Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Foundations of Physics, Vol. 25, No. 12.

1995

An Einstein Addition Law for Nonparallel Boosts


Using the Geometric Algebra of Space-Time

B. T o m King ~

Received May 30. 1995

The modern use oJ'algebra to describe geometric ideas is discussed with particular
reference to the constructions of Grassmann and Hamilton and the subsequent
algebras due to Clifford. An Einste#1 addition law for nonparallel boosts is shown
to follow naturally from the use of the representation-independent form of the
geometric algebra of space-time.

1. INTRODUCTION

Of those points where mathematics and physics touch, the oldest is


geometry. Great interest naturally surrounds the interface between the
axiomatic approach of mathematics and the phenomenological approach of
physics. In particular the way in which we use algebra to describe
geometric ideas has been a regular theme. The modern field of study was
initiated by Grassmann ~t~ who attributed algebraic labels to points in
space. The combination of two such points with labels a and b into ab is
interpreted as a directed line segment connecting the two points. Now the
term ba represents the same line segment but with opposite direction. This
is stated algebraically as ab = - b a and hence such an algebra is anticom-
muting. Further, if c labels a third point then abc translates as a plane with
an orientation given by a screw, usually right handed, turning through the
points in the given order, as illustrated in Fig. 1. Consequently bac defines
the same plane, but the orientation is reversed. Higher-dimensional objects
are thus constructed, the relative orientations of the objects being con-
tained within the rules of combination.

t Department of Physics and Astronomy, University of Canterbury, Private Bag 4800,


Christchurch, New Zealand.
1741
0015-9018/95/1200-1741507.50/0~/ 1995PlenumPublishingCorporation
1742 King

b
Fig. I. Left: Grassmann's view of geometric objects described by algebraic elements. Right:
Hamilton's quaternions associated with the basis vectors of Euclidean 3-space. Grassmann's
technique distinguishes objects of different dimension and can be tailored to suit any linear
space whereas Hamilton's quaternions are specific to E3.

While Grassmann worked on his algebra, Hamilton was studying


spatial rotations. ~- He discarded the imaginary unit, x / / ~ ] -, preferring three
new elements i, j, and k each associated with an axis of Euclidean 3-space
E3. Each has a square equal to negative one and anticommutes with the
others. Hamilton named these elements quaternions because together with
the scalar there are four. Hamilton successfully fashioned this invention
into the famous quaternion description of rotations/21 Hamilton's quater-
nions resemble Grassmann's algebra in that they possess an anticommuting
structure (e.g., 0 = - j i ) . The two approaches differ though in that the
product of two quaternions is not interpreted as a higher-dimensional
object, the relation ~k = - 1 allowing the reduction of all such products to
a third quaternion.
Hamilton and Grassmann never collaborated; and it was left up to
Clifford to combine their approaches. Clifford classified a family of algebras
which attribute an element ei to each of the basis vectors of a linear space,
as did Hamilton. The product of two elements is given by eiej + eje~ = 2g u

2 One of the most elegant examples of the use of a geometric picture to visualize a difficult
mathematical concept is Hamilton's theory of turns, t2~ which provides a means of envisaging
SU(2), the covering group for rotations. Consider two points n and n' on the surface of the
unit sphere S=. Such points define a unique great arc of least length connecting the two
points, provided that they are not diametrically opposite. This great arc defines a directed
axis, an angle and thus a rotation. The equivalence class of such arcs A(n, n') generated by
sliding the arc around its great circle was called a turn by Hamilton and is identified as a
member of SU(2). A second element A(n', n") may visually be seen to combine with the first
by sliding the two arcs around their respective great circles until they lie end to end giving
the rule of combination A(n',n") A(n,n')=A(n,n"). This useful mental picture may be
carried over to give a geometric interpretation of other algebraic structures by using the idea
of turns on manifolds other than S 2. This has been done for the groups Sp(2, R) t24~ and
SU( 1, 1). c=5~Turns may also be used to describe Berry's geometric phase. (26~
Einstein Addition Law for Nonparallel Boosts 1743

