A L'UNIVERSIT PARIS-SUD

CHVTRE D'ORSAY
pour obtenir v
LE GRADE DE DOCTEUR ES-SCIENCES PHYSIQUES
par
Marietta MANOLESSOU-GRAMMATICOU
Cont r i but i on a l ' Et ude de Quelques Pr opr i t s Fondamentales
des Ampl i tudes de Di f f usi on e t de Fonct i ons
,^Je n-po1nts Associ es -
Soutenue le 12 Dcembre 1977 dev ant la Commission d'Examen
Prsident ' . -F. Llirat-
Rapporteur R. Stpra
f J. B r
< J.M. Bo
Bros ,-_,
Examinateurs \ J.M. Sony
L a s c o u x
. M e m b r e I n v i t '< D . I a g o l n l t z e r
H* d'ordre : 1925
THSE
prsente
A L'UNIVERSIT PARIS-SUD
CENTRE D'ORSAY
pour obtenir
GRADE DE DOCTEUR ES-SCIENCES PHYSIQUES
par
Marietta MANOLESSOU-GRAMMATICOU
nt r i but i on l ' Et ude de Quelques Propri t s Fondamentales
des Ampl i tudes de Di f f usi on et de Foncti ons
de n-poi nt s Associes
Soutenue le 12 Dcembre 1977
devant la Commission d'Examen
Prsident F- Lurat
Rapporteur R. Stora
| J. Bros
Examinateurs s J.H. Bony
|^ J. Lascoux
Membre invit D. Iagolnitzer
Truth is not yours or mine. What is yours aan be organized,enshrined,
exploited. That is what is happening in the world. But truth cannot be
organized. Like beauty and love truth is not in the realm of possessions.
J. Krishnamurti "The only revolution"
A mes parents
A mon meilleur ami, Basile
REMERCIEMENTS
Je remercie vivement C. De Dominicis et R. Balian po ur m 'a v o i r accueillie
a u Service de Physique Th o r i que du CEN Saclay, en tant que collaboratrice tem-
poraire de thse et au sein du groupe de Physique M a t h m a t i que .
Je tiens remercier particulirement J. La sco ux qui m'a offert la po ssi -
bilit de prolonger ma carrire dans la recherche en m
1
accueillant au l a b o r a -
toire de Physique Thorique de l'Ecole Po l yt e chn i que et dont l'intrt po ur
m e s travaux a t trs stimulant
Que M . le Professeur F. Lurat. qui a b i e n v o ul u accepter de prsider le
Jury de cette thse et M . le Professeur J.M. Bony, y participer trouvent ici
l'expression de mes r e m e r ci e m e n t s.
Je suis profondment reconnaissante pour l'intrt que R. Stora a montr
au prsent travail, et pour avoir accept de participer au jury de cette thse.
La plus grande et importante partie de cette thse est le fruit de m o n
travail en tant qu'lve et collaboratrice de Jacques Br o s. C'est difficile
de trouver des mots pour exprimer ici toute m a reconnaissance pour la co n -
fiance qu'il a montr quand il a accept de m 'a ppr e n dr e travailler ; pour la
gnrosit avec laquelle il m 'a touj'ours fourni les fruits de sa pense et sur -
tout pour m'avoir appris (et il m'apprend e n co r e ) voir un problme avec
honntet et sans pe ur .
Un des sujets de ce travail tant une application d'un rsultat m a t h m a t i que
de Bros-Iagolnitzer , j'ai eu l'occasion de profiter d'excellents conseils de
Daniel lagolnitzcr ; je n'oublierai jamais que m e s durs dbuts en physique m a t h m a -
tique ont t encourags par sa prsence amicale et les stimulantes discussions
avec l ui . Je lui exprime ici ma profonde r e co n n a i ssa n ce .
Je remercie vivement D. Atkinson et G. M a ho ux de m 'a v o i r propos un sujet
de travail (qui fait partie de cette thse) et de m 'a v o i r conseill tout le
long de son laboration.
A des moments trs difficiles psychologiquement et moralement, j'ai
pu apprcier la prsence amicale de Gilles Cohen-Tannoudji, An dr M o r e l ,
Andr Vo r o s, Madeleine Forneuf qui je dois rendre hommage.
Avec Michel Bergre et Henri Ca m i l l e , j'ai eu d'intressantes di scus-
si o n s. Je les en remercie v i v e m e n t .
Je do i s exprimer m e s remerciements les plus profonds a toutes celles
qui ont contribu la matrialisation de ce travail : Chantai Le g a l l ,
Eliane Cotteverte et surtout Jeannine Rigou et Danielle Bunel qui ont t
tortures par mes formules.
Je remercie tous ceux, appartenant ou pas au domaine de la recherche
scientifique, qui m'ont encourage tout au long de ma carrire.
Je remercie Ficus pour son existence.
Je remercie Basile pour son amiti profonde, unique, pour son e n t ho u-
siasme et sa confiance en tout ce que je pense et je fa i s.
T A B L E DE S M A T I E R E S
P R E M I E R SUJET
PRODUIT PE G-COHVOLUTTOH ASSOCI E AUX FONCTIONS A (t-POINTS
EN THEORI E AXIOMATIQUE SES CHAMPS
I . LE CAS EUCLIDIEN
INTRODUCTION
Section 1 Quelques classes de fonctions croissance lente.
Section 2. La gnralisation de 1'integrand renonnalis de Z inane rmann.
Section 3. La formule de "forts co m pl t e s" de R par rapport un ensemble
de sous-espaces emboif.s de E,.. .
Section 4. Intgrabilit de 1'integrand renormalis R .
Section 5. Indpendance de H,; ?
n
"^ par rapport au choix de l'ensemble a dm i ssi -
Q
b l e .
Appendice A
Appendice B
R f r e n ce s.
I I . COMPLEMENTS DE LA PARTIE I : (Prolongeaient analytique de R
G
) .
INTRODUCTION
Section 2.
2.1. Ensembles de lignes internes indpendantes pour C et tous ses so us-
graphes .
2.2. Cho i x admissible.
2.3. Eclatement en a r b r e s.
Section 3. Quelques oprations sur les fonctions gnrales n -po i n t s.
Section A. Proprits de l'intgrand gnralis renormalis de Zimmermann Kg .
Rfrences.
DE UXI M E SUJE T
CONDITIONS VE MACROCAUSALTE OBTENUES A PARTIR DES PROPRIETES D' A M U T I C T T E
EU THEORIE AXIOI.KriQ.UE PES CHAMPS
1. Introduction.
2. Causalit - Analyticit
3. A nalyticit par rapport au transfert des moments et conditions exprimant
la courte porte des forces.
4. Analyticit par rapport l'nergie totale
4.1. Conditions de causalit
4.2. Conditions de temps de relaxation
Appendice
Rfrences
TROISIME SUJET
CONSTRUCTION D'UNE AMPLITUDE UNITAIRE ET ANALYTIQUE
VANS LA DIFFUSION DE PARTICULES DE S P I N S j
1. Introduction
2. Cinmatique et analyticit
3. Thormes des applications contractantes
4. Ambiguts discrtes
5. A mbigut inlastique et mthodes de Heuton Kancorovich
6. Conclusion
1
P R E F A C E
Le prsent travail est constitu d'un ensemble de trois sujets di st i n ct s.
Le premier sujet contient deux parties : la premiere (I) prsente sous forme
d'article est le rsultat d'une collaboration avec Ja cque s Bros et concerne
la dmonstration de la convergence d'un Produit de G-convolution renormalis
(au sens de Bros-Lassalle) dans la rgion Eucl i di e n n e .
La deuxime partie (II) appele "complments de la partie I" concerne la
dmonstration de proprits analytiques et algbriques de 1'integrand r e n o r -
malis introduit dans I et co n st i t ue la premire tape d'une tude poursuivie
par l'auteur, visant l'extension dans l'espace des impulsions complexes du
produit de G-convolution rehormalis.
Le deuxime sujet a pour but d'tablir (par application du thorme de la
Transforme de Fourier gnralise de Bros-Iagolnitzer) une quivalence
prcise entre les proprits d'analyticit des noyaux de di ffusi o n obtenues
dans le cadre de la thorie axiomatique des champs et certaines proprits
de dcroissance exponentielle des amplitudes de diffusion dans l'espace-temps
(conditions de macrocausalit au sens de I a g o l n i t z e r -St a pp). Cette tude a fait
l'objet d'une publication dans Comm. of Ha t h. Physics (1976).
Le troisime sujet est une tude sur la possibilit de construire une amplitude
unitaire et analytique partir des donnes exprimentales dans le cas de la
diffusion N-ft. Ce travail constitue un e extension des investigations prcdemment
entreprises et dveloppes par Atkinson-Mahoux-Yndurain, et il a fait l'objet
d'une publication dans Nuclear Physics B (1975).
P R E M I E R S U J E T
5
FREMIER SUJET
PRODUIT DE G-CONVOLUTION RENORMALISE ASSOCIE AUX FONCTIONS A n-POINTS
EN THEORIE AXIOMATIQUE DES CHAMPS
.I..LE CAS EUCLIDIEN
Il est couramment admis par les thoriciens des particules l m e n -
taires que les divers mcanismes de diffusion peuvent tre dcrits dans
le langage des amplitudes de Feynman QiwiluJUiiiA. Ces quantits sont
associes des graphes qui gnralisent l'image traditionnelle des graphes
de la thorie des perturbations pa r le procd suivant. Chaque v e r t e x
ponctuel d'un graphe de Fe yn m a n est remplac pa r un e "b o ul e " qui a un e
certaine "structure" et correspond une interaction fo r t e . (Une telle
description intervient par exemple en thorie des interactions profondment
i n l a st i que s). Suivant le processus considr dans chaque ca s, ces "b o ul e s"
sont supposes reprsenter soit des amplitudes compltes de diffusion (ou
de pr o duct i o n ), soit des resommations partielles d'amplitudes de Fe yn m a n .
D'autre pa r t , en ce qui concerne les variables de quadri-impulsions
utilises, ces amplitudes sont parfois dcrites dans l'espace de Hinkouski et
alors la prescription d'intgration dite du "i e " est employe par analogie
avec la thorie des perturbations ; quelquefois cependant on se sert par
commodit de la restriction de ces amplitudes l'espace Euclidien, et alors
le problme du prolongement analytique sur la couche de m a sse des quantits
qui en rsultent reste en gnral ouvert. Cette situation assez confuse a
incit J. Bros et M . Lassalle introduire et explorer systmatiquement
dans l'espace des quadri-impulsions la notion de produit de G-convolution
associ un graphe donn G. La proprit la plus intressante dmontre
par ces auteurs est la suivante : considrons un graphe connexe G ayant m
boucles indpendantes, n lignes externes et n lignes e xt e r n e s ou internes)
incidentes chaque vertex v . Au graphe G est associ un produit dit de
"G-convolution" par la construction suivante :
a ) A chaque v e r t e x v de G est a sso ci e un e fonction g n r a l e n
n
v
v 4(n -1)
v
points H (K ) dans l'espace complexe t
v
des qua dr i -i m pul si o n s
K po r t e s par les lignes incidentes v .
b ) A chaque ligne i de G est a sso ci e un e fo n ct i o n g n r a l e 2 po i n t s
H (.) dans l'espace C, . de la qua dr i -i m pul si o n &. co r r e spo n da n t e .
c) Le s fo n ct i o n s H
V
( K
V
) et H* (.) satisfont ce r t a i n e s pr o pr i t s
d'i n t g r a b i l i t l'infini dans l'espace e ucl i di e n . Al o r s, les a ut e ur s
pr ci t s ont pu g n r a l i se r la n o t i o n d'amplitude de l'intgrale de
Fe yn m a n , en dfinissant un e fo n ct i o n g n r a l e n-points de l'ensemble
de s quadri-impulsions e xt e r n e s complexes de G pa r la fo r m ul e sui v a n t e de
"G-co n v o l ut i o n "
H C K ) - / H H
v
( K
v
( K , k ) ) n H
( 2 )
(i ,.(K,k))d, k (1)
w
r(K) v i
x

Da n s ce t t e fo r m ul e k est l'ensemble de 4 m v a r i a b l e s i n d pe n da n t e s
d'intgration et K
v
(K,k) &.(K,k) sont l e s'fo n ct i o n s l i n a i r e s d t e r m i n e s
pa r les lois de co n se r v a t i o n d' n e r g i e i m pul si o n _tous les v e r t e x de
G:F(K) est un contour 4 di m e n si o n s dans l'espace complexe C... d fi n i
dans le do m a i n e d'a n a l yt i ci t de 1'integrand et dpendant co n t i n ue m e n t
des qua dr i -i m pul si o n s e xt e r n e s K. Dans les r sul t a t s o b t e n us pa r Bros et
La ssa l l e les pr o pr i t s sui v a n t e s ont t explicites :
i ) Si on suppose que K v a r i e da n s la r g i o n Eucl i di e n n e :
E
A(n -l )
{ K
.
(
*
> i Q O ) a y e c
+
R
3(n -1)
Q
o
R
( n - 1 )
} e (
.
q u e r ( K ) e s (
.
a u s s i
identifi la r g i o n EuclidienneE>lk - (?,iq ) i J E K ", i " s!*} a l o r s
(!) a un se n s. Ce rsultat est un e co n s que n ce triviale du fait que le
n
v
do m a i n e axiomatique pr i m i t i f de chaque fo n ct i o n n -po i n t s H (resp.
( 2) v
H (*)) contient, la r g i o n Eucl i di e n n e de l'espace a sso ci K (resp. l^) ;
E"
n
'(K) est donc b i e n d fi n i e et a n a l yt i que dans E
G
* Un e fonction H (K), K'e 'C^. est di t e fonction g n r a l e ' n-points ' si
e l l e est analytique dans un do m a i n e pr i m i t i f de la thorie a xi o m a t i que et si
ses v a l e ur s au bord Vp( ?) pour K tendant vers las rels sont de s di st r i b ut i o n s
et satisfont aux r e l a t i o n s de St e i n m a n n (Four une d fi n i t i o n pl us d t a i l l e
v o i r I I ) .
7
ii) En utilisant des deformations de contours appropries du contour
R K ) qui gnralisent les rotations de Hi ck , o n montre que lu (K)
G
admet un prolongement analytique dans tout le domaine pr i m i t i f. Les
diverses valeurs au bord de Hg dans l'espace des rels (espace de
M i n ko wski ) sont ainsi dfinies et satisfont les relations de Steinmann,
iii) La x-valeur au bord xi de H^ (qui concide avec la transforme
de Fourier de la valeur moyenne dans le vide du produit chronologique de
n-opSrateurs de champs dans le cas o H? est associe un champ qua n t i que )
est dfinie par la prescription de Ruelle-Araki qui gnralise la "prescrip-
tion - i s" des amplitudes de Feynman. La formule (1) fournit alors un e
expression intgrale de r i dans l'espace de Minkowski ; dans cette
expression les fonctions H ^ et H de l'intgrand sont remplaces par
les t-valeurs au bord correspondantes T
v
et! ',
iv) Poui que i ) , ii) et la premire partie d iii) soient v a l a b l e s, il
suffit que les conditions de convergence de (1) I l'infini soient satisfaites
dans toute rgion quasi-Euclidienne ; E(WJ W + E o H e st un e r g i o n
compacte arbitraire de C .L a seconda partie de III ncessite certainement
des conditions .de Co n v e r g e n ce dans l'espace de M i n ko wski . Du point de v ue
physique, la ncessit d'imposer des conditions de convergence conae ci-dessus
dans i ) , n'est pas tout fait satisfaisante ; en'effet les fonctions de
Gr e s H:,J intervenant dans une thorie des champs interaction polynoaiale
(par rapport a ux champs) da n s l'espace quadri-dimensiotnel peuvent avoir
des proprits de croissance l'infini qui produisent des divergences ul t r a -
violettes dans les intgrales du type (1) comme c'est le cas en thorie des
pe r t ur b a t i o n s.
En fait, le produit de G-convolution dfini par (I) s'est rvl trs
satisfaisant pour l'tude d'analytic!t dans le cadre du programme n o n
linaire effectue dans les travaux de Bros et Lassalle "Many-Particle
5 )
Structure An a l ysi s" : e n utilisant des rgularisateurs analytiques du
type Pauli-Villars,, ces auteurs ont pu bien dfinir sous un e forme r g ul a -
rise toutes les quations de-structure du type Bethe-Salpeter gnrales o
des noyaux "p-particules irrductibles" interviennent, et ainsi donner un
sens a ux produits de G-convolution a sso ci s. M a i s il s'avre impossible
avec cette mthode de- contrler le comportement asymptotique de la thorie
8
et de .dfinir les n o ya ux p-pa r t i cul e s irrductibles "e xa ct s". Po ur
pa r v e n i r ce but une "r e n o r m a l i sa t i o n " de s qua t i o n s de st r uct ur e est
n ce ssa i r e et de ce t t e fa o n tous les pr o dui t s de G-co n v o l ut i o n di v e r -
g e n ce s ul t r a v i o l e t t e s po ur r o n t a pr s r e n o r m a l i sa t i o n do n n e r un e pr e scr i p-
tion d'i n t g r a t i o n fo ur n i ssa n t t o ujo ur s un r sul t a t fi n i .
Un a ut r e ca s i m po r t a n t o la v e r si o n du pr o dui t de G-co n v o l ut i o n
r e n o r m a l i s se pr se n t e co m m e i n di spe n sa b l e se r a i t v e n t ue l l e m e n t l ' t ude
de s qua t i o n s dyn a m i que s d'un e t h o r i e La g r a n g i e n n e de s cha m ps qua t r e
di m e n si o n s a v e c i n t e r a ct i o n po l yn o m i a l e ; e n e ffe t ce s qua t i o n s so n t
qui v a l e n t e s un syst m e i n fi n i d' qua t i o n s de G-co n v o l ut i o n r e l i a n t l e s
di v e r se s fo n ct i o n s de Gr e e n de la t h o r i e .
Pa r m i les m t ho de s de r e n o r m a l i sa t i o n e xi st a n t e s en t h o r i e de s
pe r t ur b a t i o n s, ce l l e de Zi m m e r m a n n e st la m i e ux a da pt e l ' cr i t ur e
d'a qua t i o n s da n s l'espace de s i m pul si o n s, et ce l a d'a ut a n t pl us qu'e l l e
n 'ut i l i se pa s de pa r a m t r e -de r g ul a r i sa t i o n ; e l l e co n si st e e n e ffe t
r e m pl a ce r 1'integrand de Fe yn m a n i n i t i a l I_ pa r une n o uv e l l e fo n ct i o n
r a t i o n n e l l e >il.gla.b. R a ppe l e l'"intgrand r e n o r m a l i s ".
Da n s l'tude' pr se n t e o n se pr o po se de d fi n i r un pr o dui t de
(2)
G-co n v o l ut i o n r e n o r m a l i s , po ur un e cl a sse a ppr o pr i e de fonctions H ,
( I L ) , . . . .
11 qui o n t un e cr o i ssa n ce lente 1'i n fi n i (dans un se n s qui se r a
pr ci s ci -de sso us) ; po ur ce fa i r e , o n a ppl i que r a la pr e scr i pt i o n a l g -
b r i que de Zi m m e r m a n n po ur d fi n i r pa r t i r de I-(K,k) un i n t e g r a n d r e n o r m a -
lis R_(K,k) do n t o n po ur r a a ssur e r les pr o pr i t s de co n v e r g e n ce . O n
a i m e r a i t a ussi d m o n t r e r que l 'i n t g r a l e H ^
n ) r e n o r ,
( K ) - / R-(K,k)d._k
c
r ( K )
G m
est un e fo n ct i o n a n a l yt i que qui sa t i sfa i t a ux pr o pr i t s i ) i i ) i i ) i v ) , m a i s
da n s le pr se n t trai^il o n se r e st r e i n t la d fi n i t i o n de R et n
n
M r e n o r .)
da n s la r g i o n Eucl i di e n n e (c'e st --di r e i ) ) . Le s pr o pr i t s d'a n a l yt i ci t
de Hg ^
r e n o r
' ' et e n pa r t i cul i e r le pa ssa g e de l 'e spa ce Eucl i di e n
l'espace de M i n ko w ski font l'objet d'un t r a v a i l e n co ur s do n t que l que s
r sul t a t s sont d j e xpo s s i ci da n s la pa r t i e I I .
La pr se n t e t ude (cas Eucl i di e n ) fait l'objet d'un e co l l a b o r a t i o n de
l'auteur a v e c J. Br o s. Le s di v e r se s t a pe s de so n d v e l o ppe m e n t so n t les
sui v a n t e s : Da n s la se ct i o n I o n d fi n i t de s cl a sse s de fo n ct i o n s a ya n t un
co m po r t e m e n t a sym pt o t i que a ppr o pr i a uxque l l e s les fo n ct i o n s ut i l i s e s
9
(2)
(
V
H H R doivent appartenir et on dmontre certaines proprits
mathmatiques de ces classes utilises dans les sections sui v a n t e s. Ces
classed sont caractrises par leur type de croissance l'infini en
termes de puissances des quadri-impulsions dans l'espace Euclidien, et
contiennent comme classe particulire les fonctions rationnelles de la
thorie des perturbations. Une classe de ce type, particulirement b i e n
choisie, a t initialement construite par Weinberg en 1960. Cet auteur
a dmontr un thorme trs utile (thorme du "comptage de s pui ssa n ce s")
qui fournit un ensemble d'ingalits ncessaires et suffisantes auxquelles
une fonction appartenant S. cette classe doit satisfaire pour que la
convergence de son intgrale soit a ssur e .
En thorie des perturbations Zimmermann a dmontr qu'une v e r si o n
simplifie de ce thorme tait suffisante pour tablir la convergence des
intgrales des fonctions rationnelles. I ci , pa r contre on a besoin d'a ppl i -
quer le thorme de Weinberg dans sa versio.i la plus g n r a l e . On ut i l i se r a
e n fait des classes de fonctions un pe u moins gnrales que ce l l e s'de
Weinberg, car il sera ncessaire d'imposer galement des conditions de
croissance l'infini pour toutes les drives successives des fonctions
considres. Comme dans certains domaines de 1'analyse(comme la thorie
des oprateurs pseudo-diffrentiels) o l'on veut viter un comportement
oscillatoire l'infini, on impose que les puissances qui caractrisent le
comportement asymptotique des fonctions, dcroissent de r un i t s quand o n
leur applique un oprateur diffrentiel homogne d'ordre r. Ce type de
conditions sera justifi dans le travail qui suit par un rsultat technique
mais crucial concernant l'abaissement du degr de certains coefficients
asymptotiques (au sens de We i n b e r g ) quand une certaine opration du "reste
de Ta yl o r " est applique : la preuve de cette proprit est donne dans
l'appendice B. Da n s la section IV on utilisera essentiellement ce rsulat
pour appliquer le procd de renormalisation de Zimmermann au cas gnral
de la G-convolution tudi i ci .
Da n s la section II nous donnons une dfinition de 1'integrand gnralis
renormalis R_ e n suivant de prs l'algorithme de Zimmermann : o n introduit
un e somme de contretermes associs 3 toutes les forts (au sens de Zimmermann)
composes de sous-graphes de G.
10
Dans la section III on donne la notion de "fort complte" par
rapport un ensemble de sous-espaces emboits (ou "drapeau".) dans E,. ,,
notion qui tend celle introduite par dans le cas d'un seul so us-
e spa ce . Cette notion est uti'-'se pour obtenir de nouvelles expressions
de R , importantes pour la section sui v a n t e . La section IV contient la
preuve du thorme principal de cet article : R_ satisfait au critre de
Weinberg, donc son intgrale dans la rgion Euclidienne est absolument
convergente. Dans la section V nous dmontrons que l'intgrale H ^
r e n o
est indpendante du choix "a dm i ssi b l e " (au sens de Zi n m e r m a n n ) des v a r i a b l e s
d'intgration, choix dont dpendait la dfinition m m e de 1'integrand.
11
R E N O R M A U Z E O G-CONVOLUTION OF n-POINT FUNCTIONS
IN QUANTUM FIELD THEORY
I - THE EUCLIDEAN CASE
by
Jacques BROS and Marietta MANOLESSOU-GRAttlATICOU
13
A B S T R A C T
The notion of Feynman amplitude associated with a graph G in
perturbative quantum field theory admits a generalized version in which
each vertex v of G is associated with a gene/uc'. (non-perturbative)
v
n -point function H , n^ denoting the number of lines which are incident
to Y in G. In the case where no ultraviolet divergence occurs, this has
been performed directly in complex momentum space through Bros-Lassalle's
G-convolution procedure.
In the present work we propose a generalization of G-convolution
n
v
which includes the case when the functions H are not integrable at
infinity but belong to a suitable class of slowly increasing functions. A
"finite part" of the G-convolution integral is then defined through an
algorithm which closely follows Zimmermann's renormalization scheme. In
this paper, we only treat the case of "Euclidean" four-momentum confi-
gurations .
IS
I N T R O D U C T I O N
It has become commonly accepted in Particle Physics that various:
collision mechanisms-may b e conveniently described in the language of
generalized Feynman amplitudes. These quantities are associated wi t h
"fat" Feynman graphs, in whi ch the orthodox point-wise v e r t i ce s of
perturbation theory are replaced by "bubbles", whenever strong interaction
processes have to be taken into account (For an up-to-date example of such
description, see the well known deep-inelastic scheme ) .
According to the situation which is considered, these bubbles are
supposed to represent either complete scattering or production a m pl i t ude s,
or partial hypothetic "resummations" of Feynman amplitudes. M o r e o v e r , as
far. as the momentum configurations are concerned the latter are sometimes
taken in Minkowski space and then the Feynman i e -pr e scr i pt i o n of i n t e -
gration is used, by reference to the perturbativ case ; but sometimes for
convenience r e a so n s, they a r e also restricted to the Eucl i de a n spa ce , and
the problem of analytic continuation on the mass-shell of the resulting
quantities is left unsolved in g e n e r a l .
2)
It was this conceptually confuse situation which led one of us and
3)
M . Lassaile Co introduce and investigate systematically in complex fo ur -
mooentum space the notion of G-convolution product associated w i t h a graph
G.
The m a i n property which w a s proved b y these authors is the following.
Let G b e a connected graph in whi ch each vertex v is incident to n (external
or internal) lines of G, and let c b e the number of external lines o f G.
Assume that wi t h each vertex v of G is associated a general n -po i n t function
(nv)
H(KY) : under this name is m e a n t an analytic n -po i n t function who se domain
is the pr i m i t i v e * axiomatic do m a i n in the complex space .v. of the four
momenta K
v
carried by the lines incident to v. Assum e that wi t h each
*
Moreover the correct axiomatic analyticity domains for the absorptive parts
of H
v
, or equivalently the Steinmann relations for H ^ are also assumed
to hold (see Ke f. 3 ) .
16
(2)
internal line i of G is associated a general two point function H(.) of
the corresponding complex four momentum Z..
(2)
T hen under suitable integrability assumptions on the H(&.) and on
{n, )
the H (K), it is always possible to define a general n-point function
Hg of the set K of (complex) external momenta cf G, through the followin
formula which generalizes the Feynman integral :
H < K) - J n H
( n v )
( K
v
( K, k ) ) n H
< 2 )
a . ( K , k ) ) d, k <n

r(K)
v
i '
l
*"
In this formula, k denotes a set of 4m independent complex integration
variables, m being the number of independent loops of G .; K (K,k) and
i.(K,k) are linear functions which are determined by taking into account
the four-momentum conservation equations associated with all vertices of
4m
G J r(K) is a 4m-dimensional contour in the complex space . . which lies
in the analyticity domain of the integrand and depends continuously on the
external configuration.
In Bros i-Lassalle's result, Che following properties of IC are
specified :
i) Formula (1) is meaningful when K varies in the "Euclidean region"
E
4(n-1)
( i e < R m i
f
ji
q-
it w i t h

e
R
3
^ - ' ) , Q g E <
n _ , )
;, provided that T (K)
be also chosen as the E uclidean region {k (p,iq) p R ,q R J. T his
4)
is a trivial consequence of the basic fact that the primitive axiomatic
domain of each n -point function H (K ) (resp. H (&.) contains the
V 1 / \
"Euclidean region" of the corresponding K
v
-space (resp. *>^ space). Hg (K)
is thus primitively defined and analytic in E
ii) By using appropriate distortions of the contour T (K) which gene-
ralize the Wick rotations, ni (K) is shown to have an analytic continuation
4)
in the whole primitive n-point domain. The various standard boundary values
of H on the real space (i.e. on "Minkowski-space") are thus defined, and
G
they satisfy the Steinmann relations.
17
iii) The T-boundary value* T *
n )
of H ^
n )
is defined by the Ruelle-Araki
prescription which generalizes the ie-prescription of Feynman amplitudes.
Moreover formula (1) yields an expression for Tri
1
" as an integral over
(n
v
) * (2)
Minkowski space in which each factor H (resp. H
v
') has been replaced
by the corresponding T-boundary value T ^ (resp. T ) in the integrand.
iv) For i) ii) and the first part of iii) to hold, it is sufficient
that the integrability conditions of (1) at infinity by satisfied in
every "quasi-Euclidean" region E(W) - W + E , where W is an arbitrary
compact region of C . (Of course the second part of iii) requires integra-
bility in Minkowski space) .
From the physical point of view, the necessity of imposing integra-
bility conditions as in iv) is not fully satisfactory, since the Green
functions H of a polynomially interacting Field Theory in four-dimen-
sional space time may very well have increase properties at infinity which
produce ultraviolet divergences in integrals of the type (1), as it is the
case in perturbation theory. As a matter of fact, the G-convolution product
defined above revealed itself a sufficiently good notion for carrying out
the study of the analyticity properties implied by Many Particle Structure
Analysis (see Ref. 2,5) : it was becaus" by using analytic cut-offs of the
Pauli-Villars type, all structural equations involving p-particle irre-
ducible kernels would be written in a "regularized" form in which all the
G-convolution products were well-defined in the above sense. However in
this procedure, it was impossible to keep under control the asymptotic
properties of the theory and to define the "exact" p-particle irreducible
kernels. To do this it would be necessary to write "renormalized" structural
equations in which G-products including ultraviolet divergences should be
given a precise prescription of integration in order to yield in all cases
a finite result.
* ,
For the n-point function associated with a quantum field this is the Fourier
transform of the vacuum expectation value of the time-ordered product of n
field operators.
**
Of the general Bethe-Salpeter type.
18
As another important example for which it could be helpful to have
a renormalized version of G-convolution product, we have in mind the study
of field equations of a definite four-dimensional Lagrangian theory with
polynomial interaction. Indeed in momentum space, these field equations
are equivalent to an infinite system of G-convolution equations linking
together the various n-point Green functions of the theory* ; there again,
a renormalized form of the equations is required, as it is already indicated
by perturbation theory.
Among the known rigorous treatments of renormalization in perturb-
ation theory, Ziramermann's procedure is specially attractive for writing
equations in momentum space, since Cat least in its Euclidean version/it is
exempt from regularization parameter and prescribes to replace the primitive
Feynman integrand 1 associated with the graph G by a new rational inte-
grable function R , which is called the "renormalized integrand".
It is our purpose in the present work to define a renormalized
G-convolution product for an appropriate class of functions H
l
' ,H ^ ,
having slow increase at infinity, by adapting the algebraic procedure of
Zimmermann, namely by replacing the primitive integrand I(K,k) of formula
(I) by a new normalized integrand R_(K,k) ; the latter should be built up
with the same ingredients H , H * as I, but its integrability should
be assured. He should also like to prove that the resulting integral
(n)(renor.)
H(K) J R_(K,k)d.k is an analytic function satisfying the above
G ^ G 4m
properties i) ii) iii) i v). However, in the present paper, we shall restrict
ourselves to the definition of R
G
and jj^Hrenor.) .^
R e
E
u c
i i d
e a r l r e g
i
o r l
(i.e. to i )). The analyticity properties of -g^'
{
~
xen
-
r
> and in particular
the passage from Euclidean to Hinskowski space along the lines of [3] will
8)
be investigated in a further paper .
As a preliminary step, it will be necessary to define appropriate
classes in which the various functions H , H , R, will
have to belong ; such classes will be introduced and studied in section I .
* An approach of this type was proposed by J.G. Taylor
19
These classes must of'course contain the class of rational functions
(corresponding to the perturbative case) and be characterized by their
precise type of majorization at infinity in terms of powers of the four-
momenta in Euclidean space.
A very clever realization of classes of functions of this type had
9)
been introduced by S: Weinberg in I960 . This author proved a very
useful theorem ("power counting" theorem) which provided a set of
necessary and sufficient inequalities to.be fulfilled by a function of
such a class to be an integrable function.
In perturbation theory, it turned out that a simplified and weaker
form of the Power Counting Theorem was sufficient for dealing with the
case of rational functions. Here however, we will actually need to apply
Weinberg's theorem in its most general version. As a matter of fact, we
shall use classes of functions which are almost as general as those in-
troduced by Weinberg, except for the following important restriction : we
shall also impose conditions at infinity to the successive derivatives of
our functions. As in other mathematical contexts where one wishes to avoid
-functions having an oscillatory behaviour at infinity, we shall assume that
the powers or "asymptotic coefficients" (in the sense of Weinberg) which
govern the behaviour of our functions at infinity decrease by r units when
a derivative operator of order r is applied to them. This type of condition
will be justified by a technical but crucial result concerning the lowering
of certain asymptotic coefficients when an appropriate "Taylor's rest
operation" is applied : the proof of this statement which is rather tedious
will be given in an appendix. It will be made clear in section IV that it is
fhCb resulc which allows Zimmermann's renormalization scheme to be applicable
to the present case as well as to perturbation theory.
In section II, a definition of our generalized renormalized integrand
R_ is given : this definition closely follows Zimmennann's algorithm and
in olves a sum of countercerms which are associated with all the G-forests:
a G-forest is a subset of "non-overlapping" subgraphs of G.
* See Che definition of symbols in the theory of pseudodifferential operators
20
In section III, we introduce the n o t i o n of "complete forest
wi t h respect to a nested set of subspaces of E,. ." (this is also an
extension of a notion defined in 7 )). This notion allows to write n e w
expressions of R whi ch are used in the following section IV. The latter
contains the proof of our m a i n theorem R. satisfies Weinberg's convergence
criterion, and thus the renormalized integral H - (K) is a well-defined
function in the euclidean r e g i o n .
In a final section V, it is proved that ! is independent
of the choice of "admissible variables"(in the sense of ') whi ch ha s b e e n
used throughout the preceding se ct i o n . In fact the definition of the i n t e -
grand R was actually dependent on this choice of v a r i a b l e s, and it w a s
therefore necessary to establish this final result.
21
1. SOME CLASSES OF FUNCTIONS WITH SLOW INCREASE
1. 1. The cl as s es A^
a )
:
Let E be an N-dimensional vector-space on R. We shall make an
extensive use of certain classes A A ' of complex valued functions with
slow increase on E which have been introduced in 1960 by S. Weinberg i
Each class A/j ' is characterized by a bounded real-valued function a on
the set of all the linear subspaces Sf! { 0} of E. a will be called the
"asymptotic indicatrix" of the class A A T , by reference to the following
fact : each number, or "asymptotic coefficient", a(S) will describe the
power of increase (or decrease)which majorizes every function in A A ,
in "almost all" the directions of the corresponding subspace S ; this is
made precise in the following definition.
A complex-valued function f on E belongs to the class A A ' if, for
every set of m_<N independent vectors L ...L and every bounded region
W in E there exists a* set of numbers b,...b >1 and a constant M> 0
1 m
(depending on L , . . . , L and W) such that :
, m
N
m a({L,.. . L. })
|
Ul
L
j
n
j*"
n
m
+ C
)l
M
.J, "j
( ,
-
when the real variables n .(j 1-,.. .,m) belong to the region { l-^b.} and
when C W.
In 1.1, {L.,...,L.} denotes the linear closure of the set of vectors
{ L , , . . . , ^ } .
From this definition, we easily derive :
Proposition 1.1
a) A A is a vector space (on E. or C)
(
a
,) (a,) fa.-KX,)
b> If f, A
1
.
f
2
V
C h e n
l
f
2

*li
c) If a < a-
:
A ^
a )
e tf*'>
* In the cxlgiuat classes A^ introduced by Weinberg, logarithmic factors
were also written at the right-hand side of (1.1) ; however
this refinement is not needed for the present work.
22
Let us now consider two vector spaces E, E' with respective dimensions
N, N ', and a linear mapping p from E to E'.Then we have the following
property.
Lemma 1.1..
Let f be a function on E' which belongs to a certain class AJJ, ' .
Then its inverse image f - p * f on E (namely : V K E, f(K) - f'(P(K)}
belongs to the class Ai which is determined as follows : for every
subspace S of E, let us denote by S' its image in E' : S' p(S) ; then
a(S) a'(S '). In the case when S belongs to the Kernel of P(i.e.
S' " { 0} ), cc'(S') has to be put equal to zero.
The proof of this statement makes use of the following auxiliary
lemma A : if g is a Weinberg function in AQ ', it still satisfies the
bound of formula 1.1. for aJibii/uxJiij sets of vectors L ,...,L (i.e. even
when the latter are not linearly independent, or when some of them are
equal to { 0} ). The asymptotic coe.$.i.eJ.ZYii a({L,L,. ..L.}) still have the
same meaning as before : {L,,L-...L.} denotes the subspace S which is
generated by the (possibly redundant) set of vectors LpLj.-.-L.. In the
special case S * { 0} , a(S) is defined as being equal to zero.
In view of its technical character, the proof of this auxiliary
lemma is given in the appendix A.
Proof of Lemma 1.1.
Let (L,,L
2
...L ) be an arbitrary set of independent vectors in E and
let C belong to a compact subset W of E. We put : L'. p(L.) ( I j m )
and C p(C ). Then we have :
f
( j
1
L
i
t ,
i V
n

+ c
) "

' ( j
1
L
' i
n
J -
n

+ C
') '
2 )
But in view of lemma A there exist numbers b,,...,b > 1 and a constant
M > 0 such that :
;
( _ ,
] L l j n
. . . . n
m +
C . ) | < M
j
S
i
n
m
. a'( L' ...L'. )
TT
J
J
(1.3)
23
provided that : 0. b. . I f j <.m and C W - P (W).
B y noting that {L '
I
...L
1
.} - p'({Lj ...L .} ), and by taking (1.2) into
account we immediately interpret the bound (1.3) as the announced result
for p*f. I n particular, if L . Ker p for lj m', the corresponding
asymptotic coefficients a({L ?7T 7L T }) are equal to a'({0}) - 0.
q.e.d.
T he most trivial examples of functions in classes A/jj 'are Che
polynomials on E . I n fact, if E - IR it is clear from definition l.a.
that the coordinate function K is in the class A . with (unique) coefficient
o( E ) I. B y applying lemma 1.1 to the case when E ' - IR, we then obtain
that every linear form S. on E with null-hyperplane H belongs to the class
A ^ s u c h that :
CC(S) - 0 if S = H and a(S) - 1 if S H.
M ore generally, by applying proposition 1.1(b, c and a sucessively)
we obtain :
Proposition 1.2.
a) L et {. ; i 1} be any finite set of linear forms on E with
respective null hyperplane H. ; T hen each monomial ir I- belongs to the
(a)
X
.
i 1
corresponding class A_~ which is defined as follows :
V S c E ; a(S) t v.
{iIjSiHj.}
x
b) Let ft, be any polynomial of degree V on E ; thenQ
v
belongs to the
class A i whose asymptotic indicatrix is the constant v(V S c E , v(S) - v ) .
(o)
The following non t r i v i a l property of the cla sses AjJ has been proved
by S. Weinberg in 6% and wi l l be used as a ba sic tool in our section 4.
24
Lemma 1.2.
For a function f to b e integrable on E, it is sufficient that it
(a)
belongs to a class A,, whose asymptotic indicatrix a satisfies the
following inequality :
sup (a(S) + h(S)) < 0 (1.4)
S<=E
In ( 1 .4 ) , the sup. extends over all the suhspaces S of E, and h(S) denotes
the dimension of S.
n
1.2. The classes E
n
Let C be an n-dimensicnal vector space in which a certain n o r m (denoted
by ||. || ) has been chosen. We consider the class C () of all the functions
which are infinitely differentiable on ,
a
**d we denote b y ?
V
( U) any
homogeneous polynomial of degree V in the derivatives with respect to the
coordinates of the variable i n O > In a more intrinsic way (independent of
the choice of a system of co o r di n a t e s), F
v
(D) is a linear operator o n the
space C (E) which is associated wi t h a given element P
v
of the symmetrized
v
tensor product
s
and is defined as follows ; if f C(E) and if f^ ' (K)
denotes the derivative application of order V of f at K, w hi ch is a linear
V
form on O then one has :
V K E E , <P
v
<D)f) (K) - f
( N
( K ) <P
V
) (1.5)
He then define the following classes Z of functions on C > w hi ch are
contained in the general class of symbols introduced in to) and II ) .
Definition 1.b.
Let y be an arbitrary real number. A function f on G is said to belong
to the class E if it belongs to Cf(> and if for every integer V > 0 and
n
every homogeneous polynomial Py(D) , there is a constant C such that :
25
|.<P
v
CD) f ) (K)| C^ IH.PJH (I.
+
H K I D ^ (,.6)
V
here ||| F ||| denotes a certain norm of P i n , which we do not need to
specify.
Remark. is independent of the choice of the representative ||.|| in the
class of all the (equivalent) norms o n C . Indeed in the physical applications
of section 2, there will be no canonical choice of the norm o n ) .
Let now E denote an N-dimensional vector space (as in I.l) and let X
be a linear mapping from E to ; He shall prove :
Lemma 1.3.
For every function f on which belongs to the class S , the inverse
image X f of f belongs to the class Avj whose asymptotic indicatrix a ^'
is defined as follows :
a
( l I >
(S) i 0 if S<= Ker X ") (1.7.a)
a
C p )
( S ) - U if S oi Ker A I (1.7.b)
Moreover, for every integer v-> 0 and every homogeneous polynomial
Q.(D) of degree y on E, the function Q,(D) (A f) belongs to the class
Proof
Let {L.,...,L } be an arbitrary set of independent vectors (m < _ N)
and H a bounded region in E.
Let J < m be" the integer such that :
V i < J , X(Lj) - {0} , and X ( L
j + 1
) jl {0} (1.8)
If J - m, we have :
26
CX*f) ( s L
j n j
. . . T i
m
+ c ) | - | f a < C | _ < M ( 1. 9)
where M - sup | f ( A ( 0 ) |
c w
I f J < m, we have :
. m
|(\*f)(_z L
j n j
. . . n
m
+ c)| - |f(_ i xttj) n
r
. . n
n
+ A( o) | ( i . i o)
Then under the condi ti ons : V j <. m T| > 1, and C W (11)
the norm i nequal i t y yi el c. s :
J Xt t , ) n. . . . n_ + X(c>|| < ( | | X(L , )| | + s u
P
| | X(C )| | ) n n
j - J+i
J J m
M - j +i
J
c e w ' j - j + i
J
(1.12)
Then by taking into account definition 1 .b (namely formula 1.6 in the case
\ i -0, P 1), we easily derive from (1.10), (1.12) the following m a jo r i z -
ation :
|(X*f)^
i
L.
n
....n
m +
c ) | < M j
+ [
(
n
5 C..3)
which holds under Che conditions (1.11).
In (1.13), the constant M is given by the formula
M - C (i + ? | | xa . ) | |
+ S U P
||x( 0||)
u
\ j-j+i
J
c e w '
No w if we introduce the asymptotic indicatrix a by the fo r m ul a e (1.7),
we can easily check that the right-hand sides of both inequalities (1.9) and
B
/
H
, a <IL....L.iX
(1.13) can be rewritten in a unique way as M H [r\.
M
}
j-P
J
'
(since in view of (1.8) the inclusion {L....L.} c Ker X holds if and only
if j < J ) . Thus the first part of lemma 1.3 is proved.
27
To pr o v e Che second pa r e , w e use Che fa ce Chat for every Q in
V v
v
E
s
, there exiscs a n element P in O * such thaC :
Q y C DX X *^ A*0?
v
(D)f) (1.14)
indeed, if J ^ , de n o ce s Che linear m a ppi n g fr o m E Co G
s
w hi ch is
ca n o n i ca l l y induced b y X, F is determined b y the e qua t i o n :
P
v
X .
v
. CQ^), and (1.14) ca n b e de r i v e d from (1.5) and fr o m the
No w , b y a ppl yi n g the sa m e a r g um e n t as a b o v e to the fun ct i o n P
v
(D)f
instead of f (namely, b y usi n g fo r m ul a e (1.9 ), (1.10), (1.12)) and b y
taking (1.6) into a cco un t , w e o b ca i n Che second r e sul t of lemma 1.3.
q.e .d.
a
(a,a)
.... ...w -....., ,. of a dm i ssi b l e We i n b e r g fun ct i o n s
I n this sub se ct i o n , a v e ct o r space & . . < 5
K
\ "
x
E . is
co n si de r e d ; the subscripts (K),(k) e t c ... w i l l always r e fe r to the n a m e
o f Che v a r i a b l e w hi ch is use d in the co r r e spo n di n g spa ce ; Che supe r cr i pCs
rL e t c... r e fe r to Che di m e n si o n of the spa ce . The n o t a t i o n s a r e just i fi e d
b y the physi ca l a ppl i ca ci o n s of Che n e xt se ct i o n : K and k w i l l co r r e spo n d
r e spe ct i v e l y to Che- e xt e r n a l and Co the i n Ce g r a t i o n m o m e n t um v a r i a b l e s
w hi ch w i l l b e associated w i t h a g r a ph G ; these m o m e n t a w i l l b e suppo se d
Co b e l o n g Co a n r -di m e n si o n a l e ucl i de a n v e ct o r spa ce (which pl a y Che .role
of Che spa ce of the e ucl i de a n fo ur -di m e n si o n a l e n e r g y-m o m e n t um v e ct o r s) .
ET. . is the space of the sets of L independent e ucl i de a n r -v e ct o r s, n a m e l y :
k (kj k ^ ) . /
1
jn ~
1
*
s
E
r n
/ E
r
is the spa ce o f the sets of n e ucl i de a n
r-vectors K (K. ...K ) linked b y the r e l a t i o n : K. + K- + ...+K - 0
n
o r N
("conservation of ..Cotal m o m e n cum "). In Q, .., N L + n - 1 de n o ce s Che
Cocal n um b e r of independent r -v e ct o r s. He shall also m a ke use of the ca n o n i ca l
pr o je ct i o n s it (resp. X) of (K,k) o n t o E* (resp. 6 ( ^
_ I )
) 5 n a m e l y w e
ha v e : ir(K,k) - (0,k) ; x(K,k) - (K,0).
We s hal l always i dent i f y E
f
. . with the subspace W/ i n *
E
n , \
o f
Qfg u\
28
As in subsection 1.2, we shall also consider homogeneous polynomials
in the derivatives with respect Co Che variables K, which we shall denoce
by
p
m
( D
K
) ! for every f C"(*
fc)
) W
f i s a w e l 1 d e f i n e d
function
" (K ki
u h i c h d o e s n o t
depend on Che choice of Che coordinates (see Che
beginning of 1.2).
Similarly we shall also consider the Taylor expansion of degree d
of a funccion f with respecC Co Che variables K ; we denoce it by t,
v
.f
C
d \
I - t ,y. Jf : these functions are
always considered as intrinsically defined on G
f K
v}> even if their explicic
expression in coordinates has to be used in some of Che arguments given
below.
Definition 1.d.
PrN
Let o be a given set of subspaces of C ^ which are contained in
rL ~~ ^ '
E..., and let a be an asymptotic indicatrix such Chat for every subspace
S with ir(S) a, one has :
a(S) - a(ir(S)) (1.15)
Then we define a class C C M of admti&iblt WninbeAg rfuncXioni f (K,k)
by the following conditions :
i) f belongs to the class A^.
ii) for every homogeneous derivative polynomial P (D ) of degree m, the
function
p
_(D
K
)f belongs to the class k

^which is defined as follows :

- V s with ir(S) a : ot (S) a(S) m .
m
- V S with TT(S) ? a : a
m
(S) a(S)
From Che above definicion and from proposicion ! . 1 , we easily derive
Che following
Proposition I. 3.
i s a veccor space.

i f f
i
e a
r N
( )
'
f
2
e Q
- r N '
c h
-
f
l

CL
r
N
(

29
d) if a* >.t (i.e. V S : a'(S) >a(S)) then : Q ^ '
a )
=C^ H''
< , )

We now consider two spaces >(v kl G(K}
x
^fk
1

a n <
'
following linear mapping s from /
K
v> into G/v> t-i}
!
K' - p(K,k)
(K'.k) - s((K,k)) : I (1.16)
k' - k

?N Cr(n'-l)
(K kl
l n t 0
GfK'* ! ^
a n d
ir' denote the respective projections : (K,k) * (0,k) and (K',k') * (0,k').
I he linear mapping K' p(K,0) from (Sri? into < ? ? . ' induces
W m
v
l * > ,gj
m
a'well-defined linear mapping p. . from ((_),? )
i n t 0
(6(K) ] '
So, to every derivative polynomial P (D), represented by an element P in
. m & m
m
1 (5, )
S
> there'corresponds a derivative polynomial P
m
'(D
K
r>. represented
by the element.
(
W (
sf
<
K
W "
s
<
p
V
D
K
, ) f )
(
K
''
k
'> <'-
17
>
How we can prove
L emma 1.4.
L et f(K',k') be an admissible function in a class Ui' . T hen
the inverse image (s f) (K,k) f(s(K,k)) of f is an admissible function in
a class Clij,' which is defined as follows :
a) if E ^ r. is identified with E^ . ,. (by taking into account that
s(0,k) - (0,k)), then a - a'.
Namely, the tangent mapping to. p at any point (K. ,k >.
30
b ) for every subspace S of J , ~ , . wi t h image S
1
s(S) in jSLi
k
'\ >
one has : ct(S) -ct '(S').
Remark. One easily checks the following identity whi ch results from the
special form (1.16) of the mapping s :TT = S O T T I T ' O S ; it implies :
s[ir(S)] - ir(S) * ir'(S') . (1.18)
Proof of lemma 1.4.
* (CO
In v i e w of lemma 1.1, s f belongs to the class A \ / such that :
V s, cs(S) a '(S').
Let a b e defined by a) , and let 3 b e such that TT (S) a . Then we ha v e ,
in v i e w of definition 1.d and formula (1.18) :
tx(S) =ct '(S') = a'(ir'(S')) = a(7TC5)) (1.19)
so that the indicatrix a satisfies the requirement (1-15) of definition
l.d.
Let us n o w consider an arbitrary derivative polynomial ? (D ) and the
corresponding polynomial P
1
(D
T
,,) ; then in v i e w of condition ii) of
n
(<*'m)
definition !.i ., we can say that
p l
_ (
D
K
t ) f A , , w i t h :
o ' (S') -ct '(S')-m if ir'(S') o \
and a '
m
(S') -<x*(S') if ir'(S') a'.
Formula (1.17) and lemma 1.1 then imply that P (D )(s*f) belongs to the
(a )
m K
class A ' where
rN
- O
m
( s ) - a '
m
( S ' ) - a <(s') - m - a (s) - m, if TT(S) a
- aJS) ' ct '(S') > a'(S') - a (S) :f ir<S) t a-
q.e.d.
31
Fo r proving the following lemma 5 o n Taylor expansie-'s of admissible
functions, w e need the two following auxiliary propositions.
Proposition 1.4.
Let f belong to a class 1C?' , and let ir f b e the inverse image of
by the projection (i.e. f(k) considered as a function of (K,k), constant
in K ) . Then for every set T of subspaces of ff.
v
., ir f belongs to the class
Qi j.' whi ch is defined as follows :
- for every subspace S of <5^
K fc
. : S(S) - a(TT(S)) (1.20)
Proof :
In view of lemma 1.1, the function ir f belongs to the class A Jr ,
w i t ha de fi n e d b y (1.20). It is clear that for every set T,S satisfies the
requirement (1.15) of definition l.d. . Moreover, since P (D)(TT f) * 0
for every operator P (D), the condition ii) of definition l.d. i~ trivially
satisfied.
Proposition 1.5.
Let Q_(K) be an arbitrary polynomial of degree m of the coordinates of
C4N
K, considered as a function on >
rv
^\ (constant wi t h respect to k ) . The n for

4 N
I*-,*; (Tfrt(
m
>T) -rl
. . ., <LXK) belongs to the class Ujjj *~'
- if u(S) T : a
< m , T )
( S ) > m
- if S * E *
L
, and TT<S) t T : a
( m , T )
( S ) - m (1.21)
"00
"00
- if S = E ?
L
, and S T ! a
( m
'
T )
( S ) - 0
Proof
Let X be projection of <^K,k)
o n t o
( K ) ~
0 < x ( K
'
W
* <
K
> >>-
W e h a v e
,(m,T)
to prove that x*<L belongs to Q*i'
N o w
'
i n v i a w o
pt
;
' si t i o n 1.2.b.,
32
0 . belongs to the class A ^ . _
n
, where the index m denotes the constant
function : m(S) - m for every S
c
C / i ?
Due to lemma 1.1., we can say that X <L belongs to a class Q'
1
such that :
fi(S) - m if S P E*j
fi(S) - 0 if S c E *^
Since m^ 0, i t is clear that :
VT , VS : ffl(S) <. cc
( m , T )
(S) , so that
x
\ i j j '
)
.
V x , a
t m , T
' satisfies condition (1.15).
Finally every function P
t
(D)(<L ) is a polynomial of degree m-m'
and obviously belongs to a class A which fulfils condition ii) of
definition l.d. . We are now in a posicion to prove :
L emma 1.5.
L et i\K;k) be an admissible Weinberg function in a class Cl^ [ , and
let h(K,W - ft^ gfWk).
rL
For every set T of subspaces of E
f
. such that T = C, there exists a
a
( a,_-sx)
^
T J
vhich contains h and which satisfies the following properties
*) V S, with ir(S) a-
a
( x )
(S ) - a(S).
b) V S, with it(S) t a , (S) T ;
a/
x)
(S) - o(ir(S)) + d.
33
c) Y S = E J \ . wi t h S E T. :
"(T)
< S )
"
a ( S )
*
Proof i
By choosing a set of coordinate? (K) {K, ; l ^J. 4( n- l ) } for represent-
ing K, we can write :
h(K,k) - 2 i S_ D^f ( 0 i k ) (1.22)
a
;
q<|x|<d}
v ! K
where we have used the notations : V ' {Vi ; lA4(n-l)} , |v| X V-, ,
V! - *(V
X
! ), (K )
u
' l \
X
, and D f - (jD^)f.
In view of proposition -O 3) a), we are led to show that each product
(K) ,D
K
f(0,k) occurring at the r.h.s. of (1.22) belongs to the class (J, ..
V
By applying proposition 1.4. to the function D_f(0,k) which belongs to
< a
|v|>
the class A .' ' (see definition l.d. ii) such that :
0|
V
|(S) -a(S ) - [v| if s e a ,
ci|
v
| (S) - a(S) if s t o ,
we conclude that for every set T, the function Tr*/DYf \(X,k) D^f(O.k)
belongs to the class<-A^ , such that :
L(S) - <x(ir(S)) - [v| if ir(S) e a ,
V
(1.23)
6\,CS) -o(ir(S)) if n(S) Sa
By applying proposition l.S to the function (K) , we now see that for
every T, (K )
u
belongs to the class W <
r i )
, defined by formulae similar
to (1.21) (with m * |v| ) .
34
Then, in view of proposition 1.2.b., we conclude that the product
V V
(K) Df(0,k) is an admissible Weinberg function in ctie cl a ss
Q^ H'
T
>.T )_
Formula (1.21) and (1.23) allow us to compute the function

V
( r ) - W
V ,
'
T )
.
a) if ir(S) CT :
a
V
( T )
( S ) - (a(Tt(S)) - |v|) + |v| - a ( S ) (1.24)
b ) if a (S) S a , ir(S) T :
ai
T>
( .S) -a(7T(s)) + )v| a (i r (S)> + d (1.25)
c) if S c E ^ K , wi t h S T :
o ^
T )
( S ) - a ( S ) . (1.26)
V
Le t us call ct
f
, sup
a
f-ri then in v i e w o proposition 1.3.d,
we have :
W
t v
;
^|<
d
} CO
V V with 0 < |v| <. d L (K)
V
D^f(0,k) c d
r N
( T )
' .
Moreover, formulae (1.24), (1.25), (1.26) show that a
( T
. satisfy the required
conditions a ) , b ) , c ) , of the lemma.
q.e.d.
Lemma 1.6.
Le t f(K.k) b e a n admissible Weinberg function in a class (i ' and
C
d \
1 - t A f. The n for
every set T of subspaces of E*|\ with T <= CT, there exists a class
35
a
(
a
(_)>T)
' j, which contains g and which satisfies the following properties :
a) V S c S" with (S) T :
O
( T )
( S ) - a
(T )
(i r (s )) - a (S )
b) Y S C E ?\ with. S $ T and S cr. :
a
( T )
(S ) : a(S) - d - l
c) y S e E ^ with S f d :
a
( T )
( S ) -cc(S).
In the proof of this statement, it will clearly appear that T can be chosen
to be the empty set 0. With this special choice, properties b) and c) can be
equivalently formulated as follows.
Lemma 1.6.
If f belongs t oU /' , then g - I 1 - t,, )f and all the derivatives
of g with respect to the variables (K) belong to a certain class A such
that
a) V S c E ^. with S a
00
<*i(S) - a(s) - d - 1
b) V S = E'y with Si a :
a
0
(S ) -<*<S)
Proof of lemma 1.6.
It is a straightforward application of lemma B.2. Let us consider a
fixed set T e a , and an arbitrary derivative D^g of g, with order |v|;our
argument will include the case of g itself.
36
D
K
g belongs to the Weinberg class A ' ' ', with ' ' defined by-
formulae (B,43) ...(B.46).
We now notice that the following inequalities hold for every value
of |v| :
a) V S, withir(S) T , we have (due to (B.43), (B.44) :
O j
v
|(S) -ct(S) - |v| if So ! f
w
rL ? <">
O j
v
|(S) <a (S) - |v| if S = E
( f e )
b) V S with IT (S) S T, anriir(S) 6 c , we have (due to (B.43), (B.44)
O j
u
|(S) -a (S) - |v| < a (S), if S g! E *^
ai
u
i (S) < o(S) - d - 1, if S c E ^
0 2 s )
a j
v
| w ^ o i s ) - o - i, ii a c - .
c) V S. withir(S) . a , we have (due to (B.45), (B.46))
a. ,(S) < sup [a(S')] + d + 1, if S St E ^
l
V
i -{S'ir(S')-(S)}
W
a|
v
|(S) - a ( S ) , if S c E ^
}
Now let us define the index function a , -, as follows :
- a ^ S ) - O
( T )
( I T ( S ) ) -a (S) for S such that IT (S) e x (1.29>
-o c,
T)
(S) -<x(S) for Sij! E*j wiehir(S) e O VT .
- a
( T
) ( S ) -a (S) - d - 1 for S c E* and S a - T (I.30)
-ci(y>(S) - sup [<x(S')l + d + I, for S i E.^.
with ir(S) t a .
37
- a
( x )
( S ) - a (S) for S e E *
f c )
., and S t a. (1.31)
The above analysis clearly shows that (in view of proposition l.l.c. and
definition !.d.)i g belongs to the class Cl -
E
j/
T
^' . In particular,
formulae (1.29), (1.30) and (1.31) exhibit respectively the properties
*),b) and c) of a, . which we r e stated in lemma 1.6.
q.e.d.
We shall end this section by proving the following lemma w hi ch relates the
classes Z of subsection 1.2 wi t h the classes of admissible Weinberg
functions Q i
(a.o)
Lemma 1.7.
Let X be a linear mapping from the vector space &.,, into an
e n - ~
l
'
n-dimensional euclidean vector space C f v ) ^
K
" ^CK,k)).

n U
,5, whi ch belongs to the class E~,
* * '
n
the corresponding inverse image X f of f is a n admissible Weinberg function
o n Cri, i.\ whi ch belongs to the following class Ct !
^ N , K ; rN
a) a is the set of all subspaces S of E... such that S < Ker \.
b ) the indicatrix a is such that :
- V s c ^
k )
withir(S) a :
<x(S) - a(ir(s)) - U . (1.32)
- V s = E
T
^\ with S ? a a(S) - 0 (1.33)
- Y StfE^Jj with ir(S) t at a(S) - sup (0,u) (1.34)
38
Proof :

rH
l v
.. playing the role of E) : we
obtain that X f belongs to the class A V j' with a defined by (1.32),
(1.33), (1.34) j in fact , when ir(S) u, i.e. ir(S) Ker X, then
necessarily S / Ker X and (1.7.b.) implies that o(S) c r
w
(S ) y ;
similarly (1.33) is implies by (I.7.a). Now both formulae (1.7.a.) and
(1.7.b.) are taken into account together with proposition l.l.e. to yield
(1.34) (since in the latter case S can be. either contained or not in Ker A ).
It remains to check that X f satisfies condition ii) of definition
l.d. : but this is entailed by the second part of lemma 1.3 (together
with proposition 1.1. c for the case ir(S) ( Cf).
39
2. A GENERALI ZATI O N OF ZI M M ERM AN'S RENO RM ALI ZED I NTEGRAND
2.1. The unrenormalized integrand I
Let us consider a general connected graph G with n external lines
and m independent loops. Let < JL denote the'set of internal lines of Q,JY~
the set of its vertices, X the set of its external lines : | X | n. Each
internal line is considered as oriented ; it means that a sign . is
prescribed for each couple (i o , v tJf), which is + 1 or - 1 according
to whether the line i points towards the vertex v or not.
With each external line of G, we associate an "Euclidean" r-vector
K.6 E , which is also represented* as :
K
j "
?
j
+ i Q
j
?
j
e
*
r
~ " Qj e ft . < i ' e (2-0
and whose norm is ||K.|| - [<?.,?-> + ( Q ? )
2
]
, / 2
; each vector K. respresents
the external momentum which is carried by the line j, and these vectors
satisfy the total energy-momentum conservation relation : E K. 0
I ii
j e X J
We denote by K *) the set :
K - {K. R ^ + i l R , j X- ; Z K. - 0} (2.2)
jex
J
Definitions 2.a.
) For every vertex v of G, we call X the set of the lines of G which
are incident to v, and call their number n - | X | .
M a E X . it concides either with an internal line i or with
an external line j X ; we shall then write : a -<*(i) o r a (j).
With each vertex v, we associate a set of r-momenta :
K
V
- ( K ' J 0 6 X ; S K* - 0 } (2.3)
a v
ex
v
a
This representation is inessential in the present work, but refers to the
usual complex four-momentum Minkowskian space C
r
in which E
r
is imbedded s
analytic continuation problems in C
r
will be considered in a further work.
40
which varies in the Euclidean space R with dimension r(n - 1 ).
ii) With each internal line i 6<c , we associate an r-momentum .
which varies in the (r-dimensional) euclidean space C .
A set of independent "internal" (or "integration") euclidean
r-momenta of G is defined as a set of r-vectors
k
- <
k
. y
e E
w
whf.ch. satisfies the following property.
For every vertex v (resp. internal line i c ) there exists
a timan. mapping A (resp. A .) from the space <. . . (5fv? *
E
n c 1
onto C (resp. .) such that the corresponding substitutions :
A
(K,k) -& K
v
(K,k) (2.4)
A .
(K,k) <f-i-> ^i(K,k) (2.S)
be solutions 'of the following system :
- V a X
y
: i
( j )
(K, k) - K. , if a - a (j) (2.6)
is an external line, i.e. j X ;
V i )
Z r (K,k) - 0 (see (2.3)).
a x. ^
Definition 2.b.
.(2.7)
internal line, i.e. i e C ;

r
_ . .. .
u
- - - (Kk)
b y t h e
following procedure :
i ) With each vertex v of G we associate a general n
v
~poi n t function
defined on the primitive analyticity domain of Q.F.T. (see introduction)
41
contained in C" and denoted by H (K ) . V/ith each line 1 of G
(2}
ue associate a general 2-point function K.UJl .) defined on the
corresponding primitive analyticity domain contained in
r
. Here w e
shall only consider the restrictions * of the above functions H (K
v
)
and H> ' (.) to the Euclidean regions , ., the latter being
1
- v i r(n -1)
respectively realized as the following subspaces of e and C
F v
r ( n
v
_ 1 )
* v o ,
C - fr
v
C
v
: ImK - 0 Re (K
v
) - 0}
<2.8)
U . C
r
: Imt, - 0 Re*? - 0}
This is meaningful since, as it is known -) 3 ) , the Euclidean region
Q (resp. C O is contained inside the primitive domain of analyticity of
H " (K
v
) (resp. H / ^ C ^ -
w
e shall put :
C\* H
v
(K,k) - H
v
(K
v
(K,k) (2.9)
&*,. H<
2 )
)(K,k) - H P c ^ C K .k ) (2-10)
Ve moreover define the aompZateZy ampufatzd general n -po i n t functions by :
< n ) ( n )
m
_ i
H
v
(K
V
) - H
V
(K
V
) x n [ H^'( < ) ] (2.11)
where , is equal to + 1 if a " <*(j) with j' X, and S - .
v
if
<x - ce(i) with i ^ :
We then also put ( **Jk "" )(K,k) - H "
v
(K
v
(K,k) (2.12)
ii) We assume thaL for each vertex v (resp. line i .J- ) , there
exists a class of functions t7. ., (resp. S,
1
) to which H "
v )
(r e sp. H
( 5
)
(.rm
v
-l) r
belongs ; in the following u and y. will be always considered as integers.
In the following we shall only consider these restrictions and omit for
simplicity the subscript
42
From (2.12) (resp. from (2.10) and lemma (1.3) it then follows
that H
C
V(K
v
(K,k))) (resp. H ^ ( *
i
( K , k belongs to the class A ^
(resp. A ) with corresponding asymptotic coefficients :
<*
V
(S) - 0 if S =Ke r X
y
(resp. a. (S) - 0 if S e Ker X.)
(2.13)
O
v
(S) - U
v
if S Ci Ker X
y
(resp. a..(S) - ^ if S Ker X
i
(2.14)
erN
lii) We define the function I on .~ v\ ca-xied the unrenormalized
integrand by the following products :
I(K,k) II H
v
(K
v
(K,k)) H H.
(
?'u.(K,k)) (2.15)
We show the following properties :
Lemma 2.1.
a) H
v
(K
v
(K,k)) (resp. H
( 2 )
(. (K.k)) belongs to a class CL,
V
'
V
(r e sp.U ) of admissible Weinberg functions. The set CT (resp. a.)
^^ rK
v
^
is the set of subspaces S <= E .
R
.. satisfying : S 4- Ker X
v
(resp.
S< f Ker X .) , and a
v
, a. are defined through formulae (2.13),, (2.14).
a class GL.
M
of admissible Veinberg functions.
The set a is given by :
b) I
G
(K, k) bel on
c
to ..* ^
"c " < ^ j D ( D O (2.16)
crN
The asymptotic coef f i ci ent for each subspace S e Q . i s :
V
S
> - ^
S
\
+
= ^ '
l
(2.17)
i
In -this equation the sum Z ^ ' ( ^ ' ) runs over all vertices v (internal
v i
lines i) of G for which : S Ker 1 (resp. S d Ker \ . ) .
43
Proof :
a) is a direcc application of lemma 1.7 to the case of the mappins X
and A .,
i
b) By taking the result of a) into accounC and applying proposition
1.3.b. and d. Co Che product (2.15), the result is obtained.
'2 0)
In perturbation theory, Zimmermann ' has proved that for a Feynman
graph Bogoliuiov - Farasiuk - Hepp's method of normalization in configur-
ation space can be worked out in momentum space independently of any
ultraviolet rgularisation. L et us denoteby R_ the renormalized Zimmeruiann
F
integrand associated with a Feynman graph G, and E the corresponding finite
part of the Feynman amplitude such that :
H<K) - J R *(K,k) dkj...dK
m
(2.18)
V
T he function-R_ is defined by Zimmermann's method as a rational function of
u
Che incernal and external momenta which is obtained by subtracting appropriate
Y
counterterms from the integrand I
G
of the divergent Feynman graph.
In the present work we shall extend Zimmermann's precription of
renormalization to the case of a general integrand I_ defined by (2.15),
and we shall show that in the E uclidean region fv? ~
a
certain
"renormalized convolution product" H. can be defined through an integral of
the form :
H (K) - / R -(K,k) dk....dk
m
, (2.19)
(
V (2)
where R . is a functional of the H and H: 'obtained by the following
V 1
procedure.
2.2. A dmissible sets of momenta for G and its subgraphs
Let us recall some usual notions in the matter. A graph G is called
one-particle irreducible or "proper" if it is connected and cannot be
separated in Cwo parts by cutting a single line. A graph y is a iabg/uzpti
44
of a graph. G, written y G, if the vertices of y form a subset J f of
the set J f of vertices of G and if all the lines of Y are lines of G
ending to vertices of y; moreover some of the internal lines of G may be
cut to form external lines of y. T wo graphs y., y, are called non-
overlapping if * either yj n y
2
' i or yj c y
2

r
y* c y,.
A G forest U is a set of non trivial nonoverlapping one-particle
irreducible subgraphs of a graph G. U say also be the empty set. A G
forest U is called full if G U ; if G & U then U is called a normal
forest. If y is any subgraph of 'J then U(y) denotes the set of all
y' G U with y' <= Y O'n particular U(G) = U). In the following, all the
momenta which we introduce belong to the euclidean space E
r
= R
r
+ i R .
Definitions 2.C
i) For any subgraph Y of G, we denote by J l the set of vertices of
Y, by ccL (resp. X ) the set of internal (resp. external) lines of Y ;
J f, (resp. <ZY ) is a subset of J f (resp.od ) . We alro call n |X | the
number o'f external lines of y>
a n (
* m(y) the number of independent loops
of y (note that m(G) m, n_ - n, J K
m
J f etc . . . ) .
Y y
ii) With V, cne can always associate two sets of variables, K and k ,
C G
which play the same role as the variables K and k for G (K' - K, k k).
- Kf is the set of "external" r-momenta of Y, namely :
K
Y
- {kT ; : ; . KT 0 (2.20)
J f
jXy
J
-r^ -1)
it varies in a space ,'
(KY)
- k {kT ,...,k
Y
,
v
} is a set of independent "internal" (or
"integration") r-momenta of y; varies in a ' iace E ^ '.
(K
Y
)
T he choice of the variablesk^ contains a large arbitrariness, but
satisfies tne following requirement.
For every vertex v .J I (resp. internal line i J -J i, there exists
a Unto*, mapping V' (resp.A') from r J ' x C
v
to r (resp. .)
I *" (KT) (kT_> _ _
* T he intersecL-.on of graphs is understood as being taken on the set of
vertices and on the set of internal and external lines.
45
such that the corresponding substitutions ;
(K
Y
, k
Y
) J . K
V
(K
Y
.k
Y
)
(K
Y
, k
Y
) ^ S . d.(K
Y
,kY)
be solutions of the following sustem :
- K' (K
Y
,k
Y
) - 0 (see (2.3))
ax
a
v
- K
v
a ( i )
(K
Y
,k
Y
) = .
v
*.(KY,k
Y
), (2.21)
if a = a(i) is an internal line o f (i <^
v
)
- K
v
(K
Y
,k
Y
) - K
Y
, (2.22)
(j)
J
if a - ce(j).is an external line of Y(j X ).
"*G ~G G
Note that X - X , X. , X. X. (see definition 2.a.),
v v ' i ' i i
iii) I t has been shown by Zimmermann '' through a constructive procedure
that it is possible to define simultaneously for all subgraphs Y of G
(including G) the variables (K
Y
,k
Y
) and the corresponding mappings X., xY
in such a way that the following properties hold :
- for every couple (Y>Y') with Y'
c
Y (where Y can be G itself), there
exists a linear mapping gL from ' * E ^ " to Q ! x E
1
-,' '
Y
(KY) (fcY) (KY) (kY)
which has the following form :
K
Y
' - K
:Y
'(K
Y
,k
Y
) (2.23)
% '
k
Y
' -k
Y
'(k
Y
) (2.24)
46
- V (Y.Y.'.Y") wi t hy" = Y*
C
Y > <e has :
gy,i - B
Y
o j j , (2.25)
. The important point here is that k
Y
is a function of kf but does
not depend on K
Y
. Following the terminology of Ti
f w e
say that such a set
of variables { (K ,k ), V y = G } is an admiii.iJbZe. izZ o& baitc momenta. The
choice of this set still contains a large arbitrariness. Such a choice
will play a crucial role in the definition of R
r
and in the proof .'of the
convergence of the integral (2.19 ). We shall suppose that we have chosen
once for all a certain admissible set in the following sections 3 and 4 ;
after the convergence proof we shall show that H is independent of the
ta
choice of any admissible set of basic momenta (See section 5).
Definitions 2.d.
i) For every Y <= G, the mapping 8 allows to reexpress 0& ,kf) as a
function of (K^ ,k) through the linear substitution k
Y
- k
Y
(k) (see (2.24)).
(K
Y
,k) varies in a space /** - g ^ ' ^ *
E

Y >
G '
V
" ^
+
\ " >>
(K',k) (K
Y
)
u ; Y Y
Then with every couple y ,y ' such that y ' c y , we can associate a
linear mapping sj , from ^ to ^J which is defined as follows :
^ (K
Y
,k) (K
Y
,k)
(K
Y
,k) < * (K
Y
* - K
Y
'(K
Y
,k
Y
(k)),k) (2.26)
In (2.26) the substitutions are those defined by B ,, and By (see
(2.23),(2.24)). It is easy Co check that due to formula (2.25), .the set
{ S
Y
, ; V Y , Y '
C
G } satisfy the similar property :
if Y " c Y ' = Y :
s Y
" ^'"
s T
>
( 2
-
2 7 )
1
Y Y Y
~Y ~Y
ii) With each mapping X (resp. X^) introduced in definition 2.b. i i ),
we can associate a mapping A (resp. X.) from ' to (\ (resp.C.^
(K
Y
,k) ^
47
through the following formulae ;
(K
Y
,k) 2-> K
V
- K
V
(K
Y
,i J(k)) (2.28)
X
(K
Y
,k) *-> K
l
- 4
1
(K
l r
,k
r
(k)), (2.29)
In (2.28), (2.29), the substitutions are those defined by X
Y
,X
Y
., g?.
Definition 2.e.
C
V (2)
Being given the sets of functions H , H of definition 2.b.,
we shall now associate with them for every subgraph Y o G a function
LXK^.k) on
Y
which is similar to I. (see formula (2. IS)).
Y
(K
Y
.)
We put :
I (K
Y
,k) - II H
V
(K
V
(K
Y
,k
Y
(k))) x n H?
2 )
(i .(K
Y
,k
Y
(k)))
Y
vt/f ^ ' ie<y
x
^
L
/
By writing :
H
<n
*V(K
Y
,k
Y
(k)) -(xf H ^ . k ) )
and Hf>(*.(K
Y
,k
Y
<k)) - (x
Y
* H<
2
>)(K
Y
,k)
and applying lenma 1.7 to the mappings X, X , we deduce from the
assumption H
v
, . and H: '
l
the following lemma whose
r (n
y
-U i r
proof is exactly similar to that of lemma 2.1. (one now applies proposition
1.3.b. and d. to the product (2.30)).
Lemma 2.2. .
Y
.For every Y <= G, the corresponding function I(K ,k) belongs to the
(a. 0" J '
class Q
Y
which is determined as follows :
r N
Y
48
where 0? - S c E
}
; S Ker xTi (2.32)
(2.33)
a
v '
{ S
=
E
" )
: S
*
K e r X
v
}
b ) V S, c y , one has
(K
Y
,k)
I U +
V V ", I
Y
V
v
+
v V; (2.34)
2.3. Ihe renormalized integrand R-
Definitions 2.f.
We first need to recall the following notions which are relative to
forests of subgraphs in G.
i) If Y is an element of a given forest U(G), we shall introduce the
set {Y-sKa ^c,} of all subgraphs Y U(y) which are maximal in Y. We
also consider the associated "reduced graph" f of , which is obtained
from Y by contracting each Y to a single vertex in Y. (see 7) ) . We then
call m(Y) the number of independent loops of "?. In view of the definition
of m (Y) (see definition 2.b.))> we obviously have :
n(Y) - m(Y) + I m(Y.) (2.35)
l<ac
a
Note that the definition of Y depends on the forest U U(G). When several
forests U are insolved, we shall use the more precise notation Y(U) in order
to avoid any ambiguous meaning (it will be the case in section 3 ) .
ii) With the reduced graph G (resp. ?) of G (resp. Y) in the given forest
U(G), we associate the following functions Ig (resp. I-) :
49
n. . , v, , . , ,
( 2 < 3 6 )
I
?
( K, k) - H 2
v
< K( K, k ) j[
H
<?)(Jl . (K, k))
G
v e ^ ies&
w 1
'a J*
a - I . . . c
Y
r . . . e
( res p. : lT.(K
Y
k) - H V l l V O t ) ) " H^( a . ( K* , k? ( k) )
' V-C, i e
w x
v"*
jf ,-rj"Y
Y
a
l C
* Y
a
( 2. 37)
a - l . . . c
Y
* - 1 - . . c
v
By applying again lemma 1.7 and proposi t i on ! . 3 . b . and d. we then obtai n a
property which i s analogous t o lemmas 2 , 1 . and 2 . 2 .
Lemma 2 . 3 .
Let y be the reduced graph of Y in a forest U(G). Then the correspond-
ing function I_(K',k) belongs to the class ( J, ' ' which is defined as
y rtr
rN
f ol l ows :
-Y n ^nr n 4\ )'f-r M \\n( J * i\ (2.3 8 )
l l ^ . ' Y
Y
' \*.'*.
G
r '
a
where aL oy are defined by (2.32), (2.33)
b ) V S <=
t N
one-has :
y V ? k )
o - ( S
Y
) - I y
v
+ X
u
( 2. 39)
l<a<Cw. ' a ^ ^ . S -
a
t ( S
r
) e a
Y
t r( S
y
) eo/
Def i ni t i ons 2. g.
i ) The s e t s { j j
v
i vCi*} and {ji . ;i%} being gi ven, we as s oci at e wi th
every subgraph Y of 0 the f ol l owi ng number which we c a l l the "dattUUion
O&Y" ( r e l at i ve to the l at t e r s et s ) :
50
<*M - I K * I t * M <
2
-
4 0
>
vSJC
y
v
i e
Y
If we similarly put :
d(Y) ". I V
v
* I P, + rm(Y) (2.41)
vec^v- u JT
y
> i e <
Y
\ u J> )
l
l
l
a
l
c
t
a
'
a
l
c
y
for the reduced graph Y of Y (relative to a certain forest U(G) containing
Y), we have the obvious relation :
<3(Y) - d(Y) + 2 d(YJ (2.42)
ii) For every function F(K
Y
,k) on &
r N
v
, ( t
d ( Y
M (K
Y
,k) will
(K
Y
,k)
V J
denote the Taylor expansion of F of degree d(Y) with respect to K' at
Y
K
1
0 (this function is intrinsically defined : see section 1.3). Of
course this definition only holds if d(Y) 2. " which is not necessarily
the case (if this is the case Y is called a "renormalization part" as
in
7
' ) . If d(Y) < 0 we put by definition t ~ ' F - 0.
Remark. The mapping (K
Y
,k) * (K
Y
,k
Y
- k
Y
(k)) defined by (2.24) commutes
with the.operation t
d ( Y
\ ely if F(K
Y
,k) - F(K
Y
,k
Y
(k)), then
( t
d
F ) (K
Y
,k) - ( t
d
<
Y
>F ) (K
Y
,k
Y
)|
k
Y
-k
Y
(k)
Definition 2.h.
We propose the following generalisation of Zimmermann's renormalized
integrand. Let Ubethe set of all the forests of G. Then we put :
R<K,k) - I F(K,k) (2.43)
SI
where : V U IL : F(K,k) - Y^
U)
(K,fc) if U is normal
- (- t
d < G )
M
D )
( K , k ) if U is full
Ihe function Yg (K,k) and all the auxiliary functions Y (K^,k) correspond-
ing to every y U are recursively defined as follows :
Y (K
Y
,k) - I-(K
Y
,k) (s*(-C " Y f W , k ) . (2.4S)
In (2.45), I- is the funccion introduced by (2.37), and s* is a short
T y *
notation for (s' ) , namely the inverse image operation induced by the
Y
a Y
linear mapping s
v
defined through (2.26) (definition 2.d. (i)).(Indeed,
d < Y
a >
a
Ya
- t Y
v
is a function of (K ,k) ; s denotes the substitution :
a
(K
Y a
,k) - (K
Y a
(K
Y
,k),k).
To see that (2.41) actually provides a complete recursive definition
of Fjjj.it is sufficient to notice that if Y is a minimal subgraph of
U(i.e. c
f
-0), then : ^ ( K ^ k ) - I^(K
Y
,k) - I,.(K
Y
,k).
So the recursion works by inclusion (of subgraphs) inside the forest,
starting from the minimal subgraphs and ending at Y-G.
Remarks.
i) Due to the definition of t ; if Y is not a normalization part
(d(Y) < 0), it yields no ean&UJbation to the formulae (2.43), (2.44), (2.45).
ii) Formulae (2.43), (2.44), (2.45) are the exact analogs of Zimmermann's
integrand R (see l) ) although in the latter the sets of variables (K^.kT)
instead of (K',k) had been used in the corresponding expressions (namely
for the operations s and t ^ ) ; however, our last remark at Che end of
definition 2.g.,shows that this slight change in Che presencacion causes no
perCurbacion in Che algorithm which defines R
Q
(o r f O .
From the analyticity of H
v
, H^ ' in the regions , . (Definitions
2.b.i.) we can now derive the following propercy :
52
Lemma 2.3.
(rN)
R(K,k) is a (real) analytic function on /
K
^ ) This comes
from the fact that all the operations (including the linear mappings s* .
and the Taylor expansions t "*') involved in the above definition of R
preserve the analyticity property.
The above analyticity property allows us to take ^/\as an
integration contour (inside the analyticity domain of R,(K ,k)). In order
to show that the Eenormalized Convolution product defined by (2.19) makes
sense, we now have to prove the absolute convergence of this integral ;
this will be possible if we ensure that the conditions of the Power
Counting Theorem of Weinberg (lemma 1.2.) are satisfied ; this is the
scope of the two next sections.
53
3. THE "CO M PLETE FO REST" FO RM ULA FOR R HI TH RESPECT TO A
'j
NESTED SET & OF SUBSPACES OF E *
m
.
(k)
As a first step toward the proof of the integrability of the function
R
G
(K,k) with respect to k it is necessary to transform the original expression
of R
Q
(see definition 2.h) and establish some combinatorial results which are
similar to those of j) and make an essential use of the notion of "complete
forest". Here however, we need to introduce the notion of complet forest not
only with respect to a given subspace S of E,,. (as in 7 )), but more
generally with respect to an arbitrary nested set * of subspaces of EV\ .
Definitions 3. a.
3.a.l. An internal line i of Y is called constant (resp. variable)
in a subspace S c E., . relatively to Y. if the corresponding subspace
Y * ' rN v
{K
1
0, k S} of y is contained (resp. is not contained)in Ker X:
(see definition 2.d.ii).
(K
'
k )
3.a.2. A G-forest IT is called complete with respect to a subspace S
i) G U
ii) For every Y U the internal lines ol the reduced graph Y(U) are either
all variable in S relatively to Y or all constant in S relatively to Y-
3.a.3. For a G-forest U which is complete with respect to S one defines
~~~~"~* S S
two sets of subgraphs H (U) and 8 (U) by the following conditions :
) W
S
(U)i s the set of all Y U for which all the lines i n ^(Y(U)) are
constant in S relatively to Y,. If Y W ^UJa l l the lines inrf(Y(U)) are
variable in S relatively to Y.
b ) B
S
(U) is the set of all T V satisfying :
T ?W
S
(U); T is a maximal element of 'J(Y), with Y W
S
( U ) . One then has
the following property of R, similar to thi>. one proved for R. in 7) :
* in french : "drapeau".
54
L emma 3.1.
Being given a subspace S, E admits the following corresponding
expression :
(S)
*-l*v
C 3
-
U
in which-the sum extends to all the forestswhich are complete with respect
to S and where
Y; * ' and the following auxiliary functions Y* are defined recursively
G
(for every y <= G) by
(3.3) '
and
\
l
r
n
a]
S* f,
a
*a
\
f
r
0
- t
d
< ) i f Y 6
B
S
(U)
f
v
- t
d(Y)
if Y t
B
S
(U)
(3.4)
We omit the proof of this lemma which goes exactly as that of l) for R ? :
indeed, it amounts to regroup the various terms of (2.40) in clusters, each
of which is associated with a complete forest (with respect to S); it will
be generalized in proposition 3.2. below.
Definition 3.b.
Let U be a complete forest with respect to an arbitrary subspace
S c: E .JJ which is fixed ones for all in the following. For a given graph
Y 0 we define the integer M
(
'(Y) by i
M
( S )
( Y ) - r- X mCOJ (3.5)
UU(Y)
U?W(U)
I n (3.5) the summation extends over all u e U such that y <= Y and U t W(U);
55
So two cases occur :
- if Y I W (U) , then : M
( S )
( Y ) - rm(Y.(U)) + I M
( S )
( Y J (3.6)
<S),, . r (S),
l<a<cj,
- if Y e W(U) , then : M
W
( Y ) - I M
W
( Y J (3.7)
Remark. It can easily be seen that the dimension h(S) of S satisfies the
inequality :
M
( S )
( G ) h(S) (3.8)
Definitions 3.c.
3.C.I. We now consider an arbitrary set of subspaces S^' c E ] ? ,
j - 1...L which satisfy : S
( 1 )
c S
( 2 )
=...<= S
( L >
.
We denote by f? this nested set of subspaces :
$ - { S ^ <= E
}
; S
< j _ 1 )
c S
( 3 )
I < j - < 1 } (3.9)
3..2.
Let 'U(S
lJ
'') be the set of all the complete forests of G with respect
to the subspace S
t 3 j
$ (J-1...L). We can then show the following :
Starting from any forest IT ' U(S
l
'), it is possible to associate with
it a unique minimal forest which contains it and which is complete simultaneously
with respect to all subspaces S^-" J : this will be achieved by a recursion
procedure which makes use of successive completions as follows :
A. By taking into account the precise prescription of 1J. we can
complete with respect to s'
J
' an arbitrary forest U
J
in t l ( S
t j J
) by the
formula :
* See in 'j)* lemmas (4.1) ...(4.5).
56
<j<j
+1
> - u
( j )
v A
( i + ] )
( U
( j )
) (3.10)
We recall briefly the definition of (U
J
') as given in fz / . He consider
et T
J +
^(t P
J
0 of subgraphs in U
J
such that at least one line in
)) i s C o n s t a n t i n S ^ (rp"lat-iwlv fn V l .' w Hpfl'n f-h coc cf *. T P-J'
nstant lines in S
t J
' of all tl
set ^ * " ( l ^ ' ) 1 i then defined by
the set W
J + J
( t T
J
0 of subgraphs in U
J
such that at least one line in
a ^yfl j^)) is constant in s ^ *
1
' (relatively t o p ; we define the set s(y',IT
JJ
)
of constant lines in S
t J
' of all the subgraphs y e W
J
0 T
J )
) ; the completion

( j + 1
V
j )
> -
I cG ; r i u
( j )
; 3 a ^ f f ^
+ I
' ( U ^ ) such that r is a connected |
component of y\ s(,tT
J
') (3.11)
It has been shown in ^ that U
( j + 1 )
e ( S ^
+ 1 )
)
Lemma 3.3
i ) Every T a A
C 5 + I
W
j )
) sati sfi es :
i.a)
T e
B ^
+ 1
VJ
+ 1
> )
i.b) cfif(D
J
) only contains constant lines in S ^ relatively to x-
i i ) U
( j + I )
given by (3.9) sati sfi es : V
li
*% 11 ( S '
C j )
) , and
jfi*(U*) <= W'i'CU'^ V .
Proof
i.a) Let T S f Y s t y .U ^ ) , with y e W
C j + ,
' ( U
( j )
) . We note that all the lines
of J\S(/,U ) are variable in S
l J +
' and by Definition (3.11) r is a maximal
subgraph of/in U ^
+ I )
.
i.b) The inclusion S
(
^ C S^
+1
^ implies that for each y e w
(
J
+ 1 )
( U
(
-
3 )
) all
constant lines in S^
J
' of the set S(/,D ) are also constant in S^-
1
' ;
so , if we suppose that there is one line variable in
with T maximal subgraph of y\ 5(Y(u-^
3
) and f ^ U ^
5
;
then ./(/(U^
5
) should
contain simultaneously variable and constant lines in S
J
; but this is
contrary to the hypothesis that U ^ f e It( S^). Consequently
0
/(fU
(
-'
+ 1)
)J
contains only constant lines i n S^ .
This set can be empty when V^h lUs'^*
1
^
Notice that here the set W ^
J J
(U^J^) has a different meaningfrom that of
Lemma 3.1 because U ' is not complete in general wich respect to sfr"J
57
ii) From i.b) and the hypothesis that U
J
S U.CS ) we obtain that For
every y U ^
+ 1 )
:
either dC(y(V ) contains only constant lines in S ^ '
or Z(V (u
( : i + 1 )
) " variable " "
We conclude that U
J
satisfies the conditions (cf . Def. (3.a.2) for
being complete with respect to S *' and in view of i.b) t
r e i "
+ 1
> (o<; implies
x&
w^ ( u
(
J
+ , )
.
B) S t art i ng from an arbi trary V G LAS .) we repeat the above completion
procedure successi vel y for a l l subspaces S S . . . S by recursi on.
We suppose that we have constructed the f ores t U ' , which i s complete wi t h
respect t o S , gi ven by :
U
W
W > U < UU
(
' *' > ( U
( f )
) (3.12)
! - i
and satisfies : h.l) V W l W ^ ) {-1...M
re6<
f
*'>(U<
f+,)
) and 4 ffl/
M )
;; contains
only constant lines in every S ^ C s "
+ 1 )
c= 5^(S)
h.2) U < % li(S >) S
(
^ S'*
We then complete U (given by 3.11) with respect to S ' using formula
(3.9). We finally obtain :
U<
i+1
> -
U
< U ( A
a + l )
W
0 )
(3.12)'
l-l
B y the recurrence hypothesis and L emma 3.3 we also see that U satisfies :
h.l for l - 1... and h.2 for S
< ! )
S S
( i + 1 )
.
So we have proved that if (3.11) h.l) h.2) hold for ji they are also
verified for j-C+1. We apply this result for j-L and obtain the following
statement.
L emma 3.
Let F-
l ?<,... S i 8,c'... CS<
L
>j .
58
With each forest U < it (S. ) one can associate a forest I T ' given
by :
^ . ^ " t u /
1 1
(0
{
") (3.13)
-1
Where each U
(
" (resp J t
f f + 1 )
( u "
5
) is defined recursively by <3.i0)
(resp. (3.1 I )).
is called the completion in forest of U^ ' with respect to a and enjoys
the following properties :
1) VreA
a+l
'> <u
(/)
> t-u..i-i
TSB ( ff
u ,
' ; (3. 14)
conta ins only constant l i nes in every S
3
'
such tha t j < I .
2) V S < " e f , u W

t o
( / !
) . ( 3. . 5)
Remark
We note that in general the sets .4 \ u ) ^ 1...L-I in (3.13) are
non empty ; so there are in general several forests u, g Zt(
s
) which
admit the same completed forest I T ^ fe l&S f V l"l...h. M ore precisely
all these forests U; { result from applying the above recursion procedure
(I ) ,) ( M
pletion of a minimal forest ^ ^ U(S ') with respect to all subspaces
s'JJe^ " with j-2...L and following the properties (3.11) (tf.I) (J?.2)
they satisfy :
of com-
%eU^ h and u g
B
g S D g .
T his enables us to classify all the forests U '< t(
s
) into equivalence
classes defined as follows :
A class C is the set of all forests U ' in tt(S ) which have the same
completion forest B with respect to
59
Definition 3.d.
Let ./ /s^ -
1
' ; l^ j iL f be a nested set of subspaces of E^f|. , as defined
in 3.c.l.
3.d. I - A forest U of G is called complete with respect toy if it is com-
plete with respect to each subspace S^
J
i n / , namely :
V j , ]< j ^ L : U ^ ( S
( j )
)
Following definition 3.a.3., we can associate with such a forest U , the
sets of subgraphs H
s J
0 0 , B
S J
(U) (for every j ^ L ) : for simplicity
c(j) /:\ qCj) (i\
we shall put : W (U) - tr
J ;
(U) , B (U) - B
w
( 0 ) ; all these sets
are subsets of U .
which belong to at least one set B ^
&
r
m - U B> (U) <3.16)
3.d.2 - If U is complete with respect to J , we call J O (U) the set of all
subgraphs Yin U which belong to at least one set B ^
J
'(U) (sor some integer
j ^ L ) ; namely :
ij S L
Lemma 3.5
L et be a nested set JS^ ; 1 <i <L j and U be a complete forest with
respect to
i) Let V-eU j if there is some integer i<L such that yB
( l ,
(U) ,
then :
V [<-i : ye W
( P )
(U)
ii) if_ ye$(v) and if i is the minimal integer (i-~V) such that
r e B ^ ^ U ) . then !
V i> i , Y ^ v
t f)
(U>"
60
Proof
i) If ^ 6 W
( l )
( U ) ,/ has only constant lines on S
( l )
, and since S ^ ' c S
( l )
for every <i , V also has only constant lines on all these subspaces S '
(see remark ) : this proves i) .
ii) Let y S B ^ ( U ) ; then ^ O (0) and in view of ), W
( 1 )
(U) for every i .
Moreover there exists a graph -t in V (U) such that r be a maximal sub-
graph of y' in U ; but in view of i ), we also have :
V f< i , f'
e
W ^'CU* (3.17)
Then if i is the smallest integer such that f<B (U) , V cannot have
variable lines in any S ' such that < i (indeed if i-t were the case, one
would have y = B"'(U) in view of ( )). So necessarily ySW "'(U) , for
every (<i .
* T
3. d. 3 - For every Yll we defi ne the subset OlCj as f ol l ows :
^- [ s ' J ' f xf W<J)(U)J C3.18)
We can then prove :
P roposi ti on 3.1
Let V ; a l . . . C y be the s e t of maximal subgraphs of Y i n U(y) . Then :
,5*
a) i f V e &< V) Y
a
e JD (U) then <r o <X.,
b) if jr. ^ S^ Oli then 4
3
^
Proof
a) I f Y . ? (0) , then i n view of Lemma 3 . 5 , i i ) , there i s an i nt eger
L such that :
j ; e B
( i)
(U)
( 3
.
1 9 )
Vht Kj^ m
( 3
.
2 0 )
V !<i Y ew
c / )
( m (3.20
St
1 1
; ' * *
61
Formulae (3. 19) and (3.20) show ( in view of Definition 3. d. 3) chat :
^ -psf, i<fstj . (3.22)
Now since J is a maximal subgraph of V in U , we conclude from (3.(9 )
that Y e w ^ ' ( U ) j Lemma 3.4,i) then implies that :
V (<i ^ eW
( P )
(D )
Then in view of Definition 3.d.3, we can say that :
^
r
c[s
a)
ef; i*t<zj (3.23)
Comparing (3.22) with (3.23) yields the desired result a ) ,
b) Let . He have the two following cases :
b.l) Ys^e?' , y
a
e w
< x )
( U ) . That means ' 0 a
n d
property b ) is
trivially obtained.
b.2) Let I be the minimal integer ( 1 < /L ) for which J ^ W ^ " ( U ) .
Lemma 3.5,i) implies that :
and .from Definition 3.d.3 , we /e :
Let us show that necessarily if j > , then y ^ i r - ^ ( O ) ; indeed since
Y is maximal in Y and since Y has variable lines in S. (in view of (3.24)),
la f\
a
t '\ 3
the relation Y e
l w
( U ) would imply Y B
U ,
( U ) , which is not possible
since we assumed that Y ft > C ) So we have proved that if S
l J
^ Cy
then y p " ' ( U ) , or in other words S
t J
^ c r
r
, and b ) is thus pr o v e d.
3
Definition 3.e.
We shall call the set of all forests U which are complete with
respect to the nested setj~~
- 62 -
He shall now characterize, for every forest D in LoF) the class C.. of
all forests U in U(S ) which admit U as their completion forest
with respect .to .Fin the sense of lemma 3,4. For every Vlltr)> the class
C_ is certainly non empty since 0 itself, as an element of 2/(S^ ' ) ,
coincides with its completion forest, and therefore belongs to C
|(
.
We can prove :
L emma 3.6
i) For every forest O in the associated class C contains a minimal
forest which is :
U
min " < P ^
U
U B
0 )
(U ) (3.25)
ii) L is the set of all forests IP ' in. (S
(
') such that :
"min^
u C

C D ( 3
-
2 6
>
iii) For every IP ' C^ , one has :
B C O Q J C O J . B
( 1 )
( U ) (3.27)
Proof
We shall only indicate the idea of the proof which is the following.
I t is possible to show by recursion over decreasing j that the following
set of forests :
j U, - (UST l) (J ( L) B
C )
(U)) ; \ i j L ] (3.28)
satisfy the properties :
i) V j ; 2 . j i L i U
5
is the completion forest of U. . with
respecc to S .
ii) V j ; l ^ j S I j U. is complete with respect to the nested
set [ S
U )
; I < I <i } and :
Vi<i : B
(t>
(U.j) - B
( e )
(U) (3.29)
Note that for j-L, II, > II which satisfies ii) , and that for j*l ,
one obtains U, u_._.
1 min
T he purpose of the end of this section is to prove the following property
for the renot-malized integrand R_ :
- 63 -
Proposition 3.2
Given any nested set there exists a corresponding
expression of R which is defined as follows :
R
G
(K,10 " H , , Xa(K.,k) (3.32)
E ach term X^ (for T ielltf)) is given by :
S-, - <i-t ) o)
< 3 < 3 3 )
and g ' is defined together with the sjt of auxiliary functions jYy ' jfyeUJ
by the following recursion formula ;
f > - i
f
TT
( D)
< >
( 3
.
3 4 )
where T1.L "j / . >
1
i
a
^
c
/ | denotes the set of maximal subgraphs ofjr in U,
(1-t
a
) if ^^O T (3.35)
if y
a
e S w (3.36)
L et us consider the expression (3.1) of R_ in terms of complete forests
(II
with respect to S :
and where
!
f(U)
a
= (1
J(U)
a
-
<
- t
Proof
R<3
" uOfel S
1
X , (3.37)
CD, V O
In (3.36) we can regroup the terras X ... which belong to the same classes
I T '
Cj, ; in view of lemua 3.6, we then obtain a summation over all Che forests U
in U(T ) and (3.36) can be rewritten as follows :
R_ - ^ _ ^ (2 - X, ) (3.38)
G ** \
( 1 )
(!) J
u-fec
In order to obtain the desired result (3.32), we shall prove that
V VG.U.&) : %, - X X ...
(1) IT
;
0
U ;
< C
- 64 -
or equivalently (see ( 3 .0 and (3.33) that :
~(U) , X v
< l )
>
(
3.39)
f ( U
C l )
) ( i l l
To show t hi s , l et us repla ce the set of functions < Y.; ' j r u
v
M of
f~fuOh 7 < ?
J
lemma 3. ! by the following set I Yi ' ; yU J :
Y J-eU , Yj, - I
r ( u )
I l ^
( 0 )
s
a
f
a
Y^ (3. 40)
- t ^ i f Y ^ and y ^ ' V ;
? f
( 0 )
- f d -
C
d C V
if f
a
U<'> and v ^
1
V
1 }
) | ( 3. 4. )
1 if Y<h and ^ U
It can be checked by recursion over the subgraphs SU
>
tha t :
Y ^
1
' , Yf
0 )
> - Y< ,
D C 0
> , and in pa rt i cul a r : Y<
u 0 )
> - lf^ (3. 42)
Remark
In view of lemma 3.6, the relations :
J
a
^(r e sp.e ) B
( 1 )
( U
( I )
) can be replaced by Y.&resp.e) B
( 1 )
(U)i n ( 3-4-1),
To prove (3.39), we are then led to prove by recursion over the subgraphs
r e U that :
Vya, Y). XI
5
<D
0 )
) (3.43)
' U
U ;
( v )
7
where the U (f) which occur in the summation denote all the forests in
U(y) which are of the following form :
O ^ C j O - TJ
( 1 )
D U(}0 for at least one u ' ^ ^ .
When Y is a minimal subgraph of U, there is only one forest
U^^(-y) \J( p -[y] , and both sides of (3.43) are equal to I
f
.
Let us now prove that if (3.43) holds for all the subgraphs y s H , ,
(n 'a i
then it holds for f itself . Every forest U ( p , determines in a unique
way a set of forests flP
l
'(]r ) ; ^e &l ?'? . a
n d t h e
forests
65
B
y>(u) - B ' W ^
u )
, &
y
o -J8*cu>fl%<
u
> j ^o - u
(
') (
r
) n4
u
\ j f
(
u)
i n such a way that :
u<%) -f U, , 0
( , )
< v ) U8^
I )
C D) ^OTUj ^ ) (3.44)
From lemma 3.6, we deduce that when IK ' varies in ( L. ,Jt]S varies in the
set of all subsets of .L(tr)\B.(U).
So we can write in view of (3.40) (3.4I)(3.44) and of our recursive assumption
(3.43) for the subgraphs y S.fnt :
.jf-f - i
m
j -fr s-
a
(-t
d(ra)
)r7> | x
B
>(
r )
' TOO Va ^4
( u ) 4 J
r
a
J
f
j [ s *( l - t
d ( r a
) Y?' I x (3.45)
[ "K" s * ( - t
d < | r a )
w|
x
f s' OJ:
' ^ '
a
r a i ' L ^>
s
*
Y
r
a
7a<^<
U
>
[v
a
eBC O(u)J
B / not i ci ng that the summation 2 - can be repl aced by
s " ( 1- t
< ,
'
a J
) , anoT by usi ng the recursi ve de f i ni t i on
of Y
T
(U) ( see formulae ( 3. 34) , ( 3. 35) , ( 3. 36) ) , we f i nal l y obtained
the desi red res ul t ( 3. 43) .
q. e . d.
66
4. I N T E G R A B I L I T Y OF T H E R E N O R M A L I Z E D I N T E G R A N D R
G
Ibis section is devoted to the proof of our main theorem :
Theorem 4.1.
a) For every fixed value of K in< 5 the normalized integrand
R
Q
(K ,k) (defined by formulae (2.*3), (2.4 V)) and all the derivatives D-Jt., of
Che latter belong to a class A Y. ' whose asymptotic indicatrixa satisfies
the following condition :
sup (a(S' ) + h( S' ) ) - - 1 (4.1)
U' i KOhSt eE^}
where h(S') rt denotes the dimension of the subspace S'.
b) The integral
I
R
G
(K ,.< )d
ra
fc (4.2)
-.rut
E
< k )
is absolueely convergent. The function H "
n
that if defines belongs to the
class C
r)-
The proof of this theorem is based on the possibility of writing
decomposition formula* of the type (3.25) for R, corresponding to a variety
of nested sets T
m
{ S, e S,C..<= S -} ; for each such decomposition (3.55),
the various term 3L, with 0 e ViCF) will be proved (in proposition 4.2) Co
belong to appropriate classes
t
^ Bue since Xy is itself defined through
recursive formulae (see (3. ) , (3. )) which involve auxiliary functions
Y (Y U(7")) it will be first necessary to prove (in proposition 4.1) that
each of these functionsY^. belongs itself to an appropriate classe
Q?VV
67
Proposition 4. 1.
Let X
m
{Sj . . . S-} be a nested set of subspaces in E^* and U be
a forest in VfJ ^ ) lm< m) ,
Y
Then for any subgraph Y S U , the function L ( K k ) in formula
(3.23) belongs to a class v-i,7 ' of admissible Weinberg functions with
the following properties :
a ) S, -6
y
QS.,f) is defined by formula (3.24)
b ) i) V S^
}
S.r such that S,*' t 3 , the asymptotic coefficient
a ( S " 0 satisfies :
either : a ( S
( j )
) 0 if V u U(Y) S
0 5
? 8 (4.3)
or : ^ ( S ^ ) < - M
( j )
( y ) " 1 (4.4)
if there is at least one y U(y) wi t h S
3
a
ii) V S ^ ' E.F" such that S " '
v
t he asymptotic coefficient
c
r
"r fi)
with respect to every subspace Sy c , ' with ir(S ) S
J
, satisfies :
' (K
Y
,k) f
Y <
s
Y
) i
d
< Y ) - M
( J )
< Y > <-5)
For the proof of this statement we shall use the recurrence hypothesis that
proposition (4.1) holds for every function %, , where {y :al...c } denotes
'a
a
the set of all the maximal subgraphs of U(y). We shall need to prove the
following auxiliary Lemmas 4.1.,4.2., 4.3.
Lemma 4.1.
.T and U lltf*) being given, the function >-(K
v
,k) belongs to a
,' ' of admissible Weinberg functions with the following properties :
i) For every S
J
^ C T s u c h that S
J
? 5 ,the corresponding coefficient
satisfies :
o ^ ( S
( j )
) - 0
ii) For every such that S
J
3 , the coefficient corresponding
to every S =
r
'.
Y
such that ' ( S ) - s'
J
^, satisfies :
Y
(r
Y
,k)
T
68
<*-(S ) - d(Y) - rm(f) (4.7)
Y Y
Proof
v ^
a
v >
0
v )
From lemma 2.3 .b. we obtain that : I -(K',k) ^ i with
CT - and a- defined by formulae (2.38) and (2.39).
Y
Now it has been proved in section 3 ((3.13) ) that the set
8 -XU,.?") - { S
C : i )
6 / \ such that y t V
W )
(U)} is contained in^ a-, so
that in view of proposition . 1.2. c. we also have I -
I t remains to check (by using formula (2.39) and the definition (2.41)
of d(Y)) that :
- Y S <= I with ir(S
v
) 9 (U.f) = a- , one has :
Y
(K
Y
,k) Y T T
OyCS) " I , "
v
+ , X , u
{
- d'(Y) - nn(Y)
I<a<c 'a ' !a<c 'a
(4.8)
f(S)
V- S. - S
(
^ e/ ' (l
i
<j<m), such that S
( j )
t 3 ( V,f), one has
(4.9)
The latter result comes from the fact that for such an S " ,
YEW. (U), which means that
S
( i )
C : ( f l ,
1 /
teUlK ,D , , Ke r ^)
<v./K>- U / j
v
.(iZ> U Z )
X
therefore formula (2.13) applies t o
a
- ( S
C j )
) .
69
Lemma 4.2.
For every maximal subgraph Y a =l ...c of UCY), with Y 35(U) the
function S*(; - t ^ ^ J ^ belongs to the class fe f * * V of admissible
Weinberg functions with the following properties :
i) V S^' ff with S
1
-
3
-* 6 <J the asymptotic coefficient corresponding
to every subspace S c with ir(S) - S
V J
' satisfies :
Y .-(K
r
,k) Y

y
C a )
( S
y
) d(Y
a
) - M
C j )
( Y
a
) (4.10)
ii) V S
< j )
& with S
C i >
? *
either : o ^ ^ C S ^
5
) - 0 - - M
( j )
( Y
a
) if P U(Y
a
) S
( j )
? ff
u
(4.11)
or : ^'''(S*^) <, - M
( j )
( Y
a
) - 1 if 3 at least one U U(Y
a
)
with S
( j )
t . (4.12)
Remark, Ihe last equality in (4.11) is consequence of definition ( 3 .D
which yields K
( j )
( Y
a
) - 0 if V <= Y, V Wj(U), i.e. S
( j )
? S
p
.
Proof
sA
}
From the recurrence hypothesis we have that Yy Q
4 N
; from
(4. 3 ) , (4. 4 ) , (4. 5 ) the corresponding coefficients OL. satisfy :
a
- for every : S c with ir
a
(S ) - S
K3J
, and
Y
* ~( K
Y a
, k)
S
( j )
e ft; i o (S ) - o ( S
C j )
) < d(Yj - M
( j )
( Y
a
) (4.13)
'a 'a 'a
Y
a
a a
for S
C i )
t 8
'a
70
( a
v
( S
( j )
) ' < - M
( j )
(
Y
) - i
" a
a
(4.14)
if 3 a t l e a st o n e n U(y.) w i t h s ' ^ o
a ( S
( j )
) - 0 if V y D( y
a
) S
( i )
tf rf
(4.15)
M o r e o v e r t he pr o pe r t y y & ( U ) yi e l ds b y pr o po si t i o n (3.2)
'r

\ <
4
-
I 6
>
o-^-K-
W e n o w a ppl y l e m m a 1.6 t o t he fun ct i o n [ 1 - t ' J.^L ; i n v i e w o f pr o pe r t y
(4.16 ), S (r e sp. 8 ) ca n pl a y t he r o l e o f T (r e sp. a) i n t hi s l e m m a a n d
w e o b t a i n t ha t :
d (Y X
C a
'^Y*
(
1 - t *yy ^ Cv
a
where the coefficients S of t hi s cla ss
/ a
W
Y
Y
a
a re specified as follows :
a) If S^ 8 property a) of l.enma 1.6 yields :
l N
Y
Y S c * withir( S ) - S
( : l )
,
Y a
<K
T
\ k)
(S ) - 0 ^ , ( S
( J )
) - O y (Sy ) (4.17 )
01
* " ' 'a 'a
M o r e o v e r b y pr o pe r t y (4.16) a n d t he hypo t he si s S
i3i
S w e ha v e S
t J J
E
'a
t o o ; so , w e i n se r t (4.13) i n (4.17) to o b t a i n :
\ (S
y
) d ( Y
1
) - M
( j )
( Y
a
) (4-18)
b) If S^ ? V, but S^ e
y
, then property b) of lemma 1.6 yields
71
a (S
( j )
) - a
Y
(S
( j )
) - d(y ) - 1 (A.19)
Inserting (4.13) in (4.19) we obtain :
6^, (S
( j )
) < - M
( j )
(Y
a
) - 1 (4.20)
c) If S^
1
' t $ property c) of lemma 1.6 yields
5
Y
.(S
V J
')-0 ,(SV')
( 4
^ , )
'a
We insert now (4.15) (resp. 4.14) in (4.21) to obtain :
s (s
( j )
) - o if v v e u(y ) s
( j )
? 5 (4.22)
.T a "u
a
H
cL (S
( j )
) <_ - M
(
^ (Y
a
) - 1 if 3 at least one u U(Y
a
)
* with s " '
e
5
Y
. >
(
dC(
a>\
By applying, lemma 1.4 to the function I I -t lit, and to the mapping
(
rN riL ..
N
x
*V a d(Y
a
)\
from Y to 'a \ we obtain that s f I - t Y is an
-0?.k) V * k / \a) '
Y
*
(
r ^ ^"t y
admissible function on in the class Q. such that :
(*\k) '
* ,(a)
V SL c Y : a<
a J
(S } - 5
Y
(Sy ) with S - s (s ) (4.24)
' (K',k) Y Y 'a ' Y
a
a Y
Since the condition ''(S ) *(.$ ) holds (in view of formula (1.16)) it i
'a Y
easy to check that' in view of (4.24) the conditions (4.13) (resp. (4.20),
(4.22), (4.23)) entail property (i) (resp (ii)) of the L emma.
q.e.d.
72
(4.25)
L emma 4.3.
For every maximal subgraph y al...cl of \J (y) with Y ? S(U),
*( W o
( a < a ) , 3
v>
the function sl-f J Y^ belongs to a class Q, > of admissible
Y
Weinberg functions whose asy "otic coefficients satisfy the following
properties :
i) L et S
( : i )
e 5
v
; V S
v

Y
with ir(S ) S ^
T T
(K
Y
,k)
T
either : a
( a >
(S .) - d(
Y
J - d(y) - K
(i)
',yJ
y j a a a
if v u e U(Y) s
( j )
?
u
or
Y
( a )
( y i d ( Y
a
) - M
C i )
( Y
a
)
if 3 at least one V U(Y) with S ^ 5 . (4.26)
ii) L et S ^ ? 3,
f
; then
either : a^ ^ (S
(i)
) - 0 - - M
( : i )
(Y
a
) if V M S U(Y) S
( j >
? 3^
(4.27)
or : a
( a )
( S
(
^ ) - ^ ^ ( Y ^ - I if 3 at least one V tf(Y)
wi t h s < J >e 8
Y ( 4
_
2 8 )
Reaark. For the last equalities in (4.25) (4.27) see the remark at the end
of lemma 4.2.
Proof .
(ct ,6 )
From the recurrence hypothesis we have Y w . M oreover the
Y
a
Y
a
property y 53W yields by Proposition (3. 2) :
L = 8 (*-29)
"a i
73
We apply directly lemma 1,5 to the function [-t JY ; now in view of
(4,29) the role of the set T (resp. o" ) is played by 3 (resp. a ) and
we obtain that :
a
I-t JY e (X.K wi'
1
<* defined as follows :
* ' Y a Y 'a
'a
a) If S
( j )
S 3 , S
( j )
S , then Y S with TT (S ) - S
( j )
Y Y
a
Y
a
a Y
a
we have by property a) of lemma 1.5 :
3
Y
(S
v
) -a (S ) (430)
'a 'a 'a 'a
From the recurrence hypothesis (4.5), the eq. (4.30) yields :
5
V
(S ) < d(Y.) - M
( j )
(
Y
J (*-31)
Y
a
Y
a
- a a
b) I f S
( j )
3 S
< j )
t S , then V S
v
with IT ( S
v
) = S
C j )
;
Y Y
a
Y
a
a Y
a
-
by property b) of lemma 1. 5 we have :
cL ( S
v
) - a . ( S
( j )
) + d (
Y
) . ( 4. 32)
Y
a
Y
a
Y
a
a
From the recurrence hypothesis (4.3 ), (4.4 ) the eq. (4.32) yields :
either : 5
V
(S
v
) - d (
Y
.) if V u U (
Y
.) S
C j >
5 (4.33)
Y
a
Y
a
* a u
or : 8L (S > <,
d
(Y
a
> " M
( j )
( Y
a
) , if 3 at least one u 6 U (
Y
)

a
with S
( j )
E S ,,. (4.34)
c) If S ^ ? S by property c) of lemma 1.5 we obtain
\ '
(S
( j )
) - a , (S
< j )
) W- 35)
Taking into account (4.25) wc also have S
J
? o i so from the recurrence
Y a
*rv-.
.., / ^ g ^ ' ^ h ^ - %."
;
th* above equation (4.35) yields :
#- f r R^f f e
J : , ,
*
; &
'
#
"i*-'V .u 6 OCT) S
0
^ (4.36)
", ;& - S W
( j )
) < - , *^M >- - 1 . if 3 least one u tT(y), (y ^
Y
J
; vith S^'e S . (4.37)
%~
;
; * ' /
d
<Y
a
>\
Wit-nov apply lemaa L.2i to the function ( -t ) Y and to the mapping s ;
as at the end of 'Che proof of lemma 4.2 , we easily check that the property
i) (resp. ii)) of thj,present lemma is directly implied by the above
inequalities (4.31), (4.33), (4.34), (resp. (4.36), (4.37)) if one again
defines the class Ci .
N
T
' by (4.24).
Proof of ProDosition 4.1.
Ve apply the results of lemmas (4.1), (4.2), (4.3) to each of the
factors in formula (3.23) :
\ " *Y "
S
a
f
v \
( 3 l 2 3 )
' ' Kot<c
T
a 'a
(a ,&)
By proposition 1.2.b, we obtain that : 1 6 CL ^ *
w h e r e t h e s e t o
asymptotic coefficients a(S) is precisely obtained as follows :
i)' V S
- J
' 6 ir with S
t J J
f. S , we take into account properties (4. 6 )
(4.11) (resp. 4.12), (4.27) (resp. 4.28) and proposition 1.2 b , and this
yields :
<S
Y
(S
( j )
) - 0-(S
< j )
) + Z a ( S
( j )
) - 0 ; if V u S U(
Y
) S
(
^ (t S
(4.3)
or respectively : a (S
(
J>) < -Y M
C j )
( v ) - 1 - - M
( j )
( v )
I a
a
if 3 at least one v (J(y) with S " ' 6
1
> e 6
(4.4)
75
For writing, the equality at the right hand side of (4.3),(4.4) we have made
use of definition 3.b and of the fact, that y w '
J
'(U).
ii) V S
( j )
G & with S ^ 3 from properties (4.7 ) , (4.8),
(4.25), (4.26) and proposition 1.2.b we have that V S e
r N
Y with
Y -(K.k)
(S ) S
< j )
:
Y
(^(S ^ - Z a^ (S
y
) + a-(S
Y
) < Z d(Y
a
) - Z M
( j )
( Y
a
) + <HY) - raff)
(4.3 8 )
Let us use the relations :
d(Y) - d(Y) + d(Y) (2 A2)
a
a
M
C j )
Cr) - 2 M
( j )
(Y ) + rmCy) ( 3. ) and H^'CY) - S M'"" ( Y. ) + rm(Y)
a
a
(the latter trivially follows from definition (3.b) since T ? W "'(U)).
So, (4.38) yields finally :
O
v
(S ) < d(
Y
) - M
( j )
(
Y
) <*'
5
>
q.e.d.
Proposition 4.2.
For every U c VLtf), the function
^ -(l-t^ Yfe (3,26)
and every derivative ,..>
D
(K)
x
0 *
X
U
belong to a class A^r of Weinberg functions such that :
V S
( j )
f , a
( u )
( S
C j )
) _< - h
< j )
- 1 (4.39)
here h ' means the dimension of the subspace S " .
76
Proof
When proposition 4.1 is applied to y - G , it yields :
(ct
G
,CT
G
)
Y
&
0j.
K
with the properties :
i) V S
( j )
6 such that S
C j )
5
G
:
a
G
( S
( j )
) - M
C j )
( G ) - 1 (4.40)
because there is always at least one p Z U(G) such that S ^ 5
(since S
( j )
t {0}).
ii) V S ^ y such that S
( j )
CTg and V S c jfL
fc
. with ir(S) - S
a
c
(S )
:
< d(G) - M
( j )
( G ) * (4.41)
We can now apply lemma 1.6
1
to the function X expressed by (3.20) (the
cases i) and ii) correspond respectively to the cases b) and a) of lemma
1.6');we then obtain that there exists a class A^ which contains JL
and all the derivatives D 3L, and which satisfies the conditions :
V S
< j >
T , a
( ( J )
( S
< j )
) <_ - M
( j )
( G ) - 1 (4.42)
Moreover taking into account definition 3.b., we have : M
J
(G) >_ h
J
where h " is the dimension of S^
J
. It then follows from (4.42) that :
(i).
V S
< j )
f , a ( S
C j )
) - h
< j )
1
q.e.d.
77
Proof of theorem 4.1.
Let {L.,...,L~} be an arbitrary set of independent vectors and W
an arbitrary bounded region in E.,.. With the ordered set (L. ,...,L-} ,
we associate a unique nested set of subspaces 3^ {S. S-} by the
following definition :
V j ; 1 < j.< m : S.. - {L,,.. ..L.}
We
then write the corresponding expression (3. ) of R :
R(K,k) - Z X..(K,k) (3.25)
G
ueuaF)
T
Then in view of proposition 4.2 and of definition l.a, we can say that for
each forest U in U C f f ) , and for each bounded region re.
4
(
n - 1
> there
exist numbers b.(U,U7)l (Uj<5) and M ( ur) such that X satisfies the
following bound : . ,
,- /
S
\,

OnCtL.-.-L,})
v K e or : l*u(
K,
-:

r
':"-
n
j
+ c
j ' - V ji
n
i
provided that : V., n . >_ b -CU^a n d C W.
Let us then put :
M(UJ) - M..CU9
ueu<3=)
b.(uy>- sup b.OJU*-)
1
uencf)
J
y
(4.44)
From (3. ),(4.43), (4.44) it follows that :
V K E ur : I R -L K, E L . n,...n. + c)| < M (K) n n."
j -1
(4.45)
78
provided that : V., n. .> b . (uT) and C e S.
let us cow introduce the class Ar whose asymptotic indicatrix a
is defined as follows ; for ever; subset S in E^j. with dimension h(S),
one puts :
a(S) - - h(S) - 1 (4.46)
Since formula (4.45) has been.established for an arbitrary set of inde-
pendent vectors {L ,...,L-} and an arbitrary bounded region H in E^* it
expresses the fact (see definition 1.a) that for every K in
rta- ) ^ g/j^fcj beXongs as a function or k, to the above-defined class
A<>
n
Now, by definition, a satisfies the following property :
sup (o(S) + h(S)> - - i; (4.1)
S
and therefore R- satisfies the Heinberg integrability criterion ( 1. + ) ;
thus in view of lemma 1.2, the absolute convergence of the integral (4.2)
is then ensured.
To achieve the proof of theorem 4.1., it remains to show that the
function Hjj
e n
defined by (4.2) is infinitely differantiable e n fp
n
~'^.
This will result from the t o following points.
i) In Weinberg's proof of his convergence theorem, (whose details are
essentially reproduced here in the proof of our lemma B.I in Appendix B),
it is clear that the uniformity of the input bounds on the integrand
(such as (4.3) for Rg) with respect to the external variables (namely K)
varying in a bounded set U3", entails the following property : there exists
an ine.Q>ia.bf.z positive function g(k) on the integration space E . such
that :
V Ke uT, |R
G
(K,k)| _< g(k) .
79
i i ) E v e r y d e r i v a t i v e D ,
v
, R ( K , k ) a l s o b e l o n g s t o t h e c l a s s A . '
W u Tin
( s a t i s f y i n g c o n d i t i o n ( 4 , 1 ) ) a n d a l s o f u l f i l s a b o u n d o f t h e t y p e
( 4 . 3 ) f o r K v a r y i n g i n W " . I n f a c t , f o r e v e r y s e t ? a n d e v e r y f o r e s t
U i n U t F " ) , p r o p o s i t i o n 2 a p p l i e s t o D L a s w e l l a s t o J L , a n d o n e
j u s t h a s t o u s e t h e v a r i o u s d e c o m p o s i t i o n s o f D . . R
G
w h i c h c o r r e s p o n d
t o t h o s e o f R w r i t t e n a b o v e ( a s a n a p p l i c a t i o n o f f o r m u l a ( 3 . 2 5 ) ) .
B y n o w a p p l y i n g t h e s a m e a r g u m e n t a s a b o v e ( i n i ) ) t o t h e f u n c t i o n s
D . . R , w e d e d u c e t h a t t h e r e e x i s t i n t e g r a b l e f u n c t i o n s g ^ ( f c ) s u c h t h a t :
V . K U 5 - , | D ^
K )
R
G
( K , k ) | 1 g y ( f c )
F r o m a k n o w n t h e o r e m o f i n t e g r a t i o n t h e o r y , w e t h e n c o n c l u d e t h a t
d e r i v a t i o n u n d e r t h e s i g n / w i t h r e s p e c t t o v a r i a b l e s K i s l i c i t i n
f o r m u l a ( 4 . 2 ) , s o t h a t
r '
'( O
q . e . d .
80
SECTION 5
5 - H_ INDEPENDENT OF THE CHOICE OF T HE ADMISSIBLE SET
Definition 5.a.
5.a.I - Given a graph G and all subgraphs j( of G , let K.j. i (resp. Yt?\)
&,*.(resp. k(n\) be the set of independent intern-il momenta of G and every ^
corresponding to a certain choice (0 (resp. (2)) of admissible set of basic
internal momenta (see D ef.C). We define the following mapping of , ,
onto P., , (respectively of C/I fy^ <"t E / n A
f o r
every j^ eG) :
j> : K-*K
^ ( l ) - ^ * ( 2 > "
A k
( l )
+ B K
(5.1)
(respect, p : K
s
- K
*<o "^w) "
A
**co
+ B
*
K
*
V
*
C G
>
( 5
' ' '
)
Here A and B (resp. A
1
', B*) are L x L , n x L (resp. l"'!l
l f
, n*x iX )
matrices with constant coefficients.
For G and for an arbitrary subgraph V cG the momentum associated with every
line ifiS ( l*y) we define the following transformation properties :
(1)
rtJ
t . (Kfc
( ] )
) -/>*i
(
f (K,!f
(2)
) V ' ^ (5.2)
resp.
^ " ^ .^ " f ^ f ' ^ ^ ) Vj #* <3.a-)
5.a.2 - In the following we shall use the following definition for
corresponding to a choice (i) i - 1,2 :
R 5i'(K, ,.,) - 2_ F
(
,V(K,fc..,) (the sum extends over all forests of G)
G (1) y U (1)
with F*
1
.
5
- (-t?^ ) Y ^ if U is a full forest
U (1J G
F
C
,V - Y
(
^ " " a normal "
U b
(5
81
The functions Y y
a r e
defined recursively by :
(i) ' (i) T T *(i) .
f
d(ya) (i)
(5.4)
If ft
and the convention is used to put t^?' " 0 if y" is not a renormalization part.
Remark
Following definitions (2. ) all functions occuring in definitions (5.3) and (5.4)
are of the form f - f (t.(K, K...)) (resp. f * - r ^ l - V ^ K ) ) ) ; we shall prove
that the operators s* p* (-t^Cjp) which act on the corresponding functional
spaces satisfy the following properties in view of (5.2) (5.2') :
Lemma 5.1
Y j
Given a forest U of G , for every jfgU(G). I5(K'K*) satisfies :
(5.5)
b) For every maximal subgraph y a - 1...C of U(^) the operators p and s
satisfy :
. C O *
<'
W
\'h ^
(2) *
(0 rif rv L (2) ,
M d ( y ) K
\
2)
.
| 4s
d
<
u
>
(5. 6)
(5. 7)
Here K *'" is a monomial of degree j{ in the variables ft" and D jj?x is
a differential operator of order :.} with respect to the variables f\',|,.
Proof
a) From Definitions 2. and properties (5.2') we obtain easily equation (5.5).
Xfl, 43. 0 3 D "3.
b) We consider an arbitrary function f,. f,,\ (l/)\) Here L/^ means the
set of momenta corresponding to all lines of Y . From definitions (2. ) it
follows that :
.(2),*. , 'a ,*. A
a
t
( 2 )
l L
( 2 )
l K , K
( 2 ) "
I
( 2 )
u
( 2 )
l K , R
( 2 ) "
(5.8)
82
and from (5.2') : f
( 2
^ (l
(
|j) - f j j (L
( l
)) (5.9)
I n v e r se l y: /^
f
( 2 ) ^ ( 2 )
) _
^ O ^ O )
3 a n d
s
a
f
( l )
a
< l )
; f
( l )
C L
( l )
) - ( 5 - 1 0 }
A comparison of (5.9) and (5.10) yields the proof.
c) From definition (5.1') we have :
p
f
(2)
( K
^(2V "
f
( 2 X * <2)' V o
C 5
-
1 0
We then use the definition of Taylor operation (-t,;V - * K
4 W
" -2y
together with the rules of derivation of composite functions :
-2-7- - S-g + i ^ - - - r - and the following relations which trivially hold :
4 f <$ f*> (K*. *, ( y , ,^ )) - p* -^ f
(
*
2)
ff*. **>) (5.12)
-v</ ;*& > *. & < $ / v i S -
f
>
<K
* *&> <
5
->
3
>
< } K
( 2 ) "T O
So, (5.7) is obtained.
We shall now establish a relation between the functions F (K, fc,.) ,
F y (K, k,,. ) . We prove first an auxiliary recursive statement.
an explicit parameter inside f .
83
Lemma 5.2
Inside a forest U -of G- , for every subgraph U , the corresponding function
Y satisfies :
Y
< D .
n
*r v <
2
) . T<i> T (
2
)
fft?-*?*?!
(2)
Here J is given recurrently by
j <
2
>- Z s <
2
> ( -
t
d W j
V
2 )
s<
2 )
\-t
dCiV
i- J5
Z K
l
A (Y. + fc J !
2 )
) | (5.16)
if Y is not minimal inside U
X
T<2>
and J
V
J' - 0 (5.16')
if ^ is minimal inside U. In this case (5.15) coincides with (5.5) (in v i e w
of Y ^ - I j ) .
In equation (5.16) we denote by Xy the set of indices of maximal subgraphs
J...Y;
C
of U(}() : Xy {l ...c
v
$ and i^ ; M *,Xy\ M y$ is a partition
of Xtf in two proper subsets M J , X^\ M^ of Xy, (Mv ) X ^ ) ; the sum has
to b e taken aver all such partitions.
Proof
He use a recurrence hypothesis. He suppose that formula (S.IS) holds for every
maximal subgraph V a - 1...c^ of U(^) in other words :
^-^HM''?]
(5.17)
(2)
where Jv is given by the analogs of definition (5.16) (or (5.16')). In
view of property (5.7) the application of Taylor operator on Yi given
by (5.17) yields :
84
(
-
C
( l ) >
Y
y
a
- / V
a
[ [
C
(2) }*U
+
P(2) , \ \
+
I *
V K
Y off * f; jg>>]
M<<<lf.> "(2)
*(1)
We apply the operator s on (5.18) taking into account property (5.6) ;
we Chen Cake the product over all Y a " l...cv and obtain :
(5.18)
4
Z)
*
3 a
|Aa Ud(Ya ) k ^ j )
a X
'
a
J
$ i f'.?']} ">
(2)
The sum in equation (5.19) yields J as given by (5.16). So finally :
T
(l ) * -(2) r-TT- *( 2) /_
t
d(Ya)\ (2) (2)"] _ *L(2)
T
(2)
T
(2)-|
Y
v 7V*; [ J * I
( 2 )
) * * J v *| . < j
q
Proposition 5.1
For a given forest U of C
a) F ^ - / * (Vjp*
l (
g
) j (
c
>
l if is normal forest (5.20)
b) , y [*. [^ i<v
2
>
+
^ ^ ^
2 ) (
vC 2>
+
^( n
( 5
,
85
Proof
a ) W e apply Lemma 5.1 to ]= G and obtain directly (5.20).
b ) Fr o m a) and property 5.7 for the Taylor operation we obtain the pr o o f.
Definitions 5.b
5.b.l - We define the analytic regulators [B.l] \J \ ! (k ) associated w i t h
every subgraph /o f G b y the following :
*
he r e V*J, -0 are large positive constants; the product is taken over all internal
lines of v" and Jc?( ftf) is the internal m o m e n t um associated wi t h . ? according
to a given m o m e n t um assignment (see De . (2. ) ) . One easily verifies that :
l i y ^
0 )
( ! ; ) - 1 (5.23)
U)r M
In v i e w of Definitions ( l .b ) one can easily v e r i fy that J r^\ Z . f 'Vl *
moreover b y Lemma (1.3 ) and the admissibility property one finds that the
'(<a,
product T T 1/ Sji belongs to a Weinberg class A J whe r e :
*S6
V S C E
( f e )
! <*,)(S) - -j.V (5.24)
wi t h 0 < ] X 2 ; the positivity of Jf results from the fact
*
S G

that : H ( , Ker . k. (k) ?f 0 .
In all wha t follows we shall suppose that 0 is arbitrarily large so that
all regularized functions used b e l o w wi l l decrease rapidly at infinity (!<->=")
and all needed integrations wi l l b e absolutely convergent b y Weinbeg's cr i t e r i o n .
Here m ^ a mass parameter a n dy?|j is equal to I on the m a ss-she l l . For
sufficiently large 2/<v "
a n a
l
v C
*
c
">
c n e
Euclidean region.
86
5.b.2 - T he "regularized" renormalized integrand R ^ ' ' defined by
(5.25)
In the following wa use also the definitions
Y,
(u),0) _ Ti fl/(u),d)
Y /SX J</0
^ f e
J )
* V *G (S-")
^
U , V > )
- T T t yfe" ' B
0
'
0
( 5 , 2 7)
*
G
We now consider the sequence of functions fv*' when m--> ; We then
state the following :
Proposition 5.2,
lin R ^ '
0 )
( K , k ) -JL(K,k> ' (5.28)
I R ^
, t f )
( K , : ) | $ ^ ( K . k ) ! (5.29)
Proof
a) From property (5.23) we obtain (5.28) by taking the lim in formula (5.25).
b) We first note that :
&*> >
^ ) "
> ) <
^ ty'*
N
z>i
(5
-
30 )
Using formula (5.25) and taking into account properties (5.23) (5.30) we obtain
q.e.d.
l R
K , )
U <
|< ^| fouipy
87
He c o n s i d e r t h e r e g u l a r i z e d i n t e g r a n d s i n t h e t w o d i f f e r e n t c h o i c e s o f a d m i s s i b l e
s e t s a n d p r o v e t h e f o l l o w i n g :
Pr o p o s i t i o n 5 . 3
J . * - " .%, : J t f >< d f e
(2)
(5 .30
L e t u s s h o w f i r s t t h e f o l l o w i n g :
L e m m a 5 . 3
F o r e v e r y Y U ( G ) t h e c o r r e s p o n d i n g f u n c t i o n J \ ' ( K , KiO s a t i s f i e s :
/
a
(^)(2) ( K ^ W ^ - 0 ( 5 . 3 2 )
Pr o o f
He p r o v e t h i s p r o p e r t y b y r e c u r r e n c e : W e s u p p o s e t h a t ( 5 . 3 2 ) h o l d s f o r a l l
m a x i m a l s u b g r a p h s .X a l . . . Cy o f U ( ^ ) . ( W e n o t i c e t h a t w h e n ^ i s m i n i m a l
i n s i d e U ( 5 . 3 2 ) t r i v i a l l y h o l d s i n v i e w o f ( 5 . 1 6') ) I n o t h e r w o r d s :
/
4? %*%\4h- <
5
-
3 3
>
W e s h a l l n o w u s e t h e f o l l o w i n g c h a n g e o f i n t e g r a t i o n v a r i a b l e s :
He r e V./i\ i s c h o s e n s u c h t h a t e v e r y I y . d e p e n d s o n l y o n t h e v a r i a b l e s
K J
a n d n o t o f t h e v a r i a b l e s f c , , . w i t h i i* j ; I n o t h e r w o r d s e a c h s y s t e m
c a n w r i t e :
1 Si .
k(i\ " *
n o n e t 0 o n e
c o r r e s p o n d e n t e w i t h t h e i n d e p e n d e n t l o o p s o f %. . So w e
/ ^ Vi &> *& -/35-*
>
ff<y
)
>
d
> TT ^ (3.35)
88
T ak ing into account the hypothesis that for sufficiently large V all
integrations are absolutely convergent,we integrate the different terms o f
formula (5 .16) ; so the essential contributions of (5 .35 ) are finally :
(2)
t V
k
\ > D ^ y
In the above integrals w e h ave performed the transformations / (2) * -
(5 .36) is obtained j n v i e w of the fact that the integral of the d erivatives
fir
w i t h r e s p e c t t o h'.J; y i e l d s f u n c t i o n s r a p i d l y d e c r e a s i n g a t i n f i n i t y ( s e e
t h e r e g u l a r i z a t i o n h y p o t h e s i s O e f . 5 . b . l ) . T h e i n t e g r a l ( 5 . 3 7 ) r s u l t e d i r e c t l y
f r o m t h e r e c u r r e n c e h y p o t h e s i s ( 5 . 3 3 ) . T h e a d d i t i o n o f a l l s u c h c o n t r i b u t i o n s
y i e l d s f i n a l l y t h e p r o o f o f ( 5 . 3 2 ) .
P r o o f o f P r o p o s i t i o n 5 . 3
a ) Let U b e a normal f o r e s t . By formula (5 .20) w e have :
f
r
^
l
h k
{ l )
- | F * " '"
2
' * k
< 2 )
* M - *
> ( 2 )
J
G
6
-
0 M 2 )
d k
( 2 )
(5 .38)
He consider the change of v a r i a b l e s K
2
)~* ?2 ) " V 2 ) U ^ U ' ^ )
5
* i
C U ( G )
as a b o v e . *"<
H e t h e n a p p l y p r o p e r t i e s ( 5 . 3 2 ) a n d ( 5 . 3 5 ) t o X G a n d t h e p r o o f i s o b t a i n e d .
b ) Let U be a full forest : B y formula. (5 .21) and p rop erties (5 .32) (5 .36) (5 .37)
(applied t o ^ G ) w e o b t a i n in a w a y analogous to a ) the proof o f ( 5 . 2 8) .
M e are now in to prove the following :
T h eorem 5.I
T h e renormalized Convolution Product given by (2.10) is the same for any
choice of admissible set of basic internal m o m e n t a .
89
Proof
By definition (2.!0) we have
H Q '(K) - | R.r/ (K.W dK i - 1,2 (5.39)
We shall, prove chac H
(
^ ( K ) - H ^ O O .
Taking Proposition (5.2) into account we obcain from definicion (5.3) :
J-F C-. HUoc, | j
o
,
d k ( 1 )
- J R < ^
S
> <
2
V , ! <
( 2 )
> dK
( 2 )
(s.AO)
Now Che absolute convergence of I Rgdif proved by theorem 4.1 Cogecher with
properties (5.28) and (5.29) of proposition (5.2) allow us Co apply Lebesgue's
theorem of analysis and obcain that :
UmJ>.
d
-/[ fV^ -/**
- H
G
(5.41)
He Chen cake the lim when
l ; J
in boch si de s of (5.40) iyly(5.41 ) and
(5.39) we finally conclude :
H^ (K) - H<
2 )
(K) q.e.d.
Acknowledg ementi
One of us (M.H.G.) is grateful Co Professor J.H. Lowenstein for
communicating her a sketch of Che proof of the above theorem 5.1. in the
case of perturbation theory.
91
A PPE N DI X
In the definition Kb of functions belonging to a class AJ| ', i t i s
assum ed (as in V ) that inequalities of the type (1.1) are satisfied for
al l sets of JtineMUy inde.pe.ndmt vectors {L. , . . . , L }. As a m atter of fact,
we need to use such inequalities even in cases when L. , . . . , L are not
i m
independent. He shall then show that this property can be derived from
definition 1.b.
Lemma A
Let f be a function in IR , which belongs to a definite class AA .
Then if {L,,...,L } is an ajihjjtluvuj set of vectors (m being an arbitrary
integer), and if H is an arbitrary bounded region in GL,, there exist positive
numbers M and b.(l < j ) , such that :
V C W , V n. > B- , l < j < m ,
j - J - -
f( f. L. n. . . . n
m +
c)| < M n
n
Cti. ,. . . i. j
( A > ] )
Rem arks :
i) As in formula (1.1), a ({L, ...L.}) denotes the asymptotic Weinberg
coefficient of the subspace S which is generated by the set {L.....L.} ;
but now the dimension of S can be smaller than j.
ii) It is useful to note that formula (A.l) still holds true when some
he vectors {L.} are chos
adds the condition : a({ 0
of the vectors {L.} are chosen to be equal to {0}, provided that one then
92
Proof :
From the set L, ,L,,. ..I"
m
} , we extract a well-defined subset of n
independent vectors {L. ,L ,L,} wi t h i. < i , < ... i < m , through
ij i
2
i
n
i i a -
the following recursive construction :
a ) i, is the smallest integer such that L. i
4
{0} .
b) Suppose that i .,...,i ,. have been determined. Then i. is the
smallest of the set of integers j > i. , such that (L. ,L. ,...,L. ,L.}
be a set of independent vectors.
c) n is the largest value of the index k such that b ) be applicable,
i.e. n is the dimension of the subspace generated by {L, L }.
1 m
For any integer j < m , there exists a unique set of real number a.,
such that :
L
3 "I * \
From the above construction, we even deduce that
a
j k
- I if j - i
k
(1 < _ k <, n )
a n d a
jk "
i f j K
\ '
so that we can write :
m n a.. \
E L.n '...i U- S A + I S - U . ...n _L.
(A.2)
We now wish to determine numbersn.'. > -O
1
as functions of n = (rii > -1^1),
in such a way that :
m n
i L.n....n_ + c - L. n\ -..n' + c .
j-i
J 3 ta
k-i *k
k n
93
By taking (A.2) into account, we are led to ,a unique solution which is
defined as follows :
!SL
nyn>
/,
+
z i\\
*k *k
as_
v."
1
x 'It
"J
^ k
"J
(A .3)
L et us now apply definition 1.b to the set of independent vectors
{L . ; 1 < k < n). T here exist constants M ' and b'. , for I < k < n, such
x
k ~ ~ K - -
Chat :
V C E W , V n - (rij n
m
) such that
n'
k
<n) >.b'
k
for ! k < n,
one has
m n *
t( l i . n n + cM W K n'.
\j-l
n
" J k-1
K
a({L, ...L. })
1 T t
(A. 4)
Now, if we put a 2 j a. ]' (note that a > ]) and if we impose
(j,k)
J k
the conditions :
r|. ^ 2a for 1 _< k n
n .>.
for every j(l <_ j <_ v
We easily deduce from (A.3) the following inequalities :
, \
W
ffcfi
- 1
T
n
i I T
n
\ in'
t
(n) < 3 n n. .
3
\
3
j - i
v
>
k
j-i
t
J
(A. 5)
(A. 6)
94
From the left inequalities (A.6), we infer that the conditions
l V(l ) b ' are fulfilled as long as (A.5) holds together with tha
additional conditions :
n. > 3 b ' for 1 < k < n .
l
f c
- . K -
From the right inequality (A.6), we conclude that the right-hand
side of (A.4) is majorized by :
m a({L. ...L.}) '
3V n n. ' '
3
To obtain this, we have also noticed that for L < _ j < i .
+
. one has
a
({L.!,}) a({L. ...L. } ), and that for j < i,, one has :
I J i , i
k
i
a( .{.
l
,...,L.}) a({0}) 0 (see remark ii) above).
To summarize, we have proved the following result :
if we put
0 j
- I for j + i
k
b . - sup (2a, 3b' ) for 1 < k < n
\ K - -
and M 3' ,
then the conditions : C W , Tl .^b . , ' .1 j i,
m
imply the
.inequality (A.l.)
q.e.d.
95
A P P E N D I X B
T A L OR R E ST S OF GR A DE D WE I NB E R G FUNCT I ONS
In this appendix, we consider Weinberg functions in a definite class
A-, on a space E , , fR? .
x
E l . (N p + q), which are infinitely
differentiable and satisfy the following additional conditions.
L et f be such a function ; we assume that each partial derivative
D f of f with respect to the variables x, with total order n k> belongs
x
(n)
to a definite class A . So, each asymptotic indicatrix a (n0,I ,2 ...)
governs, in the sense of formula (f.t), the behaviour at infinity on every
N V
subspace S of E , . of aZt the derivatives D f of f with total order n.
\XtY) x
We shall also say that is g/taded viith KUpiLCt to x, and that it is
a&yinptotlciUZy govz/tnzd by the. -sequence {o , a,,...a ...} .
Of course this sequence is by no means tinique, since in particular f
is governed by every sequence {o' } such that : V n ^ 0, a' > a (see
proposition 1.1.c)
We shall prove a general property of the T aylor rests with respect to
the variables x of these graded functions (leama B .l). T his result will then
be applied to the case of admi&iiblz. Weinberg functions (see definition ! .d)
which are graded functions of a special type (lemma E .2).
L emma B .1.
L et i be a Weinberg function in E , which is graded with respect
to x, and asymptotically governed by the sequence { a} .
T hen the T aylor rest g - (I - t, -)f of order d of f is also a Weinberg
function which is graded with respect to x ; it is asymptotically governed
by a sequence {a } which is determined as follows :
n
a) For n > d, one has :
96
V S = , R
( x , y ) ' 5
n
< - V
s )
<
B
-'>
b) For n ^ d, two cases occur :
- if S t <
y )
: <
B
'
2
>
N o
here TT denotes the projection of 1R, * onto R ; . (i.e. 7r(x,y)- (o,y)),
' N ^'
and the "sup"runs over all subspaces S' of TR, which are such that
ir(S') - ir(S).
Remark. Since each indicatrix a is a bounded function (see section l.I)
a
taking this "sup" always yields a finite number.
Proof
There will be two parts in this proof. The longer part (A) will be
concerned with proving that for every function f(x,y) which is graded with
respect to x and governed by a sequence { a } , the corresponding function
g - (1 -
E
/ O f belongs to a class A^, with the indicatrix a ci , given by
formulae (B.2) and (B.3), namely :
a(S) - a
d + ]
(S) if S c R ^
y )
a(S) - sup [a CS')] + d + 1, if S * R?,
{S';ir(S')ir(S)} d
+
I (y)
In the shorter part (B), this first result will be used to show that
each derivative D g of g, with order |v| n belongs to the corresponding
class A , with a given by formulae (B.I), (if n > d) or (B.2), (B.3)
(if n < d ) .
97
A) To prove that g belongs to AJj', we shall first treat the case
whe n x is a iigle. variable, namely : p 1, q arbitrary. Then w e shall
treat the general case through a recursive argument over p.
1. The case p 1
To prove that g (x,g ) is a Weinberg function, w e have to consider an
arbitrary set of m independent vectors {L,,...,L } in R. , x R (
v
\ and
an arbitrary bounded region W in this space. Let (R,(L, ,L-. ..L ; W ) b e
th region of E . . x 1R? , which is parametrized as follows :
(x.y) - L j v - n ^ * ... + L j ^ . . . 1 ^ + ... + L
m
n
f f i
wi t h C e W and V j _< m : T]. >^ 0
+ C
(B.4)
We have to prove that in R J ( L ., ..., L ;W) the function g satisfies a
majorization of the following form :
, /*> \, a ({Lj} )
tt
({L ...L.}) a({L,.'..L:
J"
1
. fa
P
(B.5J
provided that n . > b . ...n > b ; where b,,...,b are suitably cho se n 1
1 i 'm m 1 m
positive numbers.
We shall use the following basic expression of the Taylor's rest of f
d+1 fl d+1
d!
J
o
d+1
f
l d+1
g(x,y) - - D
x
f (Cl-t)x,y) t
a
dt (B.6)
rl f Jf>
We thus need to extend the parametrization (B.4) to the region
{((l -t )x,y), wi t h 0 < t < 1, (x,y) CR,(L,,.. .L
m
;W)>
Let L be the unit vector of
R
/ w identified with (x=I,y=0) s for every
j <:m , we can write in a unique way :
L. -=
v
. L + V. with V. R
y )
98
and C - uL + C' with C' S* .
For (x. y) in 3UL, , . . . L ;W) . we then have :
i m
( ( l - t : >x, y ) - Z L. Tij. -. r^ - Lex + C
jl
m
with x 2 y. n-. . . n + u
j-1
J J a
(B-7)
(B. 8)
and if ve choose v
a
tx as a new integration variable in (8.6), the
latter can be rewritten :
(B.9)
Following closely the method used by Weinberg in the proof of his
general "power counting" theorem (see [W]), we shall cover the integration
interval [0,x] of (B.9) by a finite number of sets, in each of which a
particular majcrrization of the Weinberg function |D f| holds. These sets
can be clustered into families according to the following recursive scheme.
One first covers [0,x] by an open set T )
u
- U I. : h is a fixed
i-E^h.
1
integer , m which will be determined below ; each I. is an open interval
whose label varies in a finite family <j, . One then successively defines
sets- L ([ , 3(L. of [0,x] for h < r ^ m, in such a way'that :
3
r
- % u %,
# r
c
^r+1 Cfor r n - I)
with 1 U I. j
(i
h
...i
t
)a
r
V
l
r
<i - U J.
* (i
h
...i
r
)ea
r
V - r
(B.IO)
(B.Il)
(B.12)
%' . U
'i, ..... J ^W
h r' r
H.
(i....i)eo- V
- , 1
r
s
Each set I. . is an open interval, from which a closed interval
or "hole" H. . i s taken away ; the corresponding hollowed interval is
V '
- 1
r
J. . - I. . v H.
V
. . i
r V
. . i
r V
. . i
r
Each finite family a , of labels (i. ...i ) is determined recursively by
conditions of the following form :
V i
k
...i . , H. . c .1 1 I. .
h r-1 ^ ...i^ i^ Ci^ .i^ ) V W
'r
(the latter obviously imply (B.ll), if (B.12) is taken into account).
Let F(v) be the integrand in the right-hand side of (B.9). From
(B.10) , (B.ll), it follows that :
[ o, x]
C
3
h
c:< u j
r
) u ^e
m
;
h<r<m
and correspondingly :
| / F ( v ) d v | _ < I / | F ( v ) | d v + / | F ( v ) | dv. (B . 13)
[0, x] h<r<m <J
r
^g
m
I n each s e t J. . , ( f or r<m) , and K. . , a de f i ni t e Weinberg
1 . 1 l . l
h r h m
majorization of |F(v)| will be produced ; then by taking(B.12) into account,
one will obtain bounds on all the terms of the r.h.s. of (B.13) : all these
bounds will turn out to have the required form (i.e. that of the r.h.s. of
(B.5)).
It is only for r-h, that $ needs a definition which is slightly
different from that of 8-).. In a) we shall define ^. and derive the
corresponding majorization of J|F(v)|dv. In b ) , we shall rest upon
8
'
*h
for the detailed definition of all the other sets *J- , ^g , ^ and indicate
briefly the majorizations of the corresponding terms in. (B.13),
100
a) Contribution of the set fr.
Let h be the integer (h^m ) 'for which :
VL i* 0 , and V. - 0 , for every j h-1 ;
h is also the integer for which each subspace S. {L.....L.}, with
j ^ h - l , is contained in E ,, while each subspace S., with j^h, is
not contained in R ..
la the argument of D f in formula (A.9), we replace v by a new
variable z which is defined by :
v - n. ...n u.+n......n z , ( B .1 4 )
h m h+1 m
u being a fixed number.
This allows us to write :
. !,
L
j
n
r "
n
m -
L v + c
J ,
L
J
n
j-V
1
( T5r) I- IVJ - ' W
+ c
m
where e(z) denotes the sign of z.
This expression is convenient for taking into account the assumption
that D f belongs to the Weinberg class A- and therefore satisfies the
following majorization (note that here it may be necessary to use the result
of the auxiliary lesima proved in Appendix A, if sets of non independent
vectors {L,,...L.,L} are produced) :
IRV
l z l
j -
h + i J
(B.I5)
101
provided that :
1 j l b , ( u ) for j l...m , j * h (B.16)
pj ->b
h
(u) (B.17)
and |a| >b
o
( a ) ; (B.18)
here M. (u), and the b .(u) (0 < _ j <^ m ) denote appropriate positive numbers
which - a priori - depend o n u ; the b .(u) a r e always supposed to b e larger
than 1.
The set of point v for which the majorization (B.IS) is satisfied is
given by plugging conditions (B.17), (B.18) into (B.H) : condition (B.17)
yields the interval :
K u ) - {v ; n
h
...n
n
<" ~ \ ' <
u
<
v
< V ' V
U + b
~ b
< u )
'
( B , 1 9 )
adding condition (B.18) yieldsthe "hollowed interval"
j(u) - i(u) n {v ; |v - \...n
m
"I
>
V i - V o
( u ) (B
-
2 0 )
Let us determine a positive number u such that the integration
interval [0,x] in (B.9) b e covered by U I (u). This is readily obtained
m (Xu<u
b y choosing u J y. + u i in fact, Tn view of (B.8) we have :
jh
J
m
y.
and i. we impose the conditions : T\. >_ 1, .. .T) >_ I (B.21)
w e obtain the inequality :
which yields (in view of (B.IS)) :
[0, x] <= [ 0, n
h
. . -ruig <= U K)
0<u<u
o
102
By applying the Heine-Borel lemma, we can extract from Cl(u) ;
0<u<u } a finite family of intervals I. - I(u.) (i belonging to a
finite set 0". ) ; we also introduce the corresponding family of "hollowed"
intervals J^ J(u.) through formula (B.20).
We then conclude that when v is restricted to vary in the set
^_ " U J^, the following majorization can be deduced from (B.15)
i < J
h
(the latter being applied to each case : u u.) :
J
h-l a < { L, . . . L. } ) m a <{L . . . L. , L + d + 1
| F(V )| dv< i m , '
3
* n
d + l
' i
V h"< V
< B
-
2 2 )
j - i J j - h
+
i J
h
x
j
2 |
d
+
i
T
<i
L
h
L
" d
+
i
L
. V i V
L
V>
x-Oi^u.+z) dz
In view of (B.I6) and (B.21), this majorization holds provided that
one has :
a > I and V j i* h , I < j < m : n-> sup (b.(u.)) (B.23)
h - - -
3
~ia.
J l
a
I t rem ains to esti m ate the dependence of the right-hand s i de of
(B. 22) with respect t o tl. , by perform ing the i nt egrat i on over z. The
dom inant contri buti on of the term l abel l ed by "i " i n the sum m ation
can have the al t ernat i ve form s : Tl
l
or TV^- irfl. j we tak e t hi s
r
a b a
h
as a definition of S-, the occurence of either form being linked with the
sign of the difference of the exponents :
i&J.
103
a* o
d t l
L , ....! . >
f (B.24)
a
'i d
+
l
C { L
l " -
L
h - ! '
L
h "
L u
i
} )
It is easy to check that :
i) if o ' > a \ c^ - a ' + d + 1
ii) if a-' < a'.- : cL - sup ("(a*. + r) + sup (] , a ' - a'. + d -r +1)1
0<r<d
L
i J
which yields a . sup (a
1
,a'.) + d + 1
From this it follows (in view of (B.24) that in all cases :
sup . .< s u
P
r a
d + I
({L L
h
, L , sup o
d + ]
({
L
,... L ^ .I ^ - L u .} ) ! + d + 1
h * h
CB.25)
< a CL,....!^})
For deriving the latter inequality, we have made use of (B.3) and of
the fact that for every number u, one has :
ir({L,... ,L
h
} ) - irCtt, ,L
h
,L - lT({Lj,.. .L
h
_j ,1^-Lu}) .
Similarly, we notice (by refering to (B.2), (B.3)) that :
- for j < h-l, O
d +
, ({Lj.. .Lj}) - o({L,.. .Lj, (B.26)
:ince then {L....L.} c ^ (
y
)
- f o r j >.h + l , 0
)J+ 1
({L
1
,...,L.,L) + d + 1 < ({^.....L.} ), (B.27)
since then ir({Lj,...,t.,t}> - ir({Lj,. ..L.}) .
104
From (B.25, B.26, B.27) we conclude that the majorization (B.22)
entails the following one :
| F(v)| d v 1^ _!I n '
hi
m a.(L, Lj})
(B.28)
for a certain constant H. , and provided that the parameters n . satisfy
the conditions (B.23).
b) Contributions of the sets * , for h < r a, and of the set %_

n
r -*m
To define an interval I. . , one considers a certain sequence U
of r - h + 1 positive numbers: {u, *...u. '. ,...u- ...i > and the
associated change of variables :
v - v
u
<z) i I n
r
. . n
r
. . . n
r a
u
t
* Vr"V
< B > 2 9 j
r j-h
J
h J
>
This definition extends to the case r m, provided that one adds the
convention that " n

, - * - x \ " is then equal to 1. From (B.29), one obtains :

j ^j V 'W "
l v + c
. [,
L
i
n
j--V . (T^)
z
Vi-%
+
j L^ v - i - K- v . (
1
fir)
t z S
v
1
-
n
- t.)Uiv, '.
j-r+1
J J
The assumption D
d + 1
f A^*' then implies the following majorization :
For v- v^Cz) , i > b
r
( U
r
) . , | * | V V '
(B. 30)
tij >.i>jCV . i - ' i *
r

105
j . . m h-1 0. . ' ( { L. . . . L. } )
Ix \ l
L
j y-\- ^
+ C
)l W "J '
J
* -
r-1 d
+
l
L
l V l 'V
1
"!. Y
L
\ . . 1.
n TI. h
J
h j
i - h
J
>
T
1
_
N
d
+
i i
v
- - V i ' V
L u
i
h
- - ' V
L u
i
h
. . . i
r
} >
( B
,
3 I )
j-r+l
J
The numbers M (U ) which occur in (B.30) are appropriate constants
such that : b.(U ) _> 1. One then defines :
h r r
J
t
, - I, j n { v- v
D
.(z) ; \z\ > b-(ff )}
V ""
l
r V
, - 1
r r
so that (B.3I) holds in J.
vv
Assuming that a sequence U , -.{u. ...u. . } has been defined
recursively, a set of numbers {u. . , i ; i a(i....i _.)} is determined
through the Heine-Borel lemma in such a way that :
H. - I. . > J.
i ....i _ , 1....1 . 1....1
I ) I. . .
V^r-l V
#i
r-1 VSr-l
{
-,.
U
. . V
, , l
H
x
t
:
Taking as its starting point the set {u.,icr, } of a ), this procedure achieves
the definition of all the sequences {u. u. } and sets I. . . J.
H. . , for every r < m.
V "
l
r ~
106
From (B.30), (B.31) and from the definition (B.12) o f - J , we then
conclude that :
f. , b-1 a C-CLJ . . . L. >) m a ({L. . . . L, , L}) + d + I
V r n ! [ F(v)| dv I n . "
1
'
J
x n
n
.
d + I
' J
I j "l
J
j-r+1
J
x
I M (U ) ( n n
l h , , - 1
j ) (B. 32)
( i
h
. . . i
r
) <7
r
r r
^j-h
J
/
a ' - a .
I ,
| z | V
" *
r
(I,;Y""
u
i . . . !. W)
d
l*l
WiM<Vr~' <V
J h j
where we have put :
''V-X/Vl^. V ! V
L
\- - -
L
j-
L u
i
h
.- - i
} )
(BJ3)
)
(B.32) holds provided that all n- >_ b :
r
' 1, the b. being appropriate
The estimate of the powers of T\. for h < j r in the r.h.s. of (B.32 )
are given by an analysis which is similar to the one described in a ) . It
yields the desired result
f
m ce<{L....L.})
|F(v)| d v < M H T\.~ * (B.34)
I
To obtain this, we had to take into account (as in a)) the fact that :
107
V. ; h < j < r : IT ({ L. . . . L. } ) - TT ( {L, . . .L ,L -Lu. , . . . l , - Lu . . } )
so that
f sup Co'. . Yl + d + l < a_({L. . . . L. }) (B. 35)
' (i
h
. . . i
J
. )eo-A V " Y - I -
J J
and sim ilarly : a
d + 1
({Lj L
r
,L}) + d + 1 ^ ^({Lj. . . L
r
>)
Finally , i t rem ains to m ajorize / | F ( V ) | dv. On each interval H.
we have !
v - Vy (z) ; with | z |<;b fly (B.36 )
m
which implies :
m h1 ' m / \
7 Lj n--"Ti_ - L V + c J L . n.-.-n + M L . - L U . . ) n-..-n +
where C C - Lz varies (in view of (B.36)) in a compact set W .
The majorization (B.31) is thus replaced by :
.., / m \ h-l a.., ({L. ...L.})
m a... ({L. ...,L. .,L.-Lu. ,...L.-Lu. . })
x n n- ' h
J l
h " -
1
j
j-ri
J
Since no z-integration remains in this bound (Valid on H. . ,) the end
V " ra
of the above argument (see (B.35)) applies directly and allows us to
conclude that :
108
m o({L....L.})
/ |F(v )| d v M H T\,
J
(B.37J
He can now conclude that in the case p 1, the announced result (B.5)
is a trivial consequence of (B.13), (B.28), ( B 3 4 ) , (B.37).
2. The general case
tfe split arbitrarily the space R? * into two factors :
n r s ^
R
( x ) "
R
( u )
x R
( v )
( i , e >
* "
( v l
'
v ) w i t f a t
< P
- 1
' s p - 1 .
We then use the following identity which links together the Taylor's
rests of f at x 0 and at u 0, and the Taylor's rests of the partial
v
derivatives D at v 0 :
[u0
[<l -4> ] <*> -[(I-t
u
)f](u,v,y) >..! ^ [ ( l - ^ -
| v |
) D^f 1 Cv,,)
<BJ8)
Since all the functions f (u,v,y) ,D^f. .(
v
.y) a r e .graded Weinberg functions
with respect to variables u and v and since r ,s<p'-l , it is possible to
apply - as a recursive assumption - the statement of the present lemma to
each Taylor's rest which occurs at the right-hand side of formula ( B 3 8) .
We will show that each term of this formula belongs to the class A T
with :
o { - a
d +
, ( S )
i f S = R
( y )
a (S) - sup [o r ,
+
,<S')]+ (d+I) if S R? .
(s'sTrCSM-irCS)}
a
'
w
'
i ) (I -t )f is a Weinberg function in the class
A
T
0 + a
\
s u c h
that
a ' C 8) -
d t l
( s ) .
i f S
=
R
( v ' y )
a'(S) - sup [a ,
+ 1
(S')] + d + I, if S < P K
i+q
t S'r t ^CS'J-i r ^CS
;
d + 1
w ,j - o T ., ^ - T - ^
( V i y )
109
Hhere IT, .. denotes the projection of &> onto R, . parallel to
r
W VLtjJ tv,y;
s +
R, . If we also introduce the projection Tt, -, of K ,
s
. onto
R? . parallel to R ^ ., we have : ir i r ,. o ir . ; from this w e deduce
that the equality :
T
( U )
( S ' ) - T T
( U )
( S ) implies ir(S') -i r (S).
Then we can easily check that for every S, one has :
c'(S) < a (S)
i i ) For each multiple index V with [v| < d, D f is a Weinberg
function on R ^ which is graded wi t h respect to v and asymptotically
Cv.y)
governed by the sequence {a. i, a. ,
+1
,..., a, ,
+
...} (the functions
cti i. (S) being here restricted to the set of subspaces S of B.. ' , ) .
Then from our recursive assumption, we can sav that (1 - t ' ) D
v
f
<cf,
v u
-|u-0
belongs to the class A. .'^ which is defined as follows :
|
w
| C S > - a
d + 1
( S ) if S e R ^
(B.39)
j V | (S) - sup r
a
d + l
( S , )
1 + d - |v[ + 1
if S c
y )
and S * R
y )
No w, we have to consider the function fi - e ' ')D f. " h
v
as a
+
\ v / u |
u
_
0
function on R? ' \ (constant in v ) , namely ir*,. h : lemma 1.2 implies that
te'|v|>
ongs to cne ciass J*
this function belongs to the class A. such that
*
S c R
E y > *
a ,
| v |
( s )
"
a
| v |
(
( u ) ^ >-
110
Thus, in view of (B.6), we certainly have :
a* |
u
i(S )< sup [a...(S')3 + d - |v| + 1 (B A O )
(here we made use again of the identity : ir ir
f
o IT, . ) .
Finally, since u is a Weinberg function with asymptotic coefficients
a(S) - [v| for every S t R ^
q
. and cc(S) - 0 for S = */*
q
j. we can
conclude from (B.39), (H.40), and from proposition l.I.b, that each terra
^j- ( 1 - '
T
) D^f (v,y)occurring in (B.38) is a Weinberg function
which belongs to the'"class
A
T .
+ a
) - Therefore, in view of proposition 1.1.a,
we conclude that (1 -t ) f (x,y) belongs to t& , ..
x
(P+q)
B) For every derivative operator iJ' with total order |v| n, the
function D z is related with the corresponding derivative D
v
f of f through
the following formula :
>- ( ' - < )
(B
-
4

Notice that for n >d , (B.41) reduces to D^g - b^f. Now ri^f is, as f,
a graded function ; it is asymptotically governed by the sequence ( V a ^ O :
a' a . } . So when n >d , we readily obtain that D^g belongs to the class
m m+n x
fa' ) (a )
Aj|
a
Aj, a ; namely, formula (B.I) holds true.
When n_<d, we can apply the result of part A to the function D
x
f and to
its Taylors rest of order d - n . In view of (B.4I), we conclude that D^g
belongs to a class K, with ci expressed as follows :
*n
(S
> -
a
'd - n
+ 1
<
S )
- < W
< B
-
2 )
- i f 8 * **
y )
:
a^(S) - sup r
a
' r f - n + | (
s
' >l
+ d
- n
+
I " sup ra.^.CS'll+d-
^ {S';Tr(S')^r(S)}L
d a + l
J {S';7r(S')=i(S)}L
d + 1
'l
(B.3)
Ill
T his achieves the proof of lemma B .1.
A pplication Co admissible Weinberg functions
(see definition 1.d).
L et a be a set of subspaces of K"! . and a an index function on
S
y
R . ., such that :
v
x
yj
V S with TT(S) E d ., (S) - <*{(S)) . (B .42)
T he class Q A ' can be alternatively defined as follows : it is the
class of all functionsf(x,y) which are graded with respect Co x, and
asymptotically governed by Che following sequence {ct^} :
V n > 0 : a
n
(S) - a(S) - n if ir<S) a
(.43)
a(S) - a(S) if ir(S) a. )
Applying lemma B.l to admissible functions yields the following property of
the T aylor rests of Che latcer.
L emma B .2.
II f belongs to a class ClJj ' , then its T aylor rest g - (1
- t
/
x
) )
f
is a graded function with respect to x ; it is governed by the following
sequence {a } :
V n > 0 : aJS ) - a(S) -n,
i s
- *
&
%) *-* >* a
a (S) - "(S)-sup(n,d+l), if S = K^ . and S a
a (S) - sup t a ( S') ] + d - n + 1 , i f S < S.f .
{s-, ir<S')- C S)>
a n d

( s )
,
^( S ) - <x(s), i f S c R^
y )
and S ( o\
112
Proof
Lemna B.I entails that g is a graded function with respect to x
which is governed by the following sequence {a } :
a) If S e E? one has (in view of (B.I),(B.2),(B.3) and (B.43))
. for S o : a ^ S ) - a(S) - d - 1 if n <, d
a ^ S ) -a (S) -n if n > d.
. for S $ (J ! a ^ S ) a(S) for every n >. 0.
b) If S * E5 ., one has (in view of (B.2),(B.3), (B.42) and (B.43))
. When TT(S) O :
a (S) = a .^Or (S)) + d-n + .l - (a (S)-d-l ) + (d-n + 1)
- o(S) -n , if n d
^(S) - a (S) a(S) -n , if n > d.
When TT(S) $ a :
a (S) - sup ta(S')] + d - n + 1 if n < d
" {S';ir(S')=ir(S)}
a (S) =< a(S) , if n > d.
q.e.d.
113
REFERENCES
1) Deep inelastic scheme, Pevnaan J.G., "Photon Hadron Interaction" W.A. Benjamin
Hew York 1972
2) Bros J., Analytic methods in mathematical physics. New York, Gordon and
Breach 1970
3) Lasalle M , Commun, math. Phys. 36_, 185-226 (1974)
4) Ruelle D., Nuovo Cimento J , 356 (1961)
Araki H., J. Math. Phys. . 163 (1961) ; Suppl. Progr. Theor. Phys. _lj[ (1961)
5) Bros J., Lassalle M., Il Commun, math. Phys. 43, 279-309 (1975)
III Commun, math. Phys. 54, 33-62 (1977)
IV Preprint Ecole Polytechnique to be published in Ann.IHP
6) Taylor J.G., Theoretical Physics Institute University of Colorado, Summer
1966.
7) ZimmarmannW., Commun, math. Phys. JL5, 208-234 (1969)
8) Weinberg S., Phys. Rev. vol. 118. N 3 (1960)
9 ) Hahn Y., Zimmermann W., Commun, math. Phys. 2.
3
3 (1968)
10) Horraander L., Survey Lectures on the Existence and the regularity of
solutions of Linear Pseudodifferential equations (1.971)
11) Duisterraaat J.J., Fourier Integral Operators (New York University)
115
I I . COMPLMENTS DE LA PARTIE I
INTRODUCTION
Une tude systmatique des divers travaux, concernant les amplitudes
de Feynman en Thorie des Perturbations, montre qu'il n'existe pas un
procd explicite de prolongement analytique des intgrales renormalises
dans l'espace des quadri-impulsions complexes. En effet, presque toutes
les mthodes de renormalisation proposes pour liminer les divergences
ultra-violettes utilisent des techniques de rgularisation, et ne four-,
nissent pas d'expression analytique pour les intgrales de Feynman la
limite correspondant l'limination du paramtre de rgularisation. La
seule mthode qui vite ce dsavantage est celle qui a t propose par
Zimmermann ; en effet, dans . 0 une dfinition explicite de 1'integrand re-
normalis est donne,pour tout graphe de Feynman G., sous forme d'une fonc-
tion rationnelle dans l'espace de Minkowsky de toutes les quadri-impulsions
internes et externes de G. Il reste pourtant encore un paramtre de rgula-
risation prcisment le -ie des propagateurs de Feynman qui ncessite une '
dmonstration supplmentaire du caractre distribution -^ de l'intgrale
normalise'(obtenue quand e * 0) .
On a malgr tout l'intuition qu'une utilisation de l'intgrand renorma-
lis de Zimmermann dans l'espace des impulsions complexe sans paramtre de
rgularisation c devrait conduire, en choisissant un contour d'intgration
appropri, une dfinition explicite de l'intgrale analytique renormalise,
au moins dans un certain domaine de l'espace des impulsions complexes.
Jusqu' prsent les mthodes utilises pour l'tude des intgrales de
Feynman ont eu surtout pour but de dterminer directement les singularits
de l'intgrale, savoir les "singularits de Landau". Pourtant, dans les
diverses tudes mme i^s plus rcentes ,.
3
^ qui prtendent s'appliquer au
cas de la renormalisation, il n'existe pas une expression explicite de
l'intgrale de Feynman analytique dans le complmentaire des singularits
de Landau ; en particulier l'intgrand renormalis de Zimmermann n'a pas
t utilis.
116
Il est important ici de remarquer que le caractre local de chaque
terme dans la srie perturbative tait pendant longtemps une conjecture
qui n'a reu sa dmonstration rigoureuse que par la m t ho de de r e n o r m a l i -
sation d'Epstein et Glaser ; ces auteurs ont montr qu' chaque ordre
de la thorie des perturbations un systme causal de produits chr o n o l o g i -
ques peut tre dfini ; d'autre part ils ont tabli l'quivalence de
l'existence d'un tel systme avec celle d'un ensemble de fonctions a n a l y-
tiques n-points qui peuvent do n c s'identifier des sommes d'intgrales
de Fe yn m a n correspondant l'ordre considr de la thorie des pe r t ur b a t i o n s.
Cependant du point de v ue des proprits d'analyticit des intgrales
de Feynman, cette dmonstration est extrmement i n di r e ct e .
Le but initial de n o t r e programme a t d'essayer, pa r un e m t ho de
directe de dformation de contour dans l'espace des quadri-impulsons co m -
plexes! d'obtenir l'amplitude de Fe yn m a n comme une valeur au bord au sens
des distributions d'une fonction analytique dfinie par une v e r si o n dans
les complexes de l'intgrale de Zimmermann. M a i s, la possibilit d'tendre
la notion d'amplitude de Fe yn m a n n celle, plus g n r a l e , d'un produit de
G-convolution au sens de Bros et Lassalle (expliqu dj dans l'intro-
duction de I) associ un graph G - dont les v e r t e x v ("boules") co r r e s-
pondent des fonctions gnrales a ppo i n t s et les lignes des fonctions
gnrales 2-points - nous a incits largir les perspectives de n o t r e
programme. Nous dfinissons donc un integrand gnralis renormalis R e n
nous inspirant du procd de Zimmermann ; ayant assur dans la partie I du
prsent travail que R_ satisfait au critre d'intgrabilit du thorme de
Weinberg, dans la rgion Eucl i di e n n e , notre but est de pouvoir dmontrer
qu'a u m o ye n de l'opration de la G-convolution - qui gnralise la formule
intgrale des amplitudes de Feynman' - l'intgrale de R satisfait toutes
les proprits analytiques et algbriques conformes au programme linaire
de la thorie axiomatique des champs, c'est--dire plus prcisment, la
Structure d'une fonction gnrale n-points (qui v a tre brivement rappele
ci -de sso us). De ce t t e m a n i r e l'tude de s intgrales de Fe yn m a n . renormalises
constituera une application particulire de ce rsultat g n r a l .
Le s rsultats inclus dans le prsent travail concernent seulement une
tape du programme entrepris, savoir, les proprits analytiques et a l g b r i -
ques de 1'integrand renormalis gnralis R. Rappelons d'abord la notion
de fonction gnrale n-points :
* ^ ^ ^ _ ^ _ ^ _ ^ _ _
Po ur une prsentation complte et dtaille de ces proprits et de toutes
les rfrences associes, v o i r : Bros (Thse)
5
.? Lassalle &' et Epst e i n et
Glaser : les Houches 1975 ft,
117
Considrons l'ensemble d'indices X {l ... n } ; appelons 5^ (X)
l'ensemble de tous les sous-ensembles propres de X et (I, X \ I) toute
partition de X en deux sous-ensembles propres t, X \ I de X. Une fonc-
tion H (K) dfinie dans l'espace C.-1 de l'ensemble
K - {K. P. iQ. C... i X, K. - 0} est dite fonction gnrale
i l i (i)
i
*
x
1
n-points si elle satisfait aux proprits suivantes :
i) Analyticitg dans l'union d'une certaine famille de tubes :
{7 ^ - E ' + i C , A A
t n
' } , et croissance lente lorsque K tend., vers
~o 4(n-1)
les rels ; ici ( . est un cone dans R ' dfini par :
C
x
- J Q K
4 ( n _ l )
: X(I )Qj.ev
+
VI ^ (X)J , (1.1)
X(I) est une fonction dfinie sur IP( Xi prenant ses valeurs dans {1,-1}
et telle que j soit non v i de . L'ensemble de telles fonctions est dsign
par A .
Les valeurs au bord W J (P) ; X A } sont des distributions.
ii) Les valeurs au bord HO (P) et H," (P) associes a deux tubes adja-
.m- 11- A * A
cents v 7 'A spars par la partition (I, X \ I ) concident ( au sens
des distributions)dans une rgion relle Sz\ dfinie par :
fl, - {p
4(n
-'> : *l< *l , P * ^ } (..2)
Dans (1.2) M est un seuil de masse appropri dans la voie (I, X \ I ) et nt-
est use nasse dans le spectre discret de la thorie.
i) et ii) constituent le domaine primitif d'analyticitS d'une fonction
gnrale n-points.
ii) Les valeurs au bord {H, (P) , A A } satisfont des "relations
de Steinmann" gnrales. Plus prcisment pour un quartet de tubes adjacents
{%" I < i < 4} correspondant a un couple de partitions transverses*"
<I', X \ I ) (J, X\ .J), savoir 5^"
+ +
" ^ J^_ _ ^
+
_ , les quatre valeurs
au bord correspondantes satisfont :
H^(P)
+
x -
( P )
*
H
X ? -
( P ) + H
A ^
( P ) ( , , 3 )
Transverses si gni fi e que : I n J i 0 I fl( x\ l)t 0 , ( X\ J) n J i< 0
(X \ I) n (X \ J) il 0 .
118
iv) La T-valeur au bord de H appele- IP (P). est dfinie par la
"prescription de Rue l l e " :
&>m - H H
Cn)
CK) (i.4)
K+ P
P n
x
,
Q
e q
Les ouverts S2, sont dfinis par ;
a
x
- | p e n
4 < n _ 1 )
: XWP JC VJ VI ^(X)j ((1.5)
o
Vj - v ; - {p K
4
: p > ? 4 } u {psm
4
=
P
- / ? $
(1.6)
Quand H (K) est la fonction n-points physique, alors HJ; (P) est le
produit chronologique connexe T (p).
Le plan du travail qui suit est le suivant :
Dans la section 2 nous introduisons une technique qui associe de faon
cohrente un ensemble de graphes simplement connexes ou arbres ({T ; V y c G } .
un graphe G donn et tous ses sous-graphes ; nous appellerons cette o p -
ration "clatement" ("blowing up pr o ce dur e "); elle nous permet d'tendre
1
r
integrand'renormalis gnralis (TU -* fL) un espace complexe de di m e n -
sion suprieure.
Da n s la section 3 nous exposerons un ensemble de rsultats m a t h m a t i que s
utiliss par la sui t e . On tablit certaines proprits algbriques et a n a l y-
tiques pour des fonctions n-points du type "a r b r e " ou "fort".
Da n s la section 4 on applique les rsultats de la 3rae section 3 K
Q
et on m o n t r e qu'elle est une fonction gnrale n-points "au sens fa i b l e "
savoir elle est analytique dans le domaine primitif et elle satisfait des
relations de Steinmann seulement par rapport certaines partitions t r a n sv e r se s.
L'tape suivante, n o n abordeici, concernera l'intgration de t par le
(s
pr o c d de rcurrence sur le nombre de boucles utilis par M .La ssa l l e dans
l'tude de la G-convolution '
J19
SECTION 2
2.1 - Sets of independent internal lines of G and of subgraphs Y of G
Definitions 2.a
2a.1 - On the set J. of internal lines of a given connected graph G, we decide
once for all to put an order Ci. which we denote by the sign > , so that :
/" {'.- '-&] . C O
with A* IA I (the number of elements oiJL) and :
t ,>
2
>-...|
2a.2 - The order Q will allow us to associate with G a well defined tree
graph G as follows.
We first define the following subset K of internal lines of G (as
a subset of JL , it is necessarily ordered by Q) :
satisfying the properties :
a ) G\ ', connected and V C, } [, G\ l disconnected.
b ) V j 4 , G\lj\...\l. connected and
V l
k
: J j - i ^ f j a \ ,\ ...\
i
_
l
\ e
k
disconnected
c)V l
k
:
L
>f
k
, G\ ,\ ...
L
\ f
k
disconnected
W c a l l ^ "the set of independent lines of G induced by the order " .
We then state :
Lemma 2.1
The subgraph G\?C of G is a tree graph G (S2J (depending on the order 52).
Proof
From properties a ), b ) , c), it follows that G\JC is connected and that
G\ Xj\ '
k
is disconnected, for every [ G V L . These properties define a
tree graph and L li/L,l is the number of independent loops of G. When the
order 2 over the internal lines of G is changed ( jQ . SI"' i
120
another JL, is obtained and G,WL = G'(fl') is a new tree graph.
Since the order ft induces an order ft on the set of internal lines
of every connected subgraph Y
c
G, we can define in an analogous manner (by
properties a ), b ), c)) the set of independent internal lines of Y :
{, >...L } ; similarly the set of independent internal lines of G\X,
Y
vA
Y Y
denoted by/f". One then verifies (by application of the properties a ), b ) , c)
for the latter sets) that :
*
Lemma 2.2
The subgraphs Y u L and GvTfyT' are tree graphs depending on the
order ft.
He shall now prove the following :
Lemma 2.3
Let Yi Y
2
o e
two connected graphs with Y i
c
Y,
a n d l e t
& x be a line
in Y,, which links two vertices v v. of Yi M Y,
x
{
h
J is connected,
then y Ail . } is connected too.
Proof
Let us suppose tha t y.^ -ll . } i s disconnected ; tha t means Chat there
b a h
a re two di sj oi nt subgraphs Y?
c
Y
2
Y
2
c
Y
2
wh
e r e
Y
2
( resp. Y
2
) conta ins
the v ertex v ( resp. v . ) such tha t : Y
2
" Y
2
u
^
a
k
U Y
2 '
T h e n t h e i n c l u s i o t l
Y, <= Y
2
allows us to write : Y, - Y*UY
b
U U
a b
} with Yf - Yf f> Y, , Y* - Y
2
f Y, .
the two l a t t er graphs being di sj oi nt ; this ent a i l s tha t y.^ ll >, }"Y
a U
Y^ i s
disconnected which i s contra ry to the hypothesis.
Lemma 2. 4
Let Y't Y" be two connected graphs, with Y'
c
Y" and let the following
hypotheses be satisfied :
i) There exists a set A.' of internal lines of y' such that : V f g A , y'\{( }
is disconnected,
ii) There exists an internal line
a
( p&) of y' such that : y'\ f
0
J
is connected,
iii) A" is also such that Y"\ A" is connected.
Then 'Y"\A\IL] *
S
connected.
121
Proof
Because of i) we can write :
r'=r
c
;,
u
7>--r>A" .
with Y'.. connected, Y,:. f\ "trj\ -$ V,j,k:L*l and the reduced graph
Y / y ' n v ' H ' is a tree graph whose internal lines are the elements
of A" : we say that A" is a "skeleton tree" of Y . Now there exists an
index j such that t.
a
Y,:. Moreover Y.'. \ ^f
s
] is connected : in
fact, if this were not the case, we would be able to write !
and therefore A" U J 'J would also be a "skeleton tree" of y'; this would
entail that y'\ !J
5
1 is disconnected which in contrary to condition i i ).
The inclusion V,- \< Z( Y"\ A ) and condition iii) allow us then to apply
Lemma (2.3) and conclude that Y"\A\IU] *-
s
connected.
Lemma 2.5
For two connected graphs y., y, with the inclusion K T , the associated
sets 7C* , 7C-, satisfy :

. ,
Proof
He shall prove it by recursion. He suppose that the j-1 maximal lines
of \K-t belong to Jiy too. We shall then prove that the j maximal line
of J( belongs to J( . Let us denote by J{Y the set of these common
He now call A*
J
' the set of all lines I in
X
such that >..
J
He can vrite :
A
( /
L A , V A, U % l
122
with: vW/ ^r, ) nti
v
For each line - in A. , there exists a number K(i) and a corres-
ponding line %v(!\ in J i-t such tha t , . , >A. >
K f i

+
, >. . Then from the
definition of J ty i t follows tha t yX l.\ \ i j , , . \ i i s disconnected,
and therefore tha t :
Vli & A
(
j V\ J { \ l- i s disconnected (a)
Moreover, the definitions of c/L
v
and sl
7
a lso ent a i l tha t :
7 \ Ay \ h i s connected (b)
y \ A

= ^\ A
J
\ A

\ X
r
i s
connected (c)
Properties a ) , b) , c) , allow us to apply Lemma ( 2. 4) to the connected graphs
As a r esul t , the subgraph Y,\A
J
^4 . i s connected, and therefore C yS/tv -
To end the proof we j ust have to notice tha t the f i r s t l i ne l,&J it belongs
v . . . ~t/(i'ii ''
to yt y , since the previous argument holds with yt
Y
^ the empty s et .
II
Lemma 2 . 6
For a given connected graph G and an arbitrary connected subgraph ycG
the following decomposition formula holds for the set </[(. :
X
g
=K
r
v X <2.5)
(so that QsG\J C
t
\ Xf)
123
Proof
We first note that the inclusions v cff, CSVTL./C (? imply according to
Lemma 3.5 that :
^ c %
e
Siy C J%Q
Moreover by the definition of the sets J\y, shy it holds :
t/T,y( \ Sity 0 . From these considerations we can write :
, where A means a certain set of internal lines
not contained in JCy ,nor in JVy . Lemma (2.1) (2.2) state that :
G.= G\%
e
=G\J (f\'J {?\Z- is a tree graph and G\X
r
\7Cf
is a tree graph too. It follows that A 0 j this ends the pr o o f.
For an arbitrary couple of connected subgraphs ^^f of G, the order Qy
induced by Q allows us to define in a similar way the sets :
. Therefore Lemma 2.6 can b e immediately generalized
into the following statement:
Lemma 2.6
For a given connected graph G wi t h the order Q. on its set of internal
lines, and for every couple of connected subgraphs JAcy of G the sets
!Ky Kf> , A t satisfy :
Ay- < &L. U IK (2.6)
W e shall n o w prove a general decomposition formula for the set Jig inside
each forest U of G. Let Y he a connected subgraph y G with L independent
loops and let {y ;a!.... c} be the set of maximal disjoint subgraphs of y
in U(y). Each Y has L independent loops. Let A b e a given order over
the set of internal lines of Y- We consider the union of the sets Jly
u
: ^ - ^ i
with I V Ji
v
I : Z. i-i
w
e define then the set of independent lines of
Id. a.
y\U X
y
(induced by ) as the set of ordered lines :
' 4. Id.
j \ l
u
= [I, > l > . , ,>l J , nrl .Z U (
2
.7)
124
analogously to definition 2.a.2
We then prove the following generalization of-Lemma 3.6.
Lemma 2.7
With the above definitions, and for the set of disjoint subgraphs y
al...c of a connected graph y, the following formula holds j between the
- X
7t
X
U
,X^ :
^r*fy%rJ y^uu *=i...< : . <z-8>
Proof
W e suppose that (2.8) is true for a-l,...,j-l :
J
h
and w e prove that it also holds for aI,...,j . We note that Y;
c
7 ^
moreover by definition t/L,, * w i t h a*l,,..,j-i, is the set of independent
ines of y\t) X .
An application of the decomposition formula (2.6) of Lemma (2.6)' yields then :
* -,,
J
r
_ is the set of independent lines of V \ " /!
i from (2.8) and (2.9)
-.bx
r
uXL
Here
'y . ^ i v
we obt a i n t hen : Z*X uXL ( 2 . 10)
125
Application of (2.10) for every j-l...c yields (2.8).
Lemma 2.8
For a general graph G and a given order on the set of its internal
lines, inside each forest U the following decomposition holds for the set
% of G :
Zf. = ( uX) u^ (2.H)
with y , a*l...c, maximal subgraphs of U(G) ; for every y in U , the
set Jiy also satisfies the decomposition :
Xy = (UX
r
)UX
UY
(2.12)
(Y , Tl d, are the maximal elements of U ( Y ) ) .
Proof
We use the recurrence hypothesis that the statement holds for all maximal
elements y a-l...c of an arbitrary subgraph Y belonging to the forest U.
Application of Lemma 2.7 then yields directly (2.12) for every Y- If "e put
Y " G formula' (2.11) is obtained.
2.2 - An admissible set of basic internal momenta : the fl assignment
With a fixed order fi on the internal lines of a graph G , we are going
to associate a special "admissible" set of basic internal momenta for G and
for all connected subgraphs Y of G ; this special choice will be called an
"fi-assignment" of internal momenta. We choose as a particular solution of [2,10
the one corresponding to the choice :
*aba "
f o r e v e r
y
U a e
* a b o -
e
^
( 2
-
1 3
>
The set of equations ( HAO ) then redi-<-^s to a system of 4(N - P equations
for the H -1 momenta p
a b o
associated with the li.ies i=G\ ?Lg (the
latter being a tree graph). Furthermore, with every line ".ujSJ^
w e
In the sense of Zimmermann [Z] as already explained in []] .
126
associate one of the L-independent variables k. (whose sec is denoted by k).
k
a h T "
k
j ' 1<JS- , for every 1^ 6% (2.14)
So in the following the lines ojC, will be in one-co-one correspondence
wich Che L independenC variables of incegration. Similarly, for every
connected subgraph Y of G, we choose as a particular solution of (!?")
the one corresponding to the choice :
the set jpJt_'V I ^ -ST A ^ L f being then determined as the solution of a system
of 4(N -I ) equations. We also choose the set k' of independent variables
of integration in Y as the set of four momenta k . associated with all
the lines { , in J {~> T aking into account L emmas 2.5 and 2.9 we can
then put :
y
T hat means the set of independent variables k is a subset of the set k
defined in (2.I t). We are now in a posiion to show the following :
Proposition 2.1
For a given graph G the above defined -assignment of internal momenta is
an admissible one. T hat means it satisfies for every subgraph YCG :
i)l
a b f f
(k,P ) - ^ ( k
r
, P
y
) (2.17)
i i 3
P ab*" P abtf
( K
'
W
<
2
-
1 8 )
k
I b ^ -
k
L
(
<
2
-
, 9
>
(p* (resp.p) denotes the set of basic internal aomenta of Y (vesp. of G).
iii) For every couple of subgraphs Y, U. of G such that u.c. Y '
tib* "fib* < ***>
(2
'
20)
k
a b
f f
"
k
abff
( k Y )
<
2
'
2 1
>
Proof
i) T he 2-assignment (equ. (2.15) (2.16)) ensures that for all lines / .
127
belonging Co the sec/L , (2.17) is satisfied :
U,-t\
ba
-**, ' (2-22)
M oreover the solution of the system of -1) equations provided by the
energy momentum conservation laws at each one of the N vertices, yields
Jv_(V {'.eyvL ) as a function of the known parameters k (equ. (2.22),
and the external momenta to V :
A " r ***
< 2 , 2 3 >
where the sum extends over all lines c . G\Y incident to the vertex v' .
abo* ' a
On the other hand the corresponding momenta [. carried by the same lines,
now considered as lines of G satisfy equivalent equations at all vertices v^ .
T he solution being unique we conclude that :
ii) For every line c . e/iy we have because of the S2-assignment :
V &% 4o-" "abo-"
k
L " W
k
> <
2
-'*>
On the other hand for t uJ ^ Y\-'y
:
^
e
solutions of the system of 4(N -1) equati
of type (3,10") and of the system of 4 (^ -1) equ. of type (fff'") yield:
V&r e ' y\ft
Y
4 a " *L,<*1 ' Pl .W <
2
'
2 5
>
v
k
I b < r -
k
L
( k
> <
2
"
2 6
>
i i i ) I n view of Lemma 2. 6' the above arguments hold i n a si mi l ar way for the
rel at i ons between the i nt ernal momenta of y and u, ( r,t being an arbi trary
couple of subgraphs of G such that U c Y
123
2.3 - The "Blowing up procedure" associated with Q
Definition 2c.1
For an arbitrary graph Yff M! consider now the following procedure. He
"cut" the lines of the set JC* so that from each internal line e Jw e
obtain a pair of external lines . We then define for G a set :
&'-{*;>#] ,
and for a l l yG we define the corresponding sets :
"Ty-Jl. with f. eylv a nd L /Jg , V i - I. . .L j ; so tha t the following
conditions a re sa t i sfi ed \KcJ ia
a n <
* ^
o r e v e r v
couple y,fc with ycltCG :
lemma 2. 8 then allows us to st a t e the following :
Lemma 2. 9
Given a graph G with n external lines and given an order 2 o n * , the "cutting"
of all lines Ij&JCgCQ) inside every forest U yields simultaneously :
From G a tree graph T. having n+2|./Lj external lines and from every subgraph y
of U?) a tree graph T with n+2l^trl external lines. We call this opera-
tion a "blowing up" procedure for the forest U of G.
After this "blowing up" procedure we define the new renormalied integrand
R , in a way analogous to I.
Definitions
2.C.2) With every pair of cut lines (j l^)B
we associate momenta (Jfc. fc.)SC
8
(resp (ff/ jf')fi <C ) . We use the notation
(K,fc, k) (resp K k' k' ) to denote the set of external and internal momenta of
T- (resp. Tv ). These momenta satisfy :
2.C.3) In a way analogous to [I] we define the mappings :
See in section 2 of [I] all exact definitions given there
129
Y bel ; CK.K.10 <K l; (KX faty) A ^ CKWAD
j(resp. wi t h every line .)
and w e associate wi t h every v e r t e x U of T./a completely amputated general
n -point function H (resp. Hj|j ) satisfying :
3?y(h) : YhCW (
r
.f* fifth) zRtftCKW))
(2.29)
He shall suppose that :
{[""Mm-))]
t
>flV l;lJ C2.30)
K 'A
C l < k
K ' U , 0 --C'rt?
K
^ (2.30
=0
res?
.j fi'V'rKW;/^ . 's/T VfeV;; I
Hfttf& w) ^ .fl*Uo<w
;
(2.32)
W e then de fi n e the "integrand" corresponding to T_ (resp T.) by ;
- , ,- 2, ____ -i - (2.33)
L( M/ r t =TT H 7K
V
CKS*O; TT H
ffJ
ft-ow;)
4-.
This is a shorthand notation to indicate that we take the restriction of
the function \\"
v
(K k JO on the submanifold k. k' \< i.)X~\>
1 1 ' Kt '
130
Here *C_ denotes the set of internal and external lines of T .
T
G
G
Respectively: -T T i V c K W,) J T H ^ C U K V A O ) (2.34)
7
ve/T
r
i*A
T
'
For the "integrand" corresponding to a tree graph T r which results from a
"blown up" reduced subgraph J 7^ r,..,K. with K aiKdisjoint maximal
subgraphs of U(v) we have the analog of formula (.2, ) of [ 1 ].
L=7T ^ (MKWMIIT Ha\(!i(if.K
r
Kn) (2.35)
From properties (2.22) (2.23) we can show the following :
Le m m a 2.10
A* t.y
(resp.
2.C.4 - Taking into account the order i2
y
on the internal and coupled blown
up external linesof T we select the first 2 ^ 1 -1 momenta :
corresponding to the first ordered "cut" lines : J q h fs.J (v U'^ r ^ 'lX I' ,
We define on this way a set of independent internal momenta of and we denote
it by :
131
K
T
= [k] , */',/Sj \Kf\-l , ana *L , j -(2.38)
From Lenma 2.5 and the momentum 2 -assignment (sec. 2.2 equ. (2.18) (2.19))
we then obtain :
Lemma 2.1T
Y
For any two graphs Y. , Y- with Y- c Jfj the corresponding sets /C ar/,.2
of independent internal momenta satisfy :
fc
Tl
C k
r
> (2.39)
2.C.5 - Given two graphs T,c Y, we consider a function g(k >P (K *)) of
the internal and external momenta of 7\,. The inclusion property (2.39) allows
y . y I
us to define the functions k
i
( k ' ) by :
*f
sK
/' IS-jSl^l ) *f'=tf'' l<)il%
Y
J -i (2.40)
The basic internal momenta denoted by P L,-^ * ) (for an internal line ' .^
of Tv ) are defined as in sec.2 of [I] through the momentum conservation
equations at all the vertices of the tree graphs T- , Ty by (see prop ..2.1) :
r l - - r l ^ ' > p
r , )
<
2
-
4 l
>
y
Finally the substitution operator si [see [I] ] acting on the functions
of the variables of any graph JT.C f , is defined as follows :
S* g ft\ p
T l
) = g (i$
Yl
0i
X
) , fttf, f) ) (2.42)
132
and satisfies the following properties :
2.C.6 - T he renormalized integrand R
fi
(analog of equ.(l.) after the "blowing
up" procedure is defined by :
U4jf\x
r
.reUCs)
(2.45) with : for a full forest U
u 'a ff
T y Y ^ " " normal " " (2.46)
Here the functions Y^ are defined recursively by :
r
r
.I
?
T T ^ f-t^ ?
4
(2.47)
and; J
r
= l t H
f 2 ,
f ^ CKJ; (2.48)
the product in (2.48) is taken over all external lines of G . We state the
following :
Proposition 2.2
For a given general graph G with |/U ' independent loops. T he G.R .l.Z. R

(K, )
(equ.C ) is related to R g (equ. (2.44)) by :
G
k +K =0
133
Proof
We use the following recursion hypothesis. We suppose that for each y^ maximal
subgraph of U(v) the function Xfa. satisfies :
K-K. . .
(2.50)
We then prove that a similar relation holds between Y_. and Yv .
The operator -T '* acts only on the external variables.and S
a
satisfies
property (2.43) ; therefore taking into account the recursion hypothesis (2.50)
we can write (in view of (2.Ik)) ,
J A
1
We combine Lemma(2Ao) with (2.51) and formulas (2.12) (2.14) inside, the formula
(2.4?) to obtain :
%\
M
- - T T , H
f
fy;Y/KV; (2.52)
We then apply the result (2.52) to / G and obtain :
TUe sum over all forests U yields (2.49).
The above statement allows us to study all properties of fuCK , ' ) (this
will be the scope of the two next sections) and then to try a procedure (to
be presiced in III/ of successive integrations of R<y|u.!j' over all
internal four momenta . kj [<Jf?2^l > in order to obtain the extension in the
complex space of the finite part H of Zimmermann (equ.2, ) and prove all
its analytic and algebraic properties.
134
SECTION 3
3. Some operations on general n-point functions
We shall prove some useful properties of the general n-point functions.
Lemma 3.1
let H
( n )
( K ) , K - ( K j j e c
4 ( n - t )
, (iex - (l...n J), be a general n-point
function. The partial derivatives with respect to arbitrary directions :
; H
( n
' ( K )
'YK- "\U.
'1 ' ' ' 'r (for any choice of 4(n-l) independent
components of the vectors K.) are general n-point functions of the same set
of- variables K e t ^
n _ 1 )
.
Proof
The analyticity inside the tubes of every derivative of the analytic function
H is evident. The Steinmann and coincidence relations follow from the li-
nearity of the derivation operator together vith the fact that the real boundary
value of any derivative of the analytic function H is equal to the corres-
ponding derivative (in the sense of distributions on real spaceR.
x
)
of the boundary value of 11 .
Lemma 3.2
a) Let H
( n )
( K ) be as in L.3.I. The product of H
( n )
( K ) by a polynomial B(K)
in the same variables is a general n-point function.
b) The sum 2 H . (K) of general n-point functions is a general n-point
function.
The proof is trivial.
Lemma 3.3
Let lT
n
'(K) be as in L.3.I. For every I X the restriction :
H C n )
<
K
>l {
V
o, i ei j
is a general n-ll| point function.
135
Proof
It has'been proved in [7] and [14] that a general n-point function is analytic
at all the boundary points of the tubes J?" which do not belong to the union
of the family of sets [ A T , J& P*( X)\ '
It is then easy to check that in the linear m a n i fo l d;
n-IIJ A*"'*'
every tube T
u
( V& f\ ) belongs to the intersection of the b o un -
n . >-*(h-0 . . A
daries of several, tubes T w y )
l n
"-X
a n <
* ^
a s n 0
P
0 1 n t l n a n v s e t
^ T
defined above. Moreover the coincidence relations for H
between adjacent tubes (j , Ty ' ) are the restrictions to C ^ "
of the coincidence relations for H* ' between the corresponding sets of tubes
(
T
3 6)1 ^3 '< ') 5 Steinmann relations for H
l n
are also restrictions
of corresponding Steinmann relations for H .
Definition 3.a
A partition (I, X\ l ) is called trivial for a general n-point function if
the corresponding discontinuity function is identically zero ; more precisely
if for every pair of adjacent tubes jj, , \ A ' separated by the 'face q_ 1
X
\ T * >
one has H ^
n )
- H ^ P in the whole space t
( n - 1 )
.
Definition 3.b
Let (3 be a partition (I.....I ) of X : X - ,U I. wi t h M /I ... m f.
n J. m Je M J 4(
n
_|) f
l
'
We define the following linear mapping i . of C < (K....K ) K ,
a
m
x L i n
1
" J 1
c
4(n-l) . I ,
(
.
+
.
j f l 0 t
Xk 0} :
:
V
0
"fa " ?
K
i '
V i
' H * *
(P)+i
^
<Q) ( 3
-
136
With every proper subset N of M , we associate in X the' subset :
I(N) - U I I(N) < = X (3.2)
jSN
J
He thus have : p ; Z p. Z p. ; *T( K)
Lemma 3.4
Tor every general m-point function H (k) the inverse image of H
^ V
H
( n
^(K) - H ^ ( i _ (K)} is a general n-point function,
^m
Moreover the following specifications hold :
a) For every partition (N,M\N) of M which is trivial (resp. non trivial)
for H
( l n )
((Q, the corresponding partition (I(S), X\I(N)) of X is trivial
(resp. non trivial) for H
( n )
( K ) .
b) For a partition (I, X\I) of X to be non trivial for Hi (K) the following
condition in necessary :
There exists subset N such that I I(N) (in the sense of (3.2)).
Proof
The proof of this lemma is obtained by the following analysis :
i) Tubes :
We note that for every given cone C j
n >
^ e A
< n )
in the space j R *
( n - 1 )
:
^
n >
-{Qfi l R*^"
0
: ,}(!) Q
X
V
+
j Jl" 8 function on P\x) J
there exists a well defined cone C, in \ , such that :
. ^ 4
m )
) 3 e j
n >
(3.3)
here It is a sign function on IPiti), which is defined as follows :
V N e j P ( M ) , 11(H) - ^ ( K N ) ) . (3.4)
137
Since H^ ' (k) is analytic in the union of all the tubes ,*
5
( m )
- { feeR
4 ( m
" + i t
(
f } C/ieA
(ffl)
) H
(t
(K> is analytic in the union
of all the tubes C'ct/f
5
) " " "
in view of (3.3) contains the union of all
the tubes 5 - R ^
n
- ^
+
i ^ .
ii) Coincidence relations
L et (I,X\I) be a partition of X and let Cj+ , j- be two adjacent
cones separated by the face Q_ Q\T " ^ T hen ^
v o
cases are possible :
a) I f I is not a union of sets in , then in view of (3.4) and (3.3),
2 J j determine the same u, , and :+ > Ctf-
t n u s
belong to the same
cone i (Cj ? ) T he boundary values of H
J
(K) from the corresponding
tubes ^T- 5J- therefore coincide on all R y
a n d t h e
partition
(I,X\I) is trivial for H' (K) : this proves property b)
B ) If I is a union of sets I . in S , i.e.
J m
if I I (N) , with N C H , .then 3 ,3 determine (respectively) two
sign functions U. , u. such that
.-l,.s(m) . />(n)
I,he boundary values of H from ffi
t
, X -
}
coincide in a region of
the form : #
M
-[ p S ^ ^
- 0
; P* < M ;" P* + m
2
j ; and since
P ( i
= P ^ ,
the corresponding bo""dary values of H ' from 5\} >5j- coincide in
the region :
^.{p *<*-'> ; r *< . P * * .
2
]
I n particular (I,X\I) is trivial for H
( n )
(K) if and only if (N,M\N)
is trivial for H (R) , i.e.
if J ty, is empty (property a).
138
iii) Steinmann relations :
Le t ! % , , + , ' + ( be a quartet of tubes mutually
L

y l
" ~
J
**~( n) w
separated by the two faces <J_ - 0 , Q.., - 0 (in Ujc-r QjS V ,
Q
(
e V
4 ,
) : (I,X\ I), (I',X\ I') are two transverse partitions of X
(i. e. m i ' M , (XM) -n r #0 , i n (x\r) * , (X\D n (xu
1
) * 0 ) .
The boundary values of H ' from the tubes Ju"
e
, (on R ) are
called Hj?' (P) and we want to check that :
H > + H^> H > - H f
a )
(3.5)
a ) At least one of the two partitions (say (I,X\ I)) is a trivial partition-
for H
( n )
( K ) . It follows that : "
H
( n )
( P
) .
n
( n)
m
H ^ ( P ) - H ^ ( P )
and relations (3.5) are trivially satisfied.
g) Two partitions are non trivial for H (K) : From property b ) already
established (in ii),' there exist two subsets N, S' of M such that I = I (N),
I' I'(N') ; (N.M AN) and (N',M \ N') are transverse partitions of M which .
separate the quartet of tubes ji ^JV') I
i n
Si
I h e b o u n d a r v
( *\ t \ *~
values H ' , of H^
m
'(|(;) from these tubes then satisfy :
H
( m )
+ H
( m ) . <)
+ H
( m )
o n
^ ( m - I )
( 3 6 )
(3.5) then follows by taking the inverse image of (3.6) in IL,
Lemma 3.5
Let H |
n )
( K ) , H ^
n )
(K) .. .H^
n )
(K) b e a set of general n-point functions.
In order that the product : H
( n
' (K) - T U T ? ' (K) be a general n-point func-
tion it is sufficient that every pair of transverse partitions (I, X\ I)
(J,X\J) be non trivial (at most) for one factor H . and trivial for all
other factors H . j 1...C , j j* i.
139
T he analyticicy inside the tubes jjj <* the coincidence relations are
evident for the product without any supplementary condition. On the other
hand if two transverse partitions are trivial for all factors H . except
for one (say j i), that means :
$><
H
#- >
( P ) =
H
K V
P )
-
H
l ( - )
( p ) v
^ 4
< n _ 1 ) v
J - ' -
j * i , (3.7)
for any set of boundary values a
;
? e
B
i\ (&* ' " t? taken from a
quartet of tubes ST }., , separated by the two partitions (I ,X\I ), (J,X\J).
Now H . satisfies the Steinmann relation :
(n1 <n) (n) (n)
H
i(++)
H
i(~ ) !(+->
H
i<-0 '
then multiplying each side of this relation by
J 1
c
j * i
obviously yields (in view of(3.7)) :
(n) .
H
(n) (n) (n)
H
(++)
H
( ) (+") <"+)
which is a Steinmann relation for the product H
T
'
}
.
Dfini " as 3*c
I n the foil -ig, we shall consider a tree-graph T whose external lines l^
are labelled by the elements of the set X, and carry the respective (complex)
four-momenta K. ( .2- Ki " 0)
1
iex
Definition 3.C.I
We say that (I,X\I) is a line parti Lion .
:
or a tree graph T or a partition
associated with a line of T , if there exists at least one line I of T such
that by "cutting" this line w*> separate ' in two subtrees T j , T
2
with external
lines carrying momenta j K., iSI ; -K- } for T . and !.., iex\l ; -K^ J for
In the following we have to consider tree graphs in which the number n
v
of
of lines incident to every vertex v is larger or equal to 2. Such trees can
possibly contain "chains" of lines and vertices for which n
y
= 2. T hen a "line
partition" (I,X\I) for such a tree can obviously be associated with all the
lines of a certain chain.
140
(When 2 is an external line, one of the two subtrees is the s'ngle line I
and the other is identical with T itself).
Note that every line of the tre T defines (through the above procedure)
one and only one partition (I ,X\l) of X.
Definition 3.C.2
We say that (I ,x\l) is a vertex partition for T (or a partition associated
with the vertex v) if there exists a vertex v of T and a partition (I J \I )
" . - v' v v
of the set of the lines ending to v such that :
I - U I (,v) (3.8)
61..
Here I (P,v) is a subset of X which is specified by the following rule :
i) I f le< is an internal line of T then (I (S,v), X\l(,v)) is the partition ,
of X obtained (as in 4.c.t) by cutting the line &J I ; from the two subtrees
thus obtained the one which does not contain v has its external lines labelled
by l(P,v)U[? .
ii) I f iieZ is an external line f. of T then I ( f ,v) = JiJ .
Note that every line partition (associated with line I ) it at the same time
a vertex partition for T (associated with any end-vertex of ) .
iii) We shall say that a vertex partition is a "pure" vertex partition (asso-
ciated with a vertex v ) when it is not simultaneously a partition associated
with a line (ending at v ) i.e. when |l I Z. 2 and ]c \l | > 2.
Definition 3.c 3
We say that a general n-point function HJ, (K) (K t ) is of "tree type"
if it is a tree-graph convolution in the sense of [7] , which is associated
with a tree T with n external lines.
We recall the definition of such a function H,
(n)
T '
(n
v
) . . 4 (n
v
-
a) With every vertex v of T a general n -point function H- ( k ) (k
V
C
141
is associated ; here X denotes the set of indices which label the lines
incident to the vertex v ; n is the number of these lines. Each line i.(jeX )
carries four momentum K- and K is a set ](f. , jeX i satisfying
b) Let be the partition of X whose elements are all non-empty subsets
K j t v ) of X for j varying in X . (Each element ifiC belongs to one and only
one subset I(j,v), since the external line . of T is connected to v through
a unique path which contains a unique line c: with jSC ) .
4(n-0
4 < n
v
_ 1 )
Let then !_,> be the linear mapping from I
v
' onto C associated
Sr ' v
with the partition 0 of X (see formula (3.1)). In view of Lemma 3.4,
H* ' (i . . (K)) is a general n-point function, and we then define a tree con-
"? (ai
volution product H
v
' as follows :
H
< n )
(K) - T T H
v
(i.(K))Tf fH
(2)
(K^)] (3.9)
1
,elf * k
The first product is taken over the set of vertices of T. The second one
is taken over all internal lines Ci ((I,X\I) denotes the line partition of T
which is associated with the lineu'J; the inverse of the general two point
(2) .
function H associated with the internal line ii, is used in order to produce
2 2
a simple pole at Pj - m in the corresponding channel (I,X \I).
We can use the definition of a completely amputated general n -point function
n
H
V
( k
V
) (see Def.(l.il) . ef I ) for every vertex v :
t "V ) - H \K* ) ~]X W
Z
\K\)] "' (3.10)
A n
The inverse image J-) (i (K)) is again a general n-point function. Moreover,
we associate with every line f. (internal or external) of T a general 2-point
. function H J . ^ ( K _ ) and define the linear mapping i...(K) (from C onto
C ) associated with the partition (I,X\l) (see def. 3.1) ; in view of Lemma 3.4
(2)
we obtain that Hi.^(i>_. (K)) is a general n-point function. We finally write
(3.9) in an equivalent form :
H
(
hc) - 7 T H "
V
(i (K))T T
H
m
( i
m <
K
<
3
-
! 0
>
142
He state the following :
Lemma 3.6
a) Every partition of X which is not a vertex partition of T is a trivial
partition for IP .
b) If two transverse partitions of X are non trivial partitions for H
1
"' ,
they are pure vertex partitions which are connected with the same vertex v
of T.
Proof
a) If (I,X\I) is not a vertex partition of T then none of the functions
H
V
veJK H,l (&< in (3.10) satisfies the condition b ) of Lemma 3.4 ;
in other words (I,X\I) is a trivial partition for every H and H J . W
b ) It is sufficient to note that two partitions (I.XYE), (J,X\J) connected
with two different vertices of T satisfy : either X\I C J or I C J (resp.
J C I or X\J C I ). That means they can never be transverse partitions.
From the above Lemma and by applying Lemma 3.5 we obtain that :
*
Lemma 3.7 *
H G O defined by (3.10) is a general n-point function.
Definition 3.d.l
Let us consider a function H ( K ) of tree type as given by def. 3.C.3. A
vertex v of T is said to be a virtual vertex for Hi ' (K) if every pure vertex
partition connected with v is a trivial partition for H (K).
Definition 3.d.2
A general n-point function H of tree type is said to be of "special tree type"
if at least one vertex v of T is virtual for H " , in the sense of Definition 3.d.l.
Definitions 3.e
3.e.l - Let X be a set of indices with N elements : !x[ = N. We consider a
In [7] an analogous proof of this statement had been given.
143
family F of subsets of X : F j X
a
C X SLSAJ such that fa , one has
either X f) X, 0 or X C X. ; V a e A we put Ix I - n . Moreover we
suppose that there is a subset X. C X and one subset X , cfi7C such that
X , C X
c
and j ^ b j C , .j for every X^ with b )< c ; we shall call X
the' normal subset of X.
3.e.2 - We define an order 2 over the set X of indices and we denote by
Kt
v
' the following ordered set of variables :
K ] K .S C
4
iX with 2 1 K. - 0 (3.
L 1
iex
J
For every a.J\ we denote by > the order induced by ic on the set X of
a
a
4(n -1)
a
indices and we denote by K C_
a
the set :
A
a
K
3
-1 K
a
iSX
a
; with 2 1 K
a
- 0 j
(3.12)
iex
Let n denote the maximal (following 2 ) index of X . We define the following
linear mapping of C *
( N - I )
onto c
( n
a ~ ' :
i
a
(
K
) . f
K
a
- K Vie X
a
\ { n
a
} , K
a
- K
n
j l K.j (3.13)
1 a a
ix\ X
a
J
n
3.e.3 - With every set X F we associate a general n -point function H (K )
4
<
n
a
_ 1
>
in C . We then consider the inverse image of such a function :
X
a
H^(K) - H
a
(i
a
(K)) (3.14)
By application of Lemma 3.4 we obtain the following :
Lemma 3.
H(K) defined by (3.14) is a general M-point function.
Definitions 3.f
3.f I - We define functions which are products (over the family F) of general
144
N-point functions given by Definition 3.e.2 and formula (3.14) ;
f
( N )
(K) - T T H
n
(K) (3.15)
X&T
a
We call them functions of foresftype. The factor H (K) which corresponds
to the normal set X W
(
c X
e
) is called the principal factor of f' ' (K).
3.f.2 - A partition (I,X\l) is called normal partition if :
IflX , i
1
I fix. i X, (3.16)
Proposition 3.1
A function of forest type is analytic inside the primitive domain of a general
K-point function and satisfies the Steinmann relations for every-couple of normal
transverse partitions.
Lemma 3.9
For a partition (I,X\I) to be non trivial for one of the factors H ( K )
(formula (3.15)) it is necessary that either :
or X\l - I
fl
X
a
\ \ n j (3.17b)
Proof
Ue apply directly Lemma 3.4.b) in Definition (3.14) and the proof is obtained.
Lemma 3.10
(ji) . . .
Let f (K) be a forest type function. Every normal partition is a trivial
partition for eve.ry'factor H
1
(K) (form. 3.15) which is not the principal
factor of f
( N )
( K ) (X
a
jt X
c
) .
Proof
Let ft,X I) be a normal partition and let H^
n
*(K) be a factor of f ^ ( K )
145
d i f f e r e n t f r o m the p r i n c i p a l (X i* X ) . W e d e n o t e b y I . ( r e s p . I ) the set
I. ; I f) Xj ( r e s p . I, ; ( X\I ) f] Xj ) . F r o m D e f i n i t i o n 3.e.l it f o l l o w s that
I, /) X
a
- 0 and I, 0 X
a
- 0 so : I <ji X
f l
\ [ n j , x \l t X
&
\ \ n
a
}
T h at m e a n s c o n d i t i o n s ( 3 .1 7 ) a ) b ) of L e m m a 3 . 9 ) a r e not satisfied and c o n s e -
qu e n t l y ( I, X\I) is a trivial p a r t i t i o n f or H ^ K ) .
Pr o o f of Pr o p o s i t i o n 3.I
j
L e m m a 3.10 . e n s u r e s that two n o r m a l t r a n s v e r s e p a r t i t i o n s c a n b e n o n t r i v i a l
p a r t i t i o n s only for the p r i n c i p a l f a c t o r H ( K ) so L e m m a 3.5 for the p r o d u c t
o f g e n e r a l n - p o i n t f u n c t i o n s c a n b e a p p l i e d f or f (K.), to o b t a i n the p r o o f
of the s t a t e m e n t .
!
146
' SECTION 4
- PROPERTIES OF T HE Q ft \7 R-,
4.1 - T he general case
In a given G forest U we consider the function Yv corresponding to a given
yeU(G) (after the "blowing up" operation of sec.2) following the recursion
formula :
Y
r
- If nSSfc-t
a
Y^ (2.4?1
with Y a * 1.. .c the maximal elements of U(y) .
In what follows we shall apply all general results of sec. 3 concerning the
forest type functions on the formula (2 .4 T ) when )'= G , which yields recursively
the contribution of every forest term in equ. (2.49) of R .
Definitions 4.a
4.a.l - L et X , " |l...n.( (resp. X, - [l...n J) denote the set of indices
of external momenta K
T
' of T , (resp K of T g). L et X y U r j ' (resp. (X
2
c/X
?
)
denote the set of indices of internal independent momenta (k
r
K
r
) of Ty
(resp. f; K'of T ) as given by the Q. -assignment (equs. (2.7... (2.10)) ; more
precisely : with the ordered (following Qy) set of lines
rof T y (of T _) we associate the ordered set of indices
X ^
r )
U X ^
0
(resp. X
2
U X
2
>
*P - X ^ U X ^ U X ^ |xj^ |-n
y
|x/-|X*"l -L , (4.1)
( resp. X - X, (J X
2
U X
2
In what follows we sha ll use the following sets of v a ria bles :
<K', k\ ifi - ( K , fcrf , J ^j X< , k f p j - ex?> with
so we write
XW| - n
y
Ix^Nlx^l = L,
, | X, | - n J X
2
| - | X
2
I - L)
Z K.
T
+
Z k+W o
.i'ex<7>
147
<tf * ' ) - ( # i*4
r)
k$> j'ex< , ith I *$ k], -o] ( 4. 2)
I jex<
r)
J
J*!*
Analogous definitions hold for the sets (K.fc.K') (PS,K') when J- G.
We notice that the role of variables K t ' i n section 3 is held now by
the set (K,lc, (C) and the role of the sets K
a
C ^ 'aert is held by the sets
<K,R,fc') , U V ) Vfe i KG) and J'- G .
4.a.2 - Let [ L I ! (resp. L') denote the maximal following i2y (r e sp.a O
index in X,"' (resp. in X , ) . We define the analogs of linear application
(3.13) for every ye u(G) and G :
i
;
(K,*,n -W-ki i^ Kk[V-\, l'l*U\U'
r
) I
Ly LJ. i - 1 Z _ ..*j i^-frt J
' iex, j e x
2
\ x y
;
j'e x^x^''
i
G
(K,k,k') -\\\ - fc i e x
2
, R ? , - fc., i 'e x
2
\ ^L'] -
;
( 4 . 3 )
.1>":X, J
*, " ^r .
+
^-
K
^ <*>
I
For a given forest U we then state the following :
Proposition 4.1
The function F in formula (2.45) isageneral n+2L point function in the
weak sense that means :
a) It is an analytic' function inside the primitive domain of analyticity of
a general n+2L point function.
b ) It satisfies the Steinmann relations for every couple of transverse partitions
(I,X\I) (J,X\J) such that :
148
iflx, i 0 iriXj 1* X,
(4.5)
J flX, i* 0 JflXj i X,
For Che proof of this property let us first show two useful statements
Lemma 4.1
For every subgraph ySU(G) the corresponding fun ct i o n I s(K'k' MJ ) is a
general (n ^.+ 2Lv)-point function of special tree type.
Proof
Fr o m Le m m a 2.9 of the "blowing up" o pe r a t i o n and De fi n i t i o n s 2.26 (for I j)
and 3.C.3 of tree type fun ct i o n s, it follows that Ij(K It K' ) is a general
v il
t
+ ZL
r
-point function of tree type in the space C
l
I 1 of the
i r r'
variables (K' k k
1
) . Notice that a s w e can v e r i fy from formula (2.35), the
corresponding tree graph Tj. (with n-,+2Lj, external lines) has the following
special property : the general n' -point functions (resp. the general 2-point
functions) whi ch correspond to all vertices v e VVC such that vc fl ,f' &)l'f),
(resp. fsC(j'
a
)> y s U j *) ^ are equal to the constant 1.
Lemma 4.2
For each subgraph yU(G) the corresponding function Yv. is a finite sum
of products of Is by general 2Lr point functions corresponding Co Y ,
with YeTJ(jr). Precisely :
5
r
-
f
d'. i i '. cr 'j EpOVxV) 1T
(J)
iJV( /
a
'/) ; <4.
' ' j r
a
e u(y)
T
a '
he r e P^
J
' are polynomials and H_ are general 2L^ point functions of
special tree t ype .
a
Proof
For the proof we use che following recurrence hypo t he si s. We suppose that the
statement holds for every m a xi m a l element Y of U(r ) ; we Chen show that it
is true for Y .
149
a ) Action of (-1 ) :
From the recurrence hypothesis (4.6) and the property of Taylor operator to
act only on the external variables K '
a
we obtain :
(-t
d
^\ i
Y
. I TT V*, ^ & 21 ^
U a
J
' a j a 1j6tl(y
a
) * j a facial M
, * (C
4
JJS P^'V'*"* A (4.7)
' V
a
- 0
ia
From Lemmas 3.1 and 3.3 the derivations and the restriction to K 0 yield
y y*
products of polynomials and general 2L point functions IL, ( K
a
K
a
) ; more
a a
precisely HL are of (special) tree type because the linearity of the derivative
-
a
f
operators conserves the (special) tree type character of I y (Lemma 4.1) given
'a
by formula (2.3 7) ; the corresponding tree graph T results from T
v
whe n
/a
all the external lines of the latter are eliminated, so :
111
^r. (, P
( j a >
) - T
( ! a )
vhfi k
:
i (4.8)
b ) Action of s J*
Following definition (2.42) and property 2.43 a ) w e obtain :
j * ( s
a
* H
T
( utr, krh) - sjr* H J
T
( **, **') ,
- r *
K
( r x i . .
p
<;>
( K
Y
k
r
k
r . , '
(
4. 9)
where p ^ a )
i s
, polynomial of degree Ja in ( K ' R
r
R
T
) .
Finally, b y combining Lemma 4.1 wi t h properties (4.9) w e obtain formula (4.6).
Proof o f Proposition 4.1
We apply Lemmas 4.1 and 4.2 "/- G to obtain :
150
Y
G
<KAfc') -
5
(K,C,k') Y. P
< n
( K , M ' > T T ^ H ^ U
7
, k>") (4.10)
' yeu<G) '
where Ig is a general n+2L point function in 3 V K \ in
a n d e a c h
H.j ( K ' W ) is a general 2L point function in * A r j y
W e t a k e t h e
inverse images of every. IL. by the mapping (4.3) to obtain a sum of products
of polynomials by forest type functions :
?
r
- Z
p < f )
? <
K
'M '> " f H (i,(K,*,k')) (4.11)
G
t
G
7eu(G)
T
/ '
We notice that in the present case the normal set of Definition 3.e.l is
the set JK. ieX.j of external variables of G . We then apply Proposition 3.1
of forest type functions and Lemma 3.2 for the sum to obtain that Y^K.j^ft')
or F
T
. satisfies Proposition 4.1 , when we have a normal forest (equ. 2.47).
On the other hand, application of the proof of Lemmas 4.1, 4.2.c) for I
p
and (-i
d ( G )
) Y
Q
yields that :
T
* -
G
(K.fc.K') Y. ?
(i)
M.k') H<
j )
(fc
G
K
G
) I f ^ '4'
)
((t
r
, *
r
) (4.12)
T
i
G
yeu(G)
f
By talcing the inverse images of fL, H by de f. (4.3) (4.4) respectively and
by applying Proposition 4.1 and Lemma 3.2 w e obtain the annonced result for a
full forest.
We note now that the inverse 2-point functions in formula (2.45)
are analytic functions which change nothing in the previous results so we apply
Proposition 4.1 inside the formula (2.45) for every forest U ; by Lemma 3.2
then we obtain finally for the sum :
Theorem 4.1
For a graph G of general forest structure, the G.R.I.Z. Rg(K, Kl?) defined
after the "blowing up" procedure of G, is a general n+2L point function
in the weak sense.
151
REFERENCES
I). Zimraermann W, Comm. Math. Phys. J5_, 208-234 (1969).
2) Zimmermann W, Comm. Math. Phys. U_ , 1-8 (1968).
3) Speer E, Generalized Feynraan amplitudes, Ann. of Math. Studies
Princeton Un. Press.
4) Epstein H, and Glaser V, Ref. TH. 1400 CERN (1971).
5) Bros J, Thse.
Bros J, Analytic Methods in Math. Phys. New York, Gordon and Breach
1970.
6) Lassalle M, Comm. Math. Phys. 3_6, 185-226 (1974).
7) Epstein H, and Glaser V, Les Houches (1975).
D E U X I E M E S U J E T
j
DEUXIME SUJET
CONDITIONS DE MACROCAUSALIT OBTENUES A PARTIR DES
PROPRITS D'ANALYTICIT EN THORIE AXIOMATIQUE DES CHAMPS
INTRODVCTIOH
La prsente tude a pour b ut d'tablir des conditions de macrocausalit
sur les amplitudes de particules lmentaires (comportement dcroissance
exponentielle dans 1'espace-temps) du type Iagolnitzer-Stapp, en co r r e spo n -
dance avec les proprits d'analyticit globales des n o ya ux des amplitudes
de diffusion, telles qu'elles rsultent de la thorie des champs locaux.
Dans les dix dernires annes plusieurs auteurs se sont intresss au
problme d'une interprtation directe dans l'espace temps des proprits
d'analyticit des amplitudes de diffusion des particules l m e n t a i r e s. On
peut, parmi toutes cf.s tentatives, distinguer des investigations de de ux
caractres diffrents :
i ) Fartant de proprits physiques dcrivant le comportement dans l'es-
pa ce temps des amplitudes de diffusion qui sont universellement acceptes
comme v a l a b l e s, (courte po r t e des fo r ce s, causalit) o n essaye d'en dduire
des proprits d'analyticit. Ce point de v ue a t initialement soutenu par
R. Omnsjcet auteur a propos un e dmonstration de l'analyticit des n o ya ux
de diffusion de ux corps dans les ellipses de M a r t i n , en admettant comme
hypothse de b a se la courte porte des fo r ce s. Un e forme plus rigoureuse de
ce rsultat a t donne ultrieurement pa r Kugler. et Ro sKi e S. Le m m e point
de v ue a t adopt par certains des fondateurs de la thorie de la matrice
S, savoir Stapp, Cha n dl e r , l a g o l n i t z e r , Pha m (voir les rfrences associes
dans la publication qui sui t ). Dans ce contexte lagolnitzer et Stapp ont
dfini un ensemble de conditions de macrocausalit et ils ont dmontr que
ces conditions sont quivalentes des proprits d'analyticit locales des
noyaux de diffusion n -co r ps. Ces conditions de macrocausalit expriment
dans un language mathmatique appropri le fait que dans les processus de
diffusion tous les transferts d'nergie - impulsion qui ne sont pas portes par
des particules st a b l e s, donnent lieu des phnomnes de courte po r t e dans
l'espace temps. Ces proprits de courte porte sont toujours exprimes mathraati-
156
quement par des croissances exponentielles qui sont importantes pour l'obten-
tion de proprits d
1
analyticite.
ii) A partir des proprits d'analyticit dans l'espace des impulsions
qui ont t dmontres en thorie axiomatique des champs pour les noyaux
des amplitudes de diffusion, on essaye d'obtenir des proprits quivalentes
dans l'espace des configurations. Ce peint de vue a t adopt d'une certaine
faon par Williams et Hepp qui ont (pour certaines configurations dans l'espace
temps) obtenu des conditions de macrocausalit des amplitudes de diffusion
directement partir des axiomes de la thorie des champs. Il faut seulement
remarquer que leurs conditions mettent en jeu des proprits de dcroissance
rapide, mais non exponentielle ; de plus et de ce fait mme dans ces tudes
les rapports macrocausalit-analyticit n'ont pas t traite. C'est notre but
dans le prsent travail de dmontrer des proprits de tnacrocausalit prcises
quivalentes la taille et la forme exactes des domaines d'analyticit obtenus
en thorie axiomatique des champs.
L e thorme d'quivalence dmontr par Iagolnit2er et Stapp n'tablissait
pas en fait la correspondance prcise entre domaines (foiuie et taille) et
conditions dtailles de dcroissance exponentielle. Cependant les ides math-
matiques utilises par ces auteurs ont t dveloppes par Bros et lagolnitzer
sous le nom de la transforme de Fourier gnralise et les ont conduit? une
gnralisation du thorme de la transforme de L aplace. Ce thorme tablit
l'quivalence entre le fait qu'une fonction f est valeur au bord d'une fonc-
tion analytique F dans un certain domaine spcifi (appel "tube local") et
certaines proprits de dcroissance exponentielle de la transforme de Fourier
gnralise de f.
L a ncessit d'une telle "g :alisation est suggre par le fait que le
thorme de la transforme de L
4
. la.ee ordinaire est applicable seulement
des domaines invariants par tran-lation dans l'espace rel ("tubes").
Dans le prsent travail nous considrons les amplitudes de diffusion a
2 -* n corps pour le cas des proprits d'analyticit par rapport aux transferts
de moments et les amplitudes 2 corps seulement pour le cas de 1'analyticite
par rapport l'nergie totale (s).
Dans la section 2 nous rappelons la notion de "condition de macrocausa-
lit" d'aprs le formalisme de lagolnitzer - Stapp et nous prsentons deux
versions du thorme de la transforme o^ L aplace gnralise de B ros-
Iagolnitzer qui s'appliquent au cas des distributions.
157
Dans la section 3 nous appliquons le thorme ci-dessus au cas des
amplitudes 2-* n corps en considrant leur analyticit (Ascoli-Lehmann
Martin) par rapport au moment relatif des deux particules entrantes. Nous
obtenons alors une dcroissance exponentielle des amplitudes de transition
dans l'espace des configuranions qui exprime la courte porte des. forces.
Dans la section 4 nous exploitons l'analyticit de l'amplitude 2-*2
corps par rapport au carr de l'nergie totale. L'application du mme thorme
fournit :
i) Une condition de causalit qui correspond 3 1'analyticit dans le
feuillet "physique", plus prcisment dans des sous-domaines du plan coup
de la variable s obtenus en thorie axiomatique des champs (cf. Omns, Bros-
Epstein-Glaser).
ii) Une condition du type "relaxation" qui correspond au domaine d
1
a n a -
lyticit dans le feuillet "non physique" du plan de la variables borddnt la.
rgion lastique (cf. Martin, Zimmermann).
158
Communications in
com m un. muiiLPhys. 4 6 .2 :3 :4 3 (i9 7 6 i M a t h e m a t i c a l
Physics
by Spr i n gc r - Ve r t a e 19 7 6
M a c r o c a u s a l i t y Co n di t i o n s from An a l y t i c i t y P r o pe r t i e s
o f Sc a t t e r i n g Ke r n e l s in Fi e l d Th e o r y
Marietta Manolcssou-Grammaticou
Se r vi c e de P h y s i qu e Th u r i qu c , Ce n t r e d' Et u de s Nu c l a i r e s de Sa c l a y . F- 9 119 0 Gi f - s u r - Yve t t e . Fr a n c e
Abstract. From the analyticity properties or the scattering kernels in
axiomatic field theory' we derive macrceausality conditions for the scattering
amplitudes in the sense of Iagolnitzcr-Stapp. By applying the generalized
Laplace transform theorem of Bros-Iagolnitzer. we show the equivalence of
such conditions with the exact size and shape of the analyticity domains.
1 . Introduction
In the last ten years, several authors have looked for a direct space-time interpreta-
tion of the analyticity properties of scattering amplitudes of elementary particles.
As a matter of fact one can adopt two different attitudes concerning investigations
of this type:
i) One believes that analyticity should be derived from "physically-admissible"
space-time properties of the collision amplitudes; the latter properties should in
particular express the short-range character' of interactions together with a
certain form of causality. This point of view was supported for the first time by
Omncs in [1 ], where a derivation of the analyticity of two-body amplitudes in
i-ellipses, was proposed on the basis of a certain short-range hypothesis. An
analogous derivation was given in a more rigorous form in [2]. The same point
of view had also been adopted by the tenants of axiomatic S-matrix theory [3].
In this context, a set of general "macrocausality conditions" was defined by
Iagolnitzer and Stapp in [4 ], which was proved to be equivalent to local analyticity
properties of the H-body collision amplitudes. These macroeausalhy-conditious
express in an appropriate mathematical language the fa-.'t that in collision pro-
cesses, all energy-momentum transfers which are not carried by stable elementary
particles give rise to short range phenomena in space-time.
W note that in the above formalism (as wellas in [1 ] and [2]). the short range
properties are always assumed to hold in the sense of expmvniial decrease, the
latter being essential to yield analyticity.
ii) Starting from the analyticity properties of the scattering kernels which have
been proved in axiomatic field theory, one tries to derive equivalent space-time
properties of transition amplitudes. This is the point of view which we adopt in
the present work. To the same field theoretical concou belong the works by
59
224 M. Munolcssou-Grammalicou
Williams and Hcpp [6], in which these authors derive certain macrocausality
conditions of scattering amplitudes (for special space-time configuration..^
directly from the axioms of field theory. However in the latter works, macro-
causality was expressed through conditions of rapid and not exponential decrease.
and the links between macrocausality and aualyticity properties of the scattering
amplitudes were therefore not considered there.
In the present work, our purpose will be to deris-e precise mncrocausalily
conditions of the type introduced in [4] which correspond to the analytieity of
the scattering kernels in global regions of the mass-shell, obtained from field
theory. As a matter of fact, the equivalence which had been proved in [4] did not
provide a precise correspondence between the shape and size of analyticity
domains on the one hand, and detailed conditions of exponential decrease on the
other hand. However the mathematical ideas used in [4] ' were developed in [J]
under the name of "generalized Fourier transformation" and a generalization of
the Laplace transform theorem could be proved in this context. This theorem
states the equivalence between the fact that a function / is the boundary value of an
analytic function F in a certain (specified) domain (called a local tubel and corre-
sponding exponential decrease properties of the generalized Fourier transform of/.
The necessity of such a generalization is suggested by the fact that the ordinary
Laplace transform theorem
2
only applies to domains which are invariant by
translations in the real space (namely "tubes").
In Ihe present work we restrict ourselves to considering the analyticity
properties of two-body scattering amplitudes except in the case of momentum
transfer analyticity.where 2-n collision amplitudes are also treated.
In Section 2 we recall what a "macrocausalily condition" means in the sense
of the lagolnitzer-Stapp [4] formalism, and wc present two versions of the
generalized Laplace transform theorem of [5] which apply to the case of distribu-
tions; the proof of the second one will be indicated in the Appendix.
In Section 3 we apply the above theorem to the 2rr collision amplitudes, by
considering the analyticity of the latter with respect to the relative momentum
of the two incoming particles (as obtained in field theory [8-10]). There we
obtain an exponential decrease of transition amplitudes in the space variables
which expresses the short range character of forces. The results obtained are then
compared with those of [2].
In Section 4 we exploit the analyticity of 2->2 scattering amplitudes with
respect to the squared total energy J. Application of the same theorem yields:
0 a causality condition corresponding to unalyticity in the physical sheet
of J, namely in bounded subdomains of the j-cut plane ' obtained in axiomatic
field theory [11,12],
ii) a condition of the type "relaxation" corresponding to the analyticity
domain in the elastic unphysical sheet of s las it was obtained in [133V.
1
The main ingredient of ink method was in fact the use of Gaussian wave packets as introduced by
Omnes in [ [ ] . Such an idea had also been proposed independently by Kroissart [7].
? See for example: Strcater and Wiehtman [14] and [5],
J
For simplicity, we ha\e restricted ourselves to the case of one scalar field producing panicles lof
sinfle type) of mass m.
160
Macrocausaliiy 225
2 . C ausality and Anutj ticity
2.1. T he IS. Maerocausality Conditionson Connected Scattering Amplitudes[ 4]
Let:
SiAM. {j>,}}=<5,., p . - L . i T ,A
Pi
. P)) (2.1)
be the kernel of the connected S-msitrix for a process / - J involving m ( resp. nm)
incoming (rcsp. outgoing) particles with masses m
t
( resp. 'it}) and momenta :
{p, ; f/ , | i | ~m}( resp. {pj-;jei. | J| = iimj) such tha t:
? =" >' : p f > 0 , (2-2)
<>.?=' ?; P ' > . (2-3)
' IiiPi=LwPJ- (2-
4
>
The conditions (2.2). (2.3). (2.4) express tha t the point p=( {p, }( rt) belongs to
the mass shell manifold J l
lt
of the . process l-*J . For arbitrary test-function
<Pj,tpjeSfilR.
1
), with iel.jeJ , the corresponding scatterina amplitudes a re defined
by:
SilV. W) = \ SlifP, }. P}}) rii. / %(P.) [1. J <PUWiPiPt>!Mil>'P>t W
where
t o, =J/ pf+ mf, ( O}= | / J ) | ' + H I}
2
.
We a lso introduce the space E
4
of "multiple displacements in space-time"
=( {( ] , ief; {u'jhje J ) svhere the M,'S a nd }'s are a rbitra ry four-vectors in M in-
kowski space. Let P-({P,;iel];{Pj:jeJ }) be a fixed point in . / / a nd u an
arbitrary displacement. With every configuration ( P. u) one ca n associate in space-
time the set of displaced "trajectories" [in :he sense of classical relativistic ki-
nematics] or the particles (0 (resp. {f whose equa tions a re
J i - i + t , ; ^ ,
, a nd (J being the "proper time" pa ra meters.
A configuration ( P. n) is said to be "norma lized" if (in a chosen Lorentz Frame .)
Il
s
E((M V+ "?)+ /lt"i"">'+;
3
) -l
w h c r e
i=l !'- i' j-(}"".}).
Following [ 4] , we call the configurations ( P. it) causal if the initial a nd final
displaced trajectories form a "causal set": that means, if there exists at least one
multiple-collision classical process with possible exchange of (stable) physical
intermediate pa nicles, connecting il: - initial and final particles. If such a process
docs not exist. [P. u) is called "non causal".
Being given an a rbitra ry normalized configuration iP.u). we say that the
qua ntum scattering process IJ satisfies a n "IS condition" CIP.ti.i.y
0
) if the
following bound holds: V-; with 0<7<W
If niPi). WMlPtl {r i l l *- "- * V" * ' " '
Pr-Pi) n<
u
/'
5
( Pf- 'i
!
)
0
IPf) ''
J
P. lC -. , e-'", 11 + r
v
) (2.7)
161
226 M. Nfanolcssou-Grjniitmtici :
where:
Ie/ uJ , p, -= , . , p . p i = j , j p }; t = M
t
Ci= 1 (rcsp. +1) if fis initial (resp. final).
*({/>,-P, ) is an analytic function which vanishes at />,=?,. lias there a
"critical point of spherical type" (it is strictly positive at all other real points and
is such that the level surfaces #({Pi - P,})=C keep a nested st ructure for sufficiently
small values of C).
The simplest example of such a function <P(which we shall use later), is:
*at,-P,})=L Ap,-P,)
3
- (2-8)

In this definition the bound (2.7) is assumed to hold for every test function %
in y which is locally analytic at P(i.e. analytic :a a suitable neighbourhood of P).
Furthermore, we note that the scattering amplitude" is expressed in terms of a
localizing "sequence" of states labelled by the parameter r, which localize the
particles in the classical configuration when r- cc. The actual displacements of the
wave packet x occurring in (2.7) are x,**m
:
, with the dilatation parameter -"X.
The macrocausality requirements of [4] amount to state that for every non
causal configuration (P. u) a certain condition Qp, u, x, y
0
) is fulfilled (without
specification of z and ;<,).
In the following such precise conditions C(p. u. x. -/
0
) (i.e. with x. ; specified)
will be derived from the analyticity properties of the scattering kernels in the
field theoretical framework.
22. T he Generalized Laplace T ransform T heorem
Let us recall some geometrical definitions given in [5a] and [5b] :
i) Given a fixed point Pin fR", we consider in the complex space C,", ( =p + iq)
the analytic function*
* , ( - ! ? . i (k,-P,P (2-9)
ind call Q, the real open ball:
{p;0S( p-P) =E %.,<j>,-PU
2
<x}. (2.10)
In the following, we drop the subscript Pof*, for simplicity.
ii) Let Bbe a bounded convex domain in an auxiliary -dimensional real
space, such that the closure of Bcontains the origin. Bis supposed to be described
in polar coordinates (s. w) by an inequality of the form t> < Hull In the following
we always assume that 0< r( ea ) < l , 2 ^i for all points ia in the unit sphere j " "
1
.
We now consider the set <f of points k"p+iq such that:
|d+K<e)(Re#{/>+j)-)<0 j
4
We still use this term, although z is nol ntfecssarily lactorized a* in (15).
*> In [5] more general funetions <p. local tubes 7 ^ . and gencraliTed Fourier transforms .,
considered and yield similar rcsuU-.
162
Macrous.i)ity 227
We define a /(iar/ tube T
BOx
with basis Bas the interior of the connected com-
ponent pf S whose edge on the real is Q,; since r(tu)< l/ j/ it is easily checked
that (his domain is bounded.
iii) The #
! m
transform of a function;/ (or distribution) is defined in lhc( n+ II-
dimcnsional space of the variables x=(.V[ xj,x
v
through the following formula:
?
m
(g'Ax,x
0
)=VK)-"
l2
$e-"
:
""g(p)ilp. (112)
iv) Fora distribution / which is the boundary value of an analytic function F
in the local tube T
Bfla
it is convenient to introduce the following set S
B
in (.v. .v
0
)-
space which is called the essential support of f over 0
I
(this name beiag justified
by the theorem belo\v)._S
s
is the convex_cone with apex at the origin, whose basis
is the set of points (.xsB, . v
0
= 1} where Bis the polar set of the basis B. defined by:
B={xeR";x-'to+r"'O for every point {r,o>)sB}. (2.13)
v) For every (a, s i with i '< o, we define ff,(S3J as the space of all functions;;
in ?(J2J.
6
which are moreover analytic in ,- ' and have an analytic continuation
in the closed local lube f
BJ
^- [for instance the latter condition is satisfied if z(p)= I
for p in fi,.].
The following theorem has been proved in (5b). (A slightly weaker version or
it can be found in (5a)J
Theorem I, T here isequivalence between the two following propertiesof the
distribution f:
f) / / i ?
a
isthe boundary value of a function F, which isanalytic in the local
tube T
Kr
ii) the generalized Fourier transform &"{z-f)(x, x
0
) off{p)y.{p) satisfiesthe
set ofexponential bounds:
W<a:\0'<
X
-f)(x.x<M<C
!la
-e-'
,
lXl + M
K
), for ( x. x
o
) #S. , . x
o
0. (2.M)
More precisely these boundshold with uniform constantsC y
a
. in every closed cone
of the halfspace (.vfeOJ tV'""J outside S .
Let us notice that the left-hand side of the "IS. condition" (2.7) C(P, , a, ><,)
is precisely the generalized Fourier transform &* of:
n
i
( 2 tu, ) -
1
p, -p
J
) T,
J
( {p, }
t o
, ; {p-faMM. (?'/}) (2.15)
evaluated at (.*tit. x
a
=yt).
Therefore I.S. conditions (2.7) have exactly the same form as the inequalities
(2.14) of the above theorem: in view of the latter, there will be equivalence between
appropriate sets of I.S. conditions and the analyticity of the T
u
kernel in corre-
sponding local tubes.
In the following, it will be convenient to make use of the analyticity of the
scattering kernel T
( J
, with respect toa subset |p,)of the energy-momentum variables
p. According to the above theorem, this "partial analyticity" yields conditions
of the type (2.14) written in terms of a "partial" &* transform corresponding to
the variables p, .
ff(QJ a the space of C" functions with compact support in i ?
r
*" The closure of 1?,..
163
228 M. MunolcssoU'Crammaticou
However those conditions can a'so be equivalent^ formulated in terms of a
"total" ?* transform, i.e. with respect to all variables. Indeed, two equivalent sets
of conditions can be given, either one of which expresses the fact that a distribution
f( Pt,Pi) '
s
(he boundary value of an analytic function of p, in a fixed local tube
71, when p
2
list the neighbourhood of an average position P,. More precisely.
we can state:
Theorem 2. Let fbe a distribution in the space of the variables (JV p
2
) defined
in an open neighbourhood of the set :
{( Pi, PjleR" x R"'; p,e Q,;
P l
= P
2
) (Z16)
S2, being the open ball: {p, ; (p, .P,)
z
<}.
The following properties of f tire equivalent:
i) For every positive number a' < a. there exists an open neighbourhood ui of P,
in R", such that: V< peS{( o), the distribution:
/, 0>i)- fiPi.Pi)fz)i>Pi " tymed in fl,.. (2.17)
Moreover, it is the boundary value on il
t
of a function F ir/nVA is analytic in the
local lube T
M l
. [n-irfc *( p, ) - ( p, - /
,
l
)
1
] .
ii) For every (a', x") with 0< x' < x" < x. there e.vfsre ait open nehjhbmirhimd of P,
in V!" such that:
V&e^-tJV), V( pe3( to), the "partiar generalized Fourier transform:
*w(Xr/,XX|.*o>
-(2n)-'"'
+
""
2
Je-""'-
,
"-"
1
/(P.P
J
te.(Pi)<P(P2Mp
1
''P: (21S)
satisfies the following set of bounds:
VN>0 lF
r
,
n
taf,)( x
l
,x)\iC^4< pUl + l< J''>-
,
'-'^" (2-<9)
valid in the region:
{{
Xl
,
X< >
HS xgO}. (23)
Jn (2.19) HI is a fixed integer which expresses the order of the distribution f
with respect to the variables p
2
.
The semi-norm Sp| is defined by:
M sup |DW,)|
1JS "
"
tlie constants C
v
can depend on x\ a" HI/ Hx*1|.
iii) For nwy fc*'.30 with 0< x'< x"< x and every XiipJ analytic in a neigh-
bourhood of P, with z
2
(p.)*0 r P j - F j . Xi( Pi) "* '" '9 ''"' """' generalized
Fourier transform:
*m( XfZi-fl( x,.x
1
.x
0
)
-(2ll)-
,
"'*
1
"
2
J<;-""*
,,
'"*"
|(
""''
,
'*
l
""'"
i|
'
M
/(P^PJ)ZllP)^Jl/'il'PM';:
(2.21)
164
Macrocausality 229
satisfiesthe following setofbounds: W> 0
\&J X, *, /) ( *!. x *o)l S C,
%
<1 + |.v,l|-J(l + |.v, | | T '- (2.22)
wi/iV/ m f/ie region
{lx
l
,x
!
,
Xa
):ix
l
,x
<>
)iS,;x
a
iO;x
!
eW} (2.23)
fivi/i m the same asin )].
The proof of this theorem is given in the Appendix: here we shall rather
indicate its physical interpretation when the distribution ./' is chosen to be the
scattering kernel of a certain process: P, (resp. P,) can for example refcr to the sot
of relative (resp. absolute) four momenta of a first (resp. second) group of particles.
(These two groups can be those of incoming and outgoing particles as it will be
the case in Section 3, but this is not necessary.) The set of conditions (2.22) is a
set ofmacrocausality conditions of the type IS.; each of them requires the localiza-
tion of all the particles near the average configuration l P
(
. .v^'llx, !!), [P
2
. x
z
/l\x
2
;
D-
Theorem 2 essentially says that theanalylicity of the scattering kernel with
respect to p, in the specified domain T
m3
. is equivalent to a precise exponential
damping of the scattering amplitude with respect to relative space-time dis-
placements of the first group of particles.
The equivalence of statements ii) and iii) means that for the short-range
behaviour to hold it is immaterial to consider the second group of particles either
ina localizing "sequence" of states in the sense of i.S.. or in an arbitrary fixed state tp
with momentum spectrum localized in a neighbourhood of P
2
.
3. Analyticity with Respect to Momentum Transfer and Conditions
of Short Range of Forces
We consider the scattering of two initial and ;i final identical scalar particles with
mass m. Let us denote by r={r
i
...r,} the final four-momenta which satisfy the
mass-shell conditions:
r j - m\ i}
o >
>0 y=l a . (3.1)
Wc shall represent by i| =p
y
+i?i, i - 1, 2 , f.',eC* the corresponding momenta
Tor the two initial particles.
The total energy momentum conservation:
kt+k
2
"J ^ .
l
r
J
sr <
3
-
2
'
leaves us with only+1 independent four vectors. We choose the set
{knp+ia^ -kJ IZ; r, j=l }
where k is the relative momentum of the incoming particles, which is considered
as a complex four-vector. One also puts:
r
2
-s (3.3)
(squared total energy).
165
230 M. ManoWssou-Gnimnmicoii
In the context of axiomatic field theory [S. 9] it has been proved that hen the
rj,;'= 1 are kept fixed and real, the scattering kernel T(p; {r,}) is the restriction
to the physical region of an analytic function H[k) inside the Jost-Lehrrunn-Dyson
domain of holomorphy D 153* The latter has a non-empty intersecrion with the
complex mass-shell manifold .If (which is a two-dimensional complex sphere):
u r - {
t e
c
s
*Zf~
S/
*} W
This intersection DnJJ' is of spherical type, described by:
or(equivalen.lv): l ' J < J ; _
r a 2 + c
, } (3.5)
In a more rigorous way, T(p, r) is a distribution in the variables r= {rj which
is analytically valued in k; it means that for every test functions 1~L tpjfjtelAW)-
the function:
T{< PjHk)'ink.r)Yl< i>
t
{r^r
i
/2a
i
, < Oj~\^fi^? (3.6)
is analytic in a J.L.D. domain of f.
We shall always choose the functions e>, so as to have their support in an
arbitrarily small neighbourhood of a fixed configuration (r,=RjV of the final
momenta. Then, the corresponding domain Dn.ff' of the function 7~{</>jH) is
defined by the constant:
C , - 3m
2
/ l - s (3.7)
where the number e>0, takes into account the size of the support of the wave
packets \\j Vi-
The scattering amplitude for the process A, +.-l---B
1
+ ... + B
>
when the
initial (/.|,. 1,2) and the final {Bj,]~ 1 n) particles are translated by the four
vectors .* x x'j, /-= 1,..., it respectively reads :
S p}, {#})(.<, . t}*)- JI1..J Vj'tr^fpD-f-"" * ^^'^-*
I
"T(p, {/j})
S( prWp'+r
1
J4-nt)< lpUj.,'l
1
r
J
/2ioj (3.8)
wi t h. . - , . . - . . }
r^- ai j - m
(r
J
,p)6./r:jpr0 (3.9)
for arbitrary test functions f]
j

>J
(r
J
)
v
,(p)ei?(.//T with a suitable parametrization
of.//" described in the following).
In (3.S) the energy momentum conservation p, +p. = YJ- I *t
n:ls
been taken
into account besides the change of variables:
" lx.+xJ/2-X, . T . - . X, - * . (3.10)
166
Mncrocausality 231
By using Theorems 1 and 2 we shall be able to write conditions C(P. , x ;) of the
type "IS." for the scattering amplitudes (3.S); these conditions will be equivalent
to the above analyticity domain.
The average momentum configuration "P" is here specified by fixing:
r >
=K j , r=R, Jt=P(real) with P = 0 , P
2
- m
:
- R
2
/ 4 -
It is convenient to choose axes of coordinates such that:
R=( R
IC
", 0) , />=(0,O,0,P
< :
"), k=(k",k">,k'
2
\k<) (3.11)
we also put:
*=( #". itf
11
) {k transverse)
The space time (normalized) translation ""*="-v"/Il"-^"S is here specified by-
fixing the set
V= W e J ; X^ +xJ /2; . x. =. x
1
-x
2
=( 0. . v , 0) }. (3.12)
The tsvo dimensional vector x is the "impact parameter" with respect to the
classical momentum configuration {R. P).
We shall use the vector variables k
u
(resp. p ) as a parametrization of.//'
(resp. .W). With the above choice of "x" the matrix element (3.S) can be written
more precisely:
S({>}, {>}})(*, { * } } . * )
-iUl.jVni-'MP^ -^ '^ '-^ '^ ^ np.ArfiUj.ji'P^ oy-p,, (3.13)
where <pe3(R-) has a sufficiently small support.
We now construct a maximal domain in .11' whose projection in the variables
k will be a local tube with real open set Q (centered at P) defined by:
a =0>: Otf(jJ)-P?,<:t(s)i- (3.14)
When p
u
varies in 12 the point p=[p ;]/3--pi) varies in an hemisphere of. //'
(with pote P).
In an auxiliary space with dimension two, we define the basis Bof the local
tube 7i , as the set of points ((?. <o) with Q < r(o).
Then (see Section 2 and [5]) 7"
Wl
is the union for all values of g < r{ai) of the
bounded components (with edge cQ) of the following "cycles":
r,:{k,
r
=p
lt
+i
u
eC-
to,,l+<Klft,f-k,l
2
-*)=0;
G, -I< h>. M
2
<*}- (3.15)
Such a definition is meaningful provided that rfw) <l/2 |/ . We shall now fix
t[()r
mM
(independent of <o) in such a way that T
M
, be maximal and contained
in the axiomatic domain Dr\. II'.
Looking for a local tube in which the maximal value of |g| is the one given by
the domain (3.5) [or (3.7)]. we must impose that when | j>|=0
" b | " I Af - ( l / V- ) "
2
S C , . (3.16)
167
252 M. Manolessou-Granmiaiieou
One then easily verities the following two possibilities:
i) for energies J/s such that: C /<a, the inequality (3.16) is satisfied when
ii) for energies sud. t^t Cj _a the condition e_ V-V
1
is sufficient for (3.16)
to be true.
Wc conduis that th> laximal local tube T
Ma
with image in .//' contained
in the domain (3.5) (or (5 ' (where T(j>.r) has an analytic extension) admits as
basis:
-"reCAc;**) for c,<\fi.
( 3
-
1 7)
Let us now choose arbitrary functions -/&p
u
). YljX.'MJ
1
J~
l
with the
appropriate properties described in Section! and consider the generalized
Fourier transform:
^*f c {*;))(*.*<>)= Je-'
4 w
e-""*"rO.. {rjlMft,) n ^ W V ^ ^ ,
(3.1S)
Following Theorem 1, f% satisfies bounds of type (2.14) for {. V. .
U
#S
S
): where
S , is the convex cone in the space {x
u
. x
Q
) with apex at the origin, (with .\\>=yl|.vil.
yconst) whose intersection with the plane x
0
1 is the polar set S of B (3.17):
i =
L . W 2 l A for C , ^1
r-| | x
1
, | l _(C i +
I
VC, for C,<1^J-
l 3 I 9 )
Let us suppose that in the expression (3.13) of the scattering amplitude the wave
functions are gnussian type wave packets of the form:
V'fTJ>='^
,
-*'*""&>% < eM=e-'
hT
M- P-20)
Application of Theorem 2 then leads to the following bounds of the scattering
amplitude:
M< P), {< l>'j}) ( x.{Xj}, X)| S j
l C i
(3.21)
l c
K
r^"*"'
!
/ ( l + ll.v|-
v
); C,< \^)
V/V. V(
X
,
nX< 1
HS( X,.yeR."*"
We notice that the intersection of S
s
with the x
0
0 plane is only a point, that
means, the origin in x space.
The choice of P (centre of the local tube T^J being arbitrary on the sphere
./r (3.9) we can repeat the same procedure for all points P in ./<". It follows,
that by taking the union of all these local tubes, we reconstruct the total domain
of analylicity (3.5). The set of conditions of t\ pe (3.21) obtained, represent then the
"IS. conditions" corresponding to this domain.
The directions .v for each of these conditions, lie always on the plane ortho-
gonal to each vector P; it follows that the bounds of type (3.21) are valid in every
,* Remark: T te rigorous application ofThcorem 2 saj* that a has to be replaced by some i ' a - t :
wiihc-*0 when the size of the support o f ? , contained iu R? lend to zero.
168
Macrocausality 233
direction in R
3
. The exponential decrease of the transition ar.plitude is thus
expressed with respect to the relative distance of the translated trajectories of the
two initial particles {or impact parameter) independently of the translation .Y + .Vj
of the final particles. The latuT behaviour expresses the short range character
of strong interactions and we have proved that this behaviour is equivalent to
the J.L.D. domain of analylicity of the scattering kernel, with respect to the
variable of relative momentum. The precise rate of exponential decrease is given
by the size of this domain.
Because of the invariance under the Lorentz group, the analytic extension
F(k, r) of the scattering kernel defines an analytic function in the space of the
invariants k-r
s
j= I n (the invariant k
1
being a function of 5 in the manifold. //').
In the case of the scattering A^ +At fl, + S
:
(two final particles), the re-
maining independent invariant is A'-fi^-r,) (at fixed total energy) which by a
linear transformation is related to the scattering angle r = cos0. The image of the
domain Dr,.//' in the space of this variable is the Lehmann ellipse [S], Taking
into account unitarity Martin [9] obtained a larger ellipse whose inverse image''
is a spherical domain analogous to (3.5) (or (3.7)), with corresponding constant:
Cl
t
=m
2
s/(s/4-m-)>C;. (3.22)
What follows then is that "I.S. conditions" of type (3.21) equivalent to Martin's
ellipse may be obtained in the same way as described above. By inserting the value
of the constant C
v
(3.22) in the formulae (3.21) we have explicitly the following
bounds:
a) For 4m
2
Ssgnr(l2 + S[/2)
l W} . MIX-*,,, {A.}). -Y)l
S^( l + ||. v||")-' exp[-( s/ 4- r) "
2
| | . v l l / 2 ] . (3.23)
b) Form
!
( 12 + Sl / 2 ) Ss < +
|S(fo},{<pJ})fe {*,}.*)!
^ d + IWI")-' exp[ -
m
( s/ ( s/ 4- m
J
) ) >'>V( V4- r )
3
+ i r ' IK II] (3.24)
VJV; V{.v,.v
0
#5,.(.V..x;)elR'
l
'"
|,
, W = {|.v|
5
+ . t 5}'".
One then verifies that for j - x the exponential decrease reads:
| S( { p} . W} ) l , - , <c. ve-
M M
/ ( l + a*!|-
v
) (3.25)
which expresses the physical idea that for very large energies the exponential
damping is independent on the energy.
The corresponding bounds (3.22) for the case of the Lehmann ellipse give for
s- oo:
|S({p}.{v}})l,-:.v/(I + l|.x||-
v
)exp(-(const,y7)i|.v
lr
||-.0). (3.26)
We should expect these different results as a direct consequence of the qualitative
difference between the two domains of analyticity: the Martin ellipse has always
fp>x""4 r while the Lehmann ellipse shrinks to zero as f,_
x
~onst,'s- 0.
* In fc-jpacc.
169
234 M. Manolcssou-Grammuiicou
Concerning the behaviour at the threshold energy s- 4;ir, it is, as one expects.
similar in both cases: the rate of exponential decrease vanishes as:
~l / 5- ( s/ 4- ni
2
) "
2
. (3.27)
Some Remarksaboutthe O.K.R. Formalism
In the works of Omns. ICuglcr-Roskies. Finley [1. 2] , one starts from the short
range hypothesis expressed in terras of exponential decrease properly of the
scattered wave for any production process (a). Such a property is immediately
rewritten in terms of the extlusive absorptive part A
a
of the two-body scattering
amplitude associated with the channel z as follows:
J/t
1
( ,p-p)jJ(rtm,<m
(
,.<-
2
" (K.R.)
with s>
u
( p) =( / ui / n)
w
e-"-"
! ,
''
5
e-"" similar to our ipc'"-' in (3.31.(3.20).
The parameter a is the impact parameter (in semi-classical approximation)
and the rate of decrease a in this formalism is justified from intuitive potential
theory considerations. By exploiting conditions (K.R.). these authors derive
analyticity properties of the absorptive parts A
r
The analyticity of the scattering
amplitude is yielded afterwards by unitarity. In these studies the principal scope
was the understanding of analyticity properties on the basis of the short-range
character of the interaction in spam-time. It prefigured in a special situation the
general S-matrix approach of lagolnitzer and Stapp [3. 4] .
Our method and results differ from those of O.K.R.F. in the following respects:
a) The method of the generalized Laplace transform allows us to state a
precise equivalence between the size "(and the shape) of the analyticity domains
and the corresponding rate of exponential decrease in space-time.
b) We applied our method to the derivation of IS. conditions for the (2n)
scattering amplitudes. However it could have been applied as well to the corre-
sponding scattered waves (or ^tj and would liave then led to inequalities of the
type (K.R.).
c) In our formulation, the fixed total energy is not necessarily fixed as in (K.R.).
Testing with wave packets in all variables, including the total four momentum.
is a little more satisfactory from both (physical and mathematical) points of view
(see the general I.S. formulation [3.4]).
d) Concerning the physical interest of the above mathematical equivalence.
we preferred to choose the point of view according to which space-time asymptotic
properties can be derived from the analytic )' properties of field theory, rather
than the converse. In fact, without discussing here the general motivations which
can lead one to prefereithcr the field theoretical approach or the S-matrix approach
to the analyticity properties on the mass shell, we can make the following remark:
in the S-matrix approach it seems difficult to justify with physical arguments a
precise "quantitative" formulation of the macrocausality conditions lincluding
all types of "short range assumptions"). For example in [1. 2] the rate of ex-
ponential decrease a which was chosen was only justified by an analogy with
potential theory. On the contrary, if one accepts the global analyticity domains
of axiomatic field theory as a firm basis, the above method yields a derivation of
"quantitative" macrocausality laws in the field theoretical framework.
170
MaCTOousatity 235
4. Analyticity with Respect to the Toial Energy
4.1. Causality Conditions
Let us consider now the scattering
Ai+Ai-^ Ai + Aj (4.1)
of identical scalar particles with mass m. The four momenta p, , pj are real and
satisfy the mass shell conditions: p
2
=m
:
, pj
o l
>0, i = 1,3 while the corresponding
momenta for A
z
. A particles arc:
ifc
2
, /(
4
C .
The total energy-momentum conservation: p, + l -
2
=p
3
+ J:
4
'leaves us with three
independent four-vectors: we choose the following ones:
r _ P l P l
W k^ +kJ ft^ Hk^ p+iq).
l 3=P l +P 3j
(4.2)
[p-r=Q
The mass-shell conditions expressed in terms of those variables are the following:
J4p
2
+ r
2
=4i
2
r
2
+ r
2
3
=4 i
2
We suppose that the wave packets {<?} o all the particles have been translated
by corresponding Tour-vectors x
it
i = I,..., 4 and we define:
^- f Xj + x . R I - X J - X , .
The scattering amplitude for this process then reads:
i(rfj+r
J
-4m
2
)
1
(pr)'(r
1
3r)(5(4p
2
+ r
2
-4ra
2
)</pi/r
l
j(;r (4.5)
where we have put {0}(p. fu, r) = {(p}(Pi, p*. p
3
. PJ) [with (p, i^
3
, r) satisfyir ; ( :.3)]
Let us now define the average momentum configuration "P
n
of the scattering
process by fixing:
'13*^13. r =R. t =P( r ea l )
p. R=0 / >
2
=m
2
- i
2
/ 4
W , h :
i !. R
l 3
-0 ' + , - * , '
( 4
' '
{} is then assumed to have its compact support containing [P, R
X5
.R) and such
that its projection in the space K^^, lies in an arbitrary small neighborhood of R
iy
Wc now choose for convenience axes of coordinates such that:
Ru-( R. O)
P m ( p a p ) ( 4 7 )
R=( 0. P, ) '
171
M. Maoolettiiii'Gntmfflattcut!
Fig. t* Space time displacements. The relative translation { lies inside the light cone (sec text)
and we restrict our set of "normalized translations" "iT=".i:'7||".v"||, with ".*"=
( Jf. X. f. l) to be of the following form: t t
1
Xj , i
1
s , with ( x
t
- x
4
)
I
>0) or:
X= X= ( r
0
> . , =- | =K " , {) (4.8)
wi t h^
0 J
>
!
.
Such a choice of *.x" ensures that the trajectories of the two final particles
intersect each other at a point which lies in the forward (for p
m
<0) or backward
cone (for "">0) with apex at the intersection of the trajectories of the two initial
particles. In this subsection, we stick to the case if
01
< 0 (sec Fig. 1) whose physical
interpretation will be a test of causality.
(The detection apparatus, represented by the average trajectories of the
outgoing wave packets should not detect any scattering in this situation.) With
this choice of "*.*", the scattering amplitude (4.5) can be rewritten:
Now using the Lorentz covariance of T. the integration over rfr, j and <tt"' can be
performed and yields [with a new test function {<,} (p
101
.p.r<?}, r) ] :
S( {r t ) ( f l - I[ T(
Vl
}] (
?
', p. < J. r ) f - W"- "
J

, , , :
'< 'O>r )
S(4p
}
-r
1
- m'W - r
1
-Am')dpi^ lr. (4.9)
Let us introduce the usual invariants:
j - ( Pi + k
2
) ' =( Ps+W
!
-( ft
3
/ 2 + k)' : squared total energy in the C M . system.
l ( p
1
- P3)
,
- ( f c
4
- J . j )
1
r
I
: squared energy-momentum transfer
C M . system. (4.10)
In if , the complex mass-shc'.l maniiold. / / ' is defined by:
-0
HEf-*}
(4.11)
172
Macrocausalily 237
Taking into account the definitions (4.10). (4.11) \vc notice that the variable js is
related to the four-vector ke./F by the following mapping IT, (rjj and r being
kept fixed):
ke.//'-**s- 2m
2
+2J/r - r/4 J.<' -1/ 2. (4.12)
It has been proved inRcf. [11] that when r, r
1 3
are kept fixed, real, and such t hat
, 0
:
- 2Si
2
gr S0 (4.13)
then the scattering kernel T(p. p, r. rf
3
) considered as a function of k=*(k, k)
has an analytic continuation in "' its analyticity domain " in.//' is the inverse
image by s, of the s-cut-plane with cuts se[4ni
2
, +ao) se[r, -co) and r fixed
in (4.13).
It follows that the intersection of the manifold. /F with the domain of analytic
continuation of the scattering kernel T{k
a
. k. r. r\
3
) when r, r^
3
are kept fixed, is
represented by the J--cut-plane with the following cuts:
p
, 0
' e r>'+r/4)/(i= - r/4) ' , +00)
(4.14)
p""e[-^(m
2
+r/4Wm
2
-r/4)"
!
, - co) .
Let us call Po' = l /r -r /4t hc threshold of the physical region of(4.1). We shall
exploit partially this analyticity domain of 7* by considering the maximal local
tube inside (lmfc>0} which is bounded on the real by the open interval =
lPo'.pS'l whos
5
nier is P.
More precisely, let:
f"=(p<? + P<?)/2, -pS' -po
0
' )/?.- <"P
,
)=(P
1 0 ,
-P
l 0 ,
)
2
(4.15)
fl= {p":0g $(p<
m
- f " ) < a}.
From the equation of "cycles":
*><>; q><+e{(p<>-F>Y-
l
""
2
-*}=0\
r ( 4 1 6 )
q""+S{(p""-P"")
1
-q""
2
-^=0\
"+tq", - (p"-P")
!
-<0 I '
we conclude that the basis B of the minimal local tube T
M
, is defined by:
B={
e
:Q<Q<l/2]fi). (4.17)
7J, is the bounded connected component '- of the union ofcycles" f such thai
QBB. By taking arbitrary functions -/^{p). j;,(p
( 0>
). Zj(rfi). n M with the appropriate
i a
Here wc -.lick to the case when no pseudo-threshold exists in any channel ICor insuna* the . T-rr
scattcrin!-).
" Mow precisely this is 3 consequence of Ref. [II] and of the com plex Lorenu invariance ot the
anah/ttcity dom ain of the four point function. Kef. [12].
1 1
Whose cdfc on the real is ft.
173
M. Manolessou-Gramiinuiuou
5T5o-
Fe. 2a and b. Essential support of T(/>
t0,
);fe(i>
w
(see text), (a) For the analyticity on the first sheet;
(b) For the annlyciciiy on the second sh-^t
properties described in Section 2 (in particular y.
z
analytic in Ti
e
J we define ttie
following generalized Fourier transform ^J, in the space:
'"'.XoJeiR
2
(4.1S)
- J T( p">\ p. r>ft r) rif-1 *
/i
e-"
,

,
?"->< "-'< "
H^i
1
~r
,
-in^)( 4p
1
-r
1
-4>,r)( p-r),lp'>ilpd
1
'^ilr. (4.19)
The essential support S^
10
'. x) of I"(P)A'1'
0
) is defined as the convex cone in
({"" *o) space with apex at the origin, having as equations .
{*

-2^'
:
*
0
}
whose intersection with the line x
a
m 1 is the set (Fig. 2a)
B - K
| O ,
: ?
0 ,
S _^-
(4.20)
Application of Theorem 1 then provides bounds of type (2.14) for
7
^, whenever
({"", x
a
)iS,. If we further suppose that the wave function {<(>,} in (4.17) contains
Gaussian type wave packets, precisely:
{9i}(P.r
l
ij.'-)-4'i(l')<Cj(p"")<(>
1
(r',>4(r) (4.21)
with:
< P
i
( p)-t-"-"''"^p)
< p
1
( lf
m
)-e-"
m
-'
<
">''z> ;"")
9Jti)-r#zJ.r),
Theorem 2 then yields the following bounds of the scattering amplitude (4.9):
_ |S(v}(<)I Sc.
v
*-
| 5
""
l , n
/(14-[
1
||) (4.23)
" VW and K"".x)#S {
6
&V U . J- K""' *- ^] "
1
.
(4.22)
174
Macrocuusality 239
oTO.0 ,1,0
p'S" P p'S,' p"? P A ' P'S
^o\y ,-uo \^<y
a b
Fig. 3a &nd a. Local tube T
t
on the second sheet (see texl)i (a) Without pole: {b) With n pole inside it
When :co=0. S, contains the positive i
1 0 1
axis, that means the bounds (4.23) are
valid for all " " < 0 .
We have to remark that the only space-time variable which contributes to the
exponential bound is the time j '
0 1
(conjugate variable to the squared total energy s '
with respect to which we have exploited the analyticity of the scattering amplitude).
In conclusion, we can interprte the bounds (4.23) as giving a precise expression
or causality (in the usual sense). Qualitatively it indicates that the probability of
the scattering process .4, -T -A
2
->A
3
+ AI decreases exponentially with respect
to the negative time-shift of detection of the final particles with respect to the
initial ones. The constant in the exponent of (4.23! corresponds to the size of the
bounded analyticity domain we have considered in the half-plane lms>0 for
s>4m
2
. Taking more and more analyticity into account, the exponential decrease
is more and more rapid and approaches in some sense the classical limit (i.e. an
.infinite rate of decrease): as a matter of fact, due to the space-time spreading of
wave packets, one could not expect a "sharper" expression of causality in the
quantum case.
42. Relaxation T ime Conditions
Under an additional smoothness, postulate, it has been proved from axiomatic
field theory [13] that the two-body scattering kernel has at least locally an analytic
extension in the second Riemann sheet of the. j-plaue. when the real part of s is
in the elastic region
l 3
|4 r $s< \6itr). We can then consider a maximal local tube
on the second sheet: i/ < 0 with |(/| < C (see Fig. 3).
This local tube will be defined os a union of manifolds r
s
with end points:
j
0
4nr t, iI6 r. The constant C will measure the "distance" of the nearest
singularity from the real axis. (More precisely: C indicates to which cycle r
t
the
nearest singularity belongs.) The equation of "cycles r," for this case reads:
|n""l+eC(p
, <
"-f"")
=
-""
2
-*']<o
4<
0 ,
< o to
10l
i<c
where the constants: *'((/>'-/>)/2)
:
, P'-,j>2'+riJV2 and /.J,'. p
0
correspond
'o the valuess= I6ur, j4m
:
t respectively: more explicitly from (4.12) we have:
ye?( 3wi
J
-K/4) , '( r-r/4) "
2
P"
l
'-4i n
!
/ ( i n
1
-r/ 4) "
!
-2Snri0. (4.25)
The conditions ensuring the boundedness of the domain (4.24) define the basis B
of the local tube T ,#,- and one verifies that:
B = ( fl:0?< C } C -( C / ( C
2
+ i ') . forC
:
< 3l , [/ 2 ^7( rorC
:
Sx-) }. (4.26)
Here we considered the rase of an even theory.
175
240 M. Manotessou-Cmmnialicou
The polar set is then given by:
B=K
, 0 l
:< ""l / C }. (4.27)
By analogous considerations to those of Subsection 4.1 we then define the essential
support Si of T {p")yJ ' . ' Fig. 2b) as the convex cone in {.;"". .\) space with
x> C
m
; . x
0
O. Application of Theorem 2 then leads to die following bounds for
the scattering amplitude (4.9):
\S{<pHaZc
s
e-'
e
w',V + Kill") (428)
VW and for (c?
m
, x
0
)$S* eR'. K , || - K
< 0 > 2
+ . x
:
}"
2
.
We obtain consequently (for . \
0
=0) an exponential decrease of the scattering
amplitude for all times
l o ,
>0.
This property expresses the fact that the system of interacting pa rticles A,. A
:
can produce the final particles A
2
, A
t
during an a vera ce time which is bounded
by 1/'C.
According to the above result we can therefore say that the annlytiuty of the
scattering amplitude in a domain of the second sheet whose size is measured
by C (Fig. 3) is equivalent to the existence of an upper bound 1/s'C on the relaxa-
tion time of the "composite system" due to the interaction of A, and A
z
.
We cap refine the above result in the case where the extension of the two-body
scattering kernel on the second Riemann sheet is known to be meromorphic in a
bounded domain, with a single pole inside this domain at a point 5 =S
t
- i T.
T>0. We can separate then the pole contribution from an analytic background
in the following way:
W \ p. r, r<) - Ti(p" j , . r, rft') + Ttf< , p. r. r<J) (429)
where:
T(p'
<
'
,
,f,r,rffl)-i(p.r,r','S)/(p'-
l
,g'+<')
-[2| / m
2
-r/ 4(s -s -HTB-
l
6. (4J0)
6 is a constant '* and T J Lp,p, r. r
l3
) an analytic function inside the maximal local
tube (hat we can construct as previously discussed. That means that for <;<0
we have: | g
0
| < C
t
(see Fig. 3 b) and the constant C, determines again to which
cycle r, the nearest singularity (apart from the pole) belongs. Similar equations
i s (424,...,427) then hold with the replacements CC\ and the generalized
Fourier transform J ? , reads:
J V. - ^ + ^ m - ( I)
The explicit evaluation of the pole contribution gives:
^junaH*""-*.)
_
cnst e
- ) ,
, l
. i
, l
-3r ) f'-f<"n!f''e-l'if
,
-'"
l l
'-'
0
"I'<'-
,
'S'i"". (4.32)
The scattering amplitude (4.9) can be separated in the same way:
sfoKa-s'toKa+s'toHO. (4.33)
* I n^swct.
176
Macrocuusality 241
S* satisfies bounds of type (4.2S) and on the other hand one verifies that:
for|/)'-P|||,,f|:|S "(^l(s
r
'
l , ,
.x
0
)l*cnstf-<
,
-''''
,
'
,
'
f
-I CI li''',
s
=i'>i
>
0
Wf - I * ! *
0 1
! : |S*(
V
}(i
, 0,
,.Vo)l=cnst
c
-<'i:""; i"">0
I rff'-fl^lA
0
'!: |S"{e.}U=
l 0,
..t
0
)|s:cnst
e
-l'SW'"-'l<
1
ii
i
(x' ' - 4"",
< o l
>0). (4.34)
The last case in (4.34) expresses the fact that the "relaxation time" contribution
of the "resonance" is limited as the isolated pole approaches the boundary of the
local tube.
JcknowlcdfinHnrx. This work has been carried through under the direction of Dr. J. Bros. The
author has profited from his constant encouragement during the completion of thw work whose
present form is mainly due to his most helpful advice and suggestions. She wishes to cypress him all
her sratitudc.
She is also very grateful to Dr. D. lagolnilzcr for many fruitful discussions and illuminating
informations.
Appendix
Proof of Theorem 2. (i);=(ii) is a direct corollary of Theorem 1 since in view or the
latter it is equivalent to state that Vpe<?(ej), _/(pi) is analytic in T
s

r
or that:
V'<a"; VN>0, V
Zl
in SJflr)
l*
r
H..Cc.-/JI <e>-
,
"/(I + ll-tiB'')- (A.!)
Let us show that each constant C
x
can be rewritten C,
v
J>!m. with C
v
independent
of if. In Theorem 1, the constants C are proportional to sup|t)(ji)| g being a
bounded primitive of / : here similarly, the function J'JpJ admits a bounded
primitive (of a certain order):
SM" ilAPi- PviPiVPi. (A.2)
where g is a distribution or order) in p.: we thus have an inequality of the form:
su
Plfl(j>iM<Cil<f>||, and this accounts for the e> dependence of ihe right hand
side of (A. 1).
Conversely, this dependence is necessary to be able to prove in (ii)->(i) the fact
that < pf, a a continuous linear functional (i.e. a distribution).
A proof that (i)|jii) could be given along the same line as the proof of Theo-
rem 1 (by making use of a contour deformation in the variables p,): however it is
simpler to prove directly that (ii)^(iii).
(ii)-(iii): Let ji be such that =t" </( <a" <a, and such that Xi [chosen in 'J,\Q,.)
also belongs to ff/fi,-)].
Wc then choosea sphere ui centered at P* such that condition (2.19) be satisfied.
but with the exponential e~'
r
" instead of ."*'*. Let us now choose the test function
< p as follows:
< p( P2)-XilP:)< :~"
,
~''
,
'"c~"""
(AJ)
177
242 M. Manolcssou-Grummatib-ou
with XilPi) s in (iii)and .\
0
. x
2
given constants. With this choice we obviously have:
&m&XfXrf)lxx,.x,-f^Ax
l
-fJlx
l
.Xa) (A.4)
and assumption (2.19) readily entails:
l ^f ci - *j \ n( Xi . s*XoM 3ei M , . e"^A l +l l *i O (A.5)
[valid in the region (2.23)].
In view of (A.3), evaluation of ||v? yields the majorization:
h>l (l+.| Xim-.X*o). (A.6)
^(.to) is a certain fixed polynomial in x
0
which can be absorbed in the exponeiunil
if we majorize e"'* by e"*
-
": inequality (122) is thus established in the ration
(2^3):
(iii)-(ii): Let us define:
F(x Xj, . oJ- ^rt b, ; / . . , . ^ -.)(x x
0
)
For every test function Xi satisfying (iii). with supp^asuppp. and Xi positive in
the interior of its support, we can write:
F ( x
l
, ' x
J
, x
< )
) =f / (
P l
p
2
)
2 l
(
P l
)
Z:
( p, >- ""*- ' - i ". - ' ' '
!
*' - ' ='
!
w
{\zMr
l
>l*PiX!"-"
y
'*WPi (A.7)
Then putting:
G(x
2
,x<>- je-
,
""bMr>< p( p
2
)'>-
r
'>
,
dp
l
(A.8)
it follows that
J*lX|.Xa.*o)- S2
r
m( XrX2-f)< .xx
2
-x
2
.x
0
)Glx
2
.x
< >
),tx~. (A.9)
Because of the properties of <(p
:
)~' and < p( pi\. Eq. (A.S) yields the following
majorization for G(x
2
,x
0
):
\fXx
1
,x< X< l
ll
l< pi
m
'*
M
'A.l + Xi") (A.10)
where s is the radius of the sphere u>.
' On the other hand by choosing x x inside the region (2.23) and taking into
account (2.22). we have:
t ^( x
1
. x
2
- x
2
, *
0
H *yi + Hx
1
- x
2
n>- ' ". * *'<J<'<a (A.!!)
such that 2,eff(fl,-).
By inserting the bounds (A.IO). (A.l 1) into (A.9) we obtain:
|F(x,.x
2
.x
0
)(
S *-
,
' -' "*
,
M >I Jl +| x, l ' ' )-
|
J*cyi + Jx, -x;! )(l + |!.v
2
i!
v
)-
,
J.v,. (A.12)
The integral on the right hand side of (A.12) is bounded by a (finite) constant
nd we can always take e us small as we wish and such that l!-a
:
>x. Noting
.(hat F[x
l
,0.x
o
)"^
rall
{x,-fj{x
l
.x
c
). we see thai the bound 12.19) readily follows
from (A.12) and holds ill region (120).
178
Macrocausality 243
References
t. Omncs.R.: Phys. Rcv. 140. Ii23 (I9fi6)
2. Kugfer.M.. Roskies.R.: Pfiys. Rev. (55. \-S5tPJS7)
F nlcy.J.P.: University of (California. Berkeley Thesis (1968)
3. Stapp.H.: al tUfli Energy Ph>ics, Imermuimial Center for Thcor. Phyj, Trieste I.A.G.C. I (1965);
b) J. Main. Pins. 9. 154 (I9M1: e) S-nialri* tlieorv
CiawikT.C.. Stapp.H,: i; Math. Phys. it). SM (19693
Pham.F.: Ann. Inst, H. Poineari 2. S9 (1967)
4. lajolniizer. D.. Siapp.H.: Commun, math. Phys. 14, 15--S5II969)
Ia$ o!m'tzer,D.: Lectures in Thcor. Physics {ed. Mahanthappa.K.T.. Brittin.W.E.): New York:
Gordon and Hreaeh 1969
fagoInitzcr.D.: ^traduction fo 5-matri.x theory
5. Bros. J-, Iiisolnitzer. D.: a) Ann. tnsl.-H. Poinear 17.147 1S4 (1973), b) Sminaire Goulaouic -
Schwartz 19741975; and papers in preparation
6. Williams. D.N.: Theoretical Physics Institute, University of Colorado Summer 1967
Hcpp.K.: Hdv. Phys. Acts 37. 659 1964}
7. Froissart.M.; Priww communication to O. hcolnitzcr
. Ascoli.R.. M iiguzz .A.: Phys. Rev. 118. 1435(1960)
sea t.R.: Nuovo Cimente IS. 754 (I960)
9. LebroanruH.: Nuovo Cimemo 10. 579 (195S)
. Mariin.A.: Phys. Rev. 129. 1432(1963)
It. Sec for instance and for further references: Les Houchcs Summer School 1960 "Dispersion
relations".by Omncs.R.
12. BrosX Epstein.H.. Gloser. V.: Nuovo Ciracnw 3U 12651302 (1964)
13. Martin,A.; Problemsof Theoretical Physics. Moscou: Ba ka[Iflp9)
Bros,J,; Symposium on Anal>iic Method in Math. Phys, Btoomin^ton. Indiana. June L96S
Zimmerman. W.: Nuovo Ctmcnto 31. 249 (1961)
14. Stttater,R.F, Wi^jtman.AS.: PCT Spin and Sia&ttes and all ibat New York: W,A.Benjamin
1964
15. iost.R.. lehmarm.H.:NuovoC memo5. 1598(1557)
Dyson.F.: Phys. Rev. 110 (195S)
Bros.1. Mcssiah.A., Stora.R.: J. Math- Phys. 2..639-65I (1961)
Communicated by R. Haaf
Received July 22,1975
Not* A&t In Proof, In the Hac which follows farmuia 3.2t (acid similarly for (3-4)) replace
V^JC t f S. by:
la the Uoe which follow formula (4.23) (respectively (4JS1) replace: *(**'. .*i.)*S* by:
2 J/*
(r
P
.v.t.-eKn4'
, ,
>o).
T R O I S I E M E S U J E T
181
TROISIME SUJET
CONSTRUCTION D'UNE AMPLITUDE UNITAIRE ET ANALYTIQUE
DANS LA DIFFUSION DE PARTICULES DE SPINS \ .
INTRODUCTION
Da n s ce travail nous nous proposons de montrer conent l'amplitude de
diffusion N+N * N+N (et en gnral de spin ^- -j) peut tre dtermine
partir d'un ensemble complet de donnes nergie fixe (section efficace
diffrentielle, polarisation, corrlations de spin e t c ..) sachant sur le
pl a n thorique que c f t e amplitude obit a ux conditions d'unitarit et
d'analyticitl en cos 8 rsultant de la thorie axiomatique des champs.
Lorsque l'nergie est fixe une v a l e ur infrieure l'ouverture des
seuils inlastiques nous sommes e n mesure de m o n t r e r que la construction
de l'amplitude est unique sous certaines conditions et part certaines
ambiguts discrtes-
Le problme de construire un e amplitude unitaire partir des donnes
exprimentales a proccup, depuis longtemps, tous les thoriciens ainsi que
les exprimentateurs, et il prsente plusieurs aspects et difficults : une
difficult vidente est celle des incertitudes sur les donnes exprimentales
lles-nots et les rpercussions <jue ces incertitudes peuvent entraner sur
la dtermination des amplitudes. Cependant le problme thorique ne peut Stre
abord qu'aprs avoir mis au point une mthode permettant la dtermination
de l'amplitude e n fonction de "donnes idales". La deuxime difficult se
produit lorsque l'nergie est fixe au-dessus du seuil des canaux inlastiques
t que les donnes exprimentales considres n e fournissent aucune informa-
tion sur les canaux inlastiques.
Ces premiers rsultats concernant le cas scalaire (spin 0 } , et pour la
rgion lastique, ont t obtenus pa r M e v t o n et M a r t i n . En tenant compte de
la section efficace diffrentielle donne par l'exprience et de l'unitarit
lastique, ils ont o b t e n u des conditions sur l'existence t l'unicit de
l'amplitude de di ffusi o n . Martin en particulier a considr la relation
d'unitarit comme dfinissant une fonctionnelle non linaire sur l'espace
des fonctions continues 41(8) reprsentant la phase de l'amplitude :
182
F(12) - |F(12)| e x p ( i *C 1 2
Ce t t e fo n ct i o n n e l l e est d fi n i e pa r la fo r m ul e sui v a n t e :
, I dn
3
|F(13)||F(23)| cos [$(13) - *(23)]
si n * * <12)
4ir J |F(12)|
La pha se de l 'a m pl i t ude che r ch e do i t t r e un po i n t fi xe (6' 6) de
ce t t e fo n ct i o n n e l l e .
En a ppl i qua n t le pr i n ci pe de Le r a yScha ude r , M a r t i n a t r o uv co m m e
co n di t i o n suffi sa n t e de l 'e xi st e n ce de .ce po i n t fi xe :
, dfl
3
|F(I 3)||F(23)|
|ain u| = m a x^jp j -
F(12)|
et co m m e co n di t i o n d'un i ci t : |si n u| < 0,79 .
D. At ki n so n , P.W. Jo hn so n , R.L. W a r n o ck ont a m l i o r les t e chn i que s
e m pl o y e s pa r Ne w t o n et M a r t i n et e n pl us ils o n t e xpl o r la r g i o n i n l a s-
t i que . I l s ont d m o n t r que da n s ce t t e r g i o n o n pe ut a v o i r un e i n fi n i t
d'a m pl i t ude s a v e c la m m e se ct i o n e ffi ca ce di ff r e n t i e l l e do n n e .
Ce r sul t a t est trs i m po r t a n t ca r il m o n t r e b i e n qu'e n l 'a b se n ce de
do n n e s e xp r i m e n t a l e s sur l 'a n a l yse de s ca n a ux i n l a st i que s, il e xi st e un e
a m b i g u t co n t i n ue da n s la d t e r m i n a t i o n de s d pha sa g e s et de s pa r a m t r e s
d'i n l a st i ci t .
D'a ut r e part un e s r i e de t r a v a ux pa r R.F. Al v a r e z -Est r a da , B. Ca r r e r a s
a t fa i t e sur les di ffusi o n s ir-N et N-S. Ils o n t e xpl i ci t de s co n di t i o n s
sur les o b se r v a b l e s e xp r i m e n t a l e s po ur que l 'un i ci t de la so l ut i o n soit
a ssur e .
Un a ut r e a spe ct de ces i n v e st i g a t i o n s a t do n n pa r D. At ki n so n ,
G. M a ho ux, J.F. Yn dur a i n qui e n ut i l i sa n t a ussi les m t ho de s de po i n t fixe
o n t r ussi t e n i r co m pt e de s pr o pr i t s d'a n a l yt i ci t de s pa r t i e s a b so r pt i v e
et di spe r si v e de l 'a m pl i t ude da n s le pl a n co s 8, telles qu'e l l e s sont do n n e s
pa r la t h o r i e a xi o m a t i que de s cha m ps.
Le s pr i n ci pa ux a v a n t a g e s de ce t t e t ude sont les sui v a n t s :
a ) L'a n a l yt i ci t e n cos 8 da n s la g r a n de (pe t i t e ) e l l i pse de M a r t i n de
la pa r t i e a b so r pt i v e (di spe r si v e ) a ssur e la d cr o i ssa n ce e xpo n e n t i e l l e de s
o n de s pa r t i e l l e s po ur les g r a n de s v a l e ur s de Z ce qui i m pl i que qu'i l est
po ssi b l e d'o b t e n i r un e b o n n e a ppr o xi m a t i o n de l 'a m pl i t ude a v e c un n o m b r e fi n i
d'ondes pa r t i e l l e s.
b ) On peut accepter des zros de la section efficace dans l'ellipse de
M a r t i n , m m e dans la rgion physique.
c) M m e en dehors du domaine o les thormes du point fixe sont valables
o n peut dmontrer la convergence d'une itration Newton-Kantorovich et explorer
numriquement l'ambigut continue dans la rgion inlastique.
Da n s l e cas scalaire A.M .Y. ont dmontr l'unicit d'une amplitude un i -
taire, analytique partir de la section efficace diffrentielle, en utilisant
le thorme du point fixe sous la forme suivante :
Ils ont dfini un espace de Banach B. de fonctions analytiques de cos 9
(dans l'ellipse de M a r t i n ) des parties absorptives A(x) et en vue de sa t i s-
faire la relation d'unitrit, ils ont introduit l'application fonctionnelle
dfinie par le systme (S) suivant :
A'(x) . I ( .21* t) P.(x) [D* + A*] (1.1)
Jt*0
U j l
- i f d x P ^ x ) [o(x) - A
2
( x ) ]
I / 2
(1.2)
1 f
1
A^ - j I dx P

(x) A(x) (1.3)

Le s "points fi xe s" de cette application (c'est--dire les couples (A,A')
tels que A(x) - A'(x) fournissent alors b i e n une solution D+ iA satisfaisant
a ux conditions d'unitrit et d'analyticit et dont les quations ci-dessus
fournissent les dveloppements en ondes pa r t i e l l e s.
Utilisant des techniques de dformation de contour dans le plan cos 8 ces
auteurs ont pu trouver des conditions faisant intervenir les bornes suprieures
et infrieures de la section efficace o(z) telles que l'application fonction-
n e l l e A * A' soit contractive dans une certaine b o ul e S. de l'espace B.
Si ces conditions sont satisfaites l'amplitude (ou la srie des d pha -
sages) est dtermine de faon unique dans la boule S. par la do n n e de o (x).
Av e c des conditions plus faibles le thorme de Schauder est applicable
L'existence de l'amplitude est alors garantie mais n o n son unicit.
Po ur la rgion au-dessus du seuil inlastique la seule modification
apporter aux quations (l,a) a * 1>2,3 est l'introduction de la fonction
184
d'inlasticit (n,x) dans (1.0 savoir :
A '(x) - l (2*+l) P,(x) [D
2
+ A
2
]+ i(n;x) (2.!)
1*0
o I - ti
2
A (n,x) = (2)1+1) j ^ P^ x) (3.1)
et les n sont les paramtres dit d'inlasticit, non mesurs exprimenta-
lement et qu'on peut donc choisir arbitrairement dans l'intervalle 0 < n^ < 1
D. A tkinson, P.W. Johnson; A .L . Warnock avaient dmontr que gnralement
l'amplitude est une fonctionnelle continue de la fonction inlastique (n,x),
par consquent une infinit d'amplitudes (ou de dphasages) peuvent corres-
pondre la mme section efficace diffrentielle.
A .M .Y. dans (1) ont obtenu cette ambigut continue en tenant compte de
l'analyticit et d'autre part ils ont explicit des conditions sur (n,x) et
o(x) pour que certains thormes de point fixe soient applicables.
L e rsultat essentiel de cet.ce tude est que la "drive de Frechet"
(application linaire tangente) de l'application fonctionnelle dfinie par les
les quations (2.1) , (1.2),(! .3) est un oprateur compact ce qui permet d'uti-
liser les thormes de Fredholm et une mthode d'itration peut tre applique
pour explore!; numriquement l'ambiguit continue.
Ces calculs ont dj t faits par D. A tkinson, P.W.-Johnson, L .P. Kok,
M . de Roo pour le cas de la diffusion a-a . A .M .Y. ont tendu les rsultats
.mentionns du cas scalaire la diffusion TTN. Ils ont montr que sous des con-
ditions favorables la seule connaissance de la section efficace a(x) et de la
polarisation P, suffit dans la rgion inlastique dterminer l'amplitude
une ambiguit discrte prs. De cette faon on peut trouver les autres obser-
vables non mesures A et R .
Dans la rgion inlastique l'ambiguit continue reste toujours malgr le
fait qu'on puisse'utiliser un ensemble plus grand d'observables mesures
savoir a, P, A et R . On a donc une libert infinie varier les paramtres
d'inlasticit sans changer la section efficace a(x), la polarisation P(x) et
les paramtres A et R , et dans (2) des conditions:- ont t trouves pour que
l'itration de Newton Kantorovich soit applicable.
Dans ce qui suit nous.tudions par la mme mthode la diffusion de spin
K - y. L'tude se rapporte gnralement aussi bien la diffusion N-N, qu'a
celle d'autres particules ayant spin j .
185
A part les complications dues au spin (expressions compliques des fo n c-
tionnelles qui entrent dans le pr o b l m e ) et des difficults dues au principe
de Fault (particules identiques) nos conclusions constituent d'une certaine faon
l'extension un plus grand nombre de dimensions des rsultats pr c de n t s.
M a i s d'autre part c'est le grand intrt physique qui nous a amen
cette t ude , puisqu'une analyse en dphasage des interactions N-N est n ce s-
saire l'tude de s forces n ucl a i r e s.
Dans la premire partie nous dcrivons la cinmatique employe, le d v e -
loppement en ondes partielles de l'amplitude de diffusion, et les observables
mesures par l'exprience qu'on utilise par la sui t e . Dans la deuxime partie
nous dfinissons l'espace de Banach et l'application fonctionnelle construite
partir de l'unitarit. Cette application fonctionnelle sera dfinie par un
systme d'quations ayant la m m e structure que le systme de (1) que nous
avons dcrit ci-dessus j cependant chacune des quations (1.!), (1.2), (1.3)
sera remplace par un systme d'quations, -faisant intervenir la place de a
un ensemble d'observables qui sera prcis plus loin. D'a ut r e part les lments
A(x), A'(x) de l'espace B seront remplacs pa r des ensembles d'amplitudes
invariantes (A
a
(x)), (A^(x)).
Nous dmontrons que lorsqu'on fait des petites perturbations sur les
6 (ou 5) observables au voisinage de certaines v a l e ur s, le thorme de l'appli-
cation contractante est applicable, et que nous obtenons un systme unique
d'amplitudes satisfaisant l'unitarit et l'analyticit.
Nous considrons sparment la diffusion des particules non identiques
de celle des particules identiques (par e x. n-n, ou avec invariance d'isospin
p - n ) .
Dans la troisime partie nous dmontrons l'applicabilit du thorme de
Schauder qui assure l'existence (et pas l'unicit) de l'amplitude de diffusion
analytique et un i t a i r e .
Dans la quatrime partie nous explicitons des ambiguts dues certaines
symtries que les observables prsentent si on applique certaines transforma-
tions l'amplitude de di ffusi o n .
Dans la cinquime partie n o us obtenons les conditions d'une itration de
Newton-Kantorovich convergente.
On verra dans ce qui suit que dans la rgion l a st i que , la mesure de
toutes observables conduit un problme d'analyse en dphasages sur d t e r m i n .
Pour avoir un problme non surdtermin, on doit donc faire un choix parmi les
186
donnes exprimentales qui entreront dans l'analyse en dphasage. Ce choix est
bien sur arbitraire et celui que l'on fait ici ne coincide pas ncessairement
avec celui que fait l'exprimentateur.
187
Nu c t a i l ' h y i f c t 09 S {19 7 5 ) 2 9 3 - 3 2 S
O No t l l M l o l l a n d P u bl i s h i n g Cu m n u n y
CONSTRUCTION 01* A UNITARY ANAL YTIC AM P L ITUDE
FOR THE SCATTERING OFSP IN- * P ARTICL ES
M . M ANOL ESSOU- GRAM M ATICOU
Service de Physique Thoriitiie, Centre d'Etudes Nuclaires de aclay, ffPno, 2~ 9U9Q
Gif'Sar-Yvctte. Fronce
Re c e i ve d 14 Apr i l I9 7 S
( Re vi s e d 4 Ju l / 19 7 5 )
We s t u dy t h e e xi s t e n c e , t h e u n i qu e n e s s a n d [h o c a n s t r u c t i o n o f u n i t a r y a n a l y t i c a m pl i -
t u de : ; Tor s pi n { - j s c a t t e r i n g, i n t h e f r a m e wo r k pr e vi o u s l y i n t r o du c e d by At ki n s o n -
M a h o u x- Yn du r i i n f o r s pi n 0- 0, a n d s pi n 0- j s c a t t e r i n g.
1. In t r o du c t i o n
Th e pr e s e n t wo r k c o n s t i t u t e s a n e xt e n s i o n t o t h e c a s e o f t h e s c a t t e r i n g o f s pt n - j
pa r t i c l e s , o f t h e r e c e n t r e s u l t s o f At ki n s o n . M u h o u x a n d Yn du r din ( AM Y [1,2 ]) c o n -
c e r n i n g t h e c o n s t r u c t i o n o f a n a l y t i c u n i t a r y s c a t t e r i n g a m pl i t u de s t h a t c o r r e s po n d
t o a gi ve n s e t o f o bs e r va bl e s at f i xe d e n e r gy .
We s t u dy h o w t h e s c a t t e r i n g a m pl i t u de o f pr o c e s s e s N + N * N + N ( n u c l o n -
n u c l o n ) c a n be dc t e r n i i n e d f r o m th e kn o wl e dge o f a r e s t r i c t e d s e t e xpe r i m e n t a l
da t a a t f i xe d e n e r gy , r e qu ir in g t h e c o n di t i o n s o f u n i t a r i t y a n d a n a l y t i c i i y o n t h e
c o s 0 pl a n e th a t f o l l o w f r o m a xi o m a t i c f i e l d t h e o r y . At e n e r gi e s be l o w t h e i n e l a s t i c
t h r e s h o l d u t i i t a r i t y i s e xpr e s s e d a s a n o n - l i n e a r m a ppi n g o n a B a n a c h s pa c e o f f u n c -
t i o n s wh i c h arc a n a l y t i c i n t h e s m a l l M a r tin e l l i ps e . We s h o w t h a t t h e r e e xi s t s a
f i xe d po i n t o f t h i s m a ppi n g wh i c h is u n i qu e , a pa r t fr o m s o n i c di s c r e t e a m bi gu i t i e s
i n a c e r t a i n ba l l o f t h e B a n a c h s pa c e , i f t h e o bs e r va bl e s s a t i s f y a ppr o pr i a t e c o n di -
t i o n s . It f o l l o ws t h e n t h a t t h e pr o bl e m o f ph a s e - s h i f t a n a l y s i s , gi ve n a c o m pl e t e
m e a s u r e m e n t o f a ll o bs e r va bl e s n e c e s s a r y t o de s c r i be t h e s c a t t e r i n g pr o c e s s N + N *
N + N i n t h e e l a s t i c r e gi o n i s o ve r do t e rut i tie d i n a m a t h e m a t i c a l s e n s e . As a m u t t e r
o f f a c t , i t i s s u f f i c i e n t t o kn o w f ive o bs e r va bl e s ( s i x f o r n o n - i de n t i c a l pa r t i c l e s ) i n
o r de r t o de t e r m i n e c o m pl e t e l y t h e a m pl i t u de a n d a ll t h e r e m a i n i n g u n kn o wn o b-
s e r va bl e s . Th e a bo ve r e s u l t a ppl i e s t o oven .- s c a t t e r i n g pr o c e s s o f s pi n 4- - fo r i de n t i -
c a l a n d n o n - i de n t i c a l i n t e r a c t i n g pa r t i c l e s . In t h e i n e l a s t i c r e gi o n t h e c o n t i n u u m a m -
bi gu i t y f o u n d i n r e f s . [3 .1*2 ] f o r t h e s c a la r a n d s pi n 0- j pa r t i c l e s r e m a in s de s pi t e
t h e c o n s t r a i n t s i m po s e d o n ( t ie i n e l a s t i c i t y pa r a m e t e r s . We h a ve a l wa y s a n i n f i n i t y
o f s o l u t i o n s t h a t c o r r e s po n ds t o t h e s a m e s e t o f f i ve o bs e r va bl e s . Ne ve r t h e l e s s a
n u m e r i c a l i n ve s t i ga t i o n o f t h e s e c o n t i n u u m a m bi gu i t i e s i s po s s i bl e by a n i t e r a t i ve
m e t h o d.
188
M, fiaiizitcsou-Crauuiuttkoti / Unitary analytic amplitude 299
The problem of constructing a unitary amplitude from experimental data has al-
ready been examined [4, 5] . For an historical survey on this subject one should refer
to the introduction of ref. [6] .
We mention particularly the work of Alvarez-Estrada and Carreras [7] , who
treated rr-N and M-N scattering, but without taking cos 0 annlyticity into account.
They used a space of functions continuous in the physical region. The method of
the present work is a generalization of the one of AMY [1. 2 ] . Aside from the fact
that much more complicated expressions occur and aside from certain technical dif-
ficulties that enter into the problem, because of spin and the Pauli principle, our re-
sults constitute the extension to a larger number of dimensions of the previous re-
sults by the authors of refs. [1, 2 ] . What stimulated us in this study is the great
physical interest that a phase-shift analysis of N-N interactions presents for the
understanding of nuclear forces. In sect. 2 we describe the kinenutical situation, the
partial wave decomposition and' the choice of the five observables.
In sect. 3 we define the Banach space and the non-linear mapping constructed
from unitarity. We show that when the \~we observables vary in the neighbourhood
of some fixed vaiues explicitly given, the contraction mapping theorem of Bauach-
Cacciopoli can be applied inside a certain ball: the existence and the uniqueness of
a unitary analytic amplitude is then ensured. The Schauder theorem, implying only
the existence of a solution but not its uniqueness, is also applicable under less strin-
gent conditions.
In sect. 4, we investigate certain discrete ambiguities emerging from the symme-
tries of the observables, when the scattering amplitude undergoes appropriate trans-
formations. These ambiguities are found to be analogous to those, present in the case
of T-N scattering [2 ] . At the end of this section we give a list of different possible
choices of the set of observables which allow a simplification of the mathematical
structure of the problem.
In sect. 5, we discuss the continuum ambiguity and following the method of
AMY [2] for a practical calculation scheme, we show how the Newton-Kantorovich
method works for the inelastic unitarity equation.
Finally, some conclusions are presented in sect. 6.
2 . Kinematics and analyticity
The scattering of two spin-* particles is described by the amplitude T [S] ,
7 *
T
+(3)
T
+<4)
x
+( 3)
x
+( 4)
i l / x
( l )
x
( 2)
T
( l )
T
<! )
(
where x
( l )
(f = I . . . 4) arc the spinors for initial ( 1) , ( 2) and final ( 3) ( 4) particles: r
( , )
are the corresponding isospinors;jl/ is a 16 X 16 matrix depending on the vectors q,
q', the initial and final relative momenta in the cm. system, and on the operators
of spin a^ and isospin t W (j- \, 2 ) .
189
300 M. Manotcssoii-Gramitiaticott / Unitary analytic amplitude
Rotational invariance in isospin space allows us to write: M"MQPQ + M|P
t
whore
P
T
( r * 0,1) arc the projection operators in tlie total isospin r states. [Because of tit-
variance under spatial rotations, parity conservation and time reversal invariance
each one of il/
0
. , 1/, isoscalar amplitudes has the general form
, '. -E^we. tf'. fl. *=cos0, j5j, (2.i)
where F(
x
)
a r c
complex functions of the energy and the scattering angle. By using
the vectors
>- ! * . . * . = * , . i-JM. (2.2)
IWI W -l lxl
the operators Q
a
(q', q) can be chosen as follows:
0 , - i .
fi
2
-(o,-p)(
2
-p) + (<ii -k)(c
2
-k),
Qj - ( '1 P) (o-
2
P) - (I *> (2 " *) (2.3)
fi
4
-(l')(<i2').
Gj=('(o-[+<r2) ,
having the property: Q
a
(q", q) Q*W, ?').
For non-idemical particles, M has the general form (2.1) plus an additional term:
/(e^ 52) * ^g. In the case of non-conservation of isospin the Fp i - 1... 6 are
complex matrices in isospin space, functions of energy and the angle 0.
From now on we will be interested only in spin space. We define the absorptive
part of the amplitude M(q\ q) by
5.6
Abs Miq'.q) = jr Wiqq) - <(*(, q')\ = j- S [?(.*) - fJM ] fl
a
'j). (2.4)
This defines the absorptive part of the scalar amplitudes ?
a
M
^. w- ^ [f.cv)-?;wi. (2.5)
In the same manner the dispersive part will be
s>)-i i ^w
+
?:wi . (2.6)
From axiomatic field theory we know that the amplitudes F
a
,a~ 1,..., 4
Vl - 2
2
?j 6 at fixed energy are anlytic inside the following domain in the varia-
ble z * cos 0,
VC(?)\<E(z
0
), where i t ) i + ( i
2
- l ) i , - o ' l + T ^. (2.7)
2q-
190
M. Afanolcisoit-Granmialtcou / Unitary analytic amiililtiijtt 301
with t
0
being the nearest r-channcl singularity. This is in fact the small Martin ellipse.
The function (z- 1)2 is a onc-vulucd function in the complex z-plane, cut from
J = -1 to z = +1, thai takes positive values for J real and greater than +1.
We can then continue analytically the A
a
fy), RJ x) in the complex i-plaue by
the formulae
4.W=37 fcw-?*(**)],
(2.8)
Wi'HW + W)]-
The A
a
(z) a - 1,..., 4 (a = 5, 6) are analytic inside the large Martin ellipse
(
r
l ) * M ~ 2rfl - 1 (cut ellipse) below the inelastic threshold and the R
a
(=)
a~ 1, ...,4 (a -5,6) are analytic inside the small Martin ellipse (z$) (cut ellipse).
For the inelastic region the ellipse of analyticitv oM
a
( r ) is in general smaller than
For practical reasons we will use later the following scalar amplitudes:
F, = 2 ( F, +P
4
) , F
2
=4F
S
, F
3
'2Fi, F
4
= -2 f
2
,
F
5
= 2 ( ?, - f
4
) , F
6
=4F
6
, (2.9)
with
4,<3 = 27FC--)-r(**)].
*(*) = H ^O + F-Cz*)].
(2.10)
In ref. [6] we have explained in detail how the partial wave decomposition of ampli-
t udes^ is obtained. We give here the final result,
'.% *{,,/&. an)
where/is .the total angular momentum: J?^<^ are functions of Legendre polynomials,
fp-p is the partial wave.which corresponds in a configuration p"',3 = 0,01,,+-, 1 of
total spin and orbital angular momentum for the final (j3') and the intial (pi state
respectively; this definition becomes clearer from table 1 giving the matrix elements
offpp Explicitly (2.11) gives
'
,
|-cS2[(2/t|)P,_
1
-J] /i .
+
2[c/H)P
/ t |
-
3
j /i ;t(yi )
3 1,, ^
1 + - 2 - V - ^ ft
| v y( y+ i ) |
;
i y/jy + iy*-
191
302 M. Mattolciiou-Gramutaticou / Unitary analytic amplitude
Tiblcl
MalrE* dementi o
e
ji
Q
r -o sr*\
i ' j r-j*i t ' j t'J-\
S-0 l-J
fi
a
//,
0
S'I l'J
l-J-x
0
/ /
0
fi,
0
0
ri
0
fU
0
i fia" fit
2 4u y y
r
j-v~ j-n
e
J *i>** jy + \.)
r
j'v
p/(/ /*.,-et *fytf_,
*OJ*i)Pjf^-
V)
>sW+)
fi.
By inverting (2.11) we obtain
tf" J PjCF
s
+2F

)(.aaa),
(2.11)
/ / - '
( 2 / t l )
^ J 7 T T ) ^ ' ^ - ' - ' Wi > +W^O ^- cos M ^F j Jdf cus J) .
192
M. MaitolcssoU'Grar.:malicou / Unitary analytic amplitude 303
I
/
i
r =
J y T i ) SW+lVAP
i
-2F
i
)-cos0l'
J
lF
s
-2F
A
-F
l
)*2f
J
F
i
*PiF
l
] d(cos ),
^ "sVi _/ [ c ^i Vj ^" ^- ^ - ^r ^3
fL =^i _/ [<V+I V
/
.
l
f,+ytfi-
2 f
4-^-24
=i
^
- e ^ . ^ c o s e ) ,
Because of the invariance principles already mentioned, only 9 (or II) of 256 [8]
different possible experiments in a scattering process of spin-i particles, are necessary
in order to reconstruct the scattering amplitude at fixed energy below the inelastic
threshold. In otlwr words we have to determine Rve (or six) complex invariant ampli-
tudes/
1
^, a general phase remaining arbitrary. The choice of a set of 5 (or 6) observ-
ables is not unique and in particular the one adopted for a theoretical study is not
always the same as the one measured by the experimentalists. In the following list
we define the observables used in the present work. The criteria for our choice have
been rclativistic invariance and the existence of experimental data. As a nutter of
fact most of the given observables have been measured for the interactions p-p, p-n,
n-n P J.
a : Non polarized differential cross section cr cTff/dn - o(x) - JTLUJI/ *;
P< : Polarization of particle ( I) opMfx) = iTr. t/. iro, :
/O' : Polarization of particle (2) oP&ix) = iTr. U. UVi '':
D : Depolarization aD(x)-\T tM{p^ n\\t'(p^ n\\
D, : Transfer of polarization oD,(x) = $ TW( < T, V) W*( O, ) :
C
m
: Spin correlation parameter J C , ( I ) = ^Tr.l/;l/*(<r[ /?) ( , ) .
(2.13)
For identical particles af
fl)
= aP
n)
= oP.
From definitions (2.13) we obtain the expression of the observables in terms of
the 5 (or 6) scalar amplitudes. If follows that a. P. D. D
v
C
lm
are functions of x -
cos 9, defined on the real axis 1 <x < I and they can be continued analytically in
the small Martin ellipse ( r
0
) of the c-plane by the formulae
.a=iSu*)^;c)!, ...f
for
-'
2
-
i
-
s
-
for a = 3, 4,
19 3
3 04 M itanolcttoU'Cramimtko / Unitary analytic amplitude
0(
I
)f"'(r)>J[2(-->
1
|W-'Rl(--)'2C=) + 5(--)'1
6
(--)-
6
(--)^5fr)l,
oCryG'M ' I lR
1
Afi)-R,
L
&A
l
(:) - R
S
(.:)A
6
(:) + K
6
( r ),t jf r )I.
<K )(l -*. (; . E / # . - ) +*;[(.-), a - 3, 4 ,
oWO - fl.(i)) > J S W; ( 0 + **(:)), a = 3, S, (S (factor 4 for a * 3),
<>G0O-C0O)*i S (/ij;(s)+ ;(*)), a = 4 ,5,6 (factor 4 for a 4 ).
(2.1 4 )
In order to perform the non-linear mapping it will be necessary to invert eq. (2.1 4 )
(see eq. (3.1 5) later).
3. Contractive mapping theorems
3.1. Non-fa tear mapping
The elastic imitarity reads
\mf
J
=f
J
U
J
=f
J
f
J
K ( 3 .1)
where/-' is the partial wave matrix (table 1 ). Time reversal gives
!'-t
n
. (3.2)
The matrix Im/ ^ is real symmetric: while f^ f
3
, f
J
f** are hermitian matrices but
not symmetric. By separating symmetrical and'antisymmetrical part in (3.1 ) we ob*
tain
. m^- i O^, / - ' } , (3.3.1)
r / '
, ,
, /
y
] =0 ( 3.3.2)
For the fixed point theorems, only eqs. (3.3.1 ) will appear in the definition of the
mapping. The constraints (3.3.2) have to be satisfied by the solution, and it has been
demonstrated [1 0] tint for a sufficiently small differential cross section, they are
automatically satisfied. For the inelastic region the equivalents of (3.3) are
! m/
/
= \ V
s
*
t
f
J
] + \{K
J
+ L
J
\ (3.4 .1 )
\f
J
K f
J
\ + iC-tf
7
- A
y
) = 0, (3.4 .2)
and only (3.4 .1 ) appear in the dfinition uf the mapping. The A', Lr are hermit ton
positive semi-definite matrices defined by
t94
if. Manolcsiott-Granwuuicoit f Unitary analytic amplitude 305
V
fl
-tfA/,|7*t .T \J M& \ ,_ ,
?
v
* . witliA. '^-i-', (3.5)
where J ?
i n
is the projector on all inelastic states.
In practice K
J
+ L
J
are given arbitrarily and once the problem of fixed points is
resolved we calculate K
J
-L
J
from (3.4.2). What we must then verify is whether
the calculated A'-', L
J
arc positive semi-definite. WIuiI wc can say is thai generally
the last property is true for some K
J
+ l
J
, and we expect in general [21 a continuum
imbiguily the K
J
+ L
J
varying for every / in a certain domain.
From the unitarity we will construct the non-Unear mapping,
A -+A' = WiA, a, p) + /(T>), . (3.6)
using the notations A = (A
a
} for the set of absorptive parts of the invariant ampli-
tudes, o is the differential cross section, p is the set of four remaining observables,
while 7(77) is the non-elastic contribution. Wc define in fact a 4 X 4 matrix of inelastic-
ity, having the same structure (2.1) as the scattering amplitude,
5.6
/(z, ) = S I (s. n) QJ , q), (3.7)
where thc/
a
( j, ) are analytic functions of cos 8 =z in an ellipse C(z2) with j , > r
0
fora = 1,3,4, 5 and a cut ellipse for a = 2, 6. They depend on the inelasticity para-
meters r/ (defined in detail in ref. [6]), and their partial wave decompositions are
analogous to eqs. (2.11) for the F
a
/
.
f c
*>"j
i
rt
w
*irt"
( 3 8 )
with
jCff' + i
7
) * / ' . (3.9)
In order to write explicitly the mapping (3.6) we define the absorptive and dispersive
parts of partial waves
V e 2 7 ' . w 2 *
( 3 , l 0 )
In the following we will treat the problem of identical particles only, the applica-
tion of the method to non-identical particles being the same, except for the number
of dimensions: six instead of five. Nevertheless, as we have explained before, there
ire no constraints of Pauli symmetries for non-identical particles, and we can use
the method that was applied to the case of spin 0 5 [2], as we showed explicitly
in rcf. (6].
The symmetries that the Pauli principle imposes upon the five invariant ampli-
tudes are
195
306 H. itattolcssou-Grammaticott / Unitary analytic amplitude
F
2
(8) -tF
2
(iT-6),
F^O) - e f j ^ - 8 ) . (3. 1!)
where e* 1 for isospinJ* l , ande
s
1 for/ "0.
If we use the notations J2
t
, I = 1,. . . . 5 for the successive functional entering in
the expression of the non-linear mapping, it has the following fonn:
A-* A;
A' = J2
l
(a
r
) (the absorptive part of equations (2.11)), (3.12)
a
r
*
2
(a
J
,r
,
,Tj) (the unitarity, cq. (3.4.1)), (3.13)
r
J
=
3
lR) (the dispersive pari of eqs. (2.12)) (3.14)
R 4(0/, p, A) (the dispersive part of amplitudes [F
a
] in terms
of the observables and the absorptive parts), (3.15)
a
J
= J2
s
{A) (the absorptive part of eqs. (2.12)), (3.16)
which are given explicitly in appendix E.
3.2. Banach space - contractive mapping theorem
We define a Banach spaie B of sets F = {F
a
;u 1,. . . , 5} of analytic functions by
means of the following norm:
nFH supm ax-^-r, (3.17)
where the partial waves/^ ate defined by eq. (2.12). This norm ensures that the
functions F
tt
(a 1, 3,4, 5) are analytic in the ellipse <5(f
t
): (cut-ellipse (?[) for
a = 2). In fact if the norm Mil is finite, the partial wave decom positions (eq. 3.12)
converge inside the ellipse C(({) which means that A
a
(or - 1,3,4. 5) are analytic
in the large Martin ellipse (cut ellipse for.4;) d(f|), where fj = 2fj - 1.
Besides that, we show that in appendix A the functions A
a
are bounded in the
following manner:
|X
a
(z)l <;*'(?) Mil, (3.18)
where
a
(?) is a real function of z which becomes infinite on the border of the ellipse
d(Ji). As remarked in ref. [4] the form of the singularity of the absorptive part on
the border of the ellipse of convergence is not prescribed by general theory. But we
can always take f
0
to be smaller than z
a
and. in the favourable case when there are
no zeros in (:
0
) it can be as close to : as we wish.
196
M. Manotcssou-Gramnwticoii I Unitary analytic amplitude 307
By applying the Banaeh-Caceiopoli contraction mapping theorem [11] we investi-
gate some arbitrarily chosen regions in the Banach space 8, and show that they con-
tain fixed points of the mapping defined by the system of eqs. (3. !2)-(3. 16). We
now state the Eanach-Cacciopoti theorem ( BC).
IT: b)AeS
b
A'eS
b
where S
b
* {A( :); IHII < 6} for some convenient 4:
(b) ll/l'i - A'
2
\\ < KWA, - A
2
\U \ M , . A
2
<= 5
6
with K < I , then the mapping has
one and only one fixed point into the ball S
b
.
Before examining the conditions tltat the observables a, P, D, >,, C must satisfy
for (a) and (b) to be true, sve remark that the mapping is singular at the origin be-
cause of the denominators in eqs. (3.9), that is,
lim fi
1
04, o, p) l l =~, A SB.
A-0
Consequently we must centre the ball S^ around a point different from the origin,
such that there will be a neighbourhood of zero-free functions. By considering as
necessary the properties of unitarity, absence of zeros in the Martin ellipse, the
Pauli symmetries, and seeking simplicity, we have chosen among other possibilities,
as centre of the ball 5^, the amplitude which has all of its partial waves zero except
the first three: ; , / J
+
, /
L
'. Let us denote by F this amplitude; it has the form
F
r
[271 + 37}]*,
^ = - 8 ^2/ ^3 / 7] (I -**)},
*i-i5Fl-?&*37?I.
s-S&K + W-
(3.19)
One readily verifies that the symmetries (3.11) are satisfied by the f . The partial
waves/J, /}, 7 j
t
satisfy elastic unitarity automatically; therefore we can define
parameters tj, 3^<., s[ by
].i<l-..MV /?
+
4(l-c
Z
'^), /,'-<!-Mi).
2 SO
r =
S
inScos6,
sin
1
5"
n of,
,so r an* Q i.ua Wg, '
+ +
=
Sm 6
+ +
COS
The mapping is then performed on the ball
S
i
= {A( .:):U-\\< b),
wit the norm of A defined by
[si n
2
5, sin
2
Z\ sin
2
5?.
L,
IWn = max
\GoCfi) ' 2i<?i) ' <?(?,)
(3 .20)
(3 .21)
(3 .22)
197
308 M Maiiolettott-G/alwitaticou / Unitary analytic amplitude
The corresponding Inelastic matrix !(x. i)) is of tlie form (3.7) 7(., rf)
2 & ir
a
( x, il) QJ s, 9) where tlte functions T jx, r;) have the structure given by ecis.
(3.19).
The observables corresponding to the amplitude F* [F
a
} are
o ? - 0 ,
| < 1- B - B, C) - ^ l -2 / ?
+
+ 3/, ' I
2
,
U'C itS-B. -e. J-^iatftJtfl*.
i ( l + 5 + B, ? > - ^ R & + 2 7/I
1
. (3.23)
Let us define
M, sup |irf|,
up
M
* ' *gfo 'Hi - o+a
t
- q) - iff(i - B+B, - O .
*S " ^ ^ W l + B - C, - C) - 2 t7
( 1
+D-D
t
- CJ \. (3.24)
The above quantities measure the deviation from the values of the observables (3.23)
which correspond on the point .4.
In order to obtain sufficient conditions for the mapping to be contractive inside
Sjfiwe have to ensure the exponential decrease of partial waves a^ . For this pur-
pose we need (because* of unttarity (3.13)) the exponential decrcase*offf^ already
given by the analyticity of A
a
and of r ^ which will be obtained by exploiting the
analylicity of R^ (eqs. (3.14) (3.15)). This step will be realised by transforming the
integrations (3.14) into contour intgrais on the boundary 3(Q) of the ellipse ((,<)).
We must therefore ensure that the square roots in eqs. (3.15) are analytic inside this
ellipse (absence of square root branch points) and the denonir ;urs in the same
equatigns arc zero free. Sufficient conditions for the analyttctty of the square roots
are then obtained:
(i) for-the analyticity of 3 :
i nf| A, | s inf U o a - D - ^+ C j - ^t l X i e r r r ^- ^s i n S O c o s ^
i
iE (t
f t
> '
m J
+ 3 sin | cos 5! I -Af
3
- K,b > 0, AT
3
= const. (3.25)
198
A/. Manotcssoa-Gramttiath'oii I Unitary analytic amplitude
By imposing
1-2 sin 5* cos 6 + 3 sin5[ coseJl^O, (3.26)
and A/
3
. o sufficiently small, (3.2S) is satisfied: tnf | i
3
l * 0.
(H) For the analytieily of 4, R$:
i \ i i
5
: 6
^ M * **>,- c) - <
3
i
>2|sinS5cos5?| sup 3|isin!cosS !l-Af. - K . ,b, Kj.A'
s
= const.,
u
" iee-
0
> ' '
, J
eo>
then forA/j, Af
4
, 4 sufficiently small,
inf | 4 ,
5
t # 0 ,
if
j e e
0
)
'4.5'
(3 .27)
W
2 iinagcoseg
3
sin 5,
L
cos
(3 .2S)
0") For the analyticityof^, / ?
2
:
Inf |A, , | > (2 sin
2
5 + 3 sin
2
S!) 12 sin 5 cos 3.. + 3 sin 8"! eos s!l
- 41^ - CJA/J - A'i, C
2
, A* = const.
ForA/j, Af
2
. ft sufficiently small we will have
i nf |A, , 1 * 0,
ze e(f) ' *
i f
|2 sin 6J
+
cos 6
+
+ 3 sin S{ cos 5[| * 0 .
For the denominator Q
2
w e n a v e
inr I <Bjl > 12 sin
2
3 + 3 sin
2
all
2
- K'b,
i6e(f)
2 + + l
and for b -* 0 as long as we have
K' = const.,
12 sin' 4?
a so
+ 3 sin.
2
Sj l
2
* 0,
inr |C>,|5*0.
ree(r
0
) *
(3.29)
(3.30)
(3.31)
(3 .3 2)
(3 .3 3 )
If, given the observables a. P. D, D
v
C, there exist o
a
, 5!J
+
, S | satisfying (3.26).
(3.28), (3.30), (3.32), such that the Af
r
- are small enough to ensure the existence of
4 - 0 fulfillius conditions (3.25), (3.27). (3.29), (3.31 ), Q.22) then the fi ( =
1,3,4,5) are analytic inside the ellipse ( Q) (the cut ellipse for a = 2). Therefore
we can convert eq. (3.14) into the form
I
'i'T i $ 2/--)(/?
s
(.-) + 2^(.-))d:,
3< la)
19 9
3 10 M Manolessau-Grammaiicnu / Unitary analytic amplitude
+ 2 (2 7 (7 +1)2 / - iSjCOt f jOO] d=.
- ^ T ^ O O + ^ 7 ^ v P - i e ;_i
2
f r )l (L- . ( 3 .3 4 )
L et us first lo o k at the co n ditio n (a) o f the theorem (B C) and prove that the bail
Sjf is mapped in to itself. We write this:
U'-\\<b, when I U- ^ | | < 4 . ( 3 .3 S)
A stronger co n ditio n to satisfy would be the following:
| | %( /I,f f .p)- ^ | | + II/- r i i + ||7 l |< *, when | H- I | | < 4 . ( 3 .3 6 )
In order t o calculate the contributions s u p/ lajj^ - 3%jQjQ;(j, we separate the
partial waves OQ, a
t
, a\ from the other a^. As will be clear later, the m odulus
iA
a
\ must satisfy appropriate constraints in order that the mapping be con tractive.
and it appears o n ly in the estim ations o f the quantities \n - I|, |ff*
+
- a ' ,!.
|d{ - }l . We have then
Q
a
i) SoCf,)
v
ff"rr c
0
(,)
|> *0- O,++, ( 3 .3 7 )
t f - j, I j- | l l l +l | l l r j - r {| | r i
+
r [ l
e, (f, )
s
e,(f,) a,,)
- t 3 , 3
'
200
M. Mauolessoii'Grammalicou / Unitary analytic amplitude 311
We obtain directly the following bounds:
! | ^bJ
1
1
+ i
f l j | <
W +
2!ft). M = 0
1 + +
, 1. '(3.39)
lrW + r
0
%' | < 2 | s i n^cos
1
,
1
| , flJ-O.t+,1. (3.40)
From q. (3.34) and taking into account the bounds (3.2S), (3.27), (3.29), (3,33)
we demonstrate in appendix B the following inequalities:
' K "
F
2 l < l < W t
c
4,S
M
4.S
+ C
0*
2 +
*^.
S
( ) ] i .
|ri - r Jl < I2
0
( f
0
) i [ S c,.Aj; + r
0
ft
2
+ 4( F
4 >5
( ) *i'
3
(8) t ^( 6 ) ) ] L .
Ir
+
- ?
+
l <l<2
0
(r
0
)l [ S c,,V, + cj? + i( ^i,
2
( ) + ^
3
( i ] i . (3.41)
where c
;
, i" = 1, . . . . 5, c
0
are constants and /. the length of ellipse d(jr
0
) and
j| 2 sinS
+
cos j
+
+3sin5"| cos5}I
2
+ K
2
12sin
2
5 + 3sin
2
S[i
2
| 2 sin
2
5
t
+3sin
2
S| l| 2 sin5 , . cos5
+
+ 3siiiJ}cos6[l
E
*<(3 $-sin
2
S} + 2si n
2
)
4 , 5
[2|sin 5g cos Sgl - 3 f
0
lsin s\ 'cos S| I] 5
A ^M s i n ^ + Ssi n
2
^!
jj t,A
(
= const. i= t, . . . , '
>ll' from the bounds (3.IE
(3.42)
, ( 5) = ^ - --r (Kj = const, i = I, . . . . 4
l-2 sin6
+
cos5, +3sin5"[ cos5|| from the bounds (3.IS)).
For the remaining afo, we verify easily that
Because of bounds (3.39)...(3.43), the condition (3.36) reads
11/ - 7|| + c
g
b
2
+ S c
t
M
f
+ (C - 1)6 < 0, (3.44)
withcq',Ci = const and C* C(5) given explicitly in appendix B. For the condition
(b) of the (BC) theorem which gives the continuity and the contractivity of the non-
linear mapping,
U
t
W( /1i, o, p) -%( . 4
2
, o, p) | | < A-| | . -l
1
- /Jjll. K<\,
we proceed in the same manner. We separate the contributions la^'J^) - jfj^l
201
312 M. Manotniou-Crammalicoit / Unitary analytic amplitude
flB 0, ++, 1 from the others: | f l /
w l )
- j^ri)!- V'B * 0, ++, 1 and finally after a
long but straightforward calculation (sec rf. [fil, app. C) wc Hud that
sup "ff" n ?
W M
<6V> + C+ S f. il/. ) I M. -/ M. ' 0*0-0,++. i, (3.4S)
up "'"g'g'^'"
2
' < ( t |
|
H&
r

l
) l i l | - i )
1
l , 0*0*0,++,!. (3.46)
It follows tliat (b) is verified with
*3cf> + C+ S
C /
A f
(
<l , (3.47)
where Co. c,- = const C = C(5), the same as in (3.44).
IfweiinposeonSj.J^.j the condition
C ()<l, (3.48)
it is dear that fority-, 1=1,...; S and H/- 7I small enough there exists a positive b
satisfying simultaneously (3.44), and (3.47). We then conclude that the existence
and uniqueness of a fixed point inside S
b
is ensured.
3.3. Schmtdcr's theorem
The conditions of continuity (3.45). (3.46) together with the requirement of
relative compactness of the image ^( S j ) of the ball5
6
, are sufficient to ensure the
existence of a fixed point in S
b
(but not the uniqueness) by Schauder*s principle.
We proceed to the proof of this supplementary condition through the use of the
following criterion [12]:
If in a Banach space of sequences {a
lt
} with norm fall- sup,, lal for anye>0
there exists an integer iV(e), such that |a|<e(n>A
/
(e)) for every {a,} S SCB then
S Is compact. To this end wchave to ensure first a domain of analyticity of/?,-(-) .
greater than cT(f
0
) in order to be able to shift the integration path from 3(f
0
) to
Sfftj), in order that a bound of the form I jjfyl < const 2y(f
0
) may be obtained. The
application of the criterion becomes then straightforward. The ellipse C(i"
n
) is con-
tained iu the domain of analyticity of/fy and .4,-. J* <T Q <J
I f
and is such that the
square roots in (fy) are zero free, (the conditions (3.261. (3.2S). (3.30). (3.32) be-
trig satisfied). Since the zeros are depending on the pointez) in .Sg. we have to
prove that we can always choose a point f
u
e ( :
a
); r
0
>f
0
> f
u
such'that inde-
pendent of .4 the square roots in R
t
are zero free when : S ( ^t. In fact by differ-
entiation wc find that the |A,| / I,..., 5, a^ir/or are uniformly bounded because
1/1,1 <u,IMII,
& " * " ' " >
fif. J*2
c o n s
t - (
S C l a
P P - A)*
(3 .49)
202
M. Afanatcssau-Grammatfcou ( Unitary analytic amplitude 313
Consequently, there exists a point YQ > ?o independent of A such that the square
root of R( is always zero free, if the path of integration is shifted to 3(?)and
For Ihc image of the ball S
b
we then obtain
sup max
n
... . < i-J sup ... . + sup . < sup - ... . ,\.
+ sun X; rt,70,l.,+-, X^Xj = cons'. (3.50)
But
A T
[e#o)]
2
[e,(r
0
)]
2
[Q^,)i
"C AT *& w "e^J "* -
(3
-
52)
*
AT-
Because of (3.51), (3.52).we obtain from(3.50) that for every e > 0 there exists
sup max
p p
< e . (3.53)
It follows that the set ^ C{S
b
) is compact, and the existence of a fixed point inside
S
b
isensiyed.
4. Discrete ambiguities
We have shown that if the five observables (six for non-identical particles) vary
in a neighborhood of the values Q, D, D
t
, C
m
, P, there is locally a 'jnique solution
of the mapping inside a compact arbitrarily chosen region of the Bunach space.
Now, because of the- double signs of the square roots in cqs. (3.15) there are
eight different balls (sixteen for non-identical particles), every one containing one
fixed point, in other words we have S(16) sets of invariant amplitudes satisfying
analylicity and unitarity, and corresponding to the same set of values of the ob-
servables a, P, D, D
{
. C
mr
Independent of this trivial ambiguity, we have some other discrete ambiguity
arising from the invariance that the 5(6) observables present under suitable trans-
formations of the scattering amplitude. In a manner analogous to that used by AMY
In ref. [2] we can define the following transformations:
* i cwi '/ e/ fi ) < K WI ' / C I W <* '' m. *w-m.
203
M, Manolcssou-Crammaticou J Unitary analytic amplitude
r, -JM ( 4 *) - -*/*( * **). (4.i)
T
2
= M(q, ft * (c
{
w')(<r
2
*q)A/tf, )(fl
t
)f>
2
'*). (4-2)
where 7*2 corresponds to the "Miuami" transformation in the case of spin 0 \ [2\.
We can verify that the scattering amplitude transformation by 7% always satisfies
the conservation taws of probability (unitarity) of parity and of time reversal, and
of course the differential cross section; consequently our conclusions arc in accord-
ance with the analysis done by Ryndin and Smorodinski [I3j, concerning the
"Minami" type transformations in N*N. The law of transformation of the observ-
ables under the operations 7*p 7*
2
and 7*j X 7% - r
3
is given in the following table,
where we have listed in addition the symmetries of some other useful observables.
We remark that the 5(6) observables used, arc even under r
3
. which means that
we have a double ambiguity of the total amplitude. We have thus finally 16 differ-
ent sets of invariant amplitudes (32 for the non-identical particles) that correspond
to the same values of the 5(6) observables. We can eliminate this supplementary
ambiguity due to the 7*
3
symmetry, if we measure one of the sets,
{Cpp. Cpk, Ckk} or {Dpp, Dpk, Dkk}.
Observable T
l
T
2
T
i
0 + +
*
pm _ _ *
em
- -
*
D + +
*
*>t
+ .
+
*
Sm
+ + +
r-.-.
+ + +
<v*-
-
+
-
< %t
+ + +
D
i
+ + t
tyj
-
+
-
Dib
+ + +
with the definitions
qXk',
i
xk'
" i
kXA'-q, n Xq -*',
RXq'i, k XA
*'
Ctf.*-\sinBC
pp
+{sineC
klc
+cos6 C
pl!
Q* i ( I- c o s 9 ) C
w
-sinS C ^+ i ( l + eoj fl ) C
t t
Dtf jO+cosfl) D + sin 9 D
pk
- l ( l -cosa ) D
kk
D
*?~i " O^ + isin O
t t
+ cos0 D
kp
D
ik --l (I -cosfl ) D
pp
+ sinff D
pk
*\(HcosO)D
kk
204
M. Afanotcssoti'Grammaticott / Unitary analytic amplitude 315
4.1. Different choices of observables
If instead of observables a, P. D, D
lt
C
m
we choose a different set always "com-
plete" (taking unitarity into account) in order to construct the scattering amplitude,
the form of fuactionals describing [he non-linear mapping is changed; tlte problem
will require a different approach in practice. It is therefore interesting to explore in
these other cases the constraints imposed on the various possible sets of observables.
Moreover, sometimes Ihe structure of the non-linear mapping is simpler. For ex-
ample the singularities at the origin may disappear. In appendix D we have listed
six different sets of observables (the first being the one we have used for the contrac-
tive mapping theorem)- We have taken.into account their dependence as functions
of the invariant amplitudes, using the definitions given in the literature [SJ. For
every "complete" set we distinguish three columns; in the first one we have written
the set of observables experimentally measurable. The second gives the definition
of bilinear or quadratic forms used as a new observable, expressed as functions of
the five invariant amplitudes; the relation between these observables and the original
ones is linear. In the third column we have expressed the dispersive parts as func-
tions of the "new observables" and the absorptive parts. We have obtained the
various forms of the mapping for the different choices of observables and we remark
that when an observable of the form Re i' (F
(
/^), (as is the case of the polarization;
P- Re i(Ff F2)) does not appear among the selected set, we have no singularities
arising from the zeros of denominators. Consequently we have the possibility 10
centre the ball S
b
at the origin A -Q and to obtain directly the conditions for the
validity of fixed point theorems; therefore such cases are simpler for numerical ap-
plications.
Among the six possibilities considered here two are simpler for the above reasons,
namely set (3): R, A, R\A'), o(i+D-D
t
- C
m
), 0, and set (5): R
t
, .4, R\(A[),
o(l C ~D
t
- D), a. (It should be pointed out that R, A are observables here and
not the absorptive and dispersive parts of amplitudes).
5. Inelastic ambiguity and Newton-Kantorovich method
In the present we consider the research of fixed points outside the domain that
can be handled by Schauder's theorem. The essential point here is to find out the
exact solution of the problem, given an approximate solution. The mathematical
technique for this purpose is the Newton-Kantorovich (14] method. In the inelastic
region there is a continuous ambiguity as was the case for scalar [3, 1 ] particles. In
other words, there exists in general an infinity of solutions corresponding to the
same set of observables a, P, D. /?,. C
mr
We have found that for sufficiently small
Afj and 11/ - 7||, there is always a solution in the neighborhood ai A. For a given
7(r), the knowledge of o
i n c
i only r est r i ct s/- 7 in the forward direction.; - l . but
there is still a continuum of solutions corresponding to different !(:). In fact even
205
316 M. Manolcssou-Grznnmaticait t Unitary analytic amplitude
if one could measure all the possible observables one would not in general remove
lllis ambiguity, as all observables arc invariant under the transformation
M(ij.q') ~e< nf(< j.), (S.1)
where c *
f c
' is an analytic function, becoming a simple ::-dependent phase factor
inside the physical region. The inequalities imposed by unitarity (in lire inelastic
region), allow in general &) to vary continuously inside a certain domain. We ob-
tain thus a continuous ambiguity. We will study the solutions of the mapping
(3.12)-(3.16) where the only know quantities are the observables: a. P, D, O
t
,
C
m
. Let us use the notation S
J
fA, I) for the non-linear operators: QJ = 0, 1, , +- )
sft<./)-<f-(f)*-0f)
2
-^.
sijA. i)'"U- (fiiJ
2
-<fU)
2
-<.iJ
2
-of J
2
-ii (5.2)
( live given explicitly in eqs. (3.15)). We have to solve the equations: S$( A, I) = 0
which correspond to the equation
lmf
,
= i( f
J
*.f
J
} + l( K
J
-H/), with \( K
J
+ L
J
) = I
J
, (3.4.1)
and with the supplementary condition (3.4.2). The operators K
J
, L/ have certain
positivity properties as previously described.
Let us suppose that we know the solution AQ of eqs. (3.4.1) for an inelasticity/Q.
To first order in the variations SA = A - Ar,, SI = I - 1
0
, one writes
S - S a { - 6 l { * 0 , (0 = a = O, l , , ' +-). (5,3)
We remark that the Frcchet derivative dS( 5a docs not depend on / and can be written
| | 1 -M( A
0
), (5.4)
where
Ho &' - a j j t r . , - ^ ! ^ - <
5
-
5
>
We show in appendix C that iU(.4|)) is a compact linear operator on the space B,
which means that eq. (5.3) is a Fredholm equation. If the 1 is not an eigenvalue of
M( A$, cq. (5.3) has a unique solution SA which is a continuous function of 5/. If
1 b an eigenvalue of J/, its multiplicity is finite and eq. (5.3} has a solution only [I ]
if 67 is orthogonal in thestibspacc of ff corresponding on the eigenvalue 1 of the
operator JU*1.-0. The conditions of orthogonality being always of finite number, it
" remains always a continuous ambiguity.
206
M Ahnolcssou'GramitietKtMt / Unitary analytic amplitude 317
When the variation is finite, eq. (5. 3) which is ouiv to first order in 5/1, is not
sufficient, and we then use the equations of the Ncwtou-Kantorovich method:
A
*l "
A
n - [' - W i ) ]
- 1
^ ' ) . (5. 6)
where-4;, is the nth order iteration of the procedure a nd^o is the approximative
solution.
As in rcf. [2] wc can easily verify that for 5/ sufficiently small the convergence
conditions of the procedure are satisfied and the limit oA
n
is the solution of ( 5. 3) .
(We assume that unity is not an eigenvalue of Af). What remains to be verified is that
K
J
, U have the required positivity proporties. What we can say is that this would not
be true for every 67, but generally we can expect the existence of a domain of 5/
where these posiiivity conditions are satisfied. .Moreover, it may happen, during the
Iteration, that singularities of dispersive-parts R
{
appear inside the ellipsc' ( f
0
) ; it
is possible that the double zeros of square roots <\ may be separated inside the ellipse
and create branch points (parasites [2] ) . The problem has been treated extensively hi
rcf. [15] and following the same line we find analogous conditions [6J.
6. Conclusion
. In this work we have shown tiiat the results of the scalar and spin 0 - { processes
by Atkinson-Mahoux-Yndurain [1, 2 ] can be generalized in a natural way to the
case of spin 5 -{ scattering. However the peculiarities of the higher spin and Paul!
principle must be taken into account.
In the elastic case we have been able to show, under certain restrictive conditions
on the 5 ( 6) observables ( a, D, D
t
. P, C
mi
), the local uniqueness of a solution (apart
from an eight-fold trivial ambiguity) in the neighbourhood of a point A B which
has only the first three partial waves different from zero.
This local uniqueness can be proved in other suitable neighborhoods if wc choose
different sets of observables. Nevertheless because of certain symmetries of the ob- ^
servables under a transformation analogous to that of Minami for the :rN case, the
discrete ambiguity is 16-fold (32-fold for the non-identical particles).
For the elastic region this result enables one, by measuring a set of only five ob-
servables to find all other useful ones for a N-N scattering process. However this
method might be of small practical interest because of the stringent conditions im-
posed on the observables, for the applicability of the contraction mapping or
Schauder theorems.
A better way to numerically investigate the fixed points in the elastic region is
the Newton-Kantorovich iteration. The same method is the main exploratory tool
in the inelastic region. In this region the ambiguity is continuous in spite of the con-
straints imposed in order to ensure the convergence of the iteration. One could thus
apply theNewton-Kantorovich method and investigate these ambiguities [15] . The
work presented in this article has been the subject of a "Doctora t do Troisime C ycle"
207
318 il. llaiiottiSUHi-Griillmatictxt / Unitary analytic ainplitittjc
thesis of the author, supported at Universit" Paris XI, Centre d'Orsay (June 1974)
and this is the reason why we have referred to it tor many detailed explanations and
references. Moreover the problem of non-identical particles (absence ol'Pauli sym-
metries) is treated there separately following the AMY approach for spin 0-j.
The author would like to thank Prof. D. Atkinson and Dr. C. Mahoux for an in-
troduction to the subject and for their help and advice during the course of this
work. She is also greatly indebted to Drs. R. Alvarez-Estrada, J. Bros, J. Raynal,
M. Chemtob and G. Cohen-Tannoudji for many fruitful discussions.
Appendix A
We prove the bounds,-
\AJz)\< Kjr)U\\, o 1 . . 5 . (3.18)
From eqs. (3.12) we obtain
M,K Ea i ci / H ) ^ ~/ill_liC2/+i)/>.
w
- T J X J
*W*WJ *J UT T )^
+
^ " - ' *
From the definition of the norm it follows that
i<r/,i<iwnfl/r,j. (A-2)
Moreover, we have
\PJ( Z)\< PJ( ), (A.3)
1^(2)1 < lfj( X)i (A-4)
By inserting (AJ) (A.3) (A.4) in (A.I) we get
M, l <l | / l l [S s
j
(
l
){2(l / +l )(P
/
.
1
(5)t/ >
y t l
(=' ) + r'/5)) + 6(a/+[)l?l^(;)},
J
(A.S)
and the recurrence relations of P/.) allow one to write
W, | < \\A\\ S ( 6( 2y+l ) P , ( ) + 2(27 t DPjiz) + (2/+l)(2 H)f
J
G))Qj( ii)
'
J
(A-6)
Using the fact that
2 ( 2 / ^) ^) 2 ^, ) = ^,
203
At. AfaiiolessM'Granmiatfcau / Unitary analytic amplitude 319
S ( 2 /
+
l ) ^( ) Q/
? 1
) = ^ - ^ . (A.7)
we obtain finally from (A.6)
W, K M B l ( 2 i * j p f ^ ) - * , e > M l . (A.8)
In the same manner we have
Uj K IUi l T - H i +;r?-=: =K.m\A\\. (A.9)
w, i < M I I (
4 ? +
' + jr^"-) = K, <?) IM II, ( A. IO)
M
4
I< M i l f
3 + r
,
+
r ^
T
) =A'. ( ) IWII, (A.11)
H
5
I < 11-411 f -
1
^ + r - ^r ) = A'
S
(=)IUII. (A.12)
We remark that A
(
(?) are bounded functions of ?when?< f
0
. The bounds
\dAfi:l< j)IMII, follow easily in the same way.
Appendix B
We prove bounds for the | ^ - pj^l. We have_
Ir g - f j K I < t l S
0
t t
0
) ( 2 | f i
4
- g
4
l + | R
s
- R
s
l ) . (B.l)
Taking into account the bounds (3.18), (3.27), (3.28) and the definitions (3.24)
v/e obtain
__
| j
,ji(r*oto
r
c
M
)-^
t
-tG('
;
g*B,-g,.)*^5i
< 5 4
'
S
[2|sin6cosSg| -3r
0
l si n5| cosjlji
2 A'
4
, ( 3f
o S
m
2
6[+2 sin
2
3<J)
4
'
3 4
'
5
pisinSgcossOl-JrolsmS^osslU!
<:,<:, i * I . . . 5const. , (B.2)
and the result announced in eq. (3.41) follows directly.
209
3 2 0 M NaitoksuU'Gramimtlcou t Unitary amtytic amptitiul
In an analogous manner,
If j - ?{l < kf>d:| {1 2,(01 [\R
S
-
5
I+2|R
4
- R
4
\)+1 Q',(.-)| [Ifij- S
s
l+2|/J
4
- g
4
|
+ K, - JJ
1
l ] + IG;(.-)|[2|3-S
3
l + ! 2
: !
- I l i | , - ^, I I }. (BJ)
Using 'he bounds (3.1 3), (3.25), (3.26), (3.29), (3.30) and the definitions (3.23),
(3.24 ) we have
&.IA.
1
-.,\ l/l
L
,vS;il<B
2
-<5,| |X
L
,llv/5T-v/S,
w-*w' < %
t
+ ^ Ai -
+
^Wj
l
' .
K. Ps i n ^cos S^ + SsinS'cosSjl
2
+ K
2
|2 sin
2
0 ^ + 3 sin
2
?}!
1
<C
r
,4
2
+c,M, tc,*f,+6 r; r; , j
" * ' ' * | 2s m
2
c ^ + 3sin
2
s[iesmS
t
cos6
+
+3sin5}cos5|l
(B.4 )
1-2 sin
2
o* + 3 sin
2
Sil
\R

-R^<e

M
3
*c
0
b
2
*K
3
b -z . , _, i . (B.S)
Furthermore,
is^)i<ie,(f
0
)i2
0
(i-
0
), (B.6)
From inequalities (B.2)-(B.7), we obtain for lf[ - 7 J| and l r j
+
- fj
+
l the announced
result in eq. (3.4 1 ).
Hie explicit fonn we obtain for C(5) in eqs. (3.4 4 ), (3.4 7) is tnen the following:
C-max{C<>.Cj
+
.C}}, (B.S)
with
qf-2sm 4
0
.+ g
o ( f i )
+
4
Q ( ) f f l
) ' t '
C
,
2 s m
a
t + +
_
t
- _ . ( B..0)
-r. . -r. , 12 Si<l 5} COS } | 1 - r j '
2
e}-a.i
B
8i + iy
t
gg.{)+s
I
tg.,g|)]
Q
^ *3S^ j - <
B1 I
>
and
S
l
(Sg.5;) = 2/.0
u
(r
0
)/f
4 i S
CS). (B->2)
5, (S;
t
, [) = 2LQ
a
d
a
)[K
3
<,S) + A-, ,()], (^(5), I - 1 ... 5 given in (3.4 2)).
210
M. Manolcssoit-Grammaticoti / Unitary analytic amplitude 321
Appendix C
We prove the compactness of the operator M( A). We have defined
iri
W%i'i<>i*jr
i
tv-
2,
i--f- W-o, i, , +-. (c.i)
bag
We must ensure that the image M( A) of every closed bounded subset of B is com-
pact. We use the same criterion as for the Schauder theorem. We have then to prove
that
ff
Wit'
sup " - ... <e VJ>N( e), JV-, (C.2)
for &a{ such that
I S ^KC Q, ^, ). (C.3)
Because of conditions (3.49), lijl and I0
(
/3;| ( i = 1,... 5) are uniformly bounded
Thus independent otM we can again choose a point f
0
S ct(i"
0
): r
0
> f
0
> f
0
such
that (lie square roots ^ of ,- are zero free, and we are able to shift the integration
path from 3(f
0
) * 3(?
0
) and analogously C (J [) -* <5(?i): i\ >f
t
. Consequently we
can obtain the following bounds:
I^KM lie/?!) , (C.4)
|r/l<c2
y
( F
0
) , (C.5)
*i
-~ < c'Qj$
a
)U\\, c.c' = const. (C.6)
da
b
.
Because of (C.3)-(C.6) we finally obtain
r _ ,G
y
(F,) e/F. h
< c
1
M i i [ c
2
(
G
/ r
0
) ) - ^y
+
c
3
^^o .
C|, Cj, c
3
" const. J>N,N-*,
and (C.2) follows.
Appendix D
'In all the following lists we have used the notations for the invariant amplitudes
which arc most frequently fuund in the literature:
211
M. Maiioleuoit-Graimitaticou t Unitary analytic amplitude
F
l
, F j t. ^3 " /l, ^ m, >j C.
(0 .
V. fl
r
ft< ^
Re (d+0* ic*p
i
tsl
2
+ l A |
2
- p
2
l ^ l - l ^ - P j
|<i|

+li|
2
+2|c|
2
+2tst
2
+l''l
2
=P4
(2)
R e( a- m) c*p
2
R ef / i p
3
|Re(f(a+m)c)*p
4
|al
I
+|ml
2
+2kl
2
+2lsl
2
+2l''l
2
=Pj
Re-i !
,+-4j)p
2
+-4
s
(14|-p,) ' "4 J
j r (p,-/l
2
-"/l|)p
4
( A
l
+A
2
)p
1
*A
i
OA^-p
l
)l
Rent's-I ;
.
2
L (A , t/l
2
)p
2
+/l
5
(2. 4
2
-p
1
) "4 J
y>2-( A
l
-A
1
)A
s
]p
i
R ec- r
(Ai+AfatAf-pJ
| .R ej-Je, VA' + 2P 3+3e
2
VA '-2pj.
2/: - p
s
- T/A\-Re
2
a - Re3 m - ZRe-c
jRcA - e, VK+2pj - T
2
VA'-2p'
3
p
3
- p
3
- , (
3
, l
4
212
M. Maitolessou-Gratimtaticott / Unitary analytic amplitude
( 3 )
R.A,R\A'l
l
0
(.l+D-D
r
C
m
ir
a
l a P - l m l ' a p!
Rc ( a - i H)*
=
P2
Rc /i * = p
3
|a l
2
+|i l
2
t 2 |e pt 2 1g|
2
+2 |/i |
2
H
/
i
s
* p
4
- ( / i
1
- / i , ) -
(4)
Re (< !+/ )?* =P j
Rc s p
2
Re ( a - m )A*s pj
Re i ' ( j+ 0*c = p
4
|a |
2
+| i |
2
+2 |< : l
2
+2 l jr l
2
+2 l /i |
2
=p
5
( p
t
- /l , , 4
4
)p
4
Re ( a +m ) =
"i
A
l-Pl
A
2
Re c =
(P
2
--^; ^
4
)P4
i 4 ,
s
Im ( a + i )
yi
3
= Im A
/ J
4
" l m ;
<4<"l m(<i -i )
Re ( a - M )
s
[
l
Vp
1
+4 p,+e
2
> /p
1
- 4 P }]
p-,A. - p. A,
Re t
Rc 'i "
,
i [e
1
Vp^ +- 4 p3 - f
2
VP 4 - 4 p3 ]
213
M. Afanotcssau-Cranwialicoti f Unitary ana
^ [4( pj / l
1
- p
1
/ ( j ) *
rf
2 p
s
p ' j - p j - ^ / i j
W W
Re(a + Mi)s* = pj
ReC*sp
2
Re( a- m ) / i *sp
3
l *l
2
*P
4
|a|
2
+|m |
2
+2|c|
2
+2lsl
2
+:i/i|
2
=p
3
Re (+m )--
Rec
RA- i [e
2
\^- <
3
v_]
R * " J V P
4
- ^
Re(fl-m)'{ [2-v^V + 3\'^.
p
+
- 2 ( Pj - 2 p
4
+2 p'
3
- / l
2
- / ) ;
- M| - 4 / l
2
- 2 ^ 4 - 2 ^ j / l 3 )
( p
r
^
t
>l
4
)
2
+ 4 ( p , - / M
Pi.~
A
l
2 14
()
if. MaimtasoU'Grammaticott / Unitary analytic amplitude 3 2 5
| p. . 2 Co
5
- 2 p
4
- 2 p
3
- ^ - / l |
- 4 ,4 !- 4 /i 5 - 2 ,1j- 2 ,t
s
,1
3
)
j) k p
Re i "( a - m )*c = P j
Ifi+D-DrCJ
Re i a m * s p
2
'o
Re i g/ ' *
3
P
3
Ia - m |
2
=p
4
|a|
2
+|B>|
2
+2|el
2
+2!sl
2
+2[/.l
2
=Pj
Re wi -
p2~A
1
e
l
\/pp<-^r
A,-A,
R5 C=- . - r -
A
L
-A
2
, / i jVp4 - ( . 4
l
- /
:
)
: !
Re /i =
"l
-
"2
- Z4
4
p
3
+e
2
Vw
2
p5 - 2 M
2
+..l
2
)C
2M
2
+/lf)
"3
A
* .
with
X, = Im a
X* Im m
A
3
= Im c c 2 p5 - P
s
/l | *2 p
2
- . 4
2
[ p
s
- , 4
2
- / i
2
- 2 / 4
2
- l . |
2
- 2 / l
2
- Re
2
s - Rs
2
w
A
*~
Im
* - 2 Re
2
c ]
/l j Im A
Appendix E
^ , 4 ?
2
[
( , J t | ) f
' - i - 7 ] - - *
2
[
( , J t l ) P
- ^ ] '
+( V+1
215
326 M. AanolcaohGramiriaticau / Unitary analytic amplitude
^2 4ir 7 y^- i
4
T fF v+i "~37+ J v
a
i -
^3 " g? ^ - T - "- - -7TT "
+
3(7Ti7 ^"i
->f Cf)
2
+frf)
2
+Ki-<<>
2
),
< (<)
2
+ C d)
2
+ tfj
2
+ i'iJ - + id - (if

)
2
).
il ">{. <si+*L..*>iM+ **i-)*iaL i{ J ^P- ' *- ' . <3.n>
where
Xs!n ( #- - / ) .
ind it^-Sna arc the inelasticity parameters and the phase shifts for the correspond-
ing/*,; ((J-', vr
7
) arc tlic modulus and phase of mixing parameters between states
W - l - W + i [IS].
r$'x / p, (
s
+2K
4
) d( cos0) ,
-I
t
f t W
{ -r'~ * = = fVUR.-2R,-K.)
+ 2(2/ (/ +i y
/
- cos d/jJRjldCcos ),
216
M. Matiolitstoii-Gramttiatlcott / Unitary analytic amntirtttlc 327
1
+ 2f^
3
+PjR
2
]d(cosfl),
- ^J , - f i ! J ^* , ] d(cos6), (3.14)
4aB4. 2 f M , , > , _
R
A
-
fi
lic(.l-D*D
t
-C
m
)-Al)i, e
3
.
^..jpodtiJ-D.-C ^-^li (3.1S)
- fPj(A
s
+ Z4J i(cc*6),
- 1
*P
l
/i
2
] d(cosff).
217
328 M. Atattotnuou-Gratttmaticott ( Unitary analytic amplitude
(3.16)
References
111 D. Atkinson, G. Mahoux and F.J. Yndurn. Nucl. Phys. D54 (19731 3S3.
121 D. Atkinson, G. Mahoux and F.J. Yndurn, Nucl. Pliys. B66 (1973) 429.
( 3| D. Atkinson, F.W. Johnson and A.L. Warnock, Connu. Math. Pliys. 28 (1972) 133.
[4| A.Martin,NuovoCimcnto59A(l969) 131.
| 5| R.G. Newton. J. Malli. Phys. 9 (1968) 2050.
(6) M.Munolcssou-Graimmticou, Thesis, Universit" Patis XI. Orsay. (June 1974).
[7| R.F. Alvarez-Estrada and B. Carreras, Nucl. Pliys. B54 (1973) 237.
R.F. Alvarez-Estrada, B. Catrctas and M.A. Goni, Nucl. Pliys. B62 (1973) 221; II, JEN
Preprint FT CR 115.
( S| N. Hoshizaki, Projtr. Thcor. Phys. Suppl. 42 (1968) 1.
[91 J. Bystricky. F. Lehar and Z. Janout, Elastic NN scattering data 270-3000 McV; Ncte
CEA-N-1547 (E) (August 1972).
[101 R.F. Alvarez-Estrada. B. Carreras and G. Mahoux, Nucl. Pliys. BS8 (1975) 2S9.
[11] M.A. Krasnosd'skii, Topological methods in die theory of non-linear integral equations
(MacMillan, New York, 1964).
[12] D. Atkinson, P.VY. Johnson and R.L. Warnock. Coram. Math. Phys. 33 (1973) 221.
[131 R. Ryndin and J. Smorudinsky, CERN Symposium Proc.. vol. 2 (1956).
[14] M.M. Vainocre, in L.V. Kantorovich and G.P. Akilov, Variational methods for the study of
non-linear operators (llotden-Day, 1964), see ch. 7.
| IS] D. Atkinson. P.W. Johnson, L.P. Kok and M. De Roo. Nucl. Pliys. B77 (1974) 109:
D. Atkinson and Nl. Kackebcke, M. De Roo, J. Math. Phys. 16 (1975) 6S5;
M. Dc Roo,Thesis, University of Groningcn (December 1974).