Possible Kinematics

Cite as: Journal of Mathematical Physics 9, 1605 (1968); https://doi.org/10.1063/1.1664490

Submitted: 27 May 1967 . Published Online: 28 October 2003

Henri Bacry, and Jean-Marc Lévy-Leblond

ARTICLES YOU MAY BE INTERESTED IN

Galilei Group and Nonrelativistic Quantum Mechanics

Journal of Mathematical Physics 4, 776 (1963); https://doi.org/10.1063/1.1724319

Getting the Lorentz transformations without requiring an invariant speed

American Journal of Physics 83, 338 (2015); https://doi.org/10.1119/1.4901453

Classification of ten-dimensional kinematical groups with space isotropy

Journal of Mathematical Physics 27, 2455 (1986); https://doi.org/10.1063/1.527306

Journal of Mathematical Physics 9, 1605 (1968); https://doi.org/10.1063/1.1664490 9, 1605

© 1968 The American Institute of Physics.

 GREEN'S FUNCTIONS FOR THE ISING CHAIN 1605

tk = t~ = ... = t~ give a complete statistical descrip- Note added in proof: After having submitted this
tion of the Ising chain. In particular, we can obtain paper, we came across a publication by C. Blomberg
in this manner the spin-correlation funCtions calcu- ["Some fundamental aspects of many-body problems
lated earlier by Marsh.' We find a complete agreement in statistical thermodynamics" (Stockholm, 1966)1,
between his and our results. where some of our results have been obtained. We
As we have already mentioned in the introduction, should also mention that some general properties of
the Ising chain does not exhibit any nontrivial kinetic Green's functions for the Ising model have been
behavior. Our investigation of the time-dependent discussed in the following series of papers: F. Goro-
Green's functions serves therefore as an illustration of detzky et al., Phys. Letters 2, 14 (1962); B. G. S.
the general features of the Green's functions formal- Doman, ibid. 4, 156 (1963); A. H. Muir, Jr., E.
ism, rather than as a description of any new properties Kankeleit, and F. Boehm, ibid. 5, 161 (1963).
of the Ising chain. 4 J. S. Marsh, Phys. Rev. 145,251 (\966).

JOURNAL OF MATHEMATICAL PHYSICS VOLUME 9, NUMBER 10 OCTOBER 1968

Possible Kinematics
HENRI BACRY*
The Institute for Advanced Study, Princeton, New Jersey
AND
JEAN-MARC LEVy-LEBLOND
Physique Theorique, Faculte des Sciences, 06-Nice, France
(Received 27 May 1967)

The kinematical groups are classified; they include, besides space-time translations and spatial
r~tations, "inertial tr~nsformations" conn~cting different inertial frames of reference. When parity and
~lme-reversal are requ!red to be automorphlsms of !he groups, and when a weak hypothesis on causality
made,. the only pOSSible groups are found to consist of the de Sitter groups and their rotation-invariant
IS
contractIOns. The scheme of the contractions connecting these groups enables one to discuss their physical
meaning. Beside the de Sitter, Poincare, and GaliJei groups, two types of groups are found to present
some interest. The "static group" applies to the static models, with infinitely massive particles. The
other type, halfway between the de Sitter and the Galilei groups, contains two nonrelativisticcosmologica!
groups describing a nonrelativistic curved space-time.

I. INTRODUCTION inertial transformations, leaving invariant the laws of

It is almost a fact of everyday life that the laws of nature. Due to the isotropy of space (valid at nuclear
physics are invariant under space and time "trans- as well as astronomical dimensions), these inertial
lations" 1 as well as under spatial rotations. The transformations form a three-parameter set, so that on
homogeneity of space and time and the isotropy of the whole there is a ten-parameter group of kine-
space were probably recognized a very long time matical transformations, the relativity group of
before being explicitly stated. But there exists another nature, or kinematical group.3
type of kinematical invariance, expressing the equiv- But the existence of these inertial transformations
alence of a larger class of frames of reference, the does not suffice to determine the precise structure of
so-called inertial frames. This fact, much less obvious, the kinematical group. One still has to tell (i) how the
was clearly realized for the first time by Galilei. 2 In laws of physics transform under inertial transforma-
addition to time "translations," space "translations," tions; (ii) if there exists some natural curvature of
and rotations, there is thus an additional set of space, or, more generally, of space-time. These two
transformations, which we will call, in general, aspects of kinematics are described by the group law.
* Present address: Faculte des Sciences, Marseilles, France. Work 3 Of course, it might be that such an invariance is only approxi.
supported by the National Science Foundation. mat~. [n the context. of general relativity, for instance, only very
1 "Translations" ~e understood here in a generalized sense;
special mod~ls of UnIverses p~ssess. such a la;ge invariance group.
they do not necessarily commute when large distances are involved. Indeed, ~ Rlemanman four-dImensIOnal marufold has an isometry
2 G. Galilei, Dialogo sopra i aile massimi sistemi ael monaD
~roup WIth at most ten parameters, and this only when its curvature
(Edizione Nazionale, Firenze), Vol. 7, J.n, pp. 212-214. IS a constant.
1606 H. BACRY AND J.-M. LEVY-LEBLOND

That this is not a trivial question is proved by the fact way. To sum up, we require the two transformations
that inertial transformations have been thought for
D:{H-H,P--P,J-J,K--K}, (2)
centuries to be pure Galilean transformations, al-
though we now believe them to be pure Lorentz e:{H- -H, P~P, J -.J, K~ -K} (3)
transformations. 4 Because the Lorentz group recently to leave invariant the Lie brackets defining the Lie
has been brought in question,5 it is natural to ask, algebra of the kinematical group. Let us notice that
Are there any other possibilities? We intend here to the combined operation r = De,which inay replace
answer this question under very general assumptions. 6 n and is given by
r:{H-.-H,P-.-P,J-.J,K-.K}, (4)
II. ASSUMPTIONS AND RESULTS
is exchanged with e when P and K are interchanged.
Our investigation is based upon the use of Lie-
(3) The one-dimensional subgroups generated by
algebraic methods. Supposing the kinematical group
the K;'s are noncompact. If this were not the case, the
to be a Lie group is a very natural physical requirement,
boost parameter u of a general boost exp (uKi ) would
expressing the supposed continuity of space-time.
be defined modulo some Uo such that exp (UOKi) is the
We call H, Pi> J i , and Ki (i = 1,2,3), respectively,
identity transformation. In other words, a sufficiently
the generators of time "translations," space "trans-
large boost would be no boost at all. This is an utterly
lations," spatial rotations, and inertial transformations
unphysical situation. Notice that we do not require
along the i axis. We now assume:
the Same compacity for the one-dimensional trans-
(1) Space is isotropic, meaning that the infinitesimal
lation subgroups: Indeed the universe might very well
generators transform correctly under rotations, i.e.,
be a closed one. The present hypothesis also may be
H is a scalar and P, J, K are vectors. In other words,
seen to be equivalent to the very weak causality re-
all the Lie brackets involving the angular momentum
quirement that if, in a certain reference frame, two
J have their standard form:
events take place at the same point, no kinematical
transformation should alter the temporal order of
[J, H] = 0, (la)
these events. 6 f!owever, the equivalence of these two
[J, J] = J, (lb) conditions can only be proved when hypotheses (1)
and (2) have been exploited and a suitable notion of
[J,P] = P, (1c) space-time has been introduced.
[J, K] == K, (ld) Theorem: Under the assumptions that:
where the notation [A, B] = C is a shorthand for (1) space is isotropic (rotation invariance);
[Ai' Bil = €iikCk' (2) parity and time-reversal are automorphisms of
(2) Parity and time-reversal are automorphisms of the kinematical groups;
the kinematical group. This is a natural assumption (3) inertial transformations in any given direction
which greatly simplifies the final result, but is by no form a noncompact subgroup,
means compelling, especially in view of the failure of then there are eight types of Lie algebras for kinematical
parity invariance, and probably, although to a much groups corresponding to eleven possible kinematics.
smaller extent, of time-reversal, in weak interactions. These algebras are:
However, we stick to this conservative approach
throughout the present paper. From the interpretation Rl. The two de Sitter Lie algebras 7 isomorphic, re-
of inertial transformations as "boosts" in a certain spectively, to the Lie algebras of SO(4,1) and SO(3,2);
direction, we conclude that their generator K must R2. The Poincare Lie algebra;8
be odd under the parity operation n and the time- R3. Two "para-Poincare" Lie algebras, of which one
reversal e, while H, P, and J transform in the standard ;s isomorphiC to the ordinary Poincare Lie algebra but
4 A. Einstein, Ann. Physik 17,891 (1905). 7 W. de Sitter, Amsterdam Proc. 19, 1217 (1917); 20, 229 (1917);
• M. Dresden, in Noncompact Groups in Particle Physics, Y. C. Moller, The Theory of Relativity (Oxford Vniv. Press, London,
Chow, Ed. (W. A. Benjamin, Inc., New York, 1966), and private 1952), Sec. 134; F. GOrsey, in Group Theoretical Concepts and
communication. Methods in Elementary Particle Physics, F. Giirsey, Ed. (Gordon &
S V. Lalan [Compt. Rend. Acad. Sci. 203, 1491 (1936); and Bull. Breach Science Publishers, New York, 1964), p. 365 and additional
Soc. Math. France 65, 83 (1937)] has led an investigation in the references therein.
same spirit. However, his hypotheses are very different and much 8 E. P. Wigner, Ann. Math. 40, 149 (1939). A. S. Wightman,

stronger than ours, since he assumes the relativity group to operate "Lectures on Relativistic Invariance" in Relations de dispersion et
linearly on space-time and the space-time translations to be an particules elt!mentaires, Les Houches, 1960 (Hermann & Cie., Paris,
invariant subgroup. ]960).
 POSSIBLE KINEMATICS 1607

physically different 9 and the other is the Lie algebra of where the structure constants ~, (J, A, fl, P are real
an inhomogeneous SO(4) group; numbers and [P, K] = pH is a shorthand for
R4. The Carroll Lie algebra 10 ; [Pi' K j ] = PCJiiH.
A 1. The two "nonrelativistic cosmological" Lie Corresponding to the symmetry already noticed of
algebras; the automorphisms e and r = I1e under the ex-
A2. The Galilei Lie algebra lI ,- change P ~ K, we remark that the expressions (5)
A3. The "para-Galilei" Lie algebra 9 ; are invariant under the symmetry S defined by
A4. The "static" Lie algebra.
S:{P ~K, ~ ~A, (J ~fl, p ~ -p}. (6)
While the Lie algebras of class R have no nontrivial
central extensions by a one-parameter Lie algebra, We now require the Jacobi identity to be satisfied by
those of class A each have one class 0/ such extensions. all the triples of distinct basis elements of what must
be a Lie algebra. Due to the fact that we already took
We start by giving the proof of this theorem. In the into account rotation invariance, the Jacobi identity
following section, we discuss the physical meaning of is automatically obeyed by all the triples containing
the various Lie algebras by considering their con- at least one component of J. Of the seven types of
traction scheme, that is, the limiting processes leading remaining triples [PPP], [PPK], [PKK] , [KKK], and
from one to the other of these algebras. [HPP], [HPK], [HKKJ, only four need to be con-
sidered if we use the symmetry S. In fact, only three
Proof" The very simple and straightforward proof
types of triples give rise to constraints on the structure
consists merely in:
constants. The Jacobi identity for [HPK] requires
(i) writing the unknown Lie brackets as linear forms
on the Lie algebra in our standard basis H, P, J, K; ~fl + (JA = 0, (7)
(ii) taking into account Assumptions 1 and 2;
while for [PPK] and [PKK], respectively, it requires
(iii) imposing Jacobi identity;
(iv) using Assumption 3; (J - ~p = 0, (8a)
(v) looking for nontrivial central extensions.
fl + Ap = 0. (8b)
We have to examine the five types of Lie brackets
not involving J, that is, [H, P), [P, P), [H, K), Since the constraints (8) imply (7), we are left with
[K, K), and [P, K]. Consider, for instance, [H, Pl. only these two conditions (8) to analyze. Due to the
This bracket is odd under I1 and e. According to peculiar form of the brackets (5), namely, the homo-
Eqs. (2) and (3), it can only be a linear function of K geneity of their left-hand sides with respect to each
(Assumption 2). Since it has a vector character, the structure constant, it is seen that any nonzero structure
bracket [H, Pi] will be proportional to Ki (Assumption constant may be normalized arbitrarily by a scale
1), a property which may be written in the form change, in particular, to unit absolute value. Sign
[H, P] = ~K. More generally, when Assumption 2 changes, however, are not always permissible since
is used, one term at most survives for each bracket, they could give different real forms of the same com-
and Assumption 1 requires the admissible linear plex algebra and consequently lead to different groups.
combinations of generators to be formed only with Disregarding the question of signs leads to a classi-
the covariant tensors J, bij , Eijk' We may finally write fication of the physical algebras in the follOWing eight
types (characterized by their null structure constants) :
[H, P] = ~K, [H, K] = AP, Class R ("relative-time" Lie algebras); p ;;6 0. We
(P, P] = (JJ, (K, K] = flJ, (5) have the following types:
[P, K] = pH, °
Rl. ~ ;;6 0, A ;;6 [hence, by (8), (J ;;6 0, fl ;;6 0].
One recognizes the Lie algebras of the groups
• We emphasize that two relativity groups may be isomorphic SO(5), SO(4, 1), and SO(3, 2). More precisely, one
but completely different from the physical point of view, since the
basis elements (H, P, J, K) of their Lie algebras have a well-defined obtains two groups isomorphic to SO(4, 1), depending
physical meaning and cannot be transformed arbitrarily, except by on whether the subgroup SO(4) is generated by J and
scale changes. Moreover, even a given real Lie algebra may have
many interpretations by a change of normalization (for the para- P or by J and K. This last group, as well as SO(5),
Galilei groups for instance) or by making the corresponding group must be rejected in virtue of Assumption 3. Conse-
acting on a space which is not a homogeneous space of the group.
10 J.-M. Levy-Leblond, Ann. Inst. Henri Poincare 3, I (1965). quently, one is left with the Lie algebras of the two
See also N. D. Sen Gupta, Nuovo Cimento 44, 512 (1966). de Sitter groups which will be denoted (dS).
11 M. Hamermesh, Ann. Phys. (N.Y.) 9, 512 (1960); J.-M. Levy-
Leblond, J. Math. Phys. 4, 776 (1963). °
R2. ~ = 0, A ;;6 (hence (J = 0, fl ;;6 0).
1608 H. BACRY AND J.-M. LEVY-LEBLOND

TABLE I.. Characteristic L!e brackets for the Lie algebr~s of the various kinemat!cal groups. The terms inside parentheses refer to
the possible central extensions of the algebras. They vallish for the nonextended Lie algebras. To each Lie algebra corresponds one
physical interpretation except for the para-Galilei algebra endowed with two physical interpretations.

(dS) 1 (P)2 (P')3 (C)4 (N)! (0)2 (0')3 (SI)4

de Sitter Inh. Para- Expanding Oscillating Para-

SO(4, I) SO(3,2) Poincare SO(4) Poincare Carroll Universe Universe Galilei Galilei Static

[H,P] K -K 0 K -K 0 K -K 0 K 0
[H,K] P P P 0 0 0 p P P 0 0
[P, P] J -J 0 J -J 0 0 0 0 0 0
[K,K] -J -J -J 0 0 0 0 0 0 0 0
[P,K] H H H H H H (I) (I) (I) (I) (I)

They are the Lie algebras of the Poincare group mechanical applications. 8 Using standard methods,12
(P)8 and of inhomogeneous SO(4). This last algebra one easily shows that the four algebras of Class A
does not satisfy Assumption 3 and must be discarded. each have one type of central extension. They are very
°
R3. ex y!: 0, ;. = (hence f3 y!: 0, p, = 0).
One obtains in that case two Lie algebras iso-
simply characterized: the central element I of the
extended Lie algebra (Le., the one-dimensional
morphic respectively to those of R2, but where the extending Lie algebra) only appears in the Lie bracket
roles of space translations and inertial transforma- [P, K) which is zero for the no.nextended Lie algebras
tions have been exchanged. In opposition with the
case R2, both algebras agree with Assumption 3.
°
(p = defines the Class A). This element I thus takes
in this bracket the place occupied by the Hamiltonian
We will denote these groups by (PI). H in cases R (see Table I). This property justifies
°
R4. ex = 0, ;. = (hence f3 = 0, It = 0).
One recognizes the Lie algebra of the Carroll group
our classification of possible Lie algebras in two classes
even from a physical point of view since the nonrela-
(C)9 which satisfies Assumption 3. tivistic groups will be characterized by the property
Class A ("absolute-time" Lie algebras): p =
(hence f3 = 0, It = 0). We have also four types in this
° of additivity of masses. Another justification is pro-
vided by the fact that time will have an absolute
class, all of them satisfying Assumption 3. character for every group of Class A (see below).
Ai. ex = 0, ;. y!: 0.
III. GENERAL PHYSICAL DISCUSSION
It is shown below that the two Lie algebras of this
type are those of two "nonrelativistic cosmological There is no need to comment on the appearance in
groups," which we feel appropriate to call "Newton our classification of the de Sitter, Poincare, and Galilei
groups" (N). groups. These are well known and have been thor-
A2. ex y!: 0, ;. y!: 0. oughly studied,7.s.10 especially from the point of view
This is the Lie algebra of the Galilei group (G).10 of their role in quantum physics, for the last two ones
A3. ex y!: 0, ;. = O. at least. The surprising fact is perhaps that there are
One recognizes a Lie algebra isomorphic to the other possibilities, which we discuss now.
Galilei Lie algebra. The situation is similar to cases The Carroll group has already been introduced as a
R2 and R3 above. The group associated with this Lie second nonrelativistic limit of the Poincare group,9
algebra will be called para-Galilei group (G/).11 describing low-velocity transformations of large
A4. ex = 0, ;. = 0. spacelike intervals. This explains why we devote a
This is the Lie algebra of what we call the static whole section to the static group and another one to
group (St) for reasons to be discussed below. the Newton groups.
The above classification is summarized in Table I, In this section, we intend to examine the relation-
which clearly exhibits the Lie brackets defining the ships between all the Lie algebras listed in Sec. II
various Lie algebras. from the point of view of their physical interpretation.
We must now give the proof of the last part of the Such an approach is required by the fact that the real
theorem concerning the possible central extensions of group of symmetry is either the Poincare group or a
these Lie algebras by a one-dimensional Lie algebra. group from which the Poincare group is an approx-
In fact, this is necessary for constructing the projective imation. The mathematical aspect of the relationships
representations of the kinematical groups, repre- we are looking for is described by the process of
sentations which are of interest for the quantum- 12 V. Bargmann, Ann. Math. 59. 1 (1954).
 POSSIBLE KINEMATICS 1609

contraction, a concept which has been defined Speed-Time Contractions. One puts
by Inonu and Wigner13 in order to give a precise
meaning to some singular limits encountered when
H - EH, K - EK, E- 0, (lOa)
trying to replace "exact" kinematical groups by contracting with respect to the three-dimensional
"approximate" ones. For instance, the Galilei group Euclidean group. The physical interpretation of the
is obtained by contracting the Poincare group with contracted group is obtained in considering only low
respect to the direct product of the rotation and time- velocities and large spacelike intervals. This contraction
translation subgroups; leaving unchanged the gener- leads from the Poincare group to the Carroll group.9
ators J and H of these subgroups, one substitutes According to its very definition, such a process only
P -- EP and K -- EK in the Lie algebra and considers yields groups describing intervals connecting events
the (singular) limit E - - 0 of the Lie brackets. One is without any causal connection, hence without much
left with the Lie algebra of the Galilei group. The physical applications. The mapping corresponding
physical meaning of the contraction is very simple: to this contraction is
the factor E has affected the generators K and P, so
that the contracted (Galilei) group will describe a RI - R3, R2 -- R4, Al -- A3, A2 - A4.
situation where velocities (parameters associated to (lOb)
K) and space translations (parameters associated to Let us notice that under a speed-time contraction,
P) are small. More precisely, taking the light speed as space intervals become absolute.
the unit speed, velocities have to be small compared Space-Time Contraction. By choosing
to the unit, and spacelike intervals small compared to
timelike intervals. This is why we call such a contrac- P - eP, J -- EH, E- 0, (Ila)
tion a speed-space contraction. On this particular one contracts with respect to the group of rotations
example, one sees immediately that a given kinemat- and inertial transformations (Lorentz group). Physi-
ical group may have, beside its "intrinsic" interpreta- cally, this means that we consider very small units of
tion (for insta.nce Newtonian mechanics for the space and time. A group obtained in this way describes
Galilei group), one or more "approximate" in.ter- local transformation properties, but, contrary to the
pretations (the Galilei group is an approximate preceding contractions, for arbitrarily large inertial
symmetry of special relativity but also of a de Sitter transformations. The space-time contraction realizes
universe). the following mapping:
The contraction of a group always being defined
with respect to a particular subgroup, one can con- RI -- R2, R3 -- R4, Al -- A2, A3 -- A4.
sider four types of physical contractions for general (Ub)
kinematical groups, since there are four possible
rotation-invariant subgroups of any kinematical Because of the local character of the contracted
group; indeed, it is easily seen, looking at Table I, groups, it seems natural to call RI, R3, AI, and A3
that the only rotation-invariant Lie subalgebras cosmological groups and their contracted groups R2,
common to all eight types of kinematical Lie algebras R4, A2, and A410cal groups.
are generated by (J, H), (J, P), (J, K), and (J). Let Before examining the fourth kind of contraction,
us discuss in turn the corresponding four types of we will refer the reader to Fig. 1, where the eight
contraction: types of Lie algebras have been put at the eight
Speed-Space Contraction. It is defined by vertices of a cube, the edges of which describe the
contraction processes we have just discussed. Each
P--EP, K--EK, E-O. (9a) upper face of the cube is transformed under a given
It is the contraction with respect to the rotation and type of contraction into the opposite face.
time-translation subgroups which leads from the General Contraction. Here we put
Poincare group to the Galilei group.13 This contraction
process corresponds to the passage from relative to P -+ EP, H -+ EH, K -- EK, E-- 0, (12)
absolute time and maps in a one-to-one correspond- contracting with respect to the rotation group which
ence the groups of Class R onto the groups of Class A : alone is unchanged.
RI -- A I, R2 -- A2, R3 -- A3, R4 __ A4. (9b) This contraction combines the features of the pre-
ceding ones and will lead to a description of local and
13 I. Segal, Duke Math. J. 18, 221 (1951); E. Inonu and E. P.
Wigner, Proc. Nat!. Acad. Sci. U.S. 39, 510 (1953); 40, 119 (1954); small inertial transformations. As we will see, it is a
E. Saletan, J. Math. Phys. 2, 1 (1961). rather drastic operation from a physical point of view.
 1610 H. BACRY AND J.-M. LEVY-LEBLOND

Thus they describe the transformation properties of

large spacelike intervals, forbidding causal relations
betw~en events. The physical interest of these groups
then IS very much reduced. One may notice that, for
these two groups, [H, K] = 0 (see Table I), which
means that inertial transformations do not act
effectively on time translations; this is consistent with
the large spacelike character of the considered inter-
vals, but underlines the unphysical nature of the
groups, since ordinarily an inertial transformation
fol~owed by a time translation yields a space trans-
FIG. 1. The contraction scheme for the relativity groups. latIOn (waiting a certain time after boosting an object,
Relative-ti~e groups: (dS), (P'), (P), (C).
Abso~ute-tlme groups: (N), (G'), (G), (St).
one finds it displaced !). The same considerations
Relative-space groups: (dS), (N), (P), (G). could seem to apply to the para-Galilei and static
Absolute-~pace groups: (P'), (G'), (C), (St).
CosmologIcal groups: (dS), (N), (P'), (G').
groups. However, we show that these groups are
Local groups: (P), (G), (C), (Sf). endowed with special significance.
Remark: The effect of th.e symmetry S of Eq. (6) is equivalent to a (b) The para-Galilei group (G') can be obtained
symmetry of the cube wIth respect to the plane containing the from (dS) in the following way:
vertices (dS), (N), (C), and (Sf).
P-+€p, H-+€H, K-+€2K, €-+O,
All groups are contracted in this way into the static and so describes a situation where timelike and
group. spacelike intervals are reduced but the velocities are
We notice that geometrical concepts are naturally completely negligible, that is to say, a situation of a
associated to each kind of contractions. For instance, static "de Sitter" universe. Looking at the cube of
the group obtained via a speed-space contraction Fig. 1, one can expect an analogous interpretation for
describes space-time properties in the neighborhood the static group relative to the Poincare group but the
of a worldline. This is due to the fact that the con- fact that this group can be reached directly through a
traction is made with respect to the subgroup which general contraction provides it with even more
leaves invariant a given worldline. On the other hand, general properties.
space-time contractions are associated with events, (c) The groups (N) acquire an obvious meaning:
points in space-time. they are obtained from the de Sitter groups by per-
We may now state the main result of this section: forming a speed-space contraction. Thus they describe
the eight possible types of kinematical Lie algebras the low-velocity transformations of the universe at
may all be obtained from one among them (Rl) by a large endowed with an absolute time, hence their de-
suitable sequence of contractions14 (i.e., a contraction nomination as "nonrelativistic cosmological groups."
in the sense given by Saletan13). The detailed con- (d) It is a remarkable fact that, without any
traction scheme is much better understood by looking additional hypotheses, each of our groups can be
at Figs. I and 2 than by reading a verbal description. considered as a group of transformations acting on
We note immediately that the Poincare group is a space-time coordinates. Indeed, it results from our
contracted group of the two de Sitter groups only; this basic assumptions that rotations and inertial trans-
property is essential, since it follows that the only formations automatically form a subgroup of any
possible exact symmetry groups compatible with our kinematical group; this is seen on Table I, where J
assumptions are the Poincare and the de Sitter groups. and K clearly form a Lie subalgebra. As a consequence,
We examine in detail this statement in our conclusion. there exists a four-dimensional homogeneous space
The cube of Fig. 1 enables us to understand the for each kinematical group, namely, its quotient by the
relationships between all the groups we obtained and six-dimensional subgroup of rotations plus inertial
to find their interpretations as approximated sym- transformations. This inhomogeneous space may be
metries of the Poincare or of the de Sitter symmetries. identified with the manifold of the space-time trans-
(a) The groups (Pi) and (C) are obtained respec- lations 15 and gives a purely group-theoretical definition
tively from (dS) and (P) via a speed-time contraction. of space-time itself. From the group law, one immedi-
ately obtains the formulas defining the action of the
14 Conversely, the eight possible kinematical Lie algebras may
be considered as "deformations" of the static Lie algebra. For a 15 Note that the space-time translations do not form a subgroup
discussion of this concept see M. Levy-Nahas, J. Math. Phys. 8, in general. It is a subgroup in the Poincare, Carroll, Galilei, and
1211 (1967). static cases. This subgroup is, in fact, an invariant subgroup.
 POSSIBLE KINEMATICS 1611

FIG. 2. Top of the cube of Fig. 1 (detail).

group on space-time coordinates. This is where we parameter "phase group." Contrarily to the de Sitter
see that time intervals are not invariant under inertial and Poincare groups, the Galilei and static groups
transformations for type R groups, while they are for have nontrivial extensions. The Lie algebras of the
type A groups. Indeed, this results directly from the extended groups differ from those of the original
presence or absence of the time-translation generator group by the addition in the Lie brackets of scalar
H in the derived algebra, i.e., in the "right-hand side" terms, multiples of the phase-group generator I. As
of the Lie brackets (see Table I). Of course, we have indicated in Table I, for kinematical groups, such
no guarantee that the group acts effectively on terms can always be reduced so as to appear only in
space-time, i.e., that every group element induces a the commutator [P, K]. For the Galilei group, it is
nontrivial transformation. As a matter of fact, for the most important that only the nontrivial projective
static group (and the para-Galilei group as well), representations have a physical interpretation; the
inertial transformations do not act on space-time; vector representations do not allow for any sensible
this results from the commutation relations [H, K] = notion of localizability.16 The phase generator I in this
°
0, [P, K] = (Table I). But it is an advantage of our
approach not to have considered beforehand the
case physically appears as the mass operator. Since
the mass obeys a superselection rule,1O·12 we must
kinematical groups as effectively acting on space- consider systems with a definite mass value, say m.
time, since this would have eliminated the static Within their state space, one has
group, for instance, despite its real physical content.
I=ml, (13)
. (e) By identifying space-time with a homogeneous
space of the Lie algebra, we have excluded some other so that the commutation relation
possibilities offered in making the group acting on the
(14)
closure of a homogeneous space. The group proposed
by Dresden5 in his model is claimed to fulfill all our is nothing but the canonical commutation relation
three assumptions. In any case space-time cannot be
(15)
identified as a homogeneous space of the Dresden
group. between the momentum P and the center-of-mass
IV. THE STATIC GROUP position operator Q = (ljm)K. Taking Q as a basis
We saw that the static group is reached from any vector of our Lie algebra instead of K, we have to
of the other kinematical groups by performing the rewrite the Lie bracket involving the Hamiltonian
whole sequence of physical contractions in any order,
or the general contraction directly. This somewhat [H, Q] = ~P, (16)
obscures its possible physical meaning. rn
The interpretation of the static group appears most which is just the velocity operator. But then we see
clearly when we relate it to its closest physical neighbor, that by taking the limit m -+ 00 we obtain a new Lie
the Galilei group, by a slight extension of our con- algebra with the property (15) and
siderations on the contraction scheme. It has been
mentioned that, in quantum physics, the state space
[H, Q] = 0. (17)
of a system invariant under a group of transformations This is just the Lie algebra of the extended static
is a representation space for a projective (i.e., up to a group! In more technical terms, we have contracted
factor) representation of the group, that is, a represen-
16 E. Iniinu and E. P. Wigner, Nuovo Cimento 9, 705 (1952); A.
tation of a centra I extension of the group by a one- S. Wightman, Rev. Mod. Phys. 34, 845 (1962).
1612 H. BACR Y AND J .-M. LEVY-LEBLOND

a central extension of the Galilei group into a central representation space the space of (2s + I)-component
extension of the static group with respect to the square-integrable functions in momentum space, the
rotation and space-time translations subgroup; that is Lie-algebra representation is
taking the limit
H = E = const,
K -* eK, 1-* eI, e -* O. (18) P=p,
(21)
If we now point out that, for the same reasons as in K=iV 11 ,
.the Galilean case (lack of localizability), the vector J = - ip x V11 + S,
representations of the static group have no physical
where the three matrices Si (i = 1, 2, 3) generate a
meaning, the interpretation of the group is quite
(2s + I)-dimensional representation of the rotation
clear. We have just shown that this group represents
group. In (21), the generators have been redefined so
"infinitely massive" systems. One sees how, due to
as to be Hermitian, having been multiplied by (i).
the m -* 00 limit, the position operator of the system We see that:
now commutes with the Hamiltonian (17), instead of
giving the velocity operator (16). This, of course, (1) a "static elementary system" can possess a spin,
means that the system cannot move-whence the as can Galilean and relativistic systems;
name of the group. The key point is that such a static (2) the energy E of a static system is a constant,
system is invariant not only under the Euclidean independent of its momentum p;
motion group and time translations; the additional (3) under a pure inertial transformation (u), the
invariance under pure Galilean transformations, momentum of a static system changes according to
although broken by the infinite-mass limit, leaves its p-*p + u.
imprint by ensuring the existence of a position oper- All this is perfectly in keeping with the intuitive
ator, endowing the system with nice localization picture of a static system as having a constant energy,
properties. rotational degrees of freedom, and being able to ex-
It may be interesting to give the group law. Calling change momentum in arbitrary amounts while its
(b, a, u, R) = ebHea,Peu.KeD.J (19) energy does not change. Such results obviously would
not obtain if one only considered Euclidean invariance.
the generic element with b a time translation, a a space As an application, it is amusing to consider the
translation, U a pure static inertial transformation ,17 various static models of quantum field theories, such
and R a rotation, one easily computes the multiplica- as the Chew-Low modeJ.19 One easily sees that they do
tion law have the static group as an invariance group. It may
(b', a', u', R')(b, a, u, R) be noted that Feynman had proposed to modify the
Chew-Low model so as to take into account Galilean
= (b' + b, a' + Ra, u' + R'u, R'R). (20)
invariance. 2o It has been pointed out elsewhere 21 that
The peculiar feature of this law, as opposed to its this cannot be done consistently because of the
Poincare and Galilei counterpart, is that the pure Galilean mass superselection rule. Accepting, how-
inertial transformations do not act on the space-time ever, the intrinsically static nature of the model, one
(identified to the space-time translations subgroup).18 need not introduce additional interaction terms, as
Any system is at rest in any inertial reference frame and proposed by Feynman, to have a model with some
the inertial motions are no motion at all. remnant of Galilean invariance, namely static in-
Finally, we can construct the irreducible unitary variance. Anyway, the new terms had the effect of
projective representations of the static group, using badly spoiling the agreement between the predictions
the standard "little group" technique. s Taking as our of the standard model and the experimental results. 22
17 Before the contraction, the parameter u is linked to the velocity V. THE NEWTON GROUPS
v of the pure Galilean transformation by u = mY. The contraction
process amounts to taking the limits m -+ 00, v -+ 0 with u fixed. The Newton groups are characterized by a non-
18 However, for the nontrivial central extensions of the static
group, the inertial transformations act on the phase group. We have vanishing Lie bracket [H, P], proportional to the
(It, b', a', u', R')(O, b, a, u, R) 19 S. S. Schweber, Introduction to Relativistic Quantum Field'

+ +
= (0' 0 u'· R'a, b' + b, a' + R'a, u' + R'u, R'R), Theory (Rowand Peterson, Evanston, 1961), Chap. 12 and original
references therein.
where 0 is the generic element of the phase group (a real number). 20 R. P. Feynman, Proceedings of the Third Annual Conference on
It is very amusing to notice that such an extension of the static High Energy Nuclear Physics, Rochester, 1952 (University of
group is isomorphic to the (trivial) extension of the Carroll group. Rochester, 1953), p. 87.
The isomorphism is simply realized by exchanging the roles of the 21 J.-M. Levy-Leblond, Commun. Math. Phys. 4, 157 (1967).

phase group and the time-translation group (H ~ I). 22 E. M. Henley and M. A. Ruderman, Phys. Rev. 90, 719 (1952).
 POSSIBLE KINEMATICS 1613

boost generator K. Depending on the sign of the origin x = 0 for definiteness) to a pure inertial
coefficient (J. [see Eq. (5)], we have one of two possible transformation, are no longer uniform motions, but
real forms for the Lie algebra. We choose to write are given by
[H, P] = ± (1/'T 2)K, (22) . h -t for
x = V'T sm (26a)
N+
'T
where 'T is a characteristic time, which might be or
chosen as the natural time unit, but which we prefer t
to keep apparent in our expressions. Depending on the x= V'T sin- for N_. (26b)
'T
sign in (22), we will have a group N+ or N_.
In order to acquire some familiarity with the new Inertial motions thus are either exponentially accel-
groups, let us first compute the group law by direct erated (N+) or oscillating (N_). Correspondingly, N+
exponentiation of the Lie brackets. We denote the and N_, respectively, describe expanding and os-
generic element by cillating universes.
It seems, then, that both groups have features very
(b, a, v, R) = ebfIea,Pev.KeD.J (23)
much like the de Sitter groups? from which they
(the vector n defines the axis and angle of the rotation detive by speed-space contraction, SO(4, 1) leading
R). Multiplying together two such elements, re- to N+ and SO(3, 2) to N_. In other words, despite the
ordering the product in the same normal form (23) nonrelativistic limit, the effects of space-time curva-
by repeatedly using Baker-Hausdorff formula and the ture are present, due to the fact that we are still
Lie-algebra properties of the infinitesimal generators, considering the universe on a large scale of time.
we obtain These effects disappear when we take the local limit,
(h', a', v', R')(b, a, v, R) i.e., 'T ~ 00, which immediately yields the Galilean
formulas, as expected, for both N+ and N_. Notice
= (h' h,
+ cosh ~ a' + 'T sinh ~ v' + R'a,
'T 'T
that in such a universe there is an absolute time: as in
the Galilean case, simultaneity of two events is
preserved by an inertial transformation. We see, then,
cosh!? v' + ! sinh!? a ' + R'v, RIR) for N+,
'T T 'T that the groups N may be interpreted as describing
(24a) the kinematics of nonrelativistic' universes at large. 23
We propose to call "Newton groups" these nonrela-
( h' + b, cos!?'T a' + 'T sin!?'T v' + R'a, tivistic cosmological groups which bear to the
Galilei group the same relation as the de Sitter groups
cos!? v' + ! sin!? a' + R'v, R'R) for N _ . to the Poincare group.
'T 'T 'T
It is of interest for possible applications to quantum
(24b) physics to construct the unitary irreducible representa-
The group can be considered as a group of trans- tions of the Newton group. As in the Galilei and static
formations of the space-time manifold. Under a cases, only the projective representations may have a
general transformation (h, a, v, R), the coordinates physical meaning, for lack of a proper localizability
(x, t) of a physical event are transformed according to notion for vector representations. There are two
invariants of the extended Lie algebra, which we
X' = Rx + V'T sinh ~ + a cosh ~ , write:
( 'T 'T
for N + ' (2Sa) Ql = 2IH - p2 ± (lj'T2)K2 == 2lU for N±, (27a)
t' = t + b,
or Q2 = (IJ - P x K)2 == 128 2 • (27b)
, R . t t The interpretation is very similar to the Galilean case:
X = x + V'T sm - + a cos - ,
I appears as a "mass," obeying a superselection rule;
( 'T 'T
for N_. (25b)
t' = t + h, from Ql one defines an "internal energy" U and from
Qz a "spin" S. The most novel feature of a "particle"
Notice that, as in the de Sitter case, these are nonlinear described by such a representation is that its energy is
transformations; more precisely, the new spatial
coordinates do not depend linearly on the time. This 23 A perfectly consistent, nonrelativistic, Newtonian model of
expanding universe has been described by C. CaHan, R. H. Dicke,
is in direct contrast to the Poincare and Galilei groups. and P. J. E. Peebles, AT?' J. Phys. 33, 105 (1965). We do not expect it,
Physically, it means that inertial motions, obtained by however~ to b~ mvanant under our nonrelativistic cosmological
group, smce thIS group, as the de Sitter group from which it comes
subjecting a motionless material point (located at the certainly describes an empty universe, contrarily to the cited model:
1614 H. BACRY AND J.-M. LEVY-LEBLOND

no longer invariant under a translation, as one sees The present work could be extended by relaxing
directly in Eq. (22). Also it is amusing to notice that the conditions that parity and time reversal be auto~
in the N_ case the kinetic energy of the particle, that morphisms of the relativity groups. This can be done,
is, but the number of possible cases become very large
T = H - U = 1/2I[P2 + {l/T2)K2], (28) and many of them seem difficult to interpret in physical
is quantized, which is not surprising in view of the terms.24 This is why in this paper we have presented
"compactness" of the corresponding universe. The only the simplest approach.
It seems important, in our opinion, to recall in the
oscillator levels have a separation
conclusion one of the main consequences of our
(29) theorem: With the requirements of kinematical
in agreement with the uncertainty principle, since T rotation, parity, and time-reversal in variance , there
may be thought of as the "lifetime" of this oscillating exists only one way to "deform" the Poincare
universe. Of course, bE is absurdly small (bE ~ 2 X group, namely, in endowing space-time with a certain
10-33 eV for T ~ 1010 yr). curvature. 25
ACKNOWLEDGMENTS
VI. CONCLUSIONS
We think it rather remarkable that the very simple The authors are grateful to F. Lun;at and J. Nuyts
conditions which we require so severely constrain the for stimulating discussions and their interest in this
possible relativity groups that we found only a small work. One of us (R. B.) wishes to acknowledge the
number of more or less degenerate cousins of the hospitality extended to him at The Institute for
de Sitter group. Besides the known de Sitter, Poincare, Advanced Study. Part of this work was done while the
and Galilei groups, there are two physically inter- other author (J.-M. L.-L.) benefitted from the
esting cases. The static group is appropriate to the hospitality of the University of Rochester.
description of the various static models where the
•• H. Bacry, J.-M. Levy-Leblond, and J. Nuyts (unpublished).
studied particles are supposed to have infinite masses. '0 This result is, of course, related to the statement that "the only
The Newton groups present features quite close to deformations of the Poincare group are the two de Sitter groups" (see
Ref. 14), but the quoted result concerns abstract groups, whereas our
those of the de Sitter groups, but in a nonrelativistic work is based on concrete considerations. It is not difficult to prove,
situation, so that they are perhaps closer to one's for instance, that the Poincare group can be "deformed" in groups
isomorphic to itself. An example can be found in H. Bacry and
intuition and may be used as a simple model to N. P. Chang "Kinematics at Infinite Momentum" [Ann. Phys.
understand some effects of space-time curvature. (N.Y.) (to be published)]. See also Ref. 9.