1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%%%%%%%%%%%%%%%%%%%%%%%% DEFINIZIONI DI FONT %%%%%%%%%%%%%%%%%%%%%%%%%%%
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: \font\tenmib=cmmib10
5: \font\sevenmib=cmmib10 scaled 800
6: \font\titolo=cmbx12
7: \font\titolone=cmbx10 scaled\magstep 2
8: \font \titolino=cmbx10
9: \font\cs=cmcsc10
10: \font\sc=cmcsc10
11: \font\css=cmcsc8
12: \font\ss=cmss10
13: \font\sss=cmss8
14: \font\crs=cmbx8
15: \font\ninerm=cmr9
16: \font\ottorm=cmr8
17: \textfont5=\tenmib\scriptfont5=\sevenmib\scriptscriptfont5=\fivei
18: \font\msxtw=msbm9 scaled\magstep1
19: \font\euftw=eufm9 scaled\magstep1
20: \font\euftww=eufm7 scaled\magstep1
21: \font\msytw=msbm9 scaled\magstep1
22: \font\msytww=msbm8 scaled\magstep1
23: \font\msytwww=msbm7 scaled\magstep1
24: \font\msytwwww=msbm5 scaled\magstep1
25: \font\msytwwwww=msbm4 scaled\magstep1
26: \font\indbf=cmbx10 scaled\magstep2
27: \font\type=cmtt10
28: \font\ottorm=cmr8\font\ottoi=cmmi8\font\ottosy=cmsy8%
29: \font\ottobf=cmbx8\font\ottott=cmtt8%
30: \font\ottocss=cmcsc8%
31: \font\ottosl=cmsl8\font\ottoit=cmti8%
32: \font\sixrm=cmr6\font\sixbf=cmbx6\font\sixi=cmmi6\font\sixsy=cmsy6%
33: \font\fiverm=cmr5\font\fivesy=cmsy5\font\fivei=cmmi5\font\fivebf=cmbx5%
34:
35: \def\ottopunti{\def\rm{\fam0\ottorm}%
36: \textfont0=\ottorm\scriptfont0=\sixrm\scriptscriptfont0=\fiverm%
37: \textfont1=\ottoi\scriptfont1=\sixi\scriptscriptfont1=\fivei%
38: \textfont2=\ottosy\scriptfont2=\sixsy\scriptscriptfont2=\fivesy%
39: \textfont3=\tenex\scriptfont3=\tenex\scriptscriptfont3=\tenex%
40: \textfont4=\ottocss\scriptfont4=\sc\scriptscriptfont4=\sc%
41: %\scriptfont4=\ottocss\scriptscriptfont4=\ottocss%
42: %\textfont5=\tenmib\scriptfont5=\sevenmib\scriptscriptfont5=\fivei%
43: \textfont\itfam=\ottoit\def\it{\fam\itfam\ottoit}%
44: \textfont\slfam=\ottosl\def\sl{\fam\slfam\ottosl}%
45: \textfont\ttfam=\ottott\def\tt{\fam\ttfam\ottott}%
46: \textfont\bffam=\ottobf\scriptfont\bffam=\sixbf%
47: \scriptscriptfont\bffam=\fivebf\def\bf{\fam\bffam\ottobf}%
48: %\tt\ttglue=.5em plus.25em minus.15em%
49: \setbox\strutbox=\hbox{\vrule height7pt depth2pt width0pt}%
50: \normalbaselineskip=9pt\let\sc=\sixrm\normalbaselines\rm}
51: \let\nota=\ottopunti%
52:
53: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
54: %%%%%%%%% RIFERIMENTI SIMBOLICI A FORMULE, PARAGRAFI E FIGURE %%%%%%%%%%
55: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
56: %
57: % Ogni paragrafo deve iniziare con il comando \section(#1,#2), dove #1
58: % e' il simbolo associato al paragrafo e #2 e' il titolo. Per le
59: % appendici bisogna pero' usare \appendix(#1,#2).
60: %
61: % Se nel titolo compaiono riferimenti ad altri simboli, questi vanno
62: % racchiusi fra parentesi graffe, per es. {\equ(1.2)}; in caso contrario
63: % si provoca un errore.
64: %
65: % Ogni sottoparagrafo deve iniziare con il comando \sub(#1) o \asub(#1),
66: % nelle appendici.
67: %
68: % I riferimenti a paragrafi e sottoparagrafi si realizzano con il comando
69: % \sec(#1), che produce il numero effettivo preceduto dal simbolo di
70: % paragrafo, o \secc(#1), che produce solo il numero (serve nel caso si
71: % faccia riferimento ad un sottoparagrafo, che e' un Lemma, un Teorema o
72: % altro oggetto suscettibile di una denominazione speciale).
73: %
74: % Le formule sono contrassegnate con \Eq(#1), eccetto che all'interno
75: % del comando \eqalignno, dove si deve usare \eq(#1). Nelle appendici
76: % i comandi corrispondenti sono \Eqa(#1) e \eqa(#1).
77: % I riferimenti alle formule si realizzano con \equ(#1).
78: %
79: % La numerazione delle figure utilizza il comando \eqg(#1), per
80: % contrassegnarle, e \graf(#1) per citarle.
81: %
82:
83: \global\newcount\numsec\global\newcount\numapp
84: \global\newcount\numfor\global\newcount\numfig
85: \global\newcount\numsub
86: \numsec=0\numapp=0\numfig=0
87: \def\veroparagrafo{\number\numsec}\def\veraformula{\number\numfor}
88: \def\veraappendice{\number\numapp}\def\verasub{\number\numsub}
89: \def\verafigura{\number\numfig}
90:
91: \def\section(#1,#2){\advance\numsec by 1\numfor=1\numsub=1\numfig=1%
92: \SIA p,#1,{\veroparagrafo} %
93: \write15{\string\Fp (#1){\secc(#1)}}%
94: \write16{ sec. #1 ==> \secc(#1) }%
95: \hbox to \hsize{\titolo\hfill \number\numsec. #2\hfill%
96: \expandafter{\alato(sec. #1)}}\*}
97:
98: \def\appendix(#1,#2){\advance\numapp by 1\numfor=1\numsub=1\numfig=1%
99: \SIA p,#1,{A\veraappendice} %
100: \write15{\string\Fp (#1){\secc(#1)}}%
101: \write16{ app. #1 ==> \secc(#1) }%
102: \hbox to \hsize{\titolo\hfill Appendix A\number\numapp. #2\hfill%
103: \expandafter{\alato(app. #1)}}\*}
104:
105: \def\senondefinito#1{\expandafter\ifx\csname#1\endcsname\relax}
106:
107: \def\SIA #1,#2,#3 {\senondefinito{#1#2}%
108: \expandafter\xdef\csname #1#2\endcsname{#3}\else
109: \write16{???? ma #1#2 e' gia' stato definito !!!!} \fi}
110:
111: \def \Fe(#1)#2{\SIA fe,#1,#2 }
112: \def \Fp(#1)#2{\SIA fp,#1,#2 }
113: \def \Fg(#1)#2{\SIA fg,#1,#2 }
114:
115: \def\etichetta(#1){(\veroparagrafo.\veraformula)%
116: \SIA e,#1,(\veroparagrafo.\veraformula) %
117: \global\advance\numfor by 1%
118: \write15{\string\Fe (#1){\equ(#1)}}%
119: \write16{ EQ #1 ==> \equ(#1) }}
120:
121: \def\etichettaa(#1){(A\veraappendice.\veraformula)%
122: \SIA e,#1,(A\veraappendice.\veraformula) %
123: \global\advance\numfor by 1%
124: \write15{\string\Fe (#1){\equ(#1)}}%
125: \write16{ EQ #1 ==> \equ(#1) }}
126:
127: %\def\getichetta(#1){Fig. \verafigura%
128: \def\getichetta(#1){\veroparagrafo.\verafigura%
129: \SIA g,#1,{\veroparagrafo.\verafigura} %
130: \global\advance\numfig by 1%
131: \write15{\string\Fg (#1){\graf(#1)}}%
132: \write16{ Fig. #1 ==> \graf(#1) }}
133:
134: \def\getichettaa(#1){A\veraappendice.\verafigura%
135: \SIA g,#1,{A\veraappendice.\verafigura} %
136: \global\advance\numfig by 1%
137: \write15{\string\Fg (#1){\graf(#1)}}%
138: \write16{ Fig. #1 ==> \graf(#1) }}
139:
140: \def\etichettap(#1){\veroparagrafo.\verasub%
141: \SIA p,#1,{\veroparagrafo.\verasub} %
142: \global\advance\numsub by 1%
143: \write15{\string\Fp (#1){\secc(#1)}}%
144: \write16{ par #1 ==> \secc(#1) }}
145:
146: \def\etichettapa(#1){A\veraappendice.\verasub%
147: \SIA p,#1,{A\veraappendice.\verasub} %
148: \global\advance\numsub by 1%
149: \write15{\string\Fp (#1){\secc(#1)}}%
150: \write16{ par #1 ==> \secc(#1) }}
151:
152: \def\Eq(#1){\eqno{\etichetta(#1)\alato(#1)}}
153: \def\eq(#1){\etichetta(#1)\alato(#1)}
154: \def\Eqa(#1){\eqno{\etichettaa(#1)\alato(#1)}}
155: \def\eqa(#1){\etichettaa(#1)\alato(#1)}
156: \def\eqg(#1){\getichetta(#1)\alato(fig. #1)}
157: \def\sub(#1){\0\palato(p. #1){\bf \etichettap(#1).}}
158: \def\asub(#1){\0\palato(p. #1){\bf \etichettapa(#1).}}
159:
160: \def\equv(#1){\senondefinito{fe#1}$\clubsuit$#1%
161: \write16{eq. #1 non e' (ancora) definita}%
162: \else\csname fe#1\endcsname\fi}
163: \def\grafv(#1){\senondefinito{fg#1}$\clubsuit$#1%
164: \write16{fig. #1 non e' (ancora) definito}%
165: \else\csname fg#1\endcsname\fi}
166: \def\secv(#1){\senondefinito{fp#1}$\clubsuit$#1%
167: \write16{par. #1 non e' (ancora) definito}%
168: \else\csname fp#1\endcsname\fi}
169:
170: \def\equ(#1){\senondefinito{e#1}\equv(#1)\else\csname e#1\endcsname\fi}
171: \def\graf(#1){\senondefinito{g#1}\grafv(#1)\else\csname g#1\endcsname\fi}
172: \def\figura(#1){{\css Figure} \getichetta(#1)}
173: \def\figuraa(#1){{\css Figure} \getichettaa(#1)}
174: %\def\fig(#1){\0\veroparagrafo.\getichetta(#1)}
175: \def\secc(#1){\senondefinito{p#1}\secv(#1)\else\csname p#1\endcsname\fi}
176: \def\sec(#1){{\S\secc(#1)}}
177: \def\refe(#1){{[\secc(#1)]}}
178:
179: \def\BOZZA{\bz=1
180: \def\alato(##1){\rlap{\kern-\hsize\kern-1.2truecm{$\scriptstyle##1$}}}
181: \def\palato(##1){\rlap{\kern-1.2truecm{$\scriptstyle##1$}}}
182: }
183:
184: \def\alato(#1){}
185: \def\galato(#1){}
186: \def\palato(#1){}
187:
188: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
189: %%%%%%%%%%%%%%%%%%%% DATA E PIE' DI PAGINA %%%%%%%%%%%%%%%%%%%%%%
190: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
191:
192: {\count255=\time\divide\count255 by 60 \xdef\hourmin{\number\count255}
193: \multiply\count255 by-60\advance\count255 by\time
194: \xdef\hourmin{\hourmin:\ifnum\count255<10 0\fi\the\count255}}
195:
196: \def\oramin{\hourmin }
197:
198: \def\data{\number\day/\ifcase\month\or gennaio \or febbraio \or marzo \or
199: aprile \or maggio \or giugno \or luglio \or agosto \or settembre
200: \or ottobre \or novembre \or dicembre \fi/\number\year;\ \oramin}
201: %\setbox200\hbox{$\scriptscriptstyle \data $}
202: \footline={\rlap{\hbox{\copy200}}\tenrm\hss \number\pageno\hss}
203:
204: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
205: %%%%%%%%%%%%%%% INSERIMENTO FIGURE ( se si usa DVIPS ) %%%%%%%%%%%%%%
206: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
207:
208: \newcount\driver %\driver=1 %dvips
209: %
210: \newdimen\xshift \newdimen\xwidth
211: \def\ins#1#2#3{\vbox to0pt{\kern-#2 \hbox{\kern#1
212: #3}\vss}\nointerlineskip}
213:
214: \def\insertplot#1#2#3#4#5{\par%
215: \xwidth=#1 \xshift=\hsize \advance\xshift by-\xwidth \divide\xshift by 2%
216: \yshift=#2 \divide\yshift by 2%
217: \line{\hskip\xshift \vbox to #2{\vfil%
218: \ifnum\driver=0 #3
219: \special{ps::[local,begin] gsave currentpoint translate}%
220: \special{ps: plotfile #4.eps} \special{ps::[end]grestore}\fi
221: \ifnum\driver=1 #3 \special{psfile=#4.eps}\fi
222: \ifnum\driver=2 #3
223: \ifnum\mgnf=0\special{#4.eps 1. 1. scale} \fi
224: \ifnum\mgnf=1\special{#4.eps 1.2 1.2 scale}\fi
225: \fi }\hfill \raise\yshift\hbox{#5}}}
226:
227: \def\insertplotttt#1#2#3#4{\par%
228: \xwidth=#1 \xshift=\hsize \advance\xshift by-\xwidth \divide\xshift by 2%
229: \yshift=#2 \divide\yshift by 2%
230: \line{\hskip\xshift \vbox to #2{\vfil%
231: \ifnum\driver=0 #3
232: \special{ps::[local,begin] gsave currentpoint translate}%
233: \special{ps: plotfile #4.eps} \special{ps::[end]grestore}\fi
234: \ifnum\driver=1 #3 \special{psfile=#4.eps}\fi
235: \ifnum\driver=2 #3
236: \ifnum\mgnf=0\special{#4.eps 1. 1. scale} \fi
237: \ifnum\mgnf=1\special{#4.eps 1.2 1.2 scale}\fi
238: \fi }\hfill}}
239:
240: \def\insertplott#1#2#3{\par%
241: \xwidth=#1 \xshift=\hsize \advance\xshift by-\xwidth \divide\xshift by 2%
242: \yshift=#2 \divide\yshift by 2%
243: \line{\hskip\xshift \vbox to #2{\vfil%
244: \ifnum\driver=0
245: \special{ps::[local,begin] gsave currentpoint translate}%
246: \special{ps: plotfile #3.eps} \special{ps::[end]grestore} \fi
247: \ifnum\driver=1 \special{psfile=#3.eps}\fi
248: \ifnum\driver=2
249: \ifnum\mgnf=0\special{#3.eps 1. 1. scale} \fi
250: \ifnum\mgnf=1\special{#3.eps 1.2 1.2 scale}\fi
251: \fi }\hfill}}
252:
253: \def\initfig#1{%
254: \catcode`\%=12\catcode`\{=12\catcode`\}=12
255: \catcode`\<=1\catcode`\>=2
256: \openout13=#1.ps}
257: \def\endfig{%
258: \closeout13
259: \catcode`\%=14\catcode`\{=1
260: \catcode`\}=2\catcode`\<=12\catcode`\>=12}
261:
262: \newdimen\xshift \newdimen\xwidth \newdimen\yshift
263: %
264: \def\eqfig#1#2#3#4#5{
265: \par\xwidth=#1 \xshift=\hsize \advance\xshift
266: by-\xwidth \divide\xshift by 2
267: \yshift=#2 \divide\yshift by 2
268: \line{\hglue\xshift \vbox to #2{\vfil
269: \ifnum\driver=0 #3
270: \special{ps::[local,begin] gsave currentpoint translate}%
271: \special{ps: plotfile #4.ps} \special{ps::[end]grestore}\fi
272: \ifnum\driver=1 #3 \special{psfile=#4.ps}\fi
273: \ifnum\driver=2 #3 \special{
274: \ifnum\mgnf=0 #4.ps 1. 1. scale \fi
275: \ifnum\mgnf=1 #4.ps 1.2 1.2 scale\fi}
276: \fi}\hfill\raise\yshift\hbox{#5}}}
277:
278:
279: \def\figini#1{
280: \def\8{\write13}
281: \catcode`\%=12\catcode`\{=12\catcode`\}=12
282: \catcode`\<=1\catcode`\>=2
283: \openout13=#1.ps}
284:
285: \def\figfin{
286: \closeout13
287: \catcode`\%=14\catcode`\{=1
288: \catcode`\}=2\catcode`\<=12\catcode`\>=12
289: }
290: %
291: % Se si vuole utilizzare delle macro postscript personali, contenute
292: % nel file ini.ps, togliere il commento alla riga seguente
293: %\special{header=ini.ps}
294: %
295: % Il comando seguente inserisce una scatola contenente #3 in modo che
296: % l'angolo superiore sinistro occupi la posizione (#1,#2)
297: %
298: %%\def\ins#1#2#3{\vbox to0pt{\kern-#2 \hbox{\kern#1 #3}\vss}\nointerlineskip}
299: %
300: % Il comando seguente crea una scatola di dimensioni #1x#2 contenente
301: % il disegno descritto in #4.ps;
302: % in questo disegno si possono introdurre delle stringhe usando \ins
303: % e mettendo le istruzioni relative nell'argomento #3.
304: % Il file #4.ps contiene le istruzioni postscript, che devono essere scritte
305: % presupponendo che l'origine sia nell'angolo inferiore sinistro della
306: % scatola, mentre per il resto l'ambiente grafico e' quello standard.
307: % #5 deve essere della forma \eqg("nome simbolico").
308: %
309: % Le istruzioni postscript possono essere inserite nel file che contiene
310: % l'istruzione \insertplot, racchiudendole fra le istruzioni \initfig{#4}
311: % e \endfig; inoltre ogni riga deve cominciare con "write13<" e deve finire
312: % con ">". In questo modo si crea il file #4.ps relativo alla figura.
313: %
314: %%\newdimen\xshift \newdimen\xwidth \newdimen\yshift
315: %%
316: %%\def\insertplot#1#2#3#4#5{\par%
317: %%\xwidth=#1 \xshift=\hsize \advance\xshift by-\xwidth \divide\xshift by 2%
318: %%\yshift=#2 \divide\yshift by 2%
319: %%\line{\hskip\xshift \vbox to #2{\vfil%
320: %%#3 \special{psfile=#4.ps}}\hfill \raise\yshift\hbox{#5}}}
321: %%
322: %%\def\initfig#1{%
323: %%\catcode`\%=12\catcode`\{=12\catcode`\}=12
324: %%\catcode`\<=1\catcode`\>=2
325: %%\openout13=#1.ps}
326: %%
327: %%\def\endfig{%
328: %%\closeout13
329: %%\catcode`\%=14\catcode`\{=1
330: %%\catcode`\}=2\catcode`\<=12\catcode`\>=12}
331:
332:
333: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
334: %%%%%%%%%%%%%%%%%%%%%% SIMBOLI VARI %%%%%%%%%%%%%%%%%%%%%%%
335: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
336:
337: \let\a=\alpha \let\b=\beta \let\g=\gamma \let\d=\delta \let\e=\varepsilon
338: \let\z=\zeta \let\h=\eta \let\th=\theta \let\k=\kappa \let\l=\lambda
339: \let\m=\mu \let\n=\nu \let\x=\xi \let\p=\pi \let\r=\rho
340: \let\s=\sigma \let\t=\tau \let\f=\varphi \let\ph=\varphi\let\c=\chi
341: \let\ps=\psi \let\y=\upsilon \let\o=\omega\let\si=\varsigma
342: \let\G=\Gamma \let\D=\Delta \let\Th=\Theta\let\L=\Lambda \let\X=\Xi
343: \let\P=\Pi \let\Si=\Sigma \let\F=\Phi \let\Ps=\Psi
344: \let\O=\Omega \let\Y=\Upsilon
345:
346: \def\\{\hfill\break} \let\==\equiv
347: \let\txt=\textstyle\let\dis=\displaystyle
348: \let\io=\infty \def\Dpr{\V\dpr\,}
349: \def\aps{{\it a posteriori}}
350: \let\0=\noindent \def\pagina{{\vfill\eject}}
351: \def\bra#1{{\langle#1|}}\def\ket#1{{|#1\rangle}}
352: \def\media#1{{\langle#1\rangle}}
353: \def\ie{\hbox{\it i.e.\ }}\def\eg{\hbox{\it e.g.\ }}
354: \let\dpr=\partial
355: \def\der{{\rm d}}
356: %\def\mm{{\rm m}}\def\mmmm{{\rm m}}
357: \def\mm{\hbox{\euftw m}}\def\mmmm{\hbox{\euftww m}}
358: \let\circa=\cong
359: \def\arccot{{\rm arccot}}
360: \def\sn{{\rm sn}}\def\cn{{\rm cn}}\def\dn{{\rm dn}}
361:
362: \def\tende#1{\,\vtop{\ialign{##\crcr\rightarrowfill\crcr
363: \noalign{\kern-1pt\nointerlineskip}
364: \hskip3.pt${\scriptstyle #1}$\hskip3.pt\crcr}}\,}
365: \def\otto{\,{\kern-1.truept\leftarrow\kern-5.truept\to\kern-1.truept}\,}
366: \def\fra#1#2{{#1\over#2}}
367:
368: \def\PPP{{\cal P}}\def\EE{{\cal E}}\def\MM{{\cal M}} \def\VV{{\cal V}}
369: \def\CC{{\cal C}}\def\FF{{\cal F}}\def\HH{{\cal H}}\def\WW{{\cal W}}
370: \def\TT{{\cal T}}\def\NN{{\cal N}}\def\BBB{{\cal B}}\def\III{{\cal I}}
371: \def\RR{{\cal R}}\def\LL{{\cal L}}\def\JJ{{\cal J}} \def\OO{{\cal O}}
372: \def\DD{{\cal D}}\def\AAA{{\cal A}}\def\GG{{\cal G}}\def\SS{{\cal S}}
373: \def\KK{{\cal K}}\def\UU{{\cal U}}
374:
375: \def\T#1{{#1_{\kern-3pt\lower7pt\hbox{$\widetilde{}$}}\kern3pt}}
376: \def\VVV#1{{\underline #1}_{\kern-3pt
377: \lower7pt\hbox{$\widetilde{}$}}\kern3pt\,}
378: \def\W#1{#1_{\kern-3pt\lower7.5pt\hbox{$\widetilde{}$}}\kern2pt\,}
379: \def\Re{{\rm Re}\,}\def\Im{{\rm Im}\,}
380: \def\lis{\overline}\def\tto{\Rightarrow}
381: \def\etc{{\it etc}} \def\acapo{\hfill\break}
382: \def\mod{{\rm mod}\,} \def\per{{\rm per}\,} \def\sign{{\rm sign}\,}
383: \def\indica{\leaders \hbox to 0.5cm{\hss.\hss}\hfill}
384: \def\guida{\leaders\hbox to 1em{\hss.\hss}\hfill}
385:
386: \def\hh{{\bf h}} \def\HHH{{\bf H}} \def\AA{{bf A}} \def\V0{{\bf 0}}
387: \def\BB{{\bf B}} \def\XX{{\bf X}} \def\PP{{\bf P}} \def\nn{{n}}
388: \def\vv{{\bf v}} \def\xx{{\bf x}} \def\yy{{\bf y}} \def\zz{{\bf z}}
389: \def\II{{\bf I}} \def\LLL{{\bf L}}
390: \def\ul{\underline}
391: \def\wt{\widetilde}
392: \def\oo{{\omega}}
393: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
394: %%%%%%%%%%%%%%%%% LETTERE GRECHE E LATINE IN NERETTO %%%%%%%%%%%%%%%%%
395: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
396:
397: % lettere greche e latine in neretto italico - pag.430 del manuale
398: \mathchardef\aa = "050B
399: \mathchardef\bb = "050C
400: \mathchardef\xxx = "0518
401: \mathchardef\hhh = "0511
402: \mathchardef\zzzzz= "0510
403: \mathchardef\oo = "0521
404: \mathchardef\tt = "051C
405: \mathchardef\lll = "0515
406: \mathchardef\mmm = "0516
407: \mathchardef\Dp = "0540
408: \mathchardef\H = "0548
409: \mathchardef\FFF = "0546
410: \mathchardef\ppp = "0570
411: \mathchardef\nnn = "0517
412: \mathchardef\pps = "0520
413: \mathchardef\XXX = "0504
414: \mathchardef\FFF = "0508
415: \mathchardef\nnnnn= "056E
416: \mathchardef\minore = "053C
417: \mathchardef\maggiore = "053E
418: \def\ol#1{{\overline #1}}
419:
420: \def\to{\rightarrow}
421: \def\la{\left\langle}
422: \def\ra{\right\rangle}
423:
424: \def\Overline#1{{\bar#1}}
425: \let\ciao=\bye
426: \def\qed{\hfill\raise1pt\hbox{\vrule height5pt width5pt depth0pt}}
427: \def\hf#1{{\hat \f_{#1}}}
428: \def\barf#1{{\tilde \f_{#1}}} \def\tg#1{{\tilde g_{#1}}}
429: \def\bq{{\bar q}} \def\Val{{\rm Val}}
430: \def\indic{\hbox{\raise-2pt \hbox{\indbf 1}}}
431:
432: \def\RRR{\hbox{\msytw R}} \def\rrrr{\hbox{\msytww R}}
433: \def\rrr{\hbox{\msytwww R}}
434: \def\CCC{\hbox{\msytw C}} \def\cccc{\hbox{\msytww C}}
435: \def\ccc{\hbox{\msytwww C}}
436: \def\NNN{\hbox{\msytw N}} \def\nnnn{\hbox{\msytww N}}
437: \def\nnn{\hbox{\msytwww N}}
438: \def\ZZZ{\hbox{\msytw Z}} \def\zzzz{\hbox{\msytww Z}}
439: \def\zzz{\hbox{\msytwww Z}}
440: \def\TTT{\hbox{\msytw T}} \def\tttt{\hbox{\msytww T}}
441: \def\ttt{\hbox{\msytwww T}}
442: \def\QQQ{\hbox{\msytw Q}} \def\qqqq{\hbox{\msytww Q}}
443: \def\qqq{\hbox{\msytwww Q}}
444: %\def\vvv{\hbox{\euftw v}} \def\vvvv{\hbox{\euftww v}}
445: %\def\www{\hbox{\euftw w}} \def\wwww{\hbox{\euftww w}}
446: \def\vvv{\hbox{\msytwwww V}} \def\vvvv{\hbox{\msytwwwww V}}
447: \def\www{\hbox{\msytwwww W}} \def\wwww{\hbox{\msytwwwww W}}
448: \def\GGG{\hbox{\msytw G}}
449:
450: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
451: %%%%%%%%%%%%%%%%% FORMATO PAGINA E LETTURA FILE .AUX %%%%%%%%%%%%%%%%%
452: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
453:
454: \newcount\mgnf %ingrandimento
455: \mgnf=0
456:
457: \ifnum\mgnf=0
458: \def\openone{\leavevmode\hbox{\ninerm 1\kern-3.3pt\tenrm1}}%
459: \def\*{\vglue0.3truecm}\fi
460: \ifnum\mgnf=1
461: \def\openone{\leavevmode\hbox{\ninerm 1\kern-3.63pt\tenrm1}}%
462: \def\*{\vglue0.5truecm}\fi
463:
464:
465: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
466: %%%%%%%%%%%%%%%%%%%%%% BIBBLIOGRAFIA %%%%%%%%%%%%%%%%%%%%%%%%%
467: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
468:
469: \newcount\tipobib\newcount\bz\bz=0\newcount\aux\aux=1
470: \newdimen\bibskip\newdimen\maxit\maxit=0pt
471:
472: %\openin14=\jobname.aux \ifeof14 \relax \else
473: %\input \jobname.aux \closein14 \fi
474: %\openout15=\jobname.aux
475:
476: \tipobib=0
477: \def\9#1{\ifnum\aux=1#1\else\relax\fi}
478:
479: \newwrite\bib
480: \immediate\openout\bib=\jobname.bib
481: \global\newcount\bibex
482: \bibex=0
483: \def\verabib{\number\bibex}
484:
485: \ifnum\tipobib=0
486: \def\cita#1{\expandafter\ifx\csname c#1\endcsname\relax
487: \hbox{$\clubsuit$}#1\write16{Manca #1 !}%
488: \else\csname c#1\endcsname\fi}
489: \def\rife#1#2#3{\immediate\write\bib{\string\raf{#2}{#3}{#1}}
490: \immediate\write15{\string\C(#1){[#2]}}
491: \setbox199=\hbox{#2}\ifnum\maxit < \wd199 \maxit=\wd199\fi}
492: \else
493: \def\cita#1{%
494: \expandafter\ifx\csname d#1\endcsname\relax%
495: \expandafter\ifx\csname c#1\endcsname\relax%
496: \hbox{$\clubsuit$}#1\write16{Manca #1 !}%
497: \else\probib(ref. numero )(#1)%
498: \csname c#1\endcsname%
499: \fi\else\csname d#1\endcsname\fi}%
500: \def\rife#1#2#3{\immediate\write15{\string\Cp(#1){%
501: \string\immediate\string\write\string\bib{\string\string\string\raf%
502: {\string\verabib}{#3}{#1}}%
503: \string\Cn(#1){[\string\verabib]}%
504: \string\CCc(#1)%
505: }}}%
506: \fi
507:
508: \def\Cn(#1)#2{\expandafter\xdef\csname d#1\endcsname{#2}}
509: \def\CCc(#1){\csname d#1\endcsname}
510: \def\probib(#1)(#2){\global\advance\bibex+1%
511: \9{\immediate\write16{#1\verabib => #2}}%
512: }
513:
514:
515: \def\C(#1)#2{\SIA c,#1,{#2}}
516: \def\Cp(#1)#2{\SIAnx c,#1,{#2}}
517:
518: \def\SIAnx #1,#2,#3 {\senondefinito{#1#2}%
519: \expandafter\def\csname#1#2\endcsname{#3}\else%
520: \write16{???? ma #1,#2 e' gia' stato definito !!!!}\fi}
521:
522:
523: \bibskip=10truept
524: \def\hboxto{\hbox to}
525: \def\hsizeby{\hsize by}
526: \catcode`\{=12\catcode`\}=12
527: \catcode`\<=1\catcode`\>=2
528: \immediate\write\bib<
529: \string\halign{\string\hboxto \string\maxit%
530: {\string #\string\hfill}&%
531: \string\vtop{\string\parindent=0pt\string\advance\string\hsize%
532: by -1.55truecm%
533: \string#\string\vskip \bibskip
534: }\string\cr%
535: >
536: \catcode`\{=1\catcode`\}=2
537: \catcode`\<=12\catcode`\>=12
538:
539:
540: \def\raf#1#2#3{\ifnum \bz=0 [#1] \cr\else
541: \llap{${}_{\rm #3}$}[#1]\cr\fi}
542:
543: \newread\bibin
544:
545: \catcode`\{=12\catcode`\}=12
546: \catcode`\<=1\catcode`\>=2
547: \def\chiudibib<
548: \catcode`\{=12\catcode`\}=12
549: \catcode`\<=1\catcode`\>=2
550: \immediate\write\bib<}>
551: \catcode`\{=1\catcode`\}=2
552: \catcode`\<=12\catcode`\>=12
553: >
554: \catcode`\{=1\catcode`\}=2
555: \catcode`\<=12\catcode`\>=12
556:
557: \def\makebiblio{
558: \ifnum\tipobib=0
559: \advance \maxit by 10pt
560: \else
561: \maxit=1.truecm
562: \fi
563: \chiudibib
564: \immediate \closeout\bib
565: \openin\bibin=\jobname.bib
566: \ifeof\bibin\relax\else
567: \raggedbottom
568: \input \jobname.bib
569: \fi
570: }
571:
572: \openin13=#1.aux \ifeof13 \relax \else
573: \input #1.aux \closein13\fi
574: \openin14=\jobname.aux \ifeof14 \relax \else
575: \input \jobname.aux \closein14 \fi
576: \immediate\openout15=\jobname.aux
577:
578:
579: \def\biblio{\*\*\centerline{\titolo References}\*\nobreak\makebiblio}
580:
581:
582: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
583: %%%%%%%%%%%%%%%%%%%%%%%% CORNICE %%%%%%%%%%%%%%%%%%%%%%%
584: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
585:
586: \ifnum\mgnf=0
587: \magnification=\magstep0
588: \hsize=15truecm\vsize=21.0truecm\voffset2.truecm\hoffset.5truecm
589: \parindent=0.3cm\baselineskip=0.45cm\fi
590: \ifnum\mgnf=1
591: \magnification=\magstep1\hoffset=0.truecm
592: \hsize=15truecm\vsize=21.0truecm
593: \baselineskip=18truept plus0.1pt minus0.1pt \parindent=0.9truecm
594: \lineskip=0.5truecm\lineskiplimit=0.1pt \parskip=0.1pt plus1pt\fi
595:
596:
597: %%% IMPOSTAZIONI PARAMETRI E LETTURA FILE .AUX
598:
599: \mgnf=0 %ingrandimento
600: \driver=1 %figure
601: \openin14=\jobname.aux \ifeof14 \relax \else
602: \input \jobname.aux \closein14 \fi
603: \openout15=\jobname.aux
604:
605: \let\figini=\initfig
606: \let\figfin=\endfig
607:
608:
609:
610:
611:
612: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
613: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
614: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
615: % %
616: % ARTICOLO %
617: % %
618: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
619: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
620: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
621:
622:
623: %\BOZZA
624: %\headline{{\hss {\rm DRAFT - Not for circulation}}}
625:
626:
627: %\null\vskip1truecm
628:
629: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
630: %%%%%%%%%%%%%%%%% TITOLO %%%%%%%%%%%%%%%%%%
631: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
632: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
633: \centerline{\titolone Periodic solutions for completely resonant}
634: \vskip.2truecm
635: \centerline{\titolone nonlinear wave equations}
636: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
637: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
638:
639: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
640: %%%%%%%%%%%%%%%%% AUTORI %%%%%%%%%%%%%%%%%%
641: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
642: \vskip1.truecm \centerline{{\titolo
643: Guido Gentile$^{\dagger}$,
644: Vieri Mastropietro$^{*}$,
645: Michela Procesi$^{\star}$}}
646: \vskip.2truecm
647: \centerline{{}$^{\dagger}$ Dipartimento di Matematica,
648: Universit\`a di Roma Tre, Roma, I-00146}
649: \vskip.1truecm
650: \centerline{{}$^{*}$ Dipartimento di Matematica,
651: Universit\`a di Roma ``Tor Vergata'', Roma, I-00133}
652: \vskip.1truecm
653: \centerline{{}$^{\star}$ SISSA, Trieste, I-34014}
654: \vskip1truecm
655:
656: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
657:
658: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
659: %%%%%%%%%%%%%%%%% RIASSUNTO %%%%%%%%%%%%%%%%%%
660: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
661: \line{\vtop{
662: \line{\hskip1.1truecm\vbox{\advance \hsize by -2.3 truecm
663: \0{\cs Abstract.}
664: {\it
665: We consider the nonlinear string equation with Dirichlet boundary
666: conditions $u_{xx}-u_{tt}=\f(u)$, with $\f(u)=\F u^{3} + O(u^{5})$
667: odd and analytic, $\F\neq0$, and we construct
668: small amplitude periodic solutions with frequency $\o$
669: for a large Lebesgue measure set of $\o$ close to $1$.
670: This extends previous results where only a zero-measure set of frequencies
671: could be treated (the ones for which no small divisors appear).
672: The proof is based on combining the Lyapunov-Schmidt decomposition,
673: which leads to two separate sets of equations dealing
674: with the resonant and nonresonant Fourier components,
675: respectively the Q and the P equations,
676: with resummation techniques of divergent powers series,
677: allowing us to control the small divisors problem.
678: The main difficulty with respect the nonlinear wave equations
679: $u_{xx}-u_{tt}+ M u = \f(u)$, $M\neq0$, is that not only
680: the P equation but also the Q equation is infinite-dimensional.
681: }} \hfill} }}
682: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
683: \vskip1.truecm
684:
685: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
686: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
687: \*\*
688: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
689: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
690: \section(1,Introduction)
691: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
692: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
693:
694: \0We consider the {\it nonlinear wave equation} in $d=1$ given by
695: %
696: $$ \cases{
697: u_{tt} - u_{xx} = \f(u), & \cr
698: u(0,t) = u(\p,t) = 0 , & \cr}
699: \Eq(1.1)$$
700: %
701: where Dirichlet boundary conditions allow us to use as a basis in
702: $L^{2}([0,\p])$ the set of functions $\{ \sin m x ,
703: m\in\NNN\}$, and $\f(u)$ is any odd analytic function $\f(u) = \F u^{3}
704: + O(u^{5})$ with $\F\neq0$.
705: We shall consider the problem of existence of periodic solutions for
706: \equ(1.1), which represents a completely resonant
707: case for the nonlinear wave equation as in the absence of
708: nonlinearities all the frequencies are resonant.
709:
710: In the finite dimensional case the problem has its analogous
711: in the study of periodic orbits close to elliptic equilibrium points:
712: results of existence have been obtained in such a case by Lyapunov \cita{L}
713: in the nonresonant case, by Birkhoff and Lewis \cita{BL}
714: in case of resonances of order greater than four,
715: and by Weinstein \cita{We} in case of any kind of resonances.
716: Systems with infinitely many degrees of freedom (as
717: the nonlinear wave equation, the nonlinear Schr\"odinger equation and
718: other PDE systems) have been studied much more recently;
719: the problem is much more difficult because of the presence of a
720: small divisors problem, which is absent in the finite dimensional case.
721: For the nonlinear wave equations $u_{tt}-u_{xx}+M u=\f(u)$,
722: with mass $M$ {\it strictly} positive, existence of periodic solutions
723: has been proved by Craig and Wayne \cita{CW1}, by P\"oschel \cita{P}
724: (by adapting the analogous result found by
725: Kuksin and P\"oschel \cita{KP} for the nonlinear Schr\"odinger equation)
726: and by Bourgain \cita{Bo2} (see also the review \cita{Cr}).
727: In order to solve the small divisors problem one has to
728: require that the amplitude and frequency of the solution must belong to a
729: Cantor set, and the main difficulty is to prove
730: that such a set can be chosen with {\it non-zero} Lebesgue measure.
731: We recall that for such systems also quasi-periodic solutions have been
732: proved to exist in \cita{KP}, \cita{P}, \cita{Bo3}
733: (in many other papers the case in which the coefficient $M$
734: of the linear term is replaced by a function depending
735: on parameters is considered; see for instance
736: \cita{Wa}, \cita{Bo1} and the reviews \cita{K1}, \cita{K2}).
737:
738: In all the quoted papers
739: only non-resonant cases are considered.
740: Some cases with some low-order resonances between the
741: frequencies have been studied by Craig and Wayne \cita{CW2}.
742: The completely resonant case \equ(1.1) has been
743: studied with variational methods starting from
744: Rabinowitz
745: \cita{R1}, \cita{R2}, \cita{BN},
746: \cita{BCN}, \cita{FR}, where periodic solutions with
747: period which is a rational multiple of $\p$ have been obtained;
748: such solutions correspond to a zero-measure set of
749: values of the amplitudes. The case of irrational periods,
750: which in principle could provide a large measure of values,
751: has been studied so far only under strong Diophantine conditions
752: (as the ones introduced in \cita{Ba}) which essentially remove
753: the small divisors problem leaving in fact again a zero-measure
754: set of values \cita{LS}, \cita{BP}, \cita{BB1}.
755: It is however conjectured that also for $M=0$
756: periodic solutions should exist for a large measure set
757: of values of the amplitudes, see for instance \cita{K2},
758: and indeed we prove in this paper that this is actually the case:
759: the unperturbed periodic solutions with periods $\o_{j}=2\p/j$ can be
760: continued into periodic solutions with period $\o_{\e,j}$ close to $\o_{j}$.
761: We rely on the Renormalization Group approach proposed in \cita{GM},
762: which consists in a Lyapunov-Schmidt decomposition followed
763: by a tree expansion of the solution which allows us
764: to control the small divisors problem.
765:
766: Note that a naive implementation of the Craig and Wayne approach to the
767: resonant case encounters some obvious difficulties.
768: Also in such an approach the first step consists in
769: a Lyapunov-Schmidt decomposition, which leads to two separate
770: sets of equations dealing with the resonant and nonresonant
771: Fourier components, respectively the Q and the P equations.
772: In the nonresonant case, by calling $v$ the solution of the Q equation
773: and $w$ the solution of the P equation, if one suppose to fix $v$
774: in the P equation then one gets a solution $w=w(v)$,
775: which inserted into the Q equation gives a closed finite-dimensional
776: equation for $v$. In the resonant case the Q equation is
777: infinite-dimensional. We can again find the solution $w(v)$
778: of the P equation by assuming that the Fourier coefficients of
779: $v$ are exponentially decreasing. However one ``consumes'' part of the
780: exponential decay to bound the small divisors, hence
781: the function $w(v)$ has Fourier coefficients decaying
782: with a smaller rate, and when inserted into
783: the Q equation the $v$ solution has not the properties
784: assumed at the beginning. Such a problem can be
785: avoided if there are no small divisors, as in such a case even a power
786: law decay is sufficient to find $w(v)$ and there
787: is no loss of smoothness; this is what is done in the
788: quoted papers \cita{LS}, \cita{BP} and \cita{BB1}:
789: the drawback is that only a zero-measure set is found.
790: The advantage of our approach is that we treat the P and Q
791: equation simultaneously, solving them together.
792: As in \cita{BP} and \cita{BB2} we also consider the problem of finding
793: how many solutions can be obtained with
794: given period, and we study their minimal period.
795:
796: \*
797:
798: If $\f=0$ every real solution of \equ(1.1) can be written as
799: %
800: $$ u(x,t) = \sum_{n=1}^{\io} U_{n} \sin nx \cos (\o_n t+\th_n) ,
801: \Eq(1.2) $$
802: %
803: where $\o_{n}=n$ and $U_{n}\in\RRR$ for all $n\in\NNN$.
804:
805: For $\e>0$ we set $\F=\s F$, with $\s={\rm sgn}\F$ and $F>0$,
806: and rescale $u \to \sqrt{\e/F} u$ in \equ(1.1), so obtaining
807: %
808: $$ \cases{
809: u_{tt} - u_{xx} = \s \e u^3 + O(\e^{2}) , & \cr
810: u(0,t) = u(\p,t) = 0 , & \cr}
811: \Eq(1.3)$$
812: %
813: where $O(\e^{2})$ denotes an analytic function of $u$ and $\e$ of
814: order at least $2$ in $\e$, and we define $\o_{\e}=\sqrt{1-\l\e}$,
815: with $\l\in\RRR$, so that $\o_{\e}=1$ for $\e=0$.
816:
817: As the nonlinearity $\f$ is odd the solution of \equ(1.3) can be
818: extended in the $x$ variable to an odd $2\pi$ periodic function
819: (even in the variable $t$).
820: We shall consider $\e$ small and we shall show
821: that there exists a solution of \equ(1.3), which is
822: $2\pi/\o_{\e}$-periodic in $t$ and $\e$-close to the function
823: %
824: $$ u_{0}(x,\o_{\e}t) = a_0(\o_{\e} t + x) - a_0(\o_{\e} t - x) ,
825: \Eq(1.4) $$
826: %
827: provided that $\e$ is in an appropriate Cantor set and $a_0(\x)$
828: is the odd $2\pi$-periodic solution of the integro-differential equation
829: %
830: $$ \l \ddot a_{0} = - 3 \la a_{0}^{2} \ra a_{0} - a_{0}^{3} ,
831: \Eq(1.5) $$
832: %
833: where the dot denotes derivative with respect to $\x$, and,
834: given any periodic function $F(\x)$ with period $T$, we denote by
835: %
836: $$ \la F \ra = {1\over T} \int_{0}^{T} {\rm d}\x \, F(\x)
837: \Eq(1.6) $$
838: %
839: its average. Then a $2\p/\o_{\e}$-periodic solution of \equ(1.1)
840: is simply obtained by scaling back the solution of \equ(1.3).
841:
842: The equation \equ(1.5) has odd $2\pi$-periodic solutions,
843: provided that one sets $\l>0$; we shall choose $\l=1$ in the following.
844: An explicit computation gives \cita{BP}
845: %
846: $$ a_{0}(\x) = V_{\mmmm} \, \sn (\O_{\mmmm}\x,\mm)
847: \Eq(1.7) $$
848: %
849: for $\mm$ a suitable negative constant ($\mm \approx -0.2554$),
850: with $\O_{\mmmm}=2K(\mm)/\p$ and $V_{\mmmm}=\sqrt{-2\mm}\O_{\mmmm}$, where
851: $\sn(\O_{\mmmm}\x,\mm)$ is the sine-amplitude function and $K(\mm)$ is the
852: elliptic integral of the first kind, with modulus $\sqrt{\mm}$ \cita{GrR};
853: see Appendix A1 for further details.
854: Call $2\k$ the width of the analyticity strip of
855: the function $a_{0}(\x)$ and $\a$ the maximum value
856: it can assume in such a strip; then one has
857: %
858: $$ \left| a_{0,n} \right| \le \a e^{-2k|n|} .
859: \Eq(1.15) $$
860: %
861: Our result (including also the cases
862: of frequencies which are multiples of $\o_{\e}$)
863: can be more precisely stated as follows.
864:
865: \*
866:
867: \0{\bf Theorem.} {\it Consider the equation \equ(1.1),
868: where $\f(u)=\F u^{3} + O(u^{5})$ is an odd analytic function,
869: with $F=|\F| \neq 0$. Define $u_{0}(x,t)=a_{0}(t+x)-a_{0}(t-x)$,
870: with $a_{0}(\x)$ the odd $2\pi$-periodic solution of \equ(1.5).
871: There are a positive constant $\e_{0}$ and for all $j\in\NNN$
872: a set $\EE \in[0,\e_{0}/j^{2}]$ satisfying
873: %
874: $$ \lim_{\e\to0} { {\rm meas}(\EE \cap [0,\e] ) \over \e} = 1 ,
875: \Eq(1.8) $$
876: %
877: such that for all $\e\in\EE$, by setting $\o_{\e}=\sqrt{1-\e}$ and
878: %
879: $$ \|f(x,t)\|_r= \sum_{(n,m)\in \zzz^2}f_{n,m}e^{r(|n|+|m|)},
880: \Eq(1.9) $$
881: %
882: for analytic $2\pi$-periodic functions,
883: there exist $2\p/j\o_{\e}$-periodic solutions
884: $u_{\e,j}(x,t)$ of \equ(1.1), analytic in $(t,x)$, with
885: %
886: $$ \left\| u_{\e,j}(x,t) - j \sqrt{\e/F} u_{0}(j x,j\o_{\e}t)
887: \right\|_{\k'} \le C \, j \, \e\sqrt{\e} ,
888: \Eq(1.10) $$
889: %
890: for some constants $C>0$ and $0<\k'<\k$.}
891:
892: \*
893:
894: Note that such a result provides a solution of the
895: open problem 7.4 in \cita{K2},
896: as far as periodic solutions are concerned.
897:
898: As we shall see for $\f(u)=F u^{3}$ for
899: all $j\in\NNN$ one can take the set $\EE=[0,\e_{0}]$,
900: independently of $j$, so that for fixed $\e\in\EE$
901: no restriction on $j$ have to be imposed.
902:
903: \*
904:
905: We look for a solution of \equ(1.3) of the form
906: %
907: $$ \eqalign{
908: u(x,t) & = \sum_{(n,m) \in \zzz^{2}}
909: e^{in j \o t + i j m x} u_{n,m}
910: = v(x,t) + w(x,t) , \cr
911: v(x,t) & = a(\x) - a(\x') , \qquad
912: \x= \o t + x, \quad \x'= \o t - x , \cr a(\x) & = \sum_{n\in \zzz}
913: e^{i n \x} a_{n} , \cr
914: w(x,t) & = \sum_{\scriptstyle (n,m) \in \zzz^{2} \atop
915: \scriptstyle |n| \neq |m|}
916: e^{in j \o t +i j m x} w_{n,m} , \cr}
917: \Eq(1.11) $$
918: %
919: with $\o=\o_{\e}$, such that one has $w(x,t)=0$
920: and $a(\x)=a_{0}(\x)$ for $\e=0$.
921: Of course by the symmetry of \equ(1.1), hence of \equ(1.4),
922: we can look for solutions (if any) which verify
923: %
924: $$ u_{n,m} = - u_{n,-m} = u_{-n,m}
925: \Eq(1.12) $$
926: %
927: for all $n,m\in\ZZZ$.
928:
929: Inserting \equ(1.11) into \equ(1.3) gives two sets of equations, called
930: the Q and P equations \cita{CW1}, which are given, respectively, by
931: %
932: $$ \eqalign{
933: \hbox{Q} \qquad & \cases{
934: n^{2} a_{n} = \left[ \f(v+w) \right]_{n,n} , & \cr
935: - n^{2} a_{n} = \left[ \f(v+w) \right]_{n,-n} , & \cr} \cr
936: \hbox{P} \qquad & \left( - \o^2 n^2 + m^2 \right) w_{n,m} = \e \left[
937: \f(v+w) \right]_{n,m} , \qquad |m| \neq |n| , \cr}
938: \Eq(1.13) $$
939: %
940: where we denote by $[F]_{n,m}$ the Fourier component of the function
941: $F(x,t)$ with labels $(n,m)$, so that
942: %
943: $$ F(x,t) = \sum_{(n,m)\in\zzz^{2}} e^{i n\o t + m x} [F]_{n,m} .
944: \Eq(1.14) $$
945: %
946: In the same way we shall call $[F]_{n}$ the Fourier component of
947: the function $F(\x)$ with label $n$;
948: in particular one has $[F]_{0}=\la F\ra$. Note also that the two
949: equations Q are in fact the same, by the symmetry property
950: $\left[ \f(v+w) \right]_{n,m}=-\left[ \f(v+w) \right]_{n,-m}$,
951: which follows from \equ(1.12).
952:
953: We start by considering the case $\f(u)=u^{3}$ and $j=1$,
954: for simplicity. We shall discuss at the end how the other cases
955: can be dealt with; see Section \secc(8).
956:
957:
958: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
959: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
960: \*\*
961: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
962: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
963: \section(2,Lindstedt series expansion)
964: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
965: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
966:
967: \0One could try to write a power series
968: expansion in $\e$ for $u(x,t)$, using \equ(1.13)
969: to get recursive equations for the coefficients.
970: However by proceeding in this way one finds that
971: the coefficient of order $k$ is given by a sum of terms
972: some of which of order $O(k!^\a)$, for some constant $\a$.
973: This is the same phenomenon occurring
974: in the Lindstedt series for invariant KAM tori
975: in the case of quasi-integrable Hamiltonian systems;
976: in such a case however one can show that
977: there are {\it cancellations} between the terms
978: contributing to the coefficient of order $k$,
979: which at the end admits a bound $C^k$, for a suitable constant $C$.
980: On the contrary such cancellations are absent in the present case
981: and we have to proceed in a different way, equivalent to a resummation
982: (see \cita{GM} where such procedure was applied to the same
983: nonlinear wave equation with a mass term, $u_{xx}-u_{tt}+M u=\f(u)$).
984:
985: \*
986:
987: \0{\bf Definition 1. }{\it Given a sequence
988: $\{\n_{m}(\e)\}_{|m|\ge 1}$, such that $\n_{m}=\n_{-m} $,
989: we define the {\rm renormalized frequencies} as
990: %
991: $$ \tilde\o_m^2 \= \o_m^2 + \n_m , \qquad \o_{m}=|m| ,
992: \Eq(2.1) $$
993: %
994: and the quantities $\n_{m}$ will be called the {\rm counterterms}.}
995:
996: \*
997:
998: By the above definition and the parity properties \equ(1.12)
999: the P equation in \equ(1.13) can be rewritten as
1000: %
1001: $$ \eqalign{
1002: w_{n,m} \left( - \o^2 n^{2} + \tilde\o_m^2 \right)
1003: & = \n_{m} w_{n,m} + \e [\f(v+w)]_{n,m} \cr
1004: & = \n_{m}^{(a)} w_{n,m} + \n_{m}^{(b)} w_{n,-m} + \e [\f(v+w)]_{n,m}, \cr}
1005: \Eq(2.2) $$
1006: %
1007: where
1008: %
1009: $$\n_{m}^{(a)}-\n_{m}^{(b)}=\n_{m} .
1010: \Eq(2.3) $$
1011: %
1012:
1013: With the notations of \equ(1.14), and recalling that we are
1014: considering $\f(u)=u^{3}$, we can write
1015: %
1016: $$ \eqalign{
1017: \left[ (v+w)^3 \right]_{n,n} & = [v^{3}]_{n,n}+[w^{3}]_{n,n}+
1018: 3 [v^{2} w]_{n,n} + 3 [w^{2} v]_{n,n} \cr
1019: & \equiv [v^{3}]_{n,n} + [g(v,w)]_{n,n} , \cr}
1020: \Eq(2.5) $$
1021: %
1022: where, again by using the parity properties \equ(1.12),
1023: %
1024: $$ [v^{3}]_{n,n} = [a^{3}]_{n} + 3 \la a^{2} \ra a_{n} .
1025: \Eq(2.6) $$
1026: %
1027: Then the first Q equation in \equ(1.12) can be rewritten as
1028: %
1029: $$ n^{2} a_{n} = [a^{3}]_{n} + 3 \la a^{2} \ra a_{n} +
1030: \left[ g(v,w) \right]_{n,n} ,
1031: \Eq(2.7) $$
1032: %
1033: so that $a_{n}$ is the Fourier coefficient of the
1034: $2\p$-periodic solution of the equation
1035: %
1036: $$ \ddot a = - \left( a^{3} + 3 \la a^{2} \ra a + G(v,w) \right) ,
1037: \Eq(2.8) $$
1038: %
1039: where we have introduced the function
1040: %
1041: $$ G(v,w) = \sum_{n\in\zzz}
1042: e^{in\x} \left[ g(v,w) \right]_{n,n} .
1043: \Eq(2.9) $$
1044: %
1045: To study the equations \equ(2.2) and \equ(2.7)
1046: we introduce an auxiliary parameter $\m$, which
1047: at the end will be set equal to 1, by writing \equ(2.2) as
1048: %
1049: $$ w_{n,m} \left( - \o^2 n^{2} + \tilde\o_m^2 \right)
1050: = \m \n_{m}^{(a)} w_{n,m} + \m \n_{m}^{(b)} w_{n,-m} +
1051: \m \e [\f(v+w)]_{n,m} ,
1052: \Eq(2.2a) $$
1053: %
1054: and we shall look for $u_{n,m}$ in the form of a
1055: power series expansion in $\m$,
1056: %
1057: $$ u_{n,m} = \sum_{k=0}^{\io} \m^{k} u_{n,m}^{(k)} ,
1058: \Eq(2.4) $$
1059: %
1060: with $u_{n,m}^{(k)}$ depending on $\e$ and on
1061: the parameters $\n_{m}^{(c)}$, with $c=a,b$ and $|m|\ge 1$.
1062:
1063: In \equ(2.4) $k=0$ requires $u_{n,\pm n}=\pm a_{0,n}$
1064: and $u_{n,m}=0$ for $|n|\neq|m|$, while for $k\ge 1$,
1065: as we shall see later on, the dependence on the
1066: parameters $\n_{m'}^{(c')}$ will be of the form
1067: %
1068: $$ \prod_{m'=2}^{\io} \prod_{c'=a,b}
1069: \left( \n_{m'}^{(c')} \right)^{k_{m'}^{(c')}} ,
1070: \Eq(2.5a) $$
1071: %
1072: with $|\ul{k}|=k_{1}^{(a)}+k_{1}^{(b)}+k_{2}^{(a)}+k_{2}^{(b)}+\ldots
1073: \le k-1$. Of course we are using the symmetry property to restrict
1074: the dependence only on the positive labels $m'$.
1075:
1076: We derive recursive equations for the coefficients $u_{n,m}^{(k)}$
1077: of the expansion. We start from the coefficients with $|n|=|m|$.
1078:
1079: By \equ(1.11) and \equ(2.4) we can write
1080: %
1081: $$ a = a_{0} + \sum_{k=1}^{\io} \m^{k} A^{(k)} ,
1082: \Eq(2.10) $$
1083: %
1084: and inserting this expression into \equ(2.8)
1085: we obtain for $A^{(k)}$ the equation
1086: %
1087: $$ \ddot A^{(k)} =
1088: - 3\left(a_0^2 A^{(k)} +
1089: \la a_0^{2} \ra A^{(k)}
1090: + 2 \la a_0 A^{(k)} \ra a_0\right) + f^{(k)} ,
1091: \Eq(2.11) $$
1092: %
1093: with
1094: %
1095: $$ f^{(k)} = - \sum_{\scriptstyle k_{1}+k_{2}+k_{3}=k \atop
1096: \scriptstyle k_{i}=k \to |n_{i}| \neq |m_{i}|}
1097: \sum_{\scriptstyle n_{1}+n_{2}+n_{3} = n \atop
1098: \scriptstyle m_{1}+m_{2}+m_{3} = m}
1099: u^{(k_{1})} _{n_{1},m_{1}}
1100: u^{(k_{2})} _{n_{2},m_{2}}
1101: u^{(k_{3})} _{n_{3},m_{3}} ,
1102: \Eq(2.12)$$
1103: %
1104: where we have used the notations
1105: %
1106: $$ u^{(k)}_{n,m} = \cases{
1107: v^{(k)}_{n,m} , & if $|n|=|m|$ , \cr
1108: w^{(k)}_{n,m} , & if $|n|\neq |m|$ , \cr}
1109: \Eq(2.13) $$
1110: %
1111: with
1112: %
1113: $$ v^{(k)}_{n,n}= \cases{
1114: A^{(k)}_{n} , & if $k\not=0$, \cr
1115: a_{0,n} , & if $k= 0$, \cr} \quad
1116: v^{(k)}_{n,-n}= \cases{
1117: - A^{(k)}_{n} , & if $k\not=0$, \cr
1118: - a_{0,n} , & if $k= 0$ . \cr}
1119: \Eq(2.14) $$
1120: %
1121:
1122: Before studying how to find the solution of this equation
1123: we introduce some preliminary definitions.
1124: To shorten notations we write
1125: %
1126: $$ c(\x)\=\cn(\O_{\mmmm}\x,\mm),\quad
1127: s(\x)\=\sn(\O_{\mmmm}\x,\mm)\quad d(\x)\=\dn(\O_{\mmmm}\x,\mm) ,
1128: \Eq(2.15) $$
1129: %
1130: and set $cd(\x)=\cn(\O_{\mmmm}\x,\mm)\,\dn(\O_{\mmmm}\x,\mm)$.
1131: Moreover given an analytic periodic function $F(\x)$ we define
1132: %
1133: $$\PP[F](\x) = F(\x)-\la F \ra .
1134: \Eq(2.16) $$
1135: %
1136: and we introduce a linear operator $\II$ acting
1137: on $2\p$-periodic zero-mean functions
1138: and defined by its action on the basis
1139: $e_{n}(\x)=e^{i n\x}$, $n\in\ZZZ\setminus\{0\}$,
1140: %
1141: $$\II[e_n](\x)= {e_n(\x)\over i n} .
1142: \Eq(2.17) $$
1143: %
1144: Note that if $\PP[F]=0$ then $\PP[\II [F]]=0$
1145: ($\II[F]$ is simply the zero-mean primitive of $F$);
1146: moreover $\II$ switches parities.
1147:
1148: In order to find an odd solution of \equ(2.11) we replace first
1149: $\la a_0 A^{(k)}\ra$ with a parameter $C^{(k)}$,
1150: and we study the modified equation
1151: %
1152: $$ \ddot A^{(k)} =
1153: - 3\left(a_0^2 A^{(k)} +
1154: \la a_0^{2} \ra A^{(k)}
1155: + 2 C^{(k)} a_0 \right) + f^{(k)} .
1156: \Eq(2.18) $$
1157: %
1158: Then we have the following result (proved in Appendix \secc(A1)).
1159:
1160: \*
1161:
1162: \0{\bf Lemma 1.} {\it Given an odd analytic $2\pi$-periodic
1163: function $h(\x)$, the equation
1164: %
1165: $$ \ddot y = - 3 \left( a_0^2 +\la a_0^{2} \ra \right) y + h ,
1166: \Eq(2.19)$$
1167: %
1168: admits one and only one odd analytic $2\pi$-periodic
1169: solution $y(\x)$, given by
1170: %
1171: $$ y = \LLL[h] \=
1172: B_{\mmmm} \Big( \O_{\mmmm}^{-2}D_{\mmmm}^{2} s \la s\, h\ra +
1173: \O_{\mmmm}^{-1}D_{\mmmm} \left( s \, \II[cd\, h]-
1174: cd \,\II [\PP[ s\, h]] \right) + cd \, \II [\II [cd \, h]] \Big) ,
1175: \Eq(2.20) $$
1176: %
1177: with $B_{\mmmm}=-\mm/(1-\mm)$ and $D_{\mmmm}=-1/\mm$.}
1178:
1179: \*
1180:
1181: As $a_{0}$ is analytic and odd, we find immediately,
1182: by induction on $k$ and using lemma 1,
1183: that $f^{(k)}$ is analytic and odd, and that
1184: the solution of the equation \equ(2.18) is odd and given by
1185: %
1186: $$ \tilde A^{(k)} =
1187: \LLL[ - 6 C^{(k)} a_0 + f^{(k)} ] .
1188: \Eq(2.21) $$
1189: %
1190:
1191: The function $\tilde A^{(k)}$ so found depends of
1192: course on the parameter $C^{(k)}$; in order
1193: to obtain $\tilde A^{(k)}= A^{(k)}$,
1194: we have to impose the constraint
1195: %
1196: $$ C^{(k)} = \la a_0 A^{(k)} \ra ,
1197: \Eq(2.22) $$
1198: %
1199: and by \equ(2.21) this gives
1200: %
1201: $$ C^{(k)} = - 6 C^{(k)}
1202: \la a_0 L[a_0]\ra + \la a_0 \LLL[f^{(k)} ] \ra ,
1203: \Eq(2.23) $$
1204: %
1205: which can be rewritten as
1206: %
1207: $$ \left( 1 + 6 \la a_0 \LLL[a_0]\ra \right) C^{(k)} =
1208: \la a_0 \LLL[f^{(k)} ]\ra .
1209: \Eq(2.24) $$
1210: %
1211: An explicit computation (see Appendix \secc(A3)) gives
1212: %
1213: $$\la a_0 \LLL[a_0]\ra=
1214: {1 \over 2} V_{\mmmm}^{2} \O_{\mmmm}^{-2} B_{\mmmm}
1215: \left( \left( 2D_{\mmmm} - {1\over2} \right)
1216: \langle s^{4} \rangle + \left( 2D_{\mmmm} \left( D_{\mmmm}-1 \right)
1217: + {1\over 2} \right) \langle s^{2} \rangle^{2} \right) ,
1218: \Eq(2.25) $$
1219: %
1220: which yields $r_0 = (1+6 \langle a_0 \LLL[a_0]\rangle ) \neq 0$.
1221: At the end we obtain the recursive definition
1222: %
1223: $$ \cases{
1224: A^{(k)} = \LLL[ f^{(k)} - 6 C^{(k)} a_0 ] , & \cr
1225: & \cr
1226: C^{(k)} = r_0^{-1} \la a_0 \LLL[ f^{(k)} ] \ra . & \cr}
1227: \Eq(2.26) $$
1228: %
1229: In Fourier space the first of \equ(2.26) becomes
1230: %
1231: $$ \eqalign{
1232: A^{(k)}_{n} & = B_{\mmmm} \O_{\mmmm}^{-2} D_{\mmmm}^{2}
1233: s_{n} \mathop{\sum}_{n_1+n_2=0}
1234: s_{n_{1}} \left( f^{(k)}_{n_{2}} -
1235: 6 C^{(k)} a_{0,n_{2}} \right) \cr
1236: & \qquad + B_{\mmmm} {\mathop{\sum}_{n_{1}+n_{2}+n_{3}=n}}^{*}
1237: {1 \over i^2 (n_{2}+n_{3})^2}
1238: cd_{n_{1}} cd_{n_{2}} \left( f^{(k)}_{n_{3}} -
1239: 6 C^{(k)} a_{0,n_{3}} \right) \cr
1240: & \qquad + B_{\mmmm}\O_{\mmmm}^{-1}D_{\mmmm}
1241: {\mathop{\sum}_{n_{1}+n_{2}+n_{3}=n}}^{*}
1242: {1 \over i (n_{2}+n_{3})} s_{n_{1}} cd_{n_{2}}
1243: \left( f^{(k)}_{n_{3}} -
1244: 6 C^{(k)} a_{0,n_{3}} \right) \cr
1245: & \qquad - B_{\mmmm} \O_{\mmmm}^{-1}D_{\mmmm}
1246: {\mathop{\sum}_{n_{1}+n_{2}+n_{3}=n}}^{*}
1247: {1 \over i (n_{2}+n_{3})}
1248: cd_{n_{1}} s_{n_{2}} \left( f^{(k)}_{n_{3}} -
1249: 6 C^{(k)} a_{0,n_{3}} \right) \cr
1250: & \= \sum_{n'} \LLL_{n n'} \left( f^{(k)}_{n'} -
1251: 6 C^{(k)} a_{0,n'} \right) , \cr}
1252: \Eq(2.27) $$
1253: %
1254: where the constants $B_{\mmmm}$ and $D_{\mmmm}$ are defined after \equ(2.20),
1255: and the $*$ in the sums means that one has the constraint
1256: $n_{2}+n_{3}\neq 0$, while the second of \equ(2.26) can be written as
1257: %
1258: $$ C^{(k)} = r_0^{-1} \sum_{n,n'\in\zzz}
1259: a_{0,-n} \LLL_{n,n'} f^{(k)}_{n'} .
1260: \Eq(2.28) $$
1261: %
1262:
1263: Now we consider the coefficients $u^{(k)}_{n,m}$
1264: with $|n|\neq |m|$. The coefficients $w^{(k)}_{n,m}$
1265: verify the recursive equations
1266: %
1267: $$ w^{(k)}_{n,m} \left[ - \o^2 n^{2} + \tilde\o_m^2 \right] =
1268: \n_{m}^{(a)} w_{n,m}^{(k-1)} + \n_{m}^{(b)} w_{n,-m}^{(k-1)} +
1269: [(v+w)^3]^{(k-1)}_{n,m} ,
1270: \Eq(2.29) $$
1271: %
1272: where
1273: %
1274: $$ [(v+w)^3]^{(k)}_{n,m} = \sum_{k_{1}+k_{2}+k_{3}=k}
1275: \sum_{\scriptstyle n_{1}+n_{2}+n_{3} = n \atop
1276: \scriptstyle m_{1}+m_{2}+m_{3} = m} u^{(k_{1})}_{n_{1},m_{1}}
1277: u^{(k_{2})}_{n_{2},m_{2}} u^{(k_{3})}_{n_{3},m_{3}} ,
1278: \Eq(2.30) $$
1279: %
1280: if we use the same notations \equ(2.13) and \equ(2.14) as in \equ(2.12).
1281:
1282: The equations \equ(2.27) and \equ(2.29), together with
1283: \equ(2.30), \equ(2.12), \equ(2.28) and \equ(2.30),
1284: define recursively the coefficients $u^{(k)}_{n,m}$.
1285:
1286: To prove the theorem we shall proceed in two steps.
1287: The first step consists in looking for the solution
1288: of the equations \equ(2.27) and \equ(2.29)
1289: by considering $\tilde\o=\{\tilde\o_{m}\}_{|m|\ge 1}$
1290: as a given set of parameters satisfying the Diophantine conditions
1291: (called respectively the first and the second Mel$^{\prime}$nikov conditions)
1292: %
1293: $$ \eqalignno{
1294: & \left| \o n \pm \tilde \o_{m} \right|
1295: \ge C_{0} |n|^{-\t}
1296: \qquad \forall n\in\ZZZ \setminus
1297: \{ 0 \} \hbox{ and }
1298: \forall m \in \ZZZ \setminus\{0\}
1299: \hbox{ such that } |m| \neq |n| , \cr
1300: & \left| \o n \pm \left( \tilde\o_{m} \pm \tilde\o_{m'}
1301: \right) \right| \ge C_{0} |n|^{-\t}
1302: \qquad
1303: & \eq(2.31) \cr
1304: & \qquad \qquad \qquad \qquad
1305: \forall n\in\ZZZ \setminus \{ 0 \} \hbox{ and }
1306: \forall m,m' \in \ZZZ \setminus\{0\}
1307: \hbox{ such that } |n|\neq |m \pm m'| , \cr} $$
1308: %
1309: with positive constants $C_{0},\t$.
1310: We shall prove in Sections \secc(3) to \secc(5) the following result.
1311:
1312: \*
1313:
1314: \0{\bf Proposition 1.} {\it Let be given a sequence
1315: $\tilde\o=\{\tilde\o_m\}_{|m| \ge 1}$ verifying \equ(2.31),
1316: with $\o=\o_{\e}=\sqrt{1-\e}$
1317: and such that $|\tilde\o_{m}-|m||\le C\e/|m|$ for some constant $C$.
1318: For all $\m_{0}>0$ there exists $\e_{0}>0$ such that
1319: for $|\m|\le \m_{0}$ and $0<\e<\e_{0}$ there is
1320: a sequence $\n(\tilde\o,\e;\m) = \{\n_{m}(\tilde\o,\e;\m)\}_{|m| \ge 1}$,
1321: where each $\n_{m}(\tilde\o,\e;\m)$ is analytic in $\m$,
1322: such that the coefficients $u^{(k)}_{n,m}$
1323: which solve \equ(2.27) and \equ(2.29) define via \equ(2.4)
1324: a function $u(x,t;\tilde\o,\e;\m)$ which is analytic in $\m$,
1325: analytic in $(x,t)$ and $2\pi$-periodic in $t$ and solves
1326: %
1327: $$ \eqalign{
1328: \hbox{} \qquad &
1329: \cases{
1330: n^{2} a_{n} = \left[ a^{3} \right]_{n,n} +
1331: 3 \la a^{2} \ra a_{n} + \left[ g(v,w) \right]_{n,n} , & \cr
1332: - n^{2} a_{n} = \left[ a^{3} \right]_{n,-n} +
1333: 3 \la a^{2} \ra a_{-n} + \left[ g(v,w) \right]_{n,-n} , & \cr} \cr
1334: \hbox{} \qquad & \left( - \o^2 n^2 + \tilde\o_{m}^2 \right)
1335: w_{n,m} = \m \n_{m}(\tilde\o,\e;\m) \, w_{n,m} +
1336: \m \e \left[ \f(v+w) \right]_{n,m} , \qquad |m| \neq |n| , \cr}
1337: \Eq(2.32) $$
1338: %
1339: with the same notations as in \equ(1.13).}
1340:
1341: \*
1342:
1343: If $\t \le 2$ then one can require only the first Mel'nikov conditions
1344: in \equ(2.31), as we shall show in Section \secc(7).
1345:
1346: Then in Proposition 1 one can fix $\m_{0}=1$, so that
1347: one can choose $\m=1$ and set
1348: $u(x,t;\tilde\o,\e)=u(x,t;\tilde\o,\e;1)$ and
1349: $\n_{m}(\tilde\o,\e)=\n_{m}(\tilde\o,\e;1)$.
1350:
1351: The second step, to be proved in Section \secc(6),
1352: consists in inverting \equ(2.1), with $\n_{m}=\n_{m}(\tilde\o,\e)$
1353: and $\tilde\o$ verifying \equ(2.31). This requires
1354: some preliminary conditions on $\e$, given by
1355: the Diophantine conditions
1356: %
1357: $$ \left| \o n \pm m \right|
1358: \ge C_{1} |n|^{-\t_{0}}
1359: \qquad \forall n\in\ZZZ \setminus
1360: \{ 0 \} \hbox{ and }
1361: \forall m \in \ZZZ \setminus\{0\}
1362: \hbox{ such that } |m| \neq |n| ,
1363: \Eq(2.33)$$
1364: %
1365: with positive constants $C_{1}$ and $\t_{0}>1$.
1366:
1367: This allows to solve iteratively \equ(2.1),
1368: by imposing further non-resonance conditions besides \equ(2.33),
1369: provided that one takes $C_{1}=2C_{0}$ and $\t_{0}<\t-1$,
1370: which requires $\t>2$. At each iterative step one has to exclude
1371: some further values of $\e$, and at the end the left values
1372: fill a Cantor set $\EE$ with large relative measure in $[0,\e_{0}]$
1373: and $\tilde\o$ verify \equ(2.33).
1374:
1375: If $1<\t \le 2$ the first Mel'nikov conditions, which, as we said
1376: above, become sufficient to prove Proposition 1,
1377: can be obtained by requiring \equ(2.33) with $\t_{0}=\t$;
1378: again this leaves a large measure set of allowed values of $\e$.
1379: This is discussed in Section \secc(7).
1380:
1381: The result of this second step can be summarized as follows.
1382:
1383: \*
1384:
1385: \0{\bf Proposition 2.} {\it
1386: There are $\d>0$ and a set $\EE\subset [0,\e_{0}]$
1387: with complement of relative Lebesgue measure of order $\e_{0}^{\d}$
1388: such that for all $\e\in\EE$ there exists
1389: $\tilde\o=\tilde\o(\e)$ which solves \equ(2.1)
1390: and satisfy the Diophantine conditions \equ(2.31) with
1391: $|\tilde\o_{m}-|m||\le C\e/|m|$ for some constant $C$.}
1392:
1393: \*
1394:
1395: As we said, our approach is based on constructing the periodic solution
1396: of the string equation by a perturbative expansion
1397: which is the analogue of the {\it Lindstedt series}
1398: for (maximal) KAM invariant tori
1399: in finite-dimensional Hamiltonian systems.
1400: Such an approach immediately encounters a difficulty;
1401: while the invariant KAM tori are analytic in the perturbative
1402: parameter $\e$, the periodic solutions we are looking for are {\it not}
1403: analytic; hence a power series construction seems at first sight
1404: hopeless. Nevertheless it turns out that the Fourier
1405: coefficients of the periodic solution have the form
1406: $u_{n,m}(\tilde\o(\o,\e),\e;\m)$; while such functions are
1407: not analytic in $\e$, it turns out to be analytic
1408: in $\m$, provided that $\tilde\o$ satisfies the
1409: condition \equ(2.31) and $\e$ is small enough;
1410: this is the content of Proposition 1.
1411: As far as we know, in the case $M\neq0$, besides \cita{GM},
1412: the smoothness in $\e$ at fixed $\tilde\o$ was only discussed
1413: in a paper by Bourgain \cita{Bo2}. This smoothness is what
1414: allow us to write as a series expansion $u_{n,m}(\tilde\o,\e;\m)$;
1415: this strategy was already applied in \cita{GM} in the massive case.
1416:
1417:
1418: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1419: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1420: \*\*
1421: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1422: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1423: \section(3,Tree expansion: the diagrammatic rules)
1424: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1425: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1426:
1427: \0A (connected) graph $\GGG$ is a collection of points (vertices)
1428: and lines connecting all of them. The points of a graph
1429: are most commonly known as graph vertices, but may also be
1430: called nodes or points. Similarly, the lines connecting
1431: the vertices of a graph are most commonly
1432: known as graph edges, but may also be called branches or
1433: simply lines, as we shall do. We denote with
1434: $P(\GGG)$ and $L(\GGG)$ the set of vertices and the set of lines,
1435: respectively. A path between two vertices is a subset of $L(\GGG)$
1436: connecting the two vertices. A graph is planar if it can be drawn
1437: in a plane without graph lines crossing
1438: (i.e. it has graph crossing number 0).
1439:
1440: \*
1441:
1442: \0{\bf Definition 2.}
1443: {\it A {\rm tree} is a planar graph $\GGG$ containing no closed loops
1444: (cycles); in other words, it is a connected acyclic graph.
1445: One can consider a tree $\GGG$ with a single special vertex $\vvv_{0}$:
1446: this introduces a natural partial ordering on the set
1447: of lines and vertices, and one can imagine that each line
1448: carries an arrow pointing toward the vertex $\vvv_{0}$.
1449: We can add an extra (oriented) line $\ell_{0}$
1450: connecting the special vertex $\vvv_{0}$ to another point which
1451: will be called the {\rm root} of the tree; the added line will
1452: be called the {\rm root line}. In this way we obtain
1453: a {\rm rooted tree} $\th$ defined by $P(\th)=P(\GGG)$
1454: and $L(\th)=L(\GGG)\cup\ell_{0}$. A {\rm labeled tree} is a
1455: rooted tree $\th$ together with a label function defined on
1456: the sets $L(\th)$ and $P(\th)$.}
1457:
1458: \*
1459:
1460: Note that the definition of rooted tree
1461: given above is sligthly different from the one which is
1462: usually adopted in literature \cita{GoR}, \cita{HP}
1463: according to which a rooted tree is just
1464: a tree with a privileged vertex, without any extra line.
1465: However the modified definition that we gave will
1466: be more convenient for our purposes.
1467: In the following we shall denote with the symbol $\th$ both
1468: rooted trees and labels rooted trees, when no confusion arises
1469:
1470: \*
1471:
1472: We shall call equivalent two rooted trees which can be transformed
1473: into each other by continuously deforming the lines in the plane
1474: in such a way that the latter do not cross each other
1475: (i.e. without destroying the graph structure).
1476: We can extend the notion of equivalence also to labeled trees,
1477: simply by considering equivalent two labeled trees if they
1478: can be transformed into each other in such a way that also
1479: the labels match.
1480:
1481: Given two points $\vvv,\www\in P(\th)$, we say that $\vvv \prec \www$
1482: if $\vvv$ is on the path connecting $\www$ to the root line.
1483: We can identify a line with the points it connects;
1484: given a line $\ell=(\vvv,\www)$ we say that $\ell$
1485: enters $\vvv$ and comes out of $\www$.
1486:
1487: In the following we shall deal mostly with labeled trees:
1488: for simplicity, where no confusion can arise, we shall call them just
1489: trees. We consider the following {\it diagramamtic rules}
1490: to construct the trees we have to deal with; this
1491: will implicitly define also the label function.
1492:
1493: \*
1494:
1495: \0(1) We call {\it nodes} the vertices such that there is at least
1496: one line entering them. We call {\it end-points} the vertices
1497: which have no entering line. We denote with $L(\th)$, $V(\th)$
1498: and $E(\th)$ the set of lines, nodes and end-points, respectively.
1499: Of course $P(\th)=V(\th)\cup E(\th)$.
1500:
1501: \*
1502:
1503: \0(2) There can be two types of lines, $w$-lines and $v$-lines,
1504: so we can associate to each line $\ell\in L(\th)$ a {\it badge}
1505: label $\g_{\ell}\in\{v,w\}$ and a {\it momentum}
1506: $(n_{\ell},m_{\ell})\in\ZZZ^{2}$,
1507: to be defined in item (8) below. If $\g_{\ell}=v$ one
1508: has $|n_{\ell}|=|m_{\ell}|$, while if $\g_{\ell}=w$
1509: one has $|n_{\ell}|\neq |m_{\ell}|$. One can have
1510: $(n_{\ell},m_{\ell})=(0,0)$ only if $\ell$ is a $v$-line.
1511: To the $v$-lines $\ell$ with $n_{\ell}\neq 0$ we also associate
1512: a label $\d_{\ell}\in\{1,2\}$. All the lines coming out from
1513: the end-points are $v$-lines.
1514:
1515: \*
1516:
1517: \0(3) To each line $\ell$ coming out from a node
1518: we associate a {\it propagator}
1519: %
1520: $$ g_{\ell} \= g (\o n_{\ell},m_{\ell}) =
1521: \cases{
1522: \displaystyle
1523: {1\over - \o^{2} n_{\ell}^2 +\tilde\o_{m_{\ell}}^2} , &
1524: if $\g_{\ell}=w$, \cr
1525: & \cr
1526: \displaystyle {1 \over (i n_{\ell})^{\d_{\ell}} } , &
1527: if $\g_{\ell}=v$, $n_{\ell}\neq 0$ , \cr
1528: & \cr
1529: \displaystyle 1 , &
1530: if $\g_{\ell}=v$, $n_{\ell} = 0$ , \cr}
1531: \Eq(3.1) $$
1532: %
1533: with momentum $(n_{\ell},m_{\ell})$.
1534: We can associate also a propagator to the lines $\ell$ coming out
1535: from end-points, simply by setting $g_{\ell}=1$.
1536:
1537: \*
1538:
1539: \0(4) Given any node $\vvv\in V(\th)$ denote with $s_{\vvvv}$ the
1540: number of entering lines ({\it branching number}): one can have
1541: only either $s_{\vvvv}=1$ or $s_{\vvvv}=3$.
1542: Also the nodes $\vvv$ can be of $w$-type and $v$-type: we say
1543: that a node is of $v$-type if the line $\ell$
1544: coming out from it has label $\g_{\ell}=v$;
1545: analogously the nodes of $w$-type are defined.
1546: We can write $V(\th)=V_{v}(\th)\cup V_{w}(\th)$,
1547: with obvious meaning of the symbols; we also call
1548: $V_{w}^{s}(\th)$, $s=1,3$, the set of nodes in $V_{w}(\th)$ with $s$
1549: entering lines, and analogously we define $V_{v}^{s}(\th)$, $s=1,3$.
1550: If $\vvv\in V_{v}^{3}$ and two entering lines come out of end points
1551: then the remaining line entering $\vvv$ has to be a $w$-line.
1552: If $\vvv\in V_{w}^{1}$ then the line entering $\vvv$ has to be
1553: a $w$-line. If $\vvv\in V_{v}^{1}$ then its entering
1554: line comes out of an end-node.
1555:
1556: \*
1557:
1558: \0(5) To the nodes $\vvv$ of $v$-type we associate a label
1559: $j_{\vvv}\in \{1,2,3,4\}$ and, if $s_{\vvvv}=1$,
1560: an {\it order} label $k_{\vvvv}$,
1561: with $k_{\vvvv}\ge 1$.
1562: Moreover we associate to each node $\vvv$ of $v$-type two {\it mode} labels
1563: $(n_{\vvvv}',m_{\vvvv}')$, with $m_{\vvvv}'=\pm n_{\vvvv}'$,
1564: and $(n_{\vvvv},m_{\vvvv})$, with $m_{\vvvv}=\pm n_{\vvvv}$,
1565: and such that one has
1566: %
1567: $$ { m_{\vvvv}' \over n_{\vvvv}' } = { m_{\vvvv} \over n_{\vvvv}}
1568: = { \displaystyle \sum_{i=1}^{s_{\vvvv}} m_{\ell_{i}} \over
1569: \displaystyle \sum_{i=1}^{s_{\vvvv}} n_{\ell_{i}} } ,
1570: \Eq(3.1a) $$
1571: %
1572: where $\ell_{i}$ are the lines entering $\vvv$.
1573: We shall refer to them as the {\it first mode} label
1574: and the {\it second mode} label, respectively.
1575: To a node $\vvv$ of $v$-type we associate also
1576: a {\it node factor} $\h_{\vvvv}$ defined as
1577: %
1578: $$ \h_{\vvvv} = \cases{
1579: - B_{\mmmm} \O_{\mmmm}^{-2} D_{\mmmm}^{2} s_{n_{\vvvv}'} s_{n_{\vvvv}} , &
1580: if $j_{\vvvv}=1$ and $s_{\vvvv}=3$ , \cr
1581: - 6 B_{\mmmm} \O_{\mmmm}^{-2} D_{\mmmm}^{2} s_{n_{\vvvv}'} s_{n_{\vvvv}}
1582: C^{(k_{\vvvv})} , &
1583: if $j_{\vvvv}=1$ and $s_{\vvvv}=1$ , \cr
1584: %
1585: - B_{\mmmm} cd_{n_{\vvvv}'} cd_{n_{\vvvv}} , &
1586: if $j_{\vvvv}=2$ and $s_{\vvvv}=3$ , \cr
1587: - 6 B_{\mmmm} cd_{n_{\vvvv}'} cd_{n_{\vvvv}}
1588: C^{(k_{\vvvv})} , &
1589: if $j_{\vvvv}=2$ and $s_{\vvvv}=1$ , \cr
1590: %
1591: - B_{\mmmm} \O_{\mmmm}^{-1} D_{\mmmm} s_{n_{\vvvv}'} cd_{n_{\vvvv}} , &
1592: if $j_{\vvvv}=3$ and $s_{\vvvv}=3$ , \cr
1593: - 6 B_{\mmmm} \O_{\mmmm}^{-1} D_{\mmmm} s_{n_{\vvvv}'} cd_{n_{\vvvv}}
1594: C^{(k_{\vvvv})} , &
1595: if $j_{\vvvv}=3$ and $s_{\vvvv}=1$ , \cr
1596: %
1597: B_{\mmmm} \O_{\mmmm}^{-1} D_{\mmmm} cd_{n_{\vvvv}'} s_{n_{\vvvv}} , &
1598: if $j_{\vvvv}=4$ and $s_{\vvvv}=3$ , \cr
1599: 6 B_{\mmmm} \O_{\mmmm}^{-1} D_{\mmmm} cd_{n_{\vvvv}'} s_{n_{\vvvv}}
1600: C^{(k_{\vvvv})}, &
1601: if $j_{\vvvv}=4$ and $s_{\vvvv}=1$ . \cr}
1602: \Eq(3.2) $$
1603: %
1604: Note that the factors $C^{(k_{\vvvv})}=r_{0}^{1}\la a_0
1605: \LLL[f^{(k_{\vvvv})}]\ra$
1606: depend on the coefficients $u^{(k')}_{n,m}$, with $k'<k$,
1607: so that they have to been defined iteratively.
1608: The label $\d_{\ell}$ of the line $\ell$ coming out from
1609: a node $\vvv$ of $v$-type is related to the label $j_{\vvvv}$ of $v$:
1610: if $j_{\vvvv}=1$ then $n_{\ell}=0$, while if $j_{\vvvv}>1$ then
1611: $n_{\ell} \neq 0$ and $\d_{\ell}=1+\d_{j_{\vvvv},2}$,
1612: where $\d_{j_{\vvvv},1}$ denotes the Kronecker delta
1613: (so that $\d_{\ell}=2$ if $j_{\vvvv}+2$ and $\d_{\ell}=1$ otherwise).
1614:
1615: \*
1616:
1617: \0(6) To the nodes $\vvv$ of $w$-type we simply associate
1618: a {\it node factor} $\h_{\vvvv}$ given by
1619: %
1620: $$ \h_{\vvvv} = \cases{
1621: \e, & if $s_{\vