Lorentz Invariance Vs Temperature QFT

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Letters in Mathematical Physics 11 (1986) 73-80.

73
9 1986 by D. Reidel Publishing Company.

Lorentz Invariance vs. Temperature in QFT


IZUMI OJIMA
Research Institute for Mathematical Sciences. Kyoto University, Kyoto 606, Japan

(Received: 8 August 1985)

Abstract. The Lorentz invariance of a relativistic Q F T is shown to be broken spontaneously at finite


temperature with a zero-energy Goldstone spectrum which does not necessarily carry a one-particle
structure. The Lorentz behaviour of the temperature is determined and the relation between Gibbs states
m different Lorentz frames is clarified.

I. Introduction
Although the problem of how to incorporate the relativity theory into thermodynamics
and statistical mechanics is quite old [1], it seems to be far from being settled. For
instance, the Lorentz transformation property of the temperature is still one of the main
issues in dispute in relativistic thermodynamics [2]. When one picks up a Gibbs state
in a given relativistic quantum field theory (QFT), it is intuitively clear that it cannot
be a Lorentz-invariant state as the 'heat bath' maintaining its equilibrium somehow
determines an 'absolute' rest frame. But, how should it behave under the Lorentz
transformation? And, what type of symmetry-breaking mechanism is involved in this
breakdown of Lorentz invariance?
In a system with infinite degrees of freedom, such as QFT, we must be prepared for
the possible occurrence of unitarily inequivalent representations: The answer to the
second question above is that the Lorentz noninvariance of a Gibbs state should be
interpreted as the spontaneous breakdown of the Lorentz boost symmetry with a
Goldstone state but (probably) without a Goldstone particle. The present author
announced this conclusion in [3], a detailed account of which, however, has not been
given in the form of a published paper (apart from some seminar talks [4]). In view of
recent applications of QFT with thermal effects (as well as curved spacetime
background) to cosmology and also of the interesting relation of this symmetry breaking
with that of supersymmetry at T ~ 0 K [5, 6] where some confusion has occurred, this
kind of problem is not of purely academic interest, and the present Letter will be
addressed to the task of answering the above questions in some detail.
In the next section, the reason is explained in a general form why the spontaneous
breakdown of the Lorentz boost symmetry is unavoidable at a finite temperature
T r 0 K. The zero-energy Goldstone spectrum responsible for this symmetry breakdown
is identified in Section 3, which does not (necessarily) exhibit particle structure. In
Section 4, the Lorentz behaviour of the temperature is shown to be fixed from the
standpoint of statistical mechanics, on the basis of which the relation between Gibbs
74 IZUMI OJIMA

states in different Lorentz frames is clarified in terms of family of states: namely, the
Lorentz behaviour of the temperature is given by T = T O~ 1 - V2/c 2, and the Gibbs
states characterized by the inverse-temperature four-vectors/~' constitute a Hilbert
bundle over the manifold of Lorentz frames.

2. Spontaneous Breakdown of Lorentz Invariance at T ~ 0 K


In the following discussion, we need not specify a particular model for our relativistic
QFT but, just for simplicity, a scalar field q~(x) will be chosen in the illustration, if
necessary. As long as the theory is Poincar6 invariant at T = 0 K, the dynamical law
('equation of motion') governing the system should be Poincard invariant at the operator
level also at T # 0 K. Now, let 09 be a Gibbs state at temperature T = 1/kflgiven as an
expectation functional on the algebra of field operators. In the conventional approach,
it is expressed as
09(A) = Tr (e- ~z-/a)/Tr (e - t ~ ) . (2.1)

Precisely speaking, some ultraviolet and volume cutoffs should be introduced in order
for the r.h.s, of(2.1) to make sense and, after some renormalization procedure, the limit
removing the cutoffs should be taken. In this limit, each of the numerator and
denominator will lose their own meaning in general, but their ratio 09(A) will be
meaningful. Then, the essence of a Gibbs state will be condensed into the KMS
condition [7, 8]
09(A(t)B) = 09(BA(t + i[3)),* (2.2)

or in Fourier transformed form,

FT[09(AB)] (p) = I dax eip'(x-Y)09(A(x)B(Y))

