An Elementary Proof of Dyson's Power Counting Theorem HAHN ZIMMERMANN
An Elementary Proof of Dyson's Power Counting Theorem HAHN ZIMMERMANN
An Elementary Proof of Dyson's Power Counting Theorem HAHN ZIMMERMANN
Abstract. For the case of Euclidean metric an elementary proof of the power
counting theorem is given.
1. Introduction
As is well known, Dyson's power counting theorem plays an important
part in the theory of renormalization [1]. For the case of Euclidean metric
a rigorous proof of this theorem was obtained by WEINBERG as a by-
product of his work on the high-energy behavior of Feynman integrals
[2]. The purpose of the present paper is to give a short, direct proof of
the power counting theorem which uses more elementary methods. We
restrict ourselves to the Euclidean case. An extension to the case of
Minkowski metric will be discussed in a forthcoming paper.
We will be concerned with integrals of the form
( L 3 )
l=(lv . . ., Q , K=(KV . . ., Kn) , JΓ,Φ0 .
G denotes a n w x m matrix.
For integrals of the form (1.1) the following version of the power
counting theorem will be proved in Section 3.
* The research reported in this paper was supported in part by the National
Science Foundation.
Power Counting Theorem. I 331
Theorem 1. Let all masses μό =4= 0. The integral (1.1) is absolutely con-
vergent (a.c.) if
(i) the dimension
d= q-2n + &m (1.4)
are a.c. The integral (1.5) extends over the hyperplane Hό defined by
Kf ^ 1 . (1.11)
332 Y. HAHN and W. ZIMMERMANN :
such that
KξφKf for i φ / .
i = Σh
7 =1
(1.13)
K* ^ Kf ^ 1 . (1.14)
t = Ak (1.15)
satisfying
L
^ l. (1.16)
I, = dtj —i —- (1.17)
K^KM-^-Σdwk' (1.18)
U = c lίj % i = 2, . . ., m (1.21)
with
(L24)
*i=w-
The region F ' consists of all t%,. . ., t'm satisfying
K^A-H') ^ 1 . (1.25)
2. Auxiliary Lemmas
It will be convenient to consider more general integrals of the form
(2.1)
= If + μf , μj^ 0.
The following Lemma states that the values of the external momenta q
are irrelevant for the convergence of the integral /.
Lemma 1. Let all masses μι Φ 0. Necessary and sufficient for the ab-
solute convergence (a.c.) of I(q, p, μ, V) is the a.c. of 1(0, p, μ, V).
Proof. The statement follows from the inequalities
(2.4)
where
dx = dxx . . . dxm .
h in) 21 2
"IL..Φ)- Σ °ii...iMh...i»Qlp>* - ) ( )
h...3n
where the Ci1...injL...jn depend on the parameters (2.9) only and are
determined by a generalization of Lagrange's interpolation formula to
several variables. Hence
/ Q(x, z) dx
v
is either a.c. everywhere or a t d different points a t most (d denotes t h e
degree of Q in t h e variable z). Hence (b) is correct for n—1. Assuming
t h a t (b) is correct for n — 1 we will prove it for n. The divergence of
(2.8) a t (2.11) implies t h a t
f Q(x,z9 . . ,,zn^an)dx
v
is absolutely divergent (a.d.) everywhere except for
where Ό is a set of measure zero. Let now (z±. . . zn_^) $D(an) then
JQ(x9z1. . .zn-τan)dx
v
is a.d., hence
/ Q(xiz1. . .zn^.1zn)dx
V
Zn = ίf (% Zn-l) , « = 0, . . ., <X ^ ί n .
Hence (2.8) is a.d. except for either (% . . . 2W_1) ( D ( α J or zn — f^
(zt. . . 2w_i). This, however, is a set of measure zero.
The purpose of the following Lemmas 4 and 5 is to show that the
mass terms in the denominators are not relevant for the convergence of
the integrals (2.4) provided the zeros of the denominators are properly
excluded from the domain of integration.
Lemma 4.
I(S,P,μ,DQ)£l(9,P,0,I><), (2.13)
I(q, p, 0, Dq) ^ CI(q, p, μ9 Dq) . (2.14)
23*
336 Y. HAHN and W. ZIMMERMANN :
£ 1, (2.16)
2
ς 1 + -4"for e^kqjO) ^ r . (2.17)
is a.c.
(b) The integral (2.18) is a.c. if and only if
Proof, (a) I t is obvious that the a.c. of I{qpμ) implies the a.c. of
I(qpμDq). We will show that the a.c. oίl(qpμ) follows from the a.c.
of I(qpμDq) for m integration variables kx . . . km. As hypothesis of
induction we assume that the statement has been proved for all integrals
of the type (2.18) with less than m variables of integration.