and it is this expression that links the algebra to the geometry of the par-
ticular space under consideration by inclusion of the metric go" Higher-
dimensional objects are represented by the product of the appropriate
number of algebraic elements in the same manner as Grassmann. Clifford
algebra is thus a hybrid of the Grassmann and Hamilton approaches, and
as a result of their deep geometrical features Clifford gave them the name
geometric algebra. Unfortunately Clifford's algebras have remained largely
unused. Not until Pauli and Dirac formulated their algebras did the ideas
of Clifford become more widely known.
Since the 1960's Hestenes has worked toward a new language for
physicists, t3~ He has bridged the gap between the mathematician and the
physicist by using the framework of Clifford algebra to great effect/3-5~
Several works provide introductory 13"61 and advanced ~7-~31 reading for
those interested in the field.
This paper utilizes these geometric ideas in the analysis of Lorentz
transformations. The framework for performing such coordinate trans-
formations is naturally constructed on flat Minkowski space-time, the
associated algebra being called the geometric algebra of space-time. The
advantages of the eventuating scheme are illustrated by expressing the
process of combining successive nonparallel boosts as

~_, - -
~ - + ~5 (1)

an expression identical to Einstein's original rule for combining parallel


boosts. This result has remained hidden behind convoluted expressions for
some time and is unveiled by use of the correct language from the outset.
A further benefit exists as this mechanism of manipulation provides an
intuitive forum for the handling of geometric symmetries.

2. THE REPRESENTATION-INDEPENDENT FORM OF THE


GEOMETRIC ALGEBRA OF SPACE-TIME
The geometric algebra of space-time is the real Clifford algebra •,.3
constructed on flat Minkowski space-time R ''3 and can be embedded in
Dirac's 4 x 4 complex-valued matrix algebra. Explicitly it is the algebra
generated by 4 and e~, (where the Lorentzian index p = 0, 1, 2, 3) such that

e~,e,, + e,,e~,= 21h,,, (2)

and where ev are algebraic terms to be interpreted as the basis of


Minkowski space-time. The Lorentz metric t/,,, = diag( + 4, - 4 , - 4 , - 4 )
1744 King

is used and for brevity all subscripts are concatenated such that e~,e. = et....
The algebra is 16-dimensional, an arbitrary element e e NL3 being written
in accordance with Zeni tl41 as

1 1
+ a e~, + - - a~'Vej,~ + ~ i af'~Pe/'"t' + p e ~,3o
2! 3. -
(3)
s, a t', a/'v, a/m', p ~

and the element with all coefficients equal to zero is written 0. Both the
basis elements and the coefficients are antisymmetric under the exchange of
neighboring indices requiring the factors of 1/2! and 1/3! in order to
account for the various permutations.
Other authors 19-131 choose to write the elements of this algebra as
(1, y~,, {ak, i a k } , i),j,, i) where i = Yo123 and ak = Yko. I have elected not to
use this notation due to certain inherent misleading properties. Consider the
product of the terms Yo and i. We would write this y o i = - i y o , suggesting
that the two terms anticommute. This is not consistent with the fact that
the imaginary i commutes with all complex-valued matrices, a set of items
which contains representations of the space-time algebra. A clearer way to
read the above product is to bracket basis terms so that (Y0)(i)=-(iyo).
If scalar terms are not bracketed, then (?0)i=i(y0), which is surely bad
notation.
I avoid the need for such notation by avoiding the use of the symbols
i, ~, and a in favor of e with a varying number of subscripts. Thus the
algebra is spanned as a linear space by (1, ej,, e~,v, era.p, e1230). This nota-
tion has the benefit of keeping clear the geometric nature of an object while
reducing all manipulations to simple subscript handling. For example,
only with experience would the statement y3a2 =i)q be obvious, but the
alternative e3e20 = e320 is straightforward and more easily implemented on
symbolic manipulation software. Further, the results of this paper are
representation-independent, so that in the presentation that follows no
allusions are made to the well known representation of Pauli (Dirac)
implied by the symbols O'k(y/, ).

2.1. The Multivector

The elements of this algebra have particular geometric properties


which require some discussion. A scalar has no geometric ingredient, that
is, it has no space-time dimension associated with it. The corresponding
algebraic symbol thus has no subscripts allowing it to commute with all
elements of the algebra. The scalar is identified by 1, distinguishing it from
Einstein Addition Law for Nonparallel Boosts 1745

the number 1. Those elements spanned by ej, are called vectors, and the
corresponding vector space is denoted by R 1'3 which is contained in the
algebra R~, 3. Constructs of higher order involve the product of two or
more vectors. The elements e~,,, result from the product of two different
basis vectors and are thus properly referred to as bivectorsJ 3) Hamilton's
description of rotations is flawed in this respect as it fails to write the
product of two vectors as a higher-dimensional object. We adopt
Grassmann's approach and view bivectors as oriented planes. In the pre-
sent case we have a basis vector associated with the time axis and thus we
have the possibility of space-time planes. Just as Grassmann viewed a
spatial plane as defining a rotation, space-time planes provide us with time-
like rotations (Ref. 3, p. 53) which are the core of transformations from one
inertial frame to another. Third-order constructs e~,w are identified as
pseudo-vectors (or trivectors) and the element e1230 is called the pseudo-
scalar.
One of the key contributions made to the field of geometric algebra by
David Hestenes is embodied in the statement physicists have not learned
properly how to multiply vectors'. Returning to conventional 3-space vector
analysis briefly, we know that two vectors, a and b, may be combined in
different ways; the scalar product a - b and the vector product a/x b. What
Hestenes meant was that a complete vector product ought to contain all
this information in a single term. To this end the geometric product is
defined
ab=a.b+a ^ b (4)

which requires that a scalar and a vector be added! This is an example of


a multivector. Returning to the language of geometric algebra, a multivec-
tor is simply a member of the algebra which is spanned by basis elements
of different order, the general member of the algebra being written as
in Eq. (3). Henceforth when an expression is expanded in terms of
basis elements the result is not interpreted in terms of scalar and vector
products.

2.2. Geometric Symmetries

Discrete symmetries such as reflections in a plane are simple to imple-


ment using the current formalism. If l~ and v are vectors, then the operation
v ~ 1~vl~-1 has the effect of reflection in a plane perpendicular to ~ and then
reversing the sign of all components. We could then write the process of
reflection through a plane perpendicular to r~ as v ~ - ~ v l ~ -1, but the
1746 King

presence of the minus sign gives rise to problems when performing reflec-
tions on higher-dimensional objects. The minus sign is acquired via alter-
nate means. We pre-multiply by el230 and post-multiply by el230 -~ = -e1230 ,
which has the effect of multiplying every element of R 1.3 by ( - 1 ) s , where
s is the number of subscripts of that element. Thus reflection through a
plane is implemented by v ~ (1]el230) v(~e1230) -1. This can be seen to hold
for all multivectors. Consider the basis tensor e . . . . . . Reflecting through the
plane perpendicular to i~, and using R,~ = (1~e123o), we get

R,~e . . . . . Ri, l = (R,~e~,R,7, l ) R,~... R,7,1(R,ie,oR[ l ) (5)


by inserting the identities R,~-~R,~ appropriately. Thus each basis element is
reflected so that the overall sign change of the tensor under the reflection
is properly found. This extends to an arbitrary multivector as the geometric
algebra is distributive.
To illustrate the above point we m a y consider the electromagnetic field
tensor, F. This is written as a bivector 3

F = E l elo + E2e20 + E3e30 + B1 ez3 + B2e31 + B3el2. (6)

The polar and axial properties of the electric and magnetic fields can be
identified by performing a parity inversion, F--* P F P -~. A parity inversion
is the result of reflection through all three of the spatial axes so that P =
(elel?.30)(e2e1230)(e3el,_30)=eo. This reverses the sign of E but leaves B
unaltered. Time reversal operates in a similar way, F--* T F T -~ where T =
eoe1230 = -el23. It is trivial to check that this operation leaves E unaltered
whilst reversing B.
When using this mechanism sign changes due to geometry inversions
are systematically treated. A q u a n t u m mechanically consistent geometric
algebra will also give rise to sign changes and will be likely to isolate those
due to geometry inversions.

3. L O R E N T Z T R A N S F O R M A T I O N S

The framework of the algebra is now established and we turn attention


to the process of performing a coordinate transformation. Initially we
consider general orthogonal transformations of the basis vectors {e~,} and
later in Sec. 3.2 we look more closely at the restricted Lorentz group.

s Lounesto~.,7) also writes the field tensor in this way. Defining the invariant differential
operator a = ej,at, and a current vector J=fl'e~, we find that Maxwell's equation(s) is written
aF= J, a compact form originally due to M. Riesz.~28~
Einstein Addition Law for Nonparallel Boosts 1747

3.1. Orthogonal Transformations of Basis Vectors


Before continuing let us define a few key terms. The reversion
automorphism, t, of N~,a is realized by reversing the subscripts of each
basis element of N1,3 (e.g., et*23= e32~-- -e123) leading to the result

( ~ ) t = flt~t, VO~,fie ]1.3 (7)


The metric tensor g ( , ) operating on u, v ~ ~'3 is

g(u, v) = 89 vu) (8)


which is symmetric and in the special case u = v reduces to g(u, u) = u 2.
A vector u transforms to u' in the primed frame according to the trans-
formation

A n 2 = u ' e I~ 1"3 (9)

where the tilde operation is yet to be defined. The condition that the trans-
formation is orthogonal (length preserving) is stated formally by requiring
that g(u', v ' ) = g(u, v). In particular when u - - v we see that u '2 = u z. We
consider only transformations connected with the identity which may thus
be generated by exponentiation (Ref. 14, Theorem A). Such a transforma-
tion can be written as a power of an infinitesimal transformation
A A = 1 + e T , where T e I~l. 3. We apply this infinitesimal transformation in
accordance with Eq. (9) and, neglecting all but linear terms in e, we have

[(1 + e T ) u(l + e T " ) ] 2 = u -"


uTu + u'-T+ Tu z + uT"u = 0 (10)
VuER 1'3 ( T + ~')u = - u ( T + 7 " )

where we have used the fact that u 2 is a scalar and hence commutes with
all T ~ ~ , 3 . Now there are no elements of g~,3 which anticommute with
each of e/, and hence with an arbitrary u E R ~'3 so that T + I"= O. We raise
A A to the appropriate power, a general orthogonal transformation then
being written in exponential form as A = e r and A = e - r immediately
showing that . ~ A - - A , 4 = I , and the tilde operation is thus simply the
inverse for those elements connected to the identity (Ref. 14, Theorem C).

3.2. The Restricted Lorentz Group


The group of orthogonal transformations is known as O(1, 3). The
restriction to those members which are connected to the identity gives us
1748 King

the group S O ( l , 3), widely known as the restricted Lorentz group. The
elements of this group are the proper (no parity inversion) orthochronous
(no time reversal) transformations.
It is known (Ref. 14, Theorem C) that all Lorentz transformations
connected to the identity can be generated by the exponentiation of some
b i v e c t o r f ~ {ei,~, } so that
A = e cl/2~o,,,'e,,,. ( 11 )

and it is a simple matter to show that the inverse is given by the reversion
operation for such elements, A - l = A*. It is now natural to separate e~,,
into its space-space components e 0. and its space-time components ek0. This
is the case because the division separates the generators of pure rotations
from those for pure boosts. Consider the transformation generated by riejk,
namely A = e '''#, where i, j, k represent cyclic permutations of 1, 2, 3.
Expanding this, we have

9 1 .
A = ] + r ' e j , + ~ ( r ' e j k ) 2 + ...

r'ejk
=~ cos ~b+ - ] 7 - sin q~ (12)

Application of this transformation on an arbitrary vector performs a rota-


tion of angle 0 = 2 ~ b e ( - n , n) about an axis given by the direction cosines
pi= rTlrl. A rotation by angle 0 about an axis f'~ is performed by operation
of
A R = e (1/2lOei'7* = ] cos(0/2) + i~iejk sin(0/2) (13)

Similarly a boost is realized by application of

AB = e I t/zl)Y,',.,o = ~ cosh(2/2) + bieio sinh(2/2) (14)

where/~g specifies a unit vector in the direction of the boost.


We are now in a position to factor the arbitrary Lorentz transforma-
tion of Eq. (11) into the product of a boost and a rotation. However, this
cannot be done in a unique fashion as an arbitrary A can be factored any
number of ways corresponding to a series of boosts and rotations which
have a combined effect equal to that of A. In particular, A can be factored
into a boost followed by a rotation, AR, AB,, or vice versa, AB, AR, where
the parameters in either case are not equalJ 15~ The result of this paper
follows in either case so long as the proper rules of manipulation are
adopted. In order to arrive at results which will be comparable with
Einstein Addition Law for Nonparallel Boosts 1749

standard texts on electrodynamics (for example, Jackson t~61) we adopt the


convention of factorizing every A as a rotation followed by a boost and in
this case every factorization is unique. Hence each A is written

A =(cosh(2/2)~ +sinh(2/2)bieio)(cos(O/2)~ + sin(0/2) riejk) (15)

The six parameters 2, 0, b ~, r; can be expressed in terms of the six coef-


ficients a t'' of Eq. (11 ), but it is not enlightening to do so here.
As an illustration of the implementation of this scheme, the situation
discussed by Misner et al. ~7~ is considered. Two successive rotations of 90 ~
are performed, the first about the z-axis and the second about the x-axis.
We require the combined effect in the form of a single rotation. Noting that
cos(n/4)=sin(n/4)= 1/,v/2_ and using Eq. (15) we find that the two rota-
tions are performed by the elements

Al =(1/x/~)~ + ( 1 / x / ~ ) el2 and A, =(1/x/~))~ + (1/x//2))e23


(16)

The combined effect is a single rotation performed by A = A2A~ so that

A = (1/2)~ + (x/~/2)(ez3- e3, + e,2)/x/~ (17)

from which the combined angle of rotation 0 is identified by cos(0/2) = I/2


and sin(0/2)= x/~/2 to be a rotation of 120 ~ The axis is given by r;ejk =
(e_,3-e3~ +e~z)/v/3 or more simply the vector connecting the origin to
( 1 , - 1 , 1).

4. T H E EINSTEIN A D D I T I O N LAW F O R N O N P A R A L L E L BOOSTS

We will now derive an expression for combining two nonparallel


boosts which is the same in form as the famous Einstein addition law for
parallel boosts. Before doing so it is necessary to define the concept of divi-
sion within the confines of the algebra ~ , 3 , a process which is not usually
found within a noncommuting algebra. In the previous section we adopted
the convention of factorizing an arbitrary Lorentz transformation into a
rotation followed by a boost A =ABAR. The main result of Eq. (27) relies
upon being able to divide out the rotation part, and hence we define
division using the right inverse, that is,

~x
_=~fl-I
fl (18)
1750 King

for t w o a r b i t r a r y m e m b e r s ~, fl o f the a l g e b r a 0~1,3 p r o v i d e d t h a t f l - ~


exists. 4 If we h a d elected to factorize the r o t a t i o n o n the left, then the result
w o u l d still follow p r o v i d e d we then used the left i n v e r s e )
C o n s i d e r t w o p u r e b o o s t s A i a n d A 2 given b y

A I = c o s h ( 2 1 / 2 ) ~ + sinh(21/2) b~eko (19)

A~_= c o s h ( 2 2 / 2 ) ~ + sinh(22/2) b2ek


olr (20)

W e c o m b i n e the effect o f these on a n a r b i t r a r y v e c t o r u a n d e q u a t e the


result with the a c t i o n o f a single t r a n s f o r m a t i o n A 12

u'=A2AIuA~A~=A12uA~ - (21)

C l e a r l y we have A12 = A2AI, a n d w h e n A IZ is w r i t t e n in a c c o r d a n c e with


the c o n v e n t i o n a d o p t e d in Sec. 3.1 we have

(cosh(2p_/2) ~ + sinh(212/2) b~,2em)(cos(012/2)~ + sin(012/2) 89

= cosh().2/2) c o s h ( ; q / 2 ) ] + c o s h ( 2 2 / 2 ) s i n h ( ) . l / 2 ) bil eio

+ sinh(22/2) cosh(21/2) bi, em+ sinh(22/2) s i n h ( 2 1 / 2 ) [ b ~ e m ] [b{ ejo]


(22)

where 212,012, r~12, a n d b~12 are the p a r a m e t e r s o f the c o m b i n e d t r a n s f o r m a -


tion. W e a b b r e v i a t e c o s ( G 2 / 2 ) ~ + sin(01J2)r~12ejk =l~ a n d e q u a t e the coef-
ficients o f either side:

bi12e,o s i n h ( 2 , 2 / 2 ) ii = c2 s, bil em + s2c, b~e,o (23)

cosh(212/2)Ji = ~c2cl + S2Sl[ b2em][ b~ejo] (24)

