B − C)(A − C)
√
A(A − 2C)
γ3 = cn(u 1 ) dn(u 1 ). (8.83)
A−C
8.6.3 The Second Class
In this case the centre of mass of the body lies on the axis of largest moment of
inertia.

(2B − A)(A − 2C)a
p= sn(u 2 ),
(A − B)(A − C)2

A(A − 2C)a
q= cn(u 2 ),
(A − B)(B − C)(A − C)

(2B − A)Aa
r = dn(u 2 ), (8.84)
(A − B)(B − C)(A − C)

where u 2 = 2 (t − t0 ), 2 = a
A−B
and γ is given by
@seismicisolation
@seismicisolation
A sn2 (u 2 )
γ1 = −1 + ,
A−C

A(2B − A)
γ2 = − sn(u 2 ) cn(u 2 ),
(B − C)(A − C)

A(A − 2C)
γ3 = sn(u 2 ) dn(u 2 ). (8.85)
(A − C)(B − C)
8.6.4 Some Properties of the Motion
8.6.4.1 1
We have suppressed some different combinations of signs of the square roots in

(8.82)–(8.85) which are equally possible. Those may be obtained by applying the
transformations u → u + 2K or (t, ω, γ) → (−t, −ω, γ).
8.6.4.2 2
Both classes of motion are periodic. The periodic times for them are
4K (ki )
Ti = , i = 1, 2. (8.86)
i
The motions consist of vibrations with period T about x- and y-axes and unidirec-
tional rotation with variable angular speed about the z-axis with period T /2. The two
periods become zero on the line Q R (B = A) corresponding to the limiting case of
infinitely large angular velocities.
8.6.4.3 3
One way to understand the geometry of motion, as was done in the Bobylev–Steklov
case, is to examine what we call the “trajectory” or “orbit” of motion. That is the
curve traced during the motion of the body by the tip of the vector γ on the Poisson
sphere γ12 + γ22 + γ32 = 1. For the first class of motions this curve can be described
as the intersection of the sphere with the elliptic cylinder
(2B − A) (γ1 − 1) [(A − C) γ1 + C] + A(B − C)γ22 = 0. (8.87)
Noting that for all values of the parameters the point P1 (1, 0, 0) lies on that intersec-
tion, so that this point is common between all orbits of the first class. Moreover, the
@seismicisolation
@seismicisolation
Fig. 8.10 a Three class 1 orbits passing through P1 . b Three class 2 orbits passing through P2
projection of the orbit on the yz-plane in the vicinity of the point P1 is described by
the equation
A − 2C 2
γ32 − γ =0 (8.88)
2B − A 2
representing two line segments intersecting at that point. The generic orbits of
motions of the first class on the Poisson sphere are Fig. 8-shaped curves, all with node
at P1 (See Fig. 8.10a). It is easy to show that orbits of the second class of motions
are also Fig. 8 curves, but with their node at P2 (−1, 0, 0).
8.6.4.4 4
On the border line P S both classes of motion degenerate into plane (pendulum-like)
motions with p = q = 0 about the z-axis, which is taking a horizontal position.
Those are rotations with a variable but unidirectional angular velocity

(B − C)a
r =2 dn(u 1 , k1 ) for the first class of motions ,
C(B − 2C)

a
r =2 dn(u 2 , k2 ) for the second class of motions ,
2C − B
where 1 , 2 , k1 , k2 are calculated for A = 2C. The vertical unit vector is given by
γ1 = −1 + 2 sn2 (u 2 ), γ2 = −2 sn(u 2 ) cn(u 2 ), γ3 = 0.
The energy constant for the first class of motions varies on Q S from 3a at S to infinity
at Q, while for the second class of motions it varies on P Q from a at P to infinity
@seismicisolation
@seismicisolation
at Q. Note that on the energy level h = a, the plane motion becomes asymptotic to
the upper (unstable) equilibrium position.
On the line P S in the plane of inertia parameters, the two halves of Fig. 8 orbits
coincide and degenerate into circular orbits γ3 = 0, corresponding to pendulum-like
motion. Figure 8.10a, b shows two sequences of three Fig. 8 orbits of classes 1 and
2, respectively, with the last one close to the circular section γ3 = 0.
8.6.5 Exercises
1- Substituting γ2 , γ3 from (8.78) with n 2 , n 3 as given in (8.79) in Eqs. (8.75), the

last equations take the form
1
A ṗ + (C − B )qr = 0,
2
1
B q̇ + A − C pr = 0,
2
1
C ṙ + (B − A) pq = 0, (8.89)
2
in which B = 2B − A, C = 2C − A. Those equations can be written in the vector

form
1
Ġ + ω × G = 0, (8.90)
2
where G = (Ap, B q, C r ). This form shows clearly that the equations are identical
with the equations of motion about a fixed point by inertia of a hypothetical body
with inertia matrix I = diag(A, B , C ) and angular velocity ω referred to a system
of axes moving with angular velocity 21 ω. However, this analogy should be taken
with caution. The constants of motion of the hypothetical body are not arbitrary:
(a) Show that the analog of the energy integral corresponding to Steklov’s solution
has the value zero.
1
(Ap 2 + B q 2 + C r 2 ) = 0.
2
This makes no contradiction, since the hypothetical body has one negative moment
of inertia C = 2C − A < 0 for both of Steklov’s families of solutions.
(b) The angular momentum vector G of the hypothetical body is a constant vector
in the coordinate system moving with angular velocity 21 ω about O. Then its square
modulus G 2 is constant in all coordinate systems. Show also that the analog of the
integral of the angular momentum corresponding to Steklov’s solution has the value
@seismicisolation
@seismicisolation
(A − 2C)(2B − A)
G 2 = A2 p 2 + B 2 q 2 + C 2 r 2 = 2Aa .
(A − C)(B − A)
The right-hand side is positive for both classes of Steklov’s motions.
8.7 Goryachev’s Case [114] (1899)
8.7.1 Conditions and Solution
Like in the Bobylev–Steklov and Steklov cases, the centre of mass of the body is
assumed on one of the principal axes of inertia at the fixed point. Without loss of
generality we take this axis to be the x-axis, so that the motion is described by the
same equations of the previous section (8.75), (8.76) and the same integrals (8.77).
For a solution, Goryachev tries two relations of the form
γ2 = pq(n 1 + n 2 p 2 ), γ3 = n 3 pr (8.91)
and finds that this is possible only when the principal moments of inertia of the body
are subject to the restriction
A(9B − 8C) = 16C(B − C). (8.92)
In the plane of inertia parameters, the condition (8.92) is satisfied on the curve
P Q N R (Fig. 8.11), so that the character of motion of a body is determined by the
point representing the body on that curve. Denoting the ratio C/A by c and solving
(8.92) for B one can write
8c(2c − 1)
C = Ac, B = A . (8.93)
16c − 9
Regarding triangle inequalities, the parameter c ranges from 35 at P to ∞ at R (the

point at infinity).
From the equations of motion, it follows also that the areas and energy parameters
of the motion must take the values
f = 0,
(2c − 1)(9 − 56c + 64c2 )
h = 2εa , (8.94)
(3 − 4c)(3 − 8c)(3 − 4c)
where
ε = ±1.
@seismicisolation
@seismicisolation
8.7 Goryachev’s Case [114] (1899) 149
Fig. 8.11 Goryachev’s

condition (8.92) is satisfied
on the hyperbolic branch
PN QR
Under the above conditions, five of Euler–Poisson’s variables can be expressed

in terms of p as

16c − 9 32ε(2c − 1)a p2
q= − 2,
4 c(4c − 3)(8c − 5)(8c − 3)A c
1 128ε(16c − 9)a (4c − 3)(48c2 − 44c + 9) 2
r = {− + 32 p
8 c(4c − 3)A c2 (2c − 1)
ε(8c − 5)(8c − 3)(4c − 1)(4c − 3)3 A 4 1/2
− p } ,
c3 (2c − 1)2 a
(4c − 3)(16c2 − 15c + 3)A 2
γ1 = ε − p
2c(2c − 1)a
(8c − 3)(8c − 5)(4c − 1)(4c − 3)3 A2 4
− p ,
128c2 (2c − 1)2 a 2
(4c − 3)(8c − 3)(8c − 5)A (4c − 1)(4c − 3)2 A 2
γ2 = pq[−1 + ε p ],
2(2c − 1)(16c − 9)a 16c(2c − 1)a
(3 − 4c)A
γ3 = pr, (8.95)
2a
while p is determined as a function of time from the relation
p
dp B −C
√ = t, (8.96)
F( p) A
where F( p) = q 2 ( p)r 2 ( p) is a polynomial of the sixth degree in p.
@seismicisolation
@seismicisolation
It remains to ensure that the two square roots figuring in the expressions for q and
r in (8.95) take only real values, depending on the ratio c of the principal moments
of inertia of the body and the sign of ε. Regarding the first expression in (8.95), the
arc P R is divided into three parts:
(1) P Q, on which ci 0 = [ 53 , 58 ). On this arc, one should choose ε = 1 and then
h[ 26
9
a, ∞).
(2) Q N , on which c( 58 , 43 ). On this arc, if one chooses ε = −1, q can take real
values. However, in that case, all the three coefficients under the square root sign
in the expression for r become negative, so that r takes only imaginary values.
(3) N R, with c( 34 , ∞) and ε = 1. On this arc, it can be shown that h < −a, so that
the kinetic energy of the body is negative.
Thus, the motion is possible only for bodies corresponding to points of the arc
P Q with the choice ε = 1.
In fact, with the use of the change of variable

2c(2c − 1)a
p=4 v, (8.97)
(3 − 4c)(8c − 3)(5 − 8c)A
the relation (8.96) is rendered to

(4c − 1)(3 − 4c)a dv
2 t= √ , (8.98)
c(8c − 3)(5 − 8c)A (v − v0 )v(v1 − v)(1 − v)
where

48 c2 − 44 c + 9 2 9 − 64c + 160 c2 − 128 c3
v0,1 = −1/2 ∓ 1/4 √ . (8.99)
(4 c − 1) (3 − 4c) (4 c − 1) (3 − 4 c)
We first note that, as shown in Fig. 8.12, on the interval i 0 = [ 35 , 58 ), the roots of
the fourth-degree polynomial under the square root sign in (8.98) have the order
v0 < 0 < v1 < 1.
The polynomial under the quadratic root sign in (8.98) has degree four, and thus,
the integral is elliptic of the first kind. From tables of integrals, e.g. [130], we evaluate
this integral and hence obtain the inversion formula
n sn 2 (u, k)
v= , (8.100)
1 − m sn2 (u, k)
where
u = μt + u 0 ,
@seismicisolation
@seismicisolation
8.7 Goryachev’s Case [114] (1899) 151
Fig. 8.12 The order of the

roots of the fourth-degree
polynomial in (8.98)
Fig. 8.13 k v. c
√
4
(4c − 3)(128c3 − 160c2 + 64c − 9) a
μ= √ ,
2c(8c − 3)(5 − 8c)A

v1 (1 − v0 )
k=
v1 − v0
√
1 2(256c3 − 320c2 + 144c − 27)
= + √ ,
2 8(4c − 3)3/2 128c3 − 160c2 + 64c − 9
23/2 (8c − 5)(8c − 3)(16c − 9)
n= ,
16 (4c − 3)(128c3 − 160c2 + 64c − 9)
√
1 2(48c2 − 44c + 9)
m = [1 − ], (8.101)
2 (4c − 3)(128c3 − 160c2 + 64c − 9)
and u 0 is an arbitrary constant. Figure 8.13 shows the graph for k(c) as drawn from
the last expression. It ranges from 0.87064540 at c = 35 to 0 at c = 58 .
Using (8.100), we express the components of the angular velocity as follows:
@seismicisolation
@seismicisolation

a sn(u, k)
p = p1 ,
A 1 − m sn2 (u, k)

a dn(u, k)
q = q1 ,
A 1 − m sn2 (u, k)

a cn(u, k)
r = r1 , (8.102)
A 1 − m sn2 (u, k)
in which
√
2 2c(2c − 1)(16c − 9)
p1 = ,
4
(4c − 3)(128c3 − 160c2 + 64c − 9)

2(2c − 1)
q1 = (16c − 9) ,
c(4c − 3)(8c − 3)(8c − 5)

2(16c − 9)
r1 = . (8.103)
c(3 − 4c)
The components of the vertical unit vector γ take the following form:
√ (16c − 9)(16c2 − 15c + 3) sn 2 (u, k)

γ1 = 1 − 2 2
(4c − 3)3 (128c3 − 160c2 + 64c − 9) 1 − m sn2 (u, k)
(4c − 1) (8c − 3) (5 − 8c) (16c − 9)2 sn4 (u, k)
− ,
4(4c − 3) (128c − 160c + 64c − 9) (1 − m sn2 (u, k))2
2 3 2
(8c − 3)(8c − 5)(4c − 3) sn(u, k) dn(u, k)

γ2 = p1 q 1
8((2c − 1)(16c − 9)) (1 − m sn2 (u, k))
√
2 (4c − 1) (16c − 9) sn2 (u, k)
×[ − 4],
(4c − 3) 128c3 − 160c2 + 64c − 9 1 − m sn (u, k)
2
3 − 4c sn(u, k) cn(u, k)
γ3 = p1 r 1 . (8.104)
2 (1 − m sn2 (u, k))3/2
8.7.2 Properties of the Motion
8.7.2.1 The Initial Motion
To obtain symmetric views in the graphics, we give the constant u 0 the value 0. This
corresponds to the choice of the initial moment t = 0 as the one at which p = 0.
At this same moment, from (8.104), we have γ = (1, 0, 0). The x-axis carrying the
centre of mass begins motion from a position vertically above the fixed point. The
@seismicisolation
@seismicisolation
8.7 Goryachev’s Case [114] (1899) 153
Fig. 8.14 a The initial angular velocity (q , r ) as c varies. b Directions of angular velocity during
motion as c varies
initial angular velocity that should be given to the body to commence a motion of
the Goryachev type lies in the yz-plane and its direction depends on the parameter
c.
Figure 8.14 shows the variation of the dimensionless quantities (q = aA q, r =

A
a
r ) as c varies on the interval i 0 . From this figure, it is obvious that as c tends to
√
5
8
, q tends to infinity while r tends to a finite limit 4 10
5
a
A
2.5298 aA .
8.7.2.2 Periodicity of the Motion
We now clarify the general character of motion after this initial setting. We readily
note that the second component q of the angular velocity does not change sign. In
a time period 2Kμ(k) , it ranges from a minimum q0 to a maximum √1−v q0
and then to
1
q0 again. The motion of the body is composed of a rotation around the y-axis and
vibrations around the x, z-axes. The whole motion is periodic with period
√
4K (k) 2c(8c − 3)(5 − 8c)A/a
T = = 4K (k) . (8.105)
μ 4
(3 − 4c)(9 − 64c + 160c2 − 128c3 )
The period of motion decreases monotonically from its value at c = 3/5 to zero at
c = 5/8, corresponding to very fast uniform rotation about the y-axis.
In Fig. 8.14b, we show four closed curves representing the traces on the unit sphere
fixed in the body by the unit vector ωω in the direction of the angular velocity along a
period of the motion. All curves are closed around the y-axis. As c approaches 5/8,
the curve becomes a very small one shrinking to that axis.
@seismicisolation
@seismicisolation
Fig. 8.15 Space view of the trajectory for c = 0.6 on a transparent Poisson sphere
Fig. 8.16 Motion direction on two halves of an opaque sphere
8.7.2.3 Orbits on the Poisson Sphere
Now, we turn to the picture of the trajectories of the motion, the trace of the vertical
vector γ on the Poisson sphere. As we have seen above, the point P(1, 0, 0) is com-
mon between all orbits. Figure 8.15 shows the space view of the orbit corresponding
to the value c = 3/5, as an example of the generic orbits. It begins from the point
P on the x-axis and goes through the points P1 , Q, R, P2 , P, P3 , Q, R, P4 and then
closes at P. The orbit has three self-intersection points P, Q and R, so that it makes
three loops, two small and one larger. The direction of motion is shown on all arcs
of the orbit on the two halves of an opaque sphere (Fig. 8.16a, b).
Although those orbits are not simple in the space view, they have two planes of
symmetry: x z and x y. The projections of four orbits on the x z-plane are shown in
Fig. 8.17 for values of c ranging from the beginning to near the end of the interval
@seismicisolation
@seismicisolation
8.7 Goryachev’s Case [114] (1899) 155
Fig. 8.17 Projections of trajectories on the x z-plane corresponding to four values of c (View from
the top of y-axis)
Fig. 8.18 Projections of trajectories on the x y-plane corresponding to four values of c (View from
the top of z-axis)
[ 53 , 58 ). Each orbit can be seen as the intersection of the Poisson sphere with a cylin-
drical surface, which is parallel to the y-axis and touches the unit sphere at three
points. With increasing c, the projection becomes wider and encloses the orbits with
smaller values of c. As c → 5/8, the projection of the orbit approaches a circle of unit
radius. In fact, the limiting orbit at this value of c is the circle γ2 = 0, corresponding
to very fast rotations about the y-axis, which takes a horizontal position.
Figure 8.18 depicts the projections of the same orbits on the x y-plane. This pro-
jection begins wider at c = 3/5 and becomes narrower with increasing c to coincide
in the limit, as c → 5/8, with the diameter of the sphere on the x-axis. Figure 8.19
shows the same orbits on the front and the rear halves of the sphere.
@seismicisolation
@seismicisolation
Fig. 8.19 Front and rear views of the four trajectories on the Poisson sphere
8.7.3 The History of Goryachev’s Case
(1) In his 1899 work [114], Goryachev obtained the condition (8.92) and the expres-
sions (8.95), but he erroneously concluded that the integral (8.96) is hyper-
elliptic.
(2) In 1965, Kharlamov [196] derived a relation equivalent to (8.98) and concluded
that the motion is periodic. He also evaluated the time period of the motion, but
he did not proceed to invert the elliptic integral.
(3) In 1974, Stepanova [349] gave an explicit inversion formula of the integral (8.96)
in the form

1 − sn(χt, k0 )
p = p0
1 + n 0 sn(χt, k0 )
and expressed the six Euler–Poisson variables as functions of time. Unfortunately,

the solution given by Stepanova is not adequate for describing the motion, since it
involves odd powers of the auxiliary quantity p, which is not differentiable at all
its zeros and takes the wrong sign on half of each period. As a result, expressions
for p, r and γ2 are not analytic functions of time. Moreover, expressions for the
parameters m 0 , n 0 , χ and the modulus k0 of the Jacobi elliptic functions are extremely
complicated. In fact, we have established that elliptic functions used in (8.102)–
(8.104) and the elliptic functions used in Stepanova’s solution are connected by
second-order transformation of the Landen type (See, e.g. [75]).
The expression for γ2 in [349] is written erroneously. A corrected expression for
γ2 is given in [108] (Sect. 8.3). The authors of [108] remarked also that Stepanova’s
solution is two-valued and that its period is only half the period found earlier by
@seismicisolation
@seismicisolation
8.7 Goryachev’s Case [114] (1899) 157
Kharlamov based on the analysis of the quadrature in (8.98), but they did not realize
the need to modify the way of inversion of the integral used by Stepanova.
8.8 Chaplygin’s Case (1904)
As in the previous two cases, the centre of mass of the body is assumed on the x-axis,
so that the motion is described by same Eqs. (8.75), (8.76) and integrals (8.77). In
Chaplygin’s own words in his original paper [54], one reaches the conclusion that
the present case was found by certain extension of the method used in Steklov’s and
Goryachev’s cases. Generalizing (8.78) and (8.91), Chaplygin tries two relations of
the form
aγ2 = q(α p + λ p n ), aγ3 = r (β p + μ p n ), (8.106)
where n, α, β, λ and μ are constants to be determined. Equations (8.75) give
B 2μ n+1
(B − C)q 2 = (C − A − β) p 2 − p + Bk,
A n+1
C 2λ n+1
(B − C)r 2 = (A − B + α) p 2 + p + Cl, (8.107)
A n+1
in which k, l are integration constants. Proceeding to satisfy Eqs. (8.76), Chaplygin

reaches the conclusion that three choices are possible:
(1) λ = μ = 0, which leads to the case of Steklov.
(2) λ = 0, n = 3, which leads to Goryachev’s case.
(3) n = − 13 , k = l = 0, which gives the new case known now under his name.
To make the nature much clearer, we shall now describe this case in a slightly
modified way than that of Chaplygin [54] and several authors who mostly repeated
Chaplygin’s approach [108, 195].
Substituting expressions (8.106) into the integrals of motion and Euler’s equations,
we obtain the values of the parameters
h = 0, f = 0,
(B − A) (A − C) (B − A) (A − C)
α= ,β= ,
(A − 2 C) (A − 2 B)
3A − 2B 3A − 2C
λ= Cs, μ= Bs, (8.108)
2C − A 2B − A
s being a constant, determined from the equation
(2B − A)2 (2C − A)2

9A3 (2B + 2C − 3A)ν 3 = 4a 2 , (8.109)
(3A − 2B)(3A − 2C)
and the moments of inertia are subject to the restriction
@seismicisolation
@seismicisolation
9(2B − A)(2C − A) = 4BC. (8.110)
The last condition determines the curve P Q R in the plane (B/A, C/A) (Fig. 8.20).
This curve can be parametrized by a parameter c so that C = Ac, B = 9A(2c−1) 2(16c−9)
.
The edge√
point P, at which the equality B = A + C is satisfied, corresponds to
c = 1+16 73 = 0.596 5.
On the other hand, to avoid the appearance of cubic roots, we introduce an interme-
diate variable u by the relation p = u 3 . The Euler–Poisson variables can be written
in the following form:
p = u3,

(9 − 16 c) 3s (3 − 2c) − (1 − c) u 4 u
q= − √ ,
c (8 c − 3) (3 − 4 c)

36s (3 − 5c) − (9 − 14c) u 4 u
r = √ ,
(8 c − 3) (3 − 4 c) (2 c − 1)

Au 2 − (14 c − 9) (c − 1) u 4 + 9 12 c2 − 22 c + 9 s
γ1 = −1/4 ,
ca (2 c − 1)

A 3s (3 − 2c) − (1 − c) u 4 12cs (3 − 5c) − u 4 (9 − 14c) (1 − c)
γ2 = − √ √ √ ,
2a c 8 c − 3 3 − 4 c (2 c − 1)

A − (14 c − 9) (c − 1) u 4 + 9 (2 c − 1) s (2 c − 3)
γ3 = − √ √ √ ×
4a c 8c − 3 3 − 4c 2c − 1

× 36 (3 − 5 c) s − (9 − 14 c) u 4 , (8.111)
where s is now given from
4a 2 c2 (2c − 1)2
s3 = . (8.112)
27A2 (3 − 4c)2 (3 − 5c)(3 − 2c)
In order that all√ the variables be real, the parameter c must be restricted to the interval
[c0 , c∗ ] ≡ [ 1+16 73 , 35 ] corresponding to the arc P Q in Fig. 8.20.
The change of the variable u with time is determined from the equation

36s (3 − 5c) − (9 − 14c) u 4 3s (3 − 2c) − (1 − c) u 4
u̇ = √ , (8.113)
6 c(2c − 1)
so that the time can be evaluated by the expression

c(2c − 1)A du
t =6 .
a 36s (3 − 5c) − (9 − 14c) u 4 3s (3 − 2c) − (1 − c) u 4
(8.114)
@seismicisolation
@seismicisolation
8.8 Chaplygin’s Case (1904) 159
Fig. 8.20 Physically

admissible moments of
inertia lie on the arc P Q
Remark. An equation equivalent to this equation was obtained by Chaplygin [54],

who proceeded to express the last integral as the sum of two elliptic integrals (see
exercises). He also obtained the equation of the moving hodograph of the angular
velocity as a cone of the second degree and the fixed hodograph as a surface of
revolution. The motion of the body should be represented geometrically as rolling the
first hodograph on the second. This description, however, turns out to be inaccurate,
because, as we shall see below, the actual hodograph is only a part of the movable
cone. The full quadratic cone never makes any complete revolutions.
Remark. Kharlamov [196] introduced instead of u in (8.111) a variable σ by the
relation p = σ 3/2 . This led to some complications in his reasoning, but he was able
to show that only a part of the movable hodograph (cone) rolls over an equal arc of
the immovable hodograph in one direction and the body reaches an instant state of
rest and then immediately begins rolling on the same arc in the opposite direction.
Here we shall go the same way as in the last few cases. Let us write (8.114) in the
form
1 c(2c − 1)A du
t= , (8.115)
s (3 − 5c) (3 − 2c) a 4
1− u 1− u
4
u 41 u 42
where
22/3 31/4 [c(2c − 1)(3 − 5c)]1/6

u1 = ,
(9 − 14c)1/4 (3 − 2c)1/12 (3 − 4c)1/6
[2c(2c − 1)(3 − 2c)]1/6
u2 = . (8.116)
(1 − c)1/4 (3 − 5c)1/12 (3 − 4c)1/6
Note that on the interval of change of c, we have u 2 > u 1 ≥ 0. The variable u then
takes its values on the interval [−u 1 , u 1 ]. Thus, we can use the substitution
@seismicisolation
@seismicisolation
u = u 1 sin φ, (8.117)
so that (8.115) becomes

√ φ
3 (3 − 4c)A/a dφ
t= , (8.118)
[3(3 − 2c)(9 − 14c)] 1/4
0 1 + sin φ 1 − k 4 sin4 φ
2
where k = u 1 /u 2 . We have chosen φ = 0 at t = 0, so that we suppressed an imma-

terial arbitrary integration constant. Thus, φ monotonically increases with t and the
Euler–Poisson variables take their final parametric form in terms of φ
√ √ √
a 4 · 33/4 c 2 c − 1 3 − 5 c
p= √ √ sin3 φ,
A (9 − 14 c)3/4 4 3 − 2 c 3 − 4 c
√ √ √
a 2 4 3 2 c − 1 4 3 − 2 c (16 c − 9)
q = √ sin (φ) 1 − k 4 sin4 φ,
A (9 − 14 c)1/4 (4 c − 3) 8 c − 3
√ √
a 4 · 33/4 c 3 − 5 c (sin (φ))2 + 1 sin (φ) cos (φ)
r = √ √ √ ,
A 4
3 − 2 c (4 c − 3) 4 9 − 14 c 8 c − 3
√
3 3 sin2 φ[12 c2 − 22 c + 9 − 4 (3 − 5 c) (1 − c) sin4 φ]
γ1 = √ √ ,
3 − 2 c (3 − 4 c) 9 − 14 c
√ √
4 c 3 − 5 c[c − 3 (1 − c) sin4 φ]
γ2 = √ 1 − k 4 sin4 φ,
(3 − 4 c)3/2 8 c − 3
√ √
3 3 2 c − 1 cos φ sin2 φ + 1
γ3 = √ √ ×
2 c − 3 (4 c − 3)3/2 8 c − 3
×[4 (5 c − 3) (c − 1) sin4 φ − (2 c − 1) (2 c − 3)]. (8.119)
@seismicisolation
@seismicisolation
8.8 Chaplygin’s Case (1904) 161
8.8.1 Properties of the Motion
8.8.1.1 The Periodicity
The motion is periodic in time with period T

√ 2π
3 (3 − 4c)A/a dφ
T =
[3(3 − 2c)(9 − 14c)] 1/4
0 1 + sin φ 1 − k 4 sin4 φ
2
√ π/2
12 (3 − 4c)A/a dφ
= . (8.120)
[3(3 − 2c)(9 − 14c)]1/4 0 1 + sin φ 1 − k 4 sin4 φ
2
8.8.1.2 The Hanging Centre of Mass
From (8.108) we have h = 0. The energy integral gives
1
γ1 = − (Ap 2 + Bq 2 + Cr 2 ). (8.121)
2a
This means that the x-axis, which carries the centre of mass of the body, never rises
over the horizontal plane passing through the fixed point. When the centre of mass
touches that plane, the angular velocity of the body vanishes.
8.8.1.3 The Initial Motions
From the qualitative point of view, the motion varies slightly as c varies on the
admissible period [c0 , c∗ ]. Equations (8.119) tell that at t = 0(φ = 0) for all values
of c the initial angular velocity vanishes p = q = r = 0. The body begins from a rest
position. At this moment, the x-axis carrying the centre of mass occupies a horizontal
position γ1 = 0, and the positive y-axis makes with the horizontal an angle
√ √
4 3 c3/2 3 − 5c
δ = arctan √ . (8.122)
9 (2c − 1)3/2 3 − 2c
8.8.1.4 A Border Case—The Physical Pendulum Motion
For the border value c∗ we have δ = 0, i.e. the body begins motion with x, y hori-
zontal and z vertical upwards. On the following motion, we have
@seismicisolation
@seismicisolation
Fig. 8.21 Traces of γ on the

Poisson sphere from the
negative side of the x-axis.
View from top of -ve x-axis
2
ω = (0, − √ sin φ, 0),
3

γ = (− sin φ, 0, cos φ 1 + sin2 φ).
2
The centre of mass begins to drop down, moving on a vertical circle, until it reaches
the horizontal position opposite to its initial position and then reverses the direction
of motion returning again to that position. The whole motion is just a planar vibration
of the body as a pendulum about the y-axis which keeps a horizontal position. In
this case, k = 0. The integral in the relation (8.118) turns into an elliptic integral of
the first kind and, as expected, all variables in (8.119) render to elliptic functions of
time.
8.8.1.5 The Orbits on the Poisson Sphere
Figures 8.21 and 8.22 depict the trajectories of the point of intersection of the vertical
upward with the Poisson sphere for five equidistant values of c on the interval [c0 , c∗ ].
For c = c∗ , the trajectory is the semi circle in the x z-plane. As c decreases to c0 ,
the trajectories become slightly deformed from that trajectory as seen from the two
figures.
8.9 Kowalewski’s Case [239] (1908)
Kowalewski assumed, as in the previous three cases, that the centre of mass of the
body lies on the first principal axis of inertia. Thus, the problem is described, as
in those cases, by the same equations of motion (8.75)–(8.76) and admit the same
integrals (8.77).
@seismicisolation
@seismicisolation
8.9 Kowalewski’s Case [239] (1908) 163
Fig. 8.22 The same

trajectories on a transparent
Poisson sphere
After the success of Goryachev and Chaplygin in finding new solvable cases using
slight variations of Steklov’s method, Kowalewski developed an extension of that
method. With this method, he restored the three previous cases and found one more.
For a detailed presentation of that method, the reader is referred to Kowalewski’s
original paper [239] or Leimanis’ book. We shall directly use his ansatz for the new
solution of the Euler–Poisson system of equations in the same way as in the last three
cases. This expresses four of the variables in the form
γ2 = q(n 0 + n 1 p + n 2 p 2 ), γ3 = r (m 0 + m 1 p),
q 2 = q0 + q1 p + q2 p 2 ,
r 2 = r0 + r1 p + r2 p 2 + r3 p 3 , (8.123)
and the fifth is obtained from the energy integral as
1
γ1 = 1/a{h − [Ap 2 + Bq 2 ( p) + Cr 2 ( p)]},
2
so that five variables are expressed in terms of p. Inserting this choice into the
remaining two integrals of motion and the dynamical equations, we find that the
following condition on the moments of inertia must be satisfied:
B −C
A = 18C . (8.124)
10B − 9C
In the plane of ratios of moments of inertia, this condition is satisfied on the hyperbolic
branch P Q R. The point P( 27 , 10 ) = P(1. 588 2, 0.588 24) is determined by the
17 17
triangle inequality B = C + A.
It will be more comfortable to use a parametric representation of the ratios of the
moments of inertia (Fig. 8.23)
@seismicisolation
@seismicisolation
Fig. 8.23 Kowalewski’s

condition (8.124) satisfied on
the arc P Q R
9 2c − 1
C/A = c, B/A = c . (8.125)
2 9c − 5
This will help writing formulas in a unique way using two parameters A and c. The
point P corresponds to the minimum value of c, say, c0 = 10/17.
For h, f , we get the values
(2c − 1)(3c − 1)P3 (c)a

h=− √ ,
(3 c − 2) P9 (c)
√
2 Aa (3 − 4 c) P3 (c) c(2c − 1)3 (3c − 1)
f = , (8.126)
[(3 c − 2) P93 (c)]1/4
where
P3 (c) = 384 c3 − 624 c2 + 305 c − 40,

P9 (c) = 1769472 c9 − 8257536 c8 + 16831488 c7 − 19642368 c6 + 14443596 c5
−6930624 c4 + 2167313 c3 − 425427 c2 + 47520 c − 2300. (8.127)
The Euler–Poisson variables can be written in terms of an auxiliary non-dimensional

variable v as follows:

a
p=N v,
A

(9 c − 5) N a R2 (v)
q = 1/3 ,
c (2 c − 1) A (3 c − 1) (2 − 3 c)

N a
r = 1/2 R3 (v),
c A
@seismicisolation
@seismicisolation

N2 (−2 + 3 c) 72 c2 − 71 c + 16 v 2
γ1 = [2 (3 c − 1) (−2 + 3 c) v −2 3
c 2c −1

864 c − 1424 c + 767 c − 134 v
3 2
+1/4
(2 c − 1)2
384 c − 752 c3 + 561 c2 − 189 c + 24
4
−1/4 ], (8.128)
(2 c − 1)3 (3 c − 1)

N2 R2 (v)
γ2 = [ (3 c − 1) (−2 + 3 c)2 v 2
c (2 c − 1) (3 c − 1) (2 − 3 c)

(−2 + 3 c) 30 c2 − 30 c + 7 v 384 c3 − 624 c2 + 333 c − 58
− + 1/8 ],
2c −1 (2 c − 1)2
N ((2 − 3 c) N v + N )
γ3 = 1/2 R3 (v), (8.129)
c
where
1/4
c2 (−2 + 3 c) (3 c − 1)2 (2 c − 1)6
N = (64 ) ,
P9 (c)
R2 (v) = 1 − 4 (3c − 1) (2c − 1) (2 − 3c) [(2c − 1) v 2 − 2v],

(3c − 2) 168c2 − 164c + 37 v 2
R3 (v) = −16 (3c − 1) (3c − 2) v + 4
2 3
2c −1

384 c3 − 624 c2 + 333 c − 58 v
−6
(2c − 1)2
384 c − 1112 c3 + 1101 c2 − 453 c + 66
4
− . (8.130)
(3c − 1) (3c − 2) (2c − 1)3
We note first that the polynomial R2 (v) figuring under square root in the expres-
sions for q and γ2 in (8.129) has a negative leading coefficient. If we denote by v1, v2
the roots of R2 , we can write
2(9c − 5)
q= (v − v1 )(v2 − v), (8.131)
3c
where
⎛ ⎞
1 1 −36c + 36c2 + 7
v1 = ⎝1 − ⎠,
(2c − 1) 2 (3c − 1) (3c − 2)
⎛ ⎞
1 ⎝1 + 1 −36c + 36c + 7 ⎠
2
v2 = . (8.132)
(2c − 1) 2 (3c − 1) (3c − 2)
@seismicisolation
@seismicisolation
Fig. 8.24 The roots of P2

and P3
As v2 > v1, the variable v can take its values only in the interval [v1 , v2 ].
The cubic R3 figuring under root in expressions for r, γ3 has negative discriminant
on [v1 , v2 ], so that it has one real root, say, v3 and two complex conjugate roots v4
and v̄4 . We can write
√
2(2 − 3c) 3c − 1
r =− (v3 − v)(v 2 − 2 Re(v4 ) + |v4 |2 ). (8.133)
c
It follows that v ≤ v3 , and thus v changes on the interval [v1 , min(v2 , v3 )]. But
comparing the graphs of v1 , v2 , v3 in Fig. 8.24, we finally have
v1 ≤ v ≤ v3 < v2. (8.134)
From the same figure, we see that with growing c the interval [v1, v3 ] shrinks to
zero at some value c∗ , say, of c. This value at which the root v3 of R3 equals the root
v1 of R2 is a root of the resultant of R2 and R2 . This can be easily seen to be a root
of the eighth-degree polynomial
P8 = 147456 c8 − 563712 c7 + 920032 c6 − 836796 c5 + 463806 c4

−160441 c3 + 33838 c2 − 3980 c + 200. (8.135)
Numerical solution gives the value
c∗ = 0.6469380234. (8.136)
The value c∗ corresponds to the point Q(1.040241051, 0.6469380234) in Fig. 8.23,

so that the feasible points constitute the arc P Q.
@seismicisolation
@seismicisolation
The relation of the variable v with time is obtained by substituting p, q, r into the
first equation of (8.75). We obtain

c A
v̇ = qr
2(9c − 5) a
√
2(2 − 3c) 3c − 1 a
= (v − v1 )(v2 − v) (v3 − v)|v − v4 |2 ), (8.137)
3c A
so that, after separating variables and integration, we have t given by a hyper-elliptic

integral
√
3c A/a dv
t= √ √ √ . (8.138)
2(2 − 3c) 3c − 1 (v − v1 )(v2 − v) (v3 − v)|v − v4 |
In this integral if v starts from v1 , it will increase up to v3 , decreases to v1 and then

increases again. The motion is periodic.
We now introduce a new variable, the angle φ, by the relation
v = v1 + (v3 − v1 ) sin2 φ, (8.139)
so that φ = 0 at the initial time t = 0. The expression (8.138) becomes

√
3c A/a
t= √ ×
(2 − 3c) 3c − 1
φ
dφ
× . (8.140)
0 [v2 − v1 − (v3 − v1 ) sin φ]|v1 − v4 + (v3 − v1 ) sin2 φ|
2
The variable φ increases monotonically with time. Components of the angular veloc-
ity become

a
p=N [v1 + (v3 − v1 ) sin2 φ],
A

2 √
q= (9c − 5) v3 − v1 v2 − v1 − (v3 − v1 ) sin2 φ sin φ,
3c
2
r = − (2 − 3c) (3c − 1)(v3 − v1 )|v1 − v4 + (v3 − v1 ) sin2 φ| cos φ.
c
(8.141)
The components of γ can be written accordingly. The overall period of the motion
is 2π in φ. The time period T of the motion is given by
@seismicisolation
@seismicisolation
√
3c A/a
T = √ ×
(2 − 3c) 3c − 1
2π
dφ
×
2
√
12c A/a
= √ ×
(2 − 3c) 3c − 1
π/2
dφ
× .
2
(8.142)
We note that in the limiting case c = c∗ , from (8.141), we get

a
p = −1.922773239 , q = r = 0.
A
We can also show that for the same value of c
γ1 = −1, γ2 = γ3 = 0.
Thus, at c = c∗ , the motion of the body in the case of Kowalewski renders to a

rotation with uniform angular speed about the first principal axis of inertia, while the
positive half of that axis, carrying the centre of mass points, vertically downward.
Figures 8.25, 8.26, 8.27, 8.28 and 8.29 depict trajectories of the motion on the
Poisson sphere fixed in the body for some values ranging from c0 to near to c∗ . The
trajectories have two planes of symmetry x y and x z. Projections on both planes are
shown in the first four cases.
In Fig. 8.25, we note a trajectory with four loops, large enough to intersect pair-
wise.
In Fig. 8.26, the loops become smaller and do not intersect.
In Fig. 8.27, the four loops become very small.
In Fig. 8.28, the loops disappear and the trajectory becomes a simple closed curve.
In Fig. 8.29, we show trajectories for two values of c near to c∗ . Those trajectories
are closed around the negative end of the x-axis. As c approaches c∗ , the trajectory
shrinks and the motion tends to uniform rotation about that axis, which then occupies
a vertical position.
8.10 Grioli’s Case (1947): The Regular Precession About a

Tilted Axis
The regular precession is that in which the body rotates uniformly about an axis
fixed in it (the figure axis), while that axis precesses also uniformly about an axis
fixed in space (the precession axis), keeping with it a fixed angle. As we have seen
@seismicisolation
@seismicisolation
8.10 Grioli’s Case (1947): The Regular Precession About a Tilted Axis 169
Fig. 8.25 Views of the trajectory for the minimal value c0 = 10/17
Fig. 8.26 Views of the trajectory for the value c = 3/5
@seismicisolation
@seismicisolation
Fig. 8.27 Views of the trajectory for the value c = 0.614
in Sect. 4.2, a wide class of regular precessional motions can be performed by an

axially symmetric body (Lagrange’s top). In all those motions, the figure axis is the
axis of symmetry and the precession axis is vertical. Initial conditions can even be
chosen so as to give the angle between the two axes any preassigned value.
As was shown long ago by Routh [312], an asymmetric body, in which no two
of the three principal moments of inertia are equal, is not capable of performing a
regular precession, having for the axis of precession the vertical and for the figure
axis one of the three principal axes of inertia.
Grioli established, on a purely dynamical basis, the possibility of a regular pre-
cession of the heavy rigid body about a non-vertical axis under certain conditions
on the parameters of the body [138]. Gulyaev derived the full explicit solution of
this case [141] (see, e.g. [256]). We present the necessary details in Gulyaev’s direct
and transparent derivation, which uses principal axes of inertia as the body system.
Another derivation, using general axes, will be given later in dealing with motion of
the rigid body under the action of two or three skew fields.
Let the axes be arranged such that A ≥ B ≥ C, so that we deal with a body for
which the ratios of inertia are confined to the triangle P Q R in Fig. 8.30. Moreover,
assume that the conditions
√ √
x0 B − C = z 0 A − B, y0 = 0 (8.143)
@seismicisolation
@seismicisolation
Fig. 8.28 Views of the trajectory for the value c = 0.62
Fig. 8.29 Trajectories shrink as c approaches c∗ = 0.6469380234
hold for the centre of mass and the principal moments of inertia. These conditions
mean that the centre of mass of the body lies on the perpendicular from the fixed
point O to the plane of one of the circular cross-sections of the ellipsoid of inertia.
Note also that on the side P Q, A = B and from (8.143) x0 = 0. Similarly, on Q R,
we have B = C, z 0 = 0. On both sides, the body becomes axially symmetric with
the centre of mass lying on the axis of symmetry, i.e. the body turns into Lagrange’s
top. On the third side P R of the triangle, B + C = A and the body becomes a disc
@seismicisolation
@seismicisolation
in the yz-plane and thus x0 = 0, which contradicts (8.143). The side P R should be
excluded, and thus an asymmetric body satisfying (8.143) corresponds to points of
the interior of the triangle P Q R in Fig. 8.30.
As was shown by Gulyaev in [141], under conditions (8.143), the system of Eqs.
(3.29) admits the particular solution:

p= (a − c cos(t)), q = sin(t), r = (c + a cos(t)),
s s
2
γ1 = − 2 [Cc cos(t) + (B − C)a sin (t)],
2
s
2
γ2 = 3 sin(t)[(Aa 2 + Cc2 ) − (A − C)ac cos(t)],
s
2
γ3 = 2 [Aa cos(t) + (A − B)c sin2 (t)], (8.144)
s
√
where a = Mgx0 , b = Mgy0 , s = a 2 + c2 , 2 = √ s
. This
2 (A−B+C) +(A−B)(B−C)
solution corresponds to a uniform precession of the body. The angular velocity ω
can be written as the sum of two terms
ω = ζ + α, (8.145)
where
a c c a
ζ = ( , 0, ), α = ( − cos(t), sin(t), cos(t)). (8.146)
s s s s
The first vector ζ is fixed in the body along the configuration axis. This axis carries
the centre of mass of the body and is orthogonal to a circular section of the inertia
ellipsoid. The second vector, α, can be easily shown to have the following properties:
(1) α is orthogonal to ζ. In fact α · ζ = 0.

(2) α is fixed in space. In virtue of (8.146) and (8.145), it satisfies the equation
α̇ + ω × α = 0.
Thus, α is the unit vector along the precession axis. Note that the angular velocities
of the body around the configuration and precession axes are equal (each equals ).
Now, we determine the angle δ of the inclination of the axis of precession to the
upward vertical, which must be a fixed angle. Using the relation cos δ = α · γ, we
find
A− B +C
δ = arccos
(A − B + C)2 + (A − B)(B − C)
√
(A − B)(B − C)
= arctan . (8.147)
A− B +C
@seismicisolation
@seismicisolation
Fig. 8.30 Contours of the

angle δ
Thus, in the final general picture of the motion, the body rotates with the uniform
velocity around the vector ζ fixed in it, while that vector rotates with the same
angular velocity about the direction α orthogonal to ζ but fixed in space and
making with the vertical upwards a fixed angle δ. On the two sides P Q, Q R of the
triangle P Q R in Fig. 8.30, we have δ = 0. As we have seen above, on both sides,
the body becomes Lagrange’s top. As in Lagrange’s case, the figure axis is the axis
of symmetry of the body and the axis of precession is the vertical at the fixed point,
but here we have two additional conditions: the configuration and precession axes
are orthogonal and the velocities of rotation and precession are equal. We recall that
for Lagrange’s top precession, one or the two conditions may not be satisfied.
Figure 8.30 shows the contour lines of equal δ. As seen from it, in the interior
of the triangle P Q R, the inclination angle δ increases downwards and approaches a
limit π2 as the current point approaches P. This means that for asymmetric bodies, for
which A − B and C take very small positive values, the precession axis occupies an
almost horizontal position and the configuration axis precesses around it in a nearly
vertical plane.
8.10.1 Motion of the Centre of Mass
From (8.144) and (8.146), we find the angle, (say), between the vector ζ directed
to the centre of mass and the vertical. We get
@seismicisolation
@seismicisolation
Fig. 8.31 Projection of orbits on the x z-plane of symmetry
cos = ζ · γ

(A − B)(B − C)
= cos(t)
(A − B + C)2 + (A − B)(B − C)
= sin δ cos(t). (8.148)
Thus, the angle varies with time from π/2 − δ to π/2 + δ. The centre of mass
spends equal time intervals above and below the horizontal plane through the fixed
point.
8.10.2 Orbits of Motion on the Poisson Sphere
The picture of the motion can be most clearly interpreted as the motion of a right
circular cone with vertical angle π/2 rolling without slipping on a similar cone fixed
in space. This will be presented later in this book, based on the use of non-principal
system of axes. But it is of interest here to depict some orbits of the motion on the
Poisson sphere and show how the orbit changes with the change of moments of
inertia.
(1) From (8.144), γ1 , γ3 are even functions in t, while γ2 is odd. The orbits are
symmetric with respect to the x z-plane.
(2) At δ = 0, the orbit is a great circle γ1 = 0. For increasing values of δ (see
Fig. 8.30), the orbit remains simple and smooth. When δ reaches π4 , a cusp appears
on the orbit on the plane of symmetry x z. The cusp turns into a small second loop,
which grows with increasing δ.
Figures 8.31 and 8.32 show sample orbits corresponding to moments of inertia
represented by six points in Fig. 8.30 lying on the line drawn from P and bisecting
Q R. The orbits are numbered from 1 to 6 as δ varies from 0 to 79◦ .
@seismicisolation
@seismicisolation
8.11 Dokshevich’s First Case [63] (1966) 175
Fig. 8.32 Orbits of motion on the Poisson sphere a View on opaque sphere b View on transparent
sphere
8.11 Dokshevich’s First Case [63] (1966)
This case bears some resemblance to the case of Hess described in Sect. 8.4. The
centre of mass lies on a circular section of the gyration ellipsoid. It is possible
to investigate this case using the principal set of axes at the fixed point, but, as
presented in Sect. 7.9 of [108], expressions of the Euler–Poisson variables in terms
of a regularizing variable become too complicated.
8.11.1 Use of Special System of Axes
We shall use here the special system of axes fixed in the body and associated with
the gyration ellipsoid. In this system expressions are less complicated, but, most
importantly, the process of separation of variables and the explicit expressions of
the variables in terms of time is more straightforward. Our presentation is a slight
modification of those given in [63, 108].
Let, for determinacy, the centre of mass of the body lie on the x-axis. Let also the
gyration matrix at the fixed point in the chosen axes be written as
⎛ ⎞
ab0
J ≡ I−1 = ⎝b c 0⎠, (8.149)
00c
so that the equation of the gyration ellipsoid becomes
ax 2 + 2bx y + c(y 2 + z 2 ) = 1, (8.150)
@seismicisolation
@seismicisolation
and the plane x = 0 contains the circular cross-section of the ellipsoid. The compo-
nents of the angular velocity are related to the components of the vector of angular
momentum, which we denote by
G = (P, Q, R), (8.151)
by the relations ω = GJ or in expanded form
p = a P + bQ,
q = b P + cQ,
r = c R. (8.152)
Equations of the motion take the form
Ṗ = −b P R,
Q̇ = [(a − c)P + bQ]R + sγ3 ,
Ṙ = b(P 2 − Q 2 ) + (c − a)P Q − sγ2 ,
γ̇1 + (b P + cQ)γ3 − c Rγ2 = 0,
γ̇2 + c Rγ1 − (a P + bQ)γ3 = 0,
γ̇3 + (a P + bQ)γ2 − (b P + cQ)γ1 = 0, (8.153)
where s = Mgx0 and the first integrals of motion become
1
I1 ≡ [a P 2 + 2b P Q + c(Q 2 + R 2 )] + sγ1 = h,
2
I2 ≡ Pγ1 + Qγ2 + Rγ3 = f,
I3 ≡ γ12 + γ22 + γ32 = 1. (8.154)
Dokshevich [63] used the ansatz
γ2 = g0 + g1 P 2 ,
Q2
Q = Q1 P + ,
P
1
R= R0 + R1 P 2 + R2 P 4 , (8.155)
P
and γ1 , γ3 are expressed from the first two integrals of motion in (8.154). The third
integral and the equations of motion are used to obtain conditions on the parameters
of the problem. It can be found that
@seismicisolation
@seismicisolation
8.11 Dokshevich’s First Case [63] (1966) 177
√
s (a − 7/2 c) a 2 − ac + 3 b2 + c2 + a 2 − 4 ac + 6 b2 + c2
h = −2 √ ,
(2 a − 4 c) a 2 − ac + 3 b2 + c2 + 2 a 2 − 5 ac + 12 b2 + 5 c2
f = 0,
1
Q1 = (c − 2a + a 2 − ac + 3b2 + c2 ),
3b
cs
Q2 = ,
b(2bQ 1 + a − c)
cs
R0 = −[ ]2 ,
b(2bQ 1 + a − c)
cs cs (a − c)s(ac − b2 )
R1 = − 2− + 2 ,
b a(2bQ 1 + 2a − c) ab (2bQ 1 + a − c)
a
R2 = −Q 21 − ,
2bQ 1 + 2a − c
c
g0 = − ,
b
(2bQ 1 + a − c)[2bQ 21 + (2a − c)Q 1 − b]
g1 = − . (8.156)
s(2bQ 1 + 2a − c)
(bQ 1 + a)(2bQ 1 + a − c) 2
γ1 = −1 − P ,
s(2bQ 1 + 2a − c)
c (2bQ 1 + a − c)[2bQ 21 + (2a − c)Q 1 − b] 2
γ2 = − − P ,
b s(2bQ 1 + 2a − c)

γ3 = 1 − γ12 − γ22 . (8.157)
From (8.153) and (8.155), we obtain the relation of P and time

Ṗ = −b R0 + R1 P 2 + R2 P 4 , (8.158)
so that P can be expressed as an elliptic function of time. In fact, it can be shown

that [63]
P = P0 dn(ρt, k), (8.159)
where P0 , ρ, k are determined through coefficients R0 , R1 , R2 .
@seismicisolation
@seismicisolation
8.11.2 Orbits on the Poisson Sphere
Recently, the first solution of Dokshevich was isolated on the basis of a property
of the orbits of motion on the Poisson sphere [427]. To this end, we use the orbital
equation, which will be introduced in a later chapter of this book Chap. 9. This
equation, resulting from eliminating dynamical variables using integrals of motion,
is a second-order differential equation connecting two of the components of the vector
γ. From a known solution of the orbital equation, one can construct the corresponding
full solution of the Euler–Poisson equations.
It was shown in [427] that the only solution of the classical problem when the orbit
of motion on the Poisson sphere is a circular section of that sphere is Dokshevich’s
first solution. This agrees with (8.157), from which we see that eliminating P we
obtain a linear relation between γ1 and γ2 .
8.12 The Case of Konosevich and Pozdnyakovich [230, 231]

(1968)
The solution in this case was found in a trial to express it in the form of truncated
trigonometric series in an intermediate variable σ. In the system of principal axes of
inertia at the fixed point, let the centre of mass lies on the x-axis, so that the equations
of motion have the form (8.75) and let a and u denote Mgx0 and CA , respectively.
The Euler–Poisson variables have the following expressions:

a
p= u[2R1 cos(2σ) + 2 cos(σ)
A
3R1
− (3u − 2) (9u − 4) (11u − 4) (u − 1)],
2P1

aC s0
q = −2 (1 − u)(2u − 1)[2R1 sin(2σ) + sin(σ)],
Bu
au 2
r 2 = −2/3 ×
C (2 u − 1) (3 u − 2)2 (9 u − 4) 17 u 2 − 16 u + 4
×{P1 cos (4 σ)
+6 (2 u − 1) (3 u − 2) (9 u − 4) (11 u − 6) R1 cos (3 σ)

+3 (2 u − 1) (3 u − 2) 92 u 2 − 89 u + 22 cos (2 σ)

(2 u − 1) 1329 u 4 − 2915 u 3 + 2290 u 2 − 766 u + 92 cos (σ)
−
(9 u − 4) R1
(2 u − 1) P3
+1/8 }, (8.160)
(3 u − 1) (3 u − 2) (9 u − 4) P1
@seismicisolation
@seismicisolation
8.12 The Case of Konosevich and Pozdnyakovich [230, 231] (1968) 179
2 1−u P1 cos (4 σ)
γ1 = {
3 (3 u − 2) (9 u − 4)
2 2u −1
+3 (3 u − 2) (9 u − 4) (15 u − 8) R 1 cos (3 σ)

+3 (3 u − 2) 35 u 2 − 33 u + 8 cos (2 σ)

1473 u 4 − 3235 u 3 + 2582 u 2 − 890 u + 112
−1/2 cos (σ)
(9 u − 4) R1
P2
+1/8 },
P1
s0 (1 − u)2 (2 u − 1) P1 sin (4 σ)
γ2 = − {
3 (3 u − 2)2 (9 u − 4) (2 u − 1)2
+36 (3 u − 2) (9 u − 4) R 1 sin (3 σ)
+3 (3 u − 2) (13 u − 8) sin (2 σ)

309 u 3 − 464 u 2 + 222 u − 34 sin (σ)
−2 },
(9 u − 4) R1
2 A(1 − u)
γ3 = − r [(9u − 4)R1 cos(2σ) + (3u − 1) cos(σ)
c1
u(1 − u)(11u − 4)
+ ], (8.161)
4(3u − 2)(9u − 4)R1
in which
P1 = 97 u 4 − 271 u 3 + 258 u 2 − 101 u + 14,

P2 = 729185 u 8 − 3592714 u 7 + 7498373 u 6 − 8683880 u 5
+6116476 u 4 − 2687840 u 3 + 720872 u 2 − 108064 u + 6944,
P3 = 334695915 u 11 − 2239025631 u 10 + 6709632635 u 9
−11904287977 u 8 + 13908377054 u 7 − 11245436972 u 6
+6425316760 u 5 − 2596052504 u 4 + 727305680 u 3
−134634208 u 2 + 14828480 u − 736512, (8.162)
4u
s02 = ,
(1 − u)(17u 2 − 16u + 4)
P1
R12 = ,
3(3u − 2)2 (9u − 4)2
Ca
c12 = , (8.163)

@seismicisolation
@seismicisolation
and
(5 u − 2) (u − 1)3
2 = 2 ×
144 (3 u − 2)5 (9 u − 4)2 (2 u − 1)2 97 u 4 − 271 u 3 + 258 u 2 − 101 u + 14 (3 u − 1)
×(3828317502420 u 18 − 45928438243128 u 17 + 256746575766313 u 16
−888956089381273 u 15 + 2137100780408016 u 14 − 3789579706038612 u 13
+5138532374554515 u 12 − 5449736202969735 u 11 + 4584313633529960 u 10
−3082573845813260 u 9 + 1661356742951664 u 8 − 716191818324608 u 7
+245195273873024 u 6 − 65783181186816 u 5 + 13530394373376 u 4
−2059297120512 u 3 + 218459750400 u 2 − 14419084288 u + 445763584). (8.164)
Formulas (8.160)–(8.164) are slight modifications of the presentation in [108].

Using the first of dynamical Eqs. (8.75), the following equation is obtained for
the intermediate variable

1 (1 − u)(17u 2 − 16u + 4)
σ̇ = r. (8.165)
2 u
The moments of inertia of the body are subject to two conditions: a simple one
2C − A
B = 4A , (8.166)
17C − 8A
and the second is that u must be a root of the seventh-degree polynomial equation
P7 = 437511u 7 − 1822945u 6 + 3227896u 5 − 3146990u 4

+1823596u 3 − 627920u 2 + 118960u − 9568
= 0. (8.167)
Although this equation has three real roots, only two roots lead to real expressions for
r and , and hence for all (8.160)–(8.161), provided that the variable σ is restricted
for each root to certain interval [σ1 , σ2 ].
However, we cannot go further with the exact solution and must use approximate
values of the roots u 1 = 0.4119045665, u 2 = 0.7081863612, and hence replace the
whole solution by a one with approximate numerical coefficients. From now on, the
case under consideration splits into two different subcases, which differ not only in
the value of u, but also in the descriptionof motion accordingly.
It is notable here that only the factor aA figures in the components of ω, while
γ contains no parameters. This means we are dealing with a single orbit of the apex
on the Poisson sphere corresponding to each of the roots u 1 , u 2 . Orbits are shown
for the two subcases, together with the necessary values of the parameters.
@seismicisolation
@seismicisolation
8.12 The Case of Konosevich and Pozdnyakovich [230, 231] (1968) 181
8.12.1 The First Subcase
u = u 1 = 0.4119045665,
A : B : C :: 1 : 0.7064431289 : 0.4119045665,
= 0.029056173,
σ1 = 2.7431083805,
σ2 = 2π − σ1 . (8.168)
At the ends of this interval, r and γ3 change their signs and hence the motion is
periodic time.
The orbit is symmetric with respect to the x z-plane and intersects it at σ = π.

The projection shows clearly that the motion is a vibration, in which the apex of the
vertical draws the orbit from one end to the other and then σ̇ changes its sign, and
the orbit is described in the reverse direction. The motion is periodic in time. The
periodic time T is obtained from (8.165) as
σ2
u dσ
T =4
(1 − u)(17u − 16u + 4) σ1 r
2
π
u dσ
=8 . (8.169)
(1 − u)(17u − 16u + 4) σ1 r
2
8.12.2 The Second Subcase
u = u 2 = 0.7081863612,
A : B : C :: 1 : 0.4123351224 : 0.7081863612,
@seismicisolation
@seismicisolation
= 14.9952481138,
σ1 = 1.0117323905,
σ2 = 2π − σ1 .
The orbit is symmetric with respect to the x z-plane and intersects it at σ = π. The
projection shows clearly that the motion is also a vibration, in which the apex of the
vertical draws the orbit from one end to the other and then σ̇ changes its sign, and
the orbit is described in the reverse direction. The motion is periodic and the period
is given by the same formula as (8.169).
8.13 Dokshevich’s Second Case [64] (1970)
This is the last case discovered until now in the classical problem of motion of a
heavy rigid body about a fixed point. As all the cases discovered are not obtained
on an exhaustive basis, it is not known whether some new cases of solvability of the
Euler–Poisson equations for the classical problem in the near (or even distant) future
to be found.
Under the condition assumed in most of the particular cases treated above, the
centre of mass of the body is assumed to lie on one of the principal axes of inertia
of the body at the fixed body. The equations of motion are the same as (8.75)–(8.76)
and their first integrals have the form (8.77).
To obtain this solution, assume the components of the angular velocity in the form

p= a2 (u 2 − u 21 ),

q = b2 (u 2 + b1 u + b0 ),

r = u c1 (u − u 3 ). (8.170)
@seismicisolation
@seismicisolation
8.13 Dokshevich’s Second Case [64] (1970) 183
The first component of the vector γ is determined from the energy integral in (8.77)
1 1
γ1 = [h − (Ap 2 + Bq 2 + Cr 2 )], (8.171)
a 2
and the other two from third and second dynamical Eqs. (8.75)
1
γ2 = − [C ṙ + (B − A) pq],
a
1
γ3 = [B q̇ + (A − C) pr ]. (8.172)
a
The first dynamical equation is used to express u̇ as
(B − C)
u̇ = pqr
Aa2 u

(B − C)
= a2 b2 c1 (u 2 − u 21 )(u 2 + b1 u + b0 )(u − u 3 ). (8.173)
Aa2
Substituting the above expressions into the second and third integrals of motion
I2 ≡ Apγ1 + Bqγ2 + Cr γ3 = f,
I3 ≡ γ12 + γ22 + γ32 = 1, (8.174)
we obtain two polynomials in u whose coefficients must be all zeros. From those
equations, it turns out that the moments of inertia are related by the same relation as
in the case of Kowalewski
A(10B − 9C) = 18C(B − C),
and one more relation

C C
192( )2 − 184 + 41 = 0.
A A
Together, those relations give
√
649 + 71 37
B= A = 1. 274 6A,
848
√
23 + 37
C= A = 0.605 89A. (8.175)
48
Without loss of generality, we can normalize the moments of inertia so that A = 1,
and no free parameters remain in the solution. We also find the parameters
@seismicisolation
@seismicisolation
√
7 + 11 37
f = 0, a = ,
256
√ √
(2816 37 − 4411)(67801 + 11975 37)
h= = 2.955473898. (8.176)
5366592
The final expressions for the Euler–Poisson variables can be written as follows:

√
N 9327212 + 1782163 37 √
p= √ 9 u 2 − 17 − 8 37
13 068 231

√
N 9327212 + 1782163 37
= √ u 2 − u 21 ,
4356 231

√
N 1635292 − 474787 37 √ √
q= √ 36 u 2 − 3 25 + 7 37 u − 352 − 64 37
13 068 231

√
N 1635292 − 474787 37
= √ (u − u 4 )(u − u 5 ),
2178 231

√
N 1853 − 4463 37 √
r = √ u 4 u − 7 − 37
1452 21

√
N 1853 − 4463 37
√
= √ u u − u3, (8.177)
1452 21
and
√ √ √
9[24u 3 − 12(5 +37)u 2 − 3(17 + 5 37)u + 4(127 + 19 37)]
γ1 = √ √ ,
(1 + 37) 2816 37 − 4411
√
176(12u − 31 − 37)
γ2 = √ pq,
1649 + 61 37
√ √
2(53 37 − 235)(4u + 7 + 37) r
γ3 = p , (8.178)
1107 u
where
√
4
N = 11(256 37 − 401),

1 √
u1 = 17 + 8 37 = 2. 701 1,
3
@seismicisolation
@seismicisolation
Fig. 8.33 Graphs of

u̇, p, q, r

1 √
u2 = − 17 + 8 37 = −2. 701 1,
3
1 √
u 3 = (7 + 37) = 3. 270 7,
4
1 √ √
u4 = (25 + 7 37 + 8070 + 1374 37) = 8. 156 2,
24
1 √ √
u5 = (25 + 7 37 − 8070 + 1374 37) = −2. 524 6. (8.179)
24
From Eq. (8.173), the time can be expressed as a hyper-elliptic integral. Comparing
numerical values of the roots u i , we easily see that they have the order
u4 > u3 > u1 > u5 > u2. (8.180)
Graphical check (Fig. 8.33) shows that p, q, r, u̇ are all real only on the interval
u 1 ≤ u ≤ u 3 . Moreover, the q component of the angular velocity does not change
its sign on the whole motion.
To complete the description of motion, we introduce an auxiliary angle φ by the
substitution
u = u 1 + (u 3 − u 1 ) sin φ. (8.181)
Now, the angular velocities p, q, r are expressed as analytical functions of φ.
@seismicisolation
@seismicisolation
√ √ √ √
113/4 21 2 9327212 + 1782163 37 256 37 − 401
4
p= ×
12074832

√ √ √ √
× 251 − 37 − [523 + 127 37 − 12(7 + 37) 8 37 + 17] cos2 φ sin φ,
√
21 · 22(474787 37 − 1635292) √
q= δ1 cos4 φ + δ2 cos2 φ + 176 37 + 968,
6037416
√ √ √ √
N 7 −1853 + 4463 37(21 + 3 37 − 4 8 37 + 17)3/2
r = ×
639966096
√ √ √
×[2763 + 450 37 + (43 37 + 299) 8 37 + 17 − 1749 cos2 φ] cos(8.182)
φ,
where

√ √ √
δ1 = (12 37 + 84 ) 8 37 + 17 − 523 − 127 37,

√ √ √
δ2 = 2 ( 37 − 17 ) 8 37 + 17 + 123 + 15 37,
and the components of γ are changed accordingly.
8.13.1 Periodicity of the Motion
From (8.173), we see that the time can be expressed as a hyper-elliptic integral in the
auxiliary variable u. However, it turns out that the angle φ is more suited to complete
the description of motion. In fact, using (8.173) and (8.181), one can find

φ̇ = K P2 P4 ,
where
√
√ √
25549 37 + 116399 8 37 + 17 + 84(23 37 + 21514)
K = √ ,
14 256 8162

√ √ √ √
P2 = 251 − 37 − [523 + 127 37 − 12(7 + 37) 8 37 + 17] cos2 φ,
√
P4 = δ1 cos4 φ + δ2 cos2 φ + 176 37 + 968. (8.183)
It is evident that P2 and P4 have no real zeros. Thus, φ̇ is positive for all φ, so that φ
monotonically increases with time. The motion of the body is periodic with period
2π in φ. The time elapsed in motion from the initial point P (φ = 0) to a general
position φ is
@seismicisolation
@seismicisolation
Fig. 8.34 Planes of symmetry
Fig. 8.35 Two sides of the sphere
φ
1 dφ
t= √ ,
K 0 P2 P4
and the time period is

2π
1 dφ
T = √ .
K 0 P2 P4
8.13.2 Orbits on the Poisson Sphere
The present case bears some similarity to Goryachev’s case above. Graphics reveal
that the orbit of motion on the Poisson sphere is three-loop curve with the two planes
x y and x z as planes of symmetry. Two smaller loops have a common vertex P at
the point (1, 0, 0) attained periodically at φ = 0, π, 2π and so on. At φ = 2π, after
@seismicisolation
@seismicisolation
Fig. 8.36 The whole orbit. Hidden lines dashed
making two complete rotations in the same direction about the y-axis, the point
returns to P and the orbit closes (Figs. 8.34, 8.35 and 8.36).
8.14 Unsuccessful Cases and Incorrect Claims
Most of the particular solutions of the problem of motion of a heavy rigid body
about a fixed point were constructed in the last decade of the nineteenth century and
the first decade of the twentieth century. The first successful trial to build the next
particular solution was that made by Grioli in 1947. Along the period between those
times, several publications appeared announcing new solutions. However, it turned
out later that those publications carried wrong claims. Some of them are repeated or
were special cases of known results discovered earlier by other authors. Other ones
were erroneous and contained no solution at all. Erroneous solutions continued to
appear in some publications until the present time, e.g. the very recent publications
[76, 98].
In view of the fact that some of those solutions were presented in books on the
subject as valid solutions, e.g. [256, 270], we shall give here a brief account of
the ones that were published in English or used and cited in works in English. We
comment on their shortcomings and give the references, where the reader can find
more details about them. Some more incorrect results appeared in Russian, but had
no further implications. Those are not listed here. An example is Arjanikh’s claim
that the classical problem of heavy rigid body dynamics in all its generality admits
an integral of motion linear in velocities different from the areas integral [13].
@seismicisolation
@seismicisolation
8.14 Unsuccessful Cases and Incorrect Claims 189
8.14.1 Shiff’s Work [331] (1903)
Shiff’s short article [331] first introduces a transformation of the equations of motion
of a heavy rigid body about a fixed point to a certain form, which can be easily
identified with that obtained much earlier by Hess [150]. Based on those equations,
he introduces a particular solution of those equations, claiming it to be a new solution.
Hamel [147] noted that the investigations of Shiff, Stäckel and Hess, who used
equations of Hess’ type, are defective. They ignored Poisson’s equations, while the
question about equivalence of the equations they used and the original Euler–Poisson
equations was not addressed.
On the other hand, the solution introduced by Shiff satisfies the equality G2 =
const, and, in the light of Sect. 8, the check of Poisson’s equations reveals that two
of them are not satisfied.
8.14.2 Field’s Works [85, 86] (1934)
The main results of those two articles are not erroneous, but the case pointed out in
them is a special case of Steklov’s case presented in Sect. 8.6, under the additional
restriction on the moments of inertia A2 = 2BC. On the other hand, Field pointed
out an interesting property of that special version, namely the equality of the angles
of precession and proper rotation.
8.14.3 Corliss’ Works [56, 57] (1932–1934)
In the first article, Corliss announces two particular solutions of the classical problem.
The first reproduces one version of Steklov’s solution, although Steklov’s paper [347]
is cited in it. The second is a special case of Kowalevski’s general integrable case.
This also is not a new result, as it describes the class of motions of Kowalevski’s
gyroscope described much earlier by Appelrot under the name of “The second class of
the simplest motions of Kowalevski’s gyroscope” [11]. In his second article, Corliss
introduces one more particular solution, this time repeating Goryachev’s result [114].
8.14.4 Fabbri’s Works [80, 81] (1934)
Fabbri’s result in [80] also repeats Steklov’s. In the case introduced in [81], some of
the variables are not real-valued [120].
@seismicisolation
@seismicisolation
8.14.5 Concerning Mertsalov’s Work [283] (1946)
A case is included in Leimanis’ book [256] (Sect. 8.7) and listed in the table of
exactly solvable cases in Magnus’ book [270] (Sect. 3.3.1), under the name “Case
of Mertsalov”. This is a particular solution that could generalize in some sense the
conditional integrable case of Goryachev and Chaplygin A = B = 4C (Sect. 4.4)
by removing the restriction on the areas integral and imposing another condition on
the initial velocity. An integral supposed to generalize (4.68) is provided. However,
no exact solution is included in Mertsalov’s paper, which is devoted to building an
approximate solution. In fact, Mertsalov did not claim obtaining an exact solution in
his work.
8.14.6 Gao’s Work [98] (2003)
The purpose of this paper is to present two results that extend the famous integrable
case due to Kowalevski of motion of a heavy rigid body about a fixed point. Unfor-
tunately, one of these results is correct but not new. The other one is new but not
correct [282].
Case 1. As per her original work, Kowalevski’s case is characterized by the con-
ditions A = B = 2C, z 0 = 0, i.e. the centre of mass lies in the equatorial plane of
the inertia ellipsoid of revolution. As pointed out by Kowalevski, one can rotate the
x, y-axes in this plane, and they will remain as principal axes, so as to have the centre
of mass of the body on the x-axis and, equivalently, make y0 = 0. This is done in
most works on the subject, including Kowalevski’s original work [238] of 1888 and
classical textbooks like the ones by Whittaker and Leimanis. In the case introduced
here y0 is not necessarily zero. It is in no way an “extension” or “generalization” of
Kowalevski’s result. In fact, It is just an unneeded complication of the choice of the
body reference system.
Case 2. This case, characterized by the conditions A = B = 2C, y0 = 0, x0 z 0 =
0, is in conflict with the necessary conditions obtained by Husson [154] in 1906 for
the Euler–Poisson equations to have a fourth algebraic integral and cannot be true in
principle. Moreover, it is easy to find that the expression H4 given by the author can
be an integral in only two cases:
(a) z 0 = 0, and this gives Kowalevski’s case;
(b) x0 = 0, and this case can be easily identified as a special case of Lagrange’s
top.
It is noteworthy that the same incorrect generalization as Case 2 is met much
earlier in the classical monograph of Hagihara on celestial mechanics [144].
@seismicisolation
@seismicisolation
8.14 Unsuccessful Cases and Incorrect Claims 191
8.14.7 The Work of Yanxia and Keying [371] (2005)
Here is another claim to a generalization of Kowalevski’s case by assuming y0 = 0,

which is merely referring the centre of mass to another coordinate system, rotated
about the z-axis by some fixed angle. That is the same as Case 1 of Gao’s work.
8.14.8 Ershkov’s Work [76] (2014)
In his article, Ershkov announced a new exact solution of the classical problem,
which, as the author states, generalizes the famous case of motion by inertia, found
originally by Euler. Some details were further treated in [77]. In their comment
[329], Sanduleanu and Petrov showed inconsistency of this result. To satisfy all the
equations of motion, it reduces to a subcase of Euler’s case.
The source of error in Ershkov’s work is that he relied on the dynamical equations
with the three integrals of motion and completely ignored Poisson’s equations. He
also assumed that G2 = G 20 (const). As shown in Sect. 8.1, under this condition, the
three integrals of motion cannot replace Poisson’s equation. It is necessary to check
directly Poisson’s equation on the given solution. Doing that, we find that on the
Ershkov solution, only one of three Poisson’s equations is satisfied.
8.15 Particular Solutions in the Problem of Motion of a

Heavy Gyrostat
At present, fourteen particular solutions of the equations of motion of a heavy gyrostat

are known, nine of which are generalizations of their counterparts in the classical
problem of motion of a heavy simple rigid body. The other five have no analog as the
gyrostatic momentum vanishes. However, the presence of the gyrostatic terms brings
significant complications to the solution, especially to the process of separation of
variables. Therefore, cases that were expressed in terms of elliptic functions are not
brought to the same final state of the solution, due to complicated quadratures.
Here we give a brief note about each of the known cases. For clarity of information
and for the sake of comparing the original and generalized cases, we preserve the
order of those cases as in the classical problem. For each case, we give authors’ names,
references and the conditions on the gyrostatic momentum κ. For full information
about those cases, the reader is referred to the original works or one of the books,
which bear review character, devoted to the motion of the gyrostat [121, 125].
@seismicisolation
@seismicisolation
8.15.1 Generalizations of Particular Cases Known in the

Classical Problem
8.15.1.1 Generalization of Planar (Pendulum) Motion
The generalization is quite simple by adding a gyrostatic moment along the z-axis,
the axis of rotation of the pendulum. Conditions on the system parameters are
κ1 = κ2 = z 0 = 0. (8.184)
The solution is not affected by the presence of the gyrostatic moment. It is still
given by formulas of Sect. 8.2. The motion is described and its orbital stability is
investigated in [388, 420]. It turns out that the gyrostatic momentum plays a decisive
effect on the stability of the motion.
8.15.1.2 Stationary Motion (Permanent Rotations) of the Gyrostat
(A) Permanent rotations of the gyrostat in Joukovsky–Volterra’s case (The balanced

gyrostat). In the case of absence of the gravity field (or in case of a gyrostat fixed
from its centre of mass), permanent rotations satisfy the equations
ω × (ωI + κ) = 0, (8.185)
γ̇ + ω × γ = 0. (8.186)
The first equation is completely independent of γ and determines possible values of

the vector ω, constant in the body and in the space systems. Having obtained ω, the
second equation can be solved for γ.
Let us first multiply (8.185) scalarly by κ. We get
κ · (ω × ωI) = 0. (8.187)
This means that in all possible permanent rotations, the angular velocity lies along a
generator of a quadratic cone
κ1 (C − B)qr + κ2 (A − C) pr + κ3 (B − A) pq = 0. (8.188)
This cone passes through the points κ, κI−1 and the three principal axes of inertia of
the gyrostat. It is analogous to Staude’s cone (8.31) for the permanent rotations of a
simple heavy body with the gyrostatic momentum κ playing the role of the position
vector of the centre of mass and ω the role of the vertical unit vector γ.
Dynamical Eq. (8.185) can be put in the parametric form
@seismicisolation
@seismicisolation
8.15 Particular Solutions in the Problem of Motion of a Heavy Gyrostat 193
Ap + κ1 Bq + κ2 Cr + κ3
= = = s, (say)
p q r
so that one can write

κ1 κ2 κ3
p= ,q = ,r = . (8.189)
s−A s−B s−C
This parametrization plays important role in the explicit time solution of the equations
of motion of the balanced gyrostat [369].
We note also that, as in case of Staude’s rotation, the angular velocity about an
axis of permanent rotation grows indefinitely as this axis approaches any one of the
principal axes of the gyrostat. Stability analysis of permanent rotations of the free
gyrostat can be found in [369].
(B) Permanent rotations of the heavy gyrostat. This motion is characterized by
the constancy of the two vectors ω, γ in the body axes. Setting ω̇ = 0, γ̇ = 0, in
(5.3), (5.4), we get
ω × (ωI + κ) = Mgγ × r0 , ω × γ = 0. (8.190)
From the second equation, we can write, as in the case of Staude’s motion (See Sect.
8.3), ω = ω0 γ, so that the dynamical equation gives
γ × (ω02 γI + ω0 κ − Mgr0 ) = 0. (8.191)
This is the condition that the line along the vector γ intersects at right angle the
ellipsoid
1
Φ = ω02 γI · γ + (ω0 κ − Mgr0 ) · γ = const. (8.192)
2
For generic values of the parameters, the centre of this ellipsoid is displaced from
the origin and only two solutions of (8.191) are guaranteed for arbitrary ω0 .
Let a given vector γ determine a direction in the main body of an axis of permanent
rotations. Multiplying (8.191) scalarly by κ and by r0 , we obtain two expressions
for ω0 :
r0 · (γ × γI) γ · (r0 × κ)
ω0 = Mg = . (8.193)
κ · (γ × γI) r0 · (γ × γI)
Those expressions are compatible only on the fourth-degree surface
Mg[r0 · (γ × γI)]2 − [γ · (r0 × κ)][κ · (γ × γI)] = 0. (8.194)
This surface intersects the Poisson sphere γ 2 = 1 in a curve. The generators of the
cone formed by the axes of permanent rotations are drawn by connecting the origin
to the points of that curve.
@seismicisolation
@seismicisolation
A notable degeneration of (8.193) and (8.194) occurs when κ is collinear with r0 .