We first observe that the number of linearly independent forms
among Kx. . . Kn must be m. (If it were less than m the integral (2.19)
would diverge.) The factors eό of the denominator are now renumbered
such that
II • ., J? (2-22)
represent all different quadratic forms among the If. Let s be any subset
of (1, . . ., c) including (1, . . ., c) itself and the empty set. To every s we
define Xs as the set of all k satisfying
If g r2 for j ζs
1
(2.23)
If ^ r2 for j $ s .
Power Counting Theorem. I 337
= J ^ _ L ^ (2>S0)
where
EM(lcqμ)= Π ^
'~\ (2-31)
EW{kqμ)= Π e,{kqμ,).
j =α+ 1
For the integral (2.30) we will now derive an upper bound which is
independent of K. To this end we make use of the hypothesis that
HΆVμD*) =
f (k(K), q,μ) J
- P{h(KK), p)
W»(k(KK), q, μ)
(2.32)
Z
338 Y. HAHN and W. ZIMMERMANN :
is a.c. In the second line Z'lq denotes the region of all K satisfying
lf^r\ 7 = 1 , . .., α .
is a.c. for K ζZ'lq up to a set Σ of measure zero [4]. Applying the hypo-
thesis of induction to (2.33) we find that
fdK-IMS^~ (2.34)
E™(k(KK),q,μ)
J E™(k(KK),q,μ)
is a.c. in Z'lq up to the set Σ. By Lemma 1 also
fdK
E™(k(0K),q,μ)
J
is a.c. in Z'lq up to the set Σ. According to Lemma 3b (2.35) must be
a.c. for any K. Applying Lemma 1 again we obtain a.c. of (2.34) for any
K. Furthermore (2.34) is majorized (Lemma 2) in Zlq by a multiple of
(2.35). Expanding P with respect to the components of Kl9 . . ., Ka we
find (Lemma 3a) that (2.35) is bounded in Zlq. This proves that each
term I(qpμXs) in (2.24) is a.c. and hence also I(qpμ).
(b) Statement (b) follows by combining (a) with Lemma 1 and
Lemma 4.
Lemma 6. Let
(2.36)
implies the a.c. of
r P thvλ
(2.37)
D
Power Counting Theorem. I 339
We have
/ E(kθμ)
D'
/ E(k00)
D
is a.c. for all values of ρ. Applying Lemma 3a we obtain the a.c. of
(3.1)
(3.2)
The problem is to prove the a.c. of (3.1) provided the dimension of (3.1)
is negative and every subintegral (3.2) is a.c. Introducing new variables
of integration t = (tx . . . tm) by (1.15—16) we can write (3.2) in the more
convenient form
=fdta... dtm /^X^ (3.3)
according to Lemma 5b it suffices to prove the a.c. of the integral
340 Y. HAHK and W. ZIMMERMANN :
i(q)= Σ
i (3.5)
= ffdk E(kOO)
= P(kq)
= Σ ^«A...ft(?)
}«βi.-.βt
In order to prove the a.c. of the integral (3.10) we now derive an upper
bound for the inner integral
T
T(t>π\^ fat' Jt' *βi...ββ''g)
0)
J(t q ) -vJ at2.. .dtm E A
, -if 0 0)
< r"
<C fit'...dt at' \s"βi'"β*^'
f dt - -^^1
xj 2 m E{A-H,f0>μ)
The last integral (3.15) still depends on the four vector t[ which, however,
is bounded. Applying Lemma 2 with respect to t[ we obtain
i
\T(ί'n\\<Γ frit' rlt' l^«ffi •-04 (*2 • C g)l /q Ί β\
K (?i ίJI = °2 J ah - -atm W^E~777ζ~Lή (o.io;
with (72 independent of tx. ZΓ; denotes the value of E at t[ = 0.
1
1', 0, ^) for ί( = 0 . (3.17)
converges. To this end we start from hypothesis (ϋ) which states that
the integral (3.3) is a.c. Using Lemma 1 we obtain a.c. for
(319)
/^Γ.IV
By Lemma 6
μ ^ X μ ) (3.20)
is a.c. Finally Lemma 3 a implies the a.c. of
We have thus proved that the right hand side of (3.18) converges. Inserting
(3.12), (3.14) and (3.18) into (3.16) we obtain
\Rjκβl...βi(q)\^O2AB(qμ) (3.22)
with
A= \c\af\t\adt.
u
Since σ ^ — 5 this integral converges. This completes the proof of the
theorem.
For the proof of Theorem 2 we remark that the statement holds
already if H is restricted to special hyperplanes of the form
Beferences
1. DYSON, Γ. J . : Phys. Rev. 75, 1736 (1949).
2. WEINBERG, S.: Phys. Kev. 118, 838 (1960).
3. See for instance; ISAACSON, E., and H. KELLER: Analysis of numerical methods.
New York: Wiley 1966.
4. See CARATHEODORY, C.: Vorlesungen ύber reelle Funktionen, second edition.
Chapter XI. Leipzig: Teubner 1927.
W. ZlMMERMANN
Courant Institute of Math. Sciences
251 Mercer Street
New York, N. Y. 10012, USA