4 Not all elements ~ 9 R ~'3 necessarily have an inverse; for example ~ + eo cannot be pre- or
post-multiplied by any element to give 6. However, if an element cr does have an inverse fl
such that ~fl = ~, then it is also true that flc~= L that is, the left and right inverses are equal.
This follows from the widely known result that if A is a square matrix such that AB = I for
some B, then BA =I and B is unique. Since RE3 is a matrix algebra, this result is all that
is needed.
It is worth noting that the division required by the main result of the paper, Eq. (26),
requires the use of the inverse of ~ + [b~elotanh(22/2)][b~ejo tanh(21/2)], which always
exists.
s In the case where we factorize the rotation on the left, we get parameters for the combina-
tion of two boosts which differ from the standard texts. This reflects the fact that the boost
and rotation parameters for an arbitrary Lorentz transformation depend upon whether the
boost or rotation is performed first.
Einstein Addition Law for Nonparallel Boosts 1751

where we have abbreviated sinh(2~._~) and cosh(,~t.2) as s~,2 and ct.2 respec-
tively. Utilizing the definition of division given in Eq. (18), we have

bi12eiol~sinh(212/2) CzSlbileio W S2clbi~eio


= (25)
fi~ cosh(212/2) ~czcl + s2s1[b~eio][b~ejo ]
We divide out h on the right-hand side and divide through by c~ c2,

bi~eio tanh(2,/2) + b~e~otanh(22/2)


b~ze~otanh(2,z/2) = ~ + [b~e~o tanh(22/2)][b{ejo tan(2,/2)] (26)

In the tradition of geometric algebra we associate an algebraic term ~',-_


with the geometric notion of a coordinate transformation from one inertial
frame to another. In a manner similar to Hirschfeld and Metzger 1~8~ we
define elements r = bie~otanh(2_/2) so that

~/'12 - - - - (27)

which is the main result of the paper and is a result that has been passed
over when other algebras are used (most notably the algebra of complex
numbers). Physicists are always keen to see how familiar results are con-
tained within a new formalism. In response, Appendix A shows how the
Einstein addition law for parallel boosts is derived, and Appendix B derives
equations which determine the magnitude of the Thomas rotation resulting
from two successive boosts.

5. CONCLUDING REMARKS

Clifford algebras have provided a natural generalization of the work of


Hamilton and Grassmann. Hestenes has given a complete analysis of the
way in which the modern physicist can exploit the geometric properties of
these algebras. In deciding what language we ought to use to describe a
particular situation we look at the geometry we are restricted to and decide
accordingly. This paper has succeeded in generalizing the famous result of
Einstein only because the geometric algebra of space-time was used. Pre-
vious descriptions which have used the real or complex number systems
have passed over this convenient generalization because it has been hidden
in inelegant expressions.
It is therefore natural to suppose that there are other physical
scenarios where a better insight may be obtained by using the appropriate
geometric algebra. An immediate example would involve using the algebra
1752 King

~1,2 to describe systems in which particles are restricted to one time and
two spatial dimensions. Such systems are at the forefront of contemporary
physics as a result of the discovery of the integer and fractional quantum
Hall effects and the postulated existence of anyons.
Geometric algebras are being exploited in an increasing number of
applications. It is the hope of the author that in time a wider audience will
acknowledge their usefulness and reap their benefits, of which Eq. (27) is
but one.

APPENDIX A

In order to check that Eq. (27) is consistent with the law for adding
the parameters of parallel boosts, we choose two boosts lying parallel to
the x-axis which is sufficiently general due to rotational covariance. In this
case Eq. (26) reduces to

elo tanh(2t/2) + el0 tanh(22/2)


elo tanh(2,2/2)= (28)
+ [elo tanh(22/2)] [el0 tanh(;t~/2)]

and identifying eloel0=~ this is the combination law for hyperbolic


tangents and hence 212/2 = 21/2 + 2,/2. The rapidities thus add

tanh(21) + tanh(22)
tanh(212) = tanh(21 + 2_,) -- (29)
1 + tanh(2~) tanh(22)

and with v/c = tanh(2) we find

u I -Ft~ 2
vl,_- 1 + (vl v2/c 2)
(30)

as required.

APPENDIX B

The Thomas rotation may be derived in several different ways. The


rudimentary approach works from the component form of the Lorentz
transformations,~ 16) whereas later schemes I'-~ have seen notation and ter-
minology improvements. It is only with the current overhaul of the
fundamental algebra that the derivation of the Thomas rotation becomes
trivial. Consider three right-handed inertial frames, Z', Z', _r", where at
t = 0 all their origins are coincident and the pairs of frames (_r, Z") and
Einstein Addition Law for Nonparallel Boosts 1753

(S', X") agree that all axes are parallel. Now (Z', Z") agree that their coor-
dinate systems are moving relative to one another at a velocity with
magnitude v~ parallel to their mutual z-axes. This constitutes a boost per-
formed by application of the term

A ~= cosh(2 ~/2) ~ + sinh(22/2) e3o (31 )


The frames (Z", s agree that their frames are moving at a velocity v2
along an axis lying parallel to their mutual y - z plane and making an
angle q~ with the z-axis. The corresponding algebraic term is

A2 = cosh(22/2) ~ + sinh(22/2){ sin ~be20+ cos ~be3o} (32)


Again we combine the two transformations A = A2AI giving the expansion

A = cosh(22/2) cosh(2,/2)4 + cosh(22/2) sinh(2~/2) e3o


+ sinh(22/2) cosh(21/2) {cos ~be30 + sin ~be2o}
+ sinh(22/2) sinh(21/2){cos ~b~ - sin ~be23} (33)

By inspection we see that the presence of the basis element e23 implies that
this term contains a rotation part and that this rotation is about the x-axis.
The resultant boost also lies in the y - z plane and combines with a rota-
tion through e about the x-axis so that in general

A = (cosh(2/2)~ + sinh(2/2){ b2e2o + b3e30} )(cos(e/2) + sin(e/2) e23 ) (34)

Taking coefficients of the scalar ~ and e23 we find

cosh(2/2) cos(e/2) = cosh(22/2) cosh(2~/2) + sinh(2_,/2) sinh(21/2) cos ~b


cosh(2/2) sin(e/2) = -sinh(22/2) sinh(2,/2) sin ~b (35)

and eliminating cosh(2/2) we have

tanh(22/2) tanh(2,/2) sin q~


tan(e/2) = (36)
1 + tanh(22/2) tanh(2t/2) cos ~b

This is the final form for the Thomas rotation and has taken only three
steps to acquire. The expression may not be particularly recognizable as it
contains half angles and half rapidities, a feature which is common to
geometric algebra. In order to demonstrate that this is indeed the same as
the form given by other authors, a few more manipulations are required.
We first express tan e in terms of half angles

2 tan(e/2)
tan e - (37)
1 + tan2(e/2)

825/25/12-7
1754 King

We insert our expression from Eq. (36) into this equation and reduce it to
a more simple form:

2(k + cos 4) sin ~b 1


tane-(k+cos~b)2+sin_,b, where k-tanh(21/2)tanh(2,_/2 ) (38)

The identity fl = tanh 2 can be rewritten in a manner that includes half


angles and the common definition y = ( 1 + f12) _ m:

1 y+l
(39)
tanh-'(2/2) - y - 1

This allows the k term to be written

k2-yI+ly2+l k>l (40)


yl--1 y2-1'

which is exactly the form of the Thomas rotation found by Ungar [ Ref. 20,
Eq. (37a, b)]. Other authors ~2~-331 express the same result in different
fOrlTIS.

ACKNOWLEDGMENTS

This paper is the counterpart of another 1~9~ which is the outcome of


work supervised by Geoff Stedman. Much discussion with Peter Renaud
has been central to the development of these papers. I would also like to
thank Jason Harris whose comments have been helpful.

REFERENCES

1. H. Grassmann, Die attsdehmmgslehre (Leipzig, 1844; completed in 1862).


2. Sir William Rowan Hamilton, Lectures in Quaternions (Dublin, 1853), Elements of
Quaternions (posthumous, 1866).
3. David Hestenes, Space-Time Algebra (Gordon & Breach, New York, 1966).
4. David Hestenes and Garret Sobczyk, Clifford Algebra to Geometric Calculus: A Unified
Language for Mathematics and Physics (Kluwer Academic, Dordrecht, 1984).
5. David Hestenes, New Foundationsfor Classical Mechmtics (Kluwer Academic, Dordrecht,
1990).
6. T. G. Void, "An introduction to geometric algebra with an application in rigid body
mechanics," Am. J. Phys. 61,491-504 (1993); "An introduction to geometric calculus and
its application to electrodynamics," Am. J. Phys. 61, 505-513 (1993).
7. S. L. Altmann, Rotations, Quaternions and Double Groups (Oxford University Press,
Oxford, England, 1986), Chapter 12.
Einstein Addition Law for Nonparallel Boosts 1755

8. Bernard Jancewicz, Multivectors and Clifford Algebra in Electrodynamics (World Scien-


tific, Teaneck, New Jersey, 1988).
9. S. Gull, A. Lasenby, and C. Doran, "Imaginary numbers are not real--the geometric
algebra of spacetime," Found. Phys. 23, 175-1201 (1993).
10. C. Doran, A. Lasenby, and S. Gull, "States and operators in the space-time algebra,"
Found. Phys. 23, 1239-1264 (1993).
11. A. N. Lasenby, C. J. L. Doran, and S. F. Gull, "A multivector derivative approach to
Lagrangian field theory," Found. Phys. 23, 1295-1327 (1993 ).
12. S. F. Gull, A. N. Lasenby, and C. J. L. Doran, "Electron paths, tunnelling and diffraction
in the spacetime algebra," Found. Phys. 23, 1329-1356 (1993).
13. A. N. Lasenby, C. J. L. Doran, and S. F. Gull, "Grassmann calculus, pseudo-classical
mechanics and geometric algebra," J. Math. Phys. 34, 3683-3712 (1993).
14. J. Ricardo Zeni and Waldyr A. Rodrigues, Jr., "A thoughtful study of Lorentz transforma-
tions by clifford algebras," hit. J. Mod. Phys. 7, 1793-1817 (1992).
15. N. A. Doughty, Lagrangian Interaction: An hltroduction to Relativistic Symmetry in Elec-
trodynamics and Gravitation (Addison-Wesley, Reading, Massachusetts, 1990); Sec. 13.7
consider the alternative factorizations of a Lorentz transformation.
16. John David Jackson, Classical Electrodynamics (Wiley, New York, 1962).
17. Charles W. Misner, Kip S. Thorne, and John Archibald Wheeler, Gravitation (Freeman,
San Francisco, 1973), pp. 1135 et. seq.
18. A. C. Hirschfeld and F. Metzger, "A simple formula for combining rotations and Lorentz
boosts," Am. J. Phys. 54, 550-552 (1986).
19. B. Tom King, "Reflection optics, an Einstein addition law for nonparallel boosts, and the
geometric algebra of space-time," J. Opt. Soc. Am. A, 12, 773-780 (1995).
20. Abraham A. Ungar, "Thomas precession and its associated grouplike structure," Am. J.
Phys. 59, 824-834 (1991).
21. N. Salingaros, "The Lorentz group and the Thomas precession. II. Exact results for the
product of two boosts," J. Math. Phys. 27, 157-162 (1986).
22. D. G. Ashworth, "A new and simple deduction of the Thomas precession," Nuovo Cimento
40, 242-246 (1977).
23. H. Goldstein, Classical Mechanics (Addison-Wesley, Reading, Massachusetts, 1980), 2nd
edn.
24. R. Simon, N. Mukunda, and E. C. G. Sudarshan, "Hamilton's theory of turns generalized
to Sp(2, R),'" Phys. Rev. Lett. 62, 1331-1334 (1989).
25. R. Simon, N. Mukunda, and E. C. G. Sudarshan, "The theory of screws: A geometric
representation for the group SU( 1, 1)," J. Math. Phys. 30, 1000-1006 (1989).
26. R. Simon and Mukunda, N., "Hamilton's turns and geometric phase for two-level
systems," J. Phys. A 25, 6135-6144 (1992).
27. P. Lounesto, "Spinor-Valued Regular Functions in Hypercomplex Analysis," PhD thesis,
Helsinki University of Technology, 1979.
28. M. Riesz, Clifford Numbers and Spinors (Kluwer, Academic, Dordrecht, 1993). Reprinted
from Lecture series No. 38, University of Maryland, College Park, 1958).

You might also like