In that case, if we choose the uniform angular velocity ω0 such that
ω0 κ = Mgr0 , (8.195)
then (8.191) reduces to

γ × γI = 0. (8.196)
Stationary rotation is possible about any of the three principal axes of inertia, which
would be directed vertically upwards. The angular velocity ω0 has positive or negative
sign, according to whether κ and r0 are parallel or anti-parallel, so that only three
different motions of this type are possible.
8.15.1.3 Generalization of Hess’ Case (Sretensky [340, 341] 1963)
When y0 = 0, κ2 = 0, under Hess’ condition (8.36)

x0 A(B − C) = ±z 0 C(A − B),
equations of motion (5.5)–(5.6) admit the invariant relation
[(A − B) p + κ1 ]x0 + [(B − C)r − κ3 ]z 0 = 0
which generalizes that of Hess by the presence of the components of the gyrostatic
momentum. It is possible in that case to use the variables 1 , 3 from (8.40) and
obtain instead of (8.45) the relation
3
du
t= ,
2B(1 − u 2 )(h − su) − [ f + ( κA−B
1 cos δ
− κ2 sin δ
B−C
)u]2
so that the centre of mass of the body performs with respect to vertical a motion
like a spherical pendulum. A separable case analogous to the one of Hess’ case and
detailed analysis of the motion are given in [121] using the special axes attached to
the gyration ellipsoid.
8.15.1.4 Generalization of the Bobylev–Steklov Case (Kharlamov [193]

1964)
This case is characterized by the conditions
κ = (κ1 , κ2 , κ3 ),
@seismicisolation
@seismicisolation
(2C − A)κ2 x0 = (2C − B)κ1 y0 , z 0 = 0, (8.197)
and the invariant relations

p = λx0 , q = λy0 . (8.198)
As in the Bobylev–Steklov case, the solution is expressible in elliptic functions of

time.
In a slightly earlier paper [167], Keis found a gyrostat generalization of the
Bobylev–Steklov case. His result may be obtained from (8.197) by putting κ2 =
y0 = 0.
8.15.1.5 Generalization of Steklov’s and Kowalewski’s Cases

(Kharlamov [196] 1965)
In [196], Kharlamov investigated the possibility to generalize the invariant relations

(8.80) and (8.123), which express q and r in terms of p, under the simplifying
restriction that the gyrostatic moment is directed along the first coordinate axis x
carrying the centre of mass of the body. In that case, one can write
κ = (κ, 0, 0). (8.199)
It turned out that both cases accommodate such gyrostatic moment. From formu-
las provided in [195, 196], it follows that the Euler–Poisson variables can still be
expressed in elliptic functions of time in the generalization of Steklov’s solution.
However, this step was not performed due to the complication due to the presence
of terms linear in p in the two quadratic factors generalizing (8.81).
In the generalization of Kowalewski’s case, the relation of p and time is still
determined by a hyper-elliptic quadrature [195] (See also [125]).
8.15.1.6 Generalizations of Grioli’s Case (Keis 1965, Kharlamova 1969)
Keis (1963) [167] (see also [169]) discussed the existence of regular precession in
the motion of a gyrostat about a fixed point under the action of gravity. It turned
out that Grioli’s case admits generalization, in which the gyrostatic momentum is
directed parallel to the line from the fixed point to the centre of mass of the body.
Kharlamova obtained the same generalization of Grioli’s case in non-principal axes
[208] in a search for linear invariant relation. Detailed explicit analysis of the regular
precessional motion of a heavy gyrostat was again performed by Gulyaev using
principal axes of inertia in [142].
@seismicisolation
@seismicisolation
8.15.1.7 Generalization of Dokshevich’s Second Case (Kharlamova

1971)
In [209], Kharlamova constructed a new exact particular solution under the same
condition on the distribution of mass that characterizes the body in Hess’ case,
i.e. the centre of mass lies on the orthogonal to the circular cross-section of the
gyration ellipsoid. This case differs from that of Sretensky in that the gyrostatic
momentum lies in the principal plane of inertia containing the major and minor axes.
The Euler–Poisson variables were expressed in terms of an auxiliary variable, which
may be expressed as an elliptic function of time. Formulas provided in [209] are quite
complicated. The same case was later reconsidered by Kharlamov, who made some
simplification of the conditions on the parameters and established that the solution
has four different versions [199].
8.15.1.8 Generalization of Konosevich–Pozdnyakovich’s Case

(Kharlamova and Mozalevskaya [212] 1986)
In the book [212], Kharlamova and Mozalevskaya give three particular solutions
constructed using invariant relations of the type used in the case of Konosevich–
Pozdnyakovich. One of those cases contains Konosevich–Pozdnyakovich’s case as
a special case and the other two are different and have no analog in the classical
problem. However, those cases are quite complicated and as presented in [212] are
not given in the complete and verifiable form. The same content is copied in the two
recent books [121, 125] without further clarification. The reader may consult the
mentioned books for more information.
8.15.2 Solvable Cases of the Gyrostat, Having No Classical

Analog
8.15.2.1 Case of Kharlamov and Kovalev [202] (1970) and of

Kharlamova and Kharlamov [210] (1969)
This case is valid under the restriction on the principal moments of inertia
B −C
A = 18C , (8.200)
10B − 9C
i.e. under the same condition as Kowalewski’s case in the classical problem and its
generalization Sect. 8.15.1.4. The centre of mass lies on the x-axis and the gyrostatic
momentum is directed along the y-axis, i.e.
r0 = (x0 , 0, 0), κ = (0, κ, 0). (8.201)
@seismicisolation
@seismicisolation
The invariant relations are taken in the form p 2 = P2 (q), r 2 = P3 (q), where Pi is
a polynomial of degree i. The variables of the problem are expressed in terms of q,
while q is related to time by an elliptic quadrature. A somewhat simpler parametriza-
tion is given but explicit inversion of this quadrature is not performed. More detailed
presentation is given in [202].
It is notable that in the present solution the gyrostatic momentum κ figures in
denominators in most formulas, so that the solution loses its meaning as κ tends to
zero.
A generalized version of this solution due to Kharlamov and Kharlamova was
published in [210] (See also [212]). In that case, one more component of the gyrostatic
moment is added to the body along the z-direction.
8.15.2.2 Case of Mozalevskaya [291] (1970)
This case is valid for a body whose moments of inertia do not satisfy one of the
triangle inequality. Thus, it is of no physical significance neither for the problem of
motion of a gyrostat nor for the classical problem. It is noted that such a case has
its physical interpretation in the frame of the problem of motion of a rigid body in a
liquid, in which the triangle inequalities are not necessary.
8.15.2.3 Case of Dokshevich [65] (1970)
This case is valid under the conditions
B = C,
y0 = z 0 = 0,
κ = (κ1 , κ2 , 0), (8.202)
where κ2 = 0. Invariant relations are sought in the form q = P2 , r 2 = P4 , Pi is a

polynomial of degree i. It was shown in [65] that the solution can be expressed
in elliptic functions of time. The parameter κ2 occurs in denominators in several
formulas, so that the solution loses its meaning when κ2 = 0.
8.15.2.4 Two Cases of Kharlamova and Mozalevskaya [212] 1986
Together with the generalization of Konosevich–Pozdnyakovich’s case (item No. 8

in the previous subsection), the authors found two similar other cases. The solution is
given by relations of a higher level of complexity. Moreover, they lose their meaning
when the gyrostatic moment vanishes.
@seismicisolation
@seismicisolation
8.15.3 Known Cases of the Classical Problem, Which Are not

Presently Generalized to the Gyrostat Problem
Three cases are left at present without generalization by adding a gyrostatic moment.
Those are cases due to Goryachev, Chaplygin and Dokshevich’s first case.
@seismicisolation
@seismicisolation
Chapter 9
The Rigid Body in a Potential Field
In this chapter, we present some results which can not only be applied to the classical
problem of motion of a heavy rigid body about a fixed point, but can also be easily
extended to cases of motion of a rigid body acted upon by general forces which admit
symmetry about an axis fixed in space and passing through the fixed point. For such
fields, the potential is a function of the Eulerian angles θ, ϕ, and the precession angle
is a cyclic coordinate. The areas integral is preserved. The study of the Routhian
reduction of the problem opens some possibilities for deeper analysis of properties
of motion.
9.1 The Routhian in γ1 , γ2 , γ3 as Redundant Variables
Reduced equations of motion in their form Chap. 3 (3.57) are not usually convenient
in use because of lacking symmetry. It is possible to give the Routhian a more
symmetric form that allows to go further in applications, by using the components
of the vector γ. To this end, one can use a direct substitution
γ1
θ = cos−1 γ3 , ϕ = tan−1 (9.1)
γ2
in the formulas Chap. 3 (3.55 and 3.56) and making use of the constraint
γ1 γ̇1 + γ2 γ̇2 + γ3 γ̇3 = 0,
resulting from the differentiation of the geometric integral. However, we shall use a
less direct, but more effective procedure as follows.

https://doi.org/10.1007/978-3-030-96336-1_9
@seismicisolation
@seismicisolation
200 9 The Rigid Body in a Potential Field
9.1.1 Expression of ω in Terms of γ and γ̇
Multiplying Poisson’s equation (3.14) vectorially by γI and using the areas integral
(3.22), one gets
f γ + γ̇ × γI
ω= , (9.2)
D
where D is given by, the same as in (3.48),
D = γI · γ=Aγ12 + Bγ22 + Cγ32 . (9.3)
Kinetic energy in the redundant coordinates:

Using (9.2), one can write
1
T = ωI · ω
2
1
= ωI·[ f γ + γ̇ × γI]
2D
1
= [ f 2 +γ̇·(γI × ωI)],
2D
and using an identity (A) from appendix, we rewrite the last expression as
1
T = [ f 2 +γ̇(ad j (I))·(γ × !)]
2D
1
= [ f 2 +γ̇(ad j (I))·γ̇]
2D
1
= [ f 2 +ABC γ̇I−1 ·γ̇]. (9.4)
2D
The Routhian in the redundant coordinates γ1 , γ2 , γ3
Writing from (3.55) and using (9.4) and (9.1), one obtains
R = L − f ψ̇
= T − Mgr0 · γ − f ψ̇
1
= [ f 2 +ABC γ̇I−1 ·γ̇] − Mgr0 · γ
2D
1
− f. [ f − (A − B) sin θ sin ϕ cos ϕθ̇ − C cos θϕ̇]
D
ABC −1
= γ̇I ·γ̇ + f [(A − B) sin θ sin ϕ cos ϕθ̇ + C cos θϕ̇]
2D
f2
−[V + ]
2D
and we finally have, in the principal axes of inertia of the body at the fixed point,
@seismicisolation
@seismicisolation
9.1 The Routhian in γ1 , γ2 , γ3 as Redundant Variables 201
ABC γ̇12 γ̇ 2 γ̇ 2
R= ( + 2 + 3)
2D A B C
f
+ [Cγ3 (γ2 γ̇1 − γ1 γ̇2 ) − (A − B)γ1 γ2 γ̇3 ]
D(1 − γ32 )
−V1 , (9.5)
2
where V1 = [V + 2Df
] and D is given by (9.3).
Equations of motion can now be deduced in the form of three second-order equa-
tions involving a Lagrangian multiplier λ(t)—say, resulting from the holonomic
constraint
γ 2 − 1 = 0, (9.6)
which should be added to the Routhian R. Solving the resulting equations is equiv-
alent to the determination of the Eulerian angles θ, ϕ, and then one can substitute
this solution into (9.2) to get an expression for the angular velocity ω. One can also
substitute the solution into (3.54) and integrate it with respect to time to complete
the determination of the precession angle ψ.
However, we shall not write these equations down, since they will not play an
important role in the sequel. The real significance of the Routhian (9.5) (or, in other
words, the Lagrangian of the reduced system) will be elucidated in the following
sections, where it will be transformed into various more flexible forms that lead to
significant consequences.
Remark: It should be noted that after solving the Routhian equations of motion,
one can obtain the cyclic angle ψ by integrating the formula
∂R
ψ̇ = −
∂f
1 1
= {f + [Cγ3 (γ1 γ̇2 − γ2 γ̇1 ) + (A − B)γ1 γ2 γ̇3 ]}.
Aγ12 + Bγ2 + Cγ3
2 2
1 − γ32
(9.7)
9.1.2 The Case of Complete Dynamical Symmetry of the Body
When B = C = A, the Routhian (9.5) takes the form
1
R= A(γ̇12 + γ̇22 + γ̇32 )
2
f Aγ3 f2
+ (γ 2 γ̇1 − γ 1 γ̇2 ) − V − . (9.8)
1 − γ32 2A
The last term in this Routhian is now a constant and can be discarded from the
potential. But one has to remember that ψ̇ is still determined by (9.7), which for the
@seismicisolation
@seismicisolation
case of spherical dynamical symmetry takes the form
Aγ3 f
ψ̇ = (γ2 γ̇1 − γ1 γ̇2 ) − . (9.9)
1 − γ32 A
The Routhian (9.8) may be interpreted as describing the problem of motion of a

particle on the Poisson sphere
γ12 + γ22 + γ32 = 1
under the action of forces with potential V and certain gyroscopic forces. This anal-
ogy will be used more than once later in this book.
9.2 Maximal Reduction of Order of the Differential

Equations of Motion of a Rigid Body About a Fixed
Point
9.2.1 Reduction
The six Euler–Poisson first-order differential equations describing the classical prob-
lem of motion of a rigid body about a fixed point admit three first integrals. All the
equations and integrals are time-independent. This situation lead several authors,
along nearly a whole century since 1890, to consider using the three integrals to
eliminate three of the Euler–Poisson variables and thus reduce the system to three
first-order autonomic equations. Eliminating time (differential) between those equa-
tions, we are then left with a system of two equations of the first order.
The history of the reduction of order began with Hess’ work [150] (1890), in which
he used the integrals of motion to express γ in terms of ω and eliminated γ from the
dynamical equations of motion of a heavy rigid body. A system of three autonomous
first-order equations is obtained. The question of equivalence of those equations to
the original problem has led Hess to the known solvable case under his name (see
Sect. 8.4). However, the reduced equations were not used in any application. Relying
also on algebraic elimination of the vector γ, Shiff [331] and Stäckel [342] practically
repeated Hess’ main result in their works. Kharlamov [195] made the most use of this
direction for the motion of a heavy gyrostat, referring the components of the angular
velocity to the special axes of the gyration ellipsoid. He wrote the three autonomous
equations in a more compact form, which was later used by the Donetsk School of
Mechanics in the search for particular solvable cases. The ultimate reduction to a
single second-order equation was tried by Kharlamova in [206], but the equation was
so complicated that it was not written explicitly in final form and was not used in
any application.
@seismicisolation
@seismicisolation
9.2 Maximal Reduction of Order of the Differential Equations of Motion … 203
Bilimovich [23] went another way. He used Euler’s angles as generalized coor-
dinates to write equations of motion in Hamiltonian form and the integrals of areas
and energy to eliminate the variables ψ, t. Bilimovich suggested the use of variables
x = tan 2θ , y = tan ϕ2 and pointed out a sequence of operations which could lead to
a single second-order equation of the structure

y = P3 + V V1 P23 , (9.10)
where y = ddyx , V, V1 are rational functions of x, y and P2 , P3 are polynomials of

degrees 2 and 3 in y with coefficients rational in x, y. Equation (9.10) was not
explicitly written down in final form.
A similar situation was reached by Kharlamova in [207], using Poincaré–Cartan
integral invariant. The variables ψ, t are eliminated and a procedure is pointed out
to obtain an equation of second order, equivalent to (9.10), with the structure
d 2φ dφ
= F(θ, ϕ, ),
dθ2 dθ
but the final explicit form was not given. Such complicated equation was not used in
any application.
The ultimate solution of the reduction problem in the geometric direction initiated
by Bilimovich turned out to be using the components of the vector γ as coordinates
describing the motion of the reduced system, after eliminating the cyclic coordinate
ψ by Routh’s procedure. The final form of the equation produced by this method is
relatively simple and has certain symmetry properties. It turned out to be useful in
deriving simple solutions and in the study of the stability of certain known solutions.
Examples of its uses are given in the relevant chapters. A notable advantage of the
method is that it is equally applicable to obtain a reduction to a second-order equation
in the most general problem of motion of a rigid body under the action of conservative
position-dependent potential and gyroscopic forces.
The result we are going to establish here was reached by elementary, but more
lengthy steps in our 1983 and 1986 works. Here, we shall first apply Hamilton’s
principle in the form of Jacobi (see, for example, [305]) to the time-irreversible
system with the Routhian (9.5). Equations of motion are deduced from the variational
problem
δ Rdt = 0, (9.11)
and applying Maupertuis’ principle to eliminate the time differential from the vari-
ation, we arrive at the final variational problem
@seismicisolation
@seismicisolation

2 ABC(h − V1 ) dγ12 dγ 2 dγ 2
δ ( + 2 + 3)
D A B C
f
+ [Cγ3 (γ2 dγ1 − γ1 dγ2 ) − (A − B)γ1 γ2 dγ3 ]
D(1 − γ32 )
= 0, (9.12)
2
where V1 = V + 2D f
and variations of trajectories are made among the ones with total
energy h of the original system or Jacobi’s constant h of the Routhian system, the
areas integral f as well as the condition that the point γ lies on the Poisson unit
sphere.
Now we use the last condition to eliminate γ2 . Moreover, we choose γ1 (say) to
be the independent variable and γ3 as the dependent one. We reduce the (9.12) to the
variational problem
δ Sdγ1 = 0, (9.13)
where

2 ABC(h − V1 )
S= (λ + 2μγ3 + νγ32 )
D(1 − γ12 − γ32 )
f
+ {Cγ3 + γ1 [Cγ32 − (A − B)(1 − γ12 − γ32 )]γ3 },
D 1 − γ12 − γ32
(9.14)
D = B + (A − B)γ12 + (C − B)γ32 ,
λ = C[Aγ12 + B(1 − γ12 − γ32 )],
μ = ACγ1 γ3 ,
ν = A[Cγ32 + B(1 − γ12 − γ32 )]. (9.15)
The prime denotes differentiation with respect to γ1 and in V1 the variable γ2 is

eliminated by the geometric constraint.
The function S is a Lagrangian function which describes a one-degree-of-freedom
system. The equation of motion of that system can be written as
d ∂S ∂S
− = 0.
dγ1 ∂γ3 ∂γ3
After some tedious calculations, we finally obtain the equation
@seismicisolation
@seismicisolation
D(1 − γ12 − γ32 )γ3 + Cγ3 (1 − γ32 )

−γ1 [A − (A + 2C)γ32 ]γ3 + γ3 [C − (C + 2 A)γ12 ]γ32

−Aγ1 (1 − γ12 )γ33
ρ
− {Cγ3 [(A − B)(A + B − C)γ12 + B(B − C)(1 − γ32 )]
ABC D
+Aγ1 [(B − C)(B + C − A)γ32 + B(A − B)(1 − γ12 )]γ3 }
ρ ∂V1 ∂V1
+ [ (λ + μγ3 ) − (μ + νγ3 )]
2 ABC(h − V1 ) ∂γ3 ∂γ1
f ρ3/2
+ ×
ABC 2D 3 (h − V1 )
×[(A − B)(A + B − C)γ12 − B(A − B + C) + (C − B)(B + C − A)γ32 ]
= 0, (9.16)
where
ρ = λ + 2μγ3 + νγ32 ,
λ = C[B(1 − γ32 ) + (A − B)γ12 ],
μ = ACγ1 γ3
ν = A[B(1 − γ12 ) + (C − B)γ32 ].
This equation is satisfied by some orbits on the Poisson sphere, drawn by the apex
of the vector γ on the surface during motion of the body, and hence we shall call it
“the orbital equation”. It involves the two parameters h (the total energy) and f (the
areas constant of the motion). If one has a solution of this equation in the form
γ3 = γ3 (γ1 ),
then one can establish the relation of variables with time. From the energy integral,
eliminating γ2 , we write
λdγ12 + 2μdγ1 dγ3 + νdγ32

dt 2 = ,
2D(1 − γ12 − γ32 )(h − V1 )
so that the relation between γ1 and time is given by

λ + 2μγ3 + νγ32
t= dγ1 . (9.17)
2D(1 − γ12 − γ32 )(h − V1 )
Inverting this integral, we obtain γ1 (t) and immediately we also have γ3 = γ3 (γ1 (t)).
The component γ2 is obtained from the geometric constraint. This completes the
determination of γ as a function of t. Returning to the formula (9.2), we substitute γ
@seismicisolation
@seismicisolation
and its derivative γ̇ to construct an expression for the angular velocity ω in terms of
time. We have thus completed the reduction of the problem of motion of a rigid body
about a fixed point in a potential field, in all its generality, to the single second-order
ODE (9.16).
9.2.2 Applications
Although the orbital equation (9.16) is not a simple one, it is compact and, to some
extent, symmetric. It turned out to be so effective for some purposes.
9.2.2.1 Orbital Stability of Known Exact Solutions
For example, under some conditions on the potential V and under the condition
f = 0, Eq. (9.16) admits the simplest solution
γ3 = 0. (9.18)
This solution corresponds in the original problem the planar motion of the body
(pendulum motion) about the third principal axis of inertia, which takes a constant
horizontal position during motion.
These motions are either oscillation or rotation, according to the total energy of
the body, and they are periodic in time. The equations in the variations of periodic
motions are linear equations with periodic coefficients. Thus, for the study of the
stability of a periodic motion of the rigid body, we deal with six equations with
periodic coefficients, which admit three integrals. The study of such system is quite
difficult and can be performed mainly numerically. The study of the stability of such
motion seemed intractable and was never considered in the literature. The orbital
equation opened up a way to tackle this problem. The study was initiated in [374]
using the reduced second-order equation of the orbit, and thus called orbital stability.
It means that after the perturbations preserving the energy h and the areas integral
f = 0, the orbit of motion on the Poisson sphere remains near to the circle (9.18). The
equation in the variation for this periodic motion reduces in general to Hill’s linear
differential equation with a periodic coefficient, which depends on the parameters of
the body and on h.
The simplest case occurs when the body is dynamically symmetric and the centre
of mass lies on a principal axis x. The equation of the variation is transformed into
Lame’s equation:
d 2 γ3 1
+ [α(αν 2 + ) − α(α + 1)ν 2 sn2 u]γ3 = 0, (9.19)
du 2 2
@seismicisolation
@seismicisolation

for vibrational motions, where α = 2C/A, ν = h+a 2a
< 1 (the modulus of elliptic
functions), a = Mgx0 and u is the independent variable, linearly depending on time.
In virtue of the triangle inequalities for the moments of inertia A, A, C, it follows
that α is restricted to the interval 0 < α ≤ 4.
For rotational pendulum motions, the equation in the variation takes the form
d 2 γ3 1
2
+ [α(α + k 2 ) − α(α + 1)k 2 sn2 v]γ3 = 0, (9.20)
dv 2

k = h+a 2a
< 1 (the modulus of elliptic functions) and v linearly depends on t.
The zones of stability and instability of the orbit of motion in the plane of the
parameters α, ν are separated by curves in that plane, which carry primary periodic
solutions. Explicit analytical expressions for those curves were given in [374], mainly
with the help of some results of Ince [156, 157] concerning periodic solutions of the
Laméequation. The study was extended to include pendulum motions of a rigid body
in an approximate Newtonian field of a far centre of attraction [376], the rigid body
carrying a rotor (gyrostat) [377] and the case of general (not dynamically symmetric)
body [388].
Plane motions and their stability were considered in some later papers on the
basis of Euler–Poisson equations. Tkhai and Schvigin analysed numerically the case
of vibration for three values of energy parameter [355]. Dovbysh investigated some
qualitative properties of rotational motions [68]. Markeev and coworkers studied
the stability of pendulum motions not only in the linear approximation but also in
the nonlinear setting, using reduction of the Hamiltonian to the Birkhoff normal
form [274]. However, this was done mainly for bodies, whose distribution of mass
satisfies conditions of integrable cases, for example, Kowalevski’s case [275] and the
Goryachev–Chaplygin case [15, 279]. Unfortunately, the rather recent works [15–17,
279] do not mention the results of quite earlier works [375–377, 388], concerning
the stability of pendulum motions in the linear setting, although some comparison
of the results could have been useful. Note that the assertion in [17], that a countable
number of stability and instability zones appear in the plane of parameters, is dubious,
since it is incompatible with our earlier results.
Stability analysis was performed also for some other known periodic solutions of
the Euler–Poisson equations of motion. Grioli’s regular precessional motion is one
of the most famous periodic solutions Chap. 8 (8.10). Its stability was considered in
Grioli’s original 1947 paper [138]. Tkhai studied the full problem numerically and
obtained partial results [356]. The most exhaustive analysis of the stability of Grioli’s
motion was performed by Markeev in two papers [276, 277]. Zones of stability and
instability were presented in the plane of the two parameters that determine the
state of the system. Markeev also studied the stability of the periodic solution of
Euler–Poisson equations known after Steklov [33]. The problem also involves two
parameters, and results are presented in their plane.
In [420], the stability of pendulum motion of a heavy body carrying a rotor is a
generalization to the classical problem. The orbital equation was also successfully
@seismicisolation
@seismicisolation
used to investigate the stability of the particular solution of the classical problem
known after the names of Bobylev and Steklov (see Chap. 8 (8.5)). Zones of stability
were drawn in 3D space [426]. Results obtained by the use of the orbital equation
were not obtained earlier by any other method.
9.2.2.2 Search for New Exact Solutions of the Equations of Motion
In a recent paper [427], the orbital equation was used in a quite different application.
An exact solution of (9.16) is sought in the form γ3 = A0 + A1 γ1 . The forces acting
on the body are assumed to be a combination of the uniform gravity field and Newto-
nian field of attraction of a centre, which is sufficiently far from the body to be taken
in the approximate form Mgr0 · γ + 21 λγ I · γ. A particular solution is constructed,
which turned out to be a generalization of the first case of Dokshevich (see Chap. 8
(8.11)) and turns into it as λ = 0.
9.3 Reduction to the Motion of a Particle on an Ellipsoid
The form of the Routhian (9.5) suggests using the transformation of the coordinates
γ1 , γ2 , γ3 by the relations
√
γ1 = Aξ,
√
γ2 = Bη,
√
γ3 = Cζ, (9.21)
so that the geometric integral becomes
Aξ 2 + Bη 2 + Cζ 2 = 1. (9.22)
Clearly, ξ, η, ζ are Cartesian coordinates of a point on the inertia ellipsoid. The

Routhian transforms to
ABC ˙2
R= (ξ + η̇ 2 + ζ̇ 2 )
2D
f ABC
+ [Cζ(η ξ˙ − ξ η̇) − (A − B)ξη ζ̇]
1 − Cζ 2
−V1 , (9.23)
where
@seismicisolation
@seismicisolation
9.3 Reduction to the Motion of a Particle on an Ellipsoid 209
D = A2 ξ 2 + B 2 η 2 + C 2 ζ 2 , (9.24)
f2
V1 = V + . (9.25)
2D
The last Routhian may be given the interpretation as describing the problem of motion
of a particle on the smooth surface of ellipsoid (9.22), but instead of the standard
Euclidean metric this ellipsoid is endowed by a Riemannian metric conformal to it
that has the form
dξ 2 + dη 2 + dζ 2
ds 2 = 2 2 . (9.26)
A ξ + B 2 η2 + C 2 ζ 2
This alternative description of the quadratic part of the Routhian of the reduced
problem of rigid body dynamics appeared first in a paper of Minkowsky [284] in
1888, devoted to another problem, namely that of the motion of a rigid body in a
fluid.
A usual procedure, frequently used to get out from such situation, is to make a
change of the time variable t and use a new independent variable τ , related to it by
the differential relation
ABC
dt = dτ . (9.27)
D
This procedure is applicable only to iso-energetic motions, i.e. motions on which the
energy constant of the original system with Lagrangian Chap. 3 (3.44) and Jacobi’s
constant for the Routhian (9.23) keep a constant value h. The transformation renders
the Routhian to a new one given by
1 2
R∗ = (ξ + η 2 + ζ 2 )
2
f ABC
+ [Cζ(ηξ − ξη ) − (A − B)ξηζ ]
1 − Cζ 2
−V ∗ , (9.28)
where V ∗ = ABCD
(V1 − h). For a detailed account of the time transformation, see,
for example, the book of Pars [305]. Note that Jacobi’s integral for the new Routhian
R ∗ must be taken to be zero,
1 2
(ξ + η 2 + ζ 2 ) + V ∗ = 0, (9.29)
2
while Jacobi’s constant h for the pre-transformation Routhian R now enters as a
parameter in the reduced potential.
The advantage of the new Routhian R ∗ is that it describes the motion of a particle
of unit mass on the ellipsoid of inertia (9.22) of the body with the standard Euclidean
metric. The motion is driven by two types of conservative forces: potential forces
with the potential function V ∗ and gyroscopic forces characterized by terms linear
in velocities in (9.28).
@seismicisolation
@seismicisolation
9.4 The Use of Elliptic Coordinates on the Inertia Ellipsoid

[224]
The idea to use Jacobi’s coordinates on the inertia ellipsoid as generalized coordinates
in the description of rigid body dynamics appeared in Minkowsky’s work [284],
devoted to the study of the problem of motion of a rigid body in a liquid in the
integrable case due to Clebsch. In that work, Minkowsky suggested using the same
idea in the general problem of motion of a rigid body about a fixed point. Building on
Minkowsky’s result, Kolossov [224] realized this idea and used Routh’s procedure
to ignore the cyclic coordinate, the angle of precession, and thus simplifying the
calculations.
Suppose that A = B = C. The point on the ellipsoid of inertia (9.22) is
parametrized in elliptic coordinates by the following relations:
1 (u − A1 )(v − A1 )
ξ2 = ,
A ( B1 − A1 )( C1 − A1 )
1 ( B1 − u)(v − B1 )
η2 = ,
B ( B1 − A1 )( C1 − B1 )
1 ( C1 − u)( C1 − v)
ζ2 = . (9.30)
C ( C1 − B1 )( C1 − A1 )
This parametrization is valid, provided B is the middle moment of inertia. For deter-
minacy, we assume that C ≤ B ≤ A and A1 ≤ u ≤ B1 ≤ v ≤ C1 . In those coordinates
D = ABCuv.
Substituting (9.30) in the Routhian (3.57), we obtain
1v−u u v
R∗ = [ u 2 + v 2 ]
2 4 (u − A )(u − B )(u − C )
1 1 1
(v − A )(v − B1 )( C1 − v)
1
+lu + mv − V ∗ , (9.31)
where
f ABC ∂ξ ∂η ∂ζ
l= [Cζ(η − ξ ) − (A − B)ξη ],
1 − Cζ 2 ∂u ∂u ∂u
f ABC ∂ξ ∂η ∂ζ
m= [Cζ(η − ξ ) − (A − B)ξη ],
1 − Cζ 2 ∂v ∂v ∂v
1 f2
V∗ = (V + − h). (9.32)
uv 2 ABCuv
@seismicisolation
@seismicisolation
9.4 The Use of Elliptic Coordinates on the Inertia Ellipsoid [224] 211
Practically, this form of the Routhian was used by Kolossov [224], but only giving
zero value to the areas constant f and thus reducing the Routhian to a time-reversible
one.
Kolossov proceeded to consider certain potentials which lead to cases of sepa-
ration of variables in the Hamilton–Jacobi equation for the transformed problem at
f = 0, finally reaching cases solved earlier in terms of time in the works of Weber
[367], Kobb [223] and others.
9.5 The Use of Isometric Coordinates on the Inertia

Ellipsoid
A formal, but important, step was taken by Arjanykh [13], without referring to
Minkowsky or Kolossov. He replaced Jacobi’s coordinates with isometric coordi-
nates on the inertia ellipsoid. In those coordinates, the equations of motion of a
heavy rigid body about a fixed point are transformed into a problem of motion of a
particle in the plane under the action of certain potential and gyroscopic forces.
Let us define two new coordinates
√
udu
X = ,
(u − A )( B − u)( C1 − u)
1 1
√
vdv
Y = . (9.33)
(v − A )(v − B1 )( C1 − v)
1
In those coordinates, the quadratic part of the Routhian becomes 21 v−u

4
(X 2 + Y 2 ).
If, further, we modify the change of the time variable (9.27) to be
ABC (v − u)
dt = . dτ
D 4
(v − u)
= dτ , (9.34)
4uv
the Routhian is transformed to
1 2
R∗ = (X + Y 2 )
2
+l1 X + m 1 Y − V ∗∗ , (9.35)
in which
@seismicisolation
@seismicisolation
du dv
l1 = l , m1 = m ,
dX dY
(v − u) f2
V ∗∗ = (V + − h). (9.36)
4uv 2 ABCuv
The equations of motion now take the form
∂V ∗∗ ∂V ∗∗
X + Y = − , Y − X = − , (9.37)
∂X ∂Y
where
∂l1 ∂m 1
= − (9.38)
∂Y ∂X
du dv ∂l ∂m
= ( − )
d X dY ∂v ∂u
(v − u) u+v
= √ (A + B + C − 2 ). (9.39)
4 ABCuv uv uv
Those equations describe the motion of a particle in the plane X Y, while acted upon
by forces with potential V ∗∗ and a gyroscopic force
(−Y , X ),
which may be interpreted as due to Lorentz’s force exerted on a unit electric charge
by a magnetic field orthogonal to the X Y plane.
The equations of motion in the form (9.37) were written in 1954 by Arjanykh
[13] but with the function as in (9.38), unevaluated. The final form (9.39) was
published first by M. Kharlamov [171] in 1976.
9.6 Reduction to the Simplest Form of Orbital Differential

Equation
In case any two of the principal moments of inertia are equal, the transformation
of coordinates (9.30) used on the ellipsoid of inertia by Kolossov becomes singular
and the equations of motion based on that transformation become invalid. The idea
to modify this transformation to make it applicable for an arbitrary body appeared
in [378]. An examination of the integrals in (9.33) reveals that they both are elliptic
integrals of the third kind. The two variables, which we denote by σ and ρ, are chosen
so that those third-type elliptic integrals defining the isometric coordinates take their
normal (Legendre) form.
@seismicisolation
@seismicisolation
9.6 Reduction to the Simplest Form of Orbital Differential Equation 213
9.6.1 Sphero-Conic Coordinates on the Poisson Sphere [378]
For determinacy, we arrange the principal axes of inertia so that the inequalities
ABC (9.40)
are satisfied. We use new sphero-elliptic coordinates σ, ρ on the Poisson sphere and
the new isometric coordinates x, y defined by the formulas

B σ 1 − k 2 ρ2
γ1 = ,
A (1 − nσ 2 )(1 + mρ2 )

(1 − σ 2 )(1 − ρ2 )
γ2 = ,
(1 − nσ 2 )(1 + mρ2 )
√
B ρ 1 − k 2 σ2
γ3 = , (9.41)
C (1 − nσ 2 )(1 + mρ2 )
and
σ
C dσ
x= ,
A 0 (1 − nσ 2 ) (1 − σ 2 )(1 − k 2 σ 2 )
ρ
A dρ
y= , (9.42)
C 0 (1 + mρ2 ) (1 − ρ2 )(1 − k 2 ρ2 )
where
A − B 2 B −C
k2 = ,k = ,
A−C A−C
A−B B −C
n= ,m = .
A C
Note that 0 ≤ k ≤ 1, 0 ≤ k ≤ 1, 0 ≤ n ≤ 1, 0 ≤ m ≤ 1, and the intermediate vari-

ables satisfy the inequalities
− 1 ≤ σ ≤ 1, −1 ≤ ρ ≤ 1. (9.43)
Note that the transformation (9.41), unlike Kolossov’s transformation (9.30), remains
valid when B = A or B = C. This privilege is much helpful in several applications
and we shall return to this point later on.
Substituting (9.41), (9.42) into (9.5), we obtain
1 P Q
R= Bχ(ẋ 2 + ẏ 2 ) + f ( ẋ − ẏ) − V1 , (9.44)
2 M M
@seismicisolation
@seismicisolation
where
χ = B(1 − k 2 σ 2 − k 2 ρ2 )
A+ B −C 2 2
P = ρ (1 − ρ2 )(1 + k 2 ρ2 ) (1 + mρ2 )(1 − nσ 2 )(1 − k σ ),
A
A+ B −C 2
Q = nσ (1 − σ 2 )(1 − k 2 σ 2 ) (1 + mρ2 )(1 − nσ 2 )(1 − ρ ),
C
(A − B)(A + B − C) 2 2
M = 1 − nσ 2 − ρ2 + σ ρ , (9.45)
A(A − C)
and V1 now takes the form
f2
V1 = V + (1 − nσ 2 )(1 + mρ2 ). (9.46)
2B
Now, we change the time variable t by the relation
dt = Bχdτ . (9.47)
The Routhian (9.44) becomes
1 2 P Q
R∗ = (x + y 2 ) + f ( x − y ) + U, (9.48)
2 M M
in which
U = Bχ(h − V1 )
f2
= B(1 − k 2 σ 2 − k 2 ρ2 )[h − V − (1 − nσ 2 )(1 + mρ2 )] (9.49)
2B
and primes denote derivatives with respect to τ .
The equations of motion of the rigid body in the potential V on the level {h, f } of
the total energy and area’s integrals are reduced in the x y-plane and in the fictitious
time τ , to the form
∂U ∂U
x + y = , y − x = , (9.50)
∂x ∂y
in which
∂ P ∂ Q
= f[ ( )+ ( )]
∂y M ∂x M
χ
= f√ (1 − nσ 2 )(1 + mρ2 )
AC
×[A − B + C − 2(A − B)σ 2 + 2(B − C)ρ2 ]. (9.51)
@seismicisolation
@seismicisolation
Equations (9.50), obtained in [372], serve as equations of motion of a mechanical

system, comprised of a particle moving under the influence of certain potential and
gyroscopic forces in the x y-plane. Note that the motion of this system is subject to
the condition that the Jacobi integral is zero,
1 2
(x + y 2 ) + V1 = 0. (9.52)
2
The energy constant h of the original motion of the rigid body enters as a parameter
in the potential term V1 of the reduced problem (9.50).
Equations of motion in the form (9.50) turned out to be most relevant in working
with the method developed by the author in [381] and are used in numerous works for
the construction of integrable 2D natural and generalized natural mechanical systems
that admit an integral polynomial in velocities.
Because of their symmetric form, Eq. (9.50) was easily applied in [373] to con-
struct periodic solutions of the problem of the motion of a rigid body about a fixed
point in a potential field, near to equilibrium points of (9.50), which correspond to
stationary solutions of Euler–Poisson equations, usually called “uniform” or “per-
manent” rotations. In the classical problem, i.e. when the body moves under the
action of the uniform gravity field, this is the Staude motion (see Chap. 8, Sect. 8.3).
Equations similar to (9.50) but in a less finished form were used in [364, 365].
The variables x, y are uniformizing variables on the Poisson sphere. One can
easily show that the function is an analytical function in the whole x y-plane. If
V is analytic in the components of γ, then it is also analytic in the x y-plane. On
the other hand, from (9.42) one can deduce that σ and ρ are periodic in x and y,
respectively. In fact, substituting σ = sin(φ), the expression for x in (9.42) gives
φ
C dφ
x= .
A 0 (1 − n sin (φ)) (1 − k 2 sin2 (φ))
2
Thus, x is monotone increasing in φ on R. Also, a change 2π in φ adds to x a constant

C π/2 dφ
X =4
A 0 (1 − n sin (φ)) (1 − k 2 sin2 (φ))
2
1
C dσ
=4 . (9.53)
A 0 (1 − nσ 2 ) (1 − σ 2 )(1 − k 2 σ 2 )
The inverse function ϕ(x) is thus defined and single-valued (in the real sense) on R.
It is also quasi-periodic. It increases by 2π when x is increased by X. Finally, we see
that σ is an analytic and periodic function in x. The same can be said about ρ as a
function in y. It is an analytic and periodic function in y with period
@seismicisolation
@seismicisolation
1
A dρ
Y =4 . (9.54)
C 0 (1 + mρ2 ) (1 − ρ2 )(1 − k 2 ρ2 )
Thus, a real algebraic function of σ, ρ is periodic in both directions on the x y-plane

with periods X and Y .
However, it can be shown that the configuration space of the reduced problem
{ϕ ∈ [0, 2π), θ ∈ [− π2 , π2 ]} is covered twice by a rectangle of periods in the x y-
plane. The variables x, y can be considered as new angular coordinates, which enter
in the equations of motion in a more symmetric way than the Eulerian angles. In
fact, it suffices to study motion in half of the square of periods, i.e. on one of the
rectangles S1, S2 , where
X X Y Y
S1 = {x ∈ [− , ), y ∈ [− , )},
2 2 4 4
X X Y Y
S2 = {x ∈ [− , ), y ∈ [− , )}. (9.55)
4 4 2 2
9.6.2 Reduction to a Single Differential Equation [378]
Now, we proceed to obtain the orbital equation of the system described by the
Routhian (9.48), i.e. the equation satisfied by the trajectories of that system on the
Poisson sphere. From (9.52), we get

1 + ( ddyx )2 f2
dτ = d x, U = Bχ[h − V − (1 − nσ 2 )(1 + mρ2 )]. (9.56)
2U 2B
Eliminating dτ from the second equation of motion, we arrive at the second-order

differential equation
d2 y dy 1 ∂U ∂U dy dy
= √ [1 + ( )2 ]3/2 + ( − )[1 + ( )2 ], (9.57)
dx2 2U dx 2U ∂ y ∂x d x dx
in which , U are given in (9.51) and (9.49).

Equation (9.56) is formally the simplest possible form of the orbital equation
(9.16) under transformations of coordinates. This form was obtained in [378], prior
to Eq. (9.16), which is more transparent, but of a much more complicated structure.
It was used in [374] to obtain the equation in the variation to study the stability of
the orbit y = 0, corresponding to pendulum (plane parallel) motion.
@seismicisolation
@seismicisolation
9.6.3 Special Cases
In the above equations, two limiting cases are easily separated:

(1) The first case when B = A > C. Then we have k = n = 0, k = 1, m = CA −

1, X = 2π CA , σ = sin( CA x), Y = ∞ and the relevant rectangle S1 becomes an
infinite strip. The transformation (9.41) becomes

sin( CA x) 1 − ρ2
γ1 = ,
(1 + mρ2 )

cos( CA x) (1 − ρ2 )
γ2 = ,
(1 + mρ2 )

A ρ
γ3 = , (9.58)
C (1 + mρ2 )
while ρ satisfies
ρ
A dρ
y=
0 (1 + mρ )(1 − ρ )
C 2 2

C 1 1+ρ A −1 A
= [ ln + − 1 tan ( − 1ρ)]. (9.59)
A 2 1−ρ C C
(2) The second case A > B = C. In that case, the transformation (9.41) becomes

C σ
γ1 = ,
A (1 − nσ 2 )

(1 − σ 2 ) cos( CA y)
γ2 = ,
(1 − nσ 2 )
√
sin( CA y) 1 − σ 2
γ3 = , (9.60)
(1 − nσ 2 )
where σ is determined from

σ
C dσ
x=
A 0 (1 − A σ 2 )(1 − σ 2 )
A−C

1 + σ 1 − 1 − A σ √1− C
C
1 A
= ln{ ( ) A }, (9.61)
2 C 1−σ 1+ 1− Cσ
A
@seismicisolation
@seismicisolation
i.e. from
1 + σ 1 − 1 − A σ √1− C √C
C
( ) A = e2 A x. (9.62)
1−σ 1+ 1− Cσ
A

It is quite interesting that in some cases, when 1 − CA is a rational number NN ,
such that N , N (N < N ) are relatively prime numbers, the last equation becomes
1 + σ N N − N σ N √
2
) = e2 N −N x .
2
( ) ( (9.63)
1−σ N + Nσ
This is a polynomial equation in σ and can be solved for low N , N (N + N ≤ 4).

We have three cases:
(a) N = 0. That is the case of complete dynamical symmetry A = B = C. Then
σ = tanh x, ρ = sin y. (9.64)
(b) N = 1, N = 2.(B = C = 34 A)
√
1 −1
√ 3
σ = 2 cos [π + cos (tanh( 3x)], ρ = sin( y). (9.65)
3 2
(c) N = 1, N = 3.(B = C = 89 A)
2 16 √
σ = − 1 + μ + 2 − μ + 2 1 − μ + μ2 , μ = sinh− 3 ( 2x),
9
(9.66)
√
2 2
ρ = sin( y).
3
In those three cases, the functions V1 , and then the equations of motion (9.50) can
be expressed explicitly in terms of x, y, and the problem can be studied as a proper
problem of motion of a particle in the plane. However, this only illustrates the relative
degree of complication in the dynamics of a rigid body compared to the dynamics
of a particle.
9.7 Separable Potentials in Rigid Body Dynamics

(Conditional Integrable Problems)
The general reduced problem of the motion of the rigid body about a fixed point
in a potential field was transformed in some of the past sections to other problems
of motion on the surface of a sphere, ellipsoid or other surfaces. When the areas
@seismicisolation
@seismicisolation
9.7 Separable Potentials in Rigid Body Dynamics (Conditional Integrable Problems) 219
constant f = 0, the equations of motion become time-reversible and it becomes

possible to use certain methods of separation of variables known in the analogous
problem. Usually, those cases of separation are obtained by applying the Hamilton–
Jacobi equation. They will be presented here in the form already used in the above
sections, after using a time transformation to attain the Liouville separation.
9.7.1 Potentials Separable for Axi-Symmetric Body
Let the body with axial dynamical symmetry B = A be in motion about the fixed
point O, under the action of forces whose potential is a given function V (θ, ϕ). Let
also the areas constant f = 0. Equations of motion are obtained from (3.58), which
here becomes
A C sin2 θ
R= ( ϕ̇2 + θ̇2 ) − V (θ, ϕ). (9.67)
2 A sin2 θ + C cos2 θ
Using the time transformation
C sin2 θ
dt = dτ , (9.68)
A sin2 θ + C cos2 θ
on the total energy level h, we get
1 A sin2 θ + C cos2 θ 2 C sin2 θ

R∗ = A(ϕ2 + θ )+ [h − V (θ, ϕ)].
2 C sin θ
2
A sin θ + C cos2 θ
2
(9.69)
This system described by a Lagrangian of the last form becomes separable when
V (θ, ϕ) = v1 (ϕ) + v2 (θ). (9.70)
C sin2 θ
Then the Lagrangian R ∗ splits into two independent Lagrangians, each of one degree
of freedom:
1 1 A sin2 θ + C cos2 θ
R ∗ = [ Aϕ2 − v1 (ϕ)] + [ (Aθ2 + h) − v2 (θ)].
2 2 C sin2 θ
Each Lagrangian also has its own energy integral, but with the whole system’s energy
equal to zero, i.e. we have
1
Aϕ2 + v1 (ϕ) = h 1 ,
2
1 A sin2 θ + C cos2 θ
(Aθ2 − h) + v2 (θ) = −h 1 , (9.71)
2 C sin2 θ
@seismicisolation
@seismicisolation
where h 1 is the separation constant. After solving the separated equations and
expressing ϕ and θ in terms of τ , the relation with the original time variable is
obtained by integrating (9.68).
Potentials of the structure (9.70), or in the more transparent form:
V (θ, ϕ) = v1 (ϕ) + (θ), (9.72)
C sin2 θ
suffer a serious drawback. They are always singular at the two poles θ = 0, π. This
singularity cannot be removed by any choice of v1 (ϕ). It is not probable that this
class of separable potential can play some role in physical applications, but, as we
shall see later, they appear in some integrable cases obtained as generalizations of
Kowalevski’s classical case. In those integrable cases, the complementary integral
has degree four in the components of the angular velocity, but it can be rewritten as
the sum of two parts, one of which is the square of a quadratic integral resulting from
separation.
In order to write the quadratic complementary integral for this separable case in
the Euler–Poisson variables, we first note from (9.72) that
A − (A − C)γ32 γ1
V (θ, ϕ) = V (γ) = F1 ( ) + F2 (γ3 ), (9.73)
C(1 − γ3 )
2 γ2
and F1 , F2 are functions of their arguments. It can be verified that I4 can be written
as
1 γ1
I4 = Cr 2 + F1 ( ). (9.74)
2 γ2
9.7.2 Potentials Separable for an Asymmetric Body
Instead of the Jacobian elliptic coordinates on the ellipsoid of inertia, one may use
their reciprocals λ, μ in the substitution

BC(A − λ)(A − μ)
γ1 = ,
(A − B)(A − C)λμ

C A(λ − B)(B − μ)
γ2 = ,
(A − B)(B − C)λμ

AB(λ − C)(μ − C)
γ3 = , (9.75)
(A − C)(B − C)λμ
and define isometric coordinates on the ellipsoid by the relations
@seismicisolation
@seismicisolation
√
dλ
X = √
ABC ,
λ (A − λ)(λ − B)(λ − C)
√
dμ
Y = ABC √ , (9.76)
μ (A − λ)(B − μ)(μ − C)
and the change of the independent variable
1
dt = (λ − μ)dτ , (9.77)
4
then one can reduce the equations of motion on the common level {h, f } of the first
integrals to the form
∂V1 ∂V1
X + Y = − , Y − X = − ,
∂X ∂Y
X 2 + Y 2 + V1 = 0, (9.78)
equivalent to (9.37), where
λ−μ f 2 λμ
V1 = (V + − h), (9.79)
4 2 ABC
V being the original potential of the rigid body, and
f (λ − μ)
= λμ[A + B + C − 2(λ + μ)]. (9.80)
4 ABC
Note that in virtue of (9.75),
ABC
D ≡ Aγ12 + Bγ22 + Cγ32 = ,
λμ
α ≡ ABC[tr(I−1 ) − γI−1 · γ] = D(λ + μ), (9.81)
so that λ, μ are the two roots of the equation
Dλ2 − αλ + ABC = 0, (9.82)
and √
β
λ−μ= , (9.83)
D
where
β = α2 − 4 ABC D. (9.84)
On the level f = 0, the equations of motion become separable if
@seismicisolation
@seismicisolation
V1 = F1 (λ) + F2 (μ),
i.e. the original potential has the form
F1 (λ) + F2 (μ)
V =
λ−μ
Aγ1 + Bγ22 + Cγ32
2
= √ [u(α + β) + v(α − β)]. (9.85)
β
It can be checked that the fourth integral (on the level f = 0) is
1
I4 = A2 p 2 + B 2 q 2 + C 2 r 2 + √ [(α − β)v(α − β) + (α + β)u(α + β)].
β
(9.86)
The general form of the potentials was first given by Kolossov in terms of elliptic
coordinates on the ellipsoid of inertia [224]. He also considered the two cases
(1) u(x) = −v(x) = ax + bx 2 , leading to the potential
V = γI·γ(a + b γI−1 ·γ), (9.87)
where a, b, a , b are arbitrary constants. The first term of this potential is Brun’s
potential.
(2) u(x) = −v(x) = a/x 2 , leading to the rational potential
α tr (I−1 ) − γI−1 ·γ
V = a = a . (9.88)
D γI·γ
Potentials polynomial (or entire functions when N = ∞) in α and D of the form
D
N
V =√ an [(α − β)n − (α + β)n ] (9.89)
β n=1
were considered by Bogoyavlensky [31], and families of rational potentials were

given in [381].
Remark: It is interesting to note that of all potentials (9.85) allowing separation
of variables on the level f = 0, only one potential
V = const ×γI·γ
preserves integrability on arbitrary level f = 0 (see Chap. 6 Sect. 6.4.3). Of course,

there are many other potentials which allow integrability for arbitrary f , but all with
gyroscopic forces in addition to potential forces.
@seismicisolation
@seismicisolation
9.7.3 Potentials Separable for a Body of Spherical Dynamical

Symmetry
In the case of a body of complete dynamical symmetry (A = B = C) and when

f = 0, the Lagrangian (9.8) takes the form
1
R= A(γ̇12 + γ̇22 + γ̇32 ) − V. (9.90)
2
That is a problem of motion of a particle on a sphere. In elliptic coordinates u, v, the
sphere is parametrized as

(a − u)(a − v)
x= ,
(a − b)(a − c)

(u − b)(b − v)
y= ,
(a − b)(b − c)

(u − c)(v − c)
z= . (9.91)
(a − c)(b − c)
This transforms the Lagrangian to
1 u̇ 2 v̇ 2
R= A(u − v)[ + ] − V. (9.92)
8 (a − u)(u − b)(u − c) (a − v)(b − v)(v − c)
Noting that
aγ12 + bγ22 + cγ32 = a + b + c − (u + v),

γ12 γ2 γ2
abc( + 2 + 3 ) = uv, (9.93)
a b c
i.e. u, v are roots of the quadratic equation
γ12 γ2 γ2
u 2 − [a + b + c − (aγ12 + bγ22 + cγ32 )]u + abc( + 2 + 3 ) = 0.
a b c
The separable potentials in elliptic coordinates on the Poisson sphere attached to
the body have the form
@seismicisolation
@seismicisolation

[u 1 (α − β ) + u 2 (α + β )]
V = ,
β
γ12 γ2 γ2
α = a + b + c − (aγ12 + bγ22 + cγ32 ), β = α2 − 4abc( + 2 + 3 ).
a b c
(9.94)
The corresponding quadratic integral is
I4 = A(ap 2 + bq 2 + cr 2 )
1

+ [(α + β )u 1 (α − β ) + (α − β )u 2 (α + β )].
β
(9.95)
One of the simple choices
x2
u 2 (x) = −u 1 (x) =
8
leads to the potential
1
V = (aγ12 + bγ22 + cγ32 ), (9.96)
2
and the integral
γ12 γ2 γ2
I4 = A(ap 2 + bq 2 + cr 2 ) − abc( + 2 + 3 ). (9.97)
a b c
This example corresponds, on the one hand, to Neumann’s problem of motion of a
particle on a sphere acted upon by a quadratic potential in the Euclidian coordinates
[294].1 On the other hand, in rigid body dynamics it is the separable special case
on the level f = 0, of an integrable problem of motion of a body in a liquid due to
Clebsch, namely case number 3 in Table 10.1 of integrable cases in Chap. 10.
9.8 Exercises
1 - Permanent rotations of the rigid body correspond to stationary (time-independent)

solutions of the Euler–Poisson equations. They correspond to equilibrium positions
of the system described by the Routhian (9.5) and the condition (9.6). Show that the
locus of those equilibrium positions satisfies the equation
1In that work, Neumann reduced the problem to hyper-elliptic quadratures and solved them in terms
of Rosenhein Theta functions in two variables.
@seismicisolation
@seismicisolation
9.8 Exercises 225
∂V
· (γ × γI) = 0. (9.98)
∂γ
For the classical problem, this equation becomes
r0 · (γ × γI) = 0,
which gives Staude’s cone (compare with Chap. 8 (8.29)).

[Hint: Extremals of V1 on the unit sphere are obtained from the vector equation
∂V1
= λγ,
∂γ
f2
where V1 = V + 2D
and λ is an undetermined multiplier. That is
∂V f2
− 2 γI =λγ.
∂γ D
Eliminating f, λ, (9.98) follows.]

2 - A dynamically axi-symmetric body (with B = A) is in motion about a fixed
point in the field of a distant Newtonian centre. It has potential (see Chap. 6)
3g
V = aγ1 + [A(γ12 + γ22 ) + Cγ32 ].
2R
Show that Eq. (9.16) for this body admits a solution γ3 = 0, corresponding to
pendulum-like motion about the z-axis, which occupies a horizontal position. Use
(9.16) also to show that the equation in the variation about that solution can be
reduced to the Lame equation
d 2 γ3 1
2
+ {α[αν 2 + + β(α − 2)] − α(α + 1)ν 2 sn2 u}γ3 = 0,
du 2
for vibrations and
d 2 γ3 1
2
+ {α[α + k 2 + β(α − 2)k 2 ] − α(α + 1)k 2 sn2 v}γ3 = 0,
dv 2

for rotations, where ν = k1 = h+a 2a
and β = 3g A
4Ra
is a parameter characterizing the
approximate Newtonian part of the field and other parameters are as above in this
chapter.
[For zones of stability and instability of this motion, see [376, 379].]
3 - A gyrostat with constant gyrostatic momentum κ moves about a fixed point
while acted upon by potential forces with potential V (γ). Show that the reduction
of order as performed above in the present section leads to the equation [377]:
@seismicisolation
@seismicisolation
D(1 − γ12 − γ32 )γ3 + Cγ3 (1 − γ32 )

−γ1 [A − (A + 2C)γ32 ]γ3 + γ3 [C − (C + 2 A)γ12 ]γ32

−Aγ1 (1 − γ12 )γ33
ρ
− {Cγ3 [(A − B)(A + B − C)γ12 + B(B − C)(1 − γ32 )]
ABC D
+Aγ1 [(B − C)(B + C − A)γ32 + B(A − B)(1 − γ12 )]γ3 }
ρ ∂V1 ∂V1
+ [ (λ + μγ3 ) − (μ + νγ3 )]
2 ABC(h − V1 ) ∂γ3 ∂γ1
ρ3/2
+
ABC a D 3 (h − V1 )
×{ f [(A − B)(A + B − C)γ12 − B(A − B + C) + (C − B)(B + C − A)γ32 ]
+D[κ1 (2 A + B + C)γ1 + κ2 (A + 2B + C)γ2 + κ3 (A + B + 2C)γ3 ]
−(A2 γ12 + B 2 γ22 + C 2 γ32 )(κ1 γ1 + κ2 γ2 + κ3 γ3 )}
= 0,
where
ρ = λ + 2μγ3 + νγ32 ,
λ = C[B(1 − γ32 ) + (A − B)γ12 ],
μ = ACγ1 γ3
ν = A[B(1 − γ12 ) + (C − B)γ32 ],

and γ2 , still standing in few places is just abbreviation for 1 − γ12 − γ32 and V1 =
V + ( f −k·γ)
2
2D
is calculated after eliminating γ2 .
4 - The particular solvable case due to Bobylev and Steklov in the classical prob-
lem is based on the condition q ≡ 0. Determine the general form of the potential
V (γ1 , γ2 , γ3 ) of the problem of motion of a rigid body about a fixed point, which
allows the angular velocity to remain permanently in a principal plane of inertia of
the body. Find also the motion of the body.
[This problem is completely solved on the level f = 0 and unsolved yet for
arbitrary f . It turns out that for a general tri-axial body under the conditions q =
0, f = 0, the general form of the potential is
1 +Cγ3
Aγ
2 2
[A(γ12 + γ32 ) − u] A/C du

V = (Aγ12 + Cγ32 ){ F( ) 2 + G(γ2 )},
C(γ12 + γ32 ) − u u
where F and G are arbitrary functions. With this form of the potential, we can
write the expressions for the Euler–Poisson variables in their final form parametrized
by γ1
@seismicisolation
@seismicisolation
9.8 Exercises 227

2C C/A 2A √
( p, q, r ) = (− λγ1 , 0, γ1 ) ,
A C

2C
C/A
(γ1 , γ2 , γ3 ) = (γ1 , 1 − γ12 − λ2 γ1A , λγ1 ),
where
Aγ12 +Cλ2 γ1
2C/A 2C/A
[A(γ12 + λ2 γ1 ) − u] A/C du
= F( 2C/A
) 2
C(γ12 + λ2 γ1 ) −u u

C −1
2A A
F(−λ (A − C)
C ) 2C
+ 2C/A
+ G( 1 − γ12 − λ2 γ1A )
Aγ12 + Cλ2 γ1
and γ1 is determined in terms of time by inverting the integral

dγ1 Aγ 2 2C
t= √ , g(γ1 ) = −2 1 (1 − γ12 − λ2 γ1A ).
g(γ1 ) C
More details and special cases can be found in [416].]

5- A body moving by inertia about a fixed point O has an axis of symmetry Oz
and a particle of mass m moves on that axis without friction, subject to a force with
potential V (z). The problem of motion is described by the Lagrangian
1
L= [(A + mz 2 )(θ̇2 + sin2 θψ̇ 2 ) + C(ψ̇ cos θ + ϕ̇)2 + m ż 2 ] − V (z).
2
Show that the problem is integrable for arbitrary initial conditions by Liouville sep-
aration and reduce its solution to quadratures.
[For solution and a few more details, see [399].]
@seismicisolation
@seismicisolation
Chapter 10
The Problem of Motion of a Body
in a Liquid
The present chapter is devoted to the investigation of the problem of motion of

a rigid body by inertia in an ideal incompressible fluid, infinitely extending in all
directions and at rest at infinity. Strictly speaking, this problem belongs to the field of
fluid dynamics. The problem evolved namely in this way. The ordinary differential
equations of motion of the solid are simultaneously solved with partial differential
equation governing the motion of the liquid under boundary conditions satisfied on
the surface of the moving solid. In this process, the pressure of the liquid had to be
explicitly calculated at every point of the surface of the body. Nevertheless, after the
study of some simple cases, and mainly in the works of Thomson and Tait [352]
and of Kirchhoff [219], it became clear that the body and the liquid can be treated as
forming together one dynamical system of six degrees of freedom, so that the detailed
picture of the pressure of the fluid on the surface of the body is completely avoided.
This system, composed of the body and liquid, was reduced to the motion of a rigid
body with modified characteristics to compensate the motion of the liquid. When
referred to a coordinate frame fixed in the body, the kinetic energy of this system is
expressed as a quadratic form of the components of the angular and linear velocities
of the body with constant coefficients. This step was decisive in the evolution of the
subject along the next few decades.
In this setting, the present problem has six degrees of freedom: three for the
rotational motion and three for the translation of a point of the body and is traditionally
described for a simply connected body by Kirchhoff’s equations [219] (see also
[220]) or by their Hamiltonian form, mostly used by mathematicians, which are due
in their final form to Clebsch [55]. For a perforated body (a body bounded by a
multi-connected surface) the equations of motion are usually taken in the form due
to Lamb [253], or in the equivalent Hamiltonian form (see e.g. [41]).
Research in the problem of motion of a body in a liquid passed through a period of
vigorous activity in the last decades of the nineteenth century. After the formulation
of the equations of motion in their final most general form by Kirchhoff, Clebsch
and Lamb, a lot of significant results was obtained by several eminent, and mostly
https://doi.org/10.1007/978-3-030-96336-1_10
@seismicisolation
@seismicisolation
230 10 The Problem of Motion of a Body in a Liquid
Russian, scientists, including Minkowsky [284], Lyapunov [267], Chaplygin [53] and
Steklov [344, 345, 348]. For almost half a century, the research in the problem entered
a state of stagnation. As stated by Aref and Jones [12] “The Kirchhoff equations
present a most remarkable simplification of a problem that, in principle, involves
an infinite number of degrees of freedom. Surprisingly, the literature exploring these
equations from the point of view of dynamical systems theory is rather sparse”.
Half a century later, the first significant results concerning the integrable cases were
obtained by Rubanovsky [317–320] (See also books: [121, 125]) using a modified
form, due to Kharlamov P., of Clebsch’s equations of motion. In most works outside
Russia, the form of Clebsch (also Hamiltonian) was mostly used for some qualitative
studies of the motion, e.g. [151, 263] (See also references of the last paper).
It turned out that the form of equations of motion involving the variables of
Euler–Poisson type, rather than those of Hamiltonian type, enjoy some privileges
that will be explained later in this chapter. Those are equations formulated, for the
first time, in their most general form in [383]. They are in fact a form of Lagrangian
equations, using redundant non-Lagrangian variables. Such Lagrangian equations are
not completely new. Similar equations were used by Minkowski, in the special case
of Kirchhoff’s equations, to establish his brilliant theorem about the isomorphism
between the reduced problem of rigid body motion and the motion of a particle on a
smooth ellipsoid through a time transformation [284].
In this chapter, equations of motion are presented in their original forms of Kirch-
hoff, Clebsch and Lamb. Our new equations of Lagrangian, in fact Routhian, form
[383] are also presented. This form turned out to be so effective that they put the
problem in a unified context with other problems considered in this book. Those
problems form a hierarchy, ascending from the classical problem to the one of the
present chapter. This hierarchy is extended in the next part of this book to include
the most general problem of motion of a rigid body under the action of conservative
potential and gyroscopic forces which have a common axis of symmetry through the
fixed point. The last problem reduces under some restrictions on the forces, to the
problem, equivalent to that of motion of a body in a liquid. Going lower in the hier-
archy, we note that every problem in it contains all the problems considered before
it as a special case. As a result of this representation of the equations of motion, a
striking property of the equations of motion of a rigid body in a liquid is revealed.
It is the first problem which is closed under the regular precession transformation.
Referring the equations to a coordinate frame precessing with a uniform speed with
respect to the inertial frame, results in the same equations, as if in the inertial frame,
but with changed characteristics of the body. Thus, this transformation generates
from any solution in the present problem or any problem lower in the hierarchy, a
new solution that contains the precession speed as an extra-parameter. This situation
helped to re-organize the known information about the subject and to fill gaps in
it. Some recently discovered integrable cases are generalized. Tables are given for
all integrable cases, general and conditional. The most important known families of
particular solutions to the problem are discussed on different levels of detail.
@seismicisolation
@seismicisolation
10 The Problem of Motion of a Body in a Liquid 231
In our presentation of the subject, the problem of motion of a body in a liquid

plays a rather unusual role. Results obtained in this problem by various methods
and accumulated along a century have grown into a core for the advancement of
some other problems of motion of a rigid body under more sophisticated forces. In
later chapters, we shall use some transformations to obtain new integrable extensions
which are more general from the physical and mathematical aspects and which were
not subjected to any studies before.
10.1.1 Kirchhoff’s Equations
Consider a rigid body moving in an ideal incompressible liquid, extending to infinity

in all directions and at rest at infinity. Assume that the body is bounded by a simply
connected surface and is moving by inertia, i.e. under no forces, except those exerted
on it by the pressure of the liquid on its surface. Let O and O, respectively, be the
origins of the inertial coordinate system O X Y Z and another system O x yz fixed in
the body and let r = O O. Denote by ω the angular velocity of the body and by u
the velocity of O with respect to O , so that u = drdt
. The equations of motion were
derived in Lagrangian form using the Lagrangian function L (kinetic energy T , since
no external forces are present):
1
L=T = (ωA · ω+2uB · ω + uC · u) (10.1)
2
in which A, B, C are constant 3 × 3 real matrices; A, C symmetric and B is not
necessarily symmetric. Here, the state variables ω and u and all quantities (parameters
of the problem) are referred to the body system. Of course, as a quadratic form, T
must be positive definite in the six variables. For this the three matrices must satisfy
certain inequalities.
We shall not go through the explicit derivation of the matrices A, B, C from the
underlying hydrodynamical problem, because that would increase the size of this
chapter beyond preassigned limits. All this material can be found in the treatise of
Lamb [253]. It will be helpful in dealing with motion of bodies with certain symmetry
properties to borrow the cases presented in the following table from that treatise. A
similar table is presented in [41].
@seismicisolation
@seismicisolation
Table 0:
Symmetry Matrices Examples
1 Plane of symmetry x y A=⎛ diag(A1 , A2 , A3⎞),
0 0 B13
B = ⎝ 0 0 B23 ⎠ ,
B B 0
⎛ 13 23 ⎞
C11 C12 0
C = ⎝ C12 C22 0 ⎠ .
0 0 C33
2 Two orthogonal planes of A=⎛ diag(A1 , A2 , ⎞
A3 ),
symmetry x y, x z 0 0 0
B = ⎝ 0 0 B23 ⎠ ,
0 B23 0
C = diag(C1 , C2 , C3 )
3 Three orthogonal A = diag(A1 , A2 , A3 ), Tri-axial
planes of sym- B = 0, ellipsoid,
metry x y, x z, yz C = diag(C1 , C2 , C3 ) Parallele-piped.
4 Rotation through A=⎛ diag(A1 , A2 , A3⎞), Two-bladed
an angle π about B11 B12 0 ship screw.
axis Oz. B = ⎝ B12 B22 0 ⎠ ,
0 0 B33
⎛ ⎞
C11 C12 0
C = C12 C22 0 ⎠ .
⎝
0 0 C33
5 Rotation through A = diag(A1 , A2 , A3 ), Four-bladed
an angle π/2 B = diag(B1 , B2 , B3 ), ship screw.
about axis Oz. C = diag(C1 , C2 , C3 ).
6 Helicoidal symmetry about A = diag(A, A, A3 ), Helicoid.
axis Oz. B = diag(B, B, A3 ),
C = diag(C, C, C3 ).
7 Oz is axis of symmetry (or A = diag(A, A, A3 ), Spheroid,
rotation B = 0, Three-bladed
through an angle C = diag(C, C, C3 ). ship screw.
n ,n ∈
/ {2, 4}
2π
around z-axis).
8 The body is similarly A = Aδ, Cube, sphere.
related to each of the B = 0,
coordinate planes. C = Cδ.
It is usually argued that the origin of the movable coordinate system can always
be shifted so that O coincides with a certain point of the body, called the central
point, at which the matrix B becomes symmetric. It is also usually assumed that the
axes of the body system are rotated to the principal axes of the matrix A, so that the
matrix A becomes diagonal. However, we shall see soon that there is no need for
those steps for the time being, if one is not concerned in using the original variables
ω and u.
The equations of motion are [219]
@seismicisolation
@seismicisolation
10.1 Equations of Motion 233
d ∂L ∂L ∂L
+ω× + u× = 0,
dt ∂ω ∂ω ∂u
d ∂L ∂L
+ω× = 0. (10.2)
dt ∂u ∂u
Explicitly, Kirchhoff’s equations can be written in vector form
ω̇A+u̇B + ω × (ωA + uB) + u × (ωBT +uC) = 0,

ω̇BT +u̇C + ω × (ωBT +uC) = 0 (10.3)
or, if one introduces the notation
∂L
M= = ωA + uB, (10.4)
∂ω
and
∂L
p= = ωBT +uC, (10.5)
∂u
in the alternative form
Ṁ + ω × M + u × p = 0,
ṗ + ω × p = 0. (10.6)
Equation (10.3) are quite complicated. An obvious disadvantage is that they are not
solved with respect to the derivatives. Every scalar equation of motion may contain
the six components of the derivatives ω̇ and u̇. Following Kirchhoff, we also note
that those equations admit three integrals of motion:
1. The energy integral, as the Lagrangian is a homogeneous quadratic polynomial
of the velocities
1
I1 = (ωA · ω + 2uB · ω + uC · u). (10.7)
2
2. From the second equation in (10.6), it follows that the magnitude of the vector
p = ∂∂uL is conserved.
I2 = |p|2 = |ωBT +uC|2 . (10.8)
3. Also, using both Eq. (10.6), we get
I3 = M · p = (ωA + uB) · (ωBT +uC). (10.9)
The system of Eq. (10.3) was used in the treatment of certain simple cases.
@seismicisolation
@seismicisolation
10.1.2 Example: Permanent Translational Motions
For example, Kirchhoff investigated the possibility that the body performs uniform
translational motion without rotation. From (10.3), setting ω= 0, it turns out that the
condition for this motion is
u × uC = 0.
That is, the vector u must be directed along one of the principal axes of the matrix C.
Thus, a body of an arbitrary shape (with a tri-axial ellipsoid of the matrix C) always
has three mutually orthogonal axes such that if the body is set in motion parallel to
one of them along any direction in space and then left to itself, it will permanently
continue this motion with constant velocity. In case of two equal principal axes, all
axes at the equatorial plane are possible axes of permanent translation and also the
polar axis, and in case of spherical symmetry all directions in the body are possible
for permanent translation. Note that actual spherical symmetry of the body is not
necessary. It is a dynamical property that the matrix C has three equal eigenvalues.
This condition is satisfied by cubes as well as by spheres [253].
10.1.3 Clebsch’s Form of Kirchhoff’s Equations
Clebsch [55] transformed Eq. (10.6) to Hamiltonian form using the variables M, p
and the Legendre transformation
H (M, p) = M · ω + p · u − L
1
= (Mã · M + 2Mb̃ · p + pc̃ · p) (10.10)
2
where
ã = (A − BT C−1 B)−1 ,
b̃ = −C−1 B(A − BT C−1 B)−1
−1
c̃ = C−1 + C B(A − BT C−1 B)−1 BT C−1 .
Note that ã, c̃ are symmetric but b̃ is not.

The equations of motion acquire the Hamiltonian form due to Clebsch
∂H ∂H ∂H
Ṁ = M × +p× , ṗ = p × (10.11)
∂M ∂p ∂M
or, in expanded form,
@seismicisolation
@seismicisolation
Ṁ = M×(Mã+pb̃T ) + p × (Mb̃+pc̃),
ṗ = p × (Mã+pb̃T ), (10.12)
which is used, usually assuming symmetry of the matrix b̃, until recently, e.g. [41,
246, 263, 327, 328]. For the Hamiltonian form of the Kirchhoff equations see also
[12, 257, 258, 280].
The general integrals of motion take their simplest form in the variables M, p :
I1 = H,
I2 = M · p
I3 = p2 . (10.13)
10.2 Thomson-Lamb’s Equations
By the words of the contemporary of Thomson and Lamb, A.B. Basset [19] “the
general theory of motion of a ring in an infinite liquid, when there is cyclic irrota-
tional motion through its aperture, was first given by Sir William Thomson in the
Philosophical Magazine (1871), and his theory has been subsequently developed by
Professor Lamb, in his Treatise on the motion of fluids” [252]. Hence, and although
the equations of motion are direct generalization of Kirchhoff’s equations, I will
give the name “Thomson-Lamb’s theory” to the theory of equations of motion of
a multi-connected (perforated) rigid body in a liquid. This problem was not con-
sidered in its generality in the western literature for about a century. In fact, after
the works of Basset and Fawcett on the motion of perforated bodies in liquid (e.g.
[19, 83]) in the last two decades of the nineteenth century, no significant results are
seen in this area until the equations of motion were reformed by Kharlamov in the
sixties. The deduction of the Lagrangian (Euler–Poisson type) equations appeared
in 1986, a whole century later. This may have been caused by the historical nature
of that period at the beginning of the twentieth century. The period of birth of new
physical theories: atomic physics, relativity, old quantum and then quantum theories.
Research in branches of classical mechanics was significantly retarded.
It may be noted here that the generalization of Kirchhoff equations for perforated
body was given by some authors the name “Kirchhoff–Poisson equations”. As exam-
ples, see [121, 125]. This name seems to us irrational, since Poisson had no relation
at all to the present circle of problems.
Let O and O, respectively, be the origins of the inertial coordinate system and
−−→
the system fixed in the body, and let r = O O. Denote by ω the angular velocity of
the body and by u the velocity of O with respect to O , so that u = dr
dt
. The equations
as in [253] are derived from a Lagrangian function (kinetic energy, since no external
forces are present):
@seismicisolation
@seismicisolation
Fig. 10.1 Simple and perforated bodies
1
T = (ωA · ω+2uB · ω+uC · u) + ᾱ · ω+β̄ · u (10.14)
2
in which A, B, C are constant 3 × 3 real matrices; A,C symmetric and B is not
necessarily symmetric and ᾱ, β̄ are constant vectors, which characterize the multi-
connectedness of the body and the circulations of the fluid on irreducible circuits
drawn on its surface (Fig. 10.1b). Here, the state variables ω and u and all quantities
(parameters of the problem) are referred to the body system. For a body bounded by
a simply connected surface the vectors ᾱ, β̄ vanish and the Lagrangian turns into
the one used by Kirchhoff and Clebsch.
It is usually argued that the origin of the movable coordinate system can always
be shifted to a certain point of the body, called the central point, at which the matrix
B becomes symmetric if it is not so at O, and hence it is also usually assumed that
the axes of the system are rotated, so that the matrix A becomes diagonal. However,
we shall see soon that there is no need for those steps for the time being.
The equations of motion are [253]
@seismicisolation
@seismicisolation
10.2 Thomson-Lamb’s Equations 237
d ∂T ∂T ∂T
+ω× + u× = 0,
dt ∂ω ∂ω ∂u
d ∂T ∂T
+ω× = 0. (10.15)
dt ∂u ∂u
Explicitly, Lamb’s equations can be written in vector form
ω̇A+u̇B + ω × (ωA + uB+ᾱ) + u×(ωBT +uC+β̄) = 0,

ω̇BT +u̇C + ω × (ωBT +uC+β̄) = 0 (10.16)
or, if we introduce the notation
∂T
M= = ωA + uB+ᾱ, (10.17)
∂ω
and
∂T
p= = ωBT +uC+β̄, (10.18)
∂u
in the alternative form
Ṁ + ω × M + u × p = 0,
ṗ + ω × p = 0. (10.19)
Equation (10.16) are quite complicated. An obvious disadvantage is that they are not
solved with respect to the derivatives. Every scalar equation of motion may contain
the six components of the derivatives ω̇ and u̇. Following Lamb, we also note that
those equations admit three integrals of motion:
1. Jacobi’s integral, the homogeneous quadratic part of the Lagrangian
1
I1 = (ωA · ω+2uB · ω + uC · u). (10.20)
2
2. From the second equation in (10.19), it follows that the magnitude of the vector
p = ∂∂uL is conserved.
2
I2 = |p|2 = |ωBT +uC+β̄| .
3. Also, using both Eq. (10.19), we get
I3 = M · p = (ωA + uB + ᾱ) · (ωBT +uC+β̄).
The system of Eq. (10.16) was used in the treatment of certain simple cases and
is usually transformed to the Hamiltonian variables. Using a Hamiltonian (see, e.g.
[41]):
@seismicisolation
@seismicisolation
H = M · ω−T
1
= (Mã · M + 2Mb̃ · p + pc̃ · p) + α̃ · M+β̃ · p (10.21)
2
the equations of motion acquire the form
∂H ∂H ∂H
Ṁ = M × +p× , ṗ = p × , (10.22)
∂M ∂p ∂M
or in explicit form
Ṁ = M × (Mã + p b̃T + α̃) + p × (Mb̃+pc̃ + β̃),

ṗ = p × (Mã + p b̃T + α̃). (10.23)
Integrals of motion take the simple form:
I1 = H,
I2 = p2 ,
I3 = M · p.
The last form of equations is used in most recent works, e.g. [41].
10.3 On Different Forms of the Equations of Motion
The traditional equations of Kirchhoff and Lamb suffer some disadvantages that in
most cases lead to their treatment for most of their history in isolation from other
problems of rigid body dynamics. They also involve the non-symmetric matrix b.
Although this matrix can be reduced to symmetric form by shifting the origin to the
central point of the body, the presence of non-symmetry complicates the equations
either in Lagrangian or Hamiltonian forms. In most recent works some simplifying
restrictions on the parameters are assumed, such as b = 0 (e.g. [151, 263]). Equa-
tion (10.11) has also the disadvantage that their solution gives the vector quantity M,
which has no direct interpretation in terms of the motion unless transformed to an
expression involving the angular velocity ω and the vector p constant in space.
If Eqs. (10.2) (or (10.11)) are written in the frame of reference attached to the
principal axes of a matrix A (or ã), they involve 15 parameters characterizing the
shape of the body.
If Eqs. (10.15) (or (10.22)) are written in the frame of reference attached to the
principal axes of a matrix A (or ã), they involve 24 parameters characterizing the
shape of the body and, for a perforated body, circulations of the fluid along irreducible
contours on its surface.
@seismicisolation
@seismicisolation
10.4 A New Form of the Equations of Motion 239
10.4 A New Form of the Equations of Motion
Here, we present with minor modification a new form of the equations of motion
of a general body in a liquid, which was derived in our work [383]. We note first
that in the Lagrangian (10.1), the Cartesian coordinates (X, Y, Z ) of the origin of the
system of axes fixed in the body relative to the inertial system are cyclic variables,
since the resultant of forces acting on the body-and-fluid system vanishes. We now
ignore those coordinates using the vector cyclic integral
∂T
= ωBT +uC+β̃ = p, (10.24)
∂u
where p is a vector whose components are constant in space and hence satisfies the
Poisson equation
ṗ + ω × p = 0. (10.25)
Now, solving the relation (10.24) in u we obtain
u = (p − β̄ − ωBT )C−1 (10.26)
and we proceed to form Routh’s function
∂L
R = T −u·
∂u
1 1
= ωI · ω + (κ+pK̃) · ω−a · p− pJ · p (10.27)
2 2
where I,K̃, J are the constant 3 × 3 matrices and κ, a are the constant vectors given
by
I = A − BT C−1 B,
J = C−1 , (10.28)
−1
K̃ = C B,
a = −β̄C−1 , (10.29)
−1
κ = ᾱ−β̄C B. (10.30)
As seen from (10.30), the matrices I, J are symmetric but K̃, in general, is not.
Let K̃s and K̃a be the symmetric and antisymmetric parts of K̄, so that
@seismicisolation
@seismicisolation
K̃ = K̃s +K̃a , (10.31)

1 1
K̃s = [C−1 B + (C−1 B)T ] ≡ − K, (10.32)
2 2
1
K̃a = [C−1 B − (C−1 B)T ]. (10.33)
2
Here we introduced a constant matrix K = −[C−1 B + (C−1 B)T ]. Inserting (10.31)

into (10.27), we can write
R = R0 +pK̃a · ω, (10.34)
where
1 1 1
R0 = ωI · ω + (κ − pK) · ω−a · p− pJ · p. (10.35)
2 2 2
We now show that the antisymmetric part Ka does not contribute to the equations
of motion. In fact, the last term of (10.34) is
⎛ ⎞
0 −K a3 K a2
pKa · ω = (p1 , p2 , p3 ) ⎝ K a3 0 −K a1 ⎠ · ω
−K a2 K a1 0
= (p × ka ) · ω
where we introduced the vector ka = (K a1 , K a2 , K a3 ) constant in the body axes.

Thus, we have
pKa · ω = ka · (ω×p)
= −ka · ṗ
d
= (−ka · p).
dt
Thus, the extra term in (10.34) is a nugatory term and has no contribution to the equa-
tions of motion (e.g. [305]). The Routhian R0 gives full description of the rotational
motion of the body. Euler’s equation for this motion can be deduced in a simple way.
With an eye on future applications, we present that in detail. In fact, the equation of
motion about the third axis of the body system is
d ∂ R0 ∂ R0
( )− = 0. (10.36)
dt ∂ ϕ̇ ∂ϕ
This gives
d ∂ R0 ∂ω ∂ R0 ∂ω ∂ R0 ∂p
( · )− · − · = 0.
dt ∂ω ∂ ϕ̇ ∂ω ∂ϕ ∂p ∂ϕ
@seismicisolation
@seismicisolation
10.4 A New Form of the Equations of Motion 241
But from formulas of Chap. 2, we have
∂ω ∂ω ∂p
= k, = −k × ω, = −k × p, (10.37)
∂ ϕ̇ ∂ϕ ∂ϕ
and thus we get

∂ R0 ˙ ∂ R0 ∂ R0
k · [( )+ω× + p× ] = 0,
∂ω ∂ω ∂p
so that the vector equation of motion can be written as
∂ R0 ˙ ∂ R0 ∂ R0
( ) + ω× + p× = 0.. (10.38)
∂ω ∂ω ∂p
Now, inserting the expression (10.35) for the Routhian, we obtain
1 1 1
(ωI+κ − pK)˙ + ω × (ωI+κ − pK) + p × [ − ωK − (a + pJ)] = 0,
2 2 2
(10.39)
As I,κ and K are constants in the body, the last equation becomes
1 1 1
ω̇I − ṗK+ω×(ωI+κ − pK) − p×ωK = p × (a + pJ),
2 2 2
and using Poisson’s equation in the second term
1 1 1
ω̇I + ω×(ωI + κ − pK) + (ω×p)K + ωK × p = p × (a + pJ). (10.40)
2 2 2
Here, using the identity
(ω×p)K+ωK × p = ω×(p[tr (K)δ−K]),
valid for any two vectors ω, p and symmetric matrix K, we write the final form of
the equations of motion
ω̇I + ω × (ωI + κ + pK̄) = p × (a + pJ),

ṗ + ω × p = 0. (10.41)
where K̄ = 21 tr (K)δ − K, which is the same as the relation between I and Ī in

Chap. 1.
Equation (10.41) admit three first integrals:
@seismicisolation
@seismicisolation
1 1
I1 = ωI · ω + a · p+ pJ · p,
2 2
I2 = p2 ,
1
I3 = (ωI+κ − pK) · p. (10.42)
2
The vectors κ and a, resulting from the circulation of the fluid in the body per-
forations vanish for a simply connected body, in which case Eq. (10.41) reduce to a
form equivalent to Kirchhoff’s equations.
When referred to principal axes of the matrix I, Eq. (10.41) in the general case
involve only 21 parameters, compared to 24 in (10.22). The parameters of the original
problem can be expressed by inverting the relations (10.30) as:
1 1
A = I − ( K + K̃a )J−1 (− K+K̃a),
2 2
−1 1
B = J (− K + K̃a ),
2
C = J−1 ,
1
K̃ = − K + K̃a ,
2
1
ᾱ = κ+aJ−1 (− K + K̃a ),
2
β̄ = −aJ−1 , (10.43)
so that we retain, if we like, the three elements of the antisymmetric matrix, and thus
also the full set of 24 (18) parameters of the original Lamb (Kirchhoff) formulation.
Remark The observation that the antisymmetric part K a of the matrix C −1 B
has no contribution to the equations of motion, except entering into the symmetric
matrix A and the vector ᾱ, eliminates the necessity in several works to translate the
origin O fixed in the body to the so-called central point of the body, or to assume
the symmetry of the matrices and thus, unnecessarily, restricting the possible forms
of the body. Calculation of the coefficient matrices can be done at a suitable point
from the point of view of calculation and the characteristics of the rotational motion
are then constructed free of the choice of the origin.
Remark The same observation resolves once for all a situation that the Hamil-
tonian equations based on the original Kirchhoff and Lamb equations that one can
need to perform a canonical transformation to the Hamiltonian, to the equations of
motion and to the integrals of motion, so that after the transformation the integrals of
motion take a relatively simpler form. An example of such situation is the case found
originally by Sokolov in [335]. A canonical transformation introduced by Borisov
and Mamaev in [39] was used to simplify the Hamiltonian and to give the integral a
simpler form.
@seismicisolation
@seismicisolation
10.5 Steklov and Kharlamov Analogies and Their Generalization 243
10.5 Steklov and Kharlamov Analogies and Their

Generalization
The problem of motion of a rigid body in a liquid has been considered for a part of
its history in complete isolation of other problems of motion of a rigid body about a
fixed point.
As will be seen in more detail in the coming chapters, the equations of motion
in their full form (10.41) derived from the Routhian (10.27) can be interpreted as
equations of motion about a fixed point of a heavy, magnetized and electrically
charged body bearing a rotor and influenced by an axially symmetric combination
of three classical fields. More precisely, the second equation of (10.41) resembles
Poisson’s equation met in several previous chapters. This equation describes the
space time derivative of a vector p constant in space, referred to the body system.
Let us take a unit vector γ in the direction of p, so that p = p0 γ and thus equations
(10.41) take the form
ω̇I+ω×(ωI+κ + p0 γ K̄) = γ×( p0 a+ p02 γJ),

p0 (γ̇ + ω × γ) = 0. (10.44)
Let us first assume that p0 = 0. In that case one can absorb this constant in the
definitions for a, J and K̄.
10.5.1 The Equivalent Problem of Motion About a Fixed

Point
Alternatively, one can choose the units of measurement so that p0 becomes unity.
Finally, Eq. (10.41) can be written in the form of equations of motion of a rigid body
about a fixed point with the vector γ fixed in space, i.e.
ω̇I+ω×(ωI+κ + γ K̄) = γ×(a+γJ),

γ̇ + ω × γ = 0. (10.45)
This system of equations may be obtained from (10.41) by the replacement
¯ p0 , a/ p0 , J/ p02 ),
(p,K̄, a, J) →( p0 γ, K/ (10.46)
so that if for some parameters I,κ,K̄, a, J one has a solution ω = ω(t) and γ = (t)
of the equivalent system (10.45), then one can obtain a solution ω = ω(t), p = p0 γ
of (10.41) through the replacement of parameters
@seismicisolation
@seismicisolation
(K̄, a, J) →( p0 K̄, p0 a, p02 J). (10.47)
In this way, in the solution of the problem (10.41) an additional parameter p0 is

added. Returning to the problem of motion of a body in a liquid in the original
formulation, we obtain a solution containing five parameters more than the solution
of the equivalent problem.
Let us now turn to the excluded case p0 = 0. In that case (10.44) reduces to
ω̇I+ω×(ωI+κ) = 0, (10.48)
which are the equations of motion of a free gyrostat fixed from one point. Those are
the integrable equations already discussed in Chap. 5 under the name of Joukovsky
and Volterra. This justifies the use of Eq. (10.45) in the generic case.
Equation (10.45) can be derived from the Lagrangian
1 1 1
L= ωI · ω + (κ− γK) · ω−a · γ− γJ · γ, (10.49)
2 2 2
which is the last form of the Routhian R in (10.27). They admit the set of three
integrals corresponding to (10.42), which are now written as
1 1
I1 = ωI · ω + a · γ+ γJ · γ = h,
2 2
I2 = γ 2 = 1,
1
I3 = (ωI+κ − γK) · γ = f (10.50)
2
where h, f are arbitrary integration constants.
The six-dimensional problem of motion of the rigid body in the liquid is thus
reduced to another problem of motion of a body about a fixed point, having only three
degrees of freedom. This problem is described by Eq. (10.49) and has the integrals
(10.50). The Lagrangian of the new problem (the Routhian R of the original problem)
involves the angular velocity ω and the vector γ constant in space. The forces acting
on this virtual body can be interpreted as having a scalar potential
1
V = a · γ + γJ · γ, (10.51)
2
and a vector potential
1
l = κ− γK. (10.52)
2
From now on, to conform with the previous simpler problems and with future
study of more complex problems, we shall write the Lagrangian (10.49) as
@seismicisolation
@seismicisolation
10.5 Steklov and Kharlamov Analogies and Their Generalization 245
1
L= ωI · ω + l · ω − V, (10.53)
2
and Eq. (10.45) as
∂V
ω̇I+ω×(ωI+μ) = γ× ,
∂γ
γ̇ + ω × γ = 0. (10.54)
where
1
V = a · γ+ γJ · γ, (10.55)
2
μ = κ + γ K̄. (10.56)
In this form, each of the terms appearing in the equations of motion (10.54) of a rigid
body in a liquid can be given concrete alternative interpretation:
(a) The vector a constant in the body, compared with formulas in Chap. 3, can be
interpreted as the term Mgr0 , the product of the weight of the equivalent body in
a uniform gravity field g in the direction of (−γ) and the position vector r0 of the
centre of mass of that body.
(b) The vector κ, also constant in the body, can be interpreted as a gyrostatic
momentum of a symmetric rotor fixed from its axis of symmetry and rotating about
it with a constant angular rate (Compare with Chap. 5).
(c) The potential term 21 γJ · γ has a form, similar to that of the potential of a
far Newtonian centre of attraction (Compare with Chap. 6), but can be interpreted in
that way only when the matrices J and I are proportional J = λI. For an arbitrary
matrix J, this term can be given interpretation as partially due to an attraction centre
and partially as due to the electric interaction of a far Coulomb centre on the line
parallel to γ and passing through the origin O, fixed in the present analogy, on a set
of electric charges fixed in the equivalent body. In this interpretation, the matrix J is
proportional to the inertia matrix of the electric charges on the equivalent body.
(d) The term γ K̄ of the vector μ can be interpreted as a result of the Lorentz effect
of a uniform magnetic field parallel to γ on the electric charge distribution on the
body (see e.g. [139]). This effect will be considered in more detail later in this book.
Conclusion: The above considerations show that the overall effect of the hydro-
dynamic forces exerted by the fluid on the body can be replaced, as to their effect
on the rotational motion of the body, by a set of relatively simple gravitational and
electromagnetic interactions.
By analogy or equivalence between the two problems here we mean full iso-
morphism of their equations of motion.1 This analogy, pointed out in 1986 [383],
generalizes the limited earlier analogies due to Steklov and Kharlamov:
1 A weaker type of equivalence will be treated below involves isomorphism on the level of Routh-
reduced equations of motion. The full Lagrangian systems are not isomorphic to each other, but any
integrable case of one of them leads to an integrable case of the other.
@seismicisolation
@seismicisolation
10.5.2 Steklov’s Analogy
In [345] (1895) and [348] (1902) noted that if in Kirchhoff’s equations in Clebsch’s
form (10.12) one sets b̃= 0, c̃ = ã those equations become identical with the equa-
tions of motion of a rigid body about a fixed point while acted upon by approximate
Newtonian field in the integrable case when the body is fixed from its centre of mass
(Case 2 of Chap. 6). In the terminology of Eq. (10.45) Steklov’s analogy concerns
the case J = I, K̄ = 0, κ = a = 0.
10.5.3 Kharlamov’s Analogy
In 1963, Kharlamov [192] generalized Steklov’s analogy to the case of a perforated

body, allowing non-zero vectors α,˜ β̃ in (10.23) and requiring only that b̃ = 0, c̃ =
ã. For Eq. (10.54) Kharlamov’s analogy requires that:
J = I, K̄ = 0,
under which the problem of motion of a body in a liquid is analogous to the motion
of a gyrostat about a fixed point, under the action of approximate Newtonian field of
a centre (See Chap. 6).
10.6 Completing the Solution
10.6.1 Solution of the Equivalent Problem
Solving the system of equations of motion (10.45) we determine, as functions of the

time t, the vectors ω(t) and γ(t). In the alternative problem we regard the vector
γ(= pp0 ), constant in space, as the unit vector pointing vertically upwards, take the
Z -axis in that direction and measure the angle of precession ψ in the plane orthogonal
to it. As in the classical problem (see 3.9), this determines the Eulerian angles of
nutation and proper rotation θ and ϕ as
γ1
θ = cos−1 γ3 , ϕ = tan−1 (10.57)
γ2
while the precession angle ψ is expressed by the quadrature

t
pγ1 + qγ2
ψ = ψ0 + dt, (10.58)
0 γ12 + γ22
@seismicisolation
@seismicisolation
10.6 Completing the Solution 247
ψ0 is an integration constant. This completes the solution of the equivalent problem

of motion about a fixed point, which is also the solution of the rotational part of the
body in a liquid.
10.6.2 Solution of the Original Problem
Suppose that for the parameters I,κ, K, a, J the equivalent problem (10.45) has a
solution ω = ω(t) and γ = (t). The rotational motion of the body in the liquid is
the same as in the previous subsection.
Conditions on the parameters of the original parameters are obtained by applying
(10.47) to (10.43). This gives
1 −1 1
A = I − ( p0 K + K̃a )J ( − p0 K + K̃a ),
2 2
1 −1 1
B = 2 J (− p0 K + K̃a ),
p0 2
1
C = 2 J−1 ,
p0
1
K̃ = − p0 K + K̃a ,
2
1 1
ᾱ = κ+ aJ−1 (− p0 K + K̃a ),
p0 2
1
β̄ = − aJ−1 . (10.59)
p0
The velocity of the point O taken as origin is determined from (10.26) through
the formula
u = a+J−K̃ T
1
= a+J−( K − Ka )
2
1
= a+J− ωK+ × ka . (10.60)
2
In the last formula, one can easily recognize the term ω×ka as the only origin-
dependent term. It represents the velocity of a unique point of the body whose position
vector relative to O is ka . In the sequel, this point will be called the proper central
point of the body. In contrast to the settled notation of the central point as the point
at which the matrix B is symmetric, the proper central point has direct dynamical
significance. If we take this point of the body as the origin, the matrix K̃ would be
symmetric. Taking (10.47) into account, the velocity ucp of the central point is
@seismicisolation
@seismicisolation
ucp = p0 a+ p02 J− p0 K. (10.61)
That is origin-independent and depends only on the angular velocity and the orien-
tation of the body.
For a given solution of the equations of motion, the position vector of the central
point of the body can be found by quadratures
t

r = (X , Y , Z ) = + r0
ucp dt
0
t t
= p0 at+ p0 ( (t)dt)J− p0 ( (t)dt)K.
2
(10.62)
0 0
This yields the projections of the position vector of the origin of the body system
relative to the origin of the inertial coordinate system on the axes of the body system.
To express the position vector of the central point of the body referred to the inertial
system, we write
r = (r · α, r · β, r · γ).
That is
r = r R (10.63)
in terms of the rotation matrix R, which can be constructed using the expressions
(10.57) and (10.58) as shown in Chap. 2.
In the rest of this chapter we shall deal with the equivalent problem, returning to
the original problem only occasionally, when some important assertions are to be
made concerning the original problem. This is made here as a way of accommodating
the problem of motion of a body in a liquid in the hierarchy on the top of problems
of the previous chapters. A higher level in this hierarchy will be added in the next
chapter.
10.7 Uniform Translational-Rotational Motion of a Body in

a Liquid (Permanent Rotations of a Body with a Fixed
Point About a Vertical Axis)
We now put forward a more general motion than that of Sect. 10.1.2, to find all
possible permanent stationary (time-independent) motions. That is all solutions of
(10.45) with the pair (ω, γ) constant in the body and also in space. Substituting
ω̇ = γ̇= 0 in (10.45), we get
ω × (ωI+κ + γ K̄) = γ×(a+γJ),

ω × γ = 0. (10.64)
@seismicisolation
@seismicisolation
10.7 Uniform Translational-Rotational Motion of a Body … 249
From the second equation, we can express the angular velocity in the form
ω = ω0 γ, (10.65)
where ω0 is some proportionality constant and inserting this in the first equation, we
obtain
γ×[ω02 γI+ω0 (κ + γ K̄) − (a+γJ)] = 0. (10.66)
This condition determines the vector γ, which characterizes the possible stationary
motion in the sense that at any moment the body rotates about an axis parallel to γ
and passing through the proper central point.
Equation (10.66) has an obvious and direct geometric meaning:
For each, arbitrarily given, real ω0 , the vector γ characterizing the possible sta-
tionary motion lies along one of the lines drawn from the proper central point to
intersect at right angle the surface
1
Φ= γ(ω02 I+ω0 K̄−J) · γ + (ω0 κ−a) · γ==const. (10.67)
2
Here fl is an inhomogeneous quadratic function. The surface is a quadric referred to
an origin (the proper central point of the body) different from its centre. In the case
of a simply connected body κ= a = 0 and then γ becomes one of the eigenvector of
the matrix ω02 I − 2ω0 K̄−J, which are known to be three in number and orthogonal to
each other (See Sect. 10.1.2). The same conclusion can be reached also when κ, a are
non-zero parallel vectors and ω0 is chosen such that ω0 κ−a = 0. In the general case
of a Multiply connected (perforated) body no such general rule can be stated. When
this surface is an ellipsoid, for arbitrary ω0 , κ and a only two lines are guaranteed
to be drawn from the origin to intersect the surface orthogonally. Those are points
on the surface, nearest and farthest from the origin. Only one such line is guaranteed
when the surface is one-sheeted and extending to infinity.
The vector Eq. (10.66) can be written in the form of three scalar equations, but
only two of those equations are independent. In fact, multiplying (10.66) scalarly by
each of the vectors γI and (a + γJ), one obtains two different expressions for the
angular speed
γ · [γI × (a+γJ)] γ · [(a+γJ) × (κ+γ K̄)]

ω0 = = . (10.68)
γ · [γI × (κ+γ K̄)] γ · [γI × (a+γJ)]
Equality of the two expressions for ω0 determines the locus of the vector γ in the
form
{γ · [γI × (a+γJ)]}2 −
− {γ · [γI × (κ+γ K̄)]}{γ · [(a+γJ) × (κ + γ K̄)]}
= 0. (10.69)
@seismicisolation
@seismicisolation
This equation is non-homogeneous of degree six, and it represents a surface fixed

in the body. This surface intersects the Poisson sphere in some spherical curve. The
line joining the fixed point to each point of that spherical curve generates the cone of
possible axes of permanent rotations. One readily recognizes the following special
cases:
(1) From (10.66) we find that pure translations (ω0 = 0) are possible if and only
if γ is a generator of the cone a · (γ × γJ) = 0. This equation resembles that
of Staude’s cone, except for replacing the inertia matrix I by the matrix J, that
appears in the potential.
(2) For a simply connected body (κ = a = 0) Eq. (10.69) of degree six in γ
becomes homogeneous, and hence represents a cone. This result was obtained
by Minkowski [284] in 1888.
(3) For a gyrostat moving about a fixed point in a uniform gravity field, J = K̄ = 0,
(10.69) becomes, as already seen in Chap. 5,
[a · (γ×γI)]2 − [κ · (γ × γI)][a · (κ × γ)] = 0.
(4) In the special case collinear gyrostatic momentum and centre of mass and propor-
tional matrices K̄, J such that K̄ = J, κ = a, the cone of permanent rotation
axes reduces to Staude’s cone for the classical problem. Shortly below, we shall
see that this is a result of certain symmetry of the equations of motion, which
allows for a rotation transformation.
(5) For the classical problem of motion of a body (J = K̄ = κ = 0) it gives Staude’s
cone described by the equation a · (γ×γI) = 0 [343].
Remark The above analysis applies mostly to the equivalent problem of motion
of a rigid body about a fixed point under the action of potential and gyroscopic forces,
described by the equations of motion (10.45) or (10.54). In the problem of motion
of a body in a liquid, as explained above, the body rotates with the constant angular
speed ω0 about an axis parallel to γ and passing through the proper central point,
while the latter moves with the uniform velocity
ucp = p0 a+ p02 γJ − p0 ω0 γK). (10.70)
The position vector of the central point of the body can be expressed in the form
t
rcp = rcp0 + ucp dtR
0
t
= rcp0 + ucp dtR
0
⎛ ⎞
sin(ω0 t) 1 − cos(ω0 t) 0
1
= rcp0 + [a+γ(J − ω0 K)] ⎝ cos(ω0 t) − 1 sin(ω0 t) 0 ⎠.
ω0 0 0 ω t 0
(10.71)
@seismicisolation
@seismicisolation
10.8 Stationary Motions About an Axis Inclined to the Vertical 251
10.8 Stationary Motions About an Axis Inclined

to the Vertical
Unlike the classical problem and its generalization to the heavy gyrostat, the problem
of motion of a body in a liquid admits another type of motions in which ω is constant
(in space and in the body), but in a direction different from that of γ. Let us take the
z-axis of the body coordinate system along that direction. One can write
ω = k, = const . (10.72)
Equations of motion give
k × (kI+κ + γ K̄) = γ × (a + γJ),

γ̇ + k × γ = 0. (10.73)
The motion can be described by Euler’s angles in the usual notation: ψ = ψ0 , θ =

θ0 , φ = t. In virtue of the symmetry of the problem about the Z -axis and without
loss of generality, one can take ψ0 = 0. The unit vector γ can be expressed as
γ = (sin θ0 sin t, sin θ0 cos t, cos θ0 ). (10.74)
This can be easily shown to satisfy the second equation in (10.73). Substituting in
the first equation and equating coefficients of similar terms in powers of sin t and
cos t, we arrive at the following set of conditions:
J12 = J13 = J23 = 0, J22 = J11 ,

a1 = a2 = 0, a3 + K̃ 11 − (J11 − J33 ) cos θ0 = 0,
K̄ 12 = 0, K̄ 22 = K̄ 11 ,
κ1 + I13 + cos θ0 K̄ 13 = 0, κ2 + I23 + cos θ0 K̄ 23 = 0. (10.75)
In the generic case, we can write

⎛ ⎞ ⎛ ⎞
J11 0 0 K̄ 11 0 K̄ 13
J = ⎝ 0 J11 0 ⎠ , K̄ = ⎝ 0 K̄ 11 K̄ 23 ⎠ ,
0 0 J33 K̄ 13 K̄ 23 K̄ 33
a = (0, 0, − K̄ 11 + cos θ0 (J11 − J33 )),
κ = (−I13 − cos θ0 K̄ 13 , −I23 − cos θ0 K̄ 23 , κ3 ). (10.76)
where J11 , J33 , K̄ 11 , K̄ 33 , K̄ 13 , K̄ 23 , I13 , I23 , κ3 are arbitrary parameters. Note that
the axis of a permanent rotation of the present type must be a principal axis of the
matrix J, while the eigenvalues corresponding to the other two principal axes are
equal. The virtual centre of mass (the vector a) should also lie on the axis of rotation.
@seismicisolation
@seismicisolation
The last two Eq. (10.76) determine the pair of vectors a,κ depending on the angular
velocity and the angle θ0 , which can be given arbitrary values. On the other hand,
if one regards a,κ as given parameters, the last equations reduce to three equations
in two unknowns and θ0 , for the solution of which a condition on the parameters
of the body should be imposed. The parameters of the body must also satisfy the
obvious restriction |cos θ0 | ≤ 1. The solution of (10.75) exhibits several special and
degenerate cases, some of which will be summed up in the exercises.
10.9 A Several-Parameter Particular Solution
A result, which will be presented in the next chapter, was obtained in [389] in the
context of the problem of motion of a body about a fixed point under the action of
an axially symmetric combination of forces. A special case of this result reduces to
a case of motion of a body in a liquid (in fact, the alternative problem) and gives
a quite general particular solution of that problem, in the sense of the number of
parameters retained in it. That is a solution satisfying three invariant relations. It can
be formulated as the following
Theorem 10.1 Let in (10.45)
J = −MIM + αM + εδ,
K̄ = βM−αδ + tr(MI)δ − IM − MI,
κ = m(βδ − I),
a = m(αδ − IM). (10.77)
where M, m are a constant real 3 × 3 symmetric matrix and a vector, respectively,

α, β, ε are constants. Then,
(1) (10.45) admits a solution, which satisfies the relations
ω = γM + m, (10.78)
and γ is a solution of Poisson’s equation, which now takes the form
γ̇+(γM + m)×γ= 0. (10.79)
(2) In the generic case, the solution (ω and γ) is expressed in terms of elliptic
functions of time.
Proof (1) On substituting (10.77)–(10.79) into (10.45) and using the identity in
Appendix 11.1, the first equation turns into identity.
(2) Assuming that det(M) = 0 and using the relation inverse to (10.78), one can
write Eq. (10.79) as
@seismicisolation
@seismicisolation
10.9 A Several-Parameter Particular Solution 253
ω̇M−1 + ω × (ωM−1 − mM−1 ) = 0.
This is the equation of motion of a free gyrostat with inertia matrix M−1 and
gyrostatic momentum −mM−1 . This characterizes the case discussed in Chap. 5
under the name of Joukovsky–Volterra’s case. Referring equations to the prin-
cipal axes of M−1 by a suitable rotation, the solution is determined in terms of
elliptic functions of time.
This completes building for the alternative problem (10.45), which involves 21
parameters, a solution depending on 15 of those parameters. One can also go back
through (10.43) to build the relevant solution of the problem of motion of a body in
a liquid (Eq. (10.41)) and including the parameter p0 and the three elements of the
anti-symmetric part of B.
The solution established by theorem 1 generalizes by the presence of several extra-
parameters a former solution obtained by Kharlamov in [197]. It also generalizes
another solution obtained by Kharlamova [205], while studying the motion of a rigid
body about a fixed point in an approximate Newtonian field without gyroscopic
forces, except the constant gyrostatic momentum. The choices in both works [197,
205] correspond to a matrix M which is diagonal and hence commuting with I. The
relation of the variables to time was established only in some special cases, where
very restrictive conditions were imposed on the parameters. Much older partial results
were obtained by Steklov, who considered the case m = 0 and M diagonal in the
principal axes of inertia [345].
Solutions on invariant relations of the general form (10.79) (with non-diagonal
M) were considered in the much later papers [128, 129] (See also [125]). In those
papers, no reference is made to our relevant result in [389], published 14 years
earlier. Conditions that the dynamical equations of motion are satisfied along with
the given invariant relations are obtained by the (brute force) method of solving
algebraic equations. Expressions obtained in [129] are not transparent, and there is
no comparison with previous results.
10.10 Alternative Hamiltonian Formulation
Equations of motion (10.45) can be put in Hamiltonian form in two ways. On one
hand, using canonical variables, such as Euler’s angles and momenta conjugate to
them is hopelessly complicated for analytical considerations. On the other hand,
one can introduce the angular momentum of the system described by the Routhian
(10.27)
∂R
M= = ωI + κ + γ K̄, (10.80)
∂ω
as phase variable instead of ω, so that
ω = (M − κ − γ K̄)I−1 . (10.81)
@seismicisolation
@seismicisolation
The Hamiltonian corresponding to the Lagrangian (10.49) as a function in M and γ

is
1 1
H= ωI · ω + a · γ+ γJ · γ
2 2
1 −1 1
= (M−κ − γ K̄)I · (M − κ − γ K̄) + a · γ+ γJ · γ
2 2
1 −1 −1
= MI · M−(κ + γ K̄)I · M
2
1
+ (a + κI−1 K̄ T ) · γ + γ(J + K̄I−1 K̄ T ) · γ, (10.82)
2
so that the equations of motion can be written as
∂H ∂H
Ṁ = M× + γ× ,
∂M ∂γ
∂H
γ̇ = γ× . (10.83)
∂M
Note that
∂H
= (M − κ − γ K̄ )I−1 = ω,
∂M
∂H
= a+γJ − (M−κ − γ K̄)I−1 K̄ T ,
∂γ
or, in the expanded form
∂H
Ṁ = M×(M − κ − γ K̄)I−1 + γ× ,
∂γ
γ̇ = γ × (M − κ − γ K̄)I−1 . (10.84)
For the Hamiltonian equations, the integrals of motion take the simplest form
I1 = H = h,
I2 = M · γ = f,
I3 = γ · γ = 1. (10.85)
This situation makes use of the Hamiltonian form of equations favourable in certain
situations. However, in other situations and for most of our purposes, the Lagrangian
formalism of the equations of motion remains the favourable choice.
Throughout this book, we adhere to the use of Lagrangian formalism. We owe
the reader some explanation for that. In early times of Hamiltonian mechanics, the
formulation of mechanical problems stemmed directly from the physical setting. In
the Hamiltonian describing the motion of a particle under the action of certain forces,
each term of the Hamiltonian usually had its definite and unambiguous meaning. The
@seismicisolation
@seismicisolation
10.10 Alternative Hamiltonian Formulation 255
situation in modern days is different. Some integrable Hamiltonians of the structure

(10.82) were recently obtained by searching the relevant coefficients in a general
ansatz. Two factors come into play even in the simpler cases when the description
of motion is given a priori in Hamiltonian form:
(1) The Hamiltonian and the equations of motion derived from it are not unique
for one and the same physical problem. This makes classification of integrable
cases in Hamiltonian formulation rather problematic. As a matter of fact, to
decide whether two Hamiltonians describe the same mechanical system practi-
cally reduces to computing the functions V and μ as the quantities that remain
invariant under all gauge transformations (canonical transformations linear in
momenta in Hamiltonian terms).
(2) In the Euler–Poisson variables, it is possible to tell about physical interpretation
of various terms of the potential. For example, terms linear in γ represent the
potential of the heavy body in a constant gravity field. The centre of mass of the
body is uniquely determined by terms in the Lagrangian linear in γ. Other terms
can be identified as a result of gravitational, electric or magnetic potential, but in
the transformed Hamiltonian form terms of various degrees are totally disguised.
To illustrate the above points, we use as an example the case introduced by Sokolov
[336] with the Kowalevski configuration A = B = 2, C = 1. The original Hamilto-
nian describing this case is
1 2 1 1
H1 = (M1 + M22 + 2M32 ) + M3 (c1 γ1 + c2 γ2 ) + γ3 (c1 M1 + c2 M2 )
4 2 2
+ (c1 γ2 − c2 γ1 )2 − (c12 + c22 )γ32 . (10.86)
Calculating the Lagrangian corresponding to this Hamiltonian (10.86) and using the
Legendre transformation
∂ H1
ω ≡ ( p, q, r ) =
∂M
1 1 1 1 1
= ( M1 + c1 γ3 , M2 + c2 γ3 , M3 + (c1 γ1 + c2 γ2 )), (10.87)
2 2 2 2 2
we find
∂ H1
L1 = M · − H1
∂M
r2
= p2 + q 2 +
2
1 1
− c1 ( pγ3 + r γ1 ) − c2 (qγ3 + r γ2 )
2 2
9 1
− [(c1 γ2 − c2 γ1 )2 − (c12 + c22 )γ32 ] + (c12 + c22 )(γ12 + γ22 + γ32 ). (10.88)
8 8
@seismicisolation
@seismicisolation
In the last expression one can eliminate its last term, since the spherically symmetric
term does not contribute to the equations of motion. Thus, the Lagrangian L 1 has a
potential part
9
V = [(c1 γ2 − c2 γ1 )2 − (c12 + c22 )γ32 ]. (10.89)
8
On the other hand, the gyroscopic terms of L 1 correspond to the choice of the vector
l as
1 1
l = −(c1 γ3 , c2 γ3 , c1 γ1 + c2 γ2 ).
2 2
This uniquely determines the vector
3
μ = −∇[ (c1 γ1 + c2 γ2 )γ3 ]
2
3
= − (c1 γ3 , c2 γ3 , c1 γ1 + c2 γ2 ). (10.90)
2
The mechanical system under consideration is completely characterized by the pair
of functions V and μ. The complementary integral of motion in the Euler–Poisson
variables can be written as
I4 = Z 1 Z 2 , (10.91)
Z 1 = (r − 1/2 a1 γ1 − 1/2 a2 γ2 ) ,
Z 2 = 1/4 (2r − a1 γ1 − a2 γ2 ) [4 p 2 + 4 q 2 + (2r − a1 γ1 − a2 γ2 )2 ]

+ 2 (2 pa1 + 2 qa2 ) (2 pγ1 + 2 qγ2 ) + (a1 γ1 + a2 γ2 ) (2r − a1 γ1 − a2 γ2 )2
+ 1/2 (a1 γ1 + a2 γ2 )2 (2r − a1 γ1 − a2 γ2 )

− 1/2 a12 + a22 γ3 [8 pγ1 + 8 qγ2 + γ3 (2r − a1 γ1 − a2 γ2 )]. (10.92)
where we have set a1 = 3c1 , a2 = 3c2 .

An alternative Hamiltonian describing the same system was introduced by Borisov
and Mamaev [39], using a linear transformation of the phase variables, which pre-
serves the Poisson brackets and simplifies the Hamiltonian to
1 2 1 1
H2 = (M1 + M22 + 2M32 ) + M3 (a1 γ1 + a2 γ2 ) + (a12 + a22 )(γ12 + γ22 ).
4 2 4
(10.93)
The relation between ω and M for this Hamiltonian is
∂ H2
ω=
∂M
1 1 1
= ( M1 , M2 , M3 + (a1 γ1 + a2 γ2 )), (10.94)
2 2 2
@seismicisolation
@seismicisolation
10.10 Alternative Hamiltonian Formulation 257
and by direct calculation of the corresponding Lagrangian
r2
L 2 == p 2 + q 2 +
2
1
− r (a1 γ1 + a2 γ2 )
2
1
+ [(a1 γ2 + a2 γ1 )2 + 2(a12 + a22 )γ32 ]. (10.95)
8
Note that its second line gives
1
l = (0, 0, (a1 γ1 + a2 γ2 )),
2
which leads to the same μ as in (10.90). The two Lagrangians L 1,2 are in fact
equivalent. Their difference is
L 1 − L 2 = c1 (r γ1 − pγ3 ) + c2 (r γ2 − qγ3 ) − (c12 + c22 )(γ12 + γ22 + γ32 )

d
= (c2 γ1 − c1 γ2 ) − (c12 + c22 )(γ12 + γ22 + γ32 ),
dt
i.e a gauge term and a central potential term. Both terms do not contribute to the
equations of motion.
In contrast to the clarity and the physical relevance of the Lagrangian approach,
none of the Hamiltonians H1 and H2 reflects the real nature of the potential (The terms
quadratic in γi ). Physical characteristics of the mechanical system are disguised in
Hamiltonian form.
On the other hand, different Hamiltonian equations of motion are obtained using
the Hamiltonians H1 , H2 . Also, each form of the Hamiltonians corresponds to a dif-
ferent form of the complementary integral, which can be constructed by substituting
(10.87) and (10.94), respectively, in (10.91).
The change of the phase variables {M,γ} → {M̄, γ} which transforms H1 into
H2 can be obtained by comparing (10.94) with (10.87), in the form
M̄1 = M1 + c1 γ3 , M̄2 = M2 + c2 γ3 , M̄3 = M3 − c1 γ1 − c2 γ2 . (10.96)
This is identical to the (canonical) transformation given by Borisov and Mamaev in

[39] (See also [336]).
10.11 The Uniform Precession Transformation [383]
In its full final form (10.45), the problem of motion of a body in a liquid is at the
top of a hierarchy, consisting of the problems considered in the previous chapters,
@seismicisolation
@seismicisolation
involving the gyrostatic effect, the Newtonian potential term and the uniform gravity
field as special cases. Consequently, every integrable case or solution of the above
problems may have a generalization in the frame of the present one. This situation
will be made clear in the tables of integrable cases provided below in this chapter.
A remarkable feature of Eq. (10.45) for the body in liquid, which is not enjoyed
by any of the three simpler problems of Sects. 10.3–10.6, is their invariance under
the uniform precession transformation, which we are going to describe now. This
transformation was firstly introduced for the problem of motion of a body in a liquid
in [383].
10.11.1 Direct Derivation
In the equations of motion (10.45), we perform the transformation of the variables

ω to a new set of variables ω by the relation
ω = ω −nγ, (10.97)
containing the free real parameter n. Substituting in (10.45) we obtain
(ω̇ −n γ̇)I+(ω − nγ) × [(ω − nγ)I + κ + γ K̄) = γ × (a + γJ),

γ̇ + (ω −nγ) × γ = 0. (10.98)
ω̇ I+ω ×(ω I+κ + γ K̄ − nγI) = n(γ̇I + γ×ω I)+γ × (a + γJ)

+nγ × ( − nγI + κ − 2γ K̄),

γ̇ + ω × γ = 0. (10.99)
Using Poisson’s equation to express γ̇ and noting that
(ω ×γ)I + ω I × γ = ω ×γ[tr (I)δ − I]
we give (10.99) the form
ω̇ I+ω ×(ω I+κ + γ K̄+2nγ Ī) = γ×[a+nκ + γ(J − 2n K̄ − n 2 I)],

γ̇ + ω × γ = 0, (10.100)
which can be readily put in the final form
ω̇ I+ω ×(ω I+κ+γ K̄ ) = γ×(a +γJ)

γ̇ + ω × γ = 0, (10.101)
@seismicisolation
@seismicisolation
10.11 The Uniform Precession Transformation [383] 259
after introducing the notation
K = K + 2nI, K̄ = K̄ + 2n Ī,

a = a + nκ,
J = J − nK − n 2 I. (10.102)
A look at the two sets of Eqs. (10.45) and (10.101) reveals that they have the
same structure in terms of the two sets of variables {ω, γ} and {ω , γ} and that they
differ only in the values of parameters a, J, K, which are transformed to a , J , K ,
respectively, containing the extra-parameter n. When one sets n = 0, ω = ω and
the three primed matrix-parameters reduce to their original (unprimed) values. It
is an easy exercise to show that the consecutive application of two transformations
with parameters n 1 , n 2 is equivalent to the application of one transformation with
the parameter n 1 + n 2 .
10.11.2 Lagrangian Derivation
Consider the problem described by the equations of motion (10.45) derived from the
Lagrangian (10.49). Let us affect the transformation. It can be readily checked that
this transformation changes the Lagrangian (10.49) to the similar form
1
L = ω I · ω + l .ω − V , (10.103)
2
where
1 1
V = V + nl · γ − n 2 γI · γ,V = a · γ+ γJ · γ,
2 2
1
l = l−nγI, l = κ − γK. (10.104)
2
and renders the equations of motion (10.45) to
∂V
ω̇ I + ω ×(ω I + ¯ ) = γ× ,
∂γ
γ̇ + ω ×γ = 0 (10.105)
where
μ = μ + 2nγ Ī
= κ + γ K̄+2nγ Ī.
@seismicisolation
@seismicisolation
Equation (10.105) admit the integrals
1 1
I1 ≡ ω I · ω + a · γ+ γJ · γ = h ,
2 2
1
I2 = (ω I + κ − γ K̄ ) · γ = f ,
2
I3 = γ 2 = 1, (10.106)
with the constants

h = h + n f, f = f. (10.107)
10.11.3 Physical and Mechanical Significance

of the Transformation
The transformation (10.97) was used by Tisserand in [353] (See also [354]) to illus-
trate the effect of Coriolis and centrifugal forces on the motion of a rigid body with
one point fixed on the rotating earth. It was implicitly used by other authors (e.g.
[21, 51]) while studying the stability of relative equilibria in the problem of motion
of a satellite in a circular orbit. It was applied only to the integrals of motion, but the
transformed equations were not obtained and the full significance of the transforma-
tion was not revealed.
The transformation was applied for the first time to the full equations of motion
of a charged and magnetized body in [378], where all its properties were revealed. It
was also applied to the problem of motion of a satellite in a circular orbit to obtain
its equations of motion relative to the orbital frame [382], in a form that resembles
equations of motion of a rigid body about a fixed point under the action of given
potential and gyroscopic forces. The invariance of the equations of motion of the
body in a liquid under this transformation was first recognized in our work [383].
The presence of the parameter n in the transformed Lagrangian and the trans-
formed equations of motion, in the framework of the equivalent physical problem,
turns on a simultaneous combination of three physical effects:
(1) The effect of displacing the centre of mass of the body by an amount nκ, pro-
portional to the gyrostatic moment.
When a is proportional to κ, say, a = mκ, then one can choose n = −m, so

that a = 0 and thus getting rid of the uniform gravity field in the transformed
problem.
(2) The matrix K is changed by the amount nI. This can be interpreted as the matrix
of coefficients in the vector potential of a static, on the body, charge distribution
@seismicisolation
@seismicisolation
whose inertia matrix is proportional to the inertia matrix of the distribution of

mass in the body and subject to the Lorentz forces due to a uniform magnetic
field of intensity B = −2n in the direction of γ.
If the matrix K is proportional to I, say, K = mI, then the regular precession

transformation can be used to make K = 0 by taking n = m.
(3) The matrix J of coefficients of the quadratic part of the potential is modified
by adding two terms: −nK − n 2 I. The potential resulting from those terms can
be interpreted as due to magnetized (electrically charged) parts of the body
influenced by a magnetic (electric) field with second-harmonic potential.
In the next sections, we shall use the uniform precession transformation in the two
ways: to construct more general solutions containing the parameter n from known
simpler ones and to simplify some other cases of motion by using that parameter to
reduce the number of physical constants in them, in order to facilitate obtaining their
solutions.
As a quick illustrative example, we work out an explicit solution of Euler’s case
generalized by the uniform precession transformation. For the transformed motion
1
V = − n 2 γI · γ, μ = 2nγ Ī. (10.108)
2
The equations of motion for this case take the form
ω̇I + ω × (ωI + 2nγ Ī) = −n 2 γ×γI,

γ̇ + ω × γ = 0. (10.109)
They admit the complementary integral
A2 ( p + nγ1 )2 + B 2 (q + nγ2 )2 + C 2 (r + nγ3 )2 = G 2 . (10.110)
On one hand, Eq. (10.109) characterizes an integrable case of motion of the body
in a liquid, which lies on the intersection of the cases of Clebsch and Steklov (See
Table 10.2 below). On the other hand, in the framework of the equivalent problem,
they describe the motion of a body under the influence of potential and Lorentz’
forces. A family of explicit solutions of this case2 can be written down immediately
by transforming the solution constructed for Euler’s case in Chap. 4 Sect. 4.1.
2In Euler’s case we have solved only the dynamical equations of motion and adopted a very special
solution of Poisson’s equations in which the vectors γ and G are parallel. Hoever, the transformation
applies equally well to the general solution of thewhole Euler-Poisson system.
@seismicisolation
@seismicisolation

n D(D − C)
p = ±(μ − A) cnλ(t − t0 ),
D A(A − C)

n D(D − C)
q = (μ − B) snλ(t − t0 ),
D B(B − C)

n D(A − D)
r = ±(μ − C) dnλ(t − t0 ), (10.111)
D C(A − C)

A(D − C)
γ1 = ± cnλ(t − t0 ),
D(A − C)

B(D − C)
γ2 = snλ(t − t0 ),
D(B − C)

C(A − D)
γ3 = ± dnλ(t − t0 ), (10.112)
D(A − C)
where λ, D, μ and k, the modulus of elliptic functions, are the same as in Chap. 4
Sect. 4.1. The motion of the body is quite different from that in Euler’s case. For
example, choosing n = μD A
we make p ≡ 0. The angular velocity lies permanently
in the yz-plane, and it is still expressed in elliptic functions of time.
10.11.4 Uniform Precession Transformation in Hamiltonian

Formalism
The expression (10.82) gives the Hamiltonian corresponding to the Lagrangian

(10.49). Let H be the Hamiltonian corresponding to the Lagrangian L in (11.6), i.e.
the Lagrangian obtained from (10.49) by the replacement ω→ω = ω+nγ. It can
be shown by direct calculation that
H = H + nM · γ. (10.113)
The precession transformation is equivalent to adding the term nM · γ to the Hamil-

tonian, which is the precession parameter n multiplied by the areas integral (the
second integral in (10.85)). The transformed Hamiltonian is a constant of motion
H = h = h + n f, (10.114)
in agreement with (10.107). We also have
@seismicisolation
@seismicisolation
∂H
ω =
∂M
∂H
= + nγ
∂M
= ω + nγ. (10.115)
Moreover, it is easy to show that the transformed Hamiltonian H in (10.113)

produces the same equations of motion as the original Hamiltonian H. In fact, using
H in (10.83), we obtain the equations
∂H ∂H
Ṁ = M× + γ×
∂M ∂γ
∂H ∂H
=M × ( + nγ) + γ×( + nM)
∂M ∂γ
∂H ∂H
= M× + γ× ,
∂M ∂γ
∂H
γ̇ = γ×
∂M
∂H
= γ×( + nγ)
∂M
∂H
= γ× , (10.116)
∂M
which are identical to the original Eq. (10.83).
In contrast to the transformed Lagrangian (10.103), the transformed Hamilto-
nian (10.113) does not reveal any of the physical effects of the uniform precession
transformation, which we listed in the last subsection. Thus, the part of the physical
effects induced by the uniform precession transformation in the problem is completely
hidden by the Hamiltonian form of the equations of motion. The Hamiltonian for-
malism identifies the whole family of mechanical systems depending on the arbitrary
parameter n into a single Hamiltonian system. The Hamiltonian flow on the integral
manifold of that system is the same for all physical problems, which differ only in
the value of n. In the problem of motion of a body in a liquid that is a family of
bodies with differing shape characteristics, but in the alternative problem it means a
body subject to a family of potential and gyroscopic forces depending on n. As usual
in the search for integrable cases, one assumes only the Hamiltonian form of the
equations of motion and tries to determine the coefficients in a general form (ansatz)
of the Hamiltonian and the complementary integral. In Hamiltonians constructed
in this way, a term of the form nM · γ is missing and the dynamical behaviour of
the original physical system will be determined up to a precessional motion with a
constant rate.
@seismicisolation
@seismicisolation
10.12 Generalization of General Integrable Cases
The most important consequence of the form-invariance of equations of motion

under the transformation (10.97) is the possibility it opens to generalize general
(unconditional) and conditional integrable cases and also particular solutions through
adding the precession parameter n into their structure, and thus enriching the physical
problem by adding new physical effects. In the present section, we formulate this
result for general integrable cases and give concrete examples of its applications.
Conditional and particular cases are discussed in the next sections.
Theorem 10.2 Let for some set of parameters I,κ, K , a , J , Eq. (10.105) admit a
complementary integral I4 = I4 (ω , γ), so that they become integrable for arbitrary
initial conditions, and let their solution be {ω = (t), γ = (t)}. Then Eq. (10.54)
are also integrable for arbitrary initial conditions, for the set of values of the param-
eters I,κ, K, a, J :
K̃ = K̃ − 2nI,
a = a −nκ,
J = J + nK + n 2 I, (10.117)
their complementary integral is
I4 = I4 (ω + nγ, γ), (10.118)
and their general solution is {ω = (t) + n(t), γ = (t)}. It contains the addi-
tional arbitrary real parameter n. When n = 0, the generalized solution renders to
the original solution.
Any one of the hierarchy of integrable cases provided in the previous chapters
admits a generalization as a case of the motion of a rigid body in a liquid. Transformed
cases are of the same type (general or conditional) as the original ones. Examples are
given in the next subsections: In Sokolov’s case, the introduction of the parameter
n results in a new integrable case. Even when the parameter n already enters in the
structure of a known system, like in the case due to Rubanovsky [317], the regular
precession transformation can be used to simplify the process of construction of
an explicit solution of the equations of motion in terms of time. In such cases, the
solution can be found first for the simpler case n = 0 and then generalized by that
transformation for arbitrary n by the formulas in the last theorem.
@seismicisolation
@seismicisolation
10.12 Generalization of General Integrable Cases 265
10.12.1 Generalization of the Integrable Case Found

by Sokolov
In 2002, Sokolov [336] obtained an integrable case of the rigid body in a liquid,
which adds a parameter c to a former case by Yehia [380] (1986). The body has the
Kowalevski configuration A = B = 2C. The centre of mass lies in the equatorial
plane. The functions V, l,μ and the integrals I3 and I4 are given, according to the
order followed in this book, as
c2 2
V = C[kcγ1 + a2 γ2 − (γ + 2γ32 )],
2 1
l = C(2cγ3 , 0, k − cγ1 ), μ = C(cγ3 , 0, k + cγ1 ), (10.119)
I3 = 2( pγ1 + qγ2 ) + (r + k + cγ1 )γ3 ,

2
I4 = p 2 − q 2 + a2 γ2 + c2 γ22 + cγ1 (r − k)

2
+ 2 pq − a2 γ1 − c2 γ1 γ2 + cγ2 (r − k)

+ 2k(r − k + cγ1 ) p 2 + q 2 + 2cpγ3
− 2kc2 {2γ3 [2 pγ1 + cγ1 γ3 + 2qγ2 + r γ3 ]
+ kγ32 − (γ12 + γ22 + 2γ32 )(r + cγ1 )} − 4a2 kqγ3 . (10.120)
It was pointed out in [411] (2003) that the parameter n can be added to this case
to produce a non-trivial generalization, represented by the formulas
c2 2
V = C[kcγ1 + a2 γ2 − nkγ3 − (γ + 2γ32 )
2 1
n2
− ncγ1 γ3 − (2γ12 + 2γ22 + γ32 )],
2
l = C(2cγ3 + 2nγ1 , 2nγ2 , k − cγ1 + nγ3 ),
μ = C(cγ3 − nγ1 , −nγ2 , k + cγ1 − 3nγ3 ), (10.121)
I3 = 2( pγ1 + qγ2 ) + (r + k + cγ1 )γ3 + +
n(2γ12 + 2γ22 γ32 ),

2
I4 = ( p + nγ1 )2 − (q + nγ2 )2 + a2 γ2 + c2 γ22 + cγ1 (r + nγ3 − k)

2
+ 2( p + nγ1 )(q + nγ2 ) − a2 γ1 − c2 γ1 γ2 + cγ2 (r + nγ3 − k)

+ 2k(r + nγ3 − k + cγ1 ) ( p + nγ1 )2 + (q + nγ2 )2 + 2c( p + nγ1 )γ3
− 2kc2 {2γ3 [2( p + nγ1 )γ1 + cγ1 γ3 + 2(q + nγ2 )γ2 + (r + nγ3 )γ3 ]
+ kγ32 − (γ12 + γ22 + 2γ32 )(r + nγ3 + cγ1 )} − 4a2 k(q + nγ2 )γ3 . (10.122)
@seismicisolation
@seismicisolation
Comparing (10.121), (10.122), (10.119), (10.120) we note that, unlike in Sokolov’s

case, the centre of mass in the generalized case does not lie in the equatorial plane
since it has three non-zero coordinates. Also, the vector potential l and the gyroscopic
vector μ do not lie in a meridional plane as in Sokolov’s case.
Remark. On the other hand, the regular precession transformation can be used
in the reverse direction. For example, to seek the explicit solution of the equations
of motion or to study the stability of a given motion, it suffices to study equations of
motion for the Sokolov case. The solution may be extended to the generalized system
with the extra-parameter n and the conclusion about stability will be the same before
and after adding this parameter to the system.
10.12.2 Steklov’s Case and Its Generalizations
One of the first known integrable cases of Kirchhoff equations (The version of (10.45)
with κ = a = 0)
ω̇I+ω×(ωI + γ K̄) = γ×γJ,

γ̇ + ω × γ = 0. (10.123)
describing the motion of a rigid body with a singly connected surface in a liquid was
discovered in 1895 by Steklov [345]. It corresponds to the choice
J = 0. (10.124)
Using (10.123), (10.124) we calculate the derivative
d 1
{ |ωI|2 + γ K̄ · (ωI)} = −ω · [(ω×γ)K̄I].
dt 2
Under the condition
K̄I = εδ,
where δ is the unit matrix, or, equivalently,
K̄ = εI−1 (10.125)
the right-hand side of the last equation vanishes, which leads to Steklov’s comple-
mentary integral of motion
1
|ωI|2 + εω · γ = const . (10.126)
2
This is the classical case of Steklov. When ε = 0, it turns into Euler’s case.
@seismicisolation
@seismicisolation
10.12 Generalization of General Integrable Cases 267
Kharlamov [192, 197] investigated the full equations of motion which describe
the problem of motion of a body with a multi-connected surface equivalent to (10.45)
but in certain modified Clebsch variables. We write them in our form (10.45)
ω̇I+ω×(ωI+κ + γ K̄) = γ×(a+γJ),

γ̇ + ω × γ = 0. (10.127)
In those equations, the analog of the gyrostatic momentum and the centre of mass
are present. Using what can be called a “brute force” method, Kharlamov found a
generalization of Steklov’s result under the conditions
a1 + nκ1 = a3 + nκ3 = 0, a2 = κ2 = 0 (10.128)
and expressed the complementary integral as a quadratic polynomial in the variables

[197]. Somewhat later, Rubanovsky [320] by a similar method replaced Kharlamov’s
non-symmetrical conditions a2 = κ2 = 0 by the less restrictive and more symmetric
condition
a2 + nκ2 = 0, (10.129)
so that the two vectors a and κ are now proportional, i.e.
a = −nκ. (10.130)
Let us now consider a system of equations of motion containing only the arbitrary
gyrostatic vector κ, in addition to Eq. (10.123) and take Steklov’s condition (10.128)
into account. We write them as
ω̇ I + ω ×(ω I + κ + εI−1 ) = 0,
γ̇ + ω × γ = 0. (10.131)
It can be easily verified that this system admits the following complementary integral,
which generalizes (10.126):
1
|ω I+κ|2 + εω · γ = const . (10.132)
2
Now we apply the regular precession transformation to the system (10.131) and
its integral (10.132). Equation (10.131) transforms to (10.127), in which
1
K̄ = − εI−1 + n Ī, J = −n 2 I, a = −nκ. (10.133)
2
This gives at once Rubanovsky’s generalization of Kharlamov’s result. It also enables
to write the complementary integral in the very simple and transparent way:
@seismicisolation
@seismicisolation
1
I4 = [ωI + nγI+κ]2 + εω · γ = const . (10.134)
2
Although this case was noted by Rubanovsky, the method used here helped write
the integral I4 in this simple form. A notable advantage of the regular precession
transformation is that one can use it here in the reverse way. To construct the explicit
solution in terms of functions in time, it is sufficient to do that for the case n = 0.
This means to construct the solution of the system (10.131) with the fourth integral
(10.132). Having completed this task, i.e. having found ω = (t), = (t), one can
just write down the solution for the generalized case
ω = (t) − n(t), γ = (t). (10.135)
Explicit time solution of the full case (10.131) is not constructed to the present
moment. This was achieved only in two particular cases:
(1) In Steklov’s case κ = 0, by Kötter in terms of theta functions of two variables
[235].
(2) In Joukovsky’s case ε = 0, the solution was obtained by Volterra [366] in terms of
Weierstrass functions, which are complex functions in t. An alternative solution
in terms of real Jacobi’s elliptic functions was constructed by Wittenburg [369].
Despite the interest in applying methods of modern algebraic geometry (e.g. [71]),
the general solution for the full basic case ε|κ|=0 was not considered.
10.13 Generalization of Conditional Integrable Cases
Theorem 10.3 Let for some set of parameters I,κ,K̄ , a , J , Eq. (10.105) be inte-
grable on the integral level I2 = f 0 with the complementary integral I4 = I4 (ω , γ),
and let their solution be {ω = (t), γ = (t)}. Then Eq. (10.54) is also integrable
on the same integral level I2 = f 0 , for the set of values of the parameters I,κ,K̃, a, J,
K = K − 2nI,
a = a −nκ,
J = J + nK + n 2 I,
their complementary integral is
I4 = I4 (ω + nγ, γ),
and their general solution on that level is {ω = (t) + n(t), γ = (t)}. It contains
the additional arbitrary real parameter n. When n = 0, the generalized solution
renders to the original solution.
@seismicisolation
@seismicisolation
10.13 Generalization of Conditional Integrable Cases 269
There are two conditionally integrable cases known presently in the problem of
motion of a body in a liquid. We now demonstrate how the uniform precession
transformation works on one of them, namely, the Goryachev–Chaplygin hierarchy
of cases. The second hierarchy is based on a conditional subcase of Kowalevski’s
case and the integrable problem of a body in a liquid found by Chaplygin. The last
problem will be treated in more detail later in this chapter.
10.13.1 Generalization of Goryachev–Chaplygin’s,

Sretensky’s and Sokolov–Tsiganov Cases
The first and most famous conditional integrable case of Goryachev and Chaplygin
of the classical problem (See Chap. 4 Sect. 4.4) was built in 1900–1901 for a body
satisfying the conditions A = B = 4C and z 0 = 0. This case was generalized through
the addition of a gyroscope along the axis of dynamical symmetry by Sretensky in
1963 [341]. For more details, see Chap. 5 Sect. 5.3. Sokolov and Tsiganov [337]
(2002) added two more parameters. In our way of writing, the last case corresponds
to the choice
1
V = C[a1 γ1 + a2 γ2 + (c2 γ1 − c1 γ2 )2 ], (10.136)
2
and
μ = C(c1 γ3 , c2 γ3 , κ + c1 γ1 + c2 γ2 ). (10.137)
This case can be readily generalized by the regular precession transformation to

include the parameter n as follows:
1
V = C[a1 γ1 + a2 γ2 − nκγ3 + (c2 γ1 − c1 γ2 )2
2
n2
− nγ3 (c1 γ1 + c2 γ2 ) − (4γ12 + 4γ22 + γ32 )],
2
μ = C(c1 γ3 − nγ1 , c2 γ3 − nγ2 , κ + c1 γ1 + c2 γ2 − 7nγ3 ). (10.138)
The transformation adds several terms to the potential, including linear and quadratic
terms, and some linear terms to μ. The complementary integral for the generalized
case is
1 1
I4 = (r − κ + c1 γ1 + c2 γ2 + nγ3 )[( p + nγ1 + c1 γ3 )2 + (q + nγ2 + c2 γ3 )2 ]
2 2
+ γ3 [(κc1 − a1 )( p + nγ1 ) + (κc2 − a2 )(q + nγ2 )]
1
+ γ32 [κ(c12 + c22 ) − c1 a1 − c2 a2 ]. (10.139)
2
@seismicisolation
@seismicisolation
Explicit time solution for this case is not found yet. But to find this solution, it
suffices to express solution for the special case n = 0, i.e. for the case found by
Sokolov and Tsiganov [337].
10.14 Generalizations of Particular Solvable Cases
All the twelve particular solvable cases of the classical problem presented in Chap. 8
can be immediately generalized by the regular precession transformation. All those
cases produce new cases in the problem of motion of a rigid body in liquid. The same
remark fully applies for all the known solvable cases of the motion of a gyrostat (See
Chaps. 13 and 14) and also any particular solution known in the problem of motion of
a body in a liquid. The parameter n can be added to all those cases, with all possible
implications on the nature of forces acting on the body.
We shall not make a complete list of those generalized cases, but we shall provide
some of the most illustrative examples. We first formulate the following theorems:
Theorem 10.4 Let for some set of parameters I,κ, K , a , J and initial conditions
ω = 0 , = 0 , the Eq. (10.105) admit a particular solution {ω = (t), γ = (t)},
then for the set of values of the parameters I, κ, K, a, J and for the initial condi-
tions ω = 0 + n 0 , γ = 0 Eq. (10.54) admit a particular solution {ω = ω(t) +
n(t), γ = (t)} containing the additional arbitrary real parameter n.W henn = 0,
the generalized particular solution renders to the original particular solution.
Corollary. Any motion of a body in a liquid whose angular velocity has the form
ω = ω1 γ + ω 2 , i.e. involves a component ω1 in the direction of γ, can be reduced
by using the transformation (10.97) with n = −ω1 to a motion with angular velocity
ω = ω2 and vise versa. In particular,
1- A uniform (permanent) rotation about a vertical axis can be reduced to a
position of equilibrium.
2- A regular precession with a vertical precession axis can be reduced to a perma-
nent rotation about the configuration axis (fixed in the body), which becomes fixed
in space.
3- The so-called semi-regular precession (composed of pendulum-like motion of
the rigid body about a horizontal axis and a uniform rotation about the vertical) can
be reduced to pendulum-like motion, parallel to a fixed plane.
Theorem 10.5 All properties of the first solution in the last theorem, like stability
in the sense of Lyapunov, stability in (or by) the first approximation, instability and
periodicity3 are passed to the second solution.
3 Here, periodicity relates only to the Euler-Poisson variables ω, γ. The motion is periodic relative
to the body system of axes. The motion can be periodic in space only under commensurability
condition between the periods of the relative and the precessional motions.
@seismicisolation
@seismicisolation
10.14 Generalizations of Particular Solvable Cases 271
The proof follows immediately from the fact that the stability, instability or
periodicity of one of the pairs of solutions {ω = (t), γ = (t)} and {ω = (t) +
n(t), γ = (t)} implies the same to the other pair.
This theorem allows a great simplification to the study of properties of motion, as
in the last corollary, a permanent rotation reduces to an equilibrium. Also, a regular
precession (a periodic motion) reduces to a uniform rotation. Note that the equations
of variation for the precession are periodic in time, while those for uniform rotation
are of constant coefficients and hence their analysis is much simpler.
We give here only a few illustrative examples to show how the transformation
can be used in the direct or in the reverse directions, to generalize a given case or to
simplify it. We present results partly in the framework of the problem of motion of
a rigid body about a fixed point and partly in the equivalent problem, according to
our analogy described earlier in this chapter, of motion of a body in a liquid.
10.14.1 Example 1. Equilibria and Permanent Rotations

About a Vertical Axis
It is evident that a position of equilibrium of the body governed by Eq. (10.45) can be
transformed by the regular precession transformation. The image for a given finite real
n is a permanent rotation about an axis fixed in the body and taking a vertical position.
Conversely, a permanent rotation can always be reduced to a relative equilibrium in
a coordinate system moving with the same precession speed as the body.
Equilibria. Consider an equilibrium position of the system (10.45). Those are the
solutions {ω = 0, γ = γ 0 } where γ 0 satisfies
γ 0 ×(a+γ 0 J) = 0. (10.140)
1. For the classical problem, when J = 0, a = 0, there are two equilibria: the
upper and lower equilibria of the centre of mass above or below the fixed point.
2. When a = 0, J = 0, there are six equilibria, in which one of the principal axes
of the matrix J is directed along or against the vector γ.
3. In the generic case γ 0 satisfies a relation a+γ 0 J = ˘γ 0 , so that
γ 0 = −a(J − λδ)−1 , (10.141)
where λ is a root of the sixth-degree equation

a(J − λδ)−1 2 = 1. (10.142)
In this case also we have a maximal number of six positions of equilibrium and
minimum number of two.
@seismicisolation
@seismicisolation
Permanent rotations. Permanent rotations were discussed in detail in Sect. 10.7.

In a coordinate system moving with the same precession speed as the body, the
permanent rotation looks like an equilibrium position, which is determined from the
Eq. (10.140) but with the transformed parameters a and J , i.e.
γ 0 × (a +γ 0 J ) = 0. (10.143)
Substituting those parameters from (10.102) into the last relation, we get the condition
for a permanent rotation as
γ 0 × [a + nκ + γ 0 (J−nK−n 2 I)] = 0. (10.144)
This equation can now be compared with the condition for the permanent rotation
(10.66). They become identical, provided we take n = −ω0 .
10.14.2 Example 2. Permanent Rotations About a Tilted Axis

and Precessional Motions About the Vertical
Consider a precessional motion, in which the angular velocity of the body is given
by
ω = 0 e + 1 γ, (10.145)
where 0 ,1 are constants and e is a unit vector fixed in the body at the fixed point
O. The body rotates about e with angular velocity 0 , while this axis precesses
about the vertical with angular velocity 1 . Using the transformation (10.97) with
the choice n = −1 , we have
ω = 0 e. (10.146)
The precessional motion is reduced by this transformation to a uniform rotation about

an axis fixed in the body and in space and inclined to the vertical at a fixed angle. That
is the permanent rotational motion described in Sect. 10.8. Similarly, a permanent
rotational motion with angular velocity (10.146) can be transformed by the inverse
transformation n = 1 to the precessional motion.
Solutions of the equations of motion corresponding to regular precessions were
investigated in [123] (See also [125]). The conditions for existence of such precession
are obtained in a quite complicated form (conditions (18) in [123]). Those conditions
can be easily shown to be equivalent to conditions (10.75) followed by the rotation
transformation (10.97). According to the said above, one could consider only uniform
rotational motions about an axis fixed in space and inclined to the vertical. The whole
class of precessional motions generated by transforming uniform rotations about an
inclined axis using (10.97) with the parameter n taking all real values are equivalent
@seismicisolation
@seismicisolation
to that rotation and, moreover, have the same properties, for example, as concerns
stability of the motions.
10.14.3 Example 3. generalization of grioli’s precession

[402, 405]
On a dynamical basis, Grioli established the possibility of a regular precession of the

heavy rigid body about a non-vertical axis under certain conditions on the parameters
of the body [138]. Guliaev derived the full explicit solution of this case [141] (see
also [256]). We present the necessary details in brief. The solution differs from that
of Guliaev only in that we have assigned a certain value for the initial time moment,
so that the solution becomes more transparent.
Let the axes be arranged such that A ≥ B ≥ C. For
a = (a, 0, c), κ= 0, K = J = 0, (10.147)

√ √
where a B − C = c A − B, the system of Eq. (10.105) admits a particular solu-
tion (See 8.10)

p = (a − c cos(t)), q = sin(t), r = (c + a cos(t)),
s s
2
γ1 = − 2 [Cc cos(t) + (B − C)a sin2 (t)],
s
2
γ2 = 3 sin(t)[(Aa 2 + Cc2 ) − (A − C)ac cos(t)],
s
2
γ3 = 2 [Aa cos(t) + (A − B)c sin2 (t)], (10.148)
s
√
where s = a 2 + c2 , 2 = √ s
. This solution corresponds to a
(A−B+C) +(A−B)(B−C)
2
regular precession of the body. The angular velocity ω can be written as the sum of
two terms
ω = ζ + α, (10.149)
where ζ, α are two unit vectors: the first fixed in the body (orthogonal to a circular
section of the inertia ellipsoid) and the second fixed in space [141], so that in the
body system
a c c a
ζ = ( , 0, ), α = ( − cos(t), sin(t), cos(t)). (10.150)
s s s s
Note that ζ is orthogonal to α and that α is inclined to the upward vertical vector γ
at a fixed angle δ,
@seismicisolation
@seismicisolation
A− B +C
cos δ = . (10.151)
(A − B + C)2 + (A − B)(B − C)
The body rotates with the uniform velocity around the vector ζ fixed in it, while
that vector rotates with the same velocity about the direction α fixed in space.
We now consider another case of motion of the same body as above, but we will
replace V , μ by
a = (a, 0, c), κ = 0, K = J = 0,
1
V = aγ1 + cγ3 − n 2 (Aγ12 + Bγ22 + Cγ32 ),
2
μ = n((B + C − A)γ1 , (C + A − B)γ2 , (A + B − C)γ3 ) (10.152)
where, for simplicity, n is taken as a constant. It is easy to verify that applying

the substitution ω = ω + nγ transforms (10.152) into (10.147). Thus, the system
with (10.152) admits a particular solution representing Grioli’s precession uniformly
rotated with speed n about the vertical. In this solution γ1 , γ2 , γ3 are the same as in
(10.148), while
n2
p= (a − c cos(t)) − 2 [Cc cos(t) + (B − C)a sin2 (t)],
s s
n2
q = sin(t) + 3 sin(t)[(Aa 2 + Cc2 ) − (A − C)ac cos(t)],
s
n2
r = (c + a cos(t)) + 2 [Aa cos(t) + (A − B)c sin2 (t)].
s s
(10.153)
This case is a non-trivial generalization of Grioli’s result [138]. The whole picture
of Grioli’s precession about the inclined axis precesses about the vertical at an arbi-
trary angular speed n. The resulting motion admits two interpretations as a motion
of a body in liquid [383] or a motion of a charged body under potential and Lorentz
forces as described in Sect. 2.2 above. It is noteworthy that this gives a new result in
both interpretations.
The angular velocity ω = (ζ + α) + nγ no longer has constant magnitude as
was the case in Grioli’s precession. The resulting motion is not a regular precession.
Although ω and γ are periodic functions of time, the motion is not in general peri-
odic in space for arbitrary values of n. However, if n is rational the body returns
periodically to its initial position. As far as we know, such motions have not been
considered previously. This solution can be generalized by adding a rotor to the body
along the normal to a circular cross-section of the ellipsoid of inertia of the body. The
solution for the last case was found by Keis [167] and rediscovered by Kharlamova.
The resulting case of motion of a heavy gyrostat can be transformed using the regular
precession transformation to a case of motion of another body by inertia in a liquid
or a case of motion of an electrically charged body in gravity and magnetic fields.
Formulas received will generalize (10.152), (10.153).
@seismicisolation
@seismicisolation
10.14.4 Example 4. Regularly Precessing Pendulum
By this motion, we mean a generalization of the motion of a physical pendulum, such

that the axis of the pendulum rotation performs regular precession about the vertical.
A near, but different, term “semi-regular precession” was coined by Grioli [139] in
certain problems lower in the hierarchy than that of the present chapter. The same
name was used later by many authors, e.g. [357]. (See also Gorr [119] and for more
detail [126]). This name refers to the motion in which the body rotates with a time-
dependent (i.e. non-uniform) angular velocity about an axis fixed in it, while this axis
makes regular precession about an axis fixed in space. Thus, the regularly precessing
pendulum motion is a type of semi-regular precession, but the last may comprise
motions that do not fit in the pendulum type in addition to a regular precession.
Conditions for existence of semi-regular precessions of a rigid body in a liquid
involving a pendulum-like motion about an axis fixed in the body and regular pre-
cession about the (virtual) vertical were first found in [247] where solutions of the
equations of motion were sought such that the angular velocity has the form
ϕ̇e + nγ, (10.154)

where n is a constant, γ is the unit vector along the (virtual) vertical and e is a unit
vector constant in the body. This formula is substituted into the equations of motion,
compatibility conditions are found and then a differential equation is obtained for
the determination of ϕ.
Independently, and slightly later, of [247] the same motions were considered more
comprehensively in the Ph.D. Thesis [148]. In this work not only conditions for
the existence of pendulum motions and their transformed version (the semi-regular
precession) are obtained, but also a detailed study was made on the orbital stability
of certain special cases of those motions.
In our presentation of the precessing pendulum motion, we shall use the method
used in [148]. The study of the motion is made in two steps:
A) The motion is studied in a rotating reference frame in which the body performs a
pendulum motion about a horizontal axis fixed in this system. Conditions necessary
for performing this motion are found on the transformed parameters of the body
I, a , J , κ, K̄ .
B) The motion is transformed back to the inertial frame, the regular precession
will be added. The relevant conditions on the original parameters of the body are
obtained from
ω = ω + nγ. (10.155)
10.14.4.1 Pendulum Motion
Consider the motion of the body as a physical pendulum, taking place around a
principal axis of the inertia matrix of the body-liquid system, while this axis keeps a
permanent horizontal position (Fig. 10.2).
@seismicisolation
@seismicisolation
Fig. 10.2 The body configuration of pendulum motion. G the centre of mass, S the rotor with
gyrostatic moment k3
Let us choose the third principal axis to be the axis of proper rotation with variable
angular velocity ϕ̇. The solution corresponding to this motion with θ = π2 is
ω = ( p, q, r ) = (0, 0, ϕ̇),
γ = (sin ϕ, cos ϕ, 0). (10.156)
We shall find conditions on the matrices K̄, J and the vectors κ = (k1 , k2 , k3 ) and
a = (a1, a2 , a3 ) that allow the body to perform pendulum motion. Substituting into
Eq. (10.45), the Poisson equations are identically satisfied and the first two dynamical
equations give:
− ϕ̇[ K̄ 12 sin ϕ − 2 K̄ 22 cos ϕ + k2 ] − cos ϕ[J13 sin ϕ + J23 cos ϕ + a3 ] = 0,

ϕ̇[ K̄ 11 sin ϕ − 2 K̄ 12 cos ϕ + k1 ] + sin (ϕ) [J13 sin ϕ + J23 cos ϕ + a3 ] = 0,
(10.157)
while the third dynamical equation is replaced by the energy integral
1 1
C ϕ̇2 + a · γ+ γJ · γ = E, (10.158)
2 2
where E is the energy constant of the motion.
A combination of (10.157) gives
ϕ̇[k1 sin ϕ − k2 cos ϕ − ( K̄ 11 − K̄ 22 ) sin 2ϕ − 2 K̄ 12 cos 2ϕ] = 0,
which leads to the conditions
k1 = k2 = 0,
K̄ 12 = 0, K̄ 22 = K̄ 11 . (10.159)
@seismicisolation
@seismicisolation
Now, getting back to (10.157) we obtain one equation
K̄ 11 ϕ̇ + J13 sin ϕ + J23 cos ϕ + a3 = 0. (10.160)
This equation in ϕ gives a law of motion contradicting the pendulum law, and hence
should be satisfied as an identity. We obtain, in addition to (10.159), the conditions
K̄ 11 = 0, a3 = 0, J13 = J23 = 0. (10.161)
Summing up, for this version one can write the parameters of the problem in the
form:
a = (a1 , a2 , 0), κ = (0, 0, k3 ),

⎛ ⎞
J11 J12 0
J = ⎝ J12 J22 0 ⎠ , (10.162)
0 0 J33
⎛ ⎞ ⎛ ⎞
0 0 K̄ 13 K̄ 33 0 − K̄ 13
K̄ = ⎝ 0 0 K̄ 23 ⎠ , K = ⎝ 0 K̄ 33 − K̄ 23 ⎠ . (10.163)
K̄ 13 K̄ 23 K̄ 33 − K̄ 13 − K̄ 23 0
Those conditions mean that the centre of mass of the body lies in the x y-plane,
perpendicular to the axis of pendulum rotation (the z-axis), which is a principal
axis also of the matrix J. The angle of proper rotation ϕ can be determined as an
elliptic function of time by inverting the integral, obtained by separating variables in
(10.158),

dϕ
t= .
2(E − a1 sin ϕ − a2 cos ϕ − J12 sin ϕ cos ϕ) − J11 sin2 ϕ − J22 cos2 ϕ
(10.164)
This formula contains the energy constant E, which takes all real values on the
interval [V− , ∞), V− being the minimum of the potential V on the Poisson sphere.
Pendulum motions constitute a family of periodic motions of two types: vibrational
motions reversing their direction every half-period time and complete rotational
motions going on in one direction all the time.
10.14.4.2 The Precessing Pendulum
Conditions (10.163) for existence of pendulum-like motion of the body in a liquid (or
in the equivalent generalized problem) can now be generalized to generate conditions
for the semi-regular precession. One can now transform the pendulum-like motion
about its axis fixed in space to add the parameter n to the solution. Every pendulum-
like motion generates a family of semi-regular precessions, with n taking all real
values. The parameter n enters in the transformed conditions (10.154) according
@seismicisolation
@seismicisolation
to the transformation formulas (10.102). Finally, we can write the parameters of

the body, in order that the body can perform a semi-regular motion composed of a
pendulum motion and a precession with angular velocity n :
a = (a1 , a2 , nk3 ), κ = (0, 0, k3 ),

⎛ ⎞
2n(B + C − A) 0 K̄ 13
K̄ = ⎝ 0 2n(C + A − B) K̄ 23 ⎠
K̄ 13 K̄ 23 K̄ 33 + 2n(A + B − C)
⎛ ⎞
J11 − n A − n K̄ 33
2
J12 n K̄ 13
J = ⎝ J12 J22 − n 2 B − n K̄ 33 n K̄ 23 ⎠ . (10.165)
n K̄ 13 n K̄ 23 J33 − n 2 C
In the last formula, we used the relation K = tr(K̄)−K̄ to obtain the transformed
parameters from (10.102).
Conditions (10.165) are well-ordered and much transparent than conditions in
[247], where conditions for the semi-regular precession are given just as relations
between the parameters.
10.14.4.3 The Space Picture of the Motion
The integral (10.164) is elliptic and can be evaluated using formulas from [130].
When J12 = 0 and a2 = 0, i.e. when the vector a (the centre of mass) lies on a
principal axis of inertia and J, I have common principal axes, the integral becomes
simpler and γ1 = sin ϕ can be determined in terms of Jacobian elliptic function in
time. This was done in [148], where also the translational motion was studied. It
turned out that the central point can draw several types of trajectories in space.
In the special case when the parameters of the body satisfy the conditions
A = B,
J11 = J22 = εA, J33 = εC,
K 13 = K 23 = 0, K 33 = −nC,
one finds
γ1 = −1 + 2 sn2 v, γ2 = 2 sn v cn v,

2 a1
p = −nγ1 , q = −nγ2 , r = dn v, (10.166)
k C
where
a1 t
v= (10.167)
C k
@seismicisolation
@seismicisolation
and the modulus of the elliptic functions

4a1
k= . (10.168)
2h + 2a1 − A(ε + n 2 )
Note that we have chosen the case of fast pendulum (in which the pendulum rotates
with variable angular velocity in one direction).
From (10.166), we also get
pγ1 + qγ2
ψ̇ = = −n,
γ12 + γ22
so that we can write

ψ = −nt,
and thus we arrive at the following expressions for the base vectors in space
α = (γ2 cos nt, −γ1 cos nt, − sin nt),

β = (−γ2 sin nt, γ1 sin nt, − cos nt). (10.169)
The velocity of the central point of the body can now be written from (10.60) as
1
u = a + γJ − ω K.
2
The space components of the velocity with respect to some inertial system of axes
ξηζ are
ξ˙ = u · α, η̇ = u · β, ζ̇ = u · γ,
and can now be evaluated:

dξ
= F(t) sin nt + 2a1 sn v cn v cos nt
dt
dη
= F(t) cos nt − 2a1 sn v cn v sin nt
dt
dζ
= (ε + n 2 )A − a1 + nC + 2a1 sn2 v, (10.170)
dt
where
2
F(t) = n(k3 + Ca1 dn v).
k
By integrating (10.170) with respect to time, we obtain
@seismicisolation
@seismicisolation

k3 2 a1 a1 t
ξ = −C( + dn( )) cos nt,
C k C C k

k3 2 a1 a1 t
η = C( + dn( )) sin nt,
C k C C k
2a1 E(k)
ζ = [(ε + n 2 )A − a1 + nC + 2 (1 − )]t
k K (k)

2 a1 a1 t
− Z( ) (10.171)
k C C k
where K (k), E(k) are complete elliptic integrals and Z is Jacobi’s Zeta function of
the same modulus k.
The functions dn, Z have period

C
T1 = 2k K (k) , (10.172)
a1
while the trigonometric terms have period
2π
T2 = . (10.173)
n
The position vector of the central point of the body P (say) is not periodic in time
in general, but its projection on the ξη-plane can be a closed curve if the ratio TT21 is
a rational number.
Now we are ready to describe the space picture of the motion of the body. The
body performs the periodic pendulum motion about its horizontal z-axis while this
axis precesses with a uniform angular speed n in the (virtual) horizontal plane. The
motion of the central point P of the body traces a space curve of helicoidal type
about a (virtual) vertical axis. The radial distance ρ of P from the ζ-axis of the curve
2
ρ = k3 + Ca1 dn v (10.174)
k
changes periodically, while rotating about the vertical ζ-axis with the same angular
speed n of the body about the vertical Z -axis. As to its horizontal motion, the body
moves around the ζ-axis and rotates in such a way that one face of the body is always
directed to that axis. In celestial mechanics, this regime of motion is called 1 − 1
rotation.
From the expression (10.174), we note that the effect of the gyrostatic moment
k3 appears in the motion of the central point as an increase (decrease) of the radial
distance between that point and the ζ-axis. This means widening or narrowing the
lateral dimensions of the helicoidal trajectory, according to the sign of k3 .
The motion of the central point in the ζ-direction is not periodic in general, due
to the presence of a secular term (linear in t). After each (orbital) revolution about
@seismicisolation
@seismicisolation
the ζ-axis the central point elevates (or descends) a certain distance above (or below)
the horizontal plane that passed through P at the initial moment t = 0. The space
path of P is helicoidal-like.
However, if the coefficient of t in the secular term vanishes, i.e. if
2a1 E(k)
(ε + n 2 )A − a1 + nC + (1 − ) = 0, (10.175)
k 2 K (k)
then the vertical motion of P is periodic with period T1 . If, moreover, T1 and T2 are
commensurable, say, NT11 = NT22 , then the space trajectory of P closes after a number N
of revolutions N = LCM(N1 , N2 ) (The least common multiple of the two numbers).
The following figures illustrate the shapes of some space trajectories of the cen-
tral point. In all of them we suppose that (10.175) is satisfied and set C = 1 and
√
substitute a1 = 2k K (k)/T 1, so that the equation of the space curve becomes
2√ t t
ξ = −[k3 + a1 dn(2K (k) )] cos(2π ),
k T1 T2
2√ t t
η = [k3 + a1 dn(2K (k) )] sin(2π ),
k T1 T2
2√ t
ζ=− a1 Z (2K (k) ). (10.176)
k T1
The following Fig. 10.3 shows the space orbit of the central point of the body for
values of the parameters:
k3 = 1, k = 0.9, n = 1, a1 = k 2 = 0.81.
Fig. 10.3 Space trajectory of pendulum. a An upper view b A side view
@seismicisolation
@seismicisolation
Fig. 10.4 T1 = T2
Fig. 10.5 k3 = −3, a1 = 4, T1 = 2T2 (N1 = 2, N2 = 1). a Side view. b Projection on ξη-plane
Fig. 10.6 Side view for different values of k3 and a1
The shapes of some closed space curves are shown for different values of the
parameters k3 , k and the integers N1 and N2 . They show the diversity of the forms
of trajectories, even for a very limited set of initial conditions (Figs. 10.4, 10.5, 10.6,
10.7, 10.8).
Fast pendulum rotations k = 0.99,
Slower rotation k = 0.5 N1 = 1, N2 = 1
N1 = 2, N2 = 1
The case of a simple body k3 = 0 :
@seismicisolation
@seismicisolation
Fig. 10.7 k3 = −3, a1 = 4, T2 = 2T1 (N1 = 1, N2 = 2). a side view b Projection on ξη-plane
Fig. 10.8 k3 = −3, a1 = 4 a Side view b Projection on ξη-plane
It has been noticed before that the presence of the gyrostatic momentum k3 affects
only the lateral dimensions of the trajectory of the body. We now examine some
periodic motions of the body in the absence of k3 . To keep the variation of the radial
distance ρ somewhat large, we give the modulus of elliptic functions k the value
0.99. The following figures are obtained by taking a1 = 4. The values of N1 and N2
are given for each figure. The motion of the body consists of
- Complete rotations of the body as a physical pendulum, with period of rotation
T1 , about its z-axis, which is always horizontal (orthogonal to the virtual vertical γ)
and directed to the ζ-axis.
- A 1-1 regime of rotation about the ζ-axis (One side of the body always faces
that axis) of periodic time T2 = 2π n
.
- Radial displacement of the central point from the ζ-axis of periodic time T1 (The
same as the periodic time of the pendulum).
- Oscillations of the central point in the direction of the ζ-axis (the virtual vertical)
of periodic time T1 .
Figure 10.9 depicts the case of equal T1 and T2 . The space trajectory of the central
point of the body closes after a single rotation about the ζ-axis. The central point
ascends from the lowest point to the highest point on the part of the trajectory near to
the ζ-axis and then descends on the farther part to the first point. At the same time,
@seismicisolation
@seismicisolation
Fig. 10.9 T1 = T2
Fig. 10.10 T1 = 2T2
the body completes a pendulum rotation cycle in its vertical plane which rotates, in
turn, so that the axis of the pendulum motion remains all the time directed to the
ζ-axis. In Fig. 10.10, at the time of a complete pendulum rotation cycle the body
completes two rotations about the ζ-axis, giving always the same face to that axis.
The central point of the body ascends along the narrower loop and descends along
the wider one.
In Fig. 10.11, each pendulum rotation cycle corresponds to one vertical oscillation
but corresponds to ten precession cycles associated with ten loops around the ζ-axis
forming a ten-loop solenoid. The central point of the body ascends along the narrow
part of the solenoid and descends along the wider part to the lowest point. Figure 10.12
shows the reverse case T2 = 10T1 . The time of one complete precession cycle of the
body and rotation of its central point is enough for ten cycles of the vertical and lateral
vibrations of the central point and ten complete cycles of the pendulum vibration.
@seismicisolation
@seismicisolation
Fig. 10.11 T1 = 10T2
Fig. 10.12 T2 = 10T1
Remark: It is now time to repeat our previous assertion, that all regular pendulum
precessions generated by a certain pendulum motion share all qualitative properties
with that motion. For example, the condition for the (orbital) stability of the hor-
izontal axis of semi-regular precession is the same as the condition of stability of
the pendulum-like motion generating it, for n = 0. Conditions for stability of pre-
cessional motion are obtained from the former conditions by replacing the original
parameters by the primed ones, which involve the precession parameter n. Some
information about orbital stability of pendulum-like motions in the two problems of
this chapter will be included in the exercises.
@seismicisolation
@seismicisolation
10.15 Tables of Integrable Cases of Motion of a Rigid Body

in a Liquid
In this section, we provide tables of general and conditional integrable cases of motion
of a rigid body by inertia in an ideal incompressible liquid, infinitely extending in all
directions and resting at infinity. results are displayed for the case of a body bounded
by a multi-connected surface, i.e. for a perforated body. This case is characterized
by the presence of the two vectors a and κ in our equations in the framework of
the equivalent problem of motion about a fixed point of a rigid body acted upon by
potential and gyroscopic forces. To follow the same pattern as in previous and coming
chapters, we have chosen to put the problem of the present chapter in the context of
the second problem. In the tables, all integrable cases of the problem of motion of a
body in a liquid are presented in their most general form, as cases of integrability of
our new Eq. 10.45 and in terms of the parameter matrices and vectors I, J, K,κ and
a. To express the integrability conditions and the integrals of motion in terms of the
original (Kirchhoff-Lamb) parameters, one should use (10.43) or (10.30) as needed.
As to the classification of cases in each table, we organized cases according to the
degree of the complementary integral as a function in the components of the angular
velocity.
For each case, we provide
(1) The full hierarchy of earlier cases, to which the given case reduces under relevant
conditions on the parameters.
(2) The conditions on the parameters on the body and fields, under which the case
is valid.
(3) The potential function V .
(4) The vector functions l and μ , which describe the gyroscopic forces acting on
the body: The first enters in the Lagrangian and the second in the equations of
motion.
(5) The explicit forms of the first integrals I3 and I4 in the Euler-Poisson variables.
(6) A Hamiltonian function H is given for each case, together with the correspond-
ing form of the complementary integral in terms of the variables (M,γ) (See
Sect. 10.10).
Remark 16 Regarding the fact that most integrable cases are obtained by using
inverse method, different hamiltonians may be constructed for one and the same
case.
10.15.1 General Integrable Cases
The number of known general integrable cases in the two equivalent problems of the
present chapter is seven. Most of them are solutions of Thomson-Lamb equations.
They developed historically from solutions of the simpler cases of integrability of
@seismicisolation
@seismicisolation
10.15 Tables of Integrable Cases of Motion of a Rigid Body in a Liquid 287
the Kirchhoff equations and, in cases, from cases of integrability of problems lower
in the hierarchy, which are presented in the previous chapters.
Remark: The regular precession transformation parameter n figures in five of the
seven integrable cases, namely, cases 2,3,5,6 and 7. If n is introduced in cases 1 and
4, it can be absorbed in other parameters of the problem and can give no new effects.
In cases 2,3 and 5, the parameter n appeared at some stage in their development and
not necessarily from the first discovery of the case. In the remaining cases (number
6,7), the introduction of that parameter in [411] was a significant generalization of
the case found by Sokolov [336].
Table 10.1 General integrable cases

1 The case of axi-symmetric body
Generalization of Lagrange’s case
Kirchhoff [219] (1870) (see also [220]) a3 = 0, κ3 = 0
A = B,
V = a3 γ3 + 21 [b1 (γ12 + γ22 ) + b3 γ32 ],
l = (K 1 γ1 , K 1 γ2 , K 3 γ3 + κ),
μ = (−K 3 γ1 , −K 3 γ2 , (K 3 − 2K 1 )γ3 + κ),
I3 = A( pγ1 + qγ2 ) + (Cr + κ)γ3 + K 1 (γ12 + γ22 ) + K 3 γ32 ),
I4 = Cr + κ + K 3 γ3
M 2 +M 2 M2
H = 12 A 2 + 2C3 − KA1 (M1 γ1 + M2 γ2 ) − M3
C (K 3 γ3 + κ)
+a3∗ γ3 + b1∗ (γ12 + γ22 ) + b3∗ γ32 ,
I 4 = M3
where a3∗ , b1∗ and b3∗ are constants
2 Clebsch [55] (1870).

