Reciprocity Principle and the Lorentz Transformations

Vittorio Berzi and Vittorio Gorini

Citation: J. Math. Phys. 10, 1518 (1969); doi: 10.1063/1.1665000
View online: http://dx.doi.org/10.1063/1.1665000
View Table of Contents: http://jmp.aip.org/resource/1/JMAPAQ/v10/i8
Published by the American Institute of Physics.

Additional information on J. Math. Phys.

Journal Homepage: http://jmp.aip.org/
Journal Information: http://jmp.aip.org/about/about_the_journal
Top downloads: http://jmp.aip.org/features/most_downloaded
Information for Authors: http://jmp.aip.org/authors

Downloaded 19 Nov 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

VOLUME 10, NUMBER 8

JOURNAL OF MATHEMATICAL PHYSICS

AUGUST 1969

Reciprocity Principle and the Lorentz Transformations

YfITORIO BERZI AND YfITORIO GORINI

Istituto di Scienze Fisiche dell' Universita, Milano, Italy, and Istituto Nazionale
di Fisica Nuc/eare, Sezione di Milano, Milan, Italy

(Received 7 June 1968)

By using the principle of relativity, together with the customary assumptions concerning the nature of
the space-time manifold in special relativity, namely, space-time homogeneity and isotropy of space, a
simple but rigorous proof is given of the reciprocity relation for the relative motion of two inertial frames
of reference, which is usually assumed as a postulate in the standard derivations of the Lorentz transformations without the principle of invariance of light velocity. A critical discussion is set forth of the
question of eliminating the transformations with invariant imaginary velocity, which one unavoidably
obtains together with the Lorentz transformations and the Galilean ones in adopting a procedure of this
kind.

nature of the space-time manifold in special re!ativity,

namely, its homogeneity and the isotropy of physical
Since the appearance of the classical Einstein
space, point towards the existence of a universal
paper,I in which the foundations of the theory of
constant which has the meaning of an invariant
relativity were first laid down, several other derivations
velocity, so that there is no need to introduce this
of the Lorentz transformations have been published
constant into the theory at the beginning.
in the attempt to throw full light on the underlying
Without imposing from the outset the principle of
principles and to clarify both the physical content
constancy of light velocity, many of the existing
and the mathematical implications of the latter.2 In
standard derivations of the Lorentz transformations
particular, it has been shown as far back as 1911 by
make more or less explicit use of the so called reciproc2a
Frank and Rothe that the assumption of the existity principle which, as is well known, states simply
ence of an invariant velocity is not necessary in order
to arrive at the correct transformation equations. that the velocity of an inertial frame of reference S
This is rather a remarkable result, since it shows that with respect to another inertial frame of reference S'3
is the opposite of the velocity of S' with respect to S.
the principle of relativity (which establishes the
The use of this principle is not strictly necessary
equivalence of all inertial frames of reference in regard
to the scope, but it has the advantage of greatly
to the description of physical phenomena) together
with the customary assumptions concerning the simplifying the derivation of the transformation
equations, which would otherwise require rather
lengthy calculations and the resort to nonelementary
1 A. Einstein, Ann. Phys. 17, 891 (1905).
The existing literature is very wide and rather unrelated and it
results of the theory of Lie transformation groups.4
would be almost impossible to give a fairly complete summary of it.
It
appears, however, that in the existing literature no
We draw attention to the following references: (a) P. Frank and H.
Rothe, Ann. Phys. 34, 825 (191 I); (b) L. A. Pars, Phil. Mag. 42,249
sufficiently convincing arguments have been put
(1921); (c) A. S. Eddington, The Mathematical Theory of Relativity
forward to justify the use of the reciprocity principle.
(Cambridge University Press, London, 1923), Sec. 4; (d) Y. Mimura
and T. Iwatsuki, J. Sci. Hiroshima Univ. AI, III (1931); (e) V. V.
Indeed, it is generally assumed as a justification that
Narliker, Proc. Cambridge Phil. Soc. 28, 460 (1932); (f) G. J.
the reciprocity relation is a consequence of the
Whitrow, Quart. J. Math. 4, 161 (1933); (g) L. R. Gomes, Lincei
Rend. 21, 433 (1935); (h) F. Severi, Proc. Phys.-Math. Soc. Japan 18,
principle of relativity, whereas the latter merely
257 (1936); (i) E. EscIangon, Compt. Rend. 202, 708 (1936); (j) E.
implies the invariance of the relation between direct
Le Roy, ibid. 202, 794 (1936); (k) Y. Lalan, ibid. 203, 1491 (1936);
Bull. Soc. Math. France 65, 83 (1937); (I) G. J. Whitrow and E. A.
and reciprocal velocity.
Milne, Z. Astrophys.15, 270 (1938); (m) G. Temple, Quart. J. Math.
It is the aim of the present paper to give a simple
9, 283 (1938); (n) H. E. Ives, Proc. Am. Phil. Soc. 95, 125 (1951);
(0) K. D. Stiegler, Compt. Rend. 234, 1250 (1952); (p) A. W. Inglebut
rigorous deduction of the reciprocity relation,
ton, Nature 171, 618 (1953); (q) J. Aharoni, The Special Theory of
starting from the three basic postulates of the special
Relatitity (Oxford University Press, London, 1965), Chap. 1; (r) V.
Fock, The Theory of Space, Time and Gravitation (Pergamon Press
theory of relativity, namely, the principle of equivLtd., London, 1959), Chap. 1 and Appendix A; (s) H. M. Schwartz,
alence of inertial frames, the homogeneity of spaceAm. J. Phys. 30, 697 (1962); Introduction to Special Relativity
(McGraw-Hill Book Co., New York, 1968), Chap. 3; (t) E. C.
time, and the isotropy of space. 5
I. INTRODUCTION