= et~oFT[09(BA)](-p), (2.3)

where A(t) [or A(x)] is the [space-]time translation of A by t [or x]. Because of the
necessity fot the cutoffs in (2.1) which introduce the explicit breaking of Poincar6
invariance, we take (2.2) or (2.3) as our starting point in preference to (2.1).
Since a specific time axis is picked up in (2.2) or (2.3) to characterize 09 as a
temperature state, however, our Gibbs state 09 violates the symmetry under the Lorentz
boost transformations: because of the Lorentz noninvariant factor e ~p~ FT[ 09(AB)] (p)
and FT[ 09(BA)] ( - p ) in (2.3) cannot both be Lorentz invariant unless they vanish. This
violation of Lorentz boost invariance can also be checked explicitly at each order of
perturbation expansion valid in thermo field dynamics [9], where the (1.1) component
of 2 x 2 matrix propagator is given at the tree level by
i 1
FT[09(rpgo)] (p) + 2re 6(p 2 - m 2) , (2.4)
p 2 _ m 2 + i/3 e BIp~ - 1
* A(t) should be a s s u m e d to allow analytic continuation in t.
LORENTZ INVARIANCE VS. TEMPERATURE IN QFT 75

which is not Lorentz invariant:

co([Mo,, T~p(x)~p(y)]) # O . (2.5)

In this way, we encounter the spontaneous breakdown of the Lorentz boost symmetry
at T # 0 K. The essence of this symmetry breaking can also be understood in a more
general context when our relativistic QFT at T # 0 K is adapted to the algebraic
formulation of statistical mechanics [8] based upon a C*-algebra and the KMS
condition (2.3). Namely, any symmetry operation a interpreted as an automorphism of
the C*-algebra 9~ describing our physical system is easily seen to be implemented, in the
G N S representation space ($~, Trio,f~o~)of the Gibbs (KMS) state co o f % by a unitary
operator U commuting with the modular operator Ao and the modular conjugation
operator J,o associated to o9, whenever a leaves the state 09 invariant:
oJ o a = co ~ 3 U : unitary in -Do~

s.t. rroj(o(A)) = Ur%(A)U- 1 , forA ~ 9~,

and

[A~o, U] = [Jo,, U] = 0. (2.6)


This entails the commutativity of a with the time translation a t of 9~which is essentially
identical to the modular automorphism ~,o:

~,(A) = Aj'/BA --,,A"/~, for A ~ ~t,o(9~)" , (2.7)

a(~,(A)) = ~,(a(A)). (2.8)

Thus, any symmetry not commuting with the time translation, such as the Lorentz boost
symmetry, cannot leave a Gibbs state invariant, which immediately concludes the
breakdown of Lorentz invariance at T ~e 0 K.

3. Goldstone Spectrum Without Particle Structure


Now, letting ~ o , ( = Xo| - x;| with | symmetric energy-momentum tensor) be
the Noether density current of Lorentz boost Mo;, we apply the KMS condition (2.3)
to the 'Goldstone commutator' of Lorentz boost:

o # 2~ ~(po)~O~(a))
= 21z b(po)CO([ie'Mo,,A])

lim f dax eipxco([ie~ooi(X), A])


p~O d

= lim (1 - e'P~ i f dax e'PXcO(~oo;(x)A), (3.1)


p~O d
where A will be some such 'composite' operator as qg(y)~p(z) (smeared by suitable test
functions). From the above, it is clear that the density current ~oo;(X) creates some
76 IZUMI OJIMA

zero-energy spectrum responsible for the spontaneous breakdown of Lorentz boost


invariance in conformity to the Goldstone theorem.* When one considers this
phenomenon, for instance, in the massive free scalar field theory, this statement would
sound strange somehow, because we have there no massless particles nor interaction
creating massless bound states.
To resolve this paradox, we must recall the presence of the thermally-excited particle
distribution contained in the temperature 'vacuum' I0(/~)) which is just the GNS cyclic
vector of a Gibbs state 09. Because of it, an 'annihilation' operator a(p) does not
annihilate [0(/~)) but creates a 'hole' or a 'particle' with negative energy,
a(p) 10(1~)) = J,oA,o
1/2a + ( p ) 1 0 ( 3 ) ) = J~e-#a~
(3.2)
= e#P~ 10(/~)) =~ 0 ,
where g,o - -/~- ~ lnA,o is identified by (2.7) with the generator of time translations in
rio.In more general terms, the relation between J o and Ao,

JoAoJ,o = A~ t or JolgloJo = - g o (3.3)


clearly shows that the energy spectrum in ~o extends symmetrically on both the positive
and negative sides, violating the spectrum condition which is characteristic of the
vacuum state at T = 0 K. Thus, it is now easy to create, without any binding forces, a
zero-energy state by letting q approach to p in a+(p)a(q)10([3)),

Igla + (p)a(q) 10(/~)) = (Po - qo) a + (p)a(q) I0(]~)) V=-~0 9 (3.4)


This is just an example of the Goldstone zero-energy spectrum contained in cpcpwhich
causes the breakdown of the Lorentz boost invariance in (3.1) with A = cp~p.Note that,
without the spectrum condition, the presence of the zero-energy spectrum in (3.1) does
not necessarily mean the existence of the zero-energy one-particle state, in sharp contrast
to the case of the vacuum state satisfying the spectrum condition. Thus, the spontaneous
breakdown of the Lorentz boost invariance at finite temperature following inevitably
from the noncommutativity of the Lorentz boosts with the Hamiltonian, occurs in the
absence of any masslessparticles in a consistent way with the Goldstone theorem which
requires the presence of the zero-energy spectrum.

4. Degeneracy of States Due to the Spontaneous Breakdown of Lorentz


Boosts and Lorentz Behaviour of Temperature
Because of the spontaneous breakdown of Lorentz boosts at T ~ 0 K, one cannot
implement a Lorentz boost transformation in a given representation space $ o of a Gibbs
state co. However, it should still be possible to observe a system from an inertial frame
moving with a constant velocity relative to the rest frame. What happens then?
In view of the invariance of co under the rotation group H ( = SO(3) or SU(2) in four
dimensions), what is relevant to us is the induced representation of the Lorentz group
* This is in disagreementwith the claimof [6] whichdeniesthe occurrenceofbosonic zero-energyGoldstone
spectrum.
L O R E N T Z I N V A R I A N C E VS. T E M P E R A T U R E IN QFT 77

G ( = SO(3, 1) or SL(2, C) in four dimensions) obtained from the unitary representation


(P, 6o,) of H. This is a unitary representation U p of G defined on the space L2(E) of
square integrable cross-sections of a Hilbert bundle E associated to the homogeneous
bundle G ~ G/H with the Hilbert space fi,o as its standard fibre:

E-Gxp(1-i)5.~= [I E,&G/H, (4.1)


u e G/H 'l ' II

s ~EF(G/H'E);']~II2-~G/H
du 'l ~(u) H2Eu < nt- QC }, (4.2)
where the measure du on G/H is the quotient measure derived from the left invariant
Haar measure on G. Through the identification of ~ 9 F(G/H, E) with an 6~-valued
(p, H)-equivariant function r on G,

~(P(g)) = g" ~(g), (4.3)


~(g'h)=p(h) - l ~ ( g ) , VgeG, VheH, (4.4)
U~(A)r for A e G is defined by

(UP(A)~)'(g)- ~ ( A - l g ) , g e G. (4.5)

In the following, we investigate the physical meaning of the ingredients involved here,
G/H, E and L2(E), through which the Lorentz behaviour of the temperature and
temperature states will be clarified.
In the usual case of spontaneous symmetry breaking of G down to H in the vacuum
states, the homogeneous space G/H is the manifold of Goldstone fields (in the nonlinear
realization of G). In our case of the Lorentz boost breakdown at T # 0 K, however, it
seems difficult to give such a meaning to G/H because of the lack of the particle
character of the Goldstone spectrum. Instead, it represents here the totality of different
Lorentzframes. To see this, let us recall that our Gibbs state o9 is defined in reference
to the time axis specified by a unit vector uo = (1, 0) in the 'rest frame' with the rotation
group H as its isotropy group: Hu o = uo. A Lorentz transformation A 9 G takes us into
a moving frame with a velocity vector

u ~ = (1, V)/xfl - V~ = Au o = AHu o ~_ AH 9 G/H.

Now that each point u" = (1, V)/x/1 - V 2 in G/H is uniquely and smoothly charac-
terized by the relative velocity V, G/H is a contractible space diffeomorphic to ~Rd - ' (in
d - dim. spacetime). Therefore, our fibre bundles, G ~ G/H and E& G/H, over G/H are
all trivial bundles:

G ~- (G/H) x H, (4.6)
E = (G/H) x p(H)~o ~ (G/H) X S5o~ (4.7)
U U

E.-- q-'(u)-~ {u} x $~. (4.8)


From the above interpretation of G/H, it should be clear that each fibre E , of E on
u ~ G/H corresponds to the G N S representation space of the Lorentz-transformed
78 I Z U M I OJIMA

Gibbs state on a moving frame u ~ G/H. Although different fibres Eu and E~. for u # v
are mutually orthogonal in the Hilbert bundle E (spontaneous breakdown of Lorentz
boost invariance!), they are all isomorphic to the standard fibre ~ , and are related to
each other by the equivariant action of G on E. Therefore, we should find an isometry
V~ between $~o and Eu for each u ~ G/H:

V,: ~o ~ E , , (4.9)

I[ ~][,~ = I] V.~I[E. 9 (4.10)


For this purpose, it is convenient to choose and fix a global smooth cross-section B of
the trivial bundle G • G/H [(4.6)]:
B: G / H ~ G , p oB = Ida/" . (4.11)

U F-~ B u

Letting ~A(A) be the Lorentz transform of an operator A in our physical system ~ by


a Lorentz transformation A ~ G, we identify E~ with the space of G N S representation
(E,,x~,lO(fl, u ) ) of ~ associated to the Lorentz-transformed Gibbs state
r = r o 0"~ 1 :

(O(fl, u)p ~.(A)10(A u)) = ~.(A)


= ~(0"B I(A)) = (0(fl)[ /'Co(0"B I ( A ) ) [ 0 ( f l ) > , (4.12)

E . = x.(ga)[0(fl, u)) . (4.13)

The mapping V. defined by

V.x~,(A) 10(fl)) = zc.(aB.(A))10(fl, u)) , A ~ li, (4.14)

is easily seen to be an isometry between .~, and E . satisfying

Vun~,(A)VZ' = ~z,(aB,(a)), forA~. (4.15)

In view of (4.12) and the KMS condition (2.2) or (2.3) characterizing m as a Gibbs
state, we obtain

co,(A(x)B) = cou(BA(x ~' + i(B.~fl))


= eo.(BA(x ~' + iflu~')), (4.16)

or

FT[og~(AB)] (p) = e#""P"FT[~o.(BA)] ( - p ) , (4.17)

as far as A and B are supposed to obey such standard Lorentz transformation law as
af~ '(A'(x)) = S(A)~.AJ(A- ' x ) . (4.18)
LORENTZ INVARIANCE VS. TEMPERATURE IN QFT 79

Equation (4.17) says that a Gibbs state co, in a moving frame u e G / H should be
characterized by a factor e - B,~o,,instead of the 'Boltzmann factor' e - Bpoin the rest frame
u o 9 G/H, where the 'inverse temperature' Lorentz vector flu is related to fl by

flu - flu+,. (4.19)


Namely, statistical mechanics and the (generalized) Lorentz covariance principle
require that the temperature should be transformed under the Lorentz boost according
to (4.19). In other words, the temperature T = (flo)- ~ in a moving frame with a relative
velocity V should be related to the proper temperature TO -- fl- 1 by

T = T O N~]- m2/c 2 . (4.20)

In this way we have a family of Gibbs states in different Lorentz frames on which
the Lorentz group G acts equivariantly:

E C- ,E q I 6 'H

8,,, A9

'Au ~E q ~GIH. (4.21)

The effect of Lorentz transformation A 9 G on co. - coB.. = coB.can be summarized as

coB,(O'A 1(-,4)) = co(AB)u(A), A 9 91, (4.22)

in view of the rotation invariance of (oB, (4.12), (4.19), and (4.21).


Now, the task of writing down the explicit formula of U p adapted to the trivialization
(4.11) of G P~ G/H, goes exactly in the same way as the Wigner construction of the
Poincar6 group representation for a massive particle, by replacing the momentum p"
(p+'p~, = m 2) in the latter case with the unit velocity vector u" (uUu, = 1) of a moving
2 ~
frame. Namely, through the identification of Lp(G, ~ ) ~ ~<--, ~~ o B - ~ 9 L2(G/H, 5o~),
we obtain

(UP(A)r (u) = p(h(A- l, u)- 1 ) ~ ( A - l u ) , (4.23)

where

h(A, u) --- BA~1AB, 9 H (4.24)


is an H-valued G-l-coboundary: ~r

h(A 2, A, u ) h ( A , , u) = h(A2A1, u ) . (4.25)

~r If G ~ G/H were not trivial, we would encountered a nontrivlal G-l-cocycle which would lead to the
sLmdar problem to the anomalies. Cf. [10].
80 IZUMI OJIMA

In this Hilbert space

= L2(G/H, fio~) = (du) 1/2 fi~,,

the operator algebra 9a o f our physical system is represented Lorentz-covariantly as

(Tr(A)~) (u) = r~,o(a~' (A)) (~(u)), (4.26)

rC(av(A)) = UP(V)Tt(A)UP(V) - ' , (4.27)

for A ~ 9a, ~ ~ ~, u ~ G / H and A e G. Because o f the difference in the interpretation of


G/H, however, the physical interpretation is rather different from that o f the Wigner
construction of particle states: ~ ~ ~ here represents an incoherent direct sum o f states
belonging to different Lorentz frames whose velocity distribution is given by the
probability density ]r r 2/11 ~2 over G/H, and the Lorentz covariance (4.27) arises only
in the situation that the choice o f a reference frame is not sharply fixed. On the other
hand, the usual temperature state considered in a fixed Lorentz frame does not belong
to sS, but should be understood as a 'limit' of some sequence in 5 tending to
v ~, ~u/~(u - v) with ~u ~ ~o~. Similarity and dissimilarity of the above symmetry breaking
to that of the supersymmetry will be discussed elsewhere.

Acknowledgements
The author would like to thank Professor M. Flato for useful discussions and
encouragement at Universit6 de Dijon. H e is very grateful to Professor F. Mancini at
Salerno and Professor W. Z i m m e r m a n n at Munich for useful discussions and hospi-
tality. Discussions with Dr H. K a n n o and Dr T. Matsui are greatly acknowledged.
References
1. Einstein, A., Jahrbuch Radioakt. Elektronik 4, 411 (1907);
Tolman, R. C., Relativity, Thermodynamics and Cosmology, Oxford Univ. Press, London, 1934.
2. Ott, H., Z. Physik 175, 70 (1963);
Landsberg, P. T., in G. K. T. Conn and G. N. Fowler (eds.), Essays in Physics, vol. 2, pp. 93-129;
P. T. Landsberg, Phys. Rev. Lett. 45, 149 (1980).
3. Ojima, I., in K. Kikkawa, N. Nakanishi, and Nariai (eds.), Gauge Theory and Gravitation, Proc. of the
Intern. Symp. on Gauge Theory and Gravitation held in Nara, Japan, August 1982, Lecture Notes in
Physics, Vol. 176, Springer-Verlag, Berlin, Heidelberg, New York, 1983.
4. Ojlma, I, Seminar talks at Universit~ di Salerno, Universit6 de Dijon and Max-Planck-Institut f'tir
Physik und Astrophysik in 1983.
5. Girardello, L., Grisaru, M, T., and Salomonson, P., Nucl. Phys. B178, 313 (1981);
Teshima, K., Phys. Lett. 123B, 226 (1983);
Aoyama, H. and Boyanovsky, D., Phys. Rev. D30, 1556 (1984).
6. Matsumoto, H., Nakahara, H., Nakano, Y., and Umezawa, H., Phys, Rev. D29, 2838 (1984).
7. Kubo, K., J. Phys. Soc. Japan 12, 570 (1957);
Martin, P. C. and Schwinger, J., Phys. Rev. 115, 1342 (1959).
8, Bratteli, O. and Robinson, D. W., OperatorAlgebras and Quantum StatisticalMechanics, Vols. 1 and 2,
Springer-Verlag, Berlin, Heidelberg, New York, 1979 and 1981.
9. Takahashi, Y. and Umezawa, H., Collective Phenomena 2, 55 (1979);
Ojima, I., Ann. Phys. 137, 1 (1981);
Matsumoto, H., Ojima, I., and Umezawa, H., Ann. Phys. 152, 348 (1984).
10. Nelson, P. and Alvarez-Gaum6, L., Commun. Math. Phys. 99, 103 (1985).

You might also like