Euler [79] (1758). n=b=0
V = (b − 21 n 2 )(Aγ12 + Bγ22 + Cγ32 )
l = n(Aγ1 , Bγ2 , Cγ3 )
μ = n((A − B − C) γ1 , (B − C − A) γ2 , (C − A − B)γ3 )
I3 = Apγ1 + Bqγ2 + Cr γ3 + n(Aγ12 + Bγ22 + Cγ32 ),
I4 = 21 [A2 ( p + nγ1 )2 + B 2 (q + nγ2 )2 + C 2 (r + nγ3 )2 ]
−b(BCγ12 + C Aγ22 + ABγ32 ).
H = 21 (M12 /A + M22 /B + M32 /C) + b(Aγ12 + Bγ22 + Cγ32 )
−n(M1 γ1 + M2 γ2 + M3 γ3 ),
I4 = 21 (M12 + M22 + M32 ) − b(BCγ12 + C Aγ22 + ABγ32 ).
Somewhat later, after [55], the special version (n = 0) of this case case was found,
apparently independently, by Tisserand [354] (1891) and Brun [47] (1893) in the
context of the motion of a body acted upon by approximate Newtonian gravitational
@seismicisolation
@seismicisolation
forces (See Chap. 6). At that time the Steklov analogy, described above in this chapter,
between the two problems was still unknown.
3 Clebsch [55] (1870) -

A=B=C
V = 21 A(c1 γ12 + c2 γ22 + c3 γ32 )
l = n Aγ, μ = −n Aγ
I3 = A( pγ1 + qγ2 + r γ3 ),
I4 = A[c1 ( p + nγ1 )2 + c2 (q +
nγ2 )2 + c3 (r + nγ3 )2
−(c2 c3 γ12 + c3 c1 γ22 + c1 c2 γ32 )]
H = 21A (M12 + M22 + M32 ) + 21 (c1 γ12 + c2 γ22 + c3 γ32 )
−n (M1 γ1 + M2 γ2 + M3 γ3 ) ,
I4 = c1 M12 + c2 M22 + c3 M32
−A[c2 c3 γ12 + c3 c1 γ22 + c1 c2 γ32 ]
In cases 2 and 3, the parameter n invokes gyroscopic terms due to circulation of

the liquid through perforations. Setting n = 0 makes the body simply connected.
4 Rubanovsky [317] (1968),

Kharlamov (κ2 = 0) [192] (1963),
Steklov (κ= 0) [344] (1893),
Joukowsky (a = 0) [163] (1885),
Euler (κ= 0, a = 0) [79] (175
I = diag(A, B, C), Ī = 21 tr (I) − I,
J = [tr (I−1 )δ−I−1 ]
= diag( B1 + C1 , A1 + C1 , A1 + B1 )
κ = (κ1 , κ2 , κ3 ),
V = −n(κ · γ−aγI−1 · γ) − 21 n 2 γI · γ,
l = κ+aγJ + nγI,
μ = κ + 2aγI−1 − 2nγ Ī,
I3 = (ωI+l) · ,
I4 = 21 | (ω + nγ)I+κ |2 −2aω · γ
H = 21 (M − ˇ − aγJ)I−1 · (M−κ − aγJ)
−nM · ,
I4 = 1
2 | M − aγJ |2 −aγI−1 · (M−κ − aγJ).
@seismicisolation
@seismicisolation
5 Rubanovsky [317] (1968)

Lyapunov [267] (a1 = a2 = a3 = 0) (1893)
A = B = C,
V = C{a1 γ1 + a2 γ2 + a3 γ3
− 21 [(bc + b0 )γ12 + (ca + b0 )γ22 + (ab + b0 )γ32 ]},
l = − 2 C((b + c)γ1 , (c + a)γ2 , (a + b)γ3 ),
1
μ = C(aγ1 , bγ2 , cγ3 ),

I3 = pγ1 + qγ2 + r γ3 + 21 (aγ12 + bγ22 + cγ32 ),
I4 = 21 [(b + c) p 2 + (c + a) q 2 + (a + b) r 2 ] − abc( p γa1 + q γb2 + r γc3 )
+a1 ( p + aγ1 ) + a2 (q + bγ2 ) + a3 (r + cγ3 ).
H = 2C1
(M12 + M22 + M32 ) + 21 [(b + c)M1 γ1 + (c + a)M2 γ2 + (a +
b)M3 γ3 ]
+C(a1 γ1 + a2 γ2 + a3 γ3 )
− C8 [(a 2 + 2bc)γ12 + (b2 + 2ac)γ22 + (c2 + 2ab)γ32 ],
I4 = (b + c)M12 + (c + a)M22 + (a + b)M32
+C{[(b2 + c2 )γ1 + 2a1 ]M1 + [(a 2 + c2 )γ2 + 2a2 ]M2 + [(a 2 +
b )γ3 + 2a3 ]M3 }
2
2
+ C4 [(b + c)(b − c)2 γ12 + (c + a)(c − a)2 γ22 + (a + b)(a − b)2 γ32 ]
+C 2 [(a + b + c)(a1 γ1 + a2 γ2 + a3 γ3 ) + 2(a1 aγ1 + a2 bγ2 + a3 cγ3 )].
The parameter b0 has meaning in the problem of motion of a body in a liquid. In

the alternative problem it is immaterial.
6 Yehia [411] (2003)

Sokolov [336] n = 0 (2002)
Yehia [380] n = c = 0 (1986)
Kowalevski [238] n = c = κ = 0 (1889)
A = B = 2C, κ = C(0, 0, κ),
2
V = C[κcγ1 + a2 γ2 − nκγ3 − c2 (γ12 + 2γ32 ) − ncγ1 γ3
2
− n2 (2γ12 + 2γ22 + γ32 )],
l = C(2cγ3 + 2nγ1 , 2nγ2 , κ − cγ1 + nγ3 ),
μ = C(cγ3 − nγ1 , −nγ2 , κ + cγ1 − 3nγ3 ),
I3 = 2( pγ1 + qγ2 ) + (r + κ + cγ1 )γ3 + n(2γ12 + 2γ22 + γ32 ),

2
I4 = ( p + nγ1 )2 − (q + nγ2 )2 + a2 γ2 + c2 γ22 + cγ1 (r + nγ3 − κ)

2
+ 2( p + nγ1 )(q + nγ2 ) − a2 γ1 − c2 γ1 γ2 + cγ2 (r + nγ3 − κ)

+2κ(r + nγ3 − κ + cγ1 ) ( p + nγ1 ) + (q + nγ2 ) + 2c( p + nγ1 )γ3
2 2
−2κc2 {2γ3 [2( p + nγ1 )γ1 + cγ1 γ3 + 2(q + nγ2 )γ2 + (r + nγ3 )γ3 ]
+κγ32 − (γ12 + γ22 + 2γ32 )(r + nγ3 + cγ1 )} − 4a2 κ(q + nγ2 )γ3 .
M2 M2
H = 2C1
( 21 + 22 + M32 ) − cγ3 M1 + (cγ1 − κ)M3 + Ca2 γ2
−n(M1 γ1 + M2 γ2 + M3 γ3 )
M 2 −M 2
I4 = [ 1 4 2 + Cc (−M1 z + M3 x) + a2 y + c2 C 2 ]2

+[ M12M2 + Cc (−M2 z + M3 y) − C 2 a2 x]2 + 21 Ck (M3 − 2Ck) M12 + M22
+C 2 k[−2Ca2 M2 z − 2M2 (M1 y − M2 x) c − 2Cc2 M3 ]
@seismicisolation
@seismicisolation
7 The parameter n is added here to B-M-S result.

Borisov, Mamaev and Sokolov n = s = 0 [39] (2001)
Sokolov [336] n = 0 (2001)
Kowalevski n = m = 0 (1888)
A = B = 2C,
V = C{s(c1 γ1 + c2 γ2 ) + 21 m 2 [(c1 γ1 + c2 γ2 )2 − (c12 + c22 )γ32 ]
2
+nmγ3 (c2 γ1 − c1 γ2 )γ3 − n2 (2γ12 + 2γ22 + γ32 )},
l = C(2nγ1 , 2nγ2 , m(c2 γ1 − c1 γ2 ) + nγ3 ),
μ = C(−mc2 γ3 − nγ1 , −mc1 γ3 − nγ2 , m(c2 γ1 − c1 γ2 ) − 3nγ3 ),
I3 = 2( pγ1 + qγ2 ) + [r + m(c2 γ1 − c1 γ2 )]γ3 + n(2γ12 + 2γ22 + γ32 ),
I4 = A24 + B42 ,
where
A4 = ( p + nγ1 )2 − (q + nγ2 )2 + s(c2 γ2 − c1 γ1 )
+m(r + nγ3 )(c2 γ1 + c1 γ2 ) + m 2 (c22 γ22 − c12 γ12 ),
B4 = 2( p + nγ1 )(q + nγ2 ) − s(c1 γ2 + c2 γ1 )
−m(r + nγ3 )(c1 γ1 − c2 γ2 ) − m 2 (c1 γ2 + c2 γ1 )(c1 γ1 + c2 γ2 )]2 .
M 2 +M 2
H = 2C
1
[ 1 2 2 + M32 ] + m(c1 γ2 − c2 γ1 )M3 + Cs(c1 γ1 + c2 γ2 )
−Cm 2 (c12 + c22 )γ32 − n (M1 γ1 + M2 γ2 + M3 γ3 ) ,
M12 −M22
I4 = [ 4C 2
+ m(c1 γ2C+c2 γ1 ) M3 + s(−c1 γ1 + c2 γ2 ) − m 2 (c12 + c22 )(γ12 − γ22 )]2
m(c1 γ1 −c2 γ2 )
+[ M2C
1 M2
2 − C M3 − s(c1 γ2 + c2 γ1 ) − 2m 2 (c12 + c22 )γ1 γ2 ]2 .
Strictly speaking, case 7 is related to case 6 and can be considered as its special
case. We prefer, for future uses (See Chap. 12), to consider case 7 as independent
case in its most general form containing maximum number of parameters.
In case 7, the integral I4 is the sum of two squares. It is the squared modulus of
the complex quantity
A4 + i B4 = [ p + iq + n(γ1 + iγ2 )]2

− (c1 + ic2 )(γ1 + iγ2 )[s + im(r + nγ3 ) + m 2 (c1 γ1 + c2 γ2 )].
(10.177)
The quantities A4 , B4 satisfy the relations
Ȧ4 = [r + nγ3 + m(c1 γ2 − c2 γ1 )]B4 , Ḃ4 = −[r + nγ3 + m(c1 γ2 − c2 γ1 )]A4 ,

(10.178)
so that the set of conditions
{A4 = 0, B4 = 0}
define an invariant manifold.

When s = 0, case 7 renders to the case discussed in Sect. 10.10 and the integral
becomes expressible, as in (10.91), in the form of the product of two functions one
linear and the other cubic in velocities.
@seismicisolation
@seismicisolation
10.15.2 Conditional Integrable Cases on the Level f = 0
Two conditional integrable cases are known. Those cases can be generalized, as will
be shown later, using an arbitrary function ν(γ) instead of n. Nevertheless, we write
them down here with that parameter, as it adds physically significant terms to both
problems considered in this chapter: the problem of motion of a body in a liquid
and the alternative problem of motion under the action of potential and gyroscopic
forces.
Table 10.2 Cases integrable on the level

1 Parameter n is added here to the result of Sokolov and Tsiganov
n = 0 Sokolov-Tsiganov [338], 2002,
n = c1 = c2 = 0 Sretensky [341], 1963,
n = c1 = c2 = κ = 0 Goryachev-Chaplygin
A = B = 4C,
V = C[a1 γ1 + a2 γ2 − nκγ3 + 21 (c2 γ1 − c1 γ2 )2
2
−nγ3 (c1 γ1 + c2 γ2 ) − n2 (4γ12 + 4γ22 + γ32 )],
l = C( c21 γ3 + 4nγ1 , c22 γ3 + 4nγ2 , 21 (c1 γ1 + c2 γ2 ) + nγ3 ),
μ = C(c1 γ3 − nγ1 , c2 γ3 − nγ2 , κ + c1 γ1 + c2 γ2 − 7nγ3 ),
I3 = 4 pγ1 + 4qγ2 + [r + κ + c1 γ1 + c2 γ2 ]γ3 + n(4γ12 + 4γ22 + γ32 ),
I4 = (r − κ + c1 γ1 + c2 γ2 + nγ3 )×
×[( p + nγ1 + 21 c1 γ3 )2 + (q + nγ2 + 21 c2 γ3 )2 ]
+γ3 [(κc1 − a1 )( p + nγ1 ) + (κc2 − a2 )(q + nγ2 )]
+ 21 γ32 [κ(c12 + c22 ) − c1 a1 − c2 a2 ]
M2 M2
H = 2C1
( 41 + 42 + M32 ) + (−κ + 2c1 γ1 + 2c2 γ2 )M3
−γ3 (c1 M1 + c2 M2 ) + C(a1 γ1 + a2 γ2 )
−n(M1 γ1 + M2 γ2 + M3 γ3 ),
I4 = [(M3 − 2Cκ + 4C(c1 γ1 + c2 γ2 )](M12 + M22 )
−4C 2 γ3 (a1 M1 + a2 M2 )
I3 = f = 0
In [337, 338], Sokolov and Tsiganov did not give the complementary integral for
this full case. The above formulas are adjusted from [41] (English edition 2017).
@seismicisolation
@seismicisolation
2 Yehia [386] 1987,

n = κ = 0 Chaplygin [53] 1903,
n = κ = b1 = b2 = 0 Kowalevski [238] 1888 (Special case f = 0),
A = B = 2C,
V = C[a1 γ1 + a2 γ2 − nκγ3 + b1 (γ12 − γ22 ) + 2b2 γ1 γ2
2
− n2 (2γ12 + 2γ22 + γ32 )],
l = C(2nγ1 , 2nγ2 , κ + nγ3 ), μ = C(−nγ1 , −nγ2 , κ − 3nγ3 ),
I3 = 2 pγ1 + 2qγ2 + (r + κ)γ3 + n(2γ12 + 2γ22 + γ32 ),
I4 = [( p + nγ1 )2 − (q + nγ2 )2 − a1 γ1 + a2 γ2 + b1 γ32 ]2
+[2( p + nγ1 )(q + nγ2 ) − a1 γ2 − a2 γ1 + b2 γ32 ]2
+2κ (r + nγ3 − κ) [( p + nγ1 )2 + (q + nγ2 )2 ]
−4κγ3 [( p + nγ1 )(a1 + b1 γ1 + b2 γ2 )
+(q + nγ2 )(a2 + b2 γ1 − b1 γ2 )].
M2 M2
H = 2C
1
( 21 + 22 + M32 ) − κM3
+C[a1 γ1 + a2 γ2 + b1 (γ12 − γ22 ) + 2b2 γ1 γ2 ]
−n(M1 γ1 + M2 γ2 + M3 γ3 ),
M12 −M22
I4 = [ 4C 2
− a1 γ1 + a2 γ2 + b1 γ32 ]2
M1 M2
+[ 2C 2 − a1 γ2 − a2 γ1 + b2 γ32 ]2
− 2Cκ 3 (2Cκ − M3 )(M12 + M22 )
C γ3 [M1 (a1 + b1 γ1 + b2 γ2 ) + M2 (a2
− 2κ + b2 γ1 − b1 γ2 )].
10.16 Further Studies on Integrable Cases
In the set of integrable cases in the dynamics of a body in a liquid, the presence of the
complementary (fourth) integral makes it possible to perform several analytical and
qualitative investigations on each one of the integrable cases, something we are not
able to do in the generic case missing the fourth integral. Those investigations include
the final explicit solution of the equations of motion in terms of time, bifurcation and
topological classification of orbits of the integrable system on its integral manifold
and the stability analysis of certain motions like stationary and periodic motions.
In this section, we try to give a quick review of some of those investigations
performed for the problem of motion of a body in a liquid.
10.16.1 Separation of Variables, Explicit Solutions and
Separation of the variables was performed most easily in the general integrable
cases of Euler and Lagrange of the classical problem. Both cases were reduced to
elliptic quadratures and hence the explicit solution was expressed in terms of elliptic
@seismicisolation
@seismicisolation
10.16 Further Studies on Integrable Cases 293
functions of time. The general integrable case of Kowalevski and the conditional
case of Goryachev and Chaplygin were reduced to hyper-elliptic quadratures (For
more information see Chap. 4). Explicit time solution of various integrable cases of
motion of a body in a liquid was investigated by several authors. The present status
of this aspect is summarized in the following:
General integrable cases:
(1) Kirchhoff reduced the case of simply connected body (a3 = κ = 0) to an ellip-
tic quadrature and expressed some particular motions in terms of elliptic functions
[219]. Detailed analysis of the general solution in elliptic functions was performed
by Halphen [146] and Greenhill [135, 136]. The full general case 1 of Table 10.1 can
be easily solved also in terms of elliptic functions of time. This is shown in Chap. 12
Sect. 12.1 as a special version of a more general separable case of Lagrange’s type
(Case 7 of Table 10.1).
(2) The two cases 2 and 3 discovered by Clebsch were shown by Kötter [233]
in 1892 to have their general solution in terms of Theta functions in two variables.
The special version f = 0 of the asymmetric case of Clebsch was solved, in the
same set of functions by Weber, somewhat earlier using separation of variables in
Hamilton–Jacobi equation [367]. The version f = 0 of the spherical Clebsch case
is equivalent to Neumann’s problem solved in Theta functions of two variables (See
Chap. 9 Sect. 9.7.3). Equations of motion in the Lax pair form and generalization to
n-dimensional space are briefly discussed in [306]. Some later trials led to separation
of variables in a much more complicated form, e.g. [260, 273]. Recently, the full
version and Weber’s one have been reconsidered in [95, 271].
(3) Steklov and Lyapunov subcases of cases 4 and 5 are conjugate in the same sense
as the two cases of Clebsch. A solution of those subcases proposed by Kötter in Theta
functions of two variables [235] was presented in a very compact and complicated
way, which led to some controversies between his contemporaries. Tsiganov [360]
reconsidered the separation problem for Steklov and Lyapunov subcases and recently,
in [363], the full versions of cases 4, 5 due to Rubanovsky. However, no explicit
formulas are given. Thus, the separation of variables for the Rubanovsky cases cannot
be considered complete, except for the Steklov and Lyapunov subcases separated by
Kötter.
(4) As concerns case 6, only the lower level of this hierarchy (Kowalevski’s case)
is separated by Kötter [232, 234]. The status of the second level (Yehia’s gyrostat)
is described in Chap. 5. A successful procedure like that followed by Kötter has not
been found. Separation of variables is not yet achieved for that level and for the
next one (Sokolov’s generalization of Yehia’s gyrostat). Note that if the solution of
the Sokolov case (n = 0) is constructed, the solution of the last level with n = 0 is
the same, as concerns the vector γ(t). The vector ω(t) is then readily obtained by
applying the regular precession transformation (See Sect. 10.11).
Topological classification of the case of Yehia’s gyrostat in the uniform gravity
field (See Chap. 5) is discussed in detail in the book of M. Kharlamov et al. [184]
(See also [185]). The generalized version when n = 0 and for non-zero Sokolov
parameter c, was not investigated until now.
@seismicisolation
@seismicisolation
(5) Variable separation for the Sokolov case in the hierarchy 7 (without a gyrostatic
momentum) was proposed in [227]. It generalizes the one used for Kowalevski’s
case by Kowalevski and Kötter. Explicit separation and expressions for dynamical
variables were given in [70] , in terms of two intermediate variables, which are
expressed in genus-2 Theta functions. In the last level of the hierarchy, after the
introduction of the parameter n, the solution is obtained by applying the uniform
precession transformation. The same separation variables of [227] were used in
[186] for detailed investigation of the integral manifolds and their bifurcation and
also complete description of the phase topology of this case.
Conditional integrable cases
(1) Separation of variables is known for the first two levels of the hierarchy. For
Goryachev–Chaplygin’s see Chap. 4 Sect. 4.4 and for Sretensky’s level see Chap. 5.
Explicit time solution for the full case (1) of Table 10.2 is not found yet.
(2) The second case of Table 10.2 involves 6 significant parameters a1 , a2 , b1 , b2 ,
κ, n, of which the last one can be set equal to zero for variable separation and it
can be restored in the system by the uniform precession transformation. Separation
of variables and explicit expressions for the dynamical variables are known in the
following subcases:
a- The special version f = 0 of Kowalevski’s case (n = κ = a2 = b1 = b2 = 0).
By a rotation of the x, y axes fixed in the body by a constant arbitrary angle in their
plane, one can construct a solution in which both coefficients a1 , a2 are present.
b- Chaplygin [53] first established the integrability, on the level f = 0, of the case
n = κ = a1 = a2 = b2 = 0, that describes the motion of a simply connected body
in a liquid (with only one parameter b1 present in the potential). Then he achieved a
separation of variables for this case and expressed the dynamical variables in terms
of two parameters s1 , s2 , each of which can be expressed as an elliptic function of
t. This solution is presented in detail in the next section. By a rotation of the x, y
axes fixed in the body at a constant arbitrary angle in their plane, one can construct
a solution in which both coefficients b1 , b2 are present.
c- From the results of [416], it follows that the problem of motion of a rigid body
with A = 2C and arbitrary B, subject to forces with potential
V = a1 γ1 + b1 (γ12 − γ22 ),
under the additional restrictions
q = 0, f = 0,
is solvable in elliptic functions of time.
10.16.2 Topological Classification of Integrable Cases
The classical integrable cases of the problem of motion of a rigid body in a liquid
served as a fertile land for the application of methods of algebraic geometry and
@seismicisolation
@seismicisolation
10.16 Further Studies on Integrable Cases 295
topology, created specially for the study of the phase space of integrable systems.
Topological classification offers an alternative, which determines the picture of the
foliation of the Liouville tori and hence sheds some light on the general (qualitative)
features of motion that can hardly be obtained from the explicit solution of compli-
cated problems. General methods of the study of bifurcation of integral manifolds
and phase topology of the integrable cases of the classical problem and gyrostat
motion were developed by M. Kharlamov (See e.g. [183, 185]). Steklov’s case was
investigated by Bogoyavlensky and Ivakh [33].
Fomenko constructed what may be called “Morse theory of integrable Hamil-
tonian systems” [87–90], building on previous results of many authors including,
in particular, works of Smale. This theory was further developed by Fomenko, his
colleagues and coworkers (e.g. [34, 35, 92, 303]). Topological classification is made
for several two- and multi-dimensional integrable Hamiltonian systems. Most inter-
esting, in particular, are Hamiltonian systems with two degrees of freedom. Those
include reductions of higher dimensional systems with cyclic coordinates. A theory
of topological invariants of such systems was developed, which gives their classifi-
cation up to Liouville equivalence, i.e. up to deformation of Liouville tori. For basic
information about the theory and some applications to rigid body dynamics, the
reader is referred to papers in [90], the works cited above and references therein. In
this subsection, some results about topological classification are pointed out parallel
to information about explicit solution for each integrable case of the dynamics of a
rigid body in a liquid. It has to be said here that topological classification is not a
characteristic property of an integrable system. Chaplygin’s case of rigid body in a
liquid, discussed in the next section, is an example.
10.17 Chaplygin’s Case of Integrability
In [53], Chaplygin discovered the conditional case, integrable on the zero level ( f =
0) of the areas integral and like Kowalevski’s case valid under the condition A =
B = 2C and characterized by the choice
V = Cc(γ12 − γ22 ), μ = 0, (10.179)
in the equations of motion (10.54), which, in this case take the form
2 ṗ − qr = 2cγ2 γ3 ,
2q̇ + pr = 2cγ1 γ3 ,
ṙ = −4cγ1 γ2 ,
γ˙1 + qγ3 − r γ2 = 0, γ˙2 + r γ1 − pγ3 = 0, γ˙3 + pγ2 − qγ1 = 0. (10.180)
@seismicisolation
@seismicisolation
The four integrals of motion are
1
p 2 + q 2 + r 2 + c(γ12 − γ22 ) = h,
2
2 pγ1 + 2qγ2 + r γ3 = 0,
γ12 + γ22 + γ32 = 1,
( p 2 − q 2 + cγ32 )2 + 4 p 2 q 2 = K 2 . (10.181)
They involve two arbitrary constants h and K . The sign of K is immaterial and,
without loss of generality, we can assume that K ≥ 0.
Chaplygin’s case is highly interesting for many reasons. In particular,
(1) It was the second conditional case in rigid body dynamics, after the Goryachev–
Chaplygin case of the classical problem (See Chap. 3 Sect. 3.4).
(2) It turned out that separation of variables is much simpler than in the former
case and leads to explicit expressions of the Euler-Poisson in terms of elliptic
functions. In fact, it is a rare example of dynamical problem, with a relatively
simple solution that can be explicitly written in terms of two sets of elliptic
functions, which have two independent moduli.
(3) Because, as will be seen later in Chap. 2, it can be brought to equivalence with
a completely different problem. Namely, it is that of motion of a body with the
Kowalevski configuration about a fixed point, while acted upon by two irre-
ducible uniform fields.
(4) It was the subject of many later generalizations, as will be seen in Chap. 13.
Chaplygin’s case was a favourite subject for topological analysis by many authors.
Topology of the iso-energy surfaces, bifurcation diagrams in the plane {K 2 , h} and
topological classification of the Liouville tori are studied in [300, 322]. More detailed
topological analysis can be found in [295]. In [91], topological equivalence of Chap-
lygin’s case is established with two other problems, the Euler case of rigid body
dynamics and Jacobi’s problem of geodesics on an ellipsoid. Nevertheless, it seems
that not much is done in the literature dealing with the explicit analytical forms of
the solution or the qualitative properties of motion. For all those reasons, we now
give a somewhat detailed description of the solution and possible types of motion of
the body.
10.17.1 Separation of Variables
We give here the expressions for the Euler–Poisson variables in terms of Chaplygin’s
separation variables. Some more details on the separation process can be found in
[53].
@seismicisolation
@seismicisolation
10.17 Chaplygin’s Case of Integrability 297
√ √ √ √
2 K s1 − 1 1 − s2
p = 1/2 √ ,
s1 − s2
√ √ √ √
2 K s1 + 1 1 + s2
q = 1/2 √ ,
s1 − s2
√ 2
2[ s1 − 1 (β − cs2 ) + 1 − s22 (cs1 − α)]
r = ), (10.182)
s1 − s2
√ √
[ (s1 + 1) (1 − s2 ) (β − cs2 ) − (1 + s2 ) (s1 − 1) (cs1 − α)]
γ1 = √ ,
2c (s1 − s2 )
√ √
[ (s1 − 1) (1 + s2 ) (β − cs2 ) + (1 − s2 ) (s1 + 1) (cs1 − α)]
γ2 = √ ,
2c (s1 − s2 )
√
2K
γ3 = − √ , (10.183)
c(s1 − s2 )
where α = h + K , β = h − K . Note that α ≥ β and equality occurs when K = 0.

The two variables s1 , s2 are solutions of the equations

ṡ1 = − 2 s12 − 1 (cs1 − α),

ṡ2 = − 2(1 − s22 )(β − cs2 ). (10.184)
It is not hard to see that for a real solution of those equations s1 ∈ [max(α, 1), ∞],
while s2 ∈ [−1, min(β, 1)].
Now, for more visibility of the results, one can choose the units of measuring
time, so that the constant c = 1. It is essential to find the conditions of repeated roots
in the under-root polynomials in (10.184). Those are, respectively,
h = −K ± 1, h = K ± 1. (10.185)
For simplicity, we introduce the parameters
α = h + K, β = h − K. (10.186)
The bifurcation diagram in the K h-plane is shown in Fig. 10.13.

There are three admissible regions I, II and III, inscribed by solid lines. In those
regions we have
−1 < 1 < β < α in region I,

−1 < β < 1 < α in region II,
−1 < β < α < 1 in region III.
@seismicisolation
@seismicisolation
Fig. 10.13 Bifurcation

diagram for Chaplygin’s case
Inside each region the analytical form of the solution of (10.184) and its qualitative
properties do not change, and so does the topological type of the invariant two-
dimensional manifold which consists of tori on which the trajectories are wind in the
phase space. Only crossing the boundaries between those regions, those properties
can change.
By integrating (10.184), it is not hard to obtain the following formulas for s1 and
s2 in terms of time:

α+1
sn2 ( 2
t, k1 ) 2
s1 = α + (α − 1) , k1 = , α > 1,
cn2 ( α+1
t, k1 ) α+1
2

sn 2 (t, ν1 ) α+1
= 1 + (1 − α) 2 , ν1 = , α < 1,(10.187)
cn (t, ν1 ) 2
and

β+1 2
s2 = −1 + 2 sn ( 2
τ , k2 ), k2 = , β > 1,
2 β+1

β+1
= −1 + (β + 1) sn2 (τ , ν2 ), ν2 = , β < 1. (10.188)
2
Here τ = t − t0 , t0 is an arbitrary constant. Note that 0 ≤ t0 < T, where T is the

period of the Jacobi elliptic functions of moduli k2 . A similar constant appears in the
first two formulas is set equal to zero without loss of generality by choosing the initial
time moment. The generic motion is quasi-periodic, but becomes periodic when the
periods T1 and T2 of the two sets of Jacobi’s functions are commensurable.
According to the general Liouville-Arnold theorem for completely integrable
hamiltonian systems, the integral manifold of the Chaplygin system, corresponding
to a fixed pair of the parameters {K , h}, is a 2-torus or a union of such tori, each
@seismicisolation
@seismicisolation
Fig. 10.14 Iso-potentials on

the Poisson sphere, the same
as zero-velocity curves
of which is filled (winded) by quasi-periodic phase trajectories. On each torus, a

trajectory is singled out by the value of the parameter t0. The projection of a trajectory
of the system on the Poisson sphere is the trajectory of the apex of the vector γ,
during the motion of the body, on that sphere. That is what we try to clarify in the
following subsections.
10.17.2 Forms of Motion on the Poisson Sphere
A look at Eq. (10.180) reveals that they have six equilibrium positions, in which an
end of one of the principal axes of inertia is directed vertically upwards. Two positions
correspond to potential minima V = −1 at the points γ = (0, ±1, 0), two correspond
to potential saddle point V = 0 at γ = (0, 0, ±1) and the last two correspond to
potential maxima V = 1 at γ = (±1, 0, 0).
From the energy integral in (10.181), one can see that any real possible motion or
equilibrium must satisfy the condition
V = γ12 − γ22 ≤ h. (10.189)
The region determined by this condition on the Poisson sphere is called the region
of possible motions. On its boundary γ12 − γ22 = h, the angular velocity of the body
vanishes. If exists, this boundary is named the zero-velocity curve ZVC.
Figure 10.14 depicts iso-potential lines on the Poisson sphere. At the minimum
value of the energy parameter h = −1, the ZVC is composed of two opposite points,
corresponding to two stable equilibrium positions4 of the body with either ends of
4 Here we mean the alternative problem of motion about a fixed point. In the Chaplygin problem,
it corresponds to a steady translational motion of the body in the liquid.
@seismicisolation
@seismicisolation
the y-axis directed along the upward vertical (The vector γ). As h increases, namely,
for h ∈ (−1, 0) ZVC consists of two components, each of which is closed around
one of the ends of the y-axis. The region of possible motion is composed of the two
areas inside the two components of the ZVC. The value h = 0 is a critical one. At
this value, the ZVC renders to a pair of great circles intersecting on the z-axis and
the two regions meet at the two ends of the z-axis. For greater values of h ∈ (0, 1),
the two components of the ZVC become closed around the tips of the x-axis and the
region of possible motion is the whole sphere with the exception of the two regions
inscribed by the ZVC. For h = 1, the region of possible motion is the whole sphere
and the equilibrium is possible with the x-axis in vertical position. Finally, for h > 1,
the region of possible motion is the whole sphere and no ZVC exists.
In the bifurcation diagram Fig. 10.13, one can readily see that the least value of
the energy parameter for a possible motion is h = −1 at P and at this point K = 0.
Those values correspond to two equilibrium positions γ = (0, ±1, 0) at two potential
minima. The y-axis is then directed up or down the positive Z -axis fixed in space.
The trajectory of the apex of y consists of two points. If we move in the bifurcation
diagram on the boundary P Q, the trajectory becomes an arc of the great circle
γ3 = 0, corresponding to a periodic pendulum-like motion about the z-axis. As we
approach the point Q, the motion becomes asymptotic to one of the two equilibrium
positions at the two potential maxima at γ = (±1, 0, 0). At all points beyond Q,
the trajectory is the whole circle, corresponding to complete uni-directional plane
(pendulum-like) rotations about the z-axis.
A similar pattern is noted also on the line P R. The motion begins as pendulum-like
vibration about the x-axis with increasing amplitude that reaches π/2 at R, where
the motion becomes asymptotic to the equilibrium positions at γ = (0, 0, ±1), the
saddle points of the potential. Beyond R, the motion is a pendulum complete rotation
about the x-axis.
An exceptional family of motions corresponds to parameters on the segment R Q.
The motion begins as a pendulum-like vibration about the y-axis with increasing
amplitude that reaches π/2 at Q, where it becomes asymptotic to the equilibrium
positions at two potential maxima at γ = (±1, 0, 0).
Finally, on the critical line h = K + 1, the motion is asymptotic to pendulum-like
complete rotations about the y-axis.
10.17.3 Explicit Solution
Now, substituing relevant expressions for s1 and s2 from (10.187), (10.188) into
(10.182), (10.183), one can write down all the Euler-Poisson √ variables √
as functions
of time.
√ Doing that, one has to choose the signs of the radicals s 1 − 1, 1 − s2 , · ··
and s1 − s2 . However, one has to take only the combinations of signs which are
compatible with the areas integral, the second one in (10.181). To make it more
@seismicisolation
@seismicisolation
definite, we first note that the equations of motion and the integrals of motion enjoy
the property of being invariant under each of the simultaneous changes of signs of
the tuples
1){γ1 , γ2 , γ3 },
2){ p, q, γ3 },
3){r, γ1 , γ2 , t}. (10.190)
Some more changes can be obtained as products of those three. Example is the change
{ω(t), γ(t)} → {−ω(−t), γ(t)}, obtainable as the product of the three changes.
Alternatively, the last change follows from a general principle, the time-reversibility
of the motion of natural mechanical systems, acted upon by purely potential forces.
We now write down the final forms in the three zones of the primary solution,
obtained by giving all radicals in (10.182), (10.183) a positive sign. In all illustrat-
ing examples in the accompanying figures below, we have set t0 = 0, i.e. we have
projected one trajectory of the infinite number on an integral torus of the problem.
Note that those figures were plotted on different time intervals, sufficient for suitable
visualization (Figs. 10.15, 10.16, 10.17).
10.17.3.1 In Zone I (1 ≤ β ≤ α < ∞,)
√
K (α − 1) cn (ν2 τ , k2 )
p= √ ,
1
√
K (α + 1) dn (ν1 t, k1 ) sn (ν2 τ , k2 )
q= √ ,
1
√
2(α − 1)
r= [2sn(ν1 t, k1 ) cn(ν1 t, k1 ) sn (ν2 τ , k2 ) cn (ν2 τ , k2 )

1
+ (α + 1)(β + 1) dn (ν2 τ , k2 ) dn(ν1 t, k1 )], (10.191)
1
γ1 = [ (α + 1)(β + 1) cn(ν1 t, k1 ) dn(ν1 t, k1 ) cn (ν2 τ , k2 ) dn (ν2 τ , k2 )
1
− (α − 1) sn (ν2 τ , k2 ) sn(ν1 t, k1 )],
√
α−1
γ2 = [ (α + 1) sn(ν1 t, k1 ) dn(ν1 t, k1 ) cn (ν2 τ , k2 )
1

+ (β + 1) sn (ν2 τ , k2 ) dn (ν2 τ , k2 ) cn(ν1 t, k1 ) ],
√ √
2 K cn(ν1 t, k1 )
γ3 = − √ , (10.192)
1
@seismicisolation
@seismicisolation
Fig. 10.15 Examples of trajectories from zone I
where
1 = (α + 1) dn2 (ν1 t, k1 ) − 2 cn2 (ν1 t, k1 ) sn2 (ν2 τ , k2 ) . (10.193)
10.17.3.2 In Zone II (−1 ≤ β ≤ 1 ≤ α < ∞)
√
K (α − 1) dn (τ , ν2 )
p= √ ,
2
√
K (α + 1)(β + 1) dn (ν1 t, k1 ) sn (τ , ν2 )
q= √ ,
22
√
2(α − 1)(β + 1)
r=
2
√ √
[ 2 sn(ν1 t, k1 ) cn(ν1 t, k1 ) sn (τ , ν2 ) dn (τ , ν2 ) + α + 1 cn (τ , ν2 ) dn(ν1 t, k1 )],
(10.194)
@seismicisolation
@seismicisolation
Fig. 10.16 Examples of

zone II
√
β+1 √
γ1 = − [−2 α + 1 dn(ν1 t, k1 ) dn (τ , ν2 ) cn (τ , ν2 ) cn(ν1 t, k1 )
22
√
+ 2 (α − 1) sn (τ , ν2 ) sn(ν1 t, k1 )],
√
2(α − 1)
γ2 = [ 2(α + 1) sn(ν1 t, k1 ) dn(ν1 t, k1 ) dn (τ , ν2 )
22
+ (β + 1) sn (τ , ν2 ) cn (τ , ν2 ) cn(ν1 t, k1 ) ],
√ √
2 K cn(ν1 t, k1 )
γ3 = − √ , (10.195)
2
where
2 = (α + 1) dn2 (ν1 t, k1 ) − (β + 1) cn2 (ν1 t, k1 ) sn2 (τ , ν2 ) . (10.196)
@seismicisolation
@seismicisolation
Fig. 10.17 Examples of

trajectories in zone III
10.17.3.3 In Zone III (−1 ≤ β ≤ α ≤ 1 < ∞)
√
K (1 − a) sn (t, ν1 ) dn (τ , ν2 )
p= √ ,
3
√
K (β + 1) dn (t, ν1 ) sn (τ , ν2 )
q= √ ,
23
√
2(1 − α)(β + 1)
r= [sn (t, ν1 ) dn (t, ν1 ) cn (τ , ν2 )
3
+ cn (t, ν1 ) sn (τ , ν2 ) dn (τ , ν2 )], (10.197)
ν2
γ1 = [(a − 1) sn (t, ν1 ) sn (τ , ν2 ) + 2 cn (t, ν1 ) dn (t, ν1 ) cn (τ , ν2 ) dn (τ , ν2 )]
3
√
2(1 − a)
γ2 = [2 dn (t, ν1 ) dn (τ , ν2 )
23
+ (β + 1) sn (t, ν1 ) sn (τ , ν2 ) cn (t, ν1 ) cn (τ , ν2 )]
√
2K cn(t, ν1 )
γ3 = − √ , (10.198)
3
@seismicisolation
@seismicisolation
where
3 = 2 dn 2 (t, ν1 ) − (β + 1) sn2 (τ , ν2 ) cn 2 (t, ν1 ) . (10.199)
10.18 Integrability Issues
To begin with, let us note that just as in the problems considered in the previous
chapters, the equations of motion (10.45) satisfy Jacobi’s divergence condition, which
may be written as
∂ ∂
· ω̇ + · ṗ = 0
∂ω ∂p
and thus we need only one general integral (involving a new arbitrary constant) to
complete the integration of the problem of motion.
The problem of motion of a body by inertia in an ideal fluid is described by
Lagrangian and Hamiltonian equations, isomorphic by the analogy introduced in
this chapter to the equations describing the motion of a rigid body about a fixed point
under the action of an axi-symmetric combination of three classical fields. The last
problem has three degrees of freedom and thus requires for complete integrability
the existence of a fourth integral independent of the three known ones. In any set
of generalized coordinates, say, Euler’s angles, the geometric integral degenerates
into an identity and we are left with three integrals, the number of integrals required
for complete integrability in the sense of Liouville. Thus, Jacobi’s and Liouville’s
approaches lead to the same requirement.
Neither Eqs. (10.45) and (10.41) nor the equivalent Thomson-Lamb equations
were investigated in their full form for the existence of a fourth (complementary)
integral. The situation is somewhat better for Kirchhoff’s equations, which describe
the motion of a body bounded by a simply connected surface. We give here only
brief account of the various research on this matter.
10.18.1 Results Concerning Kirchhoff’s Equations
10.18.1.1 The Case of Tri-Axial Ellipsoid of the Matrix ā
Existence of a real-analytic fourth integral: One of the notable results is due to

Kozlov and Onishchenko [246] (See also [41]), who used Eq. (10.12) to establish
that when the matrices ā,b̄,c̄ are simultaneously diagonal, i.e.
ā = diag(a1 , a2 , a3 ),
b̄ = diag(b1 , b2 , b3 ),
c̄ = diag(c1 , c2 , c3 ),
@seismicisolation
@seismicisolation
and under the condition that a1 = a2 = a3 = a1 , there exists no real-analytic comple-

mentary integral of (10.12) independent of the three known general integrals (10.13),
except in the two cases when the following necessary relations hold between the
matrices:
c1 − c3 c2 − c1 c3 − c2
A) + + = 0, b̄ = 0, (10.200)
a2 a3 a1
Conditions (A) are also sufficient for integrability. They correspond to Clebsch’s
integrable case (Case 2 of Table 10.1 above) under the restriction n = 0, n the regular
precession transformation parameter.5
b1 − b3 b2 − b1 b3 − b2
B) + + = 0. (10.201)
a2 a3 a1
This condition is necessary but not sufficient. The classical case of Steklov (Case 5
of Table 10.1 above, with n = 0, κ = 0) satisfies this condition and existence of the
fourth integral is secured by the additional restriction c̄ = 0.
Branching of solution: The large number of parameters involved in the Thomson-
Lamb equations of motion of a body in a liquid has become an obstacle for further
analytical studies of those equations. In spite of its huge success in the classical
problem, the approach used by Kowalevski [238] to isolate possible cases in which
the solution of equations of motion has only poles as critical points in the complex t-
plane does not seem efficient in the problem of motion of a body in a liquid. However,
analogous result was established for Kirchhoff’s equations:
Under the condition that ā, b̄, c̄ are simultaneously diagonal and a1 = a2 = a3 =
a1 , the general solution of (10.12) is meromorphic only for the cases of Clebsch and
Steklov [316].
In both cases, the complementary integral is known and the explicit time solution
is expressed in terms of Theta functions.
Existence of a single-valued or algebraic fourth integral: The investigation of
existence of a single-valued complementary integral was performed in [36]. It turned
out that when a1 = a2 = a3 = a1 , branching of solutions is an obstacle for existence
of single-valued integrals. It is shown that
Under the condition that ā, b̄, c̄ are simultaneously diagonal and a1 = a2 = a3 =
a1 , the cases of Clebsch and Steklov are the only cases, when Eq. (10.12) admit a
single-valued fourth integral.
5 In fact, the condition b̄ = 0, is over-restrictive. The result holds when b̄ is propertional to Ī =

1
2 tr(I)δ − I = 21 tr(ā−1 )δ−ā−1 . Compare with Case 2 of Table 10.1. The full form, consistent with
that in Table 10.1, was given in [36].
@seismicisolation
@seismicisolation
10.18 Integrability Issues 307
The existence of a polynomial integral of Kirchhoff was also considered in some

recent works [263, 428], following a line due to Darboux.
10.18.1.2 Case when ā Has an Ellipsoid of Revolution
Preserving the assumption of diagonal three matrices and adding the restriction a1 =
a2 , Sadetov [327, 328] has shown that a complementary algebraic integral of the
equations of motion does not exist, except in the Kirchhoff case (Case 1 of Table 10.1)
and the special versions of the cases of Clebsch under the extra-condition a1 = a2 .
Remark: As we have seen, in all methods used to investigate integrability of
the problem, diagonality of all matrices was a common assumption. This situation
greatly reduced the efficiency of those methods. None of them has pointed out a
new integrable case. Ironically, in the case which generalizes the classical case of
Kowalevski, found later by Sokolov, the matrix K has off-diagonal elements. It was
not predicted by any method, but came as a result of the application of a brute force
method. An ansatz of an integral of degree 4 was used and a symbolic program
was used to solve the resulting conditions on the coefficients and on the system
parameters.
10.19 Remark Concerning Particular Solutions

of the Problem
The above tables of general and conditional integrable cases of Thomson-Lamb and
Kirchhoff equations give a complete up-to-date list and full identification of those
cases. Although we also know a large number of particular exact solutions, we have
not tried to make a complete list of them. At present, some of those solutions are
scattered in journal papers. We have described the most important of those solutions
in the present chapter, as examples on solutions of various forms of the equations of
motion and also in examples of application of the regular precession transformation.
The largest collection of particular exact solutions of problems of motion of a body
in a liquid may be found in books of Gorr and co-authors [121, 125, 126]. Cases
are classified by the nature of motion: permanent rotations, regular precessions,
semi-regular precessions and so on. However, those books concentrate more on the
research of the Donetsk group and in general on results published in Russian. Some
results may have been disguised by the use of various sets of variables and may need
careful revision. In general, further effort is needed to compare, complete, classify
and tabulate all existing results.
@seismicisolation
@seismicisolation
10.20 The Donetsk School of Mechanics and Its Attitude

to Competing Works
Be not proud because thou art learned; but

discourse with the ignorant man as with the sage.
For no limit can be set to skill, neither is there
any craftsman that possesseth full advantages
Ptah-Hotep (2880 BC) [72]
Although founded by Euler and developed by the basic works of D’Alembert,

Poisson and Lagrange, the field of dynamics of a rigid body acted upon by various
forces suffered from stagnation for almost a century. Over that period, the search
for integrable cases or particular solutions didn’t lead to any notable results, even in
the simplest problem, the classical problem of motion of a body about a fixed point
under its own weight, more than Euler’s and Lagrange’s cases.
The first breakthrough in the classical problem was made, in 1888, by Kowalevski,
who discovered the third integrable case. To that she was not led by a physical or
mechanical conservation law, as in the previous two cases, but was led only by a purely
mathematical property of the solution of the equations of motion. Kowalevski’s
success encouraged a number of several of the classics in mathematics and mechanics
to invest huge efforts in the same problem. Over the next two decades, the search of
such eminent scientists as Joukovsky, Lyapunov, Steklov, Chaplygin and Goryachev
produced several integrable cases and particular solutions not only in the classical
problem, but also in the gyrostat problem and the problem of motion of a body in a
liquid. We have listed those results in relevant chapters of this book. It can be noted
that of the eight solutions known up to the first decades of the twentieth century,
four cases were found by Russian authors. The next four decades have brought
no significant changes in the status of the field, but in 1948, Grioli announced the
discovery of a regular precession about an axis inclined to the vertical.
Donetsk school headed by P.V. Kharlamov has made a significant advance in the
subject of rigid body dynamics in the period extending from the mid-fifties to the
late eighties. For most of this period the Donetsk school comprised a large number of
coworkers who worked on all aspects of the classical problem and its generalizations,
and especially, the problem of motion of the gyrostat. The group made several notable
achievements: three new particular solutions of the classical problem raised the ratio
of exact particular solutions constructed by Russian-writing authors to seven cases out
of a total of twelve known at the present time. Donetsk school’s success was exclusive
in the problem of motion of a heavy gyrostat about a fixed point. As pointed out in
Sect. 15 of Chap. 5, a considerable part of our present knowledge of exact particular
solutions of the equations of motion of a gyrostat belongs to that school. Those are
mostly cases generalizing known ones of the classical problem, but a few ones have
no analogs in the classical problem. However, the Donetsk school did not find any
general integrable cases of the gyrostat problem. This is a key remark to which we
shall return later.
@seismicisolation
@seismicisolation
10.20 The Donetsk School of Mechanics and Its Attitude to Competing Works 309
The group used a “brute force” policy in the search for exact solutions. Problems
are scanned for the possibility of admitting a solution of a prescribed form. Each of
the resulting cases certainly required a high cost of manual calculations. The view of
an expression or a solution with coefficients written in one or two pages was normal
and mostly expected. In the classical field of rigid body dynamics no easy results
are left. The group earned credibility and authority in the area of constructing exact
solutions and renewed the spirit that prevailed at the turn of the 19th to the 20th
century, when the field of rigid body dynamics was, almost completely, a Russian-
language science.
Researchers from the Donetsk school have shown that the few results announced
in the thirties by Field, Corliss and Fabbri mainly repeat or are special cases of the
former results of Russian authors. Also, a result of Mertsalov (1946) was shown to
be in error. The overall performance of the research group was more than successful.
This gave the group a sense of responsibility to the Russian heritage: they kept its
competence among other schools of mechanics and turned into custodian not only
of the Russian contributions but also of the whole subject of rigid body dynamics.
In the mid-eighties, the author introduced the simple transformation discussed in
detail in Sect. 10.11, which led to an automatic generalization of all general and con-
ditional integrable cases as well as particular solutions of the classical problem and
its generalizations by inserting an additional parameter n that invokes a simultaneous
combination of potential and gyroscopic forces. Results have interpretations as new
integrable and solvable cases in the problem of motion of a body in a liquid. Nearly
at the same time, the author devised a method for constructing two-dimensional inte-
grable systems that admit a complementary integral, polynomial in velocity. This
method had two main advantages. Firstly, it produced systems living on Riemannian
manifolds and not only on flat spaces. Secondly, those systems are time-irreversible,
and thus accommodate reductions of 3D systems with a cyclic integral. Those two
advantages made the method able to obtain a new integrable system that needed some
restrictions to produce a case of motion of the gyrostat, which turned out to be the
long-waited generalization of the historical Kowalevski’s case, by adding a rotor to
the body along its axis of dynamical symmetry (For details, see Chap. 5 Sect. 5.6).
There were some other new results, like the new form of the equations of motion of
a body in a liquid, which we presented in detail in Sect. 10.4.
The new results have shocked the Donetsk school in more than one way. On one
hand, a significant contribution came from outside the Donetsk school. On the other
hand, no brute force was used, nor needed, in obtaining those results. The Donetsk
school behaved in reaction to the appearance of the new results in a strange way.
We give here few brief quotations from the publications of members of the Donetsk
school of mechanics, to show to what extent some scientific criticism can go when
a strongly overconfident group of researchers have full control over a well-known
scientific journal. A rebuttal of some of those criticizing publications was published in
2001 in [405], too late, after some comments were included in Mathematical Reviews
[281] and Zentralblatt (Zbl 1025.70007). After the publication of our article [405], it
seems that the Donetsk group, at last, realized that they were in error, nevertheless,
no one of the authors of the aggressive publications came out to declare that. The
@seismicisolation
@seismicisolation
direct aggressive series of criticism was stopped. Only in few occasions they were
resumed, and mainly indirectly (e.g. [203]).
After the detailed presentation of the problem of motion of a body in a liquid in
this chapter, one can hardly need any comments on the claims in the papers of the
Donetsk group. However, we find it necessary to pick up some of the most offensive
claims. In fact, there are some lessons here to be learned from them.
10.20.1 The Attitude to the Uniform Precession

Transformation
In a series of publications, which were brought to our attention only in the fall of
1996, the authors claimed that this method of generalization is void, meaningless
and leads to nothing new [200, 211, 216, 217, 292, 350]. We give few quotations,
referring the reader to original sources.
(A) In a rare example of unjustified criticism, one of those authors (Kharlamov
P.V. [200]) stated, after appraising the criticism in [216, 217, 350], that this criticism:
“restores the truth in the most difficult problems of the dynamics of a rigid body, and cleans
the field of study from rubbish introduced by faulty and illiterate papers of H. M. Yehia ... ”
[200].
The same quotation was included in the review (MR 93g: 01040), written for
Mathematical Reviews by Konosevich, the colleague of the authors in the same
institute.
(B) Another author writes [350]:
“H. Yehia has announced so significant results, that, in case they were true, the state of the
classical problems of rigid body dynamics could have radically changed. ...
Astonishing is the lightness with which the achievements of the greatest scientists including
prominent nationals were “generalized” in a single stroke by means of a trivial change of
variables and introducing a nonsignificant parameter to the system. But neither V. A. Steklov,
N. E. Joukovsky, S. A. Chaplygin nor G. V. Kolossov can defend themselves against Yehia’s
generalizations” .
(C) As was explained by Kharlamov in [200] (The same reasoning also in [216,
217, 350]), the main point of their criticism is the following:
“Let the system of differential equations
ẋi = X i (x1 , x2 , ..., xn ), i = 1, ..., n (10.202)

be transformed by means of the invertible substitution
yi = yi (x1 , x2 , ..., xn ; ν), i = 1, ..., n (10.203)

to the form
ẏi = Yi (y1 , y2 , ..., yn ; ν), i = 1, ..., n. (10.204)
If (10.202) admits an integral
@seismicisolation
@seismicisolation
10.20 The Donetsk School of Mechanics and Its Attitude to Competing Works 311
I (x1 , x2 , ..., xn ) = const. (10.205)

then (10.204) has the integral
J (y1 , y2 , ..., yn ; ν) = const. (10.206)

obtained from (10.205) by the substitution (10.203). Even a beginner in Mathematics can
realize that the factiously introduced parameter ν in (10.204) and (10.206) can have no
significant meaning, since it can be eliminated again by the use of the inverse transformation
of (10.203). Thus Yehia’s idea to generalize in this way all the known results in the dynamics
of rigid bodies (belonging to Joukovsky, Kowalevski and others) is empty and meaningless”.
(1) The first lesson here to learn is that a scientific journal not owned and edited by
that research group, could not allow the use of words like “rubbish, faulty and
illiterate” to describe the publications of a competing author.
(2) The second is that over-confidence caused the whole group to deny or disbelieve
scientific achievements of others.
(3) The third lesson is that the review data bases Mathematical Reviews and Zen-
tralblatt sometimes adopt the easy solution: to assign each of the members of
a certain scientific institution to review the publications of other members, and
thus allowing less probability of fair reviews and objective evaluations. In our
case, each of the members of the IAMM (Institute of Applied Mathematics
and Mechanics) reviewed other members’ works. The circle is thus closed: The
Authors are the Editors of the Journal (Mekh. Tverd. Tela) and reviewers of their
articles and, at last, the reviewers of their publications for the MR and Zbl bases.6
10.20.2 The Attitude to the Equations of Motion in the Form

(10.45)
In 2001, Kharlamov P.V., Mozalevskaya G.V. and Lesina M.E. published the paper
[203], in which the equations of motion of a body in a liquid are observed to be written
in four different forms, as per the choice of the principal variables in the equations.
The first is the classical Tomson-Lamb Eq. (10.16) using the variables ω, u. The
second is (10.23) using M, p. In fact, they use a slightly modified form due to
Kharlamov [192], praising this form as being chosen by Chaplygin and Kharlamov in
their research and giving it the term “principal (main) representation” of the equations
of motion. The third form uses ω and p(γ), which are in fact our equations presented
in Sect. 10.4 and deduced originally in 1986 [383], but they are not presented in [203]
in full form. The fourth form uses M, u and is termed as the worst choice. As the
authors tried to give references and names for the first two forms, they pass by
the third form of the equations without giving any references nor referring to any
6 In fact, articles were rejected from publication in the Russian journal PMM J. Appl. Math. Mech.
(See [200]). Namely, this rejection evoked the publication of the whole series of papers in “Mekh.
Tverd. Tela”.
@seismicisolation
@seismicisolation
authority in the field. In particular, our 1986 paper [383], which is most relevant to
this context was not mentioned nor cited in [203].
It may be interesting in this context to recall the next quotation from the Zentral-
blatt review (Zbl 1025.70007) concerning the above paper [203] and just repeating
all the claims advanced in that paper:
“It is noticed, that the objective factors, that characterize the given mechanical
object, should be separated from the subjective factors, brought by the investigator
into the mathematical model of this object. In this connection it is shown, that the
equations used by H.M. Yehia in some of his papers are not new, but they are partial
cases of Kirchhoff’s equations. It is also noticed that some of the generalizations
of known integrable cases given by H.M. Yehia are not new too, but they can be
obtained from the initial integrable cases by coordinate transformation. To avoid
such mistakes the authors suggest that all results in this area should be compared
with the corresponding results for the main form7 of Kirchhoff’s equations.”
Reviewer: Boris Ivanovich Konosevich (Donetsk)
It is notable here that the reviewer and the authors of the article [203] are members
of the same institute.
10.21 Exercises
(1) A solid of revolution moves through a liquid and its kinetic energy T is given by
1
T = [A( p 2 + q 2 ) + Cr 2 + A (u 21 + u 22 ) + C u 23 ].
2
Prove that the steady motion given by
p = q = 0, r = , u 1 = u 2 = 0, u 3 = v
is stable in the linear approximation, provided
AC (A − C )
2 = 4v 2 .
A C 2
[Lamb]
(2) Show that in the classical problem of motion of a heavy rigid body fixed from
on point the permanent rotations around a tilted axis (Sect. 10.8) is possible,
only when the body is fixed from its centre of mass, and the axis of rotation is a
principal axis of inertia of the body at the fixed point.
7 The “main form” means the second representation, i.e. the one used by Kharlamov (See the last
paragraph).
@seismicisolation
@seismicisolation
10.21 Exercises 313
(3) A body fixed from its centre of mass moves under the action of forces with
potential V = 21 Ji j γi γ j , γi are the direction cosines of a certain line fixed in
space. Show that a uniform rotation of the body about an axis inclined to that
line is possible only when the axis of rotation is a common principal axis of the
matrices J, I and this axis takes a horizontal position.
(4) A body bounded by a simply connected surface is moving in an ideal incom-
pressible fluid, infinitely extending in all directions and at rest at infinity. The
equations of motion have the form (10.45), with κ= a = 0. Show that uniform
rotation about the z-axis of the body is possible only if the three matrices have
the form:
⎡ ⎤
J11 0 0
J = ⎣ 0 J11 0 ⎦ ,
0 0 J33
⎡ ⎤
K 11 0 K 13
K̃ = ⎣ 0 K 11 K 23 ⎦ ,
K 13 K 23 K 33
⎡ K 11
⎤
I11 I12 K
2(J33 −J11 ) 13
⎢ K 11 ⎥
I= ⎣ I12 I22 2(J33 −J11 ) 23 ⎦ ,
K
K 11 K 11
K K
2(J33 −J11 ) 13 2(J33 −J11 ) 23
I33
provided J33 = J11 , K 11 = 0, the angle θ0 is chosen arbitrarily and the angular
speed of rotation
2(J33 − J11 ) cos θ0
=− .
K 11
(5) Consider the critical cases of exercise 4: K 11 = 0, J33 = J11 and K 11 = 0, J33 =
J11 .
(6) Show that the consecutive application of two transformations with parameters
n 1 , n 2 is equivalent to the application of one transformation with the parameter
n1 + n2.
(7) In Sect. 10.14.4, when K̄ 11 = 0 use Eq. (10.160) to show that the relation between
the rotation angle ϕ and time is determined from the equation

dϕ
t = − K̄ 11 (10.207)
J13 sin ϕ + J23 cos ϕ + a3
under the condition that the parameters of the body are given by
−C J13 −C J23
a = a3 ( 2
, 2
, 1), κ = (0, 0, κ3 ),
K 11 K 11
@seismicisolation
@seismicisolation
⎛ ⎞
J11 − C JK132J23 J13
⎜ 11
⎟
J = ⎝ − C JK132J23 J11 + C
2 (J13
K 11
2
− J23
2
) J23 ⎠ ,
11
J13 J23 J33
⎛ ⎞
K̄ 11 0 K̄ 13
K̄ = ⎝ 0 K̄ 11 K̄ 23 ⎠ . (10.208)
K̄ 13 K̄ 23 K̄ 33
Show that in contrast to the case of pendulum motion, this motion has a definite
energy value, which depends on the parameters of the body
1 C
E= [J11 + 2 (J13
2
+ a32 )]. (10.209)
2 K 11
(8) Starting from the Lagrangian (10.49) of the generalized problem (10.45) (the
Routhian of the problem of motion of a body in liquid after ignoring the cyclic
translational coordinates):
(a) Ignore the angle of precession retaining the Poisson variables (the compo-
nents of γ) as redundant configurational variables.
(b) Apply Hamilton’s principle in the form of Jacobi to the reduced time-
irreversible Routhian system.Equations of motion are deduced from a vari-
ational problem of the type δ Rdt = 0. Applying Maupertuis’ principle to
eliminate the time differential from the variational problem.
(c) Use γ1 as the independent variable and obtain the following second-order
differential in γ3 , to which the equations of motion of the body in a liquid
are reduced on the integral level {I1 = h, I2 = f } [384]:
D(1 − γ12 − γ32 )γ3 + Cγ3 (1 − γ32 )

− γ1 [A − (A + 2C)γ32 ]γ3 + γ3 [C − (C + 2 A)γ12 ]γ32

− Aγ1 (1 − γ12 )γ33
ρ
− {Cγ3 [(A − B)(A + B − C)γ12 + B(B − C)(1 − γ32 )]
ABC D
+ Aγ1 [(B − C)(B + C − A)γ32 + B(A − B)(1 − γ12 )]γ3 }
ρ ∂V1 ∂V1
+ [ (λ + μγ3 ) − (μ + νγ3 )]
2 ABC(h − V1 ) ∂γ3 ∂γ1
ρ3/2
+
ABC a D 3 (h − V1 )
× { f [(A − B)(A + B − C)γ12 − B(A − B + C) + (C − B)(B + C − A)γ32 ]
+ }
= 0, (10.210)
@seismicisolation
@seismicisolation
10.21 Exercises 315
where
ρ = λ + 2μγ3 + νγ32 ,
λ = C[B(1 − γ32 ) + (A − B)γ12 ],
μ = ACγ1 γ3
ν = A[B(1 − γ12 ) + (C − B)γ32 ].
and
1 1 1
V1 = a · γ + γJ · γ+ [ f − κ · γ + K · γ]2 ,
2 2D 2
= Dκ · γI+κ · γ[D tr(I) − 2|γ I|2 ]
+ |γI|2 γK · γ + D[tr(K)D − γIK · γ − tr(I)γK · γ].
(9) Under conditions (10.163) a solution of the orbital equation in the previous exer-
cise is possible in the form γ3 = 0.
[This characterizes the precessing pendulum motion, including the pendulum

motion about a fixed axis.]
(10) Let the particle of unit mass and unit electric charge moving on the fixed smooth
ellipsoid
Ax 2 + By 2 + C z 2 = 1
be acted upon by forces with potential
1 k J2
V = [ 2 2 + ]
2 A x + B 2 y2 + C 2 z2 (A2 x 2 + B 2 y 2 + C 2 z 2 )2
where k, J are constants, and effective magnetic field H whose component Hn

orthogonal to the surface is given by
[A2 (B + C − A)x 2 + B 2 (C + A − B)y 2 + C 2 (A + B − C)z 2 ]

Hn = J .
[A2 x 2 + B 2 y 2 + C 2 z 2 ]5/2
Show that this system admits in addition to Jacobi’s integral, the quadratic integral
A3 x 2 + B 3 y 2 + C 3 z 2
I = (A2 x 2 + B 2 y 2 + C 2 z 2 )(A ẋ 2 + B ẏ 2 + C ż 2 ) − k
A2 x 2 + B 2 y 2 + C 2 z 2
[BC(B − C)yz ẋ + C A(C − A)zx ẏ + AB(A − B)x y ż]
+ 2J
A2 x 2 + B 2 y 2 + C 2 z 2
[A2 (B + C − A)x 2 + B 2 (C + A − B)y 2 + C 2 (A + B − C)z 2 ]
+ J2 ,
(A2 x 2 + B 2 y 2 + C 2 z 2 )2
@seismicisolation
@seismicisolation
and is consequently integrable.

[Use the Lagrangian (10.53) with the choice
1
V = b(Aγ12 + Bγ22 + Cγ32 ),
2
l = n(Aγ1 , Bγ2 , Cγ3 ),
which characterize Clebsch’s case of tri-axial body (Case 2 of Table 10.1). After
Routhian reduction by the cyclic variable ψ perform Minkowsky change of vari-
ables. For detailed solution see [412].
11. A pendulum of unit length whose bulb has unit mass and carries a unit electric
charge is moving under the influence of forces whose potential is V (r) and a magnetic
field H(r). Show that the equations of motion on the unit sphere can be written in
the form [404]:
∂V
r×r̈ = Hr ṙ − r × , (10.211)
∂r
r = (x, y, z) is the position vector of the bulb, Hr = H · r is the radial component of
the magnetic field. The motion is completely determined by the two scalar functions
V and Hr . The two cases of motion of a dynamically spherical body in a liquid
generate the following two cases of motion of a particle on the sphere:
(1) The case corresponding to Clebsch’s case.
It is characterized by the pair of functions
V = ax 2 + by 2 + cz 2 ,
Hr = f. (10.212)
The second integral of motion for this case can be obtained from Clebsch’s integral
substiuting ω→ f r − r×ṙ (compare to (2.33)).
I = a(y ż − z ẏ − f x)2 + b(z ẋ − x ż − f y)2 + c(x ẏ − y ẋ − f z)2

− (bcx 2 + cay 2 + abz 2 ). (10.213)
This case is a non-separable generalization of the well-known separable Neumann

integrable problem [294] by the presence of the gyroscopic forces and reduces to it
when f = 0.
(2) The case corresponding to the Rubanovsky–Lyapunov case
@seismicisolation
@seismicisolation
10.21 Exercises 317
abc x 2 y2 z2
V = s1 x + s2 y + s3 z − ( + + )
2 a b c
1
+ [2 f + (b + c)x 2 + (c + a)y 2 + (a + b)z 2 ]2 ,
8
1
Hr = f + [a + b + c − 3(ax 2 + by 2 + cz 2 )]. (10.214)
2
The second integral of motion is
I = (b + c)(y ż − z ẏ − N x)2 + (c + a)(z ẋ − x ż − N y)2

+ (a + b)(x ẏ − y ẋ − N z)2 + s1 [(N + a)x + z ẏ − y ż]
+ s2 [(N + b)y + x ż − z ẋ] + s3 [(N + c)z + y ẋ − x ẏ]
− (bcx 2 + cay 2 + abz 2 ) (10.215)
where N = f + 21 [(b + c)x 2 + (c + a)y 2 + (a + b)z 2 ].
@seismicisolation
@seismicisolation
Chapter 11
The General Problem of Motion of a
Rigid Body Acted upon by a Coaxial
Combination of Potential and Gyroscopic
Forces
11.1 Introduction
In the last chapter, we have seen that the problem of motion of a body in a liquid
or, more precisely, the alternative problem of motion of a body about a fixed point,
while acted by magnetic, electric and Lorentz forces, lies on the top of a hierarchy
of problems, each of which generalizes the one below it. In this chapter, we extend
this hierarchy upwards, by allowing general axi-symmetric potential and gyroscopic
forces to act on the body. The fact that problems on that level of complication were
not treated in the literature in no way means that such problems have little phys-
ical significance. A natural reason is that the grave theoretical difficulties met in
as simple as the classical problem gave the impression that difficulties will grow
with the degree of complication of forces applied to the body. Fortunately, it turned
out that certain symmetries grow with the complication, opening wide chances to
achieve far-reaching results. In fact, one can go along the line of thinking that led
to the precession transformation in the last chapter, but this time replacing the con-
stant precession speed n with a function ν(γ). Under different circumstances, this
type of transformation keeps the equations of motion of the new problem form-
invariant, leading to construct new integrable/solvable cases from all known cases of
the previous chapters. To this end in this chapter, we shall use two different types of
transformations which can be applied to all the known integrable cases to generate
from them new ones of the most complicated structure ever seen, while preserving
integrability either general or restricted to a certain level of the areas integral. Some
of the new cases can be given definite and non-trivial physical interpretation. In this
respect a word of warning is due. As stressed in previous chapters, we are dealing
with physical models, which have their obvious limitations. Both relativistic effects
and the radiation from accelerated electric charges are permanently neglected.

https://doi.org/10.1007/978-3-030-96336-1_11
@seismicisolation
@seismicisolation
320 11 The General Problem of Motion of a Rigid Body Acted upon by a Coaxial …
Now we assume a rigid body moving about a fixed point, while subject to conservative
(time-independent) potential and gyroscopic forces of the most general form with a
common axis of symmetry O Z fixed in space and passing through the fixed point O
of the body. The Lagrangian has the form
1
L= ωI·ω + l.ω − V, (11.1)
2
in which V = V (γ1 , γ2 , γ3 ), l = l(γ1 , γ2 , γ3 ). The precession angle ψ is a cyclic
coordinate. The corresponding cyclic integral is
∂L ∂ L ∂ω
= · = (ωI + l) · γ = f. (11.2)
∂ ψ̇ ∂ω ∂ ψ̇
To write down the dynamical (Euler-like) equations of motion in the body System,
we first deduce the equation corresponding to the angle ϕ (the proper rotation angle
around the z-axis fixed in the body):
d ∂L ∂L
− = 0.
dt ∂ ϕ̇ ∂ϕ
That is
d ∂ω ∂l ∂V
(Cr + l3 ) − [(ωI + l)· + ω· − ] = 0,
dt ∂ϕ ∂ϕ ∂ϕ
and after expressing derivatives w.r.t. ϕ in terms of derivatives w.r.t. γ, it can be

written as
∂(l · γ) ∂ ∂(l · γ) ∂
C ṙ + (B − A) pq + p[ − γ2 · l] − q[ − γ1 · l]
∂γ2 ∂γ ∂γ1 ∂γ
∂V ∂V
−(γ1 − γ2 ) = 0.
∂γ2 ∂γ1
The last equation can be given the form
∂(l · γ) ∂ ∂V
k · {ω̇I + ω×[ωI + −( · l)γ]−γ × } = 0.
∂γ ∂γ ∂γ
Now, we note that nothing in the curly bracket depends on the unit vector k figuring
before that bracket. This vector can be replaced in the last equation by any of the
other two unit vectors i and j. Thus, we can write the dynamical equation in the final
vector form
@seismicisolation
@seismicisolation
∂V
ω̇I + ω × (ωI+μ) = γ × , (11.3)
∂γ
where
∂
μ = l+(γ× )×l
∂γ
∂ ∂
≡ (l·γ) − ( ·l)γ. (11.4)
∂γ ∂γ
Equation (11.3) and Poisson’s equation constitute the system
∂V
ω̇I + ω×(ωI + μ) = γ× ,
∂γ
γ̇ + ω×γ = 0, (11.5)
of six first-order equations in 6 unknowns, which generalizes the equations of motion

in all problems considered in the previous chapters. We shall refer to V and l as the
scalar and vector potentials, respectively, and to μ as the gyroscopic vector.
It is easy to check that the system (11.5) satisfies Jacobi’s condition for the last
integrating multiplier
∂ ω̇ ∂ γ̇
+ ≡ 0.
∂ω ∂γ
Hence, for its integration one needs a single additional integral of motion I4 besides
the three general integrals, which we write in the form
1
I1 ≡ ωI · ω + V = h,
2
I2 = γ 2 = 1,
I3 = (ωI + l) · γ = f. (11.6)
Those are the energy integral or, more precisely, Jacobi’s integral, the geometric
integral and the cyclic integral corresponding to the coordinate ψ. The last can be
found as
∂L ∂ L ∂ω
I3 = = · = (ωI + l)·γ.
∂ ψ̇ ∂ω ∂ ψ̇
The solution of the system (11.5) determines only ω and γ as functions of t. This
completely determines only the angles θ = cos−1 γ3 and ϕ = tan−1 γγ21 . To obtain ψ,
one has to use the cyclic integral (11.2) together with formulas of Chap. 2 to express
ψ̇ in the form
1 (A − B)γ1 γ2 γ̇3 − C(γ2 γ̇1 − γ1 γ̇2 )

ψ̇ = [f −l ·γ − ]. (11.7)
γI·γ 1 − γ32
@seismicisolation
@seismicisolation
The angle of precession is found by integrating this relation with respect to time, and
this completes the solution.
Remark: It must be noted here that the gyroscopic vector μ, which enters the
equations of motion (11.5), is unique for any physical problem, but the vector poten-
tial l is not. In fact, as was noted before in Chap. 10, a term of the type
dχ(γ) dχ dγ dχ
− =− · = ω·(γ× )
dt dγ dt dγ
can be added to the Lagrangian without changing the equations of motion. Thus, the
vector l can be determined only up to a term of the form
dχ
l 0 = γ× , (11.8)
dγ
in which χ is an arbitrary function of γ.
11.3 Relation to Grioli’s and Kharlamov’s Equations
11.3.1 Grioli’s Equations
Grioli [139] considered the system of equations of motion
∂V (γ)
ω̇I + ω×[ω I + m(ω, γ)] = γ× ,
∂γ
γ̇ + ω×γ = 0, (11.9)
as a generalization of the classical problems of motion of a rigid body about a fixed

point including a general potential function V (γ) and a general gyroscopic term
m(ω, γ). He answered the question: for which m does this system admit an areas
integral? In fact, one can use (11.9) to deduce the relation
d
(ωI·γ) + m·γ̇= 0. (11.10)
dt
If m is expressible in the form
∂ F(γ)
m= + (ω, γ)γ, (11.11)
∂γ
where F and are scalar functions of their arguments, then (11.9) admits the areas
integral
ωI·γ+F(γ) = f (arbitrary constant). (11.12)
@seismicisolation
@seismicisolation
11.3 Relation to Grioli’s and Kharlamov’s Equations 323
Grioli did not study further the system (11.9) with m as in (11.11). Although such
system preserves the sum of kinetic and potential energies, it cannot be presented in
Lagrangian or Hamiltonian form. The powerful techniques of analytical dynamics
are inapplicable to that system.
11.3.2 M. Kharlamov’s Equations
M. Kharlamov considered the same question of existence of areas integral, but

demanded that the system (11.9) had Lagrangian structure [173]. He was led to
the same form of the gyroscopic function as (11.11), but with velocity-independent
, so that
∂ F(γ)
μ= + (γ)γ, (11.13)
∂γ
where F, are two arbitrary functions of γ.

We now prove that gyroscopic terms in the equations of motion (11.5) can be
determined in two equivalent ways, either by giving the vector l(γ) or the pair of
scalar functions F and :
(1) Let l(γ) be given, then
F = l · γ, (11.14a)
∂
= · l. (11.14b)
∂γ
Note that a gauge-term vector l 0 in the form (11.8) gives no contribution to any of
those functions.
(2) Let l(γ), l (γ) be two solutions of (11.14a) and (11.14b) for given F and .
The difference
λ = l −l (11.15)
satisfies the equations

∂
λ · γ= 0, · λ = 0.
∂γ
The general solution of the first equation is
λ = γ × s(γ), (11.16)
and inserting this into the second equation we get
∂ ∂
· (γ × s(γ)) = −γ · ( × s) = 0.
∂γ ∂γ
@seismicisolation
@seismicisolation
This is a single under-determined linear partial differential equation in the three

components of s. Its solution involving two arbitrary functions χ and N is
∂χ
s= + N (γ)γ. (11.17)
∂γ
Inserting this expression into (11.16) and using (11.15), we can write
∂χ
l = l + γ×[ +N (γ)γ]
∂γ
∂χ
= l + γ× . (11.18)
∂γ
Thus, replacing l by l in the Lagrangian (11.1) adds to L a term of the form
∂χ ∂χ dχ
ω·(γ× ) = ·(ω×γ) = − ,
∂γ ∂γ dt
which is a nugatory term, having no contribution to the equations of motion. Khar-

lamov’s form (11.13) for the vector μ is equivalent to our form (11.4).
11.4 Potential of, and Torques on, a Heavy, Magnetized

and Electrically Charged Body
The model of an absolutely rigid body as such is a purely mathematical model. All
ordinary materials suffer deformation under stresses applied to them. Nevertheless,
this model has proved practical, useful and comfortable in the study of a wide spec-
trum of physical and mechanical problems. In this section, we formulate the equations
of motion of a rigid body about a fixed point in a much wider physical setting, taking
into account classical interactions, all at a time. In addition to its mass distribution
acted upon by gravitational forces, assume that the body has some magnetized parts
and carries some electric charges. The body is also subject to electric and magnetic
fields.
The picture to be drawn here for the rigid body and physical effects on it should not
be taken as literally describing a real body with usual properties as electric insulation
or conductivity, magnetic permeability or other properties that change its physical
characteristics when its orientation changes under the action of external fields. Our
aim here is to construct a mathematical model that would lead to tractable equations
of motion of the rigid body in the presence of all the classical physical interactions.
To this end, we make some necessary simplifying assumptions:
1- The main part of the body (the carrier body), which is fixed from the origin O,
has neither electrical nor magnetic properties, so that it does not interfere with the
@seismicisolation
@seismicisolation
11.4 Potential of, and Torques on, a Heavy, Magnetized … 325
interaction between the external fields and the magnets and electric charges carried
by the body.
2- The physical characteristics of the rigid body are constant in it. They do not
change with time, with the change of the body’s orientation in space, nor with the
change of internal forces in the body. Thus, the body may carry a distribution of
immovable electric charges and some permanently magnetized parts, also fixed in
it. Magnetization of the body can also arise due to the presence of steady electric
circuits in the body. An electric motor whose axis is fixed in the body generates in
its working mode a magnetic moment due to electric current in its coil, equivalent
to a permanent magnet, and a constant gyrostatic moment due to the steady rotation
of the coil.
3- It is well known that, according to the laws of classical physics, an accelerated
electric charge emits electromagnetic radiation. This was established by Larmor
[254] in 1897 (see also [161]). The total energy of motion of the body decreases with
time. The maximum acceleration attained by a point of the body will be assumed
small enough to justify neglecting this effect.
Under those conditions, the following effects on the body will be taken into
account:
(1) A torque arises due to the gravitational field g of a certain distribution of gravitat-
ing sources, fixed in the inertial system of axes O X Y Z , O being the fixed point
of the body. Gravitational forces are derivable from a scalar potential Vg (X, Y, Z )
by the relation g = −∇Vg . The gravitational potential is harmonic, i.e. satisfies
Laplace’s equation in the inertial coordinate system outside gravitating sources.
The potential of the body, due to the gravitational field, has the form

VG = Vg (X, Y, Z )dm,
where dm is the mass element at the point r(X, Y, Z ) of the body and integration
is extended on the space domain occupied by the body. Referring to the system
of axes O x yz fixed in the body, we have r = (x, y, z) and hence the potential
can be written as
VG = Vg (r · α, r · β, r · γ)dm, (11.19)
α, β, γ being the basic unit vectors in the inertial space.

(2) The external electric field E, derived from the potential Ve by the relation
E = −∇Ve , acts on the electric charges on the body in a similar way. The electric
potential of the body is

VE = Ve (r · α, r · β, r · γ)de. (11.20)
(3) The external magnetic field H = −∇Vm acts on the magnetized parts of the body.
Note that we use the magnetic field H rather than the magnetic induction B,
@seismicisolation
@seismicisolation
since the body is considered as having unit permeability. Also, for simplicity, we
consider the magnetized part of the body as composed of a set of short magnets
(dipoles). If mi is the magnetic dipole moment at the point ri (X, Y, Z ), the
potential of the body due to the scalar magnetic interaction is
VM = mi · ∇Vm (ri · α, ri · β, ri · γ)

= −mi · H(ri · α, ri · β, ri · γ). (11.21)
(4) The external magnetic field also exerts the velocity-dependent Lorentz forces on
the electric charge distribution in the body. The moment of those forces about
the origin is1
dr
M H = r × [(de ) × H],
dt
where the velocity of the point of the body in space dr

dt
= ω × r. We have

MH = r × [(ω × r) × H]de

= (r · H)ω × rde

= ω × (r · H)rde. (11.22)
This means that the vector μ in the equations of motion (11.3) may be written
in the form
μ = κ− (r · H)rde. (11.23)
For certain purposes, e.g. to construct a Lagrangian for the problem of motion, the
magnetic field can also be derived from a vector potential A, which is also assumed
time-independent, according to the formula H =∇ × A. The vector potential l of
the body may be written as

l =κ+ r × Ade, (11.24)
while μ can be derived from l according to (11.4).

For the purpose of giving a concrete example, let us consider the following phys-
ical situation.
Let the principal body of a gyrostat be carrying a permanent distribution of electric
charges and the system be subject to
(1) A uniform magnetic field H in the Z -direction, i.e. H = H γ.
1 Here MKS units are used. In Gaussian units de should be divided by the velocity of light c (e.g.
[44]).
@seismicisolation
@seismicisolation
11.4 Potential of, and Torques on, a Heavy, Magnetized … 327
(2) An electric field whose potential is a1 Z + 21 a2 Z 2 .

(3) A gravitational field with another quadratic potential b1 Z + 21 b2 Z 2 .
It should be noted that those forms of the electric and gravitational potentials
appear as a second approximation of the potentials of a general rigid body (or gyrostat)
in arbitrary coaxially symmetric electric and gravitational fields, by including the
second harmonics. The same applies for the case of fields due to a distant axi-
symmetric and symmetrically situated invariable body.
According to (11.23), we write

μ = κ− H (r · γ)rde
= κ − 2H γ Īe , (11.25)
where Ī e = 21 (tr Ie )δ − Ie , Ie is the inertia matrix of the distributions, and δ is the

unit matrix. The corresponding vector potential is
1
l = κ+ H γIe . (11.26)
2
On the other hand, the total potential of the system is (ignoring an insignificant
constant)

1
V = [a1 r.γ + a2 (r.γ)2 ]de
2

1
+ [b1 r.γ + b2 (r.γ)2 ]d M
2
1
= a.γ+ γJ·γ (11.27)
2

where J = −a2 Ie − b2 I, a =a1 rde + b1 Mrc , M the mass of the system and rc
its centre of mass. As seen in Chap. 10, formulas (11.25)–(11.27) characterize the
problem of motion of a body in a liquid.
The effect of Lorentz forces on the motion of a rigid body was considered only in
very few works (e.g. [22, 139, 140, 378, 382]). In a number of more recent works,
similar problems were considered, repeating to a great extent previous results as
introducing the inertia matrix of electric charges, e.g. [430–433].
Expressions analogous to the above ones can be derived for more complicated
forms of the magnetic field. In the case when the scalar potential of the external
magnetic field can be expressed as a second-degree harmonic polynomial,
VM = a1 Z + a2 (3Z 2 − r 2 ). (11.28)
The vector μ can be expressed as
@seismicisolation
@seismicisolation

μ= [a1 r · γ + 2a2 (3(r · γ)2 − r 2 )]rde. (11.29)
In expanded form, one may write
μ1 = −2a2 (Ix x x + Ix yy + Ix zz ) + a1 (Ix x γ1 + Ix y γ2 + Ix z γ3 )

+6a2 (Ix x x γ12 + Ix yy γ22 + Ix zz γ32 + 2Ix x y γ1 γ2 + 2Ix x z γ1 γ3 + 2Ix yz γ2 γ3 )
μ2 = −2a2 (Ix x y + I yyy + I yzz ) + a1 (Ix y γ1 + I yy γ2 + I yz γ3 ) (11.30)

+6a2 (Ix x y γ12 + I yyy γ22 + I yzz γ32 + 2Ix yy γ1 γ2 + 2Ix yz γ1 γ3 + 2I yyz γ2 γ3 )
μ3 = −2a2 (Ix x z + I yyz + Izzz ) + a1 (Ix z γ1 + I yz γ2 + Izz γ3 )

+6a2 (Ix x z γ12 + I yyz γ22 + Izzz γ32 + 2Ix yz γ1 γ2 + 2Ix zz γ1 γ3 + 2I yzz γ2 γ3 )

where, for example, Ix x = x 2 de, Ix yz = x yzde and so forth are the second- and
third-degree moments of the charge distribution.
11.5 On General and Conditional Integrable Cases in Rigid

Body Dynamics
As explained in previous chapters of this book, we call a problem general integrable if

I4 exists for arbitrary initial conditions and conditional integrable if it admits a fourth
integral I4 only on a single level f of the cyclic integral I3 (in many cases f = 0)
but for all initial conditions compatible with that level. In both types of integrable
problems, the solution can be reduced to quadratures through the application of
Liouville’s theorem or Jacobi’s theorem to the reduced two-dimensional Hamiltonian
system. It is thus sufficient to point out the fourth integral to ensure integrability in
those cases. In some cases, it becomes possible to construct a quantity constant only
under other restrictions on the initial state of motion, which do not fit as conditions
on the integral level of I3 . Then one cannot apply Liouville’s theorem to construct the
solution and a procedure for accomplishing this task should be indicated separately.
In such cases, we deal with particular solutions of the problem.
Equations of motion of the form (11.5) cover a wide range of applications in rigid
body dynamics. Special cases are the classical problem of motion of a heavy body, its
generalizations to the case of a gyrostat moving under potential and Lorentz forces.
We recall that they cover also the Routhian reduction of the problem of motion of a
body in a liquid, in which the body has no fixed point. In many cases, Eq. (11.5) with
reasonably behaving functions V can be interpreted as characterizing gravitational,
electric and magnetic interactions and μ as the Lorentz force exerted by the magnetic
field on some electric charges resting on the body. However, this is not always the case.
In some problems that happen to be integrable, such interpretation is not possible,
due to the presence of singularities that cannot be exhibited by the potentials of real
@seismicisolation
@seismicisolation
11.5 On General and Conditional Integrable Cases in Rigid Body Dynamics 329
bodies. Detailed examples of integrable problems of both types will be considered

in the next two chapters, Chap. 12 and Chap. 13.
11.6 Transformation of the Equations of Motion
In the preceding chapter, we have applied the transformation ω = ω + νγ, where

ν is a constant, to a system of the type (11.5) and its form-invariance is used to
generate integrable cases containing ν as a parameter. Here we shall develop this
idea, by replacing the constant ν by a function ν(γ). In fact, the substitution
ω = ω + νγ, ν = ν(γ1 , γ2 , γ3 ) (11.31)
leaves the invariant form of the Poisson equation in (11.5), transforming it to
γ̇ + ω ×γ = 0, (11.32)
while the areas integral in (11.6) takes the form
I3 = (ω I + l+νγI) · γ= f. (11.33)
Substituting in the Eulerian part of the equations of motion, using (11.32) and
rearranging terms, we get
∂ν ∂ν
ω̇ I+ω ×(ω I + μ + 2νγI − ν(tr I)γ + γI · γ − (γI· )γ)
∂γ ∂γ
∂V ∂ν
= γ×[ − ν μ− ν 2 γI + (ω I·γ) ]. (11.34)
∂γ ∂γ
On the level I3 = f (say), we substitute ω I · γ from (11.33) and after some manip-
ulations write the equations of motion in the final form:
∂V
ω̇ I + ω ×(ω I + μ ) = γ × ,
∂γ
γ̇ + ω ×γ = 0, (11.35)
where
∂ ∂
μ = μ+ (νI · γ) − [ · (νγI)]γ,
∂γ ∂γ
∂ν
≡ μ−2νγ Ī + γI×( × γ)
∂γ
@seismicisolation
@seismicisolation
1
V = V + ν( f − l·γ) − ν 2 γI·γ, (11.36)
2
and Ī= 21 tr (I)δ−I. From the first of Eq. (11.36) and comparing with (11.4), we can
also write the transformation law for the vector l as
l = l+νγI. (11.37)
Thus, the transformation (11.31) preserves the form of the equations of motion
on a fixed level of I3 , changing only V, μ (or l) to V , μ (or l ). The value f of I3
enters in the potential V as a parameter. The solution of the transformed equations of
motion (11.35) can be obtained from that of (11.5) through the substitution (11.31).
The system of Eq. (11.35) admits the linear integral
I3 = (ω I + l )· γ = f ,
equivalent to (11.33), and also the energy (Jacobi’s) integral
1
ω I · ω + V = h.
2
On the one hand, the transformed system (11.36) can be viewed as the equations
of motion of the original system as in (11.5), as seen by an observer resting in the
reference frame moving with the position-dependent angular velocity ν(γ1 , γ2 , γ3 ).
The new terms that appeared in the transformed system are the inertial forces due to
the rotation of the frame.
On the other hand, there is a different and more constructive way of looking at
(11.36). We shall make use of the situation that the transformation preserves the
form of the equations of motion to understand the transformed equations on their
own as describing the motion of a second body in the inertial frame under the forces
determined by V , μ . In other words, we consider the system (11.36) as formally
generalizing (11.5) to which it reduces when ν = 0. However, this will not prevent us
from relating the solutions of the two systems by the (formal) transformation (11.31).
This duality in interpretation is the key to understanding the present method. From
now on, we will mostly regard the system (11.36) as a generalization of (11.5) rather
than a transformed form of it.
Remark 17 A curious note may be in place here. In certain cases, it is possible from
Eq. (11.36) to choose the function ν so that V vanishes. This means that in those
cases, when the resulting ν(γ) is a real function, the original forces with potential V
can be replaced by purely gyroscopic forces in a properly chosen rotating coordinate
frame. However, we shall not follow this line, since it seemingly has no practical
consequences.
@seismicisolation
@seismicisolation
11.7 Maximal Reduction of the Order of the Equations of Motion 331
11.7 Maximal Reduction of the Order of the Equations

of Motion
The method used in Chap. 9 Sect. 9.2 and an exercise of Chap. 10 can be used here
in the most general case of potential and gyroscopic forces to obtain a second-order
orbital equation connecting two of the geometric variables γi . The Lagrangian of the
problem of motion will be taken in the form (11.1), namely
1
L= ωI · ω + l · ω − V, (11.38)
2
in the redundant configurational variables ψ, γ1 , γ2 , γ3 , subject to the holonomic
condition
γ12 + γ22 + γ32 = 1. (11.39)
The angular velocity may be written as
ω = ψ̇γ+N, (11.40)
where
N = θ̇n + ϕ̇k
γ̇3 γ2 γ̇1 − γ1 γ̇2
= − (cos ϕ, − sin ϕ, 0) + k
1 − γ3
2 γ12 + γ22
(−γ2 γ̇3 , γ1 γ̇3 , γ2 γ̇1 − γ1 γ̇2 )

= . (11.41)
1 − γ32
As a result of cyclicity of the Lagrangian in the variable ψ, we have the integral
∂L ∂ L ∂ω
= · = (ωI + l) · γ = f. (11.42)
∂ ψ̇ ∂ω ∂ ψ̇
Multiplying (11.40) scalarly by γI and using (11.42), we obtain
1
ψ̇ = ( f − l · γ−γI · N), D = γI · γ. (11.43)
D
Then, ignoring ψ we construct the Routhian
ABC −1
R= γ̇I ·γ̇+l ∗ ·γ̇ − V ∗ , (11.44)
2D
@seismicisolation
@seismicisolation
where
1
V ∗ = V (γ)+ ( f − l · γ)2 ,
2D
1 ∂
l ∗ = [γI×l + f (γI·N)]. (11.45)
D ∂ γ̇
Just as in the preceding chapters, applying Maupertuis’ principle to (11.44) and
eliminating γ2 , we arrive at the following second-order differential equation in
γ3 (γ1 ), to which the equations of motion of the problem are reduced on the integral
level {I1 = h, I2 = 1, I3 = f }2 [384]:
D(1 − γ12 − γ32 )γ3 + Cγ3 (1 − γ32 )
−γ1 [A − (A + 2C)γ32 ]γ3 + γ3 [C − (C + 2 A)γ12 ]γ32

−Aγ1 (1 − γ12 )γ33
ρ
− {Cγ3 [(A − B)(A + B − C)γ12 + B(B − C)(1 − γ32 )]
ABC D
+Aγ1 [(B − C)(B + C − A)γ32 + B(A − B)(1 − γ12 )]γ3 }
ρ ∂V ∗ ∂V ∗
+ [ (λ + μγ3 ) − (μ + νγ3 )]
2 ABC(h − V ∗ ) ∂γ3 ∂γ1
ρ3/2
+
ABC a D 3 (h − V ∗ )
×{ f [(A − B)(A + B − C)γ12 − B(A − B + C) + (C − B)(B + C − A)γ32 ]
+}
= 0, (11.46)
where
ρ = λ + 2μγ3 + νγ32 ,
λ1 = C[B(1 − γ32 ) + (A − B)γ12 ],
λ2 = ACγ1 γ3 ,
λ3 = A[B(1 − γ12 ) + (C − B)γ32 ], (11.47)
and
1
V ∗ = V (γ) + ( f − l · γ)2 ,
2D
∂ l × γI
= D2γ · [ ×( )]
∂γ D
∂ 1
≡ D2 · [ γ × (γIs×l)]. (11.48)
∂γ D
2 The positive sign of the square root in (11.46) corresponds the choice of positive sign of the root
in (11.49). If this choice is reversed, Eq. (11.46) is not changed, provided the signs of f and l are
reversed. This is a consequence of the invariance of the system (11.5) with respect to the replacement
t, ω, μ → −t, −ω, −μ.
@seismicisolation
@seismicisolation
11.7 Maximal Reduction of the Order of the Equations of Motion 333
As should be expected, one can verify that a gauge term l 0 (11.8) does not contribute
to the two functions V ∗ and .
Now, having a solution γ3 = γ3 (γ1 ) of the orbital Eq. (11.46), one can obtain the
dependence of γ1 on time by inverting the integral

λ1 + 2λ2 γ3 + λ3 γ32
t= dγ1 , (11.49)
2D(h − V ∗ )(1 − γ12 − γ32 )
and substituting in γ3 , the last is determined in terms of time and then γ2 is found
from the geometric integral. This completes determination of γ and hence the two
Eulerian angles θ and ϕ as functions of t. Thus, we have shown the equivalence of
the reduced Eq. (11.46) to the equations of motion (11.5) on the integral level {h, f },

provided γ3 and γ3 are well defined, i.e. excluded are only trajectories along which
γ1 takes a constant value.
It should be noticed here that the three functions V ∗ , l ∗ and , which occur in
(11.44) and (11.46), are all invariant with respect to the transformation (11.31). This
can be easily verified by replacing the pair (V, l) in them by the pair (V , l ). This
means that, on the integral level {h, f }, the Routhian (11.44), the orbital Eq. (11.46)
and the expressions of γ, θ and ϕ do not change by the variable rotation transforma-
tion (11.31). We shall use this property later in several situations.
To completely determine the position of the body in space, one has to find an
expression for the precession angle ψ by integrating (11.43), which involves the
vector potential l. Using (11.40), one can express the angular velocity ω in the form
1
ω= [γ̇×γI + ( f − l · γ)γ]. (11.50)
D
Not only all the Euler–Poisson variables are thus determined as functions of time,
but also the vectors α, β. For the transformed system (11.35), regarding (11.37),
this process gives
1
ψ̇ = ( f − l · γ−γI·N)
D
= ψ̇ − ν,
ω = ω − νγ, (11.51)
which coincides with (11.31).
11.7.1 The Case of Complete Dynamical Symmetry
For the purpose of future use, we now write down in expanded form the Routhian
(11.44) in the special case when the inertia ellipsoid of the body at the fixed point
becomes a sphere. Then, from (11.44) we have
@seismicisolation
@seismicisolation
1
R= A(γ̇12 + γ̇22 + γ̇32 )
2
f γ3
−[l1 (γ2 γ̇3 − γ3 γ̇2 ) + l2 (γ3 γ̇1 − γ1 γ̇3 ) + (l3 + )(γ1 γ̇2 − γ2 γ̇1 )]
γ12 + γ22
1
−[V (γ) + [ f − l1 γ1 − l2 γ2 − l3 γ3 ]2 ]. (11.52)
2A
11.8 Extensions of Integrable Problems
As a direct application of the transformed equations, we can readily deduce the

following theorems which construct integrable extensions of the known integrable
problems and connect the solutions of the generalized systems to those of the original
problems.
Theorem 11.1 Let the system (11.5) with certain V (γ) and μ(γ) corresponding to
vector potential l(γ), be general integrable, for arbitrary initial conditions, with the
complementary integral I4 = F(ω, γ). Then, upon replacing V, μ by
1
V = V + ν(b − l · γ) − ν 2 γI · γ,
2
∂ ∂
μ = μ+ (νγI·γ) − [ · (νγI)]γ (11.53)
∂γ ∂γ
where ν = ν(γ) is an arbitrary function and b a new parameter, the new system is
integrable on the level
I3 = (ω I+l+νγI) · γ = b. (11.54)
This theorem allows one to generate from an unconditional case (integrable for
arbitrary initial conditions) a conditional case integrable on a single level of the areas
integral I3 , but with additional potential and gyroscopic forces involving an arbitrary
function ν(γ) and an arbitrary parameter b more than the original integrable problem.
To illustrate the feasibility of the generalized problem, one can calculate for it the
reduced potential. One gets
1
V ∗ = V + (b − f )ν+ ( f − l · γ)2 . (11.55)
2D
The extra-parameter b enters in Eq. (11.35), in the equations of motion derived from
the Routhian (11.44) as well as in the orbital Eq. (11.46). The extended problem may
not be integrable for arbitrary initial conditions. However, on the single level f = b,
the reduced potential reduces to that of the original problem. The extended problem
involves one more physical parameter b and under the dynamical condition f = b,
it becomes integrable and its solution has the same number of parameters as in the
solution of the original problem.
@seismicisolation
@seismicisolation
11.8 Extensions of Integrable Problems 335
A quick example can be readily given by the simplest extension of Euler’s case
of the motion of a body under no torques. Let us take V = 0, l = 0 and choose
ν = n + n 1 γ1 , so that
1
V = bν − ν 2 γI · γ, l = νγI. (11.56)
2
A family of solutions of the transformed problem can be written down generalizing
formulas (10.111), (10.112) of Chap. 10 by replacing n by ν in that solution. The
resulting solution is valid on the level f = b.
Theorem 11.2 Let the system (11.5) with certain V (γ) and μ(γ) corresponding
to vector potential l(γ), be general integrable (for arbitrary initial conditions). Let
also V have the structure
V = V0 + b1 V1 + ... + bk Vk , (11.57)
where Vi , i = 0...k and l are functions of γ not involving any of the parameters
b1 , ..., bk and the complementary integral be
I4 = F(ω, γ; b1 , ..., bk ). (11.58)
Then, upon replacing V, μ(l) by
1
V = V0 + b1 V1 + ... + bk Vk − νl · γ − ν 2 γI · γ,
2
∂ ∂
μ = μ+ (νγI · γ) − [ · (νγI)]γ,
∂γ ∂γ
(l = l+νγI), (11.59)
where ν = n 1 V1 + ... + n k Vk and n i are new constants, the new system is uncondi-
tionally integrable with the areas integral
I3 = (ω I+l ) · γ = f, (11.60)
and for it the complementary integral is
I4 = F(ω + νγ, γ; b1 − n 1 I3 , ..., bk − n k I3 ). (11.61)
In fact, comparing the reduced potentials for the original problem characterized
by the pair {V, l} with that of the extended problem characterized by the pair {V , l }
in (11.59), we find
l ∗ = l ∗ ,
@seismicisolation
@seismicisolation
1
V ∗ = V0 + b1 V1 + ... + bk Vk + ( f − l · γ)2 ,
2D

∗ 1
V = V0 + b1 V1 + ... + bk Vk − f (n 1 V1 + ... + n k Vk )+ ( f − l · γ)2
2D
1
= V0 + (b1 − f n 1 )V1 + ... + (bk − f n k )Vk + ( f − l · γ)2 . (11.62)
2D
The potentials V ∗ , V ∗ in (11.62) are identical in form. The only difference is that
each bi is replaced by bi = bi − f n i , i = 1, ..., k, and hence follows integrability
and the form of the integral (11.61). The set of solutions of the extended problem is
the same as that of the original problem. Notable here is the coupling between the
constants which characterize the physical problem, and hence appear in the equations
of motion, and a dynamical constant of motion I3 ,which appears in the process of
integrating those equations. In fact, the phase portrait and phase trajectories of the new
integrable problems are different from their original counterparts provided f = 0.
Theorem 11.2 generates from an unconditional case integrable for arbitrary initial
conditions another unconditional case also integrable for arbitrary initial conditions.
The new system involves k + 1 parameters n 0 , n 1 , ..., n k more than the old one
and renders to it when one puts n 0 = n 1 = ... = n k = 0. According to the problem
setting, the new parameters invoke additional forces in the equations of motion,
which can be given concrete physical interpretation.
The presence of I3 in the expression for I4 in the transformed problem may lead
in certain cases to notable changes in the structure of the integral. For example, we
shall see below a case in which the degree of the quadratic I4 is raised to 3 because
of the appearance of I3 in the coefficients of the quadratic terms.
Theorem 11.3 If {ω = (t, ω ◦ , γ ◦ ), γ = (t, ω ◦ , γ ◦ )}, is the general solution of
the first system satisfying the arbitrary initial conditions {ω = ω ◦ , γ = γ ◦ }, then
for arbitrary ν(γ) the solution of the second system, satisfying the initial conditions
{ω = ω ◦ , γ = ◦ }, is
{ω = (t, ω ◦ + ν(γ ◦ )γ ◦ , γ ◦ )
−ν((t, ω ◦ + ν(γ ◦ )γ ◦ , γ ◦ ))(t, ω ◦ + ν(γ ◦ )γ ◦ , γ ◦ ),
γ = (t, ω ◦ + ν(γ ◦ )γ ◦ , γ ◦ )}. (11.63)
Theorem 11.4 If the first system admits any particular solution {ω = (t), γ =
(t)}, then for arbitrary ν(γ) the second system admits the solution {ω = (t) −
ν((t))(t), γ = (t)}.
The last theorem follows from the fact that the solution of the second system for
the Poisson variables γ is not affected by the function ν(γ).
In the following chapter, we discuss the consequences of the above theorems
in application to known solvable problems of rigid body dynamics. Theorem 11.1
ensures the integrability of the problem (11.35) on the level f of the cyclic integral and
for arbitrary ν(γ) whenever the corresponding problem (11.5) is integrable, either
@seismicisolation
@seismicisolation
11.8 Extensions of Integrable Problems 337
for arbitrary initial conditions or only on a fixed level of the cyclic integral. Theorem
11.2 relates the explicit solutions of the two problems. Theorem 11.3 enables the
generalization, by means of including the function ν, of particular solutions of (11.5),
i.e. solutions not involving any arbitrary constants or involving a number of constants
of motion less than needed to guarantee integrability.
11.9 Transformations of Cyclic Variables
In Sects. 11.6 and 11.8, we have introduced the variable precession transformations
that leave the invariant form of the Euler–Poisson equations of motion. We were also
able to use this transformation, specially designed for rigid body dynamics under
the influence of axi-symmetric forces, to construct integrable extensions of known
cases. It turns out that the same transformation can be attained in a completely differ-
ent way, applicable to any system whose structure involves cyclic coordinates. The
basic idea is that for such system to be integrable, all that matters is the structure
of its Routhian equations of motion after ignoring the cyclic coordinates. We use
a simple observation that several Lagrangian mechanical systems that have differ-
ent Lagrangian and Routhian functions can be reduced to one and the same set of
Routhian equations in the palpable part of the generalized coordinates. Clearly, this
will be the case if the Routhians of those systems differ only by constant terms that
may depend only on the cyclic constants, but not on any of the palpable coordinates
or velocities.
Consider the mechanical system of n + k degrees of freedom, of which k degrees
are cyclic, characterized by the time-independent Lagrangian
L = L(q1 ..., qn , q̇1 ..., q̇n , Q̇ 1 , ..., Q̇ k ). (11.64)
The system admits the cyclic integrals
∂L
= fi , i = 1, ..., k. (11.65)
∂ Q̇ i
Let us consider another system with the Lagrangian
.
k
L = L(q1 , ..., qn , q̇1 , ..., q n , Q̇ 1 + ν1 , ..., Q̇ k + νk ) − βi νi (q1 ..., qn ),
i=1
(11.66)
where βi are certain constants and νi are certain functions of the palpable coordi-
nates q1 ..., qn . We notice that the system (11.66) is time-independent with the cyclic
variables Q 1 , ..., Q k . This system can be considered as a transformation of (11.64)
through the linear time-independent transformation of the cyclic variable rates
@seismicisolation
@seismicisolation
Q̇ i = Q̇ i + νi (q1 , ..., qn ). (11.67)
Consider the motion of the system (11.66) on the same level of the cyclic integrals
as in (11.65), i.e.
∂ L
= fi , i = 1, ..., k. (11.68)
∂ Q̇ i
This is the transformed form of (11.65) according to (11.67).

Now, let R and R be the Routhians of the two systems, then their difference

k
k
k
.
R − R = L − βi νi − Q̇ i f i − (L − Q i fi )
i=1 i=1 i=1

k
= ( Q̇ i − Q̇ i ) f i − βi νi
i=1

k
= ( f i − βi )νi . (11.69)
i=1
The Routhian equations of motion (see, for example, [305, 368]) of the system char-
acterized by (11.64), (11.65) will be identical to those obtained for the transformed
system (11.66), (11.68) if we set { f i = βi , i = 1, ..., k}. In other words, under the
last conditions, the arbitrary functions νi do not affect the solution for the non-cyclic
coordinates.
From the above considerations we draw the following theorems:
1. For constant {νi = n i , i = 1, ..., k}. In this case the right-hand side of (11.69)
is constant, and one can take {βi = 0}. Equations for q1 ..., qn are identical from R
and R.
Theorem 11.5 If the Lagrangian (11.64) is general integrable (for arbitrary initial
conditions), then the Lagrangian
.
L = L(q1 , ..., qn , q̇1 , ..., q n , Q̇ 1 + n 1 , ..., Q̇ k + n k ) (11.70)
is also integrable for arbitrary initial conditions.

It is not hard to see that this theorem applied to the problem of motion of a
body about a fixed point under the action of axi-symmetric fields, i.e. with one cyclic
coordinate ψ (the angle of precession), leads to the uniform precession transformation
introduced in Chap. 10. Note that in this method, we have not used the property of
invariance of the form of Euler–Poisson equations.
Exercise [405]: Apply the last theorem to exercise 5 of Chap. 9, using the trans-
formation ψ̇ → ψ̇ + n, ϕ̇ → ϕ̇ + N , n, N constants. Show that the transformed
integrable Lagrangian is
@seismicisolation
@seismicisolation
11.9 Transformations of Cyclic Variables 339
1
L = [(A + mz 2 )(θ̇2 + sin2 θψ̇ 2 ) + C(ψ̇ cos θ + ϕ̇)2 + ż 2 ]
2
+n[ A + mz 2 sin2 θψ̇ + C cos θ(ψ̇ cos θ + ϕ̇)] + C N (ψ̇ cos θ + ϕ̇)
n2
−{V (z) − [ A + mz 2 sin2 θ + C cos2 θ] − n N C cos θ}. (11.71)
2
Note that the transformation engenders, among other effects, the presence of a
gyrostatic momentum C N along the axis of symmetry and uniform field potential
−n N Cγ3 .
2. For variable {νi = νi (q1 ..., qn ), i = 1, ..., k}
Theorem 11.6 If the system with the Lagrangian
L = L(q1 ..., qn , q̇1 ..., q̇n , Q̇ 1 , ..., Q̇ k ) (11.72)
is integrable for arbitrary initial conditions, then the system whose Lagrangian is
.
k
L´ = L(q1 , ..., qn , q̇1 , ..., qn , Q̇´1 + ν1 , ..., Q̇´k + νk ) − βi νi (q1 ..., qn ) (11.73)
i=1
is integrable for arbitrary functions νi and arbitrary constants {βi } on the level
∂ L´
{ . = βi , i = 1, ..., k} (11.74)
∂ Q´i
of the cyclic integrals.

It should be stressed again that the integrability of the system with Lagrangian
(11.73) in the last theorem is conditional, i.e. valid only for initial conditions consis-
tent with the restriction (11.74), even though the original system (11.72) is integrable
for arbitrary initial conditions. In application to dynamics of a rigid body about a
fixed point in an axi-symmetric field, this theorem reproduces Theorem 1 of the
previous section, which generates a conditional integrable extension from a general
one.
There are, however, very important situations when the new system can be made
integrable for all initial conditions. This depends on the structure of the potential part
of the Lagrangian.
For the sake of clarity and for future applications, we consider in detail the case of
a generalized natural system with three degrees of freedom, of which one is cyclic.
Let
1 1
L= (a11 q̇12 + 2a12 q̇1 q̇2 + a22 q̇22 ) + (c1 q̇1 + c2 q̇2 ) Q̇ + c3 Q̇ 2
2 2
+b1 q̇1 + b2 q̇2 + b3 Q̇ − V, (11.75)
@seismicisolation
@seismicisolation
where ai j , bi , ci , V depend only on q1 , q2 , so that Q is a cyclic variable. On an

arbitrary level of the cyclic integral
c1 q̇1 + c2 q̇2 + c3 Q̇ + b3 = f (11.76)
the Routhian has the form

1 1
R= (a11 q̇12 + 2a12 q̇1 q̇2 + a22 q̇22 ) − [c1 q̇1 + c2 q̇2 + b3 − f ]2
2 2c3
+b1 q̇1 + b2 q̇2 − V. (11.77)
Now we perform in (11.75) the transformation
Q̇ = ν + Q̇ , ν = ν(q1 , q2 ). (11.78)
According to the last theorem, we get the new Lagrangian
1 1
L = (a11 q̇12 + 2a12 q̇1 q̇2 + a22 q̇22 ) + (c1 q̇1 + c2 q̇2 )( Q̇ + ν) + c3 ( Q̇ + ν)2
2 2
+b1 q̇1 + b2 q̇2 + b3 ( Q̇ + ν) − V, (11.79)
integrable on the level of the cyclic integral
c1 q̇1 + c2 q̇2 + c3 ( Q̇ + ν) + b3 = f. (11.80)
Now, ignoring the cyclic coordinate Q in (11.79) with the aid of this integral,
one obtains the Routhian
1 1
R = (a11 q̇12 + 2a12 q̇1 q̇2 + a22 q̇22 ) − [c1 q̇1 + c2 q̇2 + b3 − f ]2
2 2c3
+b1 q̇1 + b2 q̇2 − V + f ν. (11.81)
Note that
R = R + f ν. (11.82)
Let the system with the Lagrangian (11.75) be integrable. This implies the inte-
grability of the system described by the Routhian (11.77), which should admit a
complementary integral, independent of the Jacobi integral (the Hamiltonian). The
transformed system with Lagrangian L is not necessarily integrable. This is clearly
seen from the relation (11.82) between the Routhians R and R . When ν is not a
constant, the two systems have different Routhian equations for the palpable coor-
dinates. The following curious situation arises, which enables us to construct a wide
class of extended integrable problems.
Let the potential V in (11.75) have the structure
@seismicisolation
@seismicisolation
11.9 Transformations of Cyclic Variables 341

V = V0 + ai vi (11.83)
where {ai } are arbitrary constants and V0 , vi are certain functions in the palpable
generalized coordinates q1 , q2 . Let, further, the system (11.75) be integrable for
arbitrary initial conditions. This means that, besides the three general integrals, the
Routhian equations of this system admit a complementary general integral, which
will depend on the set of constants {ai }, say
I4 = F(q1 , q2 , q̇1 , q̇2 , f ; a1 , a2 , ...). (11.84)
If, moreover, we choose ν in the transformation (11.78) in the form

ν= n i vi (11.85)
and substitute this and (11.83) in (11.77) and (11.81), we put the two Routhians in
the form
1 1
R= (a11 q̇12 + 2a12 q̇1 q̇2 + a22 q̇22 ) − [c1 q̇1 + c2 q̇2 + b3 − f ]2
2 2c3

+b1 q̇1 + b2 q̇2 − V0 − ai vi , (11.86)
and
1 1
R = (a11 q̇12 + 2a12 q̇1 q̇2 + a22 q̇22 ) − [c1 q̇1 + c2 q̇2 + b3 − f ]2
2 2c3

+b1 q̇1 + b2 q̇2 − V0 − Ai vi , (11.87)
where Ai = ai − f n i . The only difference between the two is that {ai } are replaced
by {Ai }.
The Routhian equations of motion for the transformed problem are also generally
integrable. They admit the integral
I4 = F(q1 , q2 , q̇1 , q̇2 , f ; A1 , A2 , ...)

= F(q1 , q2 , q̇1 , q̇2 , f ; ai − f n i , a2 − f n 2 , ...). (11.88)
Moreover, the corresponding Lagrangians are also integrable. The integrals I3 and
I3 can be obtained by substituting the parameter f from (11.76) and (11.80), respec-
tively. It is remarkable that this substitution replaces some constant coefficients of
the complementary integral of the transformed problem by ones depending on f,
which can be replaced by its expression involving the velocity variables. The pres-
ence of the added parameters {n i } changes the structure of the integral. Naturally,
the transformed system is a physical generalization of the original one and when all
{n i } vanish one goes back to the original system.
@seismicisolation
@seismicisolation
The above three ways of generalizing integrable systems with cyclic coordinates
can be applied to rigid body dynamics. In the particular case of axi-symmetric fields,
they give the same results as described in the theorems of the last section. The method
using cyclic variables furnishes a great advantage. It does not require invariance of
the Euler–Poisson equations and it will be used in various situations in the sequel,
while dealing with the motion of the body in an asymmetric combination of fields.
@seismicisolation
@seismicisolation
Chapter 12
The Most General Integrable Cases
in Rigid Body Dynamics
In this chapter, we present a small, but most exotic, set of general and conditional
integrable cases, which constitute currently the uppermost level of the hierarchy
of integrable cases in rigid body dynamics. That level is inaccessible for all direct
methods used in mechanics in the past. Methods which investigate the existence of
analytical or polynomial integrals and the existence of single-valued solutions of the
equations of motion are equally hopeless in facing such a wide class of problems.
Here, we speak about several-parameter generalizations of six of the seven integrable
cases in the dynamics of a body in a liquid in two different ways, general and
conditional, by applying Theorems 2 and 3 of the last chapter.
12.1 General Integrable Cases
Listed in Table 12.1 are the most general and most exotic integrable cases known
up-to-date of the problem of motion of a rigid body about a fixed point under the
action of an axi-symmetric combination of conservative potential and gyroscopic
forces. Their generality results from the extra number of parameters (an arbitrary
function in case 7) included in their structure. The first five of the seven general
integrable cases, namely cases occupying positions 1–5 in Table 12.1, are obtained by
applying Theorem 2 of the preceding chapter to construct unconditional integrable
generalizations of all but one of the integrable cases of Chap. 10 concerning the
motion of a body in a fluid. Depending on the structure of the potential, a number of
additional parameters, ranging up to 4, is added to the structure of each case. The case
number 6 of Table 12.1 is obtained by applying the same Theorem 2 to a general
integrable case found by Yehia and Bedwehy. The latter generalizes the classical
Kowalevski case by adding a singular term √ ε 2 to the heavy body potential.
1−γ3

https://doi.org/10.1007/978-3-030-96336-1_12
@seismicisolation
@seismicisolation
344 12 The Most General Integrable Cases in Rigid Body Dynamics
Table 12.1 General integrable extensions of general integrable cases

1 Yehia [398] (1997),
Oreshkina [301],
Clebsch [55]: n 1 = 0,
Brun [47]: n = n 1 = 0,
Tisserand [354]: n = n 1 = 0
V = (Aγ12 + Bγ22 + Cγ32 ){b − 21 [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]2 },
ν = n + n 1 (Aγ12 + Bγ22 + Cγ32 ),
l = [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]γI,
μ1 = γ1 {(A − B − C) [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]
+2n 1 [A(Aγ12 + Bγ22 + Cγ32 ) − (A2 γ12 + B 2 γ22 + C 2 γ32 )]},
μ2 = γ2 {(B − C − A) [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]
+2n 1 [B(Aγ12 + Bγ22 + Cγ32 ) − (A2 γ12 + B 2 γ22 + C 2 γ32 )]},
μ3 = γ3 {[(C − A − B)[n + n 1 (Aγ12 + Bγ22 + Cγ32 )]
+2n 1 [B(Aγ12 + Bγ22 + Cγ32 ) − (A2 γ12 + B 2 γ22 + C 2 γ32 )]}
I3 = Apγ1 + Bqγ2 + Cr γ3
+[n + n 1 (Aγ12 + Bγ22 + Cγ32 )](Aγ12 + Bγ22 + Cγ32 )
I4 = A2 { p + γ1 [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]}2
+B 2 {q + γ2 [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]}2
+C 2 {r + γ3 [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]}2
−2(b − n 1 I3 )(BCγ12 + C Aγ22 + ABγ32 )
I3 = M1 γ1 + M2 γ2 + M3 γ3 ,
M2 M2 M2
H = 1/2 ( A1 + B2 + C3 ) + b Aγ12 + Bγ22 + Cγ32
− (M1 γ1 + M2 γ2 + M3 γ3 ) [n + n 1 (Aγ1 + Bγ22 + Cγ32 )]
2
I4 = 21 (M12 + M22 + M32 )

−ABC(b − n 1 I3 )(BCγ12 + C Aγ22 + ABγ32 )
The last of those cases, number 7, is the ultimate generalization of Lagrange’s case
to the most general case in which the proper rotation angle ϕ is a cyclic coordinate
and the fourth integral is the cyclic integral.
We also give relevant supplementary information and some characteristic proper-
ties of the cases provided in the table of the last section. Some clarifications are made
about the present status of the explicit solution of each of the generalized cases as
per the progress made in solving their primitive counterparts at the lower levels of
the hierarchy.
12.1.1 Table of General Integrable Extensions of General

Integrable Cases
This case includes one parameter n 1 more than Clebsch’s case and two parameters
n, n 1 more than Brun’s problem. As established in the last chapter, the explicit solu-
tion for this case in terms of time can be obtained by the variable precession transfor-
@seismicisolation
@seismicisolation
12.1 General Integrable Cases 345
mation (11.31) from the basic solution n = n 1 = 0, given by Kötter in terms of theta
functions of two arguments for the first integrable case of Clebsch (see Chap. 10).
In [301], an integrable case of M. Kharlamov’s equations was constructed. It
admits a fourth integral quadratic in the angular velocities, with coefficients depend-
ing on γ. To this end, the author used an ansatz for the quadratic integral, and used
consistency conditions with the equations of motion. The resulting expressions are
quite complicated and lack transparency. The author has neither noted the possibil-
ity of transforming this case to Clebsch’s case of motion of a body in a liquid, nor
even any relation to Clebsch’s case as a special case. Consequently, to the end of
her paper, the author states that the existence of the fourth algebraic integral “means
the possibility, in principle, to reduce the problem to quadrature”. Our method gives
an effortless constructive way to build the explicit solution by applying the variable
precession transformation to Kötter’s solution.
2 Yehia [398] (1997),

Kharlamova L. [214] (1990): nc11 = nc22 = nc33
Clebsch’s case of spherical symmetry [55]: n 1 = n 2 = n 3 = 0.
B = C = A,
V = 21 A[c1 γ12 + c2 γ22 + c3 γ32 − (n + n 1 γ12 + n 2 γ22 + n 3 γ32 )2 ],
ν = n + n 1 γ12 + n 2 γ22 + n 3 γ32 ,
l = Aνγ,
μ1 = −Aγ1 [n + n 1 γ12 + γ22 (3n 2 − 2n 1 ) + γ32 (3n 3 − 2n 1 )],
μ2 = −Aγ2 [n + γ12 (3n 1 − 2n 2 ) + n 2 γ22 + (3n 3 − 2n 2 ) γ32 ],
μ3 = −Aγ3 [n + γ12 (3n 1 − 2n 3 ) + γ22 (3n 2 − 2n 3 ) + n 3 γ32 ]
I3 = pγ1 + qγ2 + r γ3 + n + n 1 γ12 + n 2 γ22 + n 3 γ32 ,
I4 = (c1 − 2n 1 I3 ) ( p + νγ1 )2 + (c2 − 2n 2 I3 ) (q + νγ2 )2
+ (c3 − 2n 3 I3 ) (r + νγ3 )2
−[(2n 2 I3 − c2 ) (2n 3 I3 − c3 ) γ12 + (2n 3 I3 − c3 ) (2n 1 I3 − c1 ) γ22
+ (2n 1 I3 − c1 ) (2n 2 I3 − c2 ) γ32 ]
I3∗ = M1 γ1 + M2 γ2 + M3 γ3 ,
H = 21A (M12 + M22 + M32 ) + 21 (c1 γ12 + c2 γ22 + c3 γ32 )
−(n
+ n 1 γ1 + n 2 γ2 + n 3 γ3 ) (M1 γ1 + M2 γ2 + M3 γ3 )
2 2 2
I4 = Ac1 − 2n 1 I3∗ M12 + Ac2 − 2n 2 I3∗ M22 + Ac3 − 2n 3 I3∗ M32

−A[ 2n 2 I3∗ − Ac2 2n 3 I3∗ − Ac3 γ12 + 2n 3 I3∗ − Ac3 2n 1 I3∗ − Ac1 γ22

+ 2n 1 I3∗ − Ac1 2n 2 I3∗ − Ac2 γ32 ]
In this case, the body has spherical dynamical symmetry. The basic case n 1 =
n 2 = n 3 = 0 is a case of the motion of a body in a liquid (Case 3 of Table 10.1 of
Sect. 10.15). It is closely related to the other Clebsch’s integrable case with a tri-axial
body in the same problem. The solution of this case can be expressed in terms of
theta functions of two variables [233] (see also [71]) and so, in principle, will be
the present generalization. However, this point needs a closer examination, as the
present case presents some unusual and rarely met characteristic properties.
(1) The presence of the three extra parameters n 1 , n 2 , n 3 , which we assume dif-
ferent, raises the degree of the polynomial potential V (γ) from 2 to 4 and the degree
of the components of μ from 1 to 3. The combination of forces acting on the body
has turned into a much complicated one, compared to the original.
@seismicisolation
@seismicisolation
(2) The presence of the same parameters raises the degree of the complemen-
tary polynomial integral I4 as a function of the angular velocity components in the
six-dimensional phase space {ω, γ} of the three-dimensional problem from 2 to 3.
However, on every level of the integral I3 , say, I3 = f, the complementary inte-
gral I4 becomes of the second degree. This may be reformulated in the language of
analytical dynamics in the following way: The reduced equations of motion of the
problem under consideration after ignoring the cyclic coordinate ψ admit a quadratic
complementary integral I4 in the other two Eulerian angles {θ̇, ϕ̇}.
(3) An exceptional case arises, when the two sets of constants are proportional
n1 n2 n3
= = = λ (say).
c1 c2 c3
Then I4 takes the form
I4 = (1 − 2λI3 ) {c1 ( p + νγ1 )2 + c2 (q + νγ2 )2 + c3 (r + νγ3 )2

− (1 − 2λI3 ) [c2 c3 γ12 + c3 c1 γ22 + c1 c2 γ32 ]}.
For arbitrary value of I3 , one can cancel the first factor (1 − 2λI3 ) to obtain for I4
the expression
I4 = c1 ( p + νγ1 )2 + c2 (q + νγ2 )2 + c3 (r + νγ3 )2

− (1 − 2λI3 ) [c2 c3 γ12 + c3 c1 γ22 + c1 c2 γ32 ],
which is quadratic in the velocities, but has some linear terms. This is the case of a
quadratic integral found in [214] in a more complicated and less transparent way.
3 Yehia [398] (1997),

Rubanovsky [317]: n 1 = n 2 = n 3 = 0
Lyapunov [267]: n 1 = n 2 = n 3 = a1 = a2 = a3 = 0
B = C = A,
ν = n + n 1 γ1 + n 2 γ2 + n 3 γ3 ,
V = A{a1 γ1 + a2 γ2 + a3 γ3 − 21 (bcγ12 + caγ22 + abγ32 )
+ 21 ν[(b + c) γ12 + (c + a) γ22 + (a + b) γ32 ] − 21 ν 2 },
l = A[− 21 ((b + c)γ1 , (c + a)γ2 , (a + b)γ3 ) + νγ]
μ1 = A[n 1 + γ1 (a + n − 2ν)],
μ2 = A[n 2 + γ2 (b + n − 2ν)],
μ3 = A[n 3 + γ3 (c + n − 2ν)]
I3 = ( pγ1 + qγ2 + γ3 r ) − 21 [ (b + c) γ12 + (c + a) γ22 + (a + b) γ32 ] + ν
I4 = 21 { (b + c) ( p + νγ1 )2 + (c + a) (q + νγ2 )2 + (a + b) (r + νγ3 )2 }
+ (−n 1 I3 + a1 ) ( p + νγ1 + aγ1 )
+ (−n 2 I3 + a2 ) (q + νγ2 + bγ2 )
+ (−n 3 I3 + a3 ) (r + νγ3 + cγ3 )
−(bcpγ1 + caqγ2 + abr γ3 ) − ν(bcγ12 + caγ22 + abγ32 )
@seismicisolation
@seismicisolation
I3∗ = M1 γ1 + M2 γ2 + M3 γ3 ,
H = 21A (M12 + M22 + M32 ) + 21 [(b + c)M1 γ1 + (c + a)M2 γ2 + (a + b)M3 γ3 ]
+[(Aa1 − n 1 I3∗ )γ1 + (Aa2 − n 2 I3∗ )γ2 + γ3 (Aa3 − n 3 I3∗ )]
− A8 [(a 2 + 2bc)γ12 + (b2 + 2ca)γ22 + (c2 + 2ab)γ32 ]
I4 = (b + c)[M1 + A2 (b + c)γ1 ]2 + (c + a)[M2 + A2 (c + a)γ1 ]2
+(a + b)[M3 + A2 (a + b)γ1 ]2
+(Aa1 − n 1 I3∗ )[2M1 + A(2a + b + c)γ1 ]
+(Aa2 − n 2 I3∗ )[2M2 + A(a + 2b + c)γ2 ]
+(Aa3 − n 3 I3∗ )[2M3 + A(a + b + 2c)γ3 ]
−A{bcγ1 [2M1 + A(b + c)γ1 ] + caγ2 [2M2 + A(c + a)γ2 ]
+abγ3 [2M3 + A(a + b)γ3 ]}
Lyapunov’s case s1 = s2 = s3 = n 1 = n 2 = n 3 = 0 [267] was solved by Kötter,

as well as the related Steklov case, in terms of theta functions of two arguments
[235]. This solution will cover the case s1 = s2 = s3 = 0 for arbitrary n 1 , n 2 , n 3 .
It is obvious that to express the solution in the most general case by applying the
variable precession transformation, it suffices to obtain the solution for the basic case
n = n 1 = n 2 = n 3 = 0, s1 s2 s3 = 0. This was not done up to the present time.
4 Yehia [398] (1997),

Yehia [383] n 1 = n 2 = 0,
Yehia [380]: n = n 1 = n 2 = 0,
Kowalevski [238]: k = n = n 1 = n 2 = 0
A = B = 2C,
V = C[a1 γ1 + a2 γ2 − κγ3 ν − 21 ν 2 2γ12 + 2γ22 + γ32 ],
ν = n +n 1 γ1 + n 2 γ2 .
μ1 = C −nγ1 − n 1 γ12 + 2n 1 γ22 + n 1 γ32 − 3n 2 γ1 γ2 ,

μ2 = C −γ2 n + 2n 2 γ12 − n 2 γ22 + n 2 γ32 − 3n 1 γ1 γ2 ,
μ3 = C(κ − 3nγ3 − 5n 1 γ1 γ3 − 5n 2 γ 2 γ3 )
I3 = 2 pγ1 + 2qγ2 + (r + κ)γ3 + ν 2γ12 + 2γ22 + γ32
I4 = [( p + νγ1 )2 − (q + νγ2 )2 − (a1 − n 1 I3 ) γ1 + (a2 − n 2 I3 ) γ2 ]2
+[2( p + νγ1 )(q + νγ2 ) − (a1 − n 1 I3 ) γ2 − (a2 − n 2 I3 ) γ1 ]2
+2κ (r + νγ3 − κ) [( p + νγ1 )2 + (q + νγ2 )2 ]
−4κγ3 [(a1 − n 1 I3 ) ( p + νγ1 ) + (a2 − n 2 I3 ) (q + νγ2 )]
I3∗ = M1 γ1 + M2 γ2 + M3 γ3 ,
M 2 +M 2
H = 2C
1
( 1 2 2 + M32 ) − k M3 + C(a1 γ1 + a2 γ2 )
−(n + n 1 γ1 + n 2 γ2 )(M1 γ1 + M2 γ2 + M3 γ3 )
M12 −M22
I4 = [ 4C 2
− (a1 − nC1 I3∗ )γ1 + (a2 − nC2 I3∗ )γ2 ]2
n1 ∗ n2 ∗
+[ M2C 1 M2
2 − (a1 − C I3 )γ2 − (a2 − C I3 )γ1 ]
2
k M3
+ C 2 ( 2C − k)(M1 + M2 )
2 2
− C2k2 γ3 [(a1 − nC1 I3∗ )M1 + (a2 − nC2 I3∗ )M2 ]
In view of the presence of the coefficients a1 , a2 in the linear terms of V and n 1 , n 2

in the function ν, it is evident that by rotating the x y axes one can eliminate one of
those four coefficients. Suppose we have eliminated a2 , then n 2 will remain there,
although it will not appear if we apply our method to the basic potential without a2 .
Thus, we shall keep the four terms and the resulting case will neither repeat nor be
@seismicisolation
@seismicisolation
included in the other generalization of Kowalevski’s case in the line leading to case
number 5.
The Euler–Poisson variables in Kowalevski’s case were expressed by Kowalevski
herself in terms of hyper-elliptic functions of time [238]. The solution was some-
what simplified and systematized by Kötter [232, 234] (see also [256]). Explicit
expressions for the six variables in terms of the separation variables can also be
found in [113, 256]. Many qualitative and global properties of motion are discussed
in [240] The same problem was treated in a large number of recent works using
methods of modern algebraic geometry and the inverse scattering method (e.g. [71,
145] and references cited therein). Of special interest is the work [152], which relates
the Kowalevski case to a special version ( f = 0) of Clebsch’s case by means of a
rational transformation and thus explicit solutions for the first can be obtained from
that of the other. The same idea was realized for our generalization of Kowalevski’s
case to the gyrostat by Gavrilov, who related it to the full case of Clebsch ( f = 0)
solvable in Theta functions of two arguments [110]. Thus, it becomes evident that the
generalized case under discussion here is, in principle, solvable in terms of the same
class of functions. However, a direct separation of Yehia’s gyrostat is not achieved
yet (see Chap. 5, Sect. 5.6).
5 Yehia [411]
Sokolov [336] n = n 2 = 0
Yehia [380] n = n 2 = c = 0
Kowalevski [238]n = n 2 = c = κ = 0
A = B = 2C,
ν = n + n 2 γ2 ,
c2
V = C[κcγ1 + a2 γ2 − ν(κ + cγ1 )γ3 − 2 (γ1
2 + 2γ32 )
ν2
− 2 (2γ12 + 2γ22 + γ32 )],
l = C(2νγ1 , 2νγ2 , κ + νγ3 + cγ1 ),
μ1 = C(cγ3 − nγ1 − 3n 2 γ1 γ2 ),
μ2 = C[−nγ2 + n 2 2γ12 − γ22 + γ32 ],
μ3 = C(κ + cγ1 − 3nγ3 − 5n 2 γ2 γ3 )
I3 = 2 pγ1 + 2qγ2 + (r + κ + cγ1 )γ3
+(n + n 2 γ2 )(2γ12 + 2γ22 + γ32 ),
I4 = {( p + νγ1 )2 − (q + νγ2 )2 + (a2 − n 2 I3 )γ2 + c2 γ22 + cγ1 (r + νγ3 − κ)}2
+{2( p + νγ1 )(q + νγ2 ) − (a2 − n 2 I3 )γ1 − c2 γ1 γ2 + cγ2 (r + νγ3 − κ)}2
+2κ(r + νγ3 − κ + cγ1 )[( p + νγ1 )2 + (q + νγ2 )2 + 2c( p + νγ1 )γ3 ]
−4κγ3 (a2 − n 2 I3 )(q + νγ2 )
−2κc2 [2γ3 I3 − κγ32 − (γ12 + γ22 + 2γ32 )(r + νγ3 + cγ1 )]
I3∗ = M1 γ1 + M2 γ2 + M3 γ3 ,
M 2 +M 2
H = 2C
1
( 1 2 2 + M32 ) − (κ + cγ1 )M3 + C(a2 γ2 + 2cκγ1 − c2 γ32 )
−ν(M1 γ1 + M2 γ2 + M3 γ3 )
M12 −M22
I4 = [ 4C 2
+ (a2 − nC2 I3∗ )γ2 + c( MC3 − 2κ)γ1 − c2 (γ12 − γ22 )]2
+[ 2C 2 − (a2 − nC2 I3∗ )γ1 + c( MC3 − 2κ)γ2 − 2c2 γ1 γ2 ]2
M1 M2
M 2 +M 2
+κ( MC3 − 2κ)[ 12C 2 2 + 2cγ3 MC3 ] − 2κγ3 (a2 − nC2 I3∗ + 2c2 γ2 ) MC2
2
− 2κc
C [2γ1 γ3 M1 − (γ1 + γ2 )M3 ]
2 2
@seismicisolation
@seismicisolation
Variable separation for the basic Sokolov’s case, n = n 2 = 0, was obtained in

[227] and explicit expressions for dynamical variables are constructed in [70], in
terms of two intermediate variables, which are expressed in genus-2 Theta functions.
At the present level of the hierarchy, after the introduction of the parameters n, n 2 ,
the solution is obtained by applying the relevant precession transformation.
6 Yehia [398] (1997),

Yehia–Bedwehy [419]: n = n 1 = n 2 = N = 0,
Kowalevski [238]: ε = n = n 1 = n 2 = N = 0
A = B = 2C,
ν = n + n 1 γ1 + n 2 γ2 + N 2
1−γ3

V = C[a1 γ1 + a2 γ2 + ε 2 − 21 ν 2 2γ12 + 2γ22 + γ32 ]
1−γ3
l = Cν(2γ1 , 2γ2 , γ3), N γ1
μ1 = C[−nγ1 + n 1 2γ22 − γ12 + γ32 − 3n 2 γ1 γ2 − 3 ],
(1−γ32 ) 2
N γ2
μ2 = C[−nγ2 + n 2 2γ12 − γ22 + γ32 − 3n 1 γ1 γ2 − 3 ],
(1−γ32 ) 2
μ3 = −C[(3n + 5n 1 γ1 + 5n 2 γ2 )γ3 + N γ3 ]
1−γ32
I3 = 2 pγ1 + 2qγ2 + r γ3 + ν(2γ12 + 2γ22 + γ32 )
I4 = [( p + νγ1 )2 − (q + νγ2 )2 − (a1 − n 1 I3 ) γ1
+ (a2 − n 2 I3 ) γ2 ]2
+[2( p + νγ1 )(q + νγ2 ) − (a1 − n 1 I3 ) γ2 − (a2 − n 2 I3 ) γ1 ]2
+2 (ε−N
I3 ) [( p + νγ1 )2 + (q + νγ2 )2 ] + (ε−N I23 )
2
2
1−γ3 1−γ 3
I3∗ = M1 γ1 + M2 γ2 + M3 γ3 ,
M 2 +M 2
H = 2C
1
( 1 2 2 + M32 ) − ν(M1 γ1 + M2 γ2 + M3 γ3 )
+C[a1 γ1 + a2 γ2 + ε 2 ]
1−γ3
M12 −M22
I4 = [ 4C 2
− (a1 − nC1 I3∗ )γ1 + (a2 − nC2 I3∗ )γ2 ]2
n1 ∗ n2 ∗
+[ M2C
1 M2
2 − (a1 − C I3 )γ2 − (a2 − C I3 )γ1 ]
2
N ∗ N ∗ 2
M12 +M22 ε− C I3 [ε− C I3 ]
+ 2C 2 [ 2 ] + 1−γ 2
1−γ 3 3
Separation variables and expressions of the dynamical variables in terms of them

are constructed in [218] for the conditional case 11 of Table 13.1 of Chap. 13, which
covers the Yehia–Bedwehy only on the level f = 0. To cover the present full general
case, explicit solution of the full Yehia–Bedwehy case for f = 0 is needed. This was
not achieved until now. Note that in cases 1–6, the constant I3 (or I3∗ ) figuring in the
expression for I4 can be substituted by its expression in each case as a function in
the components of ω( or M) and γ.
Remark: It would be highly interesting to study how this phenomenon of coupling
constants changes many results and conclusions obtained for all integrable cases of
motion of a body in a liquid and, in particular, the portrait of the integrals of motion,
bifurcation diagrams and the topological classifications of integral manifolds. Those
questions are presently open for all the above six cases in Table 12.1.
@seismicisolation
@seismicisolation
7 Generalization of Lagrange’s case. Two cyclic coordinates ψ and ϕ.

B = A,
V = V0 (γ3 ),
l=(γ1 , γ2 , l3 ),
μ = (−l3 γ1 , −l3 γ2 , l3 − 2γ3 + (γ12 + γ22 ) ),
V0 (γ3 ), (γ3 ), l3 (γ3 ) arbitrary functions, and
l3 , denote derivative w.r. to γ3
I3 = A( pγ1 + qγ2 ) + Cr γ3 + (γ12 + γ22 ) + l3 γ3
I4 = Cr + l3
I3 = M1 γ1 + M2 γ2 + M3 γ3 ,
M12 +M22 M32 l3
H = 21 ( A + C ) − [ A (M1 γ1 + M2 γ2 ) + C M3 ]
l32
+V0 (γ3 ) + 2 A (γ1 + γ2 ) + 2C
1 2 2 2
I 4 = M3
This is the most general case of Lagrange’s type. The body admits not only
dynamical but also physical symmetry about its z-axis. The equations of motion can
be easily reduced to quadratures. In fact, the integrals I3 and I4 can be written as
I3 = A( pγ1 + qγ2 ) + (γ12 + γ22 ) + (Cr + l3 )γ3 = f,

I4 = Cr + l3 = j, (12.1)
where f, j are the integral constants. The first can be reduced to the relation
1 f − jγ3
ψ̇ = [ − ], (12.2)
A 1 − γ32
while I4 gives
C(ψ̇γ3 + ϕ̇) + l3 = j. (12.3)
From here we find, using (12.2),
1 γ3 f − jγ3
ϕ̇ = ( j − l3 ) − [ − ]. (12.4)
C A 1 − γ32
Thus, we have expressed ψ̇ and ϕ̇ in terms of γ3 . To obtain the relation with time,
we use the energy (in fact, Jacobi’s) integral
1
[A( p 2 + q 2 ) + Cr 2 ] + V = h.
2
That is
( j − l 3 )2
A(θ̇2 + sin2 θψ̇ 2 ) + = 2(h − V ),
C
@seismicisolation
@seismicisolation
which can be given the final form
1 ( j − l 3 )2 1
γ̇32 = (1 − γ32 )[2(h − V ) − ] − 2 [ f − jγ3 − (1 − γ32 )]2 . (12.5)
A C A
Denoting the right-hand side of this relation by F(γ3 ), one can make a separation of
variables and find the quadrature

dγ3
t= √ , (12.6)
F(γ3 )
which may be used to express γ3 = cos θ as a function in time, and hence (12.2) and
(12.4) can be integrated to obtain the angles ψ and ϕ, respectively.
The above formulas are direct generalization of their lower counterparts in the
hierarchy, beginning from Lagrange’s top to Kirchhoff’s case of the motion of a body
in a liquid (Case 1 of Table 10.1 of Sect. 10.15). For Lagrange’s top, as we have seen
in Sect. 4.2, F(γ3 ) is a cubic function and γ3 can be expressed in elliptic functions
of time. In the case of a multi-connected symmetric body in a liquid, we have
1
V = a3 γ3 + [b1 (1 − γ32 ) + b3 γ32 ],
2
= K 1 , l3 = K 3 γ3 + κ, (12.7)
so that
1
F(γ3 ) = (1 − γ32 )[2h + b1 − 2a3 γ3 + (b1 − b3 )γ32
A
1 1
− ( j − κ − K 3 γ3 )2 ] − 2 [ f − jγ3 − K 1 (1 − γ32 )]2
C A
is a polynomial of the fourth degree and hence γ3 is also an elliptic function of time.
Kirchhoff reduced the case of simply connected body (a3 = κ = 0) to an elliptic
quadrature and expressed some particular motions in terms of elliptic functions [219].
Detailed analysis of the general solution of the last special case in elliptic functions
was performed by Greenhill [135]. This solution was not extended to the case of
multi-connected body, but it can be noted that the presence of the constant gyrostatic
term κ and the parameter a3 changes the distribution of the roots of F and hence
affects the picture of motion. As far as we know, this case was not studied in detail.
Thus, of all known results remains without the present type of generalization only
one case, namely the case of a body in a liquid found by Rubanovsky [317] that
includes as special versions an earlier case due to Steklov [345] and the case of a
torque-free gyrostat considered by Joukovsky [163] and Volterra [366]. Due to the
situation that in this case the basic potential is zero, no more terms can be added by
the present method. However, as we have seen in Chap. 11, a generalization involving
an arbitrary function and a parameter has been applied to this case in the next section,
but to produce a conditional integrable case from it.
@seismicisolation
@seismicisolation
12.1.2 About the Hamiltonian Formulation
We have shown in Chap. 10 that, when ν(γ) = n = const, the transformed problems
are generalizations of the original ones, without bringing any mathematical compli-
cation to the solution process. The original and transformed problems are described
by one and the same set of Hamiltonian equations. This is a consequence of the fact
that the Hamiltonian after transformation is H = H + nM·γ, i.e. a combination of
the original Hamiltonian and the areas integral.
We shall show now that when ν(γ) is chosen according to one of the two proce-
dures described in Theorems 11.1–2, the Hamiltonian flow of the generalized problem
is really different from that of the original problem and hence the first problem is a
genuine generalization of the second.
In the tables of extended integrable cases, we have adopted the choice to identify
every case by the expressions of functions V and μ, which are unique and completely
characterize the physical setting for that case. An expression for the vector potential
l is also given in the tables, so that the Lagrangian for each case can be readily
constructed. The Hamiltonian of a problem can be obtained as described in previous
chapters. To this end one can write
∂L
H =ω· − L, (12.8)
∂ω
and eliminate ω using the momentum variables
∂L
M= . (12.9)
∂ω
The equations of motion take the form
∂H ∂H
Ṁ= M× +γ× ,
∂M ∂γ
∂H
γ̇=γ× , (12.10)
∂M
and their integrals are
I1 = H,
I2 = γ 2 = 1,
I3 = M · γ = f. (12.11)
In order to illustrate our point of view described above, we now give examples of
the extended cases in the Hamiltonian formalism.
(1) The first example is case 1 of Table 12.1, which involves one parameter n 1
of the new type in addition to the uniform precession parameter n. The Hamiltonian
@seismicisolation
@seismicisolation
for that case is
M12 M2 M2
H = 1/2 ( + 2 + 3 ) + b Aγ12 + Bγ22 + Cγ32
A B C
− (M1 γ1 + M2 γ2 + M3 γ3 ) [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]
= Hc − (M1 γ1 + M2 γ2 + M3 γ3 ) [n + n 1 (Aγ12 + Bγ22 + Cγ32 )], (12.12)
where Hc is the original Hamiltonian of the Clebsch case. The equations of motion
are
∂ Hc
Ṁ = M × { − [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]γ}
∂M
∂ Hc
+γ × { − [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]M − 2n 1 (M·γ)γI},
∂γ
∂ Hc ∂ Hc
= M× +γ×[ − 2n 1 (M·γ)γI], (12.13)
∂M ∂γ
and
∂ Hc
γ̇ = γ × { − [n + n 1 (Aγ12 + Bγ22 + Cγ32 )]γ}
∂M
∂ Hc
=γ× . (12.14)
∂M
The fourth integral is
1 2
I4 = (M + M22 + M32 )
2 1
−ABC[b − n 1 (M1 γ1 + M2 γ2 + M3 γ3 )](BCγ12 + C Aγ22 + ABγ32 )
= I4c + n 1 ABC (M1 γ1 + M2 γ2 + M3 γ3 ) (BCγ12 + C Aγ22 + ABγ32 ). (12.15)
From (12.13), we see that when f = 0 and n 1 = 0, then Hc and I4c are no more
integrals of motion. The Hamiltonian flow is deformed and the overall picture of the
trajectories in the phase space of the new problem is different from that of Clebsch’s
case.
On the other hand, when n 1 = 0 (ν = n)
H = Hc − n I3 ,
so that the Hamiltonian is a linear combination of the two integrals Hc and I3 with
constant coefficients. In that case, from (12.13), (12.14), we see that the flow defined
by the Hamiltonian (12.12) is identical with the flow corresponding to Clebsch’s
Hamiltonian Hc and the integral takes the form of Clebsch. The same holds also if
consideration is restricted to the level I3 = f = 0.
@seismicisolation
@seismicisolation
(2) The second example is the extension of Clebsch’s spherically symmetric

case. That is case 2 in Table 12.1. Let Hs and I4s be the original Clebsch spherical
Hamiltonian and the corresponding fourth integral. We have
Hs = Hs (M,γ, c1 , c2 , c3 )
1 1
= (M 2 + M22 + M32 ) + (c1 γ12 + c2 γ22 + c3 γ32 ),
2A 1 2
I4s = c1 M12 + c2 M22 + c3 M32 − (c2 c3 γ12 + c3 c1 γ22 + c1 c2 γ32 ). (12.16)
For the extended integrable system with three different parameters n 1 , n 2 , n 3 ,
H = Hs − (n + n 1 γ12 + n 2 γ22 + n 3 γ32 ) (M1 γ1 + M2 γ2 + M3 γ3 ) ,

= Hs (M,γ, c1 − 2n 1 I3 , c2 − 2n 2 I3 , c3 − 2n 3 I3 ). (12.17)
The equations of motion have the form
∂ Hs ∂ Hs
Ṁ = M× +γ×[ − 2(M·γ)(n 1 γ1 ,n 2 γ2 , n 3 γ3 )],
∂M ∂γ
∂ Hs
γ̇ = γ × , (12.18)
∂M
and the complementary integral is
I4 = (c1 − 2n 1 I3 ) M12 + (c2 − 2n 2 I3 ) M22 + (c3 − 2n 3 I3 ) M32

−A[(2n 2 I3 − c2 ) (2n 3 I3 − c3 ) γ12 + (2n 3 I3 − c3 ) (2n 1 I3 − c1 ) γ22
+ (2n 1 I3 − c1 ) (2n 2 I3 − c2 ) γ32 ] (12.19)
= I4s (M,γ, c1 − 2n 1 I3 , c2 − 2n 2 I3 , c3 − 2n 3 I3 ).
The deformation caused by the presence of the three parameters n 1 ,n 2 , n 3 and I3 in

the Hamiltonian flow on any level I3 = f = 0 is obvious. The solution of the new
problem of motion is obtained from that of the original case by replacing the origi-
nal physical parameters c1 , c2 , c3 by their new values involving three new physical
parameters and the dynamical parameter f.
It is remarkable that the fourth integral (12.19) is cubic in the momenta in the whole
phase space, due to the presence of I3 in the coefficients but becomes quadratic in
momenta on any fixed level of I3 . It also reduces to Clebsch’s quadratic integral when
n 1 =n 2 = n 3 = 0.
12.2 Conditional Integrable Deformations of General

Integrable Cases
By Theorem 1 in Chap. 11, Sect. 11.6, all the general and conditional integrable
cases of integrability of the previous hierarchy of problems admit a generalization
@seismicisolation
@seismicisolation
12.2 Conditional Integrable Deformations of General Integrable Cases 355
using the transformation (11.31) to conditional cases involving the arbitrary function
ν(γ1 , γ2 , γ3 ). The explicit solution of the equations of motion in each case can be
obtained from the solution of the original case, if the last is known, by Theorem 3. We
apply this procedure here to all the general integrable cases, which were the subject of
generalization in the preceding chapter, but this time including the case of Steklov–
Rubanovsky, which was not amenable to the generalization as an unconditional case.
As the structure of potential in the basic cases is not significant for the present type of
generalization, we consider only one branch of the Kowalevski–Sokolov hierarchy,
so that the total number of cases is still 7.
12.2.1 Table of Cases
In Table 12.2, we list the deformations of the known general integrable cases as
conditional cases on a fixed level of the areas integral. For each case we provide
– conditions, if any, on the inertia matrix of the body,
– the pair of scalar potential V and vector μ, figuring in the equations of motion,
– the vector l which enters in the Lagrangian or Hamiltonian of the problem and
in the structure of the cyclic integral,
– the cyclic integral I3 itself,
– and, finally, the complementary integral.
– Each of the systems in the following table is integrable on the level I3 = β. For
each case, we also add the forms of the Hamiltonian function and the complementary
integral in terms of momenta. In the Hamiltonian formulation, the cyclic integral has
the same form for all integrable cases
I3 = M1 γ1 + M2 γ2 + M3 γ3 = β.
Table 12.2 Conditional integrable extensions of general integrable cases, valid on the level I3 = β.
ν = ν(γ1 , γ2 , γ3 ) is an arbitrary function
1 Case of Clebsch’s type
V = (b − 21 ν 2 )(Aγ12 + Bγ22 + Cγ32 ) + βν,
l=νγI,
∂ ∂
μ = ∂γ (νγI · γ) − [ ∂γ · (νγI)]γ
I3 = ωI· γ+ ν(Aγ12 + Bγ22 + Cγ32 ) = β
I4 = 21 [A2 ( p + νγ1 )2 + B 2 (q + νγ2 )2 + C 2 (r + νγ3 )2
−b(BCγ12 + C Aγ22 + ABγ32 )
I3 = M1 γ1 + M2 γ2 + M3 γ3 ,
M2 M2 M2
H = 1/2 ( A1 + B2 + C3 ) + b Aγ12 + Bγ22 + Cγ32
+[β − (M1 γ1 + M2 γ2 + M3 γ3 )]ν
I4 = M12 + M22 + M32
−b(BCγ12 + C Aγ22 + ABγ32 )
@seismicisolation
@seismicisolation
In all the cases under consideration, one can show that
d I4
= (I3 − β)(ω, γ),
dt
where is a different function for each case. Thus, I4 becomes an integral under the
condition I3 = β. As examples we give explicit results for two cases from the table.
Remark: For the conditional integrable cases presented in Table 12.2, the Hamil-
tonian of the extended system is
H = H + ν(β − M · γ),
where H is the original Hamiltonian (before the transformation). The equations of

motion are
∂H ∂H ∂ν
Ṁ = M× +γ× + (β − M·γ)γ × ,
∂M ∂γ ∂γ
∂H
γ̇ = γ × . (12.20)
∂M
Although the equations of motion in all cases of Table 12.2 depend on the parame-
ter β and the (non-constant) function ν(γ), on the single level f = β the Hamiltonian
flow of the new conditional integrable problem becomes identical with the flow in
the original unconditional integrable problem. In the last problem, the parameter f
is arbitrary, while in the former the additional parameter β is present but regarding
the dynamical condition f = β both problems have the same number of parameters.
2 Case of Clebsch’s type of spherical symmetry

A = B = C,
V = βν + 21 (b1 γ12 + b2 γ22 + b3 γ32 ) − 21 Cν 2 ,
l = Cνγ,
∂ν ∂ν
μ = C[ ∂γ − (ν + γ · ∂γ )γ]
I3 = C( pγ1 + qγ2 + r γ3 + ν) = β
I4 = C[b1 ( p + νγ1 )2 + b2 (q + νγ2 )2 + b3 (r + νγ3 )2 ]
−(b2 b3 γ12 + b1 b3 γ22 + b1 b2 γ32 )
H = 2C1
(M12 + M22 + M32 ) + 21 (b1 γ12 + b2 γ22 + b3 γ32 )
+[β − (M1 γ1 + M2 γ2 + M3 γ3 )]ν
I4 = b1 M12 + b2 M22 + b3 M32 − C(b2 b3 γ12 + b1 b3 γ22 + b1 b2 γ32 )
3 Rubanovsky–Lyapunov type [317]

A=B=C
V = C{a1 γ1 + a2 γ2 + a3 γ3 − 21 (bcγ12 + caγ22 + abγ32 )
+ν{β + 21 [(b + c)γ12 + (c + a)γ22 + (a + b)γ32 ]} − 21 ν 2 }
l = C[νγ − 21 ((b + c)γ1 , (c + a)γ2 , (a + b)γ3 )],
∂ν ∂ν
μ = C[(aγ1 , bγ2 , cγ3 ) + ∂γ − (ν + γ · ∂γ )γ]
I3 = ( pγ1 + qγ2 + r γ3 + ν) − 2 [(b + c)γ12 + (c + a)γ22 + (a + b)γ32 ] = β
1
@seismicisolation
@seismicisolation
I4 = 21 [(b + c) ( p + νγ1 )2 + (c + a) (q + νγ2 )2 + (a + b) (r + νγ3 )2 ]

+a1 [ p + (ν + a)γ1 ] + a2 [q + (ν + b)γ2 ] + a3 [r + (ν + c)γ3 ]
−[bc( p + νγ1 )γ1 + ca(q + νγ2 )γ2 + ab(r + νγ3 )γ3 ]
H = 2C 1
(M12 + M22 + M32 ) + 21 [(b + c)M1 γ1 + (c + a)M2 γ2 + (a + b)M3 γ3 ]
+C(a1 γ1 + a2 γ2 + a3 γ3 )
− C8 [(a 2 + 2bc)γ12 + (b2 + 2ac)γ22 + (c2 + 2ab)γ32 ],
+[β − (M1 γ1 + M2 γ2 + M3 γ3 )]ν
I4 = (b + c)M12 + (c + a)M22 + (a + b)M32
+C{[(b2 + c2 )γ1 + 2a1 ]M1 + [(a 2 + c2 )γ2 + 2a2 ]M2 + [(a 2 + b2 )γ3
2
+2a3 ]M3 }+ C4 [(b + c)(b − c)2 γ12 + (c + a)(c − a)2 γ22 + (a + b)(a − b)2 γ32 ]
+C 2 [(a + b + c)(a1 γ1 + a2 γ2 + a3 γ3 ) + 2(a1 aγ1 + a2 bγ2 + a3 cγ3 )]
4 Case of Steklov–Rubanovsky’s type

V = ν(β − κ · γ + a[tr (I−1 )|γ|2 −γI−1 · γ]) − 21 ν 2 γI · γ,
l = κ−aγJ + νγI, J =[tr (I−1 )δ−I−1 ]
∂ ∂
μ = κ + 2aγI−1 + ∂γ (νγI · γ) − [ ∂γ · (νγI)]γ
I3 = [ωI+κ−aγJ + νγI] · γ = β
I4 = 21 | ωI+νγI+κ |2 +2a(ω · γ + ν)
H = 21 (M−κ − γJ)I−1 · (M−κ − γJ)
+[β − (M1 γ1 + M2 γ2 + M3 γ3 )]ν
I4 = 21 | M − aγJ |2 −aγI−1 · (M−κ − aγJ)
For this case, one can show that

d I4 ∂ν ∂ν
= (I3 − β)[(ωI+νγI+κ) · ( × γ) + aγI−1 ·( × γ)].
dt ∂γ ∂γ
(1) For any function ν, I4 is an integral on the level I3 = β and the dynamics is
conditionally integrable. On the other hand, when ν(γ) = n (a constant) the
terms in the square bracket vanish and this case becomes integrable for arbitrary
initial conditions and coincides with the Rubanovsky–Steklov case of motion of
a body in liquid.
5 Case of Kowalevski–Yehia–Sokolov type

A = B = 2C,
V = βν + C[κcγ1 + a2 γ2 − νκγ3
2
− c2 (γ12 + 2γ32 ) − νcγ1 γ3
− ν2 (2γ12 + 2γ22 + γ32 )]
2
l = C(2cγ3 + 2νγ1 , 2νγ2 , κ − cγ1 + νγ3 ),

∂
μ = C{(cγ1 , 0, κ + cγ3 ) + ∂γ [(2γ12 + 2γ22 + γ32 )ν]
∂ν ∂ν ∂ν
−[5ν + 2γ1 ∂γ 1
+ 2γ2 ∂γ 2
+ γ3 ∂γ3
]γ},
I3 = C[2 pγ1 + 2qγ2 + (r + κ)γ3 + cγ1 γ3 + (2γ12 + 2γ22 + γ32 )ν] = β
2
I4 = ( p + νγ1 )2 − (q + νγ2 )2 + a2 γ2 + c2 γ22 + cγ1 (r + νγ3 − κ)
2
+ 2( p + νγ1 )(q + νγ2 ) − a2 γ1 − c2 γ1 γ2 + cγ2 (r + νγ3 − κ)

+2κ(r + νγ3 − κ + cγ1 ) ( p + νγ1 ) + (q + νγ2 ) + 2c( p + νγ1 )γ3
2 2
−4a2 κ(q + νγ2 )γ3

−2κc2 {2γ3 [2( p + νγ1 )γ1 + cγ1 γ3 + 2(q + νγ2 )γ2 + (r + νγ3 )γ3 ]
+κγ32 − (γ12 + γ22 + 2γ32 )(r + νγ3 + cγ1 )}
@seismicisolation
@seismicisolation
I3 = M1 γ1 + M2 γ2 + M3 γ3 = β,
M 2 +M 2
H = 2C
1
( 1 2 2 + M32 ) − (κ + cγ1 )M3 + C(a2 γ2 + 2cκγ1 − c2 γ32 )
+ν[β − (M1 γ1 + M2 γ2 + M3 γ3 )]
M12 −M22
I4 = [ 4C 2
+ a2 γ2 + c( MC3 − 2κ)γ1 − c2 (γ12 − γ22 )]2
+[ 2C 2 − a2 γ1 + c( MC3 − 2κ)γ2 − 2c2 γ1 γ2 ]2
M1 M2
M 2 +M 2
+κ( MC3 − 2κ)[ 12C 2 2 + 2cγ3 MC3 ] − 2κγ3 (a2 + 2c2 γ2 ) MC2
2
− 2κc
C [2γ1 γ3 M1 − (γ1 + γ2 )M3 ]
2 2
6 Yehia [398],
Yehia–Bedwehy [419]: ν = 0,
Kowalevski [238]: ν = a = 0
A = B = 2C

V = βν + C[a1 γ1 + a2 γ2 + a
− 21 ν 2 2γ12 + 2γ22 + γ32 ]
1−γ32
l = Cν(2γ1 , 2γ2 , γ3 ),
∂ ∂ν
μ = C{ ∂γ [ν(2γ12 + 2γ22 + γ32 ] − [5ν + ∂γ · (2γ1 , 2γ2 , γ3 )]γ}
I3 = C[2 pγ1 + 2qγ2 + r γ3 + ν(2γ1 + 2γ22 + γ32 )] = β
2
I4 = [( p + νγ1 )2 − (q + νγ2 )2 − a1 γ1 + a2 γ2 ]2
+[2( p + νγ1 )(q + νγ2 ) − a1 γ2 − a2 γ1 ]2
a2
+2 a 2 [( p + νγ1 )2 + (q + νγ2 )2 ] + 1−γ 2
1−γ3 3
7 Case of Lagrange’s type

B = A,
V = V0 (γ3 ) + βν − ν[(γ12 + γ22 ) + l3 γ3 ] − 21 ν 2 [A + (C − A)γ32 ],
l = (( + Aν)γ1 , ( + Aν)γ2 , l3 + Cνγ3 ),
∂ ∂
μ = ∂γ (l · γ) − ( ∂γ · l)γ,
V0 (γ3 ), (γ3 ), l3 (γ3 ), ν(γ1 , γ2 , γ3 ) arbitrary functions
I3 = A( pγ1 + qγ2 ) + (Cr + l3 )γ3 + (γ12 + γ22 ) + ν[A + (C − A)γ32 ] = β
I4 = C(r + νγ3 ) + l3
I3 = M1 γ1 + M2 γ2 + M3 γ3 ,
M12 +M22 M32 l3
H = 21 ( A + C ) − [ A (M1 γ1 + M2 γ2 ) + C M3 ]
l32
+V0 (γ3 ) + 2 A (γ1 + γ2 ) + 2C + ν[β − (M1 γ1 +
1 2 2 2 M2 γ2 + M3 γ3 )]
I 4 = M3
Note that the precession angle ψ is cyclic and the corresponding generalized
momentum is the integral of motion I3 . The proper rotation angle ϕ is no longer
cyclic in general, but becomes cyclic on the level I3 = β. In fact, one may calculate
∂L ∂ν
= (I3 − β) .
∂ϕ ∂ϕ
The cyclic integral I4 is conditional on the level I3 = β. One can easily find
d I4 ∂ν ∂ν
= (I3 − β)(γ2 − γ1 ). (12.21)
dt ∂γ1 ∂γ2
@seismicisolation
@seismicisolation
This expression vanishes when either I3 = β or ν = ν(γ12 + γ22 ), which is equivalent

to ν = ν(γ3 ). Case 7 is integrable for arbitrary function ν(γ1 , γ2 , γ3 ) on the level
I3 = β, but becomes unconditional when ν = ν(γ3 ).
12.2.2 Example of Physical Application
Consider the simple original case of Kowalevski, obtained from case 5 of Table 12.2
by setting κ = c = 0. We shall transform this case using
ν = λγ3 , β = Cλa.
The generalized case will be characterized by
A = B = 2C

V = C[a1 γ1 + λaγ3 − λ2 γ32 2γ12 + 2γ22 + γ32 ]
2
l = Cλγ3 (2γ1 , 2γ2 , γ3 ),

μ = Cλ(−2γ1 γ3 , −2γ2 γ3 , 2 − 3γ32 )
I3 = C[2 pγ1 + 2qγ2 + r γ3 + γ3 (2γ12 + 2γ22 + γ32 )] = Ca
I4 = [( p + λγ1 γ3 )2 − (q + λγ2 γ3 )2 − a1 γ1 ]2
+[2( p + λγ1 γ3 )(q + λγ2 γ3 ) − a1 γ2 ]2
H = 4C1
(M12 + M12 + 2M12 ) + C(a1 γ1 + λaγ3 )
+λγ3 [Ca − (M1 γ1 + M2 γ2 + M3 γ3 )],
I4 = [M12 − M22 − 4C 2 a1 γ1 ]2
+4[M1 M2 − 2C 2 a1 γ2 ]2
The potential is modified by the addition of two terms. The first Cλaγ3 means
a displacement of the centre of mass of the body in the z-direction (normal to the
equatorial plane) to the point (x0 = Ca1 /Mg, 0, z 0 = Cλaγ
Mg
3
) and the second term is
quartic in γ. The last may be written as
Cλ2 2 2 Cλ2 2
− γ3 2γ1 + 2γ22 + γ32 = − γ3 2 − γ32 .
2 2
The new problem involves also the gyroscopic moments described by the vector l
and μ and it is integrable on the level I3 = Ca. On setting λ = 0, one recovers the
general integrable case of Kowalevski.
Note that the parameter λ does not appear in the Hamiltonian form of I4 . Note
also that from the Hamiltonian function of the new problem one cannot recognize
the acting potential and gyroscopic forces, which are clearly defined by the potential
V and the vector μ.
@seismicisolation
@seismicisolation
Chapter 13
Miscellaneous Cases Integrable on a
Single Level of the Areas Integral
In this chapter, we collect certain sets of conditional integrable cases of different ori-
gins and of various characters, which do not belong to one of the problems discussed
in previous chapters that have definite physical interpretation, but they are unified
only by being valid on a single level of the areas integral, and mostly on f = 0.
The first set consists of the separable cases investigated in Chap. 9 above. Another
set began to appear in a work of Goryachev [117], who used an inverse method to
find potentials that admit existence of a complementary integral of the third or fourth
degrees in the components of the angular velocity, as modifications of the known
cases at that time. Those are Kowalevski’s case of a heavy body with a fixed point
and Chaplygin’s case of a rigid body in a liquid and Goryachev–Chaplygin’s case of
the classical problem. The search led in [117] to the new cases:
(1) A conditional case, on the level f = 0, under the condition A = B = 2C, with
the potential
λ
V = a1 γ1 + a2 γ2 + b1 (γ12 − γ22 ) + 2b2 γ1 γ2 + .
2γ32
This adds the singular term 2γλ2 to a former result of Chaplygin [53], which
3
combines the potentials of Kowalevski’s classical case and Chaplygin’s case of
a body in a liquid. The complementary integral for this case is of the fourth degree.
(2) Another case under the condition A = B = 43 C, with the potential
a + bγ1
V = 2 ,
γ33
which admits a cubic integral.

https://doi.org/10.1007/978-3-030-96336-1_13
@seismicisolation
@seismicisolation
362 13 Miscellaneous Cases Integrable on a Single Level of the Areas Integral
A variety of modifications and generalizations of known conditional integrable

cases have accumulated in the last three decades, mainly in the works of the
author and coworkers, as a result of the introduction of a new method of con-
struction of two-dimensional integrable systems living on Riemannian manifolds
and whose integrals are polynomials in velocities [381]. The resulting systems
usually involved a large number of parameters, which could be adjusted to iden-
tify the metric of the problem with that of the reduced system of rigid body
dynamics after ignoring the precession angle as in Chap. 9. Here we shall not
try to make any physical interpretation of the potential and gyroscopic terms
in each case, even though they are mostly generalizations of some of the cases
presented in the previous chapters in natural physical settings.
The most natural and comfortable classification of the relatively large number
(22) of known conditional integrable cases is the classification by the degree of the
polynomial complementary integral in every case. We shall follow this classification
here. We also give full-time sequence of each hierarchy of overlapping cases.
Although most of those cases do not have physical interpretation, due to strange
singularities in the potentials, we give full up-to-date list of them. As known cases
are scattered in the literature, we believe this step, made here for the first time, can
play a definite role in future development of the subject. As will be seen below, some
of those cases have already stimulated further studies on the separation of variables
and also on topological classification.
Remark 18 We have used the uniform and variable precession transformations in

some previous chapters. In the present one, those transformations will not be applied.
The reason is that in conditional cases on the level f = 0 this transformation may
be easily applied using an arbitrary function ν(γ) as discussed in Sect. 12.2 of the
preceding chapter.
13.1 Cases with a Quadratic Integral
It is well-known that a natural (time-reversible) mechanical system of two degrees

of freedom which admits an integral quadratic in velocities and independent of the
energy integral must be Liouville separable system in some generalized coordinates.
The dynamics of a rigid body acted upon by pure potential forces is time-reversible on
the zero-level of the cyclic integral. When gyroscopic forces are present, equations of
motion become irreversible and Liouville separability is lost. In the present section,
we list the three known types of potentials that admit a complementary quadratic
integral and hence admit Liouville separation and also three non-separable cases
with a quadratic integral.
@seismicisolation
@seismicisolation
13.1 Cases with a Quadratic Integral 363
13.1.1 Separable Integrable Potentials
From the Minkowski analogy between the motion of a rigid body about a fixed point
and the motion of a material point on the inertia ellipsoid of the body at the fixed
point, it follows that certain potentials exist, which allow separation of variables on
the zero level of the areas integral. Those are of three types:
(1) Potentials separable in elliptic coordinates on the tri-axial ellipsoid ( A = B =
C) Chap. 9 Sect. 9.7.2.
(2) Potentials separable in spherical coordinates for a dynamically symmetric body
(A = B), including the case of dynamical spherical symmetry at the fixed point
Chap. 9 Sect. 9.7.1.
(3) Potentials separable in sphero-conic (elliptic) coordinates on the sphere of inertia
in the case of complete dynamical symmetry ( A = B = C) Chap. 9 Sect. 9.7.3.
For all three types of separable cases μ = 0, f = 0 and u, v, F, G, g are arbitrary
functions of their arguments (Table 13.1).
Table 13.1 Conditional integrable cases. The first 20 cases are valid on the level f = 0
1 Separable in elliptic coordinates on the ellipsoid of inertia [31, 224, 381]
Aγ12 +Bγ22 +Cγ32 √ √
V = √
β
[u(α + β) + v(α − β)],
γ2 γ2 γ2
α = AB + BC + C A − ABC( A1 + B2 + C3 ) = ABC[tr (I−1 ) − γI−1 ·γ],
β = α2 − 4 ABC D, D = Aγ12 + Bγ22 + Cγ32 ≡ γI·γ.
√ √ √ √
I4 = A p + B q + C r 2 + √1β [(α − β)v(α − β) + (α + β)u(α + β)]
2 2 2 2 2
2 Separable in spherical coordinates on the Poisson sphere B = A

γ
A−(A−C)γ32 A−(A−C)γ32 g( γ1 )
V = F(γ3 ) + A(1−γ32 )
G( γγ21 ) ≡ F(γ3 ) + Aγ22
g( γγ21 ), G( γγ21 ) = γ
2
1+( γ1 )2
2
I4 = Cr 2 + 2G( γγ21 )
3 Separable in sphero-conic coordinates on the Poisson sphere [390, 391]
A = B = C,
√ √
V = [u(α − β √ )+v(α + β )]

,
β
γ2 γ2 γ32
α = a + b + c − (aγ12 + bγ22 + cγ32 ), β = α2 − 4abc( a1 + b2 + c ).
I4 = (ap 2 + bq 2 + cr 2 )

+ √1 [(α + β )u(α − β ) + (α − β )v(α + β )]
β
A special case of this separable potential, equivalent to F(γ3 ) = 0, was pointed

out also by Kolossov in [224], but the general potential seems to be unnoticed in the
literature. This type of potential appears as a part of the potential in some integrable
generalizations of Kowalevski’s case, which admit an integral quartic in velocities.
The quartic integral in those cases can be written as the square of the quadratic
integral in the separable
@seismicisolation
@seismicisolation
13.1.2 Non-separable Cases with a Quadratic Integral

4 Yehia [414]
Separable (K = 0),
Subcase of Steklov’s (s = 0)
V = √ sS ,
2 ABC S 2 −4 ABC D
S = A(B + C)γ12 + B(C + A)γ22 + C(A + B)γ32
μ = −K ( γA1 , γB2 , γC3 ),
l = 2 ABC
K
(A(B + C)γ1 , B(C + A)γ2 , C(A + B)γ3 )
I4 = 2 (A p
1 2 2 + B 2 q 2 + C 2 r 2 ) − K ( pγ1 + qγ2 + r γ3 ) + √ s
S 2 −4 ABC D
5 Yehia [394]
A(B+C)γ12 +B(C+A)γ22 +C(A+B)γ32
V =n Aγ12 +Bγ22 +Cγ32
∂F
μ= ∂γ − γ
A(B+C)γ12 +B(C+A)γ22 +C(A+B)γ32
F= J
ABC(Aγ12 +Bγ22 +Cγ32 )
2(A+B+C)v0 −3v02 + 2 2 ABC
Aγ1 +Bγ22 +Cγ32
= J
ABC(Aγ12 +Bγ22 +Cγ32 )
I3 = Apγ1 + Bqγ2 + Cr γ3 + F
I4 constructed in elliptic coordinates and not in Euler-Poisson
variables. See [394]
6 Yehia [401]
Separable (N = 0)
Subcase of Lyapunov’s (K = 0)
A=B=C
2 abc γ12 γ22 γ32
V = C[− N 2 ( a + b + c )+
√K ]
β
l = − 21 C((b + c)γ1 , (c + a)γ2 , (a + b)γ3 )
μ = C(aγ1 , bγ2 , cγ3 )
I4 = 2 [(b
1
+ c) p 2 + (c + a) q 2 + (a + b) r 2 ]
a+b+c+aγ12 +bγ22 +cγ32
−N abc( p γa1 + q γb2 + r γc3 ) + 21 K √
β
13.2 Cases with a Cubic Integral
The list of such cases comprises only three items:
7 Yehia and Elmandouh 2016 [424] (c1 , c2 added)

Yehia 2002 [409] c1 = c2 =0
Sokolov and Tsiganov 2002 [337] e0 = e1 =λ=0
Yehia [395] 1996 (Independently of [226]) e0 = e1 = c1 = c2 = 0
Komarov and Kuznetsov [226] 1987 e0 = e1 = c1 = c2 = a 2 = 0
Sretensky [341] 1963 e0 = e1 = c1 = c2 = λ = 0
Goryachev [118] 1916 e0 = e1 = c1 = c2 = k = 0
Goryachev-Chaplygin [115] 1900, [52] 1901 e0 = e1 = c1 = c2 = k = λ==0
@seismicisolation
@seismicisolation
13.2 Cases with a Cubic Integral 365
A = B= 4C,
l = C 0, 0, k + c1 γ1 + c2 γ2 + e0 ( γ24 − 1
γ32
) + γ 2e+γ
1
2 ,
3 1 2
γ1 (4−γ32 )
2e1 γ1 γ2
μ1 = C[c1 γ3 + 2e0 (γ12 +γ22 )2
−
],
γ35
γ2 (4−γ32 )
2e1 γ1 γ2
μ2 = C[c2 γ3 + 2e0 − (γ 2 +γ 2 )2 ],
1 γ35 2
γ (2−γ 2 )
μ3 = C[k + c1 γ1 + c2 γ2 + e0 2 γ 4 3 + γ 2e+γ 1
2 ],
3 1 2
V = C {a1 γ1 + a2 γ2 − c1 c2 γ1 γ2
c2 c2
+ 22 γ12 + 21 γ22 + γλ2
3
4−8γ32 +5γ34
+e0 γ32 − 2 c1 γ1γ+c4
2 γ2
− e02 2γ 8
3 3
4γ12 +4γ22 +1
+e1 k+e0 −γ 1 c1 −γ2 c2
γ 2 +γ 2
− 21 e12 2 2 2 }
1 2 γ1 +γ2
I3 = 4 pγ1 + 4qγ2
+[r + k + c1 γ1 + c2 γ2 + γ 2e+γ
1
2 + e0 ( γ 4 − γ 2 )]γ3 = 0,
2 1
1 2 3 3

e0 (2−γ32 ) e1 (8γ12 −1)
I4 = r − k + c1 γ1 + c2 γ2 + γ 4 − (γ 2 +γ 2 ) [ p + c21 γ3 ]2
3 1 2
c2 γ3 2 λ e0 (2−γ32 )
+[q + 2 ] +
2γ32
+ k( γe04 −
e1
2 )−( 2 +
e1
2γ34
)r
3
e0 (2−γ32 ) e02 (3γ34 −6γ32 +4)
+ (c1 γ1 + c2 γ2 ) [ e21 + 2γ 4 ] − 2γ38
3
e1 e1 (1−8γ12 ) 8γ12 (γ32 −2)−2γ34 +9γ32 −8
+ γ 2 +γ 2 [ 2 + 2γ34
]
1 2
c1 γ3
−γ3 [(2e1 c1 − c1 k + a1 )( p + 2 )
c2
+(2e
1 c2 − c2 k + a1 )(q + 2 γ3 )]
e0 e1 (1−2γ32 ) 4e γ 2 (e −2e γ 2 )
+k (c1 γ1 + c2 γ2 )[ γ 2 + γ 2 +γ 2 ] + 1γ 21 (γ02 +γ 21) 3
3 1 2 3 1 2
1 γ3
− γ4e2 +γ 2 ( pγ1 + qγ2 )
1 2
8e0 e1 c2 γ2 γ12 2 2e0
2
λ
− γ32 (γ12 +γ22 )
+ (a1 γ1 + a2 γ2 )[ γe02 − γ 2e+γ 1
2 ] + 4e1 γ1 [ γ 6 + γ 2 (γ 2 +γ 2 ) ]
3 1 2 3 3 1 2
8e c1 γ1 3
e0 4e0 e1 γ1
2 2
− γ 21+γ 2 [e1 + γ 2 ] + γ 4 (γ 2 +γ 2 )2 [8γ1 (γ3 − 2) − (γ1 + γ2 )(γ3 − 4)]
2 2 2 2 2
1 2 3 3 1 2
e1 (c12 +c22 ) 8e13 γ13 (1−4γ12 )
+ 9(γ 2 +γ 2 ) [18γ1 γ3 − 9γ3 + 13γ3 − 4] +
2 2 4 2
(γ12 +γ22 )2
1 2
1 γ3
+ γ2e2 +γ 2 {q[c 2 (5γ 2 − γ 2 ) − 2c γ γ ]
1 2 1 1 2
1 2
+ p[c1 (3γ12 + γ22 ) − 2c2 γ1 γ2 ]}
1 γ1 γ2
8e (γ12 −γ22 ) 2
− 16e
γ12 +γ22
pq − e12c2 γ1 + 1γ 2 +γ 2 q
1 2
The second case with a cubic complementary integral is due to Goryachev [117].
It is characterized by the following
8 Goryachev 1915 [117].
A = B = 43 C,
l = μ = 0,
V = aγ1 +bγ
2
2 +c
,
γ33
I3 = 43 ( pγ1 + qγ2 ) + r γ3 = 0,
1
I4 = 2r ( p 2 + q 2 ) + r 3 − 2aγ33 p + r a+bγ
2
1
.
γ33
@seismicisolation
@seismicisolation
Although having no obvious physical meaning, this case has received a growing
interest in the last years [45, 361]. It turns out to be the first example of a mechanical
system whose complex invariant varieties are strata of Jacobians of a non-hyper-
elliptic curve, here a trigonal curve of genus 3 [45].
Goryachev’s case 8 has been generalized to the following one involving two
(vector and scalar) potentials (Yehia 2002 [409]):
2
e1 (2+γ32 )
8* l = (0, 0, κ + 1
γ12 +γ22
[3κ + e0 γ33 + 2 ]),
γ33
aγ1 +bγ2 +c
V = 2
γ33
e (4−7γ3 )
2 2 e12 (13γ34 −8γ32 +4)
+ 1
2 [ 0 2 − 4
γ12 +γ22 6γ 3 2γ33
3
3κ2 γ32
−e0 e1 (5γ32 − 2) + 2 (γ32 − 4)
8 4
−3e0 κγ3 − 3κe1 γ3 (γ32 + 2)].
3 3
where κ, e0 and e1 are arbitrary constants. Note that the constant gyrostatic momen-
tum κ is a coupling constant for some potential and gyroscopic terms.
For this generalization, one can easily write
4 1 2 e1 (2 + γ32 )
I3 = ( pγ1 + qγ2 ) + {r + κ + 2 [3κ + e0 γ33 + ]}γ3 = 0,
γ1 + γ2
2 2
3 γ3 3
but the complementary cubic integral is not yet expressed in the Euler–Poisson vari-
ables (See [409]).
Cases 7 and 8 were constructed as special cases of an integrable many-parameter
system with a cubic integral under some restrictions on those parameters.
9 Yehia 2000 [404].

Gaffet 1998 [96, 97] (The equivalent problem for a particle on a sphere).
A = B = C,
V = K
2 .l = 0.
(γ1 γ2 γ3 ) 3
I3 = pγ1 + qγ2 + r γ3 = 0,
1
I4 = Apqr − 2K (γ1 γ2 γ3 ) 3 ( γp1 + q
γ2 + r
γ3 ).
13.3 Cases with a Quartic Integral
The thirteen presently known cases with a quartic integral are characterized by the
Kowalevski configuration A = B = 2C. They are mostly (but not all) generalizations
on the level f = 0 of the two classical cases: Kowalevski’s case of a heavy body and
Chaplygin’s case of a body in a liquid. It is curious to note that the main potential
in most of those cases is composed of the basic potential present in Kowalevski’s or
Chaplygin’s cases or both of them and some additional terms that belong to separable
@seismicisolation
@seismicisolation
13.3 Cases with a Quartic Integral 367
potentials discussed in Chap. 9 Sect. 9.7.1. When alone, the last terms give rise to a
quadratic integral, instead of the quartic one.
Those cases are divided into five types, presented in the following subsections.
13.3.1 Cases Stemming from Kowalevski’s Case
Four cases of this type are listed here:
10 Yehia 1996 [396],

λ = 0 Yehia-Bedwehy 1987, (unconditional case) [419] (unconditional case),
λ = ε = 0 Kowalevski 1888 [238] (unconditional case).
A = B = 2C,
V = C[a1 γ1 + a2 γ2 + 2ε 2 + γλ2 ],
γ1 +γ3 3
l = μ = 0.
I3 = pγ1 + qγ2 + r γ3 = 0,
λ(γ12 −γ22 ) 2
I4 = [ p 2 − q 2 − a1 γ1 + a2 γ2 − 2γ32
]
λγ1 γ2 2
+[2 pq − a1 γ2 − a2 γ1 − γ32
]

2λ γ12 +γ32
+ε[ ( p+nγ1 ) 2+(q+nγ2)
2 2
ε
+ γ12 +γ32
+ γ32
]
γ1 +γ3 2
11 Yehia 2006 [413]

ν1 = δ2 = 0: Yehia-Bedwehy 1987 [419]
A = B = 2C.
2−γ32
V = C[a1 γ1 + λ
+ ε + (δ2 + ν1 γ1 )].
γ32 1−γ32 γ22 1−γ32
l = μ = 0.
I3 = pγ1 + qγ2 + r γ3 = 0,
λ(γ12 −γ22 ) 2 2λγ1 γ2 2
I4 = [ p 2 − q 2 − a1 γ1 − γ32
] + [2 pq − a1 γ2 − γ32
]
γ2 γ2
ε+ν1 γ1 32 ε+ν1 γ1 32
γ2 γ γ2 γ
+[δ2 γ32 + 2 2 ][2 p 2 + 2q 2 + δ2 γ32 + 2 2 ]
2 1−γ3 2 1−γ3

2λε 1−γ32

+ γ32
− γ22 (a1 γ32 + λγ1 )(δ2 γ1 + ν1 1 − γ32 )
2
The main result of [413] was the construction of an integrable system of two
degrees of freedom living on a Riemannian (or pseudo-Riemannian1 ) manifold and
admitting an integral of degree four in velocities. Cases 11, 14 were obtained as
special cases under suitable restrictions of this twenty-one-parameter system that
render the metric to that of the Routhian of the rigid body dynamics. Case 18 below
was obtained by further development of the method of [413].
Separation variables and expressions of the dynamical variables in terms of them
are constructed for case 11 in [218] (See also [137]), without treating the inversion
of the resulting quadratures.
1 In differential geometry, that is a manifold whose metric is not necessarily positive-definite
@seismicisolation
@seismicisolation
12 Yehia, Elmandouh 2011 [422]

c = 0. Yehia-Bedwehy (unconditional case) [419]
λ = 0. Sokolov (unconditional case)
A = B = 2C,
V = C[aγ1 + c2 (γ12 − γ32 ) + 2ε 2 + γλ2 ],
2
γ1 +γ2 3
μ = Cc(0, γ3 , γ2 ),
λ(γ12 −γ22 ) 2
I4 = ( p 2 − q 2 − aγ1 − cr γ2 − c2 γ12 − γ32
)
2λγ1 γ2 2
+(2 pq − aγ2 + cr γ1 − c γ1 γ2 − γ 2 )
2
3
+ε[ 2(p 2−q 2) + γ 2 +γ

2 2
ε 2 + γ 2 (c2 + λ )]
2 + 2 γ1 2 γ32
γ +γ
1 2 1 2
13 Yehia, Elmandouh 2011 [422],

λ = 0: the integral I4 becomes unconditional and gives Sokolov’s [336].
A = B = 2C,
V = C[κcγ1 + a2 γ2 + c2 (γ22 − γ32 ) + γλ2 ],
2
3
l = C (0, 0, k + cγ1 ) ,
μ = C(cγ3 , 0, κ + cγ1 ),
2
λ(γ 2 −γ 2 )
I4 = p 2 − q 2 + a2 γ2 + c2 γ22 − cγ1 (κ − r ) − 1γ 2 2
3
2
2λγ1 γ2
+ 2 pq − a2 γ1 − c γ1 γ2 − cγ2 (κ − r ) − γ 2
2
3
+2κ(r − κ + cγ1 )[ p 2 + q 2 + 2cpγ3 + λ(1 + 1
γ32
) − γ3 2qa2 + c2 κγ3 ]
−2κc2 [4γ3 ( pγ1 + qγ2 ) − (1 − γ32 )(r + cγ1 )].
13.3.2 Cases Stemming from Chaplygin’s Case
Four cases of this type are listed in the following table:
14 Yehia 2006 [413]

δ1 = δ2 = 0, Yehia 2003 [411]
δ1 = δ2 = ρ = 0, Goryachev [118]
δ1 = δ2 = ρ = λ = 0, Chaplygin [53]
A = B = 2C,
V = C[b1 (γ12 − γ22 ) + γλ2 + ρ( γ14 − γ16 )
3 3 3
+(2 − γ32 )( γδ12 + δ2
γ22
)],
1
l = μ = 0,
I3 = 2 pγ1 + 2qγ2 + r γ3 = 0,
λ(γ12 −γ22 ) 2
I4 = [ p 2 − q 2 + b1 γ32 − γ32
] + 4[ pq − λγγ12γ2 ]2
3
(γ 2 +γ 2 )2
+2( p 2 + q 2 ){ρ[ γ14 − γ16 ] + γ32 [ γδ12 + γδ22 ]} + ρ 1 γ 122 (ρ − 2λγ34 )
3 3 1 2 3
2ρb (γ 2 −γ 2 )
+ 1 γ 41 2 − 2b1 γ34 [ γδ12 − γδ22 ] + γ34 [ γδ12 + γδ22 ]2
3 1 2 1 2
1−γ 2
−2(ρ + λγ34 ) γ 4 3 [ γδ12 + γδ22 ].
3 1 2
@seismicisolation
@seismicisolation
Together with case 11, this case was constructed in [413] (See comment next to
case 11). Separation variables and quadratures were constructed for this case (and
for case 11) in [137].
15 Yehia-Elmandouh 2013 [423]

K = 0 Yehia [411] (Sect. 4.2.3) 2003
K = ρ = 0 Goryachev [118] 1916
K = ρ = λ = 0 Chaplygin [53] 1903
A = B = 2C
V = C{k 2dγ1 γ2 + c(γ12 − γ22 )
2
+K 2 [2cdγ1 γ2 (γ12 − γ22 ) + d2 (γ34 + 4γ12 γ22 )
−c (γ3 (γ1 + γ2 ) + 2γ1 γ2 )] + γλ2 + ρ( γ14 −
2 2 2 2 2 2 1
γ36
)},
3 3
l = C(0, 0, K [d(γ22 − γ12 ) + 2cγ1 γ2 ]),
μ = C(2K γ3 (cγ2 − dγ1 ), 2K γ3 (dγ2 + cγ1 ),
K [d(γ22 − γ12 ) + 2cγ1 γ2 ]).

I3 = 2 pγ1 + 2qγ2 + {r + K d(γ22 − γ12 ) + 2cγ1 γ2 }γ3 ,
I4 = { p − q + ckγ3 + γ3 [K dr + cK (c(γ1 − γ2 ) + 2dγ1 γ2 )]
2 2 2 2 2 2 2
λ(γ12 −γ22 ) 2
− γ32
}
+{2 pq + dkγ32 + [d K 2 (c(γ12 − γ22 ) + 2dγ1 γ2 ) − K cr ]γ32
− 2λγγ 21 γ2 }2
3
2(γ32 −1) 2
+ρ{ γ36
[ p + q2]
− 2K
γ4
r
[2cγ1 γ2 + d(γ22 − γ12 )]
3
(1−γ32 )2
+ γ 12 (ρ − 2λγ34 )
3
(d 2 −c2 )γ12 γ22 +cdγ1 γ2 (γ12 −γ22 )
+K 2 [2c2 ( γ14 − γ22 ) + 8 γ34
]
3 3
+ γ2k4 [c(γ12 − γ22 ) + 2dγ1 γ2 ]}.
3
16 Yehia and Elmandouh 2016 [425]

κ = 0 : Special case of Yehia and Elmandouh [423]
A = B = 2C
V = C{κ[c(γ12 − γ22 ) + 2dγ1 γ2 ] + κK [2cγ1 γ2 − d(γ12 − γ22 )]
+K 2 {2cdγ1 γ2 (γ12 − γ22 ) − c2 [γ32 (γ12 + γ22 ) + 2γ12 γ22 ]
+ d2 (γ34 + 4γ12 γ22 )} + γλ2 },
2
l = C(0, 0, κ + K [2cγ1 γ2 − d(γ12 − γ22 )]),

μ = C(2K γ3 (cγ2 − dγ1 ), 2K γ3 (cγ1 + dγ2 ), κ + K [2cγ1 γ2 + d(γ22 − γ12 )]),
I3 = 2 pγ1 + 2qγ2 + (r + κ + K [2cγ1 γ2 + d(γ22 − γ12 )])γ3 ,

I4 = p 2 − p 2 + ckγ32 + cK 2 γ32 [c(γ12 − γ22 ) + 2dγ1 γ2 ]

2
λ(γ 2 −γ 2 )
+d K [2κ − γ32 (3κ − r )] − 1γ 2 2

3
+ 2 pq + kdγ32 + d K 2 γ32 [2dγ1 γ2 + c(γ12 − γ22 )]
2
+cK [γ32 (3κ − r ) − 2κ] − 2λγγ 21 γ2
3
+2κ[r − κ − K (2cγ1 γ2 + d(γ22 − γ12 ))] p 2 + q 2 + λ(1 + 1
γ32
)

+γ32 [K 2 (c2 + d 2 )(γ32 − 1) − 2dκK + κc]
@seismicisolation
@seismicisolation
−4κγ3 {[2K κ(cγ1 + 2dγ2 ) − k(2cγ2 − dγ1 )](q + nγ2 )

+γ

2 ( p + nγ1 )(2cκK + dk)}
−8κ c2 K γ32 [κK (γ32 − 1) − kγ1 γ2 ]
+c{2κd K 2 γ1 γ2 γ32 + K [ 21 kdγ34 − 2λγ1 γ2
+dkγ32 (γ12 − 21 )] − 21 kκγ32 }
−2κd 2 K 2 γ12 γ32 + K [2λdγ12 + d(κ2 + λ)γ32 ]
17 Yehia 2003 [411]

Goryachev 1916 [118], ρ = 0.
Chaplygin 1903 [53] λ = ρ = 0.
A = B = 2C,
V = C[c(γ12 − γ22 ) + 2dγ1 γ2 + γλ2 + ρ( γ14 − 1
γ36
)],
3 3
l = μ = 0,
I3 = 2 pγ1 + 2qγ2 + r γ3 = 0,
2
λ γ 2 −γ 2 2
I4 = [ p 2 − q 2 + cγ32 − 1
γ32
2
] + [2 pq + dγ32 − 2 λ γγ12γ2 ]
2 2 3 2
γ −γ 1−γ 2
+2 ρ [( γ14 − γ16 ) p 2 + q 2 + c 1γ 4 2 + 2 d γ2γγ4 1 − λ γ 83 ]
3
2
3 3 3 3
1−γ32
+ρ2 γ3
12 .
Elmandouh (2015) [73] introduced a two-parameter generalization of this case

by adding singular terms into the vector and scalar potentials:
(γ 2 −2)γ 2
17∗ V = C[c(γ12 − γ22 ) + 2dγ1 γ2 + λ
γ32
+ ρ( γ14 − 1
γ36
) + 32γ 2 3 ( γν11 + ν2 2
γ2 ) ],
3 1
2−γ32
l = C(0, 0, γ1 ( γν11 + ν2
γ2 )).
The fourth integral is also given in the Euler–Poisson variables in [73].
13.3.3 Cases Combining Kowalevski’s and Chaplygin’s Cases
Two cases are listed in the following table:
18 Yehia 2012 [415],

δ = 0 Goryachev 1917 [118],
δ = λ = a1 = 0 Chaplygin 1903 [53],
δ = λ = a2 = 0 Kowalevski 1888 [238],
A = B = 2C,
λ 2−γ32
V = 2C[a1 γ1 + a2 (γ12 − γ22 ) + γ32
+δ γ22
],
l = μ = 0,
I3 = 2 pγ1 + 2qγ2 + r γ3 = 0,
λ(γ12 −γ22 ) 2
I4 = [ p 2 − q 2 − a1 γ1 + a2 γ32 − γ32
)
2λγ1 γ2 2
+(2 pq − a1 γ2 − γ32
)
δγ34
+ γδ2 {2[( p 2 + q 2 ]γ32 − 2a1 γ1 γ32 − 2λγ12 + 2a2 + γ22
}.
2
@seismicisolation
@seismicisolation
The Goryachev subcase (δ = 0) of case 18 has led to complex separation variables

in [337, 359].
19 Yehia 1996 [396],

λ = 0. Yehia 1987 [386],
k = 0. Goryachev [118],
k = a2 = λ = 0 Chaplygin 1903 [53],
k = b1 = b2 = λ = 0 Kowalevski 1888 [238],
λ
V = C[a1 γ1 + a2 γ2 + b1 (γ12 − γ22 ) + b2 γ1 γ2 + 2γ32
],
l = μ = C(0, 0, k).
I3 = 2 pγ1 + 2qγ2 + (r + k)γ3 = 0,
λ(γ12 −γ22 ) 2
I4 = [ p 2 − q 2 − a1 γ1 + a2 γ2 + b1 γ32 − 2γ32
]
λγ1 γ2 2
+[2 pq − a1 γ2 − a2 γ1 + b2 γ32 − γ32
]
+k{(r − k)[2( p 2 + q 2 ) + λ(1 + 1
γ32
)]
−4γ3 [(a1 + b1 γ1 + b2 γ2 ) p + q(a2 + b2 γ1 − b1 γ2 )]}.
Elmandouh (2015) [74], added a parameter e, which engenders singular terms in

the vector and scalar potentials, as follows:
λ
19∗ V = C{a1 γ1 + a2 γ2 + b1 (γ12 − γ22 ) + b2 γ1 γ2 + 2γ32
eγ 2 e(2γ22 +γ32 )
− γ 23 [k − 2γ12
]},
1
e(1+γ22 )
l =C 0, 0, k + γ 2 ,
1
−2eγ3 e(1+γ22 )
μ=C γ13
(1 + γ22 ), 2eγγ22 γ3 , k + γ12
,
1
The complementary integral was also provided in [74].

Separation of variables for the Chaplygin level of the above hierarchies was
attained by Chaplygin himself. For detailed solution see Chap. 10 Sect. 10.16. The
version κ = 0 of this case was considered in Tsiganov’s work [358], where an
assertion is made that separated variables are constructed for what the author calls
“the Kowalevski–Goryachev–Chaplygin gyrostat”. However, the proposed separated
variables are complex functions of the physical variables. It remains an open problem
how to construct real solutions using complex hyper-elliptic quadratures [358]. Note
that the title and references in that work have brought certain confusion, which was
commented in our note [410].
In [416], it was shown that the problem of motion of a rigid body, with A = 2C
and arbitrary B, subject to forces with potential containing one Kowalevski term,
one Chaplygin term together with the singular Goryachev term
c1
V = a1 γ1 + b1 (γ12 − γ22 ) + , (13.1)
γ32
under the additional restrictions q = 0, f = 0, is solvable in elliptic functions of

time. The solution is the same in case 17, when A = B = 2C, n = κ = a2 = b2 = 0
@seismicisolation
@seismicisolation
under the additional restriction q = 0. Without this restriction, the solution corre-
sponding to the last potential (13.1) is not known at this moment.
The subcase with the potential
c1
V = b1 (γ12 − γ22 ) + , (13.2)
γ32
common between hierarchies #14-19, has attracted more attention. It is called by

some authors the “Goryachev system”. Ryabov found real separation variables for
this case in [323], reduced its integration to hyper-elliptic quadratures and studied the
phase topology for positive values of the parameters, i.e. when the integral surfaces
are compact. The case of negative values of the parameters, when the integral surfaces
become non-compact, is treated by Nikolaenko [297], who has also shown that
Goryachev’s system is Liouville equivalent to other integrable cases in rigid body
dynamics, according to the value of the energy parameter on the admissible energy
interval [h min , ∞) [296].
13.3.4 A Case with a Quartic Integral Outside the Above

Classification
20 Yehia 2003 [411], f = 0,

A = B = 2C,
γ3 cγ1 +dγ2 + (c2 +d 2 )(γ12 +γ22 )
aγ3
V = 3 + b + ,
(γ12 +γ22 ) 4 γ12 +γ22 γ12 +γ22
μ = 0.
This is a case of algebraic potential, which involves three singular terms of dif-
ferent fractional powers. It reminds the case of fractional power potential and third-
degree integral due to Goryachev. The fourth integral for this case can be expressed
in terms of Euler’s angles, using formulas provided in [411], but it is not constructed
yet in the Euler-Poisson variables.
13.3.5 Two Conditional Cases Valid on a Single, Not

Necessary Zero, Level of the Linear Integral [421]
This case adds to Yehia’s gyrostat the parameter m, figuring in potential and gyro-
scopic terms, and turns into it when m = 0. The quartic integral is expressed in terms
of Euler’s angles, but not in Euler-Poisson variables [421] (Table 13.2).
@seismicisolation
@seismicisolation
Table 13.2 Conditional cases on a single level of the linear integral f = α(αar bitrar y)
1 A = B = 2C,

V = aγ1 + bγ2 − 2(γ 2m+γ 2 ) 2(k − m) − 2αγ3 + γ 2 m +γ 2
,
1 2 1 2
l = (0, 0, k + m
γ12 +γ22
)

2 V = aγ1 + bγ2 − k
2(γ12 +γ22 )
−2αγ3 + γ12 +γ22
k

c2 +d 2 2 −d 2
λ+γ3 2 γ12 +γ22 + c 2 γ2 +cdγ1
+ ,
γ12 +γ22
l = (0, 0, k + k
γ12 +γ22
).
Case 2 generalizes the case of Yehia and Bedwehi. In both cases, the added new
terms are all singular at the two positions γ = (0, 0, ±1).
13.4 Integrable Extensions of Conditional Integrable Cases
As remarked in the beginning of this chapter, the method of transformation with an

arbitrary function ν(γ1 , γ2 , γ3 ) used in Sect. 12.2 of the last chapter is also applicable
to all conditional integrable cases, valid on the zero level of the cyclic integral, i.e.
to the 20 cases of this type listed in the last three sections.
We shall not give here a list of generalizations of the conditional cases. Most of
those cases involve singular terms that are not likely to get acceptable physical inter-
pretation. Physical effects of the transformation are here immaterial and will remain
at present just as parts of mathematical models. Moreover, unlike the generalized
cases introduced in Chap. 12, the flexibility offered by the presence of the areas con-
stant as an arbitrary parameter is here lost. The transformed integrable problems and
their original counterparts share the same Hamiltonian. To this kind of extension of
conditional integrable cases applies the argument of Borisov and Mamaev [41], men-
tioned in the last section of the preceding chapter and they need not to be considered
unless for some reason it becomes necessary to use a concrete form of the function
ν in the transformation.
@seismicisolation
@seismicisolation
Chapter 14
The Rigid Body Acted upon by a Skew
Combination of Fields
The difficulties encountered in the study of the classical problem of motion of a heavy
rigid body have not been a stimulant for systematically considering problems of rigid
body dynamics in more general settings. In spite of its actuality and importance for
scientific and technological purposes and although it is a direct logical generalization
of the classical problem, the problem of motion of a rigid body under the action of
asymmetric forces was left aside and was considered only occasionally.
In what may be called a historical exception, the quest for integrable cases began
with constructing the set of integrals in Brun’s problem of motion of a body in an
asymmetric gravitational field [48]. This is one of the most complicated motions
of a rigid body. More than seven decades elapsed before Bogoyavlensky made a
resurrection of this problem, completing the proof of its integrability and indicating
a way for its integration [29]. He also gave a general integrable case, in which a
spherical body moves under the action of three skew fields. The first general integrable
case of a gyrostat acted upon by two skew uniform fields was found in [380] for the
body with Kowalevski’s configuration and under a restriction coupling the intensities
of the fields. Shortly later, this problem was shown to be integrable without that
restriction.
In this chapter, we consider the most general problem of motion of a rigid body
about a fixed point in an essentially new and general setting, comprising all types of
asymmetric, but conservative force fields. We give the first ever systematic presenta-
tion of this topic together with exhausting lists of presently known integrable cases,
which far exceed in number the three cases described above. Those new integrable
cases were obtained mainly by two methods, both depending on the existence of a
cyclic coordinate. The first is based on the equivalence between two problems on the
level of Routhian reduction after ignoring a cyclic coordinate in each. The second
uses the transformation of cyclic coordinates, as explained in Sect. 11.9, to construct
a generalization of known cases. We shall get acquainted with those methods along
the way.

https://doi.org/10.1007/978-3-030-96336-1_14
@seismicisolation
@seismicisolation
376 14 The Rigid Body Acted upon by a Skew Combination of Fields
Assume that the moving body is acted upon by the most general combination of
conservative potential and gyroscopic forces, described by the Lagrangian:
1
L= ωI · ω + l.ω − V, (14.1)
2
where I = diag(A, B, C) is the inertia matrix of the body. The scalar and vector
potentials V, l depend only on the Eulerian angles through the nine direction cosines
α1 , α2 , α3 , β1 , β2 , β3 , γ1 , γ2 , γ3 .
The Lagrangian (14.1) describes a conservative system with three degrees of
freedom, which admits the Jacobi integral (the Hamiltonian of the system)
1
I1 ≡ H = ωI · ω + V = const. (14.2)
2
The equations of motion of a rigid body are usually written in the Euler– Poisson
variables ω, α, β, γ. For the present problem this form, corresponding to (14.1), is
written as
∂V ∂V ∂V
ω̇I + ω × (ωI + μ) = α × +β× +γ× ,
∂α ∂β ∂γ
α̇ + ω × α = 0, β̇ + ω × β = 0, γ̇ + ω × γ = 0 (14.3)
where I is the inertia tensor of the body at the fixed point and
∂ ∂ ∂
μ = l+(α × + β× + γ× ) × l
∂α ∂β ∂γ
∂ ∂ ∂ ∂ ∂ ∂
≡ (l · α) + (l · β) + (l · γ) − ( · l)α − ( l)β − ( · l)γ
∂α ∂β ∂γ ∂α ∂β · ∂γ
− 2l. (14.4)
Equations (14.3), (14.4) were derived in [390]. They generalize to the case of
arbitrary forces Eq. (11.3) derived in Chap. 11 for the case of axi-symmetric forces
and can be seen to reduce to them when l and V depend only on γ. Moreover, it is
not hard to show that a part l 0 of the vector l has no contribution to the vector μ, if
it is derived from a scalar function χ(α, β, γ) by the formula
∂χ ∂χ ∂χ
l0 = α × +β× +γ× . (14.5)
∂α ∂β ∂γ
dχ
This part adds a gauge term dt
in the Lagrangian. The last formula generalizes (11.8)
for axi-symmetric fields.
@seismicisolation
@seismicisolation
Exercise: Show by direct calculation that the special form of the vector potential
l 0 (14.5) gives zero gyroscopic vector μ in (14.4).
The number of integrals of motion necessary for solving the equations of

motion.
The system (14.25) is composed of 12 first-order differential equations. It can be
easily checked that they satisfy the Jacobi (zero divergence) condition. According
to Jacobi’s theorem, one must find 12 − 2 = 10 integrals, of which we know only
seven, Jacobi’s integral and six geometric integrals
|α| = 1, |β| = 1, |γ| = 1, α · β = 0,β · γ = 0,α · γ = 0. (14.6)
The areas integral, which played an important role in our study of the classical
problem, does not exist in the present problem. Thus, it remains to find three more
integrals of motion.
Another approach widely used in analytical mechanics applies only to Hamilto-
nian systems, but it has proved more effective for our purpose. That is applying a
theorem due to Liouville (see, for example, [368]), which states that “For a Hamil-
tonian system with n degrees of freedom, the knowledge of n integrals of motion,
which are in involution is sufficient for completely solving the equations of motion”.
Such systems are usually named “completely integrable systems”.
To apply Liouville’s theorem, we first note that the mechanical system under
consideration, i.e. the rigid body in the most general fields, is a Hamiltonian system
of 3 degrees of freedom. In principle, one can use the three Eulerian angles as
generalized coordinates and the generalized momenta conjugate to them to write
equations of motion in the Hamiltonian form. In those coordinates, the geometric
integrals turn into identities, and we are left only with one general integral of motion
I1 . In our case n = 3, so that three integrals are sufficient for integration of the
problem. As one integral is known, we need only two other integrals for integrability,
independent of the energy integral I1 and in involution, instead of three demanded
by Jacobi’s theorem.
It turns out to be more advantageous to keep writing equations of motion in
the symmetric and in most cases algebraic form (14.3). Compared to Hamiltonian
equations in canonical variables, this form is found to be of greater help in finding
expressions for the integrals of motion and in constructing explicit solutions to the
problems under consideration.
14.1.1 Interpretation of Forces
Different terms of Eq. (14.3) in their general form may be interpreted in most cases
in one or more of the following (or other) ways.
@seismicisolation
@seismicisolation
The potential V can be understood as due to the scalar interactions of a gravita-

tional field with the mass distribution in the body, an electric field with a permanent
distribution of electric charges and a magnetic field with some magnetized parts or
steady currents in electric circuits on the body. A constant term κ of the vectors μ
and l is the gyrostatic momentum, while the variable terms of μ may appear as a
result of the Lorentz effect of the magnetic field on the electric charges. Let B and
A be the intensity of the magnetic field and the vector potential of this field at the
point r of the body where the current charge element de is placed. In that case, one
can write the vector l as (for details, see [382])

l = κ + r × Ade,
while μ can be derived from l according to (14.4 ) or constructed directly in the form
[382]:

μ = κ − (r · B)rde

∂
= κ + (r · )rde (14.7)
∂r
where is the scalar magnetic potential. In several cases of interest for future appli-
cation, can be expressed as a sum
(X, Y, Z ) = 1 (X, Y, Z ) + · · · + N (X, Y, Z ), (14.8)
of homogeneous harmonic polynomials up to the N th degree; the formula (14.4) can

be replaced by

N
μ = κ+ s s (X, Y, Z )rde.
s=1
Now, expressing r in the moving body axes x yz, we get

N
μ = κ+ s s (r · α, r · α, r · α)rde, (14.9)
s=1
i.e. components are polynomials in the direction cosines. In most cases of physical
interest the functions V, μ are polynomials in the direction cosines, but this will
not be assumed in general, since we shall deal also with some cases involving non-
polynomial terms.
In the following sections, we survey the few known integrable cases in asymmetric
fields. In case of existence of a cyclic variable, we show some tricks that enable
@seismicisolation
@seismicisolation
constructing integrable extensions of those cases. Several cases of definite physical

interest are obtained in this way.
14.1.2 The Motion of a Magnetizable Rigid Body in an Ideal

Fluid and In a Uniform Magnetic Field
Under the assumption that the body and the fluid are linearly magnetizable, it was
shown in [50] that the equations of motion may be derived from a Hamiltonian,
which generalizes that of Clebsch’s Hamiltonian (10.10) for the Kirchhoff problem
(see Chap. 10). The equivalent Lagrangian may be written as
1 1 1
L= ωI · ω + γB · ω + γC · γ + δD · δ + J · δ, (14.10)
2 2 2
in which the additional matrix D and vector J are determined by the shape of the
body and δ denotes the magnetic field intensity, constant in space.
A direct generalization of this Lagrangian is
1
L= ωI · ω + (κ + αB + βB + γB) · ω
2
1 1 1
+ αC · α + βC · β + γC · γ + a · α + b · β + c · γ. (14.11)
2 2 2
This generalizes all physical problems discussed above in this book, including the
motion of a multi-connected body in a liquid and the motion of an electrified body
in a superposition of three classical fields, whose potentials are at most quadratic in
the direction cosines. Some integrable problems introduced below fit as integrable
versions of the last Lagrangian.
14.1.3 Example: The Motion of a Satellite in a Circular Orbit
The centre of mass of a small satellite is moving in a circular orbit around a spherical
planet with angular velocity . We now write down the equations for its rotational
motion in the orbital frame. Let R be the radius of the satellite orbit and μ be Gauss’
constant of the planet. Designate by α, β, γ, respectively, the unit vectors in the
directions of the tangent to the circular orbit, the orthogonal to its plane and the
radial in that plane from the centre of the planet. The condition of balance of the
satellite in the radial direction is
(R)2 μ
= 2,
R R
@seismicisolation
@seismicisolation
so that we have μ
2 = . (14.12)
R3
The kinetic energy of the satellite is
1 1
T = M(R)2 + ωI · ω, (14.13)
2 2
where ω is the angular velocity in the inertial frame. On the other hand, the potential
of the satellite may be written, regarding Eq. (6.5), in the form
3 2
V = γI · γ. (14.14)
2
Neglecting a constant term, the Lagrangian may be written as
1 3
L= ωI · ω − 2 γI · γ. (14.15)
2 2
This Lagrangian describes the motion in the coordinate frame with origin at the
centre of mass of the satellite and axes with fixed directions in space.
Denote by ω the angular velocity in the orbital frame. We have
ω = ω + β. (14.16)
For detailed history of the “uniform precession” transformation (14.16) in rigid body
dynamics, the reader is referred to Chap. 10, Sect. 10.11.3. Substituting in (14.15),
we obtain the new Lagrangian in the orbital frame
1 3
L= (ω + β)I · (ω + β)− 2 γI · γ
2 2
1 1
= ω I · ω + βI · ω − 2 (3γI · γ − βI · β). (14.17)
2 2
Here, one can introduce Eulerian angles in the orbital frame and obtain three
equations of motion in terms of them. Alternatively, one may derive the equations of
motion from the Lagrangian (14.17) using (14.3) and (14.4) to obtain, as in [382],
ω̇ I + ω × (ω I − 2β Ī) = 32 γ × γI−2 β × βI. (14.18)
Augmenting this equation with two others, expressing constancy of the two unit
vectors β, γ in the orbital frame:
β̇ + ω ×β = 0, γ̇ + ω ×γ = 0. (14.19)
@seismicisolation
@seismicisolation
In this way, the problem of motion of a satellite in a circular orbit is reduced to

the problem of motion of a body about a fixed point, governed by Eqs. (14.18) and
(14.19), i.e. to a special version of (14.10).
14.2 The Rigid Body (Gyrostat) Acted upon by More than

One Uniform Field
We shall first make some clarification. In classical physics, there are three distinct
force fields: gravity acts on mass, magnetic field acts on magnetized parts of the body
and electric field acts on electric charges. However, the presence of electric charges
leads to the appearance of two new effects:
(a) Accelerated electric charges produce electromagnetic radiation. This causes
attenuation of the total energy of the body carrying them, from the classical point of
view.
(b) Moving electric charges interact with magnetic field and result in velocity-
dependent Lorentz forces.
If the velocity and acceleration of points of the body are small enough to be
neglected over some period of time, then over that period we can consider the motion
of the heavy, magnetized body carrying electric charges subject to the three uniform
fields, gravity, magnetic and electric.
We first write down the potential of the system
V = Mr0 · g + m · H + Q · E
where r0 , m denote the position vector of the centre of mass and the magnetic moment
of the magnetized parts of the body and Q =qr is the dipole moment of electric
charges carried by the body and g, H, E are the three fields: gravity, magnetic and
electric, respectively. The first three vectors r0 , m, Q are fixed in the body, while the
field vectors g, H, E are fixed in space. If we fix a system of axes in space with the
orthonormal basis α, β, γ, the potential can be rewritten in the form
V = a · α + b · β + c · γ. (14.20)
This is the most general form of the potential of a rigid body in the three classical
fields. It is the potential of three irreducible fields, i.e. that cannot be reduced to two,
as long as the vectors a, b, c are not coplanar. This means that the matrix
⎛ ⎞ ⎛ ⎞
a a1 a 2 a 3
M = ⎝ b ⎠ = ⎝ b1 b2 b3 ⎠ (14.21)
c c1 c2 c3
@seismicisolation
@seismicisolation
has rank 3. Equivalently, the determinant

a1 a2 a3

det(M) = b1 b2 b3 = 0. (14.22)
c1 c2 c3
Expressing (14.20) in Euler–Rodrigues’ parameters using expression (2.25) for the

rotation matrix, one converts the potential V to a quadratic form in λ0 , λ1 , λ2 , λ3 .
In the generic case, a rotation of the four-dimensional axes, which is equivalent to
rotating both space and body axes, determines a single coordinate frame in the body,
in which only diagonal elements of the rotation matrix are present. This means that
the potential in those axes has the form
V = aα1 + bβ2 + cγ3 . (14.23)
This process is completely independent of that of determining the principal axes of

inertia and thus simultaneous use of the last formula of the potential with principal
axes of inertia requires some restrictions on the parameters of the body. However, for
a body with complete dynamical (spherical) symmetry, the general potential (14.20)
can always be reduced to that form.
When rank(M) = 2, the potential is due to two fields and in the general case can
be written as
V = a · α + b · β, (14.24)
and, without loss of generality, one can take a = (a1 , a2 , 0) and b = (b1 , b2 , 0)
provided
a1 a2

b1 b2 = 0.
Lastly, when rank(M) = 1, the potential is due to only one field and in the general
case can be chosen to be the gravity field. The problem reduces to the classical
problem discussed in previous chapters.
The equations of motion of the body (gyrostat) moving about a fixed point in the
presence of three physical effects can be written in the form:
ω̇I + ω × (ωI + κ) = α × a + β × b + γ × c,
α̇ + ω × α = 0, β̇ + ω × β = 0, γ̇ + ω × γ = 0. (14.25)
The system of Eq. (14.25) admits the energy integral
1
I1 = ωI · ω + a · α + b · β + c · γ. (14.26)
2
Thus, to obtain a general integrable case of motion of the body acted upon by two
or three uniform fields, one must know two more integrals independent of I1 .
@seismicisolation
@seismicisolation
14.2 The Rigid Body (Gyrostat) Acted upon by More than One Uniform Field 383
The classical problem has benefited from the interest of several eminent math-
ematicians, who investigated analytical properties of the solution of Euler–Poisson
equations in the complex t-plane and conditions of existence of first integrals with
various analytical structures: polynomial, analytic or meromorphic or single-valued
in the Euler–Poisson variables. In contrast to all that, no effort was made to extend
any of those lines of investigation to problems involving more than one field.
In the following two sections, we give a brief description of the integrable cases
and particular solutions in both versions of the problem.
14.2.1 The Motion of a Body Acted upon by Two Uniform

Fields
This is the case of motion of a heavy magnetized rigid body (gyrostat) about a fixed
point under the action of two uniform gravity and magnetic fields. Equations of
motion can be written in the form:
ω̇I + ω × (ωI + κ) = α × a + β × b,
α̇ + ω × α = 0, β̇ + ω × β = 0. (14.27)
The investigation of such system and construction of integrable motions seem

to be initiated by Bogoyavlensky. In [30, 31], he has shown that Eq. (14.27) under
the Kowalevski condition I = C diag(2, 2, 1) for a = (a, 0, 0), b = (0, b, 0), κ=0
is completely integrable on the invariant relation (written here in our usual notation)
z 1 = p 2 − q 2 − aα1 + bβ2 = 0,
z 2 = 2 pq − (aα2 + bβ1 ) = 0. (14.28)
It turned out that this is a particular case of the unrestricted integral (14.31) below
and reminds a generalization of the Appelrot families of solutions to the case of a
simple body in two fields (without a gyrostatic moment).
In a much later work [69], it was claimed that the conditional classical integrable
case of Goryachev and Chaplygin admits a generalization to the problem of two fields
with two conditions of the areas integral type. It was shown in [406] that conditions
of this type lead only to pendulum-like motion and that the presence of the second
field has no significance.
14.2.1.1 The Hierarchy of Yehia and Reyman and

Semenov–Tian–Shansky
We now turn to general integrable cases of the present problem. In our note [380],
we have considered the problem of motion of the gyrostat acted upon by forces with
@seismicisolation
@seismicisolation
the unsymmetrical potential
V = C(a1 α1 + a2 α2 + b1 β1 + b2 β2 + c1 γ1 + c2 γ2 ). (14.29)
The gyrostatic moment was taken in the form
μ = Cκk, (14.30)
directed along the axis of dynamical symmetry of the body. It was found that this
problem admits the general fourth-degree integral
I3 = [ p 2 − q 2 − a1 α1 + a2 α2 − b1 β1 + b2 β2 − c1 γ1 + c2 γ2 ]2
+[2 pq − (a1 α2 + a2 α1 + b1 β2 + b2 β1 + c1 γ2 + c2 γ1 )]2
+2κ(r − κ)( p 2 + q 2 )
−4κ[ p(a1 α3 + b1 β3 + c1 γ3 ) + q(a2 α3 + b2 β3 + c2 γ3 )]. (14.31)
The integral (14.31) generalizes Kovalevskaya’s integral by including two arbitrary

constant fields and a gyrostatic momentum. However, the knowledge of this integral
is not sufficient to establish complete integrability of the case (14.29–14.30) since
the potential (14.29) does not allow a linear integral in general. Another integral
is needed. In the same work [380], two integrable cases have been pointed out,
characterized by the existence of a linear integral in each case:
(a) The case of axi-symmetric potential V = c1 γ1 + c2 γ2 admits the familiar areas
integral
I2 = 2( pγ1 + qγ2 ) + (r + κ)γ3 , (14.32)
corresponding to the angle of precession ψ as a cyclic variable. This case is

discussed in detail in Sect. 5.6.
(b) The case V = a1 (α1 − χβ2 ) + a2 (α2 + χβ1 ) (χ = ±1) is characterized by the
integral
I2,χ = 2( pγ1 + qγ2 ) + (r + κ)(γ3 + χ), (14.33)
which is also linear in velocities and corresponds to the cyclic variable ϕ ± ψ,

according to the chosen value for χ. Those cases are equivalent and can be
obtained from each other by reversing the sign of the vector fi. Thus one can
consider only the choice χ = 1. Note that, unlike case a, this case does not include
Kowalevski’s case as a special version. The two fields are coupled together and
can vanish only simultaneously.
An explicit solution for case (a) is commented in Chap. 5. Case (b) in its full
generality is not solved yet. In [189], the solution of a special version of this case,
namely when I2 = f = 4κ, was reduced to real elliptic quadratures. We shall return
to case (b) later in this chapter. It is in fact closely connected to another problem
of a completely different nature, namely Chaplygin’s case of motion of a body in
@seismicisolation
@seismicisolation
liquid, modified by the Goryachev singular potential term γε2 . This connection will
3
be extensively studied in Sect. 14.5 below.
The full system with potential (14.29) and gyrostatic momentum (14.30) turned
out to be integrable. This result was established by Reyman and Semenov–Tian–
Shansky in [311] (see also [26]), where a quadratic integral was constructed for
arbitrary parameters of the two fields. In the present notation, this integral can be
written as
I2 = [(G · α)a + (G · β)b + (G · γ)c]2

+ 2(r − κ)[1 G · α + 2 G · β + 3 G · γ]
+ 4k · [1 (b × γ + β × c) + 2 (c × α + γ × a)
+3 (a × β + α × b)], (14.34)
where G = ωI + κ is the total angular momentum and 1 = b1 c2 − b2 c1 , 2 =

c1 a2 − c2 a1 , 3 = a1 b2 − a2 b1 .
The three vectors a, b, c in (14.29) are not redundant, since we deal with only
two irreducible fields, this form is symmetric and also comfortable in obtaining the
above two limiting cases. In case a) we have I2 = c2 (G · )2 , which reduces to the
linear integral (14.32). In case (b) one can easily show that
I2 = −a 2 [I2,χ
2
+ 4χκI2,χ − 4I1 + 2κ2 ],
so that I2,χ is really an integral.

To put the integral (14.34) in its simplest form, one can normalize the potential
to the form
V = aα1 + bβ2 .
The integral takes the form
I2 = a 2 (G · α)2 + b2 (G · β)2 + 2ab[(r − κ)(G · γ) + 2(bα1 + aβ2 )]. (14.35)
To construct this general integrable case, the authors of [311] introduced a Lax
representation of the equations of motion, using a Lax pair with a spectral parameter.
The solution of the complexified equations is pointed out in terms of Theta functions.
However, without a real separation of variables, those results are not useful in the
study of bifurcation and qualitative properties of motion. Some works were devoted
to the topological analysis of this case, e.g. [177, 187]. It is worthy to mention also M.
Kharlamov’s work, where a generalization of the second and third Appelrot classes
was pointed out as a two-parameter family of doubly periodic motions [175] and
generalization of the fourth class in [188]. Separation of variables for generalized
fourth Appelrot class is discussed in [179, 180]. Extensions of the Appelrot classes
for the generalized gyrostat in two fields were investigated in [182]. The stability of
the stationary motions in some special versions of this problem was studied in [149]
and recently in [159].
@seismicisolation
@seismicisolation
Worth mentioning is the work [84], in which the authors put forward a systematic
way of the derivation of the algebraic curves of separation of variables for the classical
Kovalevskaya on top and its generalizations, starting from the spectral curve of the
corresponding Lax representation found by Reyman and Semonov–Tian–Shansky.
Recently, in a conference1 the author learned from Yu. N. Fedorov that further results
were obtained including expressions of the physical variables in terms of separation
variables for the generic Reyman and Semenov–Tian–Shansky case and its special
cases of symmetry (the case a of a single field and the case b of two fields of equal
intensities). However, most of these results are not published yet.
14.2.1.2 A Generalization of the Last Hierarchy
Strictly speaking, the case presented here does not belong to the above hierarchy of
cases in two uniform fields with the Reyman Semenov–Tian–Shansky on its top. The
deformation parameter ε1 simultaneously evokes forces with quadratic potential in
addition to a change in the application centres r1 , r2 of the two uniform fields to become
r1 = ε0 a(1, 0, 0) + ε1 aκ(0, 1, 0),

r2 = ε0 b(0, 1, 0) − ε1 bκ(1, 0, 0).
Note that this change keeps the application centres of the uniform fields in the body
orthogonal to each other. The changes are coupled by the gyrostatic moment parame-
ter κ and disappear when κ = 0. We list this case here in the most suitable place for it
in this section, since it is, like its predecessor R-STS case, of three degrees of freedom
(with no cyclic coordinates). The remaining two sections will be devoted to integrable
problems with cyclic coordinates.
Sokolov and Tsiganov [337] ε1 introduced (2002).

Reyman and Semenov–Tian–Shansky [311] ε1 = 0, ab = 0 (1987),
See also [26] (1989).
Yehia [380] ε1 = 0, b = ±a and ε1 = 0, ab = 0 (1986).
A = B = 2C,
V = −{ε0 (aα1 + bβ2 ) + ε1 κ(aα2 − bβ1 ) + ε21 [a 2 α32 + b2 β32 + 21 (aα2 + bβ1 )2 ]},
l = (2ε1 bβ3 , −2ε1 aα3 , κ + ε1 (aα2 − bβ1 )),
μ = (ε1 bβ3 , −ε1 aα3 , κ − ε1 (aα2 − bβ1 )).
I2 = a 2 [2 pα1 + 2qα2 + (r + κ) α3 − ε1 (bα3 β1 + aα3 α2 − 2bα1 β3 )]2
+b2 [2 pβ1 + 2qβ2 + (r + κ) β3 + ε1 (aα2 β3 + bβ1 β3 − 2aα3 β2 )]2
+2 [r − κ + ε1 (aα2 − bβ1 )][2 pγ1 + 2qγ2 + γ3 (r + κ) + ε1 (aα2 − bβ1 − aα3 γ2 + bγ1 β3 )]
−4ε0 (bβ2 + aα1 )
+4ε1 [2aqα3 − 2bpβ3 − (aα2 − bβ1 ) (κ + r ) − ε1 ((aα2 − bβ1 )2 + 2a 2 α32 + 2b2 β32 )]
(continued)
1“Classical mechanics, dynamical systems and mathematical physics” on the occasion of Aca-
demician Valery V. Kozlov’s 70th birthday. Steklov Mathematical Institute of RAS in Moscow,
20-24.01.2020.
@seismicisolation
@seismicisolation
I3 = [ p 2 − q 2 +ε1 2 −a 2 α12 + b2 β22 − ε1 (aα2 + bβ1 ) (κ − r ) + ε0 (aα1 − bβ2 )]2

+[2 pq − ε21 (aα2 + bβ1 )(aα1 + bβ2 ) + ε
1 (aα1 − bβ2 ) (κ − r ) + ε0 (aα2 + bβ1 )]2
−2 κ (κ − r + ε1 aα2 − ε1 bβ1 ) p 2 + q 2

−2 κ{−ε1 r [ε1 (aα1 − bβ2 )2 + (aα2 + bβ1 )2 + 2 bpβ3 − 2a qα3 ]

−2 q[ε1 2 − (aα2 + bβ1 ) aα3 − 2b2 β2 β3 + ε1 κaα3 + ε0 bβ3 ]

+2 p [ε1 2 (aα2 + bβ1 ) bβ3 + 2 a 2 α1 α3 + κε1 bβ3 − ε0 aα3 ]

+ε1 2 [{ε1 (aα2 − bβ
1 ) + κ}{(aα2 + bβ1 )2 + (aα1 − bβ2 )2 }
+2κ a 2 α32 + b2 β32 ]}.
The integrability of this case was established in 2002 by Sokolov and Tsiganov who
constructed a Lax pair for it. Although explicit forms of the integrals of motion fol-
low from the Lax pair, no effort was made towards constructing those integrals, a step
taken later by Kharlamov and Ryabov. The two integrals are written here following
Ryabov [324], but as usual in this book, we normalize the two fields and do some fur-
ther simplifications. The deformation parameters ε0 , ε1 are arbitrary, so that the above
formulas unify the two cases listed in [324] as different, in a form from which they are
obtained as special cases ε0 = 1 and ε1 = 1, respectively.
This integrable case, which occupies the top of its hierarchy, has been studied only
in very few works. Neither separation of variables nor the phase topology was investi-
gated in the full case. In certain circumstances, the full system of three degrees of free-
dom can be reduced to a lower dimensional system on invariant manifolds. In [324]
two invariant four-dimensional submanifolds were pointed out, on which the original
system reduces almost everywhere to a Hamiltonian one with two degrees of freedom.
The system of equations specifying one of the invariant submanifolds is a generaliza-
tion of the invariant relations (14.28) for the integrable Bogoyavlensky case, which, in
turn, generalizes the first Appelrot class. The method of critical subsystems, devised
by M. Kharlamov in [175, 176], is used to this end. The phase topology is described
and bifurcation diagrams are given for those subsystems.
Further study of the bifurcation diagrams led in [304] to the construction of certain
periodic solutions of the problem of motion in two fields. For those solutions, phase
variables are expressed as algebraic functions of a single auxiliary variable and a set
of constants. This auxiliary variable satisfies a differential equation which can be inte-
grated in elliptic functions of time.
It can be readily seen that when b = 0, the integral I2 reduces to the simple form of
the areas integral
2( pα1 + qα2 ) + (r + κ − aε1 α2 )α3 .
It may be also shown, using the energy and geometric integrals, that in the special cases
b = ∓a, I2 renders to the linear integral
2[( p ± aε1 β3 )γ1 + (q − aε1 α3 )γ2 ] + [r + κ + aε1 (α2 ± β1 )](γ3 ± 1).
The last expression generalizes (14.33) and reduces to it when ε1 = 0.
@seismicisolation
@seismicisolation
14.2.2 The Motion of a Body Acted upon by Three Irreducible

Uniform Fields
It seems that the general problem of motion under the action of three fields has escaped
attention and is still in that state. This may be caused by the very limited advancement
in its apparently simpler version of just two fields. The problem was not investigated
for integrability, so that no integrable cases are known and it is not known if they exist.
The only exception is the case of a body whose ellipsoid of inertia at the fixed point is
a sphere.
As far as we know, the list of works dealing with this problem for an asymmetric
body consists exclusively of three papers [149, 417, 418]. The first is concerned with
the determination of equilibrium positions in the potential of the special form (14.23)
and studying their stability in the case of a dynamically symmetric body. The other
two deal with regular precessional motion of Grioli’s type, studied in Chap. 8 in the
presence of a single gravity field. It turns out that this type of motion persists to exist,
under certain conditions, in the general problem of motion in three irreducible fields
and even when one of the three uniform fields is replaced by a linear one [153].
In this subsection, we give a brief description of the only known integrable case of
a spherical body acted upon by three irreducible uniform fields and two of its special
versions of definite interest. Interested readers may find the particular solution in the
above-mentioned papers.
14.2.2.1 Bogoyavlensky’s Case of Dynamically Spherical Body in Three

Uniform Fields
Bogoyavlensky has noted that the dynamics of the body of spherical dynamical sym-
metry A = B = C is integrable for the general potential (14.20). The problem was
reduced in quaternion variables to Neumann’s problem on the sphere S 3 [31]. In the
quaternion space, the integrals of motion are all quadratic in the velocities and the
explicit solution can be found in terms of the Riemannian Theta functions in three vari-
ables, as follows from that of the Neumann system [290–294].
It is possible, as explained earlier in this section, to choose the axes of the space
frame to reduce the potential to the much simpler form
V = C(aα1 + bβ2 + cγ3 ), (14.36)
containing only three parameters. This simple form of the potential enables us to express
the integrals of the problem in the Euler–Poisson variables, which was not undertaken
in [31]. We get, as in [407],
@seismicisolation
@seismicisolation
1
I2 = [ap P + bq Q + cr R] + bcα1 + caβ2 + abγ3 ,
2
I3 = 1/2{a 2 [ p 2 + P 2 ] + b2 [q 2 + Q 2 ] + c2 [r 2 + R 2 ]}
+bcp P + caq Q + abr R

+2a b2 + c2 α1 + 2b c2 + a 2 β2 + 2c a 2 + b2 γ3 , (14.37)
where P = ω · α, Q = ω · β, R = ω · γ. Two special cases of particular interest arise.
14.2.2.2 Case of a Linear Integral [407]
When |a| = |b| , without loss of generality, we can assume the potential in the form
V = C[a(α1 − β2 ) + cγ3 ]. (14.38)
This choice makes the variable ψ + ϕ cyclic and leads to the linear integral
I3 = ω · γ + r, (14.39)
and the complementary is
1
I2 = [a( p P − q Q) + cr R] − ca(α1 − β2 ) − a 2 γ3 .
2
14.2.2.3 Case of Three Linear Integrals [380]
When the three coefficients are equal in modulus |a| = |b| = |c| = a, one can take
V = Aa(α1 + mβ2 + m γ3 ), (14.40)
where m 2 = m 2 = 1. In that case, along with the energy integral
1
I1 = A( p 2 + q 2 + r 2 ) + Aa(α1 + mβ2 + m γ3 ),
2
one can write three linear integrals
I2 = ω · α − mp = c1 , I3 = ω · β − m q = c2 , I4 = ω · γ − mm r = c3 . (14.41)
As shown in [385] (see also case 1 in Table 14.5 of Sect. 14.5.4 below), explicit solu-
tion of this case can be written in terms of elliptic functions of time. It can also be noted
that this case is super-integrable, i.e. it has one superfluous integral (say) I4 more than
the three needed for integrability. The trajectories of the integrable rigid body motions
@seismicisolation
@seismicisolation
are closed curves in the space of phase variables ω, α, β, γ, resulting from the inter-
section of Liouville tori with the integral surface I4 = c3 , and hence motion on them
is generically periodic.
14.3 Integrable Cases of a Body with a Homogeneous

Quadratic Potential
14.3.1 Brun’s Case of the Asymmetric Body in an Asymmetric

Gravitational Field
Let an arbitrary body fixed from its centre of mass be acted upon by a gravitational
field whose potential is V0 = N1 X 2 + N2 Y 2 + N3 Z 2 , i.e. a three-dimensional gener-
alization of (6.7) in the elementary Brun problem of Chap. 6. The potential of the body
in that field can be written as
1
V = (aαI · α + bβI · β + cγI · γ). (14.42)
2
This potential can be interpreted alternatively as an approximate form of the potential
due to three sufficiently distant centres of Newtonian or Coulomb interactions. How-
ever, unlike the problem of motion in three skew fields, the three terms in (14.42) are
redundant. In fact, we have
αI · α + βI · β + γI · γ= Ii j (αi α j + βi β j + γi γ j )
= Ii j δi j = tr (I),
so that the potential can be equivalently written as
1
V = [(a − c)αI · α + (b − c)βI · β]. (14.43)
2
Nevertheless, keeping all terms preserves an obvious degree of symmetry and easily
gives all special cases of axial symmetry.
Brun found three integrals of motion [48]. The first is the total energy integral
1
I1 = (Ap 2 + Bq 2 + Cr 2 ) + V, (14.44)
2
and the second is
I2 = (A2 p 2 + B 2 q 2 + C 2 r 2 ) − ABC(aαI−1 · α + bβI−1 · β + cγI−1 · γ).

(14.45)
@seismicisolation
@seismicisolation
14.3 Integrable Cases of a Body with a Homogeneous Quadratic Potential 391
As the field has no axis of symmetry, the areas integral does not exist in this problem.
Brun found the third integral which turned out to be also quadratic and can be written
in the form
I3 = a(ωI · α)2 + b(ωI · β)2 + c(ωI · γ)2

+ ABC[bcαI−1 · α + caβI−1 · β + abγI−1 · γ). (14.46)
This integral is written in a more complicated form in [41]. It is easy to check that the
integrals I2 and I3 reduce, respectively, to one integral of Tisserand’s type (6.13) and
an areas integral in cases of axial symmetry of the gravitational field, i.e. whenever
two of the parameters a, b, c are equal (or vanish).
In 1910, Goryachev [116] pointed out several cases in which all three integrals of
motion are quadratic. However, all of them turn out to be special cases of Brun’s gen-
eral problem described above.
Although Brun has indicated the three integrals of motion, sufficient for integrabil-
ity in the sense of Liouville, he did not state that the problem of motion is integrable.
This step was completed long later by Bogoyavlensky in [29] where he used Lax pair
formulation of the equations of motion to prove integrability and to obtain explicit for-
mulas expressing angular velocities of the body in terms of θ-functions of four vari-
ables restricted to a three-dimensional manifold. As may be expected, the complex
formulas obtained are too complicated to have any impact on the study of qualitative
properties of the motion. In contrast to the case of a symmetric field, no further studies
were performed on this case using the integrals of motion obtained more than a century
ago.
Remark: Brun’s problem of Chap. 6 has a curious generalization as follows [29]:
Assume that a rigid body moves freely in a gravitational field with homogeneous
quadratic potential
1
Vg = (aξ 2 + bη 2 + cζ 2 )
2
in the inertial frame. Let (X, Y, Z ) be the coordinates of the centre of mass and r be the
position vector of the element of mass dm referred to the body system. The potential
of the body is

1
V = [a(X + r · α)2 + b(Y + r · β)2 + c(Z + r · γ)2 ]dm
2
1 1
= M(a X 2 + bY 2 + cZ 2 ) + (aαi α j + bβi β j + cγi γ j )Īi j
2 2
1 1
= M(a X 2 + bY 2 + cZ 2 ) − (aαI · α + bβI · β + cγI · γ)
2 2
3
+ tr(I)(a + b + c). (14.47)
2
The last term is constant and can be omitted. The Lagrangian of the motion takes the
form
@seismicisolation
@seismicisolation
1 1
L= M( Ẋ 2 + Ẏ 2 + Ż 2 ) − M(a X 2 + bY 2 + cZ 2 )
2 2
1 1
+ ωI · ω + (aαI · α + bβI · β + cγI · γ).
2 2
Thus, the equations of translational motion of the centre of mass and the rotational
motion about the centre of mass are completely separate. The first is solved as simple
harmonic motion and the second is Brun’s case just discussed above in this subsection.
Remark: Burov and Subkhankulov [50] pointed out a new integrable case of (14.10).
Since in this case B = J = 0 and C, D are proportional to I, the effect of magnetiza-
tion brings the same term as a gravitational source and the case is equivalent to Brun’s
case in asymmetric gravitational field.
14.3.2 Case of Dynamically Spherical Body
Bogoyavlensky [30] pointed out the integrability of the motion of a spherical body in
the fields whose potential may be written in the form
1
V = (k1 αJ · α + k2 βJ · β + k3 γJ · γ), (14.48)
2
in which J is an arbitrary (nonsingular) symmetric matrix and k1,2,3 are arbitrary con-
stants. Explicit expressions for integrals were not given in [30], but were provided in
[380] in a form that keeps non-diagonal coefficient matrix. Without losing generality,
we diagonalize that matrix here to be diag(k1 , k2 , k3 ).
I2 = CωJ · ω − det(J)[k1 αJ−1 · α + k2 βJ−1 · β + k3 γJ−1 · γ],

I3 = C[k1 (ω · α)2 + k2 (ω · β)2 + k3 (ω · γ)2 ]
− [k2 k3 αJ · α + k3 k1 βJ · β + k1 k2 γJ · γ]. (14.49)
It can be easily shown that, as in Brun’s case, the three constants ki can always be
reduced to two, so that one term of the potential is redundant. Also, the integrals I2 , I3
reduce to Clebsch’s type of spherical body in a liquid and an areas integral as two of
the coefficients k1 , k2 become equal.
The Lagrangian of the last problem can be written as L = 21 C!2 − V. In quaternion
space, using (2.46), this reads
˙ 2 + V (), ||2 = 1,
L = 2C (14.50)
where V is a homogeneous quartic polynomial in , obtained by substituting (2.25) in

(14.48). This Lagrangian describes the motion of a particle on the three-dimensional
sphere S 3 , pointed out simultaneously by Bogoyavlensky [32] and in a more detailed
direct way, by Kozlov [243].
@seismicisolation
@seismicisolation
14.4 The Motion of an Axi-Symmetric Body Under the Action of Asymmetric Forces 393
14.4 The Motion of an Axi-Symmetric Body Under the

Action of Asymmetric Forces
In the last two sections, we collected the well-known integrable cases of more or less
classical problems of motion of a rigid body or a gyrostat under potential forces having
linear or quadratic asymmetric potential. In this section, we present integrable cases
of motion under more general potential and gyroscopic forces. The body is assumed to
have physical, not only dynamical, axial symmetry about its z-axis, so that B = A and
the proper rotation angle ϕ is a cyclic coordinate. In order to keep aside generalization
of Lagrange’s case, we assume that the precession angle ψ is a palpable coordinate. We
first search for versions satisfying this condition among the known integrable cases.
The next step is to apply the transformation of cyclic velocities established in Chap.
11 Sect. 11.9 to build more general integrable cases.
14.4.1 Description of the Problem
For such problem, the Lagrangian can depend only on the angles θ, ψ or, equivalently,
on the three direction cosines α3 , β3 , γ3 . It can be written as
1
L= ωI · ω + l · ω − V, (14.51)
2
where the potential V and the quantities l · γ, l · n, l · k depend only on α3 , β3 , γ3 . We
denote any vector v by v̄ when it is referred to the inertial reference frame, so that
v̄ = vR = (v · α)α + (v · β)β + (v · γ)γ.
Then L becomes
1
L= ωI · ω + l¯ · ω̄ − V (α3 , β3 , γ3 ) (14.52)
2
and the cyclic integral has the form
Cr + l3 = Cr + l¯ · k̄ = f. (14.53)
In order to describe the motion of axially symmetric body, we shall use, in addition
to the cyclic angle of proper rotation ϕ, the direction cosines α3 , β3 , γ3 of the z-axis
of symmetry of the body, as redundant coordinates. Those are the components of the
unit vector k referred to the inertial coordinate frame. We know that the space time
derivative of k̄ is
@seismicisolation
@seismicisolation
d k̄
= ω̄ × k̄,
dt
so that
d
ω̄ = k̄ × +r k̄. (14.54)
¯
dt
We also have tan ψ = − αβ33 and hence
α3 β̇3 − β3 α̇3
ψ̇ = . (14.55)
α32 + β32
After some manipulations, using both equations in (14.52), we obtain as in [390] the
Routhian
1
R= A(α̇32 + β̇32 + γ̇32 )
2
f γ3
+ l¯1 (β3 γ̇3 − γ3 β̇3 ) + l¯2 (γ3 α̇3 − α3 γ̇3 ) + (l¯3 + )(α3 β̇3 − β3 α̇3 )
α32 + β32
( f − l¯1 α3 − l¯2 β3 − l¯3 γ3 )2
− [V + ]. (14.56)
2C
The problem thus reduces to that of motion of a particle of mass A on the unit sphere
|k|2 = α32 + β32 + γ32 = 1, (14.57)
under the action of certain potential and gyroscopic forces.

Comparing the last Routhian, describing what we shall call problem 1, with another
Routhian (11.52), one easily realizes that both Routhians have the same structure. The
second Routhian describes problem 2, the reduction of the problem of motion of a
spherical body in an axi-symmetric combination after ignoring the cyclic angle of pre-
cession. The two problems can be identified by applying the transformation
f → f, t → −t,
γ = (γ1 , γ2 , γ3 ) → k =(α3 , β, γ3 ),
l = l1 i + l2 j + l3 k → l¯ = l¯1 α + l¯2 β + l¯3 γ,
A−C
V (γ) → V (k) − ( f − l¯1 α3 − l¯2 β3 − l¯3 γ3 )2 . (14.58)
2 AC
Thus, the problem of motion of an axially symmetric body in a skew combination
of fields is brought into equivalence with that of motion of a body of spherical dynam-
ical symmetry in a coaxial combination of fields. However, equations of motion are
isomorphic for the two problems on the level of Routhian reduction of each. The full
@seismicisolation
@seismicisolation
Lagrangian systems of equations are not necessarily isomorphic, and hence also the
explicit solution in terms of the time of those systems.
14.4.1.1 Applications
Maximal reduction of order of the equations of motion. As done in Sect. 11.7, the
equations of motion can be reduced, for arbitrary V and l and on an arbitrary fixed
integral level {h, f }, to a single second-order differential equation for γ3 in α3 , say.
This may be accomplished by using Maupertuis’ principle and eliminating β3 in virtue
of the geometric integral (14.57). An explicit form of the reduced equation is given in
[390], where an example of the reduced equation is given for the problem of motion
of a symmetric satellite on a circular orbit. Interested reader is referred to that work.
Equivalent integrable problems. The analogy just established can be used to con-
struct integrable cases in problem 1, which is much less studied, from well-known
integrable cases of problem 2. From the last relation in (14.58), it is clear that in general
the resulting cases will be conditional one, since the cyclic constant f figures in the
potential in a significant way. However, a general integrable case produces a general
equivalent one in two circumstances:

(a) When l¯ = 0. Then the constant potential term A−C2 AC
f 2 can be ignored, or
(b) when the body has spherical dynamical symmetry, and then the extra terms in
the equivalent potential disappear.
We shall meet both circumstances in the application below.
In the next subsection, we shall construct two integrable cases by applying the
above method. For space considerations, we write down these two cases after the
application of a transformation of the cyclic velocity ϕ̇ = ϕ̇ + ν, where ν = ν(α3 , β3 , γ3 ).
This is equivalent to the change
r = r +ν. (14.59)
This transformation, as established in Chap. 11, Sect. 11.9, generically leads to con-
ditional integrable cases, but for suitably chosen ν leads to the construction of general
integrable cases. Note that the choice ν = n = const adds a gyrostatic moment Cn
along the axis of symmetry of the body.
14.4.2 General Integrable Cases
14.4.2.1 A General Integrable Case Relevant to Brun’s Case
Among the integrable cases presented hitherto in this chapter, only one admits an axi-
ally symmetric version amenable to the application of the transformation (14.59). In
the case due to Brun (Sect. 14.3.1), the body is in general asymmetric and moves under
@seismicisolation
@seismicisolation
forces whose potential has a certain quadratic form in the nine direction cosines given
by Eq. (14.42). The restriction of this problem to the case of an axi-symmetric body
B = A gives
1
V = (aα32 + bβ32 + cγ32 )
2
l = (0, 0, 0), (14.60)
where a, b, c are certain constants. If we transform this case using
ν = n + n 1 α32 + n 2 β32 + n 3 γ32 (14.61)
we get a new integrable one that is listed in Table 14.1 case 1.

Note that I3 is, in general, of the third degree in the angular velocities, since coef-
ficients of quadratic terms involve I2 linearly. However, when n 1 : n 2 : n 3 :: a : b : c
a constant factor can be cancelled out and I3 becomes of the second degree. If we set
n 1 = n 2 = n 3 = 0 then the new case generalizes the original one merely by the addi-
tion of a gyrostatic momentum κ = Cn along the axis of dynamical symmetry. The
present case was first obtained in [407]. A restricted version of it, in which the body
exhibits spherical dynamical symmetry, was obtained in [400] by a different method.
14.4.2.2 A General Integrable Case Relevant to Lyapunov’s Hierarchy
The second integrable case is valid for a body of spherical dynamical symmetry A =
B = C. It corresponds to choice nu as in (14.61). The obtained new integrable case is
inserted in Table 14.1 case 2.
Note that the parameter n engenders an arbitrary constant gyrostatic momentum
κ = Cn along the axis of symmetry z of the body. This case was obtained in [400]
using a transformation of the case 3 of Table 12.1 in Chap. 12. The solution of this
version is not known, except in the very special case s1 = s2 = s3 = n 1 = n 2 = n 3 =
0 analogous to Lyapunov’s case of motion of a body in liquid solved by Kötter in terms
of Theta functions of two arguments [235].
Physical interpretation of the special version n 1 = n 2 = n 3 = 0 is given in [390] as
motion of an axi-symmetric body carrying electric charges fixed in it, in the presence
of a magnetic field whose potential has the form Vm = 21 (J1 X 2 + J2 Y 2 + J3 Z 2 ).
14.4.2.3 Examples of Physical Interpretation
From the considerations of Sect. 14.1, we see that a physical interpretation of the
obtained cases is possible within the framework of motion of charged, magnetized
bodies in the presence of non-uniform combination of the three classical fields. Due to
the abundance of physical parameters representing the three distributions and the coef-
@seismicisolation
@seismicisolation
Table 14.1 General integrable cases

1 Yehia [407] (2001)
C = A (The spherical body) Yehia [400] (1998)
n = n 1 = n 2 = n 3 = 0. Yehia [390] (1988).
V = 21 C[(aα32 + bβ32 + cγ32 ) − 21 (n + n 1 α32 + n 2 β32 + n 3 γ32 )2 ],
l = (0, 0, C(n + n 1 α32 + n 2 β32 + n 3 γ32 )),
μ1 = −2C(n 1 α1 α3 + n 2 β1 β3 + n 3 γ1 γ3 ),
μ2 = −2C(n 1 α2 α3 + n 2 β2 β3 + n 3 γ2 γ3 ),
μ3 = C(n + n 1 α32 + n 2 β32 + n 3 γ32 ),
I2 = C(r + n + n 1 α32 + n 2 β32 + n 3 γ32 ),
I3 = (a − 2n 1 I2 )[A( pα1 + qα2 ) + C(r + ν)α3 ]2
+(b − 2n 2 I2 )[A( pβ1 + qβ2 ) + C(r + ν)β3 ]2
+(c − 2n 3 I2 )[A( pγ1 + qγ2 ) + C(r + ν)γ3 ]2
−A[(b − 2n 2 I2 )(c − 2n 3 I2 )α32 + (c − 2n 3 I2 )(a − 2n 1 I2 )β32
+(a − 2n 1 I2 )(b − 2n 2 I2 )γ32 ].
2 Yehia [400] (1998)
n = n 1 = n 2 = n 3 = 0. Arbitrary body axes. Yehia [392] (1989)
n = n 1 = n 2 = n 3 = 0. Yehia [390] (1988).
V = C{s1 α3 + s2 β3 + s3 γ3 − 21 (bcα32 + caβ32 + abγ32 )
− 21 (n + n 1 α3 + n 2 β3 + n 3 γ3 )2
+ 21 (n + n 1 α3 + n 2 β3 + n 3 γ3 )[(b + c) α32 + (c + a) β32 + (a + b) γ32 ]},
l = C{− 21 [(b + c)α3 α + (c + a)β3 β + (a + b)γ3 γ]
+(n + n 1 α3 + n 2 β3 + n 3 γ3 )k}
μ1 = −C[(n 1 α1 + n 2 β1 + n 3 γ1 ) + aα3 α1 + bβ3 β1 + cγ3 γ1 ],
μ2 = −C[(n 1 α2 + n 2 β2 + n 3 γ2 ) + aα3 α2 + bβ3 β2 + cγ3 γ2 ],
μ3 = C[n + n 1 α3 + n 2 β3 + n 3 γ3 − (aα32 + bβ32 + cγ32 )].
I2 = r + n − 21 [(b + c) α32 + (c + a) β32 + (a + b) γ32 ]
+n 1 α3 + n 2 β3 + n 3 γ3 ,
I3 = 21 {(b + c) [ω.α + (n + n 1 α3 + n 2 β3 + n 3 γ3 )α3 ]2
+ (c + a) [ω.β + (n + n 1 α3 + n 2 β3 + n 3 γ3 )β3 ]2
+ (a + b) [ω.γ + (n + n 1 α3 + n 2 β3 + n 3 γ3 )γ3 ]2 }
+ (s1 − n 1 I2 ) [ω.α + (a + n + n 1 α3 + n 2 β3 + γ3 n 3 )α3 ]
+ (s2 − n 2 I2 ) [ω.β + (b + n + n 1 α3 + n 2 β3 + γ3 n 3 )β3 ]
+ (s3 − n 3 I2 ) [ω.γ + (c + n + n 1 α3 + n 2 β3 + γ3 n 3 )γ3 ]
−abc{[ω.α + (n + n 1 α3 + n 2 β3 + γ3 n 3 )α3 ] αa3
+[ω.β + (n + n 1 α3 + n 2 β3 + γ3 n 3 )β3 ] βb3
+[ω.γ + (n + n 1 α3 + n 2 β3 + γ3 n 3 )γ3 ] γc3 }.
ficients of the three potentials, it should be easy to adjust those parameters to match
the potential in each case and, moreover, in a variety of choices.
We shall carry out detailed examples of the less obvious adjustment of the scalar
magnetic potential and the charge distribution to meet the Lorentz effect giving rise
to the vector μ in each case. This will be done for the two cases of Table 14.1 in this
@seismicisolation
@seismicisolation
section. It is easy to verify that the most general harmonic second-degree polynomial
potential can be reduced by a rotation transformation to the form
1
= a1 X + a2 Y + a3 Z + (a11 X 2 + a22 Y 2 + a33 Z 2 ), (14.62)
2
with coefficients subject to the single condition
a11 + a22 + a33 = 0, (14.63)
insuring that is harmonic. According to (14.9), one can write

μ1 = x F(r · α, r · β, r · γ)de,

μ2 = y F(r · α, r · β, r · γ)de,

μ3 = σ + z F(r · α, r · β, r · γ)de, (14.64)
where σ is a gyrostatic moment directed along the z-axis and
F(X, Y, Z ) = a1 X + a2 Y + a3 Z + a11 X 2 + a22 Y 2 + a33 Z 2 . (14.65)
To guarantee that ϕ is cyclic, we assume the distribution of electric

charges on the body
to be axi-symmetric around the z-axis. In virtue of symmetry x ε1 y ε2 z ε3 de is sym-
metric in ε1 and ε2 , and it vanishes whenever ε1 or ε2 is odd. We denote the remaining
integrals as

x de =
2
y de = J,
2
z de = J ,
2
x zde =
2
y zde = K ,
2
z 3 de = K .
One finally gets
μ1 = J (a1 α1 + a2 β1 + a3 γ1 ) + K (a11 α3 α1 + a22 β3 β1 + a33 γ3 γ1 ),

μ2 = J (a1 α2 + a2 β2 + a3 γ2 ) + K (a11 α3 α2 + a22 β3 β2 + a33 γ3 γ2 ),
μ3 = σ + J (a1 α3 + a2 β3 + a3 γ3 ) + K (a11 α32 + a22 β32 + a33 γ32 ). (14.66)
For case 1. Comparing the components μi in case 1 to (14.66), we find that the uni-
form part (a1 , a2 , a3 ) of the external magnetic field must vanish, so that the potential
(14.62) should be homogeneous quadratic. In addition to the symmetry around the z-
axis, the charge distribution should satisfy the single condition K´ = − 21 K , i.e.

(x 2 + 2z 2 )zde = 0. (14.67)
@seismicisolation
@seismicisolation
The coefficients in case 1, in terms of the parameters of the body and field, have the
form
σ K a11 K a22 K a33
n = , n1 = − , n2 = − , n3 = − . (14.68)
C 2C 2C 2C
This completes interpretation of the first case. Note that the transformation parameters
n, n 1 , . . . have their distinct contributions to the physical problem.
For case 2. Equating the coefficients in case 2 and (14.66), we find that the charge
distribution must satisfy the following two conditions J = −J, K = K , i.e.

z de = − x 2 de,
2
(14.69)

x 2 zde = z 3 de, (14.70)
and, without any loss of generality, one can express the parameters of case 2 in terms
of the parameters of the body and external magnetic field in the form:
J a1 J a2 J a3
n1 = − , n2 = − , n3 = − ,
A A A
σ
n= ,
A
a = −K a11 , b = −K a22 , c = −K a33 . (14.71)
If we define the moments of inertia of the distribution by Ae , Be , Ce , then from

symmetry we have the equality Ae = Be . The condition (14.69) imposed on the sec-
ond moments of the charge distribution can be put in the form Ae = Be = 0. This is not
a serious restriction, since electric charge distribution, unlike mass, can take positive
and negative densities.
14.4.3 Conditional Integrable Cases
14.4.3.1 Two Cases Valid on a Single Level of the Cyclic Integral
The integrals I2 , I3 can be verified directly for each case (Table 14.2).
14.4.3.2 Conditional Case on the Level f = 0

2
For example, the special cases F2 (x) = F1 (x) = x8 and F2 (x) = F1 (x) = x1 lead,
respectively, to V = 21 1 , 21 2 . The complementary integral I3 in the two cases, respec-
tively, is
@seismicisolation
@seismicisolation
Table 14.2 Cases valid for arbitrary ν(α3 , β3 , γ3 ) on the level I2 = ε

1 B = A, ν = ν(α3 , β3 , γ3 ),
V = 21 (aα32 + bβ32 + cγ32 ) + εν − 21 Cν 2 ,
l = (0, 0, Cν),
μ is found from l according to (14.4).
I2 = C(r + ν) = ε
I3 = a[A( pα1 + qα2 ) + C(r + ν)α3 ]2
+b[A( pβ1 + qβ2 ) + C(r + ν)β3 ]2
+c[A( pγ1 + qγ2 ) + C(r + ν)γ3 ]2
−A[bcα32 + caβ32 + abγ32 ]
2 A = B = C, ν = ν(α3 , β3 , γ3 ),
V = C{s1 α3 + s2 β3 + s3 γ3
− 21 (bcα32 + caβ32 + abγ32 ) + εν − 21 ν 2
+ 21 ν[(b + c) α32 + (c + a) β32 + (a + b) γ32 ]},
l = C{− 21 [(b + c)α3 α+(c + a)β3 β+(a + b)γ3 γ] + νk]
μ is found from l according to (14.4).
I2 = r − 21 [(b + c) α32 + (c + a) β32 + (a + b) γ32 ] + ν = ε,
I3 = 21 {(b + c) (ω.α + να3 )2 + (c + a) (ω.β + νβ3 )2 + (a + b) (ω.γ + νγ3 )2 }
+s1 (ω.α + να3 ) + s2 (ω.β + νβ3 ) + s3 (ω.γ + νγ3 )
−abc{(ω.α + να3 ) αa3 + (ω.β + νβ3 ) βb3 + (ω.γ + νγ3 ) γc3 }.
I3 = A[a(ω · α)2 + b(ω · β)2 + c(ω · γ)2 ] − 2

= A[a( pα1 + qα2 )2 + b( pβ1 + qβ2 )2 + c( pγ1 + qγ2 )2 − 2 , (14.72)
and
a + b + c − 1
I3 = A[a(ω · α)2 + b(ω · β)2 + c(ω · γ)2 ] + . (14.73)
2
The above case is separable in sphero-conic coordinates on the sphere. Another

type of potential separable in spherical coordinates can be written as
v( αβ33 ) v1 ( αβ33 )
V = u(γ3 ) + = u(γ3 ) + , (14.74)
1 − γ32 β32
where u, v and v1 are arbitrary functions. This potential, analogous to that of Chap. 9,
Sect. 9.7.1, will not be considered further here (Table 14.3).
@seismicisolation
@seismicisolation
14.5 Motion of a Body with Combined (Quaternion) Symmetry 401
Table 14.3 A separable case valid on the level I2 = 0

1 Bogoyavlensky [31] (1986): F2 = F1 is polynomial
B = A,
√ √
F1 (α∗ − β ∗√
)−F2 (α∗ + β ∗ )
V = β∗
,
α∗ = a + b + c − 1 ,
β ∗ = α∗2 − 4abc2 ,
1 = aα32 + bβ32 + cγ32 ,
2 = bcα32 + caβ32 + abγ32 .
I2 = r = 0,
I3 = A[a(ω · α)2 + b(ω · β)2 + c(ω · γ)2 ]+
√ √ √ √
+ √1β ∗ [(α∗ + β ∗ )F1 (α∗ − β ∗ ) − (α∗ − β ∗ )F2 (α∗ + β ∗ )].
14.5 Motion of a Body with Combined (Quaternion)

Symmetry
14.5.1 Introduction
Hitherto, we have studied integrable problems of two types of symmetry:

1- Symmetry of the fields around an axis fixed in space corresponding to a cyclic
variable ψ and leading to the areas integral.
2- Complete or physical symmetry of the body about an axis fixed in it. The last
symmetry corresponds to a cyclic variable ϕ (the angle of proper rotation), and it leads
to conservation of the generalized momentum conjugate to this angle.
In the present section, we present some integrable cases of motion of a rigid body
acted upon by a combination of non-coaxial fields under a different condition. We
assume that the Lagrangian of the problem admits the type of symmetry met earlier
in case b of Sect. 14.2.1.1, corresponding to a cyclic variable ψ ± ϕ. For determinacy
we consider only the case when ψ + ϕ is a cyclic variable. The other case is com-
pletely analogous. A useful interpretation of this symmetry follows easily from for-
mulas (2.41). It is symmetry with respect to rotation of the plane λ3 λ0 about its origin
in the space of Hamilton–Rodrigues’ parameters (the quaternion space restricted to
the unit sphere). Similar rotation in the plane λ1 λ2 corresponds to ϕ − ψ as a cyclic
coordinate. We have used for those types of symmetry the name “combined symme-
try”, since it combines the condition of dynamical symmetry of the body and a certain
condition of symmetry of the applied fields. In several works, M. Kharlamov referred
to it as “singular symmetry” (see, for example, [189]) or S 1 -symmetry [213].
Systematic treatment of this type of symmetry was initiated in [391]. That was
based on the observation that one of the new integrable cases introduced in [380],
namely case b of Sect. 14.2.1.1, concerned a problem of motion of a heavy magne-
tized body-gyrostat with the Kovalevskaya configuration A = B = 2C in a combi-
nation of uniform gravity and magnetic fields admitting this symmetry. Following up
@seismicisolation
@seismicisolation
the general case when the problem of motion admits this type of symmetry, it turned
out that the reduced Routhian equations of motion of such problem in the space of
Hamilton–Rodrigues’ parameters can be brought into isomorphism with the reduc-
tion of the, much better studied, problem of motion of a dynamically axi-symmetric
body in a coaxial combination of fields. This situation revealed exotic relation between
known integrable cases in both problems and also furnished a way to construct new
integrable cases of the first problem based on the analogy with known ones of the sec-
ond. This will be detailed in the coming subsection. However, the lists here are not as
complete as for the former two problems. Since the isomorphism is only on the level of
reduced problems, the determination of complementary integrals of the original prob-
lem is not an easy matter. Some of the integrable cases listed below are left without this
integral explicitly written down. Though it certainly exists, its determination needs
further work.
Finally, as systems with cyclic variables are amenable to generalizations of the type
introduced in Sect. 11.9, in the following subsection we apply what we called cyclic
velocity transformations to obtain new cases involving a larger number of physical
parameters—or a function—in their structure. For such application, we have chosen
those cases for which the complementary integral is known in the Euler–Poisson vari-
ables. To have complete lists one needs some additional work.
14.5.2 Routhian Reduction
Let a rigid body-gyrostat with A = B be in motion under the action of forces with
potential V admitting combined symmetry and gyroscopic moment l3 compatible
with the present type of symmetry (This may also include a gyrostatic moment directed
along the axis of dynamical symmetry.). In terms of Hamilton–Rodrigues parameters,
we have
V = V (λ1 , λ2 , λ), l3 = l3 (λ1 , λ2 , λ),

where λ = λ20 + λ23 . We shall use the redundant coordinates λ1 , λ2 and the polar
coordinates λ, in the λ0 λ3 plane, satisfying the condition
λ21 + λ22 + λ2 = 1. (14.75)
The Lagrangian of this system is
1
L= [A( p 2 + q 2 ) + Cr 2 ] + l3r − V. (14.76)
2
In terms of the quaternions, it can be put in the form
@seismicisolation
@seismicisolation
L = 2[A(λ̇21 + λ̇22 + λ̇23 + λ̇20 ) − (A − C)(λ0 λ̇3 − λ3 λ̇0 + λ2 λ̇1 − λ1 λ̇2 )2 ]

+2l3 (λ0 λ̇3 − λ3 λ̇0 + λ2 λ̇1 − λ1 λ̇2 ) − V
˙ 2 ) − (A − C)(λ2 λ̇1 − λ1 λ̇2 + λ2 )
= 2[A(λ̇21 + λ̇22 + λ̇2 + λ2 ˙ 2]
2 ˙
+2l3 (λ2 λ̇1 − λ1 λ̇2 + λ ) − V. (14.77)
The variable is cyclic and corresponds to the cyclic integral
˙ + 4(C − A)(λ2 λ̇1 − λ1 λ̇2 ) + 2l3 } = f.

I2 = λ2 {4[A + (C − A)λ2 ] (14.78)
Ignoring the cyclic variable, we obtain the Routhian [391]
1
R= ˙
(L − f )
4
A2 C λ̇21 + λ̇22 λ̇2 1 f A
= ( + ) + [ (C − A) + l3 ](λ2 λ̇1 − λ1 λ̇2 )
2D A C D 4 2
V 1 f 1 2 2
−[ + 2 ( − λ l3 ) ], (14.79)
4 2λ D 4 2
where D = A(λ21 + λ22 ) + Cλ23 .

Reduced equations of motion in quaternions (λ1 , λ2 , λ) as redundant coordinates
may be derived simply from the last Routhian subject to the geometric condition (14.75).
However, this is of secondary importance. This Routhian may be used in various ways,
keeping in mind the obvious resemblance to the Routhian (11.44) in redundant coor-
dinates (γ1 , γ2 , γ3 ) if one adds to the latter the dynamical symmetry condition B = A.
Simple examples are
(1) The analog of Staude rotations: Stationary solutions of the system described
by (14.79) are those for which λ̇1 = λ̇2 = λ̇ = 0. This may be rewritten as θ̇ = 0, ψ̇ =
ϕ̇, which characterizes a precessional motion of the body, whose figure axis is the z-
axis and precession axis is the Z -axis fixed in space.
(2) The maximal reduction of order: Exactly as in Sect. 11.7, the equations of
motion can be reduced, for arbitrary V and l3 and on arbitrary fixed integral level
{h, f }, to a single second-order differential equation in λ, λ1 . This may be accom-
plished by using Maupertuis’ principle and eliminating λ2 . An explicit form of such
reduced equation for the case l3 = const = k3 (a gyrostatic momentum along the sym-
metry axis) is given in [391], where some applications of it are pointed out. Interested
readers may consult that work.
14.5.3 Equivalence of Two Problems
Now, we turn to the most useful application of the Routhian (14.79). Denote by R the
Routhian of the problem of motion under coaxial fields considered in Sect. 11.7, but
with potentials V (γ) and l = (0, 0, l3 (γ)). If we add Setting A = B in the Routhian
@seismicisolation
@seismicisolation
(11.44), resulting from ignoring the angle of precession, takes the form
A2 C γ̇12 + γ̇22 γ̇32 1 f Cγ3

R = ( + )+ (Al
+ )(γ2 γ̇1 − γ1 γ̇2 )
2D A C D 3
γ12 + γ22
1
−[V + ( f − l3 γ3 )2 ], (14.80)
2D
where D = A(γ12 + γ22 ) + Cγ32 .

Now, if we compare the Routhians (14.79) and (14.80), we note a curious simi-
larity in their structures. They can be made identical, through a change of variables
→ (λ1 , λ2 , λ), if we add the following conditions on the potentials of the two prob-
lems:
f A f Cλ
(C − A) + l3 = Al3 (λ1 , λ2 , λ) + 2 ,
4 2 λ1 + λ22
V 1 f 1 1
+ 2 ( − λ2 l3 )2 = V (λ1 , λ2 , λ) + [ f − λl3 (λ1 , λ2 , λ)]2 . (14.81)
4 2λ D 4 2 2D
For the time being, one would regard f, f as the constant values of the linear inte-
grals of motion in the two problems (not the integrals themselves, which depend on
the phase variables). Note that if l3 () is a smooth function on the Poisson sphere, then
the potentials l3 and V have a singularity on the great circle λ = 1(λ20 + λ23 = 1) on
the unit quaternion sphere. Since such behaviour is not favourable for a real problem,
then, for the sake of simplicity, we shall impose the additional condition f = 0.
Finally, let a case B (say) of the axi-symmetric problem be given, on the level f =
0 of the linear integral, with the pair (V , l3 ). Then the corresponding problem A (say)
with combined symmetry will be characterized by the pair (V, l3 ) given by
f (A − C)
l3 = 2l3 + ,
2A
f f2
V = 4V + l3 − . (14.82)
A 8Aλ2
Here the λ’s are substituted, using formulas of Chap. 2, from the table
1 + α1 − β2 − γ3
λ21 = ,
4
1 − α1 + β2 − γ3
λ22 = ,
4
α2 + β1
λ1 λ2 = ,
4
1 + γ3
λ2 = .
2
Thus, in problems amenable to the above construction the potentials V, l3 are func-
tions of the three quantities: α1 − β2 , α2 + β1 , γ3 .
@seismicisolation
@seismicisolation
The inverse of the transformation (14.82) may be written as
1 f (A − C)
l3 = l3 − ,
2 4A
1 f f2
V = V − l3 + . (14.83)
4 8A 32 Aγ32
The dynamical parameter f figures in the two functions, so that the integrable case is
generically conditional. However, in certain cases it is possible to construct also gen-
eral integrable cases. This occurs mostly when the potential V contains in its structure
some arbitrary parameter, so that f can be absorbed in it. Examples will be provided
below.
The isomorphism of problems A and B was first established in [391] (1988) for
the simpler case when both l3 , l3 are constant gyroscopic momenta directed along the
axis of dynamical symmetry. In the same work, the appearance of the Goryachev sin-
gular term γa2 in problem B was associated with the cyclic constant of problem A. In
3
[403], the isomorphism was extended to the case when l3 , l3 are variable quantities. In
both works, the isomorphism concerned the two problems A and B on the level of their
Routhian reductions. The original full problems, before ignoring the cyclic variable in
each, are of course equivalent, but, in general, they are not isomorphic. In [37] of 1997
(see also their book [38] of 1999), Borisov and Mamaev found explicit transformation
between the phase variables in the two reduced problems, and thus established isomor-
phism between them. However, their result was applied only in two cases of purely
potential forces (l3 = 0), because they usually do not write equations of motion for
the asymmetric fields problem A in the presence of skew gyroscopic forces (the Pois-
son bracket equivalent of Eq. (14.3)). It may be noted here that the comment in [41]
(Sect. 4.1) “... a little earlier H. Yehia [623] (1988) had used a restricted version (for
M3 ± N3 = 0)” is misleading, since the real condition was f = 0. The latter is the
areas integral in problem B.
Alternative formulation:
When written in terms of Euler’s angles, the Lagrangian (14.76) becomes
1
L= [A(θ̇2 + sin2 θψ̇ 2 ) + C(ψ̇ cos θ + ϕ̇)2 ] + (ψ̇ cos θ + ϕ̇)l3 − V. (14.84)
2
We shall now introduce a change of variables
θ = 2, ψ = − , ϕ = + . (14.85)
This will change (14.84) to
L = 2{A[ ˙ − )
˙ 2 + sin2 cos2 ( ˙ 2 ] + C(sin2
˙ + cos2 )
˙ 2}
+2l3 (sin2 ˙ + cos2 )
˙ − V. (14.86)
@seismicisolation
@seismicisolation
Note that V and l3 are now functions of and only. The cyclic integral correspond-
ing to the cyclic variable is
˙ − 4(A − C) sin2 cos2

4 cos2 D ˙ + 2l3 cos2 = const = f (14.87)
where D = A sin2 + C cos2 . Thus one can write
˙
f − 2l3 cos2 + 4(A − C) sin2 cos2
˙ =
. (14.88)
4 cos2 D
This can be used to ignore the cyclic variable and construct the Routhian
1
R= (L − f )˙
4
1 C ˙
= A( ˙ 2 + sin2 ˙ 2 ) + sin2 [ f (C − A) + 2 Al3 ]
2 D 4D
V ( f − 2 cos2 l3 )2
−[ + ]. (14.89)
4 32 cos2 D
It is curious to remark that the linear term in this Routhian (14.89) (and also in (14.79))
vanishes whenever f (C − A) + 2 Al3 = 0. That is, for a body with constant gyro-
static momentum l = (0, 0, k3 ), on the level f = 2A−C Ak3
, the reduced equations of motion
become time-reversible. This property is characteristic of the motion of the body in
pure potential forces as was shown in Chap. 9. On the other hand, it may be said that
the parameter f partially engenders a gyrostatic momentum component along the axis
of dynamical symmetry, enough to annul the effect of the existing gyrostatic moment
on the motion of the reduced system. This situation is used in [189] to get a real sep-
aration of variables and reduce the solution of the reduced equations of motion of the
body in two coupled uniform fields to elliptic quadratures. We shall return to this point
shortly later.
The Routhian of problem B in Euler’s angles is
1 C ϕ̇
R = A(θ̇2 + sin2 θϕ̇2 ) + ( f C cos θ + Al3 sin2 θ)
2 D D
( f − l3 cos θ)2
−[V + ]. (14.90)
2D
Formulas corresponding to (14.81) are
@seismicisolation
@seismicisolation
f (A − C) 2C f cos
l3 = 2l3 + + ,
2A A sin2
f l f2
V = 4V + 3 −
A 8A cos2

f 2C f
+ 2 2 [2 f (A + C) + (C f − 4 Al3 ) cos + ].
A sin sin2
(14.91)
In terms of the Euler angles of the original problem A, the last formulas read
θ ϕ−ψ f (A − C) 4C f cos 2θ
l3 = 2l3 ( , )+ + ,
2 2 2A A(1 − cos θ)
θ ϕ−ψ f θ ϕ−ψ f2
V = 4V ( , ) + l3 ( , )−
2 2 A 2 2 4 A(1 + cos θ)
2f θ ϕ−ψ θ 2C f
+ 2 [2 f (A + C) + (C f − 4 Al3 ( ,

)) cos + ]. (14.92)
A (1 − cos θ) 2 2 2 1 − cos θ
Those formulas turned out to be useful and fast in converting integrable cases of prob-
lem B to their counterparts of problem A, especially when f = 0. They were effec-
tively used in the construction of most of the integrable cases listed below.
14.5.4 Basic Equivalent Integrable Problems
The most useful application of the isomorphism established above is that for any inte-
grable case of problem B there corresponds an integrable case of problem A, and vice
versa. As problem B is much more studied, the isomorphism will mostly work in the
direction from B to A. Among the conditional integrable cases of Chap. 13, all cases
valid for B = A and l1 = l2 = 0 satisfy the requirement for problem B and all have
equivalent cases in problem A. However, not all the resulting cases are equally sound.
As noticed in Chap. 2, potentials in problem A that involve terms odd in the quater-
nion variables λ1 , λ2 are double-valued on the group of rotations S O3 . This fact ren-
ders of no physical significance the equivalent of the two classical integrable cases of
Kowalevski and Goryachev–Chaplygin. In fact, those cases have the same potential
V = aγ1 + bγ2 and hence the equivalent cases have the potential V = aλ1 + bλ2 .
It is curious that the full general integrable case of Kowalevski has as its equivalent a
conditional integrable case, valid only on the level f = 0, while its conditional ver-
sion with f = 0 containing the singular Goryachev term γε2 is a general integrable
3
case. The cyclic constant f and the arbitrary constant ε are connected by the relation
2
ε = 32f A . Integrable cases involving quaternion (double-valued) potentials are first
met in [391] (see also [403]). They are considered in more detail in [41],where they
are called quaternion integrable cases.
One of the integrable cases equivalent to the well-known case of problem B exhibits
at once several characteristics of the correspondence between the two problems. Con-
@seismicisolation
@seismicisolation
sider case 17 of Table 13.1 in Chap. 13, i.e. the case characterized by the potentials:
l = C(0, 0, κ ),
λ
V = C[a1 γ1 + a2 γ2 − nκγ3 + b1 (γ12 − γ22 ) + 2b2 γ1 γ2 + ],
2γ32
to which we applied the condition n = 0, in order to make it amenable to the above

transformation. For this case, formulas (14.82) give
f
l = C(0, 0, κ + ),
4C
ε
V = 4C[a1 λ1 + a2 λ2 + b1 (λ21 − λ22 ) + 2b2 λ1 λ2 + ]
2λ2
f f f2
− (κ + )+ .
16 4C 64Cλ2
Renaming arbitrary constants and ignoring the constant term in V, one gets
l = C(0, 0, κ),
c
V = C[a0 λ1 + b0 λ2 + 2a(λ21 − λ22 ) + 4bλ1 λ2 + ], (14.93)
2λ2
in which all parameters are now arbitrary and free of any conditions. This signifies that
(1) No restriction is there on f, i.e. the integrable case (14.93) is unconditional. It is
valid for all feasible initial conditions,2 though the equivalent integrable case of
problem B is only conditionally integrable, on the zero areas integral level.
(2) The parameter f is absorbed with κ in the arbitrary gyrostatic momentum κ =
κ + 4Cf .
(3) The two terms 2a(λ21 − λ22 ) + 4bλ1 λ2 engendered by Chaplygin’s potential for
the body moving in a liquid yield in problem A, after expressing them in direction
cosines, the terms
a(α1 − β2 ) + b(α2 + β1 ). (14.94)
This is the potential in case B of Sect. 14.2.1.1 above, the first case of its type,
discovered in 1986 [380].
(4) The last term in the potential gives 2λc 2 = 2 cos2c(θ/2) = 1+γ
c
3
. This term presents a
singularity at γ3 = −1. An interpretation will be given below for it.
(5) Finally, there remain the first two terms, namely
a0 λ1 + b0 λ2 .
2When c = 0, one has to exclude the positions that satisfy λ = 0(θ = π) from the configuration
space.
@seismicisolation
@seismicisolation
Table 14.4 Conditional integrable cases for which ϕ + ψ is cyclic. For all cases, the cyclic integral
is A( pγ1 + qγ2 )+Cr (γ3 + 1) = 0
1 Separable in spherical coordinates , .
Analog of case 2 Chap. 13, using the Routhian (14.89).
B = A, l = μ = 0,
V = F1 () + C sinD2 F2 ()
= V1 (γ3 ) + A+C+(A−C)γ3
C(1−γ3 ) F( α1−γ
1 −β2
3
).
V1 , F arbitrary functions.
I2 = A( pγ1 + qγ2 )+Cr (γ3 + 1) = 0
I3 Not constructed.
2 Separable in sphero-conical coordinates on the reduced quaternion sphere (14.75).
Analog of case 3 Chap. 13.
A = B = C, l = μ = 0,
V = v1 [V1 (u − v) + V2 (u + v)],
I2 = A( pγ1 + qγ2 )+Cr (γ3 + 1) = 0
I3 Not constructed.
If alone, those terms designate a general integrable case that imitates

Kowalevski’s potential in the quaternion space. Since those terms are not
single-valued on the group of rotations, no physical significance can be adhered
to them. In the tables of integrable cases below, we shall ignore such terms (odd
terms in the rotation quaternion) from consideration.
In the present subsection, we list some equivalent integrable cases corresponding to
different cases tabulated in Chap. 13. Cases are presented in their basic form as implied
by corresponding cases of problem B. Some generalizations are left to the next subsec-
tion. We ignore cases that degenerate under the dynamical symmetry condition or the
condition l1 = l2 = 0 and, in other cases, we also ignore terms that are double-valued
in the direction cosines. Regarding those considerations, the total number of different
cases is much less than cases of type B (Table 14.4).
V1 , V2 are arbitrary functions of their arguments,
u = a + b + c − (aλ21 + bλ22 + cλ23 )

3 c 1 1
= (a + b) + + (b − a)(α1 − β2 ) + (a + b − 2c)γ3 ,
4 2 4 4
v 2 = u 2 − 4(bcλ21 + caλ22 + abλ2 )
= u 2 − 2ab(1 + γ3 ) + c(a − b)(α1 − β2 ) − c(a + b)(1 − γ3 ), (14.95)
a, b, c are arbitrary constants. For example, the choice V2 (x) = −V1 (x) = −x 2 leads
to a potential of the form
V = a (α1 − β2 ) + b γ3 .
@seismicisolation
@seismicisolation
Another example is V2 (x) = −V1 (x) = 1/x. This yields the potential
d
V = ,
2ab(1 + γ3 ) − c(a − b)(α1 − β2 ) + c(a + b)(1 − γ3 )
where d is a constant.
The solution is reduced in [385] to quadratures, elliptic for linear V and hyper-
elliptic for higher degree polynomial potentials.
2 Yehia [407] (2001)

A = B = C,
l = μ = 0,
V = A[a1 (α1 − β2 ) + a2 γ3 ].
I2 = ω · γ + r,
I3 = a1 ( pω · α − q! · fi) + a2 r ω · γ − 2a1 [a2 (α1 − β2 ) + a1 γ3 ]
Although this is a special case of that of Bogoyavlensky with three quadratic inte-
grals (see Sect. 14.2.2.2), this case was not noticed in [32]. It was singled out in [407],
where the integrals were also given.
3 Yehia [403] (2000), Borisov and Mamaev [40]

Yehia [380] (1986), [391] (1988) c = 0.
A = B = 2C,
l = μ = C(0, 0, κ),
V = C[a(α1 − β2 ) + b(α2 + β1 ) + 1+γ c
3
],
I2 = 2( pγ1 + qγ2 ) + (r + κ)γ3 + r,
I3 = [ p 2 − q 2 − a(α1 + β2 ) + b(α2 − β1 )]2 + [2 pq − b(α1 + β2 ) − a(α2 − β1 )]2
+2κ(r − κ)( p 2 + q 2 ) − 4k[ p(aα3 + bβ3 ) + q(bα3 − aβ3 )]
+ 1+γ2c
3
[ p 2 + q 2 + κr − κ2 + 2(1+γ
c
3)
].
History: The full case 3 (c = 0) was presented in 2000 with integrals in the Euler–
Poisson variables. In 2001, it was independently found, but in a different set of Hamil-
tonian variables based on quaternions as configuration variables [40].
The special case c = 0 was found first in [380], together with its integrals I2 and
I3 . Neither the full nor the special integrable case has known general explicit solution
yet in terms of time. An exception is the case when, in addition to the condition c = 0,
one can rotate the x y-axes in order to make b = 0. Then the potentials become
V = Ca(α1 − β2 ), l = C(0, 0, κ). (14.96)
The integration of the Routhian reduction of the problem characterized by the last
potentials was investigated in [189]. It was reduced on the level I2 = f = 4κ to two
hyper-elliptic quadratures. When f = 4κ = 0, the quadratures become elliptic, and
resemble those for Chaplygin’s case of a body in a liquid (see Chap. 10, Sect. 10.17).
It can be seen easily that the equivalent case in problem B corresponds according to
the choice
@seismicisolation
@seismicisolation
l3 = 0,
κ2
V = a (γ12 − γ22 ) + + const . (14.97)
8γ32
This is the subcase of case 17 of Table 13.1 in Chap. 13 that was considered by Ryabov
in [323] and reduced to hyper-elliptic quadratures.
It should be noted that the two types of separation of problem B in [323] and prob-
lem A in [189] are quite different. In the first, all phase variables in the (generalized)
Euler–Poisson equations of motion are determined in terms of the separation variables
and then the cyclic precession angle is found by integrating equation of the type (11.7).
In the second, separation is performed only for the reduced (Routhian) equations and
the cyclic coordinate is determined by another quadrature. The angular velocity is to
be derived from the Eulerian angles for the original problem A.
Physical interpretation of the full case 3 may be given as follows [403]:
(1) Let an axi-symmetric gyroscope with axial moment of inertia IG placed along the
axis of dynamical symmetry of the body be kept rotating with uniform angular
velocity . The vector μ is the gyrostatic moment due to this rotor, provided we
take Cκ = IG .
(2) Let the body have total mass M, centre of mass r0 = (x0 , y0 , 0) lying in its equato-
rial plane. Suppose the body also contains some magnetized parts with total mag-
netic moment m = (m 1 , m 2 , 0). Let the system be moving in the presence of a
uniform gravity field g and a horizontal magnetic field H fi. Its potential will be
g(x0 α1 + y0 α2 ) + H (m 1 β1 + m 2 β2 ).
This expression can be identified with the first two terms of V if we take
m · r0 = 0 and |gr0 | = |H m|, so that Ca = gx0 , Cb = gy0 .
(3) Consider two points P(0, 0, 1) fixed in the body on the z-axis (its axis of dynami-
cal symmetry) and Q on the fixed Z -axis at Z = −1. The distance P Q = 2 cos 2θ .
The singular potential term in case 3 is thus inversely proportional to the square
of the distance P Q.
It is curious to note that the singular term √ 1 2 = sin1 (Yehia–Bedwehy term) in
1−γ3
problem B corresponds in problem A to sin1 θ , which is proportional to P1R , where
2
R is the other pole of the fixed sphere opposite to Q. Such interaction may be inter-
preted as due, for example, to Coulomb interaction.
Among the cases of Chap. 13, some have physical equivalent cases and others
correspond to purely “quaternion” cases. Example of last type is the equivalent of
case Chap. 13#8. This is a complicated conditional integrable case ( f = 0), since the
potential does not include the singular Goryachev term γ12 . This case includes as a spe-
3
cial case, κ = e0 = e1 = 0, the original Goryachev singular potential, which gives
rise to the purely quaternionic equivalent potential
@seismicisolation
@seismicisolation
aλ1 + bλ2 + c
V = 2 . (14.98)
λ3
A similar case characterized by
a
V =
(λ21 λ22 λ2 )1/3
a∗
= (14.99)
{(1 + γ3 )[(1 − γ3 )2 − (α1 − β2 )2 ]}1/3
results from case Chap. 13#9 (Gaffet’s case). Note that this singular potential is single-
valued, in the real sense, on S O3 and hence physical interpretation is possible in certain
regions, excluding singularities at θ = 0, π and positions at which sin(ϕ − ψ) = 0.
The equivalent of case Chap. 13#14 is characterized by the potential
c + dγ3 d1 d2
V = b(α1 − β2 ) + + (3 − γ3 )[ + ].
1 + γ3 α1 − β2 − γ3 + 1 α1 − β2 + γ3 − 1
(14.100)
Case Chap. 13#16 generates the general integrable case:
K
l3 = κ + [d(α1 − β2 ) + c(α2 + β1 )],
2
1 ∗ 2ε
V = [(k K d − J c)(α1 − β2 ) + (k ∗ K c + J d)(α2 + β1 ) +
2 1 + γ3
K2 2
+ [(c − d 2 )(α1 − β2 )2 + 2cd(α1 − β2 )(α2 + β1 ) + (c2 + 2d 2 )γ32 + 2c2 γ3 ],
8
(14.101)
where k ∗ is a renaming of the arbitrary constant k. This case generalizes case 3 of

Table 14.5 above and reduces to it when K = 0. For this case the complementary inte-
gral is not found yet. If one sets ε = 0, then V becomes a polynomial in the direction
cosines.
Table 14.5 General integrable cases for which ϕ + ψ is cyclic. For all cases, the cyclic integral is
I2 = A( pγ1 + qγ2 ) + (Cr + l3 )(γ3 + 1) = f
1 Yehia [385] (1987)
A = B = C, l = μ = 0,
V = F(λ21 ) = V (α1 − β2 + γ3 ),
This case is super-integrable with three linear integrals:
I2 = ω · γ + r,
I3 = ω · α + p,
I4 = ω · β − q.
@seismicisolation
@seismicisolation
14.5.5 Generalization Through Transformation
As in previous problems with cyclic coordinates, the transformation (ψ̇, ϕ̇) → (ψ̇ +
ν, ϕ̇ + ν) changes ( p, q, r ) to ( p + νγ1 , q + νγ2 , r + ν(γ3 + 1)) and renders the pair
(l, V ) to the new pair
l = l+ν(Aγ1 , Aγ2 , C(γ3 + 1)),

ν2
V = V (θ, ψ − ϕ) + ν( f − l · γ − l3 ) − [A(γ22 + γ12 ) + C(γ3 + 1)2 ].
2
(14.102)
The two systems described by the (l,V ) and (l ,V ) are mathematically equivalent.
From the physical point of view, the latter system involves several changes compared
to the first. An interesting consequence of this equivalence is that any integrable case
of (14.84) always generates a more general integrable case of (14.102) containing the
additional function ν, which can be chosen according to the free parameters in the
structure of the potential. Of special interest is the few parameter generalization intro-
duced in Chap. 11 and applied above to cases of other types of symmetry. Listed below
are the two generalized cases, for which the integrals of motion are explicitly known
in terms of the Euler–Poisson variables. Other basic cases can be generalized in the
same way.
The linear integral I2 corresponds to the cyclic variable ψ + ϕ. The integral I3 ,
though became more complicated, is still of fourth degree.
The explicit solution of this case can be deduced as described in the above sections
from that of the basic version n = n 1 = n 2 = n 3 = 0. Let the last solution be ψ =
(t), θ = (t), ϕ = (t). The solution of the full generalized case can be readily
written as ψ(t) = (t) − ν(t)dt, θ(t) = (t), ϕ(t) = (t) − ν(t)dt, where
ν(t) = n + (1 − cos (t))(n 1 cos((t) − (t)) + n 2 sin((t) − (t))). Note that
the palpable coordinates θ, ψ − ϕ are not affected by the extra parameters n, n 1 , n 2
and n 3 .
The next case is a generalization of case 2 of Table 14.5.
2 Yehia [407] (2001)

n = n 1 = n 2 = 0. Special case of [31] (1986).
A = B = C,
ν = n + n 1 (α1 − β2 ) + n 2 γ3 ,
V = A{a(α1 − β2 ) + cγ3 − (1 + γ3 )[n + n 1 (α1 − β2 ) + n 2 γ3 ]2 }
μ1 = A[−nγ1 + 2n 1 (β1 γ2 − α1 γ1 ) − n 2 γ1 (1 + 2γ3 )],
μ2 = A[−nγ2 + 2n 1 (β2 γ2 − α2 γ1 ) − n 2 γ1 (1 + 2γ3 )],
μ3 = A[n(1 − γ3 ) − 2n 1 γ3 (α1 − β2 ) + n 2 (1 − γ3 )(1 + 2γ3 )],
I2 = ω · γ + r + 2(1 + γ3 )[n + n 1 (α1 − β2 ) + n 2 γ3 ],
I3 = (a − n 1 I2 )[( p + νγ1 )(ω · α+να3 ) − (q + νγ2 )(ω · β+νβ3 )]
+(c − n 2 I2 )[r + ν(γ3 + 1)][ω · γ + ν(1 + γ3 )]
−2(a − n 1 I2 )(c − n 2 I2 )(α1 − β2 ) − 2(a − n 1 I2 )2 γ3 .
@seismicisolation
@seismicisolation
Table 14.6 Generalized integrable cases

1 Yehia [411] (2003),
Yehia [407] (2001). n 3 = 0.
A = B = 2C,
ν = n + n 1 (α1 − β2 ) + n 2 (α2 + β1 ) + 1+γ3 ,
n3
l = C(2νγ1 , 2νγ2 , κ + ν(1 + γ3 )),

μ1 /C = −nγ1 + n 1 (α2 γ2 − 2α1 γ1 + 3β1 γ2 ) + n 2 (β2 γ2 − 2β1 γ1 − 3α1 γ2 ),
μ2 /C = −nγ2 + n 1 (−β1 γ1 + 2β2 γ2 − 3α2 γ1 ) + n 2 (α1 γ1 − 2α2 γ2 − 3β2 γ1 ),
μ3 /C = κ − n 3 + n(1 − 3γ3 ) + n 1 [β3 γ2 − α3 γ1 + 4γ3 (β2 − α1 )]
−n 2 [α3 γ2 + β3 γ1 + 4γ3 (α2 + β1 )],
a3 ν2
V = C{a1 (α1 − β2 ) + a2 (α2 + β1 ) + 1+γ3 − 2 [2(γ2
2 + γ12 ) + (γ3 + 1)2 ]
−κ(1 + γ3 )[n + n 1 (α1 − β2 ) + n 2 (α2 + β1 )]}.
I2 = 2[( p + νγ1 )γ1 + (q + νγ2 )γ2 ]
+(1 + γ3 ){r + (γ3 + 1)[n + n 1 (α1 − β2 ) + n 2 (α2 + β1 )] + κ + n 3 }
I3 = [( p + νγ1 )2 − (q + νγ2 )2 − (a1 − n 1 I2 )(α1 + β2 ) + (a2 − n 2 I2 )(α2 − β1 )]2
+[2( p + νγ1 )(q + νγ2 ) − (a2 − n 2 I2 )(α1 + β2 ) − (a1 − n 1 I2 )(α2 − β1 )]2
+2κ[r − κ + ν(1 + γ3 )][( p + νγ1 )2 + (q + νγ2 )2 ]
−4κ{( p + νγ1 )[(a1 − n 1 I2 )α3 + (a2 − n 2 I2 )β3 ]
+(q + νγ2 )[(a2 − n 2 I2 )α3 − (a1 − n 1 I2 )β3 ]}
3 −n 3 I2 )
+ 2(a1+γ 3
{( p + νγ1 )2 + (q + νγ2 )2 + κ[r + ν(1 + γ3 ) − κ] + (a2(1+γ
3 −n 3 I2 )
3)
}
This case contains the extra parameters n, n 1 , n 2 . Note that the integral I3 is a poly-
nomial of the third degree in the angular velocities. It reduces to a quadratic form when
either n 1 : n 2 :: a : c or n 1 = n 2 = 0. As the case n = n 1 = n 2 = 0 reduces to a spe-
cial version of Bogoyavlensky’s case, the general solution of the last full case can also
be expressed in terms of the hyper-elliptic Theta functions (Table 14.6).
@seismicisolation
@seismicisolation
Appendix A
Some Useful Identities
(1) The following identity has proved very useful in reshaping expressions in several
places of this book, especially, in Chap. 10, dealing with the regular precession
transformation, and Chap. 11, dealing with the general precession transforma-
tion.
Let a, b be two three-dimensional vectors, K arbitrary 3 × 3 matrix and δ the
unit matrix, then the following identity holds:
(a × b)K + aK T ×b = a × (b[tr (K)δ−KT ]). (A.1)
The proof is straightforward.

The special case when K is a symmetric matrix, the above identity becomes
(a × b)K + aK × b = a × (b[tr (K)δ−K]). (A.2)
(2) Let a, b, c be three three-dimensional vectors, A nonsingular symmetric 3 × 3

matrix. Then
a · (bA × cA) = a(ad j (A)) · (b × c) (A.3)
where, for nonsingular A, ad j (A) ≡ det (A)A−1 .

The proof is straightforward, if one takes A in its diagonal form.

https://doi.org/10.1007/978-3-030-96336-1
@seismicisolation
@seismicisolation
Appendix B
Kowalevski’s Case: Appelrot’s Four Classes
of Simple Motions
A full account of the Appelrot classes and their degeneracies would require too much
space, not available in the present context. In this appendix, we give an outline of
those classes, based on the original work of Appelrot and on the review in [108].
Following Appelrot, to have less parameters in the solution, as a = Mgx0 /C = 0,
one can choose units so that a = 1.
Conditions for each of those classes are given in the following table (F is defined
in (4.59)).
The condition for class I V is somewhat complicated, and it may be useful for
some purposes to express it in the parametric form
f = −2hρ + 2ρ3 , K = 1 − 2hρ2 + 3ρ4 . (B.1)
In the three-dimensional space { f, k, h} or, more precisely, in the half k ≥ 0 of that

space, those conditions represent certain surfaces on which coalescence of roots of
the polynomial occurs. Those surfaces, called bifurcation surfaces, separate zones
of space, which correspond to ultra-elliptic solutions of one and the same type and
also the same qualitative character of motion.
M. P. Kharlamov studied the question for which values of the first integrals of
motion, i.e. for which parameters { f, k, h}, those integrals become dependent in the
sense that the rank of the Jacobian matrix
∂(I1 , I2 , I3 , I4 )
∂(ω, γ)
falls under 4, and then, according to the theory of integrable systems, determine
singular (degenerate) Liouville tori. It turned out that those values define exactly the
same surfaces listed in Table B.1, which correspond to the four Appelrot classes of
simplest motions.
A quite detailed description of the integral manifolds of Kowalevski’s case and
their bifurcation are presented in [183]. As usual in such complicated dynamical
problems, bifurcation diagrams for the present case are illustrated in the plane of

https://doi.org/10.1007/978-3-030-96336-1
@seismicisolation
@seismicisolation
418 Appendix B: Kowalevski’s Case: Appelrot’s Four Classes of Simple Motions
Table B.1 Appelrot classes of simplest motions

Class Conditions
I k = 0 (e4 = e5 )
II f 2 − 2 h − 2 k = 0 (F(e5 ) = 0)
III f 2 − 2 h + 2 k = 0 (F(e4 ) = 0), k f 2 ≤ a 2
IV [ 27
2 f − h(h + 9 − 9k )] − (h − 3 + 3k ) = 0, (F(s) has a double root)
2 2 2 2 2 2 3
two parameters k, h for fixed values for areas parameter f. Those are sections of the
bifurcation surfaces by selected planes f = const. Full presentation of those results
is not possible in view of the scope of the present book.
The nature of the solution in those critical cases depends on the way in which
roots of the polynomial coalesce together. When one of the admissible intervals
shrinks to a point at the double root r1 (say), s1 is a constant s1 = r1 at all times. The
other variable s2 varies on another admissible interval as an elliptic function of time.
The term particularly remarkable motion was coined by Appelrot to characterize
motions on which s1 or s2 keeps a constant value all the time, while the other one
changes on an interval as an elliptic function of time or one of its degenerate forms.
Appelrot classes also contain motions for which s1 and s2 are expressible in terms
of elliptic functions or their degenerations. Those correspond to the coalescence of
two roots bounding two different admissible intervals. Asymptotic motions usually
appear in this class.
B.1 The First Class of Simplest Motions (Known as

Delone’s Case)
This case, characterized by the condition k = 0, was first investigated by Delone [60].
As it contains only “particularly remarkable” motions, the solution is expressed in
terms of one variable, which is expressible as an elliptic function of time. The fourth
integral of motion (4.57) becomes
( p 2 − q 2 − aγ1 )2 + (2 pq − aγ2 )2 = 0. (B.2)
As all the variables are real, this Eq. (B.2) holds only if
p 2 − q 2 − aγ1 = 0, (B.3)
2 pq − aγ2 = 0. (B.4)
We should note that we have two integrals of the motion that do not contain arbitrary
constants instead of the Kowalevski’s case which is the general integrable case. The
three classical integrals of motion are
2( p 2 + q 2 ) + r 2 + 2aγ1 = 2h, (B.5)
@seismicisolation
@seismicisolation
Appendix B: Kowalevski’s Case: Appelrot’s Four Classes of Simple Motions 419
2( pγ1 + qγ2 ) + r γ3 = f, (B.6)

γ12 + γ22 + γ32 = 1. (B.7)
Thus, we have five Eqs. (B.3)–(B.7) in six variables, and they can be solved to express
five of those variables in terms of the remaining one. The two Eqs. (B.3)–(B.4) give
p2 − q 2 2 pq
γ1 = , γ2 = . (B.8)
a a
Inserting the two expressions (B.8) in the geometric integral (B.7), we obtain

( p 2 + q 2 )2
γ3 = ε 1 − , ε = ±1. (B.9)
a2
Substituting in the energy integral, we get

r = δ 2(h − 2 p 2 ), δ = ±1. (B.10)
Finally, Eq. (B.6) can be written as
r γ3 = f − 2( pγ1 + qγ2 ).
Squaring both sides and using the above expressions, we get
2h( p 2 + q 2 )2 − 4 f ap( p 2 + q 2 ) + a 2 ( f 2 − 2h + 4 p 2 ) = 0.
The last equation yields

a
q = ± − p 2 + [2 f p ± 2( f 2 − 2h)(2 p 2 − h)]. (B.11)
h
Thus, all the variables are expressed in terms of one variable p. Inserting the obtained
results in Eq. (4.52), we arrive at the differential equation

a
2 ṗ = ± 2h − 4 p − p 2 + [2 f p ± 2( f 2 − 2h)(2 p 2 − h)],
2
h
which, on separating the variables, gives the relation

dp
t = t0 ± 2 . (B.12)
2h − 4 p 2 − p 2 + a
h
[2 fp ± 2( f 2 − 2h)(2 p 2 − h)]
Performing the transformation

√ x
p = − 2h (B.13)
1 + x2
@seismicisolation
@seismicisolation
reduces the relation (B.12) to

t − t0 dx
=± . (B.14)
2 f2
a 1− 2h
(1 − x 4) − 2hx 2 − 2
h
f ax(1 + x 2)
The integral in the last expression is an elliptic integral and can be inverted giving x
as an elliptic function of the time t.
B.1.1 A Special Case
Appelrot noted also the following type of motion which deserves special considera-
tion.
Let the third body axis z at some moment of time t0 occupy a vertical position, i.e.
γ3 = ±1. From the geometric integral, it follows that at the same moment γ1 = γ2 =
0 and, consequently, from (B.8), p = q = 0. This means that at t = t0 , the motion of
the body is momentarily similar to that of a sleeping gyroscope with apex directed
upwards or downwards. Let the angular velocity component r at that moment have
the value r0 . From (4.54), we get
1 2
h= r , f = ±r0 , (B.15)
2 0
and hence
dp
t =2 . (B.16)
r0 − 4 p 2 − p 2 ±
2 4a
r02
p]
It can be shown using the above formulas that the subsequent motion is periodic,
i.e. the body returns periodically to its initial state of motion with vertical z-axis
position. When r02 < 8a, the body periodically passes through the two opposite
vertical positions, while for r02 > 8a the z-axis passes only through one of the two
vertical positions and gets from it a maximal inclination at an angle sin−1 8a
r02
.
The special case under conditions (B.15) enjoys another property of definite inter-
est. Using (3.32) and (B.8)–(B.11), one can express the precession velocity in the
form
ψ̇ =
ap
=
p2 + q 2
r0
= . (B.17)
2
The projection of the z-axis on the horizontal plane passing through the fixed point
rotates about that point with uniform speed r20 .
@seismicisolation
@seismicisolation
B.1.2 A Case of Rational Solution
In [285] (1896), Mlodz’evsky constructed the following rare and interesting solution,
in which all the six Euler–Poisson variables are rational functions. We present it in
a slightly modified form to conform with our choice of the parameters in this book.
√ √
ε u4 + 6 u2 − 3 ε 2u ε 2 u 4 − 6 u 2 − 15
p= ,q = 8 4 ,r = − ,
u4 + 2 u2 + 9 u + 2 u2 + 9 u4 + 2 u2 + 9
(B.18)
√ 8
3 u + 12 u + 30 u − 164 u + 9
6 4 2
γ1 = 1/3 2 ,
u4 + 2 u2 + 9
√ 4 √
3 u + 6 u 2 − 3 2u
γ2 = 16/3 2 ,
u4 + 2 u2 + 9
√ √ 4
3 2 u − 2 u 2 − 11
γ3 = −1/3 , (B.19)
u4 + 2 u2 + 9
where ε = √413 , u = √4t 3 . This solution is valid on the integral level k = 0, h =

√
3, f = 33/4
4
. It asymptotically approaches, as t → ±∞, the uniform rotation
√
ω = ε(1, 0, − 2)

1 2
γ = ( √ , 0, − ), (B.20)
3 3
√
i.e. a permanent rotation with angular velocity ε 3 about the vertical unit vector γ.
Figure B.1 shows the trajectory of the apex traversed by the upward vertical during
the motion of the body on the Poisson sphere fixed in it in the direction P Q R, P being
Fig. B.1 The path of the

apex (Hidden line dotted)
@seismicisolation
@seismicisolation
its position corresponding to uniform rotation (B.20). From the time-reversibility of

motion in a potential field it follows that another motion exists, in which the apex
traverses its√path in the opposite sense P R Q. This happens on the integral level
k = 0, h = 3, f = − 33/4 4
.
B.2 The Second Class of Simplest Motions
This class, under the condition h = f 2 /2 − k, like the first one, consists of par-
ticularly remarkable solutions. The special case f = 0 is much simpler. It can be
shown that p = r = γ2 = 0. This characterizes the pendulum-like motion of the
body around its y-axis which occupies a fixed horizontal position.
In the generic case f = 0, the solution can be written as
1 μ
p= (μ sin u − 1), q = cos u, r = 2u̇,
f f
2 2 2
γ1 = 2 sin u(μ − sin u), γ2 = 2 cos u(μ − sin u), γ3 = u̇, (B.21)
f f f

where μ = 1 + k f 2 and the auxiliary variable u is related to time by the relation

du
t =± . (B.22)
f2
4
− 1
f2
(μ − sin u)2
Using the substitution tan u2 = x, one can invert the last relation in terms of elliptic
functions of time.
Exercise: Show that in all motions of Appelrot’s second class the body precesses
about the fixed point with a uniform rate ψ̇ = 2f .
B.3 The Third Class of Simplest Motions
For this class, under the condition
h = f 2 /2 + k, (B.23)
two types of solutions exist.

The Generic Solution
The solution of the equations of motion can be written in terms of two auxiliary
variables, in the form:
@seismicisolation
@seismicisolation
√ 1
p= k sin u cos v − f cos u,
√ 2
q= k sin u sin v − u̇,
√
r = −2 k cos u cos v − f sin u,
√ √ f2
γ1 = −2 k u̇ sin u sin v + cos u(1 + f k sin u cos v) − cos2 u( + 2k sin2 v),
√ √ 2
γ2 = (2 k sin u cos v − f cos u)u̇ + cos u sin v( f k sin u + 2k cos u cos v),
√ √
γ3 = 2k sin u cos u − 2 k cos u sin v u̇ − sin u − f k cos2 u cos v, (B.24)
where u satisfies the equation
f2
u̇ 2 = cos u[1 − (k + ) cos u], (B.25)
4
and hence can be expressed as an elliptic function of time, and v is determined from
√ f
v̇ = k cos u cos v + sin u. (B.26)
2
The Particularly Remarkable Solution

It can be checked that under the condition (B.23), the equations of motion admit a
solution identical with that given for the second class by
formulas (B.21)–(B.22), but
with the change of the sign of k, so that this time μ = 1 − k f 2 . This particularly
remarkable solution is valid only under the condition k f 2 < 1.
B.4 The Fourth Class of Simplest Motions
Like in the third class, here we also have particularly remarkable motions and generic
motions. One of the first turned out to be a special case of the periodic motion
presented in Sect. 8.5 corresponding to Kowalevski’s configuration A = B = 2C.
The generic motion is partially studied in [9]. More detailed analysis of those motions
was performed in [66] (see also [67]), by means of transforming the equations of
motion to a new set of variables
p = ρ − x/M,
q = −y/M,
r = 2z + 4xγ/M,
γ1 = 2α − 4(x 2 − y 2 )γ 2 /M 2 + 2ρx(2ρx − 1)/M,
γ2 = 2β − 8x yγ 2 /M 2 − 2ρy(2ρx − 1)/M,
γ3 = 2(2ρx − 1)γ/M + 2ρz. (B.27)
@seismicisolation
@seismicisolation
Using this transformation and the integrals of motion (4.54), (4.57), the equations of
motion (4.52), (4.53) can be reduced to a somewhat simpler system whose solution
can be written in the following form.
The variable z is determined from the relation

dz 1 1
t= , F(z) = − ρ2 z 2 − [z 2 + (ρ2 − h)]2 , (B.28)
F(z) 4 2
and α, β by the expressions
1
α = z 2 + (ρ2 − h), β = F(z), (B.29)
2
provided ρ and h are determined from two conditions P(ρ) = P (ρ) = 0, P =

−ρ4 + 2hρ2 − 2 f ρ + 1 − k 2 .
In the non-degenerate cases, the variable z oscillates between two real roots of
F(z), so that z, α, β are elliptic functions of time. The remaining variables are given
by the expressions

y = η (z),
ρ βη z
x = −[ +√ + (z)η̇],
2L 2 (z) L2
ρz 1 2
γ= + (ρ + α) (z)η̇, (B.30)
2L 2 L2
where (z) = z 2 − L 2 , η is an auxiliary variable satisfying the relation

dη dt
+ = const, (B.31)
L1 − L2 L3 η2 (z(t))
and
1
L1 = (2h − 2ρ2 ),
16
1
L 2 = (2h − 6ρ2 ),
4
1
L 3 = 4ρ2 L 1 − . (B.32)
4
Classification of motions corresponding to the constructed solution according to the
values of parameters L 1,2,3 was performed in [66] (see also [67]), into eight cases
(I )L 1 > 0, L 2 > 0, L 3 > 0,

(I I )L 1 > 0, L 2 < 0, L 3 < 0,
(I I I )L 1 > 0, L 2 < 0, L 3 > 0,
@seismicisolation
@seismicisolation
(I V )L 1 > 0, L 2 > 0, L 3 < 0,

(V )L 1 < 0,
(V I )L 1 = 0,
(V I I )L 2 = 0,
(V I I I )L 3 = 0. (B.33)
A simplified particular version of this solution is also presented in [41].
B.5 Intersection of the Four Classes
It can be shown that under the conditions
k = 0, h = 2, f = 2, (B.34)
common to the four classes, the polynomial in (4.59) three equal roots = 2 and two
roots = 1.
(s) = (1 − s)2 (2 − s)3 . (B.35)
But it is simpler, since this is a particular case of Delone’s solution, to use formulas
of Sect. B.1 to obtain for p the equation
ṗ 2 = p(1 − p)2 (1 + p), (B.36)
so that we can write the generic solution of the equations of motion in the form
√ √ √2t √ √ √
((e 2t
− 1)2 , ±2 2(e − 1)et/ 2 , ±8 (e 2t + 1)et/ 2 )
ω= √ √ ,
e2 2t + 6 e 2t + 1
√ √ √ √ √ √
(e 2t
− 1)2 (e − 10 e 2t + 1)
2 2t
2(e 2t − 1)3 et/ 2
γ=( √ √ , ±4 √ √ ,
(e2 2t + 6 e 2t + 1)2 (e2 2t + 6 e 2t + 1)2
√ √
(e 2t
+ 1)et/ 2
±4 √ √ ). (B.37)
e2 2t + 6e 2t +1
Here, we note that q, γ2 are odd functions in t and the other four quantities
are even. This solution as t → ±∞ asymptotically approaches the solution ω =
(1, 0, 0), γ=(1, 0, 0), which corresponds to the solution p = 1 of Eq. (B.36). That
is a uniform rotation about the first axis (bearing the centre of mass of the body),
while this axis preserves vertical upward position.
Figure B.2 depicts the path P Q RS P of the apex of γ, corresponding to the
positive sign in (B.37), on the Poisson sphere. It begins at t = −∞ at P and ends
also at P at t = ∞. At t = 0 the z-axis passes through the vertical position. The trace
@seismicisolation
@seismicisolation
Fig. B.2 The path traced by

the apex of γ on the Poisson
sphere
of the apex for the second solution, corresponding to the negative sign, is a mirror
reflection of the first in the x y-plane.
Note that, due to the invariance of the equations of motion under the change
(ω, γ,t) → (−ω, γ, −t), two other solutions exist for the parameter values
k = 0, h = 2, f = −2.
In one of them, the apex traces the same curve in the retrograde sense P S R Q P with
reversed angular velocity. The other is interpreted similarly.
@seismicisolation
@seismicisolation
Appendix C
Particularly Simple Classes of Motions in
Goryachev–Chaplygin’s Case
Just as in the case of Kowalevski, the analytical solution of the equations of motion
in the generic Goryachev–Chaplygin case does not offer any help in imagining or
simulating the motion of the body. This situation gives high importance to the study
of the extremal cases of motion, when the six-degree polynomials in (4.77) have
repeated roots. For simplicity and without loss of generality, we set C = 1, a = 1.
The discriminant equation is the same for both polynomials
G 2 [2(E − 1)3 − 27G 2 ][2(E + 1)3 − 27G 2 ] = 0. (C.1)
Its solution (the bifurcation diagram) can be represented in the plane of parameters
(G, E) by the following three parametric equations:
I G = 0, E > −1,
II G = 21 σ 3 , E = 1 + 23 σ 2 , σ ∈ (−∞, ∞),
III G = 21 σ 3 , E = −1 + 23 σ 2 , σ ∈ (−∞, ∞).
Motions corresponding to those three branches constitute for the case of

Goryachev–Chaplygin the analog of Appelrot classes for Kowalevski’s case. The
bifurcation diagram is displayed graphically in Fig. C.1.
The (G, E) plane is divided by the three branches into 5 open regions. The motion
is impossible in the region 0 under curve III. Each of the symmetric regions 1 , 2
is composed of two components separated by line I. In each of the two regions, the
number of real roots of the under-root polynomials do not change and, consequently,
the qualitative character of the motion does not change. The bifurcation diagram was
discussed in [183] and in a slightly modified way in [166].
We shall write the solution on the two boundaries II and III of the regions of
hyper-elliptic solutions. On those boundaries the solution is expressible in elliptic
functions of time. As usual on those boundaries, it is simpler to ignore the Eqs. (4.77)
and try to find a solution directly from the equations of motion (4.64) under relevant
conditions on the parameters of the system.

https://doi.org/10.1007/978-3-030-96336-1
@seismicisolation
@seismicisolation
428 Appendix C: Particularly Simple Classes of Motions in Goryachev–Chaplygin’s Case
Fig. C.1 Regions of

possible motion 1 and 2 ,
each composed of two
components separated by the
line I
C.1 Solution on the Boundary I (The Case of Goryachev)
C.1.0.1 The Subcase of Goryachev
This is the special version of this case that was first found by Goryachev [115] (1900)
and immediately generalized to its full just presented form by Chaplygin [52] (1901).
This version is characterized by the pair of invariant relations
√
p 2 + q 2 − bp 2/3 = 0, 3
pγ3 − br = 0, (C.2)
where b is an arbitrary parameter. In fact, this parameter is an integration constant

that can be obtained by considering the following in virtue of (4.64)
d p2 + q 2 q
( ) = − 5/3 [r ( p 2 + q 2 ) − pγ3 ]. (C.3)
dt p 2/3 2p
Comparing the right-hand side with (4.65), we conclude that the quantity between
brackets becomes an integral of motion
p2 + q 2
=b
p 2/3
provided
r ( p 2 + q 2 ) − pγ3 = G = 0, (C.4)
and from the last two relations the second invariant relation in (C.2) follows. Thus,
the case G = 0 is exactly Goryachev’s version of the Goryachev–Chaplygin case.
@seismicisolation
@seismicisolation
Appendix C: Particularly Simple Classes of Motions in Goryachev–Chaplygin’s Case 429
C.1.1 The Solution
Full analytical solution for the subcase of Goryachev was first obtained in Chaplygin’s
work [52] using the above separation variables (4.77). The solution expressed in
terms of Weierstrass’ elliptic functions is not so simple to reveal properties of the
motion. Somewhat simpler reduction to quadratures was derived by Dokshevich [67].
A more transparent and finalized explicit solution was found by Gashenenko [107].
This will be written here, divided into three cases, according to the values of the
energy parameter E.
C.1.1.1 When E ∈ [−1, 1)
We have

cn3 τ
p = 2b3/4 2s13 ,
(1 + s12 cn4 τ )3/2
cnτ (1 − s12 cn4 τ )
q = b3/4 2s1 ,
(1 + s12 cn4 τ )3/2
4s1 snτ cnτ dnτ
r =− ,
1 + s12 cn4 τ
1
γ1 = [E(1 + s14 cn8 τ − 6s12 cn4 τ ) − 4s1 cn2 τ ( 1 + 4b3 − E 2 − b3/2 )
(1 + s1 cn τ )
2 4 2

+4s13 cn6 τ ( 1 + 4b3 − E 2 + b3/2 )],
−1
γ2 = [ 1 + 4b3 − E 2 (1 + s14 cn8 τ − 6s12 cn4 τ )
(1 + s12 cn4 τ )2
+4Es1 cn2 τ (1 − s12 cn4 τ ) − 2b3/2 (1 − s14 cn8 τ )],
snτ dnτ
γ3 = −2b3/4 2s1 , (C.5)
(1 + s12 cn4 τ )1/2
where
E +1 1
s1 = √ ,τ = (t − t0 ) (C.6)
1+ 4b3 − E2 + 2b3/2 2
and the modulus of the elliptic functions

E +1
ν= . (C.7)
2
The last does not depend on the parameter b and hence the period of the solution
is independent of b. Note that when the energy parameter takes its lowest value
E = −1, then the solution renders to the lower equilibrium position:
@seismicisolation
@seismicisolation
p = q = r = 0, γ = (−1, 0, 0). (C.8)
C.1.1.2 When E > 1
The above formulas become

dn3 τ
p = 2b3/4 2s13 ,
(1 + s12 dn4 τ )3/2
dnτ (1 − s12 dn4 τ )
q = b3/4 2s1 ,
(1 + s12 dn4 τ )3/2
4s1 snτ cnτ dnτ
r =− ,
1 + s12 dn4 τ
1
γ1 = [E(1 + s 4
1 dn 8
τ − 6s 2
1 dn 4
τ ) − 4s 1 dn 2
τ ( 1 + 4b3 − E 2 − b3/2 )
(1 + s12 dn4 τ )2

+4s13 dn6 τ ( 1 + 4b3 − E 2 + b3/2 )],
−1
γ2 = [ 1 + 4b3 − E 2 (1 + s14 dn8 τ − 6s12 dn4 τ )
(1 + s1 dn τ )
2 4 2
+4Es1 dn2 τ (1 − s12 dn4 τ ) − 2b3/2 (1 − s14 dn8 τ )],

snτ cnτ
γ3 = −2b3/4 2s1 , (C.9)
(1 + s12 dn4 τ )1/2

where s1 is the same as above, the modulus of the elliptic functions k = 2
E+1
= 1/ν
and τ = 1
2k
(t
− t0 ).
We note that the two classes of motion are periodic and are spanning between
two types of pendulum-like motions of the body, which correspond to the values
b = 0, b = ∞. Those are motions in which the centre of mass describes a pendu-
lum motion in the x y(or x z)-plane that takes a vertical position. For pendulum-like
motions in general, see Chap. 8, Sect. 8.2.
When E = 1: For this class of motions s1 = 2b13/2 , and the solution can be obtained
as a limiting case from either of (C.5) or (C.9) as ν = k = 1. Elliptic functions are
replaced by hyperbolic functions. It can be also verified that the solution is asymptotic
as t → ±∞ to the upper equilibrium position characterized by
p = q = r = 0, γ = (1, 0, 0). (C.10)
C.2 Solution on Boundary II
For this solution, expressions (C.13) and Eq. (C.14) are replaced by [99]:
@seismicisolation
@seismicisolation
Appendix C: Particularly Simple Classes of Motions in Goryachev–Chaplygin’s Case 431
1
p= (σ + r ) σ(r − σ),
4
1
q = −σ R1 (r ),
4
1
γ1 = 1 − (r − σ)(2σ + r ),
2
1
γ2 =− (σ − r )R1 (r ),
2
γ3 = σ(r − σ), (C.11)
1
ṙ = (σ − r )R1 (r ), R1 (r ) = −4(σ + r ) + (2σ + r )2 (r − σ). (C.12)
2
Note that the equations and the integrals of motion in the Goryachev–Chaplygin case
are invariant under the change of signs of the part of variables p, q, γ3 and hence
another solution is obtained from (C.11) by changing those signs.
The cubic R1 has the discriminant
4 16
= −4[σ 4 + σ 2 + ] < 0
3 27
and thus it always has three real roots. It is sufficient to consider the case σ > 0. One
can also notice that R1 (∞) > 0, R1 (σ) < 0, R1 (0) < 0, R1 (−2σ) > 0, so that R1
has one positive root r ∗ (say) (r ∗ > σ) and two negative ones. To have all physical
quantities real, the variable r changes on the interval [σ, r ∗ ], where it is expressible
as an elliptic function of time. In the limit as s → 0, this interval shrinks to zero,
E = 1 and the solution (C.11) becomes
ω = 0, γ = (1, 0, 0),
which represents the upper equilibrium position. That is the intersection of boundaries
I and II.
C.3 Solution on Boundary III
Five of the Euler–Poisson variables are expressed in terms of r [124]:
1
p= (2σ + r ) σ(σ − r ),
4
1
q=− σ R(r ),
4
1
γ1 = −1 + (σ − r )(2σ + r ),
2
@seismicisolation
@seismicisolation
1
γ2 = − (σ − r )R(r ),
2
γ3 = − σ(σ − r ), (C.13)
while r is determined, as an elliptic function of time, from the differential equation
1
ṙ = (σ − r )R(r ), R(r ) = 4(σ + r ) − (2σ + r )2 (σ − r ). (C.14)
2
Another solution can be obtained by changing the signs of p, q and γ3 . The cubic
polynomial R has the discriminant
2 4
= 4[(σ 2 − )2 + ] > 0,
3 27
so that it has only one real root. Let us denote this root by r1 . We have 3 cases:
(1) When σ > 0 then r1< σ and r varies on the interval [r1 , σ]. r is elliptic
σ
function with period T = 4 r1 √(σ−r dr
)R(r )
.
(2) When σ < 0 then r1> σ and r varies on the interval [σ, r1 ]. r is elliptic
r
function with period T = 4 σ 1 √(σ−r dr
)R(r )
.
(3) When σ = 0 then r1 = 0 and (C.14) takes the form
1
ṙ = (−r 2 )(4 + r 2 ). (C.15)
2
This has only the solution r = 0, which leads only to the solution ω =(0, 0, 0), γ =
(−1, 0, 0), which characterizes the lower equilibrium position of the body.
@seismicisolation
@seismicisolation
Appendix D
Gyrostatic Generalization of the Appelrot
Classes
The question arises, whether it is possible to generalize the four Appelrot classes of
motion, which we have described in Sect. 3.3, to construct analogs of them involving
non-zero gyrostatic momentum, which admits simpler solutions than the generic
case. This was investigated by M. Kharlamov. In view of the absence of variable
separation for Yehia’s case, Kharlamov searched for the critical set of parameters in
the space of parameters of the problem {h, f, k, κ}, on which the four integrals of
motion become functionally dependent. It turned out that all solutions corresponding
to the critical set were already isolated and shown to be expressible in terms of elliptic
functions long before the complementary integral I4 (in case 3 of Table 5.1 of Chap.
5) was discovered. This came as a result of the use of the method of invariant relations
as we shall see below.
(1) The first class of Appelrot motions in Kowalevski’s case (case of Delone) com-
pletely disappears in the gyrostatic analog. The integral I4(K ) = ( p 2 − q 2 − a1 γ1 +
a2 γ2 )2 + (2 pq − a1 γ2 − a2 γ1 )2 = k 2 is a sum of two squared terms and when k = 0
each of those terms is zero. The integral of motion is replaced by two invariant rela-
tions that determine Delone’s case. In the presence of the gyrostatic momentum κ,
the integral
I4 = I4(K ) + 2κ (r − κ) ( p 2 + q 2 ) − 4κγ3 (a1 p + a2 q)
contains terms of different structure and hence loses that property.

(2) The second and third classes:
The condition

2
K 2[ f 2 − 2 h − 4 K ] =

= κ2 {4κ6 (1 − h 2 ) − 4κ4 −2 f 2 h 2 − 4 h 3 + K h + 3 f 2 + 3 h

+κ2 [−16 h 4 − 16 f 2 h 3 + 4 f 4 − 32 K + 24 h 2 − 8 f 2 (K − 3/2) h
−12 f 4 + K 2 + 18 K − 27]

https://doi.org/10.1007/978-3-030-96336-1
@seismicisolation
@seismicisolation
434 Appendix D: Gyrostatic Generalization of the Appelrot Classes

+ (−16 K + 32) h 3 + 16 f 2 (K − 3) h 2 + −4 f 4 K + 24 f 4 + 20 K 2 − 36 K h

+2 f 2 −2 f 4 + K 2 + 9 K } (D.1)
generalizes cases II and III in Table 4.3 of Chap. 4 for κ = 0 and unifies them in one
case. A suitable parametrization of this condition is
f 2 = 2 h + κ2 − 4 κ2 s 2 − s −1 ,

K = −2 κ2 h − 1/2 κ2 − 2 s − κ4 + 1/4 s −2 . (D.2)
The corresponding solution of the equations of motion was described in [198]. It will
be written here in the form given in [181] and also in [184].
Let us fix a value f for the areas integral and let s = 0 be a constant. The equations
of motion (5.19) admit a solution
f
p=− − κρY,
2s
√
q= −ρ s R,
r = κ + 2κ X,
2κs X + f ρY
γ1 = − 2κ 2 Y 2 ,
2κ
√
γ2 = −2κY s R,
f X − 2κsρY
γ3 = , (D.3)
2κ
in which
f2 2κ 2
κ= + s 2 κ2 , ρ2 = 1 − ,
4 s
(cos σ, sin σ), ρ2 0
(X, Y ) = {
(cosh σ, i sinh σ), ρ2 < 0
1 κ f 2
R 2 = [(X + )2 + (ρY + ) − 1], (D.4)
2 κ 2sκ
and the relation with time is determined from the equation
σ̇ 2 = sgn(ρ2 )s R 2 , (D.5)
which can be solved in elliptic functions of time.

(3) The fourth class:
The condition on the parameters h, f, K for this class can be written in the fol-
lowing form, which reduces to (B.1) when κ = 0,
@seismicisolation
@seismicisolation
Appendix D: Gyrostatic Generalization of the Appelrot Classes 435
2 27 f 2 3
{[ − h h 2 − 9 K + 9 ]2 − h 2 + 3 K − 3 }
27 2
1
+ κ2 [+ f 2 12h 2 − 6hκ2 + κ4 (K − 1)2 κ2 − 4 h
4
+(K − 1) −1/8 κ6 + hκ4 − 3 h 2 κ2 + 4 h 3 − 9 f 2 ]
= 0. (D.6)
It is more convenient to use parametrization

f 2 = s 2 2h − κ2 − 2s ,
2
K = 1 + h − 1/2 κ2 − 4 s h − 1/2 κ2 + 3 s 2 . (D.7)
The particularly remarkable motion. The particular solution with two linear
relations (see Chap. 8, Sect. 8.15.1.4) in the problem of motion of a heavy gyrostat
generalizes the solution due to Bobylev and Steklov (Sect. 8.5 this chapter). If one
sets A = B = 2C and κ1 = κ2 = y0 = 0, then one is left with a subcase of the
Kowalevski–Yehia case. The solution of the equations of motion in this case can be
written in the form (see e.g. [184]):
p = p0 , q = 0,
1
γ1 = h − p02 − r 2 ,
2
γ2 = R(r ),
γ3 = p0 (κ − r ), (D.8)
where r is obtained by inverting the elliptic quadrature

dr
t= √ ,
R(r )
1
R(r ) = − r 4 + (h − 2 p02 )r 2 + 2κ p02 r + 1 − (h − p02 )2 − κ2 p02 , (D.9)
4
and we have, for simplicity, set units of measurement so that C = 1 and a1 = 1. The
solution can thus be expressed in terms of elliptic functions of time. This gives the
gyrostatic generalization of the fourth class of particularly remarkable motions of
Appelrot.
The Non-degenerate Solutions

Modifying the transformation of variables (B.27) introduced for Kowalevski’s case
by Dokshevich in [66], Gashenenko [101] used the transformation
p = ρ − x/M,
q = −y/M,
@seismicisolation
@seismicisolation
436 Appendix D: Gyrostatic Generalization of the Appelrot Classes
r = 2z + κ + 4xγ/M,
γ1 = 2α − 4(x 2 − y 2 )γ 2 /M 2 + 2ρx(2ρx − 1)/M − 4κγx/M,
γ2 = 2β − 8x yγ 2 /M 2 − 2ρy(2ρx − 1)/M − 4κγ y/M,
γ3 = 2(2ρx − 1)γ/M + 2ρz, (D.10)
where ρ and other constants of motion are subject to two conditions P(ρ) =
P (ρ) = 0, P = −ρ4 + (2h − κ2 )ρ2 − 2 f ρ + 1 − K .
Using this transformation, (D.7) and the integrals (5.20), the equations of motion
(5.19) can be reduced to a somewhat simpler system whose solution can be written
in the following form.
The variable z is determined from the relation

dz 1 κ 1
t= , F(z) = − ρ2 z 2 − [(z − )2 + (ρ2 − h)]2 , (D.11)
F(z) 4 2 2
and α, β by the expressions
κ 1
α = (z − )2 + (ρ2 − h),
2 2
β = F(z). (D.12)
In the non-degenerate cases, the variable z oscillates between two real roots of
F(z), so that z, α, β are elliptic functions of time. The remaining variables are given
by the expressions

y = η (z),
ρ βη 1
x = −[ +√ − (z − κ) (z)η̇],
2L 2 (z) L2
ρz κβη 1 2
γ= +√ + (ρ + α) (z)η̇, (D.13)
2L 2 (z) L2
where (z) = (z − κ)2 − L 2 , η is an auxiliary variable satisfying the relation

dη dt
+ = const, (D.14)
L1 − L2 L3 η2 (z(t))
and
1
L1 = (2h − 2ρ2 − κ2 ),
16
1
L 2 = (2h − 6ρ2 − κ2 ),
4
1
L 3 = 4(ρ2 + κ2 )L 1 − . (D.15)
4
@seismicisolation
@seismicisolation
Appendix D: Gyrostatic Generalization of the Appelrot Classes 437
Classification of motions corresponding to the constructed solution according to the

values of parameters L 1,2,3 was performed in [100, 101] into eight cases
(I )L 1 > 0, L 2 > 0, L 3 > 0,

(I I )L 1 > 0, L 2 < 0, L 3 < 0,
(I I I )L 1 > 0, L 2 < 0, L 3 > 0,
(I V )L 1 > 0, L 2 > 0, L 3 < 0,
(V )L 1 < 0,
(V I )L 1 = 0,
(V I I )L 2 = 0,
(V I I I )L 3 = 0. (D.16)
It is noted there that in cases I, I I doubly periodic are possible, in cases I I I, I V,

V I I, V I I I motions asymptotically tend to those described in (D.8)–(D.9), while in
cases V and V I motions different from (D.8)–(D.9) are impossible. The case of a
double root of F(z) is also considered. The effect of the presence of the gyrostatic
momentum on those motions was not considered.
@seismicisolation
@seismicisolation
Appendix E
The Conditional Case of Sretensky
Separation of variables in Sretensky’s case is obtained by inserting the expressions

(5.21) into the quadratures (4.77). The discriminant equation of the resulting poly-
nomial leads to three classes of critical sets [172]
I) G = 0,
2 κ κ2 3
II) = 27[G + κ(E − 1) − ( )3 ]2 − 8(E − 1 + ) = 0,
3 3 6
2 κ 32 κ2 3
III) ∗ = 27[G + κ(E + 1) − ( ) ] − 8(E + 1 + ) = 0.
3 3 6
(E.1)
Obviously, those generalize the three classes I, II, III given by (C.1) in Goryachev–
Chaplygin’s case and reduce to them when κ = 0.
Class I (G = 0) generalizes Goryachev’s case whose full solution was given
in Chap. 4, Sect. 4.1. M. Kharlamov has shown that when κ = 0, the analog of
Goryachev’s integral doesn’t exist and hence the method of solution of Goryachev’s
case cannot be generalized. As far as we know, the explicit solution of class I was
not considered.
Classes II and III ( = 0, ∗ = 0) were considered in [172], but full solution
is given only for class III. For this class, the parametrization
3 1 1
E = −1 + σ 2 + 2σκ + κ2 , G = σ 2 (1 + κ)
2 2 2
was used and formulas (C.13) and (C.14) are then replaced by
1
p= (2σ + κ + r ) σ(σ + κ − r ),
4
1
q=− σ R(r ),
4

https://doi.org/10.1007/978-3-030-96336-1
@seismicisolation
@seismicisolation
440 Appendix E: The Conditional Case of Sretensky
1
γ1 = −1 + (σ + κ − r )(2σ + κ + r ),
2
1
γ2 = − (σ + κ − r )R(r ),
2
γ3 = − σ(σ + κ − r ), (E.2)
and
1
ṙ = (σ + κ − r )R(r ), (E.3)
2
R(r ) = r 3 + (κ + 3σ)r 2 + (4 − 2σκ − κ2 )r + (σ + κ)[4 − (2σ + κ)2 ].
Note that another solution is valid that can be obtained from (E.2) by changing the
signs of p, q, γ3 .
Some details of this solution are discussed in [172]. We point out here only two
rare and quite interesting versions of the solution when all the variables are algebraic
functions of time. They occur when R(r ) has a triple root and require the value
κ = √23 . They can be written in the form
√ √ √ 2
6t t 2 + 9 2 2 3 t −6
( p, q, r ) = (± 1/9 3/2 , ± 3/2 , −4 ),
t2 + 3 t2 + 3 9 (t 2 + 3)
√ √
t 4 + 18 t 2 − 27 3t 2t
(γ1 , γ2 , γ3 ) = (1/3 2 , 8 2 , ∓2/3 √ 2 ). (E.4)
t2 + 3 t2 + 3 t +3
Those solutions represent two motions, which, as t → ±∞, asymptotically approach

the uniform rotations characterized by the solutions
√ √
6 4 3
( p, q, r )−∞ = (∓ , 0, − ),
9 9
1 2√
P = (γ1 , γ2 , γ3 )−∞ = ( , 0, ± 2), (E.5)
3 3
and approach as t → ∞ the values
√ √
6 4 3
( p, q, r )∞ = (± , 0, − ),
9 9
1 2√
Q = (γ1 , γ2 , γ3 )∞ = ( , 0, ∓ 2). (E.6)
3 3
In Fig. E.1, we visualize the motion corresponding to the last pair of solutions by
depicting the trajectory of the apex of the vertical upward unit vector γ during the
motion on the Poisson sphere. The apexes draw two halves of a Fig. 8 curve. The first
half (in red colour) begins from P(t = −∞), goes through R and ends at Q(t = ∞).
@seismicisolation
@seismicisolation
Appendix E: The Conditional Case of Sretensky 441
Fig. E.1 The trajectory of

the apex on Poisson’s sphere
(Hidden lines dotted)
The second (P P1 R Q, in blue colour) begins at Q(t = −∞) and through R ends at
P(t = ∞). Note that the motion is not time-reversible, due to the presence of the
gyrostatic momentum κ. On the other hand, since the equations of motion (5.5)–(5.6)
are invariant under the simultaneous change of signs of the quantities ω, κ, t, when
κ = − √23 two other motions exist, in which the apex traverses each of the two paths
described above in the retrograde sense.
@seismicisolation
@seismicisolation
Bibliography
1. Adler, M., & van Moerbeke, P. (1988). The Kowalewski and Hé non-Heiles motions as
Manakov geodesic flows on SO(4) - a two-dimensional family of Lax pairs. Communications
in Mathematical Physics, 113, 659–700.
2. Adler, M., van Moerbeke, P., & Vanhaeke, P. (2004). Algebraic integrability, Painlevé geom-
etry and Lie algebras. Berlin: Springer.
3. Agostinelli, C. (1935). Sopra alcuni integrali particolari delle equazioni del moto di un corpo
rigido pesante, intorno a un punto fisso. Acta Pontif. Acad. Sci. Novi Lyncaei, 88, 64–84.
4. Agostinelli, C. (1949). SuI moto di un corpo rigido pesante asimmetrico col baricentro apparte-
nente all’asse di uno dei piani cic1ici dell’ellissoide d’inerzia. Atti Sem. Mat. e Fis. Univ.
Modena, 3, 248–260.
5. Agostinelli, C. (1949). SuI moto intorno a un punto fisso di un corpo rigido pesante il cui
baricentro appartiene all’asse di uno dei piani ciclici dell’ellissoide d’inerzia. Ann. Mat. Pura
Appl. (Ser. 4), 30, 211–224.
6. Alfieri, L. (1954). Risoluzione di un problema comprendente quello di Staude. Univ. Roma.
1st. Naz. Alta Mat. Rend. Mat. e Appl., 12, 367–372.
7. Ampére, A. M. (1823). Sur quelques nouvelles propriété s des axes permanents de rotation des
corps et des plans directeurs de ces axes. A Paris, De L’Imprimerie Royale (1823). (Revised
edition of: Mé m. Acad. Sci. Paris, 5, 86–152 (1821–1822)).
8. De Angeli, E. (1937). Su alconi casi integrabili di movimento di un giroscopio asimmetrico
pesante. Rend. Acc. Naz. dei Linsei. Ser. 6, 25, 699–703.
9. Appell, P. (1911). Traité de Mécanique Rationelle. Paris.
10. Appelrot, G. G. (1892). About §1 of Kowalevski’s memoir “Sur le probl‘eme de la rotation
d’un corps solide autour d’un point fixe”. Mat. Sb., 16, 483–507.
11. Appelrot, G. G. (1940). Incompletely symmetric heavy gyroscopes. In S. A. Chaplygin & N.
I. Mertsalov (Eds.), Motion of a rigid body about a fixed point. Collection of papers in memory
of S.V. Kovalevskaya (pp. 61–156). Moscow-Leningrad: Academy of Sciences of USSR (In
Russian).
12. Aref, H., & Jones, S. W. (1993). Chaotic motion of a solid through ideal fluid. Physics of
Fluids A, 5, 3026–3028.
13. Arjanykh, I. S. (1954). On rotation equations of a rigid body about a fixed point. Dokl. AN
USSR, 97, 403–407.
14. Arkhangelsky, Yu. A. (1977). Analytical rigid body dynamics. Moscow: Nauka.
15. Bardin, B. S. (2007). Stability problem for pendulum-type motions of a rigid body in the
Goryachev–Chaplygin case. Izv. Ross. Akad. Nauk. Mekh. Tverd. Tela, 2, 14–21. English
translation: Mechanics of Solids, 42(2), 177–183 (2007).
https://doi.org/10.1007/978-3-030-96336-1
@seismicisolation
@seismicisolation
444 Bibliography
16. Bardin, B. S. (2010). On the orbital stability of pendulum-like motions of a rigid body in the
Bobylev–Steklov case. Regular and Chaotic Dynamics, 15(6), 702–714.
17. Bardin, B. S., Rudenko, T. V., & Savin, A. A. (2012). On the orbital stability of planar periodic
motions of a rigid body in the Bobylev–Steklov case. Regular and Chaotic Dynamics, 17(6),
533–546.
18. Basak, I. A. (2009). Explicit solution of the Zhukovski-Volterra gyrostat. Regular and Chaotic
Dynamics, 14, 223–236.
19. Basset, A. B. (1886). On the motion of a ring in an infinite liquid. Proceedings of the Cambridge
Philosophical Society, 6, 47–60.
20. Beletsky, V. V. (1957). Some problems of motion of a body in a Newtonian force field. Prikl.
Mat. Mekh., 21, 749–758.
21. Beletsky, V. V. (1965). The motion of a satellite about its center of mass. Moscow: Nauka.
22. Benvenuti, P., & Balli, R. (1974). Risolibilita per quadrature del problema del moto di un
solido soggitto a forze di potenza nulla. Atti. Acc. Naz. Lincei Rend. Cl. Sci. fis. math. e natur,
56(1).
23. Bilimovich, A. D. (1911). Equations of motion of a rigid body about a fixed point. Collection
of papers devoted to Professor G. K. Suslov (pp. 23–74). Kiev.
24. Birtea, P., Casu, I., & Comanescu, D. (2011). Hamilton–Poisson formulation for the rotational
motion of a rigid body in the presence of an axisymmetric force field and a gyroscopic torque.
Physics Letters A, 375, 3941–3945.
25. Bizyaev, I. A., Borisov, A. V., & Mamaev, I. S. (2016). Generalizations of the Kovalevskaya
case and quaternions. Proceedings of the Steklov Institute of Mathematics, 295, 33–44.
arXiv:1607.07982v1 [nlin.SI].
26. Bobenko, A. I., Reyman, A. G., & Semenov-Tian-Shansky, M. A. (1989). The Kowalewski
top 99 years later: A Lax pair, generalizations and explicit solutions. Communications in
Mathematical Physics, 122, 321–354.
27. Bobylev, D. N. (1896). On a particular solution of differential equations of rotation of a heavy
rigid body about a fixed point. Trud. Otdel. Fiz.Nauk Obshch-va Liub. Estestv., 8, 21–25.
28. Bogoyavlensky, A. A. (1958). On particular cases of motion of a heavy rigid body about a
fixed point. Journal of Applied Mathematics and Mechanics, 22, 873–906.
29. Bogoyavlensky, O. I. (1984). New integrable problem of classical mechanics. Communica-
tions in Mathematical Physics, 94, 255–269.
30. Bogoyavlensky, O. I. (1984). Euler equations on finite dimensional Lie algebras arising in
physical problems. Communications in Mathematical Physics, 95, 307–315.
31. Bogoyavlensky, O. I. (1986). Integrable cases of a rigid body dynamics and integrable systems
on the ellipsoids. Communications in Mathematical Physics, 103, 305–322.
32. Bogoyavlensky, O. I. (1986). Integrable cases of the dynamics of a rigid body and integrable
systems on the spheres S n . Math. USSR Izvestiya, 27, 203–218 (1986) = Izv. Akad. Nauk.
USSR, 49(5) (1985).
33. Bogoyavlenskii, O. I., & Ivakh, G. F. (1985). Topological analysis of Steklov’s integrable
cases. Uspekhi Mat. Nauk, 40, 145–146.
34. Bolsinov, A. V., & Fomenko, A. T. (2004). Integrable Hamiltonian systems. Geometry, topol-
ogy, classification. Boca Raton: Chapman & Hall/CRC.
35. Bolsinov, A. V., Fomenko, A. T., & Oshemkov, A. A. (2006). Topological methods in the
theory of integrable Hamiltonian systems. Cambridge: Cambridge Scientific Publishers.
36. Borisov, A. V. (1996). Necessary and sufficient conditions for the integrability of Kirchhoff’s
equations. Regular and Chaotic Dynamics, 1, 61–73.
37. Borisov, A. V., & Mamaev, I. S. (1997). Non-linear Poisson brackets and isomorphisms in
dynamics. Regular and Chaotic Dynamics, 2(3–4), 72–89.
38. Borisov, A. V., & Mamaev, I. S. (1999). Poisson structures and Lie algebras in Hamiltonian
mechanics. Izhevsk: Izd. UdSU.
39. Borisov, A. V., Mamaev, I. S., & Sokolov, V. V. (2001). A new integrable case on so(4). Dokl.
Akad. Nauk, 381, 614–615.
40. Borisov, A. V., & Mamaev, I. S. (2001). Rigid body dynamics (1st ed.). RCD.
@seismicisolation
@seismicisolation
Bibliography 445
41. Borisov, A. V., & Mamaev, I. S. (2019). Rigid body dynamics - Hamiltonian methods, inte-
grability and chaos. De Gruyter. (Updated translation of the original Russian published by
“Regular and chaotic dynamics. Moscow-Izhevsk: Institute of Computer Science” (2005)).
42. Borisov, A. V., & Mamaev, I. S. (Eds.). (2009). The system of Clebsch. Separation of variables,
explicit integration. Moscow-Izhevsk: RCD.
43. Bottema, O. (1945). The stability of Staude’s top motion. Koninkl. Nederland. Akad. Weten-
schap., Proc., 48, 316–325.
44. Bradbery, T. C. (1968). Theoretical mechanics. New York: Wiley.
45. Braden, H. W., Enolski, V. Z., & Fedorov, Yu. N. (2013). Dynamics on strata of trigonal
Jacobians and some integrable problems of rigid body motion. Nonlinearity, 26, 1865.
46. Bressan, A. (1957). Sulle precessioni d’un corpo rigido costituenti moti di Hess. Rend. Sem.
Mat. Univ. Padova, 21, 276–283.
47. Brun, F. (1893). Rotation kring fix punkt. Ofvers. Kongl. Svenska Vetensk Acad. Forhandl.,
7, 455–468.
48. Brun, F. (1907). Rotation kring fix punkt. II, III. Ark. Mat. Ast. Fys., 4(4), 1–4, 6(5), 1–10
(1910).
49. Bryan, G. H. (1893). A hydrodynamical proof of the equations of motion of a perforated solid,
with applications to the motion of a fine rigid framework in circulating liquid. Philosophical
Magazine, Series 5, 35(215), 338–354.
50. Burov, A. A., & Subkhankulov, G. I. (1984). On the existence of additional integrals of the
equations of motion of a magnetizable solid in an ideal fluid, in the presence of a magnetic
field. Journal of Applied Mathematics and Mechanics, 48, 538–541.
51. Capodanno, P. (1978). Sur la stabilité de certains mouvements particuliers d’un satellite arti-
ficiel de la terre soumis aux efforts aérodynamiques. Celestial Mechanics, 18, 337–350.
52. Chaplygin, S. A. (1901). A new case of rotation of a heavy rigid body with one point fixed.
Tr. Otdel. Fiz. Nauk. Obsh. Liub. Est., 10(2), 32–34.
53. Chaplygin, S. (1948). New particular solution of the problem of motion of a rigid body in a
liquid. Collected works. Vol. I (pp. 337–346). Originally: Tr. Otdel. Fiz. Nauk Ob-va Liubit.
Estest., 11(2), 7–10 (1903).
54. Chaplygin, S. (1903). A new partial solution of the problem of rotation of a heavy rigid body
about a fixed point. Tr. Otdel. Fiz. Nauk Ob-va Liubit. Estest., 12(1), 1–4.
55. Clebsch, A. (1870). Ueber die Bewegung eines Körpers in einer Flüssigkeit. Math. Ann., 3,
238–262.
56. Corliss, J. J. (1932). On the unsymmetric top. Acta Math., 59, 423–441.
57. Corliss, J. J. (1934). On the unsymmetric top. Acta Math., 62, 301–312.
58. Coulson, C. A. (1958). Electricity. Edinburgh: Oliver and Boyd.
59. Crabtree, H. (1909). An elementary treatment of the theory of spinning tops and gyroscopic
motion. New York: Longmans and Green.
60. Delone, N. B. (1892). On the question of geometric interpretation of the integrals of motion
of a rigid body about a fixed point, given by C.V. Kowalevski. Mat. Sbor. Kruzh. Liub. Mat.
Nauk., 16, 346–351.
61. Diaz, J. B., & Metcalf, F. T. (1962). On a result of Hadamard concerning the sign of the
precession of a heavy symmetrical top. Proceedings of the American Mathematical Society,
13, 669–670.
62. Diaz, J. B., & Metcalf, F. T. (1964). Upper and lower bounds for the apsidal angle in the theory
of the heavy symmetrical top. Archive for Rational Mechanics and Analysis, 16, 214–229.
63. Dokshevich, A. I. (1966). On a particular solution of the problem of rotation of a rigid body
about a fixed point. Dokl. A. N. USSR., 167, 1251–1252.
64. Dokshevich, A. I. (1970). A new particular solution of the problem of motion of a heavy rigid
body about a fixed point. Mekh. Tverd. Tela., 2, 8–12.
65. Dokshevich, A. I. (1970). A new particular solution of the equations of motion of a gyrostat
with an immovable point. Mekh. Tverd. Tela., 2, 12–15.
66. Dokshevich, A. I. (1981). Two classes of motions of the Kowalevski top. Journal of Applied
Mathematics and Mechanics, 45, 550–554.
@seismicisolation
@seismicisolation
446 Bibliography
67. Dokshevich, A. I. (1992). Solutions in a finite form of the Euler-Poisson equations. Moscow-
Izhevsk: RCD (In Russian).
68. Dovbysh, S. A. (1990). Oscillational properties of plane motions in the dynamics of a sym-
metric rigid body. Izv. R.A.N. Mekh. Tverdogo. Tela., 25(4), 11–19.
69. Dragović, V. (1997). Note on L-A pair for the Kovalevskaya gyrostat in a magnetic field. Mat.
Vesnik, 49, 279–281.
70. Dragović, V., & Kukić, K. (2014). The Sokolov case, integrable Kirchhoff elasticae, and
genus 2 theta functions via discriminantly separable polynomials. Tr. Mat. Inst. Steklova, 286,
246–261.
71. Dubrovin, B. A., Krichever, I. M., & Novikov, S. P. (1986). Integrable systems I. In dynamical
systems IV, symplectic geometry and its applications. Berlin: Springer.
72. Durant, W. (1994). The story of civilization. Vol 1: Our oriental heritage (1935) (Electronic
ed.). World Library.
73. Elmandouh, A. A. (2015). New integrable problems in rigid body dynamics with quartic
integrals. Acta Mech., 226, 2461–2472.
74. Elmandouh, A. A. (2015). New integrable problems in the dynamics of particle and rigid
body. Acta Mech., 226, 3749–3762.
75. Erdélyi, A., et al. (1953). Higher transcendental functions (Vol. II). New York: McGraw Hill.
76. Ershkov, S. (2014). New exact solution of Euler’s equations (rigid body dynamics) in the case
of rotation over the fixed point. Archive of Applied Mechanics, 84, 385–389.
77. Ershkov, S. On the invariant motions of rigid body rotation over the fixed point, via Euler
angles. Archive of Applied Mechanics, 86, 1797–1804.
78. Euler, L. (1758–1765). Recherche sur la connoissance mecanique des corps. Histoire de
l’Académie Royale des Sciences, Berlin, 14, 131–153.
79. Euler, L. (1758–1765). Du mouvement de rotation des corps solides autour d’un axe variable.
Histoire de l’Académie Royale des Sciences, Berlin, 14, 154–193.
80. Fabbri, M. R. (1934). Sopra una soluzione particolare delle equazioni del moto di un solido
pesante intorno ad un punto fisso. Atti R. Accad. Naz. Lincei. Rend. Cl. Sci. Fis. Mat. Nat (6),
19, 407–415.
81. Fabbri, M. R. (1934). Sopra un particolare movimento di un solido pesante intorno a un punto
fisso. Atti R. Accad. Naz. Lincei. Rend. Cl. Sci. Fis. Mat. Nat (6), 19, 495–502.
82. Fabbri, M. R. (1934). Sui coni di Poinsot in una particolare rotazione dei solidi pesanti. Atti
R. Acc. Naz. dei Lincei. Ser. 6., 19, 872–873.
83. Fawcett, P. G. (1893). Note on the motion of solids in liquid. Quarterly Journal of Mathematics,
26, 231–258.
84. Fedorov, I. N., & Garcia-Naranjo, L. C. (2016). A shortcut to the Kovalevskaya curve via
pencils of genus 3 curves. arXiv:1606.08331v1 [nlin.SI].
85. Field, P. (1931). On the unsymmetrical top. Acta Math., 56, 355–362.
86. Field, P. (1934). On the unsymmetrical top. Acta Math., 62, 313–316.
87. Fomenko, A. T. (1986). Morse theory of integrable Hamiltonian systems. Soviet Mathematics
Doklady, 33, 502–506.
88. Fomenko, A. T. (1988). Topological invariants of Hamiltonian systems that are integrable in
the sense of Liouville. Functional Analysis and Its Applications, 22, 286–296.
89. Fomenko, A. T. (1988). Integrability and nonintegrability in geometry and mechanics. Dor-
drecht: Kluwer Academic Publishers.
90. Fomenko, A. T. (Ed.). (1991). Topological classification of integrable systems. Advances in
Soviet Mathematics, AMS, 6.
91. Fomenko, A. T., & Nikolaenko, S. S. (2015). The Chaplygin case in dynamics of a rigid body
in fluid is orbitally equivalent to the Euler case in rigid body dynamics and to the Jacobi
problem about geodesics on the ellipsoid. Journal of Geometry and Physics, 87, 115–133.
92. Fomenko, A. T., & Konyaev, A. Yu. (2012). New approach to symmetries and singularities
in integrable Hamiltonian systems. Topology and Its Applications, 159, 1964–1975.
93. Fomenko, A. T., & Zieschang, H. (1987). On the topology of the three-dimensional manifolds
arising in Hamiltonian mechanics. Soviet Mathematics Doklady, 35, 520–534.
@seismicisolation
@seismicisolation
Bibliography 447
94. Fomenko, A. T., & Zieschang, H. (1991). A topological invariant and a criterion for the
equivalence of integrable Hamiltonian systems with two degrees of freedom. Math. USSR
Izv., 36, 567–596.
95. Françoise, J. P., & Tarama, D. (2019). The rigid body dynamics in an ideal fluid: Clebsch top
and Kummer surfaces. arXiv:1903.00917 [math-ph].
96. Gaffet, B. (1998). A completely integrable Hamiltonian motion on the surface of a sphere.
Journal of Physics A: Mathematical and General, 31, 1581–1596.
97. Gaffet, B. (1998). An integrable Hamiltonian motion on a sphere II. The separation of vari-
ables. Journal of Physics A: Mathematical and General, 31, 8341–8354.
98. Gao, P. (2003). Two new exactly solvable cases of the Euler-Poisson equations. Mechanics
Research Communications, 30, 203–205.
99. Gashenenko, I. N. (1985). Investigation of a class of motions of Chaplygin’s gyroscope. Mekh.
Tverd. Tela, 17, 6–9.
100. Gashenenko, I. N. (1991). A new class of motions of a heavy gyrostat. Dokl. Acad. Nauk
USSR, 318, 66–68.
101. Gashenenko, I. N. (1992). A case of integrability of the equations of motion of a gyrostat.
Mekh. Tverd. Tela, 24, 1–4.
102. Gashenenko, I. N. (2000). Bifurcational set of the problem of motion of a gyrostat, subject to
Kowalevski’s conditions. Mekh. Tverd. Tela, 27, 31–35.
103. Gashenenko, I. N. (2000). Invariant sets in the space of angular velocities of a heavy rigid
gyrostat. Mekh. Tverd. Tela, 30, 79–87 (Russian) Zbl 1049.70580.
104. Gashenenko, I. N. (2002). Enveloping surfaces in the problem on motion of a heavy gyrostat.
Mekh. Tverd. Tela, 32, 39–49 (Russian) Zbl 1042.70004.
105. Gashenenko, I. N. (2003). Integral manifolds in the problem of the motion of a heavy rigid
body. Mekh. Tverd. Tela, 33, 20–32 (Russian) Zbl 1120.70312.
106. Gashenenko, I. N., & Richter, P. (2004). Enveloping surfaces and admissible velocities of
heavy rigid bodies. International Journal of Bifurcation and Chaos, 14, 2525–2553.
107. Gashenenko, I. N. (2009). On D.N. Goryachev’s solution. Mekh. Tverd. Tela, 39, 29–41.
108. Gashenenko, I. N., Gorr, G. V., & Kovalev, A. M. (2012). Classical problems of rigid body
dynamics. Kiev: Naukova Dumka.
109. Gavrilov, L. N. (1987). On the geometry of Gorjatchev – Tchaplygin top. C.R. Acad. Bulg.
Sci., 40, 33–36.
110. Gavrilov, L. (1989). Remarks on the equations of motion of heavy gyrostat. C. R. Acad.
Bulgare Sci., 42(5), 17–20.
111. Gavrilov, L. (1992). Nonintegrability of the equations of heavy gyrostat. Compositio Math.,
82(3), 275–291.
112. Goldstein, H., Poole, C., & Safko, J. (2000). Classical mechanics (3rd ed.). Boston: Addison-
Wesley.
113. Golubev, V. V. (1953). Lectures on the integration of the equations of motion of a rigid body
about a fixed point. Moscow: Gostekhizdat. English transl., Israel Program for Scientific
Translations, Jerusalem, and Office of Technical Services, U.S. Department of Commerce,
Washington, D. C.
114. Goryachev, D. N. (1899). New particular solution of the problem of motion of a heavy rigid
body about a fixed point. Trudy Ob-va estest., 1, 23–24.
115. Goryachev, D. N. (1900). On the motion of a rigid body about a fixed point in the case
A = B = 4C. Mat. Sb. Kr. Liub. Mat. Nauk, 21(3), 431–438.
116. Goryachev, D. N. (1910). Certain general integrals in the problem of motion of a rigid body.
Warsaw.
117. Goryachev, D. N. (1915). New cases of motion of a rigid body about a fixed point. Warshav.
Univ. Izvest., 3, 1–11.
118. Goryachev, D. N. (1916). New cases of integrability of Euler’s dynamical equations. Warshav.
Univ. Izvest., 3, 1–15.
119. Gorr, G. V. (1979). On the precession of a gyrostat in a potential field. Mekh. Tverd. Tela, 11,
64–67.
@seismicisolation
@seismicisolation
448 Bibliography
120. Gorr, G. V., Kudryashova, L. V., & Stepanova, L. V. (1978). Classical problems of motion of
a rigid body. Evolution and contemporary state. Kiev (In Russian): Naukova Dumka.
121. Gorr, G. V., & Kovalev, A. M. (2013). Dynamics of the gyrostat. Kiev: Naukova Dumka.
122. Gorr, G. V., & Kovalev, A. M. (1988). On asymptotically uniform motions about an inclined
axis in the generalized problem of motion of a rigid body with a fixed point. Mekh. Tvjerd.
Tela., 20, 13–18.
123. Gorr, G. V., & Kurgansky, N. V. (1987). On the regular precession about a vertical axis in the
problem of motion of a rigid body. Mekh. Tverd. Tela, 19, 16–20 (Russian).
124. Gorr, G. V., & Levitskaya, G. D. (1971). On a periodic motion of the Goryachev-Chaplygin
gyroscope. Mekh. Tvjerd. Tela., 3, 101–106.
125. Gorr, G. V., & Maznev, A. V. (2010). Dynamics of the gyrostat with a fixed point. Donetsk:
IAMM.
126. Gorr, G. V., Maznev, A. V., & Shchetinina, E. K. (2009). Precessional motions in rigid body
dynamics and systems of coupled rigid bodies. Donetsk: DNU.
127. Gorr, G. V., & Shchetinina, E. K. (2006). New classes of precessional motions of a gyrostat
acted upon by potential and gyroscopic forces. Trudy IPMM NAN Ukraine., 12, 36–45.
128. Gorr, G. V., & Uzbek, E. K. (2002). On the integration of Poisson’s equations in the case
of three linear invariant relations. Journal of Applied Mathematics and Mechanics, 66(3),
409–417; translation from PMM. Prikl. Mat. Mekh., 66(3), 418–426.
129. Gorr, G. V., & Uzbek, E. K. (2004). Fractional-linear integral of Poisson’s equations in the
case of three linear invariant relations. Intern. MFNA-ANN J: Problems of Nonlinear Analysis
in Engineering Systems, 21, 54–63.
130. Gradshteyn, I. S., & Ryzhik, I. M. (2007). Table of integrals, series, and products (7th ed.).
New York: Academic.
131. Grammel, R. (1920). Die Stabilität der Staudeschen Kreiselbewegungen. Math. Z., 6, 124–
142.
132. Grammel, R. (1950). Der Kreisel, seine Theorie und seine Anwendungen. Zweite, neubear-
beitete Auflage. Erster Band: Die Theorie des Kreisels. Zweiter Band: Die Anwendungen des
Kreisels. Berlin: Springer.
133. Gray, A. (1918). A treatise on gyrostatics and rotational motion. London: Macmillan.
134. Greenhill, A. G. (1877). On the motion of a top and allied problems in dynamics. Quarterly
Journal, 11, 176–194.
135. Greenhill, A. G. (1898). The motion of a solid in infinite liquid under no forces. American
Journal of Mathematics, 20, 1–75.
136. Greenhill, A. G. (1906). The motion of a solid in infinite liquid. American Journal of Mathe-
matics, 28(71–100), 101–158.
137. Grigor’ev, Yu. A., Khudobakhshov, V. A., & Tsiganov, A. V. (2013). Separation of variables
for some systems with a fourth-order integral of motion. TMF, 177(3), 468–481.
138. Grioli, G. (1947). Esistenza e determinazione delle precessioni regulari dinamicamente pos-
sibili per un solido pesante asymmetrico. Ann. Mat. Pura Appl. (4), 26, 271–281.
139. Grioli, G. (1957). Movimenti dinamicamente possibili per un solido asymmetrico soggitto a
forze di potenza nulla. Rend. Acc. naz. Lincei, Ser. VIII, 22, fasc. 459–463.
140. Grioli, G. (1963). Qualche teorema di cinematica dei moti rigidi. Rend. Acad. naz. Lencei,
Ser. 8, 34, 636–641.
141. Gulyaev, M. P. (1955). On a new particular solution of the equations of motion of a heavy
rigid body having a fixed point. Vestnik Mosk. Univ., Ser. Fiz. Mat., 2, 15–21.
142. Gulyaev, M. P. (1973). On regular precessions of a heavy gyrostat. PMM J. Appl. Math. Mekh.,
37, 704–712.
143. Hadamard, J. (1895). Sur la precession dans le mouvement d’un corps pesant de revolution
fixe par un point de son axe. Bull. Sci. Math., 19, 228–230.
144. Hagihara, Y. (1970). Celestial mechanics. Vol. I, dynamical principles and transformation
theory. Cambridge: MIT Press.
145. Haine, L., & Horozov, E. I. (1987). A Lax pair for Kowalevski’s top. Phys. D, 29(1–2),
173–180.
@seismicisolation
@seismicisolation
Bibliography 449
146. Halphen, G. H. (1888). Sur le mouvement d’un solide dans un liquide. J. math. pures appl.
4e série, 4, 5–82.
147. Hamel, G. (1947). Uber den algemeine schweren Kreisel. ZAMM Journal of Applied Mathe-
matics and Mechanics, 25(5–6), 159–160.
148. Hassan, S. Z. (1994). Certain problems in rigid body dynamics. Ph.D. thesis. Mansoura
University.
149. Hassan, S. Z., Kharrat, B. N., & Yehia, H. M. (1999). On the stability of motion of a gyrostat
about a fixed point under the action of non-symmetric fields. European Journal of Mechanics
A/Solids, 18, 313–318.
150. Hess, W. (1890). Über die Eulerschen Bewegungsgleichungen und über eine neue particulare
Lösung des Problems der Bewegung eines starren Körpers un einen festen Punkt. Math. Ann.,
Bd., 37, 153–181.
151. Holmes, Ph., Jenkins, J., & Leonard, N. (1998). Dynamics of the Kirchhoff equations I:
Coincident centers of gravity and buoyancy. Physica D, 118, 311–342.
152. Horozov, E. I. (1989). The full geometry of Kowalewski’s top and (1,2)-abelian surfaces.
Communications on Pure and Applied Mathematics, 42(4), 357–407.
153. Hussein, A. M. (2019). Precessional motion of a rigid body acted upon by three irreducible
fields. Russian Journal of Nonlinear Dynamics, 15, 285–292.
154. Husson, E. (1906). Recherche des intégrales algebriques dans le mouvement d’un solide
pesant autour d’un point fixe. Ann. Fac. Sci. Univ. Toulouse, Ser. 2, 8, 73–152.
155. Husson, E. (1908). Sur un théoreme de H. Poincare relativement au mouvement d’un solide
pesant. Acta Math., 31, 71–88.
156. Ince, E. L. (1940). The periodic Lamé functions. Proceedings of the Royal Society of Edin-
burgh, 60, 47–63.
157. Ince, E. L. (1940). Further investigations into the periodic Lamé functions. Proceedings of
the Royal Society of Edinburgh, 60, 83–99.
158. Ipatov, A. F. (1970). Motion of the Kowalevski gyroscope on borders of ultraellipticity zones.
Uch. Zap. Petrozavod. Univ., 18, 6–93.
159. Irtegov, V., & Titorenko, T. (2017). On stationary motions of the generalized kowalewski
gyrostat and their stability. In V. P. Gerdt et al. (Eds.), CASC 2017. LNCS (Vol. 10490, pp.
210–224).
160. Ismail, A. I. (1998). The motion of a fast spinning disc which comes out from the limiting
case γ0 ≈ 0. Computer Methods in Applied Mechanics and Engineering, 161, 67–76.
161. Jackson, J. D. (1998). Classical electrodynamics (3rd ed.). New York: Wiley.
162. Jacobi, C. G. J. (1866). Vorlesungen über Dynamik. Kö nigsberg.
163. Joukovsky, N. E. (1948). On the motion of a rigid body with holes filled with a homogeneous
fluid. Collected works. Vol. I (pp. 31–152). Moscow. (Originally: Journal of Russian Physical
Chemistry Society, 17, (1885); 6, 81–113; 7, 145–149; 8, 231–280).
164. Joukovsky, N. E. (1948). Geometric interpretation of the case of motion of a rigid body about
a fixed point considered by S. V. Kowalevski. Collected works. Vol. I (pp. 316–350). Moscow:
OGIZ. Originally, reported to Moscow Mathematical Society in 1892.
165. Joukovsky, N. E. (1948). The loxodromic pendulum of Hess. Collected works. Vol I (pp.
257–274). Moscow. (Originally: Trudy. Otdel. Fiz. Nauk. Ob-va liobit. est., 5, 37–45 (1893)).
166. Karapetyan, A. V. (2006). Invariant sets in the Goryachev–Chaplygin problem: Existence,
stability, and branching. Journal of Applied Mathematics and Mechanics, 70, 195–198.
167. Keis, I. A. (1963). On the existence of certain integrals of the equations of motions of a
gyrostat with a fixed point. Vestn. MGU, Ser. Mat.-Mekh., 6, 55–63.
168. Keis, I. A. (1964). On the algebraic integrals in the problem of motion of a heavy gyrostat
fixed at one point. PMM, Journal of Applied Mathematics and Mechanics, 28, 633–639.
169. Keis, I. A. (1965). Two solutions of the problem of the motion of a gyrostat having a fixed
point. Eston. SSR, 14, 552–554.
170. Keis, I. A. (1965). On certain necessary conditions of existence of single-valued integrals for
the equations of motion of a heavy gyrostat having a fixed point. Eston. SSR, 14, 555–558.
@seismicisolation
@seismicisolation
450 Bibliography
171. Kharlamov, M. P. (1976). On a conditionally linear integral of the equation of motion for a
rigid body having a fixed point. Akad. Nauk USSR, Izv., Mekhanika Tverdogo Tela, 11, 9–17.
English translation: Mechanics of Solids, 11(3), 6–13.
172. Kharlamov, M. P. (1983). On a class of motion of a gyrostat. Mekh. Tverdogo Tela, 15, 47–56.
173. Kharlamov, M. P. (1983). Symmetry in systems with gyroscopic forces. Mekh. Tverdogo Tela,
15, 87–93.
174. Kharlamov, M. P. (1986). On an asymptotic motion of the heavy gyrostat. Mekh. Tverdogo
Tela, 18, 12–15.
175. Kharlamov, M. P. (2002). A class of solutions with two invariant relations in the problem of
motion of Kowalevski top in a double constant field. Mekh. Tverd. Tela, 32, 32–38.
176. Kharlamov, M. P. (2004). Critical set and bifurcation diagram in the problem of motion of
Kowalevski top in a double field. Mekh. Tverd. Tela, 34, 47–58.
177. Kharlamov, M. P. (2007). Critical subsystems of the Kowalevski gyrostat in two constant
fields. Russian Journal of Nonlinear Dynamics, 3, 331–348.
178. Kharlamov, M. P. (2008). One class of solutions with two invariant relations for the problem
of motion of the Kowalevski top in a double field. arXiv:0803.1028.
179. Kharlamov, M. P. (2007). Separation of variables in the generalized 4th Appelrot class. Regular
and Chaotic Dynamics, 12, 267–280.
180. Kharlamov, M. P. (2009). Separation of variables in the generalized 4th Appelrot class. II.
Real solutions. Regular and Chaotic Dynamics, 14, 621–634.
181. Kharlamov, M. P., Kharlamova, I. I., & Shvedov, E. G. (2010). Bifurcation diagrams on the
iso-energetic levels of the Kowalevski-Yehia gyrostat. Mekh. Tverdogo Tela, 40, 77–90.
182. Kharlamov, M. P. (2014). Extensions of the Appelrot classes for the generalized gyrostat in a
double force field. Regular and Chaotic Dynamics, 19, 226–244.
183. Kharlamov, M. P. (2015). Topological analysis of integrable problems in rigid body dynamics.
Regular and Chaotic Dynamics, Moscow-Izhevsk.
184. Kharlamov, M. P., Ryabov, P. E., & Kharlamova, I. I. (2016). Topological atlas of the
Kowalevski-Yehia gyrostat. Regular and Chaotic Dynamics, Moscow-Izhevsk (In Russian).
185. Kharlamov, M. P., Ryabov, P. E., & Kharlamova, I. I. (2017). Topological atlas of the
Kowalevski-Yehia gyrostat. Journal of Mathematical Sciences, 227, 241–386.
186. Kharlamov, M. P., Ryabov, P. E., & Savushkin, A. Yu. (2016). Topological atlas of the
Kowalevski - Sokolov top. Regular and Chaotic Dynamics, 21, 24–65.
187. Kharlamov, M. P., Ryabov, P. E., Savushkin, A. Y., & Smirnov, G. E. (2011). Types of critical
points of the Kowalevski gyrostat in double field. Mekh. Tverdogo Tela, 41, 26–37.
188. Kharlamov, M. P., & Shvedov, E. G. (2006). On the existence of motions in the generalized
4th Appelrot class. Regular and Chaotic Dynamics, 11, 337–342.
189. Kharlamov, M. P., & Yehia, H. M. (2015). Separation of variables in one case of motion of
a gyrostat acted upon by gravity and magnetic fields. EJBAS. Egyptian Journal of Basic and
Applied Sciences, 2, 236–242.
190. Kharlamov, P. V. (1955). On a case of integrability of a heavy rigid body in a liquid. Prikl.
Mat. Mekh., 19, 231–233.
191. Kharlamov, P. V. (1964). On the equations of motion for a heavy body with a fixed point. PMM,
Journal of Applied Mathematics and Mechanics, 27, 1070–1078; translation from Prikl. Mat.
Mekh., 27, 703–707 (1963).
192. Kharlamov, P. V. (1963). On the motion in a liquid of a body bounded by a multiconnected
surface. Journal of Applied Mathematics and Theoretical Physics, 4, 17–29.
193. Kharlamov, P. V. (1964). A solution of the problem of motion of a body with a fixed point.
PMM, Journal of Applied Mathematics and Mechanics, 28, 158–159.
194. Kharlamov, P. V. (1964). Kinematical interpretation of the motion of a body with a fixed point.
PMM, Journal of Applied Mathematics and Mechanics, 28, 502–507.
195. Kharlamov, P. V. (1965). Lectures on rigid body dynamics. Novosibirsk University (In Rus-
sian).
196. Kharlamov, P. V. (1965). Polynomial solutions of equations of motion of a body with a fixed
point. Prikl. Mat. Mekh., 29, 26–34.
@seismicisolation
@seismicisolation
Bibliography 451
197. Kharlamov, P. V. (1965). On solutions of the equations of motion of a rigid body. Prikl. Mat.
Mekh., 29, 567–572.
198. Kharlamov, P. V. (1971). A case of integrability of the equations of motion of a rigid body
with a fixed point. Mekh. Tverdogo Tela, 3, 57–64.
199. Kharlamov, P. V. (1977). Different variants of a solution of the problem of motion of a body
with a fixed point. Mekh. Tverdogo Tela, 9, 17–24.
200. Kharlamov, P. V. (1991). Commentaries to a paper of L. A. Stepanova: “An unsuccessful
attempt to defend the priority of the classics of Russian mechanics in constructing exact
solutions in the mechanics of a rigid body” [Mekh. Tverd. Tela, 22, 19–33 (1990)] and of A.
I. Khokhlov: “An unsuccessful attempt to defend a result of V. A. Steklov” [Mekh. Tverd.
Tela, 23, 26–36 (1991); MR1160719 (93g:01039)]. (Russian) Mekh. Tverd. Tela, 23, 36–43.
MR1160720.
201. Kharlamov, P. V., & Gorr, G. V. (1977). On a work of N. I. Mertsalov. Mekh. Tvjerd. Tela, 9,
58–61.
202. Kharlamov, P. V., & Kovaleva, L. M. (1970). On a new solution of the problem of motion of
a heavy gyrostat. Mekh. Tverd. Tela, 2, 3–8.
203. Kharlamov, P. V., Mozalevskaya, G. V., & Lesina, M. E. (2001). On various representations
of the Kirchhoff equations. Mekh. Tverd. Tela, 31, 3–17.
204. Kharlamova, E. I. (1959). On the motion of a rigid body about a fixed point in a central
Newtonian field. Izv. Sibir. Otdel. AN USSR., 6, 7–17.
205. Kharlamova, E. I. (1965). Some solutions of the problem of motion of a body with a fixed
point. PMM Journal of Applied Mathematics and Mechanics, 29, 868–873.
206. Kharlamova, E. I. (1969). Reducing the problem of motion of a body having a fixed point to
a single differential equation. Mekh. Tverd. Tela, 1, 107–116.
207. Kharlamova, E. I. (1969). On the canonical equations of motion of a body having a fixed
point. Mekh. Tverdogo Tela, 1, 102–107.
208. Kharlamova, E. I. (1969). On a linear invariant relation of equations of motion of a body about
a fixed point. Mekh. Tverd. Tela, 1, 5–12.
209. Kharlamova, E. I. (1971). On algebraic invariant relations of the integrodifferential equation
of the problem of motion of a rigid body about a fixed point, under Hass’ conditions. Mekh.
Tverd. Tela, 3, 28–32.
210. Kharlamova, E. I., & Kharlamov, P. V. (1969). A new case of integrability of the equations of
motion of a heavy rigid body about a fixed point. Dokl. Acad. Nauk. USSR, 188, 770–771.
211. Kharlamova, E. I., & Stepanova, L. A. (1988). On the isomorphism of certain problems of
rigid body dynamics and trials to construct new solutions by means of change of variables.
Mekh. Tverd. Tela, Kiev, 20, 1–12.
212. Kharlamova, E. I., & Mozalevskaya, T. V. (1986). Integro-differential equation of rigid body
dynamics. Kiev: Naukova Dumka.
213. Kharlamova, I. I., & Savushkin, A. Y. (2010). Bifurcation diagrams involving the linear
integral of Yehia. Journal of Physics A: Mathematical and Theoretical, 43, 105203.
214. Kharlamova, L. N. (1990). A solution of the equations of motion of a body under action of
potential and gyroscopic forces. Mekhanika Tverdogo Tela, Kiev, 22, 40–45.
215. Khlystunova, N. V. (2000). On Grioli type precessions of a heavy rigid body in a liquid. Journal
of Applied Mathematics and Mechanics, 64(4), 527–530; translation from Prikl. Mat. Mekh.,
64(4), 551–554.
216. Khokhlov, A. I. (1990). Some erroneous statements in problems of rigid body dynamics.
Izvestiya Akad. Nauk USSR- Mekhanika Tverdogo Tela, 25(2), 84–86. English Translation:
Mechanics of Solids, 25(2), 83–87.
217. Khokhlov, A. I. (1991). On an unsuccessful trial to defend a result of Steklov. Mekh. Tverdogo
Tela, Kiev, 23, 26–36.
218. Khudobakhshov, V. A., & Sozonov, A. P. (2013). Separation of variables for some generali-
sation of the Kowalevski top. Nelin. Din., 9, 247–255.
219. Kirchhoff, G. R. (1870). Über die Bewegung eines Rotationsk örpers in einer Flussigkeit. J.
Reine und agew. Math., 71, 237–262.
@seismicisolation
@seismicisolation
452 Bibliography
220. Kirchhoff, G. R. (1874). Vorlesungenū ber mathematische Physik. Mechanik. Leipzig.

221. Klein, F. (1896). The mathematical theory of the top. Congruence of sets and other mono-
graphs. Lectures delivered in Princeton in 1896. New York: Chelsea Publishing Company.
222. Klein, F., & Sommerfeld, A. (2008, 2010). The theory of the top. Vols I, II. Basel: Birkhäuser.
223. Kobb, G. (1895). Sur le problème de la rotation d’un corps autour d’un point fixe. Bull. Soc.
math. France, t., 23, 210–215.
224. Kolossov, G. V. (1903). On certain modifications of Hamilton’s principle applied to problems
of mechanics of a rigid body (76 pp.). St. Petersburg: Erlich Printing House (In Russian).
225. Komarov, I. V. (1987). A generalization of the Kovalevskaya top. Physics Letters, 123, 14–15.
226. Komarov, I. V., & Kuznetsov, V. B. (1987). Generalization of the Goryachev–Chaplygin
gyrostat in quantum mechanics. Zap. Nauchn. Sem. LOMI, 164, 134–141.
227. Komarov, I. V., Sokolov, V. V., & Tsiganov, A. V. (2003). Poisson maps and integrable defor-
mations of the Kowalevski top. Journal of Physics A: Mathematical and General, 36, 8035–
8048.
228. Komarov, I. V., & Tsiganov, A. V. (2004). On integration of the Kowalevski gyrostat and the
Clebsch problems. arXiv:nlin/0412041.
229. Komarov, I. V., & Tsiganov, A. V. (2005). On a trajectory isomorphism of the Kowalevski
gyrostat and the Clebsch problem. Journal of Physics A: Mathematical and General, 38,
2917–2927.
230. Konosevich, B. I., & Pozdnyakovich, E. V. (1970). The motion of a rigid body having a fixed
point in two particular cases of integrability of Euler-Poisson equations. Mekh. Tverdogo Tela,
2, 77–80.
231. Konosevich, B. I., & Pozdnyakovich, E. V. (1968). Two partial solutions of motion of a
rigid body having a fixed point. PMM, Journal of Applied Mathematics and Mechanics, 32,
561–565; translation from Prikl. Mat. Mekh., 32, 544–548.
232. Kötter, F. (1890–1891). Über das Kowalevskische Rotationsproblem. Jber. Deutschen Math.
Verein., 1, 65–68.
233. Kötter, F. (1892). Über die Bewegung eines festen Kö rpers in einer Flussigkeit. J. reine und
angew. Math., 109(51–81), 89–111.
234. Kötter, F. (1893). Sur le cas traité par Mme Kowalevski de rotation d’un corps solide autour
d’un point fixe. Acta Math., 17, 209–264.
235. Kötter, F. (1900). Die von Steklov und Lyapunov entdekten integralen Falle der Bewegung
eines starren Körpers in einer Flussigkeit. Sitz. Koniglich Preuss. Akad. Wiss., Berlin, 6, 79–87.
236. Kovalev, A. M. (1968). The moving angular velocity hodograph in Hess’ solution of the
problem of motion of a body with a fixed point. Prikl. Mat. Mekh., 32, 1111–1118.
237. Kovalev, A. M. (1969). On the motion of a body in Hess’ case. Mekh. Tverdogo Tela, 1, 12–27.
238. Kowalevski, S. (=Sofia Kovalevskaya) (1889). Sur le problème de la rotation d’un corps solide
autour d’un point fixe. Acta Math., 12(2), 177–232.
239. Kowalewski, N. (1908). Eine neue particulare Lösung der Differentialgleichungen der Bewe-
gung eines schweren starren Körpers um einen festen Punkt. Math. Ann., 65(4), 528–537.
240. Kozlov, V. V. (1980). Methods of qualitative analysis in the dynamics of a rigid body. Moscow:
Moscow State University.
241. Kozlov, V. V. (1980). Two integrable problems of classical mechanics. Vestn. MGU, Ser. 1,
Mat. - Mekh., 4, 80–83.
242. Kozlov, V. V. (1982). Hamiltonian equations of the problem of motion of a rigid body with a
fixed point in redundant coordinates. Teor. Primen. Meh., 8, 59–65.
243. Kozlov, V. V. (1985). Integrable cases of the problem of motion of a point over a three-
dimensional sphere in a force field with fourth-degree potential. Vestnik Moskov. Univ. Ser. 1.
Mat. Mekh., 40(3), 48–51.
244. Kozlov, V. V. (1995). Some integrable extensions of Jacobi’s problem of geodesics on an
ellipsoid. Journal of Applied Mathematics and Mechanics, 59, 1–7.
245. Kozlov, V. V. (1995). Symmetries, topology, and resonances in Hamiltonian mechanics. Berlin:
Springer.
@seismicisolation
@seismicisolation
Bibliography 453
246. Kozlov, V. V., & Onishchenko, D. A. (1982). Nonintegrability of Kirchhoff’s equations. Soviet
Mathematics Doklady, 26, 495–498.
247. Kurgansky, N. V. (1988). On semi-regular precessions of the first type about the vertical in a
problem of rigid body dynamics. Mekh. Tverdogo Tela, 20, 67–71.
248. Kuzmin, P. A. (1952). Supplement to V. A. Steklov’s case of motion of a heavy rigid body
about a fixed point. Prikl. Mat. Mekh., 16, 243–245.
249. Kuznetsov, V. B. (2002). Simultaneous separation for the Kowalevski and Goryachev–
Chaplygin gyrostats. Journal of Physics A: Mathematical and General, 35, 6419–6430.
250. Lagerborg, N. (1890). Sur le probleme du mouvement d’un corps solide autour d’un point
fixe. Bull. S. M. F., 18, 118–122.
251. Lagrange, J. L. (1788). Mécanique Analitique. Paris.
252. Lamb, H. (1879). A treatise on the mathematical theory of the motion of fluids. Cambridge:
The University Press.
253. Lamb, H. (1932). Hydrodynamics. Cambridge.
254. Larmor, J. (1897). On a dynamical theory of the electric and luminiferous medium. Philo-
sophical Transactions of the Royal Society, 190, 205–300.
255. Lecornu, M. (1902). Sur les petits mouvements d’un corps pesant. Bull. de la Soc. math. de
France, 30, 71–82.
256. Leimanis, E. (1965). The general problem of motion of coupled rigid bodies about a fixed
point. Berlin: Springer.
257. Leonard, N. E. (1997). Stability of a bottom-heavy underwater vehicle. Automatica, 33, 331–
346.
258. Leonard, N. E., & Marsden, J. E. (1997). Stability and drift of underwater vehicle dynamics:
Mechanical systems with rigid motion symmetry. Phys. D, 105, 130–162.
259. Lesfari, A. (1988). Abelian surfaces and Kowalevski’s top. Ann. Scient. Ec. Nor. Sup., 21, 4
ser., 193–223.
260. Lesfari, A. (2001). The problem of the motion of a solid in an ideal fluid. Integration of the
Clebsch’s case. Nonlinear Differential Equations and Applications, 8, 01–13.
261. Levi-Civita, T., & Amaldi, U. (1950). Lezioni di Meccanica Razionale. Bologna: Zanichelli
Editore.
262. Lewis, D., Ratiu, T., Simo, J. C., & Marsden, J. E. (1992). The heavy top: A geometric
treatment. Nonlinearity, 5, 1–48.
263. Llibre, J., & Valls, C. (2012). On the polynomial integrability of Kirchoff’s equations. Physica
D, 241, 1417–1420.
264. Logacheva, N. S. (2012). Classification of nondegenerate equilibria and degenerate 1-
dimensional orbits of the Kovalevskaya-Yehia integrable system. Sb: Math., 203, 28–59.
265. Lottner, C. (1855). Reduction der Bewegung eines schweren, um einen festen Punct rotirenden
Revolutionskörpers auf die elliptischen Transcendenten. Crelle Journal, 50, 111–125.
266. Lyapunov, A. M. (1894). On a property of the differential equations of motion of a heavy
rigid body having a fixed point. Kh. Mat. O-va., 4, 123–140. Reprinted in: Collected works,
Vol. I (pp. 402–417). Moscow: Izd. Akad. Nauk SSSR (1954).
267. Lyapunov, A. M. (1893). A new case of integrability of the equations of motion of a rigid
body in a liquid. Originally, Soobsch. Khark. Mat. O-va. Ser 2, 4(1–2), 81–85 and Lyapunov,
A. M. (1954). Collected works, Vol. I (pp. 320–324). Moscow: Izd. Akad. Nauk SSSR.
268. Maciejewski, A. J., & Przybylska, M. (2005). Differential Galois approach to the non-
integrability of the heavy top problem. Ann. Fac. Sci. Toulouse Math. (6), 14(1), 123–160.
269. Macmillan, W. D. (1939). Dynamics of rigid bodies. London.
270. Magnus, K. (1971). Kreisel. Theorie und Anwendungen. Berlin: Springer.
271. Magri, F., & Skrypnyk, T. (2015). The Clebsch system. arXiv:1512.04872.
272. Marcolongo, R. (1902). Osservazioni intorno alIa nota del Sig. KOLOSSOFF “Sur Ie cas de
M. GORIATSCHOFF de la rotation d’un corps pesant autour d’un point fixe”. Rend. Circ.
Mat. Palermo, 16, 349–357.
273. Marikhin, V. G., & Sokolov, V. V. (2010). Transformation of a pair of commuting Hamiltonians
quadratic in momenta to canonical form and real partial separation of variables for the Clebsch
top. Regular and Chaotic Dynamics, 15, 652–658.
@seismicisolation
@seismicisolation
454 Bibliography
274. Markeev, A. P. (1988). Plane and quasi-plane rotations of a heavy rigid body about a fixed
point. Izv. AN SSSR. Mekhanika Tverdogo Tela, 23(4), 29–36.
275. Markeev, A. P. (2000). The stability of the plane motions of a rigid body in the Kovalevskaya
case. Prikl. Mat. Mehk., 65, 51–58.
276. Markeev, A. P. (2003). On stability of regular precessions of a nonsymmetric gyroscope.
Regular and Chaotic Dynamics, 8(2), 297–304.
277. Markeev, A. P. (2003). The stability of the Grioli precession. J. Appl. Math. Meck., 67, 497–
510.
278. Markeev, A. P. (2005). On the Steklov case in rigid body dynamics. Regular and Chaotic
Dynamics, 10(1), 81–93.
279. Markeev, A. P. (2004). The pendulum-like motions of a rigid body in the Goryachev–
Chaplygin case. Prikl. Mat. Mehk, 68(2), 282–293.
280. Marsden, J. E., Ratiu, T. S., & Scheurle, J. (2000). Reduction theory and the Lagrange–Routh
equations. Journal of Mathematical Physics, 41, 3379–3429.
281. Mathematical Reviews. MR 93g: 01040.
282. Mathematical Reviews. MR 2003j:70005.
283. Mertsalov, N. I. (1946). The problem of motion of a rigid body, having a fixed point, with
A = B = 4C and the area integral = 0. Bull. Acad. Sci. URSS. Cl. Sci. Techn. (Izv. Akad.
Nauk SSSR, Otd. Tehn. Nauk), 697–701.
284. Minkowski, H. (1888). Über die Bewegung eines festen Kö rpers in einer Flüssigkeit. (pp.
1095–1110). Sitzungsber. Konig. Preuss. Akad. Wiss. Berlin.
285. Mlodz’evsky, B. C. (1896). On a case of motion of a heavy rigid body about a fixed point.
Mat. Sb., 18, 76–85.
286. Moiseyev, N. N., & Rumyantsev, V. V. (1968). Dynamic stability of bodies containing fluid.
Berlin: Springer.
287. Morozov, P. V. (2002). The Liouville classification of integrable systems of the Clebsch case.
Sb. Math., 193, 1507–1534.
288. Morozov, P. V. (2004). Topology of Liouville foliations in the Steklov and the Sokolov inte-
grable cases of Kirchhoff’s equations. Sb. Math., 195, 369–412.
289. Morozov, P. V. (2008). A fine Liouville classification of the Kovalevskaya-Yehia integrable
case. Moscow University Mathematics Bulletin, 63, 48–56.
290. Moser, J. (1980). Various aspects of integrable Hamiltonian systems. Progress in Mathematics,
Birkhauser, 8, 233.
291. Mozalevskaya, G. V. (1970). A particular solution of the problem of motion of a gyrostat.
Mekhanika Tverdogo Tela, Kiev, 2, 23–26.
292. Mozalevskaya, G. V. (1988). On a system of equations of rigid body dynamics. Mekhanika
Tverdogo Tela, Kiev, 20, 41–46.
293. Nekrassoff (Nekrasov), P. A. (1896). Recherches analytiques sur un cas de rotation d’un solide
pesant autour d’un point fixe. Math. Ann., 74, 445–530.
294. Neumann, C. (1859). De problemate quodam mechanico, quod ad primam integralium ultra-
ellipticorum classen revocatur. Reine Angew. Math., 56, 46–63.
295. Nikolaenko, S. S. (2015). A topological classification of the Chaplygin systems in the dynam-
ics of a rigid body in a fluid. Mat. Sb., 205, 224–268.
296. Nikolaenko, S. S. (2016). Topological classification of the Goryachev integrable systems in
the rigid body dynamics. Sb. Math., 207, 113–139.
297. Nikolaenko, S. S. (2017). Topological classification of the Goryachev integrable systems
in the rigid body dynamics: Non-compact case. Lobachevsky Journal of Mathematics, 38,
1050–1060.
298. Olsson, O. (1908). Ett integrabelt enskildt fall af fasta kroppars rotation kring en fast punkt
under tyngdkraftens inverkan. Arkiv Mat. Astr. och Fys., 4(7), 1–32.
299. Olsson, O. (1909). Tillämpning af de hyperelliptiska funktionerna inom den materiella punk-
tens dynamik. Ark. Mat. Astr. Fys., 6, 1–27.
300. Orel, O. E., & Ryabov, P. E. (1998). Bifurcation sets in a problem on motion of a rigid body in
fluid and in the generalization of this problem. Regular and Chaotic Dynamics, 3(1), 82–91.
@seismicisolation
@seismicisolation
Bibliography 455
301. Oreshkina, L. N. (1989). An integrable case of M. Kharlamov’s equations. Mekhanika Tver-

dogo Tela, Kiev, 21, 1–11.
302. Osgood, V. F. (1923). On the gyroscope. Transactions of the American Mathematical Society,
22, 240–264.
303. Oshemkov, A. A. (1991). Fomenko invariants for the main integrable cases of the rigid body
motion equations. Advances in Soviet Mathematics, 6, 67–146.
304. Oshemkov, A. A., Ryabov, P. E., & Sokolov, S. V. (2017). Explicit determination of cer-
tain periodic motions of a generalized two-field gyrostat. Russian Journal of Mathematical
Physics, 24, 517–525.
305. Pars, L. A. (1964). A treatise on analytical dynamics. London: Heinemann.
306. Perelomov, A. M. (1981). Some remarks on the integrability of the equations of motion of a
rigid body in an ideal fluid. Functional Analysis and Its Applications, 15, 83–85.
307. Perelomov, A. M. (2002). Kovalevskaya top: An elementary approach. Theoretical and Math-
ematical Physics, 131, 612–620.
308. Poinsot, L. (1851). Théorie nouvelle de la rotation des corps. J. Math. Pures Appl., 16(9–130),
289–336.
309. Poisson, S. D. (1811). Traité de Mécanique. Paris, Bachelier, Imprimeur-Libraire pour les
Math. (Vol. 2).
310. Ratiu, T., & van Moerbeke, P. (1982). The Lagrange rigid body motion. Ann. Inst. Fourier,
32, 211–234.
311. Reyman, A. G., & Semenov-Tian-Shansky, M. A. (1987). Lax representation with a spectral
parameter for the Kowalewski top and its generalizations. Letters in Mathematical Physics,
14, 55–61.
312. Routh, E. J. (1889). On a theorem of Jacobi in dynamics. Quarterly Journal of Pure and
Applied Mathematics, 23, 34–45.
313. Routh, E. J. (1897). The elementary part of a treatise on the dynamics of a system of rigid
bodies, being part I of a treatise on the whole subject (6th ed.). London and New York: The
Macmillan Co. Reprinted by the Dover Publications, New York (1955).
314. Routh, E. J. (1892). The advanced part of a treatise on the dynamics of a system of rigid
bodies, being part II of a treatise on the whole subject (5th ed.). London and New York: The
Macmillan Co. Reprinted by the Dover Publications, New York (1955).
315. Rozenblat, G. M. (2019). Estimates of the average angular velocity of the precession of
Lagrange’s top. Doklady Physics, 64, 114–119.
316. Rubanovsky, V. N. (1967). Application of the small parameter method to equations of motion
of a body in a liquid. Vestn. MGU. Ser. Mat. Mekh., 3, 80–87.
317. Rubanovsky, V. N. (1968). Integrable cases of the problem of motion of a heavy rigid body
in a liquid. Dokl. Acad. Nauk USSR, 180, 556–559.
318. Rubanovsky, V. N. (1984). On the precessional-screw motions of a solid immersed in liquid.
Journal of Applied Mathematics and Mechanics, 48, 532–537; translation from Prikl. Mat.
Mekh., 48, 738–744.
319. Rubanovsky, V. N. (1985). On a new particular solution of the equations of motion of a heavy
solid in a liquid. Journal of Applied Mathematics and Mechanics, 49, 160–165; translation
from Prikl. Mat. Mekh., 49, 212–219.
320. Rubanovsky, V. N. (1988). Quadratic integrals of the equations of motion of a rigid body in
a liquid. Prikl. Mat. Mekh., 52, 312–322.
321. Rumyantsev, V. V. (1956). Stability of permanent rotations of a heavy rigid body. Prikl. Mat.
Mekh., 20, 51–66.
322. Ryabov, P. E. (1999). Bifurcation sets in an integrable problem on motion of a rigid body in
fluid. Regular and Chaotic Dynamics, 4(4), 59–76.
323. Ryabov, P. E. (2011). Explicit integration and topology of D.N. Goryachev case. Doklady
Mathematics, 84, 502–505.
324. Ryabov, P. E. (2013). Phase topology of one irreducible integrable problem in the dynamics
of a rigid body. Theoretical and Mathematical Physics, 176, 1000–1015.
@seismicisolation
@seismicisolation
456 Bibliography
325. Ryabov, P. E. (2014). Phase topology of a special case of Goryachev integrability in rigid
body dynamics. Sb. Math., 205, 1024–1044.
326. Ryabov, P. E. (2018). Abel-Jacobi equations for the integrable case of Kowalevski-Yehia
in rigid body dynamics at zero integral of areas. In The Seventh International Conference
“Geometry, Dynamics, Integrable Systems – GDIS 2018”, 5–9 June, Moscow.
327. Sadetov, S. T. (1990). Conditions of integrability for Kirchhoff’s equations. Vestn. MGU, ser.
Mat. Mekh., 3, 56–62.
328. Sadetov, S. T. (2000). The fourth algebraic integral of Kirchhoff’s equations. Journal of
Applied Mathematics and Mechanics, 64, 229–242.
329. Sanduleanu, S., & Petrov, A. (2017). Comment on “New exact solution of Euler’s equa-
tions (rigid body dynamics) in the case of rotation over the fixed point”. Archive of Applied
Mechanics, 87, 41–43.
330. Savchenko, A. Ya. (1977). Stability of stationary motions of mechanical systems. Kiev:
Naukova Dumka.
331. Shiff, P. A. (1903). On the equations of motion of a heavy rigid body having a fixed point.
Mat. Sbornik, 24, 169–177.
332. Shuster, M. D. (1993). A survey of attitude representations. Journal of the Astronautical
Sciences, 41, 439–517.
333. Slavina, N. S. (2013). Classification of the family of Kovalevskaya–Yehia systems up to
Liouville equivalence. Doklady Mathematics, 88, 537–540.
334. Slavina, N. S. (2013). Topological classification of the integrable systems of the Kovalevskaya-
Yehia type. Candidate thesis. Moscow University.
335. Sokolov, V. V. (2001). A new integrable case of Kirchhoff equation. Theoretical and Mathe-
matical Physics, 129, 31–37.
336. Sokolov, V. V. (2002). A generalized Kowalevski Hamiltonian and new integrable cases on
e(3) and so(4). In V. B. Kuznetsov (Ed.), Kowalevski property (CRM proceedings and lecture
notes) (pp. 307–315). Providence: American Mathematical Society.
337. Sokolov, V. V., & Tsiganov, A. V. (2002). Lax pairs for the deformed Kowalevski and
Goryachev–Chaplygin tops. Theoretical and Mathematical Physics, 131, 543–549.
338. Sokolov, V. V., & Tsiganov, A. V. (2002). Commutative Poisson subalgebras for Sklyanin
brackets and deformations of some known integrable models. Theoretical and Mathematical
Physics, 133, 1730–1743.
339. Springborn, B. A. (2000). The toy top, an integrable system of rigid body dynamics. Journal
of Nonlinear Mathematical Physics, 7, 386–410.
340. Sretensky, L. N. (1963). On certain cases of equations of motion of the gyrostat. Dokl. AN
USSR, 149, 292–294.
341. Sretensky, L. N. (1963). On certain cases of motion of a rigid body with a gyroscope. Vestn.
Mosk. Univ. Mat.-Mekh., 3, 60–71.
342. Stäckel, P. (1909). Die reduzierten Differentialgleichungen der Bewegung des schweren
unsymmetrischen Kreisels. Math. Ann., 67, 399–432.
343. Staude, O. (1894). Über permanente Rotationsaxen bei der Bewegung eines schweren Korpers
um einen festen Punkt. J. reine und angew. Math., 113(4), 318–334.
344. Steklov, V. A. (1893). On the motion of a rigid body in a liquid. Kharkov.
345. Steklov, V. A. (1895). On certain possible motions of a body in a liquid. Tr. Otd. Fiz. Nauk
O-va Lub. Est., 7(2), 10–21.
346. Steklov, V. A. (1896). A case of motion of a heavy rigid body with a fixed point. Trud. Otdel.
Fiz.Nauk Obshch-va Liub. Estestv., 8, 19–21.
347. Steklov, V. A. (1899). A new particular solution of the differential equations of motion of a
rigid body with a fixed point. Tr. Otd. Fiz. Nauk O-va Liub. Estest., 10(1), 1–3.
348. Steklov, V. A. (Stekloff, W.) (1902). Remarque sur un problé me de Clebsch sur le mouvement
d’un corps solide dans un liquide indefini et sur le probléme de M. Brun. C. R. Acad. Sci.,
135, 526–528.
349. Stepanova, L. A. (1974). Dependence of the main variables on time in Goryachev’s solution
of the problem of motion of a body with a fixed point. Mekh. Tverdogo Tela. Kiev., 7, 22–24.
@seismicisolation
@seismicisolation
Bibliography 457
350. Stepanova, L. A. (1990). On an unsuccessful attempt to defend the priority of national clas-
sics of Russian mechanics in the construction of exact solutions of problems in rigid body
dynamics. Mekh. Tverdogo Tela, Kiev, 22, 19–33.
351. Tatarinov, Y. V. (1974). Portraits of classical integrals of the problem of motion of a rigid
body about a fixed point. Vestn. MGU, ser. Mat. Mekh., 6, 99–105.
352. Thomson, W., & Tait, P. G. (1867). Treatise on natural philosophy. Vol. I, II. Oxford: Clarendon
Press.
353. Tisserand, M. F. (1872). Sur les mouvements relatifs à la surface de la Terre. Comptes rendus
hebdomadaires des séances de l’Acad émie des sciences, 74, 1567–1570.
354. Tisserand, M. F. (1891). Mécanique Céleste. Vol II.
355. Tkhai, V. N., & Schvigin, A. L. (2000). In V. Rumyantsev (Ed.), Problems of investigation
of stability and stabilization of motion. (pp. 149–157) Moscow: Computing Centre of the
Russian Academy of Science, Part 2.
356. Tkhai, V. N. (2000). The stability of regular Grioli precessions. J. Appl. Math. Mehk., 64,
811–819.
357. Troilo, R. (1971). Caratterizazione delle precessioni semiregolari con asse di precessione
verticale. Rend. Semin. Mat. Univ. Padova, 46, 329–337.
358. Tsiganov, A. V. (2002). Letter to the editor: “On the Kowalevsky–Goryachev–Chaplygin
gyrostat”. Journal of Physics A: Mathematical and General, 35, L305–L318.
359. Tsiganov, A. V. (2003). Separation of variables in the Kovalevskaya–Goryachev–Chaplygin
gyrostat. Theoretical and Mathematical Physics, 135, 651–658.
360. Tsiganov, A. V. (2004). On the Steklov - Lyapunov case in the rigid motion. Regular and
Chaotic Dynamics, 9, 77–90.
361. Tsiganov, A. V. (2005). On a family of integrable systems on S2 with a cubic integral of
motion. Journal of Physics A: Mathematical and General, 38, 921–927.
362. Tsiganov, A. V. (2008). On bi-Hamiltonian geometry of the Lagrange top. Journal of Physics
A: Mathematical and Theoretical, 41, 315212.
363. Tsiganov, A. V. (2012). New variables of separation for the Steklov–Lyapunov system.
SIGMA, 8, 012, 14 pp.
364. Vagner, E. A., & Dumin, V. G. (1975). On a class of periodic motions of a rigid body about
a fixed point. PMM: Journal of Applied Mathematics and Mechanics, 39, 927–929.
365. Vagner, E. A. (1977). On a family of periodic motions of a heavy solid with a fixed point.
PMM: Journal of Applied Mathematics and Mechanics, 41, 553–556.
366. Volterra, V. (1899). Sur la théorie des variations des latitudes. Acta Math., 22, 201–358.
367. Weber, H. (1879). Anwendung der Thetafunctionen zweier Verä nderlicher auf die Theorie
der Bewegung eines festen körpers in einer Flussigkeit. Math. Ann., 14, 173–206.
368. Whittaker, E. T. (1944). A treatise on analytical dynamics of particles and rigid bodies. New
York: Dover.
369. Wittenburg, J. (1977). Dynamics of systems of rigid bodies. Stuttgart: Teubner.
370. van der Woude, W. (1923). Über die Staudischen Kreiselbewegungen. Math. Z., 16, 170–172.
371. Yanxia, H. (2005). Techniques for searching first integrals by Lie group and application to
gyroscope system. Science in China. Series A, Mathematics, 48, 1135–1143.
372. Yehia, H. M. (1976). On the reduction of order of the equations of motion of a rigid body
about a fixed point. Moscow University Mechanics Bulletin, 31(5/6), 37–39.
373. Yehia, H. M. (=Iakh’ia Kh. M.) (1977). On periodic almost stationary motions of a rigid body
about a fixed point. PMM: Journal of Applied Mathematics and Mechanics, 41(3), 571–573.
374. Yehia, H. M. (1978). Qualitative analysis of rotation of a rigid body about a fixed point in an
axi-symmetric field of force. Ph.D. thesis, Moscow State University (MGU). Moscow.
375. Yehia, H. M. (=Yahya H. M.) (1981). Qualitative investigation of plane and almost plane
motions of a rigid body about a fixed point. PMM: Journal of Applied Mathematics and
Mechanics, 45(4), 454–458.
376. Yehia, H. M. (1981). On the stability of plane motions of a rigid body about a fixed point in
a Newtonian field. Moscow University Mechanics Bulletin, 36(3/4), 41–44.
@seismicisolation
@seismicisolation
458 Bibliography
377. Yehia, H. M. (1983). On the reduction of the order of equations of motion of gyrostat in an
axisymmetric field. J. Mécan. Th éor. Appl., 2, 451–462.
378. Yehia, H. M. (=Yakh’ya Kh. M.) (1985). New solutions of the problem of motion of a gyrostat
in potential and magnetic fields. Mosk. Univ. Mech. Bull., 40(5), 21–25.
379. Yehia, H. M. (1985). Analytical and qualitative investigations of certain problems in rigid
body dynamics. D.Sc. thesis. Moscow State University, Moscow.
380. Yehia, H. M. (1986). New integrable cases in the dynamics of rigid bodies. Mechanics
Research Communications, 13(3), 169–172.
381. Yehia, H. M. (1986). On the integrability of certain problems in particle and rigid body
dynamics. J. Mécan. Théor. Appl., 5(1), 55–71.
382. Yehia, H. M. (1986). On the motion of a rigid body acted upon by potential and gyroscopic
forces. I: The equations of motion and their transformations. J. Mécan. Théor. Appl., 5(5),
747–754.
forces. II: A new form of the equations of motion of a multiconnected rigid body in an ideal
incompressible fluid. J. M écan. Théor. Appl., 5(5), 755–762.
forces. III: The reduction of order and further transformations. J. Mécan. Théor. Appl., 5(6),
935–939.
385. Yehia, H. M. (1987). New integrable problems in the dynamics of rigid bodies II. Mechanics
386. Yehia, H. M. (1987). New integrable cases in the dynamics of rigid bodies III. Mechanics
387. Yehia, H. M. (Yakhya Kh. M.) (1987). New integrable cases in the problem of motion of a
gyrostat. Mosk. Univ. Mech. Bull., 42(4), 29–31.
388. Yehia, H. M. (1987). On the stability of plane motions of a heavy rigid body about a fixed
point. ZAMM. Z. angew. Math. Mech., 67(12), 641–648.
389. Yehia, H. M. (1988). Particular integrable cases in rigid body dynamics. ZAMM. Z. angew.
Math. Mech., 68(1), 33–37.
390. Yehia, H. M. (1988). Equivalent problems in rigid body dynamics - I. Celestial Mechanics,
41, 275–288.
391. Yehia, H. M. (1988). Equivalent problems in rigid body dynamics - II. Celestial Mechanics,
41, 289–295.
392. Yehia, H. M. (1989). New integrable cases in the dynamics of rigid bodies IV. Mechanics
393. Yehia, H. M. (1989). New integrable cases in the dynamics of rigid bodies V. Mechanics
394. Yehia, H. M. (1992). Generalized natural mechanical systems of two degrees of freedom with
quadratic integrals. Journal of Physics A: Mathematical and General, 25, 197–221.
395. Yehia, H. M. (1996). On a generalization of certain results of Goriatchev, Chaplygin and
Sretensky in the dynamics of rigid bodies. Journal of Physics A: Mathematical and General,
29, 8159–8161.
396. Yehia, H. M. (1996). New integrable problems in the dynamics of rigid bodies with the
Kowalevski configuration. I- The case of axisymmetric forces. Mechanics Research Commu-
nications, 23(5), 423–427.
397. Yehia, H. M. (1996). New integrable problems in the dynamics of rigid bodies with the
Kowalevski configuration. II- The case of asymmetric forces. Mechanics Research Commu-
nications, 23(5), 429–431.
398. Yehia, H. M. (1997). New generalizations of the integrable problems in rigid body dynamics.
399. Yehia, H. M. (1997). New integrable problems of motion of a rigid body with a particle
oscillating or bouncing in it. Mechanics Research Communications, 24(3), 243–246.
400. Yehia, H. M. (1998). New integrable problems of motion of a rigid body acted upon by
nonsymmetric electromagnetic forces. Journal of Physics A: Mathematical and General, 31,
5819–5825.
@seismicisolation
@seismicisolation
Bibliography 459
401. Yehia, H. M. (1998). A new integrable problem in the dynamics of rigid bodies. Mechanics
402. Yehia, H. M. (1999). New generalizations of all the known integrable problems in rigid body
dynamics. Journal of Physics A: Mathematical and General, 32, 7565–7580.
403. Yehia, H. M. (2000). Geometric transformations and new integrable problems in rigid body
dynamics. Journal of Physics A: Mathematical and General, 33, 4393–4399.
404. Yehia, H. M. (2000). Motions of a particle on a sphere and integrable motions of a rigid body.
405. Yehia, H. M. (2001). Equivalent mechanical systems with cyclic coordinates and new inte-
grable problems. ZAMP Z. angew. Math. Phys., 52, 289–316.
406. Yehia, H. M. (2001). On certain integrable motions of a rigid body acted upon by gravity and
magnetic fields. International Journal of Nonlinear Mechanics, 36, 1173–1175.
407. Yehia, H. M. (2001). Transformations of mechanical systems with cyclic coordinates and new
integrable problems. Journal of Physics A: Mathematical and General, 34, 11167–11183.
408. Yehia, H. M. (2002). Relation between the first and second moments of distributions. Journal
of Physics A: Mathematical and General, 35, 6505–6508.
409. Yehia, H. M. (2002). On certain two-dimensional conservative mechanical systems with cubic
second integral. Journal of Physics A: Mathematical and General, 35, 9469–9487.
410. Yehia, H. M. (2002). Comment on ‘On the Kowalevsky–Goryachev–Chaplygin gyrostat’.
411. Yehia, H. M. (2003). Kowalevski’s integrable case: Generalizations and related new results.
Regular and Chaotic Dynamics, 8, 337–348.
412. Yehia, H. M. (2003). An integrable motion of a particle on a smooth ellipsoid. Regular and
Chaotic Dynamics, 8, 463–468.
413. Yehia, H. M. (2006). The Master integrable two-dimensional system with a quartic second
integral. Journal of Physics A: Mathematical and General, 39, 5807–5824.
414. Yehia, H. M. (2007). Atlas of two-dimensional irreversible conservative Lagrangian mechan-
ical systems with a second quadratic integral. Journal of Mathematical Physics, 48, 082902.
415. Yehia, H. M. (2012). A new 2D integrable system with a quartic second invariant. Journal of
Physics A: Mathematical and Theoretical, 45, 395209 (12pp.).
416. Yehia, H. M. (2014). New solvable problems in the dynamics of a rigid body about a fixed
point in a potential field. Mechanics Research Communications, 57, 44–48.
417. Yehia, H. M. (2015). On the regular precession of an asymmetric rigid body acted upon by
uniform gravity and magnetic fields. EJBAS., Egyptian Journal of Basic and Applied Sciences,
200–205.
418. Yehia, H. M. (2017). Regular precession of a rigid body (gyrostat) acted upon by an irreducible
combination of three classical fields. JOEMS (Journal of the Egyptian Mathematical Society),
25, 216–219.
419. Yehia, H. M., & Bedwehy, N. A. (1987). Certain generalizations of Kowalevski’s case. Man-
soura Science Bulletin, 14(2), 373–386.
420. Yehia, H. M., & El-Hadidy, E. G. (2013). On the orbital stability of pendulum-like vibrations
of a rigid body carrying a rotor. Regular and Chaotic Dynamics, 18(5), 539–552.
421. Yehia, H. M., & Elmandouh, A. A. (2008). New integrable systems with a quartic integral
and new generalizations of Kovalevskaya’s and Goriatchev’s cases. Regular and Chaotic
Dynamics, 13, 56–68.
422. Yehia, H. M., & Elmandouh, A. A. (2011). New conditional integrable cases of motion
of a rigid body with Kowalevski’s configuration. Journal of Physics A: Mathematical and
Theoretical, 44, 012001.
423. Yehia, H. M., & Elmandouh, A. A. (2013). A new integrable problem with a quartic integral
in the dynamics of a rigid body. Journal of Physics A: Mathematical and Theoretical, 46,
142001 (8pp.).
424. Yehia, H. M., & Elmandouh, A. A. (2016). Integrable 2D time-irreversible systems with a
cubic second integral. Advances in Mathematical Physics, 2016, Article ID 8958747, (10 pp.).
@seismicisolation
@seismicisolation
460 Bibliography
425. Yehia, H. M., & Elmandouh, A. A. (2016). A new conditional integrable case in the dynamics
of a rigid body-gyrostat. Mechanics Research Communications, 78, 25–27.
426. Yehia, H. M., Hassan, S. Z., & Shaheen, M. E. (2015). On the orbital stability of the motion
of a rigid body in the case of Bobylev-Steklov. Nonlinear Dynamics, 80(3), 1173–1185.
427. Yehia, H. M., Saleh, E., & Megahid, S. F. (2015). New solutions of classical problems in rigid
body dynamics. Mechanics Research Communications, 69, 40–44.
428. Zhang, X. (2012). Comment on “On the polynomial integrability of the Kirchoff equations”.
Physica D, 241, 1417–1420.
429. Zentralblatt (Zbl 1025.70007).
430. Zhou, G.-Q. (2009). Rotational stability of a charged dielectric rigid body in a uniform mag-
netic field. PIERS: Progress in Electromagnetics Research, 11, 103–112.
431. Zhou, G.-Q., Guan, C., & Zhang, S.-L. (2009). Charge moment tensor and the rotation equation
of a charged rigid body in a uniform magnetic field. In PIERS Proceedings: Progress in
Electromagnetics Research Symposium, Beijing, China, 23–27 March (pp. 993–996).
432. Zhou, G.-Q., & Xiao, X. (2008). Dynamical problem of a rotational charged dielectric rigid
body in a uniform magnetic field. PIERS: Progress in Electromagnetics Research, 1, 229–240.
433. Zhou, G.-Q., Zhang, S.-L., & Guan, C. (2009). Natural introduction of charge moment tensor
and the Lagrangian of a rotational charged rigid body. In PIERS Proceedings: Progress in
Electromagnetics Research Symposium, Beijing, China, 23–27 March (pp. 997–1002).
434. Ziglin, S. L. (1983). Branching of solutions and non-existence of first integrals in Hamiltonian
mechanics. II. Functional Analysis and Its Applications, 17(1), 6–17.
435. Ziglin, S. L. (1997). On the absence of a real-analytic first integral in some problems of
dynamics. Functional Analysis and Its Applications, 31(1), 3–9.
@seismicisolation
@seismicisolation

Rigid Body Dynamics, A Lagrangian Approach

Uploaded by

Copyright:

Available Formats

Rigid Body Dynamics, A Lagrangian Approach

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rigid Body Dynamics, A Lagrangian Approach

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What topics are covered in the book?

What topics are covered in the book?

Advances in Mechanics and Mathematics

Rigid Body Dynamics

ISSN 1571-8689 ISSN 1876-9896 (electronic)

Mathematics Subject Classification: 70-XX, 70Exx, 70Hxx

© Springer Nature Switzerland AG 2022

The book is divided into two parts:

Exercises constitute an essential component of the book. A large number of them

Mansoura, Egypt Hamad Yehia

Part I The Elementary Part

2.7 Space and Relative Time Rates of Change of a Vector . . . . . . . . 34

4.3 Kowalevski’s Case (1888) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

Part II The Advanced Part

8.2 Planar Motion (Motion of the Body as a Physical

8.13 Dokshevich’s Second Case [64] (1970) . . . . . . . . . . . . . . . . . . . . . 182

9.7 Separable Potentials in Rigid Body Dynamics

10.14 Generalizations of Particular Solvable Cases . . . . . . . . . . . . . . . . 270

11.7 Maximal Reduction of the Order of the Equations

Appendix A: Some Useful Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415

1.1 The Moment of Mass—The Centre of Mass

1.1.1 Moments of a Mass Distribution

where dm is an infinitesimal mass element, r is the position vector of that element

where n 1 , n 2 , n 3 are non-negative integers and n = n 1 + n 2 + n 3 . In dynamics of

1.1.2 Centre of Mass

Let M be the total mass of the body

Obviously, M is positive and finite. The vector

1.2 Second Moments and Inertia Matrix of a Mass

1.2.1 Second Moments Matrix of Mass Distribution

Define the symmetric matrix Ī = ( I¯i, j )i,3 j=1

1.2.2 Inertia Matrix of Mass Distribution

In most dynamical considerations, we more frequently meet the inertia matrix

where δ is the unit matrix. This makes

so that the inverse of the relation (1.8) can be written as

1.3 Properties of the Inertia Matrix

1.3.1 The Triangle Inequalities

1.3.2 Theorem of Parallel Axes

In a similar way, we can show that

which can be stated as follows:

1.3.3 Ellipsoid of Inertia

Fig. 1.1 Body-element

= I11 α2 + I22 β 2 + I33 γ 2 + 2I12 αβ + 2I23 βγ + 2I13 αγ. (1.15)

I11 X 2 + I22 Y 2 + I33 Z 2 + 2I12 X Y + 2I23 Y Z + 2I13 X Z = 1. (1.16)

or, in expanded form,

λ3 − (I11 + I22 + I33 )λ2

I11 + I22 + I33 = A + B + C,

corresponding to the three eigenvalues λ = A, B, C. The planes ξ = 0, η = 0, ζ = 0

and regarding the inverse relation (1.8), we also have

I = diag( B̄ + C̄, C̄ + Ā, Ā + B̄). (1.24)

Remark 7 If the axes are arranged so that A B C, then Ā ≤ B̄ ≤ C̄.

Remark 8 The condition of axial dynamical symmetry A = B is equivalent to

1.3.4 The Gyration Ellipsoid

1.3.5 Representation of Principal Moments of Inertia

1.3.5.1 The First Method

1.3.5.2 The Second Method

Fig. 1.2 Representing

Fig. 1.3 Representing ratios

1.4 Relations Between the Centre of Mass and the Inertia