Zeeman, J. Math. Phys. 5, 490 (1964); (u) R. Weinstock,Am.J.Phys.

32, 261 (1964); 33, 640 (1965); 35, 892 (1967); (v) V. Mitvalsky,
ibid. 34, 825 (1966); (w) E. Drake, ibid. 34, 899 (1966); (x) J. L.
Strecker, ibid. 35,13 (1967); (y) L. J. Eisenberg, ibid. 35, 649 (/967);
(z) H. A1mstrom, J. Phys. A (Proc. Phys. Soc.) 1,331 (/968); (aa)
Ya. P. Terletskii, Paradoxes in the Theory of Relativity (Plenum
Press, Inc., New York,1968), Chap. 2.

3 See, for example, Refs. 2(h-j, q, s, u-w, z).

See, for example, Refs. 2(a, k).
A critical analysis of the literature quoted in Footnote 2 and a
general discussion concerning the axiomatic derivation of the
extended inhomogeneous Lorentz group is the subject of a forthcoming paper.

1518

Downloaded 19 Nov 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

RECIPROCITY PRINCIPLE, LORENTZ TRANSFORMATIONS

Once the reciprocity relation has been established,

the transformation equations can easily be deduced,
as is well known, by making use of their group
property, which follows from the principle of relativity.6 Nevertheless, it is not superfluous to present
this deduction again here, since this gives us an
opportunity to discuss critically the arguments which
have been put forward in favor of excluding the
Lorentz transformations with imaginary invariant
velocity, which one unavoidably obtains together
with the Galilei and the ordinary Lorentz transformations by following a procedure of this kind.
We confine ourselves throughout this paper to the
consideration of a two-dimensional space-time and to
transformations which conserve the space-time origin.
This implies no loss ofgenerality, since any transformation can always be reduced to a homogeneous velocity
transformation along an axis by means of a suitable
space-time translation, together with suitable rotations of the space-axes of the two observers.

transformation which connects S to S' as

$' = f($)

