Notes CH 8
Notes CH 8
Notes CH 8
Chapter 8.
Lagrangian Mechanics
Examples
T=
and its potential energy is
1
1
mv 2 = m(x 2 + y 2 + z 2 ),
2
2
(8.1)
U (r) = U (x , y , z ).
(8.2)
L = T U.
(8.3)
U
L
T
L
=
= Fx ,
=
= mx = px .
(8.4)
x
x
x
x
In an inertial reference frame, Newtons second law reads Fx = p x . Differentiating
the second of the above equations and comparing with the first gives
L
d L
=
.
dt x
x
(8.5)
JRT 7.1
L
d L
=
,
y
dt y
L
d L
=
z
dt z
(8.6)
The path of the particle can be determined either by the Lagrange equations (and
the methods of the previous chapter) or by Newtons second law (chapter 2 of this
course.)
The equations (8.6) have the same form of the Euler-Lagrange equations
R (7.13 and
7.20) from the last chapter. Therefore, they imply that the integral S = L dt is
stationary for the path followed by the particle. This is stated by Hamiltons Principle:
The actual path which a particle follows between two points 1 and 2 in a given time
interval t1 to t2 , is such that the action integral
Z t2
L dt
S=
(8.7)
t1
JRT 7.1
So far, we have proved that for a single particle, the following three statements are
exactly equivalent:
1
qi = qi (r) for i = 1, 2, 3,
(8.8)
r = r(q1 , q2 , q3 ).
(8.9)
and
JRT 7.1
(8.10)
S=
t2
t1
L(q1 , q2 , q3 , q 1 , q 2 , q 3 ) dt .
(8.11)
The change of variables does not alter the value of S . Therefore, S must still be
stationary for the correct path in these new coordinates, which means that the
correct path must satisfy the Euler-Lagrange equations
L
d L
=
,
q1
dt q 1
L
d L
=
,
q 2
dt q 2
L
d L
=
q 3
dt q 3
(8.12)
1
m(x 2 + y 2 ) U (x , y ).
2
(8.13)
L
T
=
= mx ,
x
x
(8.14)
with corresponding expressions for the y derivatives. Thus the two Lagrange equations can be rewritten as:
Fx = mx
F = ma.
L
L
d
F
=
m
y
y
dt y
y
L
d L
= dt
x
x
(8.15)
JRT 7.1
Notice how in eq. (8.14) the derivative L/x is the x component of the force and
L/x is the x component of the momentum, and similarly for the y components.
When we use the generalized coordinates q1 , q2 , we shall find that L/qi ,
although not necessarily a force component, plays a role very similar to a force.
Likewise, L/q i plays a role similar to a momentum. They will be called generalized
force and generalized momentum respectively:
L
= i th component of generalized force,
q i
(8.16)
L
= i th component of generalized momentum.
(8.17)
q i
and thus, the generalized force equals the rate of change of generalized momentum.
This concept is illustrated in the following example.
Problem: Write down Lagranges equations for the same system as in the previous example, but using polar
coordinates.
Solution: As always, we start by writing down the Lagrangian, T U . In polar coordinates (see Fig. 1,) we know that the
and thus the KE is
velocity components are vr = r and v = r ,
(r , r )
= 1 m(r 2 + r 2 2 ). The Lagrangian is
T = 1 mv 2 = 1 mv v = 1 m(r , r )
2
= T U = 1 m(r 2 + r 2 2 ) U (r , ).
L = L(r , , r , )
2
Next, we examine individually the Lagrange equations for coordinates r and .
The r equation
The Lagrange equation for the coordinate r is
(8.18)
d L
L
=
.
(8.19)
dt r
r
Carrying out the appropriate derivatives of L with respect to r and r , and substituting them in the above Lagrange equation
gives
d
U
mr 2
=
(mr ) = mr .
(8.20)
r
dt
We know, however, that U /r is just Fr , and thus we can have the familiar result - a centripetal force - from chapter 1:
Fr = m(r r 2 ) = mar .
(8.21)
Fig. 1:
The velocity of a particle expressed in 2D polar coordinates - see JRT Fig. 7.1
d L
L
=
dt
(8.22)
d
U
(mr 2 )
(8.23)
=
dt
To calculate the left-hand side of this equation, we need to know the components of U in polar coordinates. This can be
determined by the methods discussed in chapter 1:
U =
1 U
U
r +
r
r
(8.24)
1 U
.
(8.25)
r
The left side of eq. (8.23) is just rF , which is equal to the torque on the particle about the origin. Meanwhile, the quantity
mr 2 = mr 2 is the angular momentum L about the origin. Therefore, the equation has reproduced the familiar relation
F =
dL
.
dt
(8.26)
Lagranges Equation for Unconstrained Motion Several Unconstrained Particles JRT 7.1
The extension of these concepts to a system of N unconstrained particles is fairly
straightforward. For N = 2, the Lagrangian becomes
1
1
2
2
m1 r 1 + m2 r 2 U (r1 , r2 ).
(8.27)
2
2
The forces on the two particles are F1 = 1 U and F2 = 2 U . Newtons second
law can be applied to each particle, yielding the six equations
L(r1 , r2 , r 1 , r 2 ) =
(8.28)
(8.30)
Lagranges Equation for Unconstrained Motion Several Unconstrained Particles JRT 7.1
One example of a set of six generalized coordinates will be used repeatedly in an
upcoming chapter, which deals with two-body central-force problems. In place of the
six coordinates of r1 and r2 , we shall use the three components of the center of
mass position R = (m1 r1 + m2 r2 )/(m1 + m2 ) and the three coordinates of the
relative position r = r1 r2 .
In the case of N unconstrained particles, there are 3N Lagrange equations:
L
d L
=
,
q i
dt q i
i = 1, 2, , 3N .
(8.31)
Fig. 2:
A simple pendulum. The position of the mass m can be specified by the single coordinate - see JRT Fig. 7.2
JRT 7.2
= T U = 1 m2 2 mg (1 cos ).
L = L(, )
(8.32)
2
Now, for a single particle we have (for a generalized coordinate q and our specific
coordinate )
d L L
d L
L
=
,
.
(8.33)
=
dt q
dt
q
d
= m2 .
(m2 )
(8.34)
dt
Referring back to the figure, we see that the left side of this equation is just the
torque exerted by gravity on the pendulum, while the term m2 is the pendulums
moment of inertia about the pivot (point mass m, a distance from the pivot.) Since
is the angular acceleration , we have shown that = I. You will recall from
qi = qi (r1 , , rN , t )
[i = 1, , n] .
(8.36)
JRT 7.3
A slightly more complex system is the double pendulum, shown in Fig. 3. This
system has two bobs, both confined to a plane, so it has four Cartesian coordinates,
which can be represented in terms of two generalized coordinates 1 and 2 . If we
put our origin at the suspension point of the top pendulum, then we have
r1 = (1 sin 1 , 1 cos 1 ) = r1 (1 )
(8.38)
r2 = (1 sin 1 + 2 sin 2 , 1 cos 1 + 2 cos 2 ) = r2 (1 , 2 ).
Notice that the components of r2 depend on both of the generalized coordinates 1
and 2 . If the relation between the Cartesian coordinates r and the generalized
coordinates qi does not involve the time t - as with the double pendulum - then the
generalized coordinates are said to be natural; later on, we will see that natural
coordinates allow certain simplifying approximations.
(8.39)
(Note that if all of the nonconstraint forces are conservative, then U is independent
of t , but we dont need to make this assumption right now.) The total force on the
particle is Ftot = Fcstr + F, and the Lagrangian, as usual, is L = T U . Since U is the
potential energy for the nonconstraint forces only, this definition of L excludes the
constraint forces.
Proof of Lagranges Equation with Constraints Stationarity of Action Integral JRT 7.4
Consider any two points r1 and r2 , through which the particle passes at times t1 and
t2 . If r(t ) is the correct path that the particle will follow, then we can denote a
neighbouring incorrect path by
R(t ) = r(t ) + (t ),
(8.40)
which defines (t ) as the infinitesimal vector pointing from r(t ) to R(t ). It is logical to
assume that both r(t ) and R(t ) lie in the surface to which the particle is confined,
and therefore so does (t ). Furthermore, since the endpoints of the path are fully
specified, (t )=0 at t1 and t2 .
Now, we denote by S the action integral taken along the incorrect path:
Z t2
t )dt
S=
L(R, R,
(8.41)
t1
(8.42)
Proof of Lagranges Equation with Constraints Stationarity of Action Integral cont JRT 7.4
We will now prove that S is stationary for variations of the path R(t ) when R(t ) = r(t )
(or equivalently, when (t ) = 0.) Another way to say this is that the difference in the
action integrals
S = S S0
(8.43)
is zero to first order in the distance between the paths.
The difference S is the integral of the difference between the Lagrangian on the two
paths,
t ) L(r, r , t ).
L = L(R, R,
(8.44)
If we substitute R(t ) = r(t ) + (t ) and
L(r, r , t ) = T U =
1 2
mr U (r, t ),
2
(8.45)
this becomes
L
=
=
i
1 h
2 r 2 [U (r + , t ) U (r, t )]
m (r + )
2
mr U + O (2 ),
(8.46)
where O (2 ) denotes 2nd- and higher-order terms in and 2 . The last step in the
above equation results from the definition of the gradient, f (r + ) f (r) f .
Proof of Lagranges Equation with Constraints Stationarity of Action Integral cont JRT 7.4
We then find that
S =
t2
L dt =
t1
t2
t1
[mr U ] dt .
(8.47)
(8.48)
t1
Now, remember that the path r is the correct path, which means that it satisfies
Newtons second law. Therefore, mr is just the total force on the particle,
Ftot = Fcstr + F. But we also know that U = F. Cancellation of these terms leaves
Z t2
S =
(8.49)
Fcstr dt .
t1
However, the constraint force Fcstr must be normal to the surface in which the particle
moves, while lies in the surface. Thus, their dot product is zero, and we have
proved that S = 0; that is, the action integral is stationary at the correct path.
T=
Fig. 4:
1
1
1
mv 2 = mv v = m(R 2 2 + z 2 ).
2
2
2
(8.50)
A mass m, confined to the surface of the cylinder = R and subject to a force F = k r - see JRT Fig. 7.7
z , z ) = 1 m(R 2 2 + z 2 ) 1 k (R 2 + z 2 ).
L = L(, ,
2
2
(8.51)
With two DOF, this system has two equations of motion. The two Lagrange equations read
d L
L
=
,
dt z
z
or
kz = mz,
L
d L
=
,
dt
or 0 =
mR 2 .
dt
(8.52)
The z equation tells us that the mass executes simple harmonic motion in the z direction, with the usual = k /m. The
The particles motion would be
equation tells us that the mass moves around the cylinder with constant angular velocity .
Examples cont
JRT 7.5
These previous examples illustrated the procedure by which you can solve problems
by the Lagrangian method. Provided that the constraints are holonomic and that the
nonconstraint forces are derivable from a potential energy, the procedure is:
1
Write down the kinetic and potential energies and the Lagrangian
L = T U , using any convenient inertial reference frame.
We cant guarantee that all of the Lagrange equations are easy to solve, but having
them in this form usually facilitates understanding of the system under consideration.
Next, we will take a look at a couple of examples for which the Lagrangian approach
simplifies matters considerably.
3
mx 2 + mgx ,
4
(8.54)
which has just the one generalized coordinate, x . The Lagrange equation reads
d L
L
=
dt x
x
mg =
3
mx,
2
Fig. 5:
(8.55)
Therefore,
mg sin = m + I /R 2 x.
x =
mg sin
,
m + I /R 2
(8.57)
(8.58)
Isph =
the sphere will win the race.
2
1
mR 2 and Icyl = mR 2 ,
5
2
(8.59)
(8.61)
We can immediately see that the difference of these two equations yields the required conservation of momentum:
Fig. 6:
(8.62)
(8.64)
= 21 m(x 2 + ( + x )2 2 ).
As for the potential energy, there are both gravitational and spring components. Using the position of the support as the
reference position for Ug , we have
1
U = mg ( + x ) cos + kx 2 .
(8.65)
2
Fig. 7:
1
1
m(x 2 + ( + x )2 2 ) + mg ( + x ) cos kx 2 .
2
2
(8.66)
(8.67)
mg cos kx = m(x ( + x ) 2 ).
(8.68)
=
mg ( + x ) sin = m( + x )2 2 + 2m( + x )x ,
dt
(8.69)
h
i
mg sin = m ( + x ) 2x .
(8.70)
This is Newtons second law in the tangential direction, F = ma . The two terms on the right-hand side represent the
angular acceleration and the Coriolis acceleration, which we wont encounter until later in the course - it occurs when an
object moves in a rotating frame, which is exactly the case here.
We cant proceed any further with this problem, since our Lagrange equations are coupled and nonlinear. In principle, they
could be solved using perturbation approaches, but one would normally use numerical techniques.
(8.71)
can be rewritten as Fi = dtd pi . That is, the generalized force equals the rate of
change of generalized momentum. In particular, if the Lagrangian is independent of
= 0, and the corresponding generalized
a particular coordinate qi , then Fi = L
qi
momentum pi is constant.
Consider a single projectile moving in the vertical direction, subject only to gravity.
The potential energy (measuring z vertically up) is U = mgz , and the Lagrangian is
1
(8.72)
L = L(x , y , z , x , y , z ) = m(x 2 + y 2 + z 2 ) mgz .
2
In this case, the generalized force is just the usual force (L/x = U /x = Fx ,
etc.) and the generalized momentum is just the usual momentum
(L/x = mx = px , etc.) Because L is independent of x and y , it follows from
Lagranges equations that the components px and py are constant, as we already
knew.
JRT 7.6
In general, the generalized forces and momenta are not the same as the usual
forces and momenta. For instance, we saw in a previous example that in 2D polar
coordinates, the component of the generalized force is the torque, and that of the
generalized momentum is actually the angular momentum. In any case, when the
Lagrangian is independent of a coordinate qi , the corresponding generalized
momentum is conserved. When the Lagrangian is independent of a coordinate qi ,
that coordinate is said to be ignorable or cyclic. It is always a good idea to see if you
can choose a coordinate system in which as many coordinates as possible are
ignorable.
Here, we will see how the laws of conservation of momentum and energy fit into the
Lagrangian formulation of mechanics. We already know that these laws hold, since
we derived the Lagrangian formulation from the Newtonian. However, much insight
can be gained from examining these laws strictly from a Lagrangian viewpoint.
JRT 7.8
U (r1 + , , rN + , t ) = U (r1 , , rN , t ),
(8.73)
that is, U = 0. Clearly, the velocities are unchanged by the translation. Therefore,
T = 0 and hence L = 0 under the translation. This is true for any . If we choose
to be an infinitesimal displacement in the x direction, then all of the x coordinates
x1 , , xN increase by , while the y and z coordinates are unchanged. For this
translation,
L
L
= 0.
(8.74)
+ +
L =
x1
xN
Since is arbitrary, this implies that
N
X
L
= 0.
x
=1
(8.75)
But using Lagranges equations, we can rewrite each derivative in the sum as
L
d L
d
=
= px
x
dt x
dt
where px is the x component of the momentum of particle .
(8.76)
JRT 7.8
x
dt
dt
=1
=1
(8.77)