Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

A Theorem Concerning The Positive Metric: Derek W. Robi Nson

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Commnn. math. Phys.

1, 89--94 (1965)

A Theorem Concerning the Positive Metric


By
DEREK W. ROBI~NSON
Max-Planck-Institut fiir Physik nnd Astrophysik, Miinchen

Abstract. I~ is proved that if the n-point correlation functions of a system


vanish for all n > N then they vanish for all n > 2. The theorem is valid for a
wide variety of formalisms and an explicit proof is given for a Bose system with
the canonical commutation relations; a proof is sketched out for a relativistic field
theory of the Wightman type. The essential property used in the proof is the pos-
itive definite metric.

1. Introduction
During the last few years a great deal of effort has been applied to
the analysis of the mathematical structure of relativistic field theory
(see for instance [1], [2]) and a number of general results have been
obtained, some of which arc valid for other physical theories. One prop-
erty which is assumed in m a n y theories, but which has been scarcely
analysed, is the property of the positive definiteness of the metric in
Hflbert space. I t is the purpose of the present note to prove one simple
theorem which is essentially a direct consequence of this property
alone. This theorem is concerned with the hierarchy of functions describ-
ing the correlations between various points or events of the system, the
so-called truncated functions or correlation functions. Physically one
would expect t h a t the correlation between n events would be small
for n large and consequently that it might be reasonable to construct
models, described by a finite number of functions, by setting all the
truncated functions for n > N equal to zero, where N is arbitra, rily large
but finite. The present theorem proves t h a t such models are inconsistent
with the positive metric unless the remaining truncated functions n > 2
also vanish identically. Thus the class of models obtained by this physical
Ansatz contains only the well known "generalized h'ee field", which has
of course been extensively applied to the many body problem and widely
analysed in relativistic field theory.
I n See. 2 of this paper we introduce several general definitions and
results and show that for a system with no internal structure the above
mentioned theorem is a direct consequence of a result due to 1V[AR-
CI~KIEWICZ.I n See. 3 this result is extended to the case of a Bose system
90 I)EI~EKW. ROBINSON:

with the canonical commutation relations and a proof for a relativistic


field theory of the Wightman type is indicated. See. 4 comprises a short
summary and discussion.

2. General formulation
A large variety of physical theories may be mathematically summarized
as consisting of an algebra 92 generated by certain elements Q~ (fields,
observables, etc.). Equations of motion, invariance properties and other
physical conditions define the internal structure of the algebra, which for
our present purposes is irrelevant. Ultimately one is interested in a
cyclic representation of the algebra in a tIitbert space and it is well
known (see for instance [3]) that corresponding to every positive linear
functional W (Q) over ~[ there exists such a representation. Thus to exam-
ine the cyclic representations of 02 we need only to examine the func~ion-
als W(Q) defined for all Q E OAand having the two properties
a) W()~Q1 + ~2Q~) = ~1W(QI) + ~W(Q2) (1)
where ~1, ~2 are constants and Q1, Q~ C 0A and
b) W(Q+Q) >=O. (2)
This result can be easily understood if we use the physical notation intro-
duced by the definition
W(Q) = (0[Q]0)
where [0) is understood to be the cyclic state in Hilbert space. (In most
applications 10) is the ground state or the vacuum.) It is perhaps im-
portant to realize that (1) and (2) are the only properties which W(Q)
must have in order to be interpretable as the expectation value of Q
in some state.
The truncated functions
WT(Q) = (OIQ]O)T
are defined from the ground state expectation values of products of the
Qi recursively through the formula
<0l Q~Q~... Qi,~[0> = ~ , <0[ Q ~ . . . t0>T... < 0 l . . . Q,, [0>T (3)
where the sum is over all possible elnstering of the Qik and the order
within each cluster is carried over from the left hand side.
We now consider a system with no internal structure described by
an Abelian algebra generated by a single hermitian element Q. In this
case a generating function exists for the ground state expectation values,

E ( ~ ) = <01 ~ Q ]0> = ~W~


~'~0
A Theorem Concerning the Positive ~etrie 91

and (3) simplifies to the identification

~ 1 Y w~ . (4)

The function E (x) has three relevant properties; it is bounded,

E* (x) = E ( - x) ,
and
f dxdy/*(x) E(x- y)/(y) >= O,
for all ] (x) from some suitable class. These properties are a consequence
of (2) and the hermiticity of Q and t h e y serve to define, in the terminology
of probability theory, a characteristic function. We now use a theorem*
due to MA~CI~KIEWmZ [4].
Theorem 1. I f exp {P(x)), where P(x) is a polynomial, is a character-
istic function then it follows t h a t P(x) = l a x - bx 2 where a and b are
real a n d b > 0.
A proof of this theorem is reproduced b y RXCRT~R [5]. Thus, from (4),
we see t h a t if W T = 0 for all n > £V t h e n E (x) is a characteristic function
of the t y p e considered b y the theorem and we m u s t t h e n conclude t h a t
Wn~ = 0 for all n > 2. I n the next section we extend this result to more
complicated systems.

3. Applications
a) Bose system
Consider a Bose system defined in terms of a field ~ (x) and its Her-
mitian conjugate ~0+ (x) which satisfy the canonical c o m m u t a t i o n rela-
tions, i.e.
[~(x), ?+(y)] = 8(x - y) etc., (5)
and from which we m a y construct the well defined operator ~ (/) given b y

q~(/) = f dax ~c(x) /(x) where f(x) E ~ •


Let us now define
Qi=q~(/i) for i=1,2 ..... n
and (6)
Qi=q~+(]i) for i=n+l ..... n+m.
F r o m (3) and (6) we define the t r u n c a t e d functions

WTnm(Pi; /i) = <0] Qi . . . . . Qin+m ]0)T


• In the course of the presen~ work ~his theorem was eonjecgured and proved
by the present author. However, before publication, ig was pointed out by Dr.
Sc~r.IrD~R thag *he theorem had been already given by ~ c ~ w x e z [4]. The
author is indebfed to Dr. Se~IED~R for this information.
92 D~R~K W. RoBiNsox:

where Pt denotes the permutation { i l , - . - , in+M} of the set of indices


{1 . . . . . n -Pm}. We need next
Lemma. The truncated functions W ~ ( P i ; ]i) are independent of the
permutation Pt except for the case n = m = 1 where
(0[ ~(]1) ~+(/~) 10)T = (0l ~+(/~) ~(h) 10)T + (/1, f~)"
This L e m m a is a direct consequence of (5) and m a y be easily proved b y
induction, using (3). As a result of the L e m m a we introduce the simplified
notation
W ~ ( P i ; fi) = W ~T, / ~r ) .

We are now in a position to prove


Theorem l a . I f the truncated functions WTnm(fi) have the property
t h a t they vanish identically for all n, m such t h a t n + m > N > 2
and every choice of the test functions fi (x) then it follows from positive
definiteness and the commutation relations t h a t they have this property
for all n, m such t h a t n + m > 2.
Proof. Define the operator Q b y
n+m
Q= ~ (e-i~'Q~ + e ~ Q +) (7)
~'=1

where the ~ are real numbers. B y assumption


W ~ = <0] QM [0>T = 0 for all M > N (8)
and as a result of theorem 1 positive definiteness implies t h a t (8) m u s t
also be valid for M > 2 and all ~ . Therefore
2~ 2 7t i ~ ¢¢r--i ~ ~¢r
"°°+*
2=,/
o o (9)
for n+ m > 2.
Using (7) we calculate explicitly from (9) t h a t
~WTnm(Pi;/i)=O for n+m>2.
Pi

The L e m m a then allows us to conclude t h a t


WnTm(fi)=0 for n-f-m>2.
As the choice of test functions and the value of n in (6) are arbitrary
the proof of the theorem is complete.

b) Relativistic field theory


Consider a relativistic field theory of the Wightman type defined in
terms of a field A (x) or A (]) where
A(/)= fddxA(x)/(x) and f(x) C ~ •
A Theorem Concerning the Positive NIetrie 93

The properties which are required for A (x) are


1. Locality; [A(x), A(y)] = 0, for (x - y)~ < 0 ,
2. Lorentz invariance,
3. Positive energy
from which properties we m a y prove*
Theorem 1 b. I f the truncated functions
W~(A) = < 0 1 A ( / 1 ) . . . A(]~)]0>T
vanish for all n > N > 2 then they vanish for all n > 2.
Proo/. We give an indication of the proof which starts from consider-
ation of the operator
N
Q = Z a~ A(]~)
~'=1
where ~¢ are real parameters and the test functions/i (x) are chosen so
t h a t the support of each function is space-like to the support of all the
remaining test functions. Now b y an argument similar to t h a t used in
the proof of theorem l a , and using the property of locality we find imme-
dia%ely t h a t
(0[ A ( / 1 ) , . . . , A(/~) [0)T = 0 for all n > 2
and with the above mentioned restrictions upon the supports of the test
functions. I n order to remove these restrictions we need the well known
analyticity properties of the Wightman functions which follow from the
properties of Lorentz invariance and positive energy [1], [2]. We do not
elaborate further upon this latter stage of the proof.

4. Summary and conclusions


The result of the present work is of a rather negative nature. I t does
demonstrate however the type of difficulties t h a t might be encountered
in attempting to find approximate solutions of physical problems. I f
one approaches the problem from a Green's function point of view it
might be tempting to make approximations reducing the infinite set of
coupled differentiM equations to a finite set but our result shows t h a t
great care must be taken, in such an approach, to avoid violating positive
definiteness. The solution of an "interacting" problem is infinitely more
complicated t h a n the solution provided b y a free field or "generalized"
free field. I n the former an infinite number of functions are involved and
although one would suspect t h a t the majority of these are small one m a y
not ignore t h e m completely.
Although the present work does not apply directly to a Fermi
system, due to the complications introduced b y anti-commutation, it
* This theorem is probably well known to most workers in the field but the
only published result [6] uses much stronger restrictions upon A(x).
94 DEREKW. ROBII~SON: A Theorem Concerning the Positive Metric

should be possible t o c o n s t r u c t a proof of a similar r e s u l t for such a


system.
The author would like to thank Prof. ttAAG, Dr. SCH~DE~ and Messrs.
RI~KE and Roos for helpful discussions and the Aspen Institute for Humanistic
Studies for their hospitality during summer 1964.

References
[1] WIGHTB~LAN,A., and R. F. STREATER: P. C. T. Spirt and Statistics and All That.
New York: Benjamin 1964.
[2] JosT, R. : General Theory of QuantizedFields.Am.Math.Soe.Public~tions(1965).
[3] I~EU~RK, M. A.: Normierte Algebren. Berlin: VEB Deutseher Verlag der
Wissenschaften 1959.
[4] M~RCU~rKI~WmZ,J. : Math. Z. 44, 612--618 (1939).
[5] RICHTER, H. : Wahrscheinlichkeitstheorie. p. 300. Berlin-GSttingen-Heidelherg:
Springer-Verlag 1956.
[6] GRE~NB~.R~,O. W., and A. L. LICHT: J. Math. Phys. 4, 613 (1963).

(Received January 1, 1965)

You might also like