(i) the principle of relativity, which states the

equivalence of all inertial systems as regards the
formulation of the laws of nature;
(ii) the homogeneity of space-time;
(iii) the isotropy of space.
We denote by x the position at which an event takes
place and by t the time at which it happens, as
viewed by an inertial observer S, and by x', t' the
corresponding space-time coordinates of the same
event, as viewed by another inertial observer S'.
The homogeneity assumption comes into our
considerations in that it implies that the transformation equations which furnish x' and t' as functions of
x and tare linear.7
In order to prove this assertion, let us employ the
notation $ for the two-vector (x, t) and write the
See, for example, Refs. 2(q, s, aa).
The question of the linearity of the transformation formulas has
long been debated in the literature [see, for example, Refs. I, and
2(d-g, k, m, r, s, y, aa)]. If one does not impose from the outset the
existence of an invariant velocity, then the principle of inertia, which
implies that a motion which appears uniform to an inertial observer
(x = all + a2) must appear uniform to any other inertial observer
(x' = a~I' + a;), is not sufficient to ensure that the transformations
are linear. To obtain this result, an additional assumption is needed;
namely, that an event of finite space-time coordinates is transformed
into an event of finite space-time coordinates. Further, one has to
require that the transformation functions be differentiable up to the
third order (compare Ref. 2r). It was Einstein (Ref. I) who first
justified the linearity property by an appeal to space-time homogeneity. The argument, however, is rather obscure. Here we give a
simple proof of linearity which utilizes a formulation of the homogeneity principle first given by Lalan in Ref. 2k and which is
particularly appealing.
6

(1)

Since we have decided to restrict our considerations

to transformations which conserve the space-time
origin, we should require f(O) = O. Here, however,
this condition is dropped for the sake of generality.
The homogeneity of space-time requires that a
space-time translation T not affect the relation
between the two observers and thus leaves Eq. (1)
invariant. Denoting by Ta and T a, the representations
of T relative to Sand S', respectively, we express this
property by the relation
(2)

or
I

+ <X) = fm + <X',

f(~

(3)

where <X = (<Xx, <x t ), <X' = (<Xx" <Xt'), and <X' depends on
fand <X but not on $.
Taking $ = 0 in (3), we get

II. THE RECIPROCITY RELATION

As indicated in the introduction, we start from the

following assumptions:

1519

I(<x) = 1(0)

+ <X'.

(4)

Substituting (4) into (3), we obtain

1($

+ <X) = 1($) + f(<x)

- 1(0).

(5)

Subtracting 1(0) from both sides and setting g( $) =

1($) - 1(0), we have
g($

+ <X)

= g($)

+ g(<x),

(6)

for arbitrary $ and <X. From this equation, provided

we only assume that g is continuous at the origin, we
get that
g(k~)

kg($),

(7)

where k is a real number. The' proof is quite standard

and is given in Appendix A.
Relations (6) and (7) state that g is linear and
homogeneous.
We thus write the relation between the pair (x, t)
and the pair (x', t') in the form
x' = a(v)x

+ b(v)t,

t' = c(v)x

+ d(v)t,

(8)

where v denotes the velocity of the frame S' with

respect to the frame S. This velocity, which we call the
direct velOCity for the pair (S, S'), is given by
v

-b(v)/a(v).

(9)

For the sake of simplicity, we confine ourselves in

the rest of this section to the consideration of the case
when the space axes ofthe two observers have the same
orientation and their times flow in the same direction,

Downloaded 19 Nov 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

v.

1520

BERZI AND V. GORINI

which implies the relations8

ox'jox = a(v)
ot'jot = d(v)

=
ot'jot =

oxjox'

> 0,
> 0,

(lOa)

+ vc(v))] > 0,
I/(d(v) + vc(v > O.

d(v)j[a(v)(d(v)

(lOb)

Then, supposing that both observers use the same

unit of time and the same unit of length, 9 the coefficients of (8) are uniquely determined functions of v
which, by the principle of relativity, do not depend
on S.
Denoting by w the reciprocal velocity, namely, the
velocity of S with respect to S', we have

w = b(v)/d(v)

lP(v) ,

(11)

and our purpose then is to show that the principles of

relativity and of isotropy of space, together with some
continuity assumptions to be specified later, are
sufficient by themselves to arrive at the conclusion
that
lP(v) = -v.
(12)
The principle of relativity implies that the set r
of the allowed velocities of S' relative to S does not
depend on S and that the reciprocal velocity is the
same function of the direct velocity for all pairs of
inertial systems. Hence, together with (11), we can
write
v = lP(w)
(13)
or
1P(IP(v = v.

-v/[I - (v/c)],

(15)

which is pertinent to one of the cynematics which are

compatible with the principle of relativity, 2k satisfies
relation (14) without having the form (12).
Note that conditions (lOb) are a priori independent of (lOa).
It is easy to devise conceptual experiments by which the standards
of length and time of the two observers can be made the same.
For example. we can make sure that Sand S' use the same time
standard. if both observers assume as unit of time the mean life of a
given unstable particle measured at rest in the laboratory of each of
the two observers.
8

x'

= a(v)x + b(v)f,

f' = c(v)x

+ d(v)f,

(16)

where v is the velocity of S' relative to S. On the other

hand, x' = -x', f' = t', x = -x, f = t, so that

= a(v)x - b(v)f,
f' = -c(v)x + d(v)f,

x'

(14)

Since w E r, it is clear from (13) that the range of the

function IP is equal to its domain r. Then IP is a oneto-one mapping of r onto r. Indeed, if lP(a) = lP(b) ,
we get from (14) that a = 1P(IP(a = 1P(IP(b = b.
Contrary to a widely held opinion,3 Eq. (14) is
the only condition imposed by the principle of relativity on the function IP. This condition is already
strongly restrictive on the possible forms of IP, but it
by no means alone implies relation (12). For example,
the equation

w = lP(v)

It is precisely with the hope of eliminating solutions

of this kind that we resort to the principle of isotropy
of space. The main result of applying this principle
is the proof that IP is an odd function of v and we see
that this property, together with (14) and a physically
reasonable assumption concerning the domain of IP
and its continuity properties, is sufficient to obtain the
result that the reciprocal velocity is given by (12).
In our case, space is one-dimensional and its
isotropy means that no one orientation along the
x axis should be considered in preference to the other.
This assertion is now made precise by stating the
isotropy principle in two equivalent forms. The first
has a more formal character and concerns the effect
that the inversion of the space axes has on the set of
transformations (8). The second one, which might be
physically more appealing, is based on simple conceptual experiments of a type frequently employed in
discussions of the theory of relativity.
We state the isotropy principle in the first form by
asserting that if two frames Sand S' are connected by
a transformation (8), then the two frames Sand S'
obtained from the preceding ones by inverting the
direction of the x axis are connected by a transformation of the same type. Therefore,

(17)

from which we conclude that v = b(v)/a(v) = -v.

Hence, r is symmetric and, by comparison with (17),
a( -v) = a(v),
b( -v) = -b(v),
c( -v) = -c(v),

(18)

d( -v) = d(v).

Then, by (11),
IP( -v) = -1P(v),

(19)

i.e., IP is an odd function of v.

Consider now the following conceptual experiments:
I. Let T be a rod at rest in S', the end points of
which occupy the positions x~ and x~ . S measures the
length of Tby marking the positions Xl and X 2 that the
end points of the rod occupy at a given time t. From
the first of the equations in (8) and, again, in (lOa) we
see that the ratio between the length [' of the rod at

Downloaded 19 Nov 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

RECIPROCITY PRINCIPLE, LORENTZ TRANSFORMATIONS

rest (as measured by S') and the length I of the rod in

motion (as measured by S) is given by
l'/l=a(v).

(20)

2. Next, let <l> be a phenomenon which takes place

at the point x and lasts from time tl to time t2 as
observed by S (e.g., we may think of the life of an
unstable particle produced at rest at x at time 11 and
decaying at time (2). By the second of the equations in
(lOa), the same phenomenon, as observed by Sf,
starts at x~ at time t~ and ends at x~ at time t~, where
(x~ , t~) and (x~, t~) are the transformed coordinates of
(x, t I ) and (x, t2), respectively (in our example, the
particle is produced in flight at x~ at time t~ and then
moves to point
where it decays at time I~). From
the second of the equations in (8) we see that the ratio
between the durations Df and D of <l> (lifetimes of the
particle) as measured by S' and S, respectively, is
given by
(21)
D'jD = d(v).

x;,

3. Finally, let <il be another phenomenon which

takes place at the point x' and lasts from time I~ to
time t~ as observed by Sf, and let (Xl' t I ) and (x 2 , t 2 )
be the transformed coordinates of (x', t~) and (x', t~),
respectively. The duration of <D as measured by S
can be determined by means of the following equations:
a(v)xl + b(v)tl = a(v)x2 + b(V)t2'
t; = C(V)xl + d(v)tI'
(22)
t~ = C(V)X2 + d(v)t 2 ,
and, by (9) and by the second of the equations in (lOb),
the ratio of the durations lY and 15 of <D as measured
by the two observers is readily seen to be

15' /15 = d(v) + ve(v).

(23)
We state the second version of the principle of isotropy of space by assuming that if v is an allowed
velocity, -v is allowed as well (hence the symmetry
of r), and by requiring that the ratios (20), (21), and
(23) are independent of the direction of the motion of
S' relative to S, provided that the magnitude of the
velocity remains the same, and thus are left unaltered
when V is changed to _V.l0 This condition implies
a( -v) = a(v),
de-v) = d(v),
d( -v) - ve( -v) = d(v) + ve(v).

(24)

Taking into account (9), we see that relations (24) are

equivalent to relations (18), and then (19) follows.
10 The same requirement, as regards only the ratio (20), was
originally imposed by Frank and Rothe in their derivation of the
Lorentz transformations (cf. Ref. 2a).

1521

We make two further assumptions before we

derive the reciprocity relation. These are:
(a) The domain r of the function cp is an interval
on the real line ;
(b) cp is continuous on r.
In other words, it is assumed that if VI and V 2 are
two allowed velocities, any velocity V which is comprised between VI and V z is again allowed, and that the
reciprocal velocity is a continuous function of the
direct velocity. The physical plausibility of these two
conditions is obvious.
Since cp is a continuous one-to-one mapping of r
onto itself and r is connected, then, from a well-known
theorem of analysis, we can state that cp is either a
strictly increasing or a strictly decreasing function
ofv.u
Suppose first that cp is strictly increasing. Let
v E r; then w = cp(v) E r. Assume that v < w; then
cp(v) < cp(w) and, by (14), w < v, which is absurd.
We can conclude in the same way v > w, so that

cp(v)

= v.

(25)

If cp is supposed to be strictly decreasing, set

= - cpo Then 1p is strictly increasing and, by (19)
and (14), satisfies 1p(1p(v = v. Applying to 1p the same
argument as before, we obtain 1p(v) = v, i.e., Eq. (12).
The choice of Eq. (25) leads to the transformation
formulas
x' = a(v)x - va(v)t,
1p

t'

= e(v)x -

a(v)t,

(26)

while the choice of Eq. (12) leads to

x' = a(v)x - va(v)t,

t'

= e(v)x + a(v)!.

(27)

Formulas (26) are incompatible with (10). Hence,

for two observers whose space axes have the same
orientation and whose times flow in the same direction,
(12) must necessarily hold and the transformation
formulas are given by (27).
Our task of proving the reciprocity relation has thus
been completed.
III. EXPLICIT FORM OF THE TRANSFORMA
TION EQUATIONS

Formulas (27) contain the two as yet undetermined

functions a(v) and e(v). However, it is seen at once
that e(v) can be expressed in terms of v and a(v).
11 See, for example, J. Dieudonne, Foundations of Modern Analysis
(Academic Press Inc., New York, 1960), Theorem 4.2.2. For the
reader'sconvenience the proof is given with some detail in Appendix B.

Downloaded 19 Nov 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

V. BERZI AND V. GORINI

1522

instead of (12), and obtain

Indeed, consider the inverse transformations

+ Ll-l(v)va(v)t',
-Ll-l(V)C(V)x' + Ll-l(v)a(v)(',

x' =

x = Ll-l(v)a(v)x'
t =

where
Ll(v)

= a(v){a(v) + vc(v)}.

By the reciprocity relation, (28) can also be written in

the form
x = a( -v)x' + va( -v)(',
(=

c(-v)x'

+ a(-v)t',

t' = - [(l/v){a-1 (v) - a(v)}x

(28)
(29)

= - [a(v)x - va(v)t),
t' = (1/v){a-1(v) - a(v)}x + a(v)t,

x' = - [a(v)x - va(v)tJ,

e(v) = (l/v){a-1(v) - a(v)}.

t' = - W/v){a-1(v) - a(v)}x

Then the transformations (27) read

+ a(v)t.

(31)

To interpret (26) we proce.::d as above, by using (25)

a(v)
O/v)(a-1(v) - a(v

-va(v) (
a(v')
a(v)
(I/v')(a- 1 (v') - a(v'

_ (

+ a(v)t].

(34)

It is the essence of the principle of relativity that the

set of all transformations (31)-(34), as v varies in r,
forms a group C. From this property one can derive
the explicit form of a(v). In fact, let us compose two
transformations of type (31), such that
-v'a(v')
a(v')

a(v)a(v') - (vjv')a(v)(a-1(v') - a(v'

(a(v')jv)(a- 1 (v) - a(v

(33)

whereas if S' is subjected to both a space and a time

inversion, then

so that

(32)

x'

(30)

+ ve(v,

x' = a(v)x - va(v)t,

[' = Ojv){a-1(v) - a(v)}x

+ a(v)tJ.

Hence, (26) is obtained from (27) by an inversion of

the time of S', and this explains why (26) corresponds
to the choice T(V) = v.
The transformation formulas, which connect S to
an observer obtained from S' by inverting the orientation of the space-axis, are

whereby, using (18),

a(v) = l/(a(v)

a(v)t - va(v)t,

+ (a(v)/v')(a- 1(v') -

-(v
a(v'

+ v')a(v)a(v')

a(v)a(v') - (v'/v)a(v')(a-1(v) - a(v

= A(v, v').

(35)

The resulting transformation must be of one of the

four types (31)-(34) for some relative velocity v".
However, since the determinant of transformations
(31) and (34) is +1, whereas the determinant of
transformations (32) and (33) is -1, then types (32)
and (33) must be ruled out.
Since the diagonal elements of the matrices of both
transformations (31) and (34) are equal, we must
have

A(v, v');

a(v)a(v') - (vjv')a(v)(a-1(v') - a(v'

and v varies in the domain r = (-c, e). Equations

(40) are the ordinary proper orthochronous Lorentz
transformations.
(B) K = O. Formulas (31) become

= a(v)a(v') -

(v'jv)a(v')(a- 1(v) - a(v,

i.e.,
Ojv 2){1 - a-2(v)} = (1/v'2){1 - a-2(v')},

(36)

v" = (v

= (v

(37)

where K is a universal constant having the dimensions

of an inverse-square velocity. Then, since a(v) is
positive,
a(v) = 1/(1 - KV2)k.
(38)

+ v')/(1 + Kvv').

(39)

x' = [1 - (v2je2)]-~x - {vel - (v2/e2)]-~}t,

t' = -{(v/e2)[1 - (v2Ie2)]-~}x

+ [1

- (V2jc 2)l-lt,
(40)

x' = x - vt,
t' = t,

(41)
+ (0). These

and v varies in the domain r = (- 00,

are the Galilean transformations.
(C) K < O. Set c = (-K)-~ and formulas (31)
become

+ (v2je2)]-!x - {v[1 + (v2je2)]-!}t,

{(vjc 2)[1 + (V2jc 2)]-!}x + [1 + (V 2 /C 2)]-h,

x' = [1

t' =

The composite velocity v" is the negative ratio between the second and the first element of the matrix

(vjv'){a- 2 (v') - I}]

Three cases are to be considered:

(A) K > O. Set e = (K)-! and formulas (31) become

whence
(ljv2){l - a- 2 (v)} = K,

+ v')/[1

and v varies in the domain

r = (- 00, + (0).

Downloaded 19 Nov 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

(42)

RECIPROCITY PRINCIPLE, LORENTZ TRANSFORMATIONS

Set Xl = X, XO = ct, and tgoc
IT/2), and (42) becomes
X'l = (cos OC)XI x'o

(sin OC)Xl

= vic

(-IT/2

< oc <

(sin oc)XO,

+ (cos oc)xo.

(43)

Hence, in contrast to the Lorentz transformations (40),

which, as is well known, are hyperbolic rotations in
the plane (x, t), transformations (42) are ordinary
circular rotations. Since oc is confined to the interval
(-IT/2, IT/2), it is clear that they do not form a group".
Ifwe let oc vary from -IT/2 to 3IT/2, so as to obtain the
full group, we can easily see that we are led to introduce also the transformations

x' = -([1 + (V 2/C 2)]-ix - {v[l + (V 2/C 2)]-i}t),

t' = -({(v/c2)[1 + (V2/C 2)]-i}x + [1 + (V 2/C 2)]-it) ,
(44)
which are obtained from (42) by inverting both the
space and the time axis of S'.
The rotation group (43) translates into mathematical
form a complete isotropy of space-time, so that the
two directions in time are completely equivalent as
well as the two directions in space. On the other hand,
if one believes that there is an intrinsic arrow in the
direction of flow of time, so that time reversal is
regarded as a purely mathematical operation which
cannot be physically realized, one obtains a strong
argument to rule out the transformations (42). Close
to this argument is the one set forth by Lalan,2k who
postulates that if two events take place at the same
point in space with respect to a given observer, their
time order must be the same for all observers. Alternatively, we could postulate that the relation which
states that the space axes have the same orientations
and that the times flow in the same' direction is
transitive, which amounts to assuming that the set of
proper orthochronous transformations (31) is by itself
a group.
Two other curious features of transformations (43)
can be read out in the formula of composition of
velocities (39) which, in the present case, has the form

v"

(v

+ v')/[I

- (vv' /e 2 )].

(45)

First, by composing two finite velocities v and v' such

that vv' = c2 , one obtains an infinite velocity v".
Second, by composing two positive velocities v and
v' such that vv' > c2 , one obtains a negative velocity
v". Some authors 2s ,w use these properties as an
argument to exclude transformations (42). In our
opinion, however, an argument of this kind is not so
convincing as the preceding ones in such a general
context, because there are not sufficient reasons of
principle to exclude the appearance of phenomena such

1523

as those described above. Besides, it is to be noted

that peculiarities of this type also appear in the
Lorentz case, for which (39) reads

v" = (v

+ v')/[1 + (vv' /c 2)].

(46)

Indeed, if, following some recent ideas,12 one conjectures the existence of faster-than-light particles
(tachyons) and interprets (46) as the transformation
formula for the tachyon velocity (v' = particle
velocity as measured by S'; v" = particle velocity as
measured by S), it is easily seen that, fixing v very small
and negative, we can transform a very large, greater
than c, and positive v' into a very large, greater than
c, and negative v". Further, there always exists a
reference frame relative to which a tachyon propagates
instantaneously. These features are just as curious as
those which have been discussed above in connection
with formula (45). Notwithstanding, this has not
prevented some authors from considering the possibility that faster-than-light particles really exist, on the
grounds that the usual objections to such particles are
ultimately found to be unconvincing when subjected
to critical analysis.
Once we agree to reject formulas (42), we are left
with the problem of the choice between the Lorentz
transformations (40) and the Galilean transformations
(41). As is well known, the Lorentz transformations
admit one and only one invariant velocity which is
equal to c. In the limit when this velocity is taken to
be infinite, one obtains the Galilean transformations.
Hence the above problem of choice can be solved
only by experience and involves the search for an
invariant velocity in nature. The experimental
evidence for the existence of signals which travel with
a finite invariant velocity (such as the electromagnetic
waves in vacuo) leads us to rule out the Galilean
transformations in favor of the Lorentz ones. In
these, of course, the numerical value to be assigned
to e is the experimentally measured value of this
invariant velocity, namely, the value of the velocity
of propagation of electromagnetic disturbances in
empty space.
Formally, the rotation transformations (42) correspond instead to the appearance of an invariant
imaginary velocity c. This is expressed by the property
that they are the linear transformations which
conserve the positive-definite quadratic form x 2 +
e2t2, while the Lorentz transformations are those
which conserve the indefinite form x 2 - c2 t 2 In a
12 O. M. P. Bilaniuk, V. K. Deshpande, and E. C. G. Sudarshan,
Am. J. Phys. 30, 718 (1962); S. Tanaka, Progr. Theoret. Phys.
(Kyoto) 24,. 171 (1960); G. Feinberg, Phys. Rev. 159, 1089 (1967),
and unpublished; R. Newton, Phys. Rev. 162, 1274 (1967); M. E.
Arons and E. C. G. Sudarshan, ibid. 173,1622 (1968).

Downloaded 19 Nov 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions

V. BERZI AND V. GORINJ

1524

four-dimensional space-time the corresponding conserved forms are x 2 + y2 + Z2 + c2t 2 and x 2 + y2

Z2 - c2 t 2 and the appropriate groups are the orthogonal group in four dimensions 0(4) and the Lorentz
group 0(3, 1). The characteristic of 0(4), that a
transformation containing both space reflection and
time inversion can be joined continuously to the
identity, corresponds to the topological property that,
while 0(3, 1) has four connected components, 0(4)
has only two.
IV. CONCLUSION

so that

By making use of the principle of relativity and of

the isotropy of space, we have deduced in a simple
but rigorous way the reciprocity relation for the
relative motion of two inertial reference frames, which
is usually assumed as a postulate in the standard
derivations of the Lorentz transformations without
the principle of invariance of light velocity. For
completeness we have then given the usual deduction
of the transformation equations by using their group
property. We have put forward some alternative
arguments to rule out the transformations with
invariant imaginary velocity. From a logical viewpoint
these arguments might seem more appealing than
those previously given by other authors.

g(k~)

APPENDIX B

We recall the following two results of general

topology [cf. Ref. (11), Theorems 3.19.7 and 3.19.1].
Proposition 1: The continuous image of a connected
topological space is connected.
Proposition 2: A necessary and sufficient condition
for a subset A of the real line to be connected is that
A is an interval.

In the following, if sand t are any two real numbers,

[s, t] will denote the closed interval {x:s ~ x ~ f}, if
s ~ t, and the closed interval {x: t ~ x ~ s}, if t ~ s.
In order to prove that the mapping rp is strictly
monotone, consider two fixed points p and q of r
such that p < q. Since rp is one to one, we can exclude
rp(p) = rp(q) and suppose, for instance, that rp(p) <
rp(q). Let r be any other point of r, r : p, r : q.
We prove
(Bla)
r < p => rp(r) < rp(p),
p

= ga)

+ gW

(AI)

that
= ng( ~).

(A2)

n,

As g(O) = 0, g( -~) = - g(
so that (A2) holds
equally well for n any integer.
Next, for any rational r = min, set m~ = Wi7. Then
mga)

= g(m~) = g(nrJ) = ng(rJ)

=

rg( ~).

(A3)

Assume now that g is continuous at the origin. This

property, together with (AI), implies that g is continuous everywhere. Then, let k be any real number
and {k n } be a sequence of rationals which converges
to k. So
and, by continuity,

But

(BIb)
(B2a)
(B2b)

: rp(p),
as implied by rp being one to one, and suppose it to be
rp(r) < rp(p). Since rp is continuous, by propositions
1 and 2, rp([r, q]) is an interval, so that rp([r, q]) ~
[rp(r), rp(q)], whereby rp(p) E rp([r,q]) because rp(r) <
rp(p) < rp(q). Then there is a p' E [r, q] such that
rp(p') = rp(p), and this is incompatible with rp being
one to one because p' : p, as implied by p
r,
p < q. (BIb) is thus proved.
(Bla), (B2a), and (B2b) are proved in a similar way.
Let now y and y' be any two points of r with y < y'.
Choose s such that y < s < y'. Three cases are
possible: s = p, s < p, and p < s. In the first case,
rp(y'). In the
apply (Bla) and (BIb) to get rp()')
second case, (Bla) implies r:p(s) < r:pCp) and we can
again apply (Bla) and (BIb) with s in place of p to
obtain rp(y) < rp(y'). In the last case, (Bib) gives
rp(P) < rp(s) , and use of (B2a) and (B2b) with s in
place of q gives again rp(y) < rp(y'). Hence y < y' =>
rp(y) < rp(y'), and rp is strictly increasing. One can
show in the same way that the alternative r:p(q) < r:pCp)
implies that rp is strictly decreasing.

<

and thus
g(r~)

< q => rp(r) < rp(q) ,

q < r => rp(q) < rp(r).
Indeed, let, for example,p < r. We have rp(r)
r

Ifn is a positive integer, we get by induction, from (AI),

g(n~)

< r => rp(p) < rp(r) ,

and

Let g be a mapping of Rn into itself such that

+ ')

(A4)

(AI) and (A4) state that g is an endomorphism of

the vector space Rn.

APPENDIX A

g(~

= kga).

<

Downloaded 19 Nov 2012 to 152.14.136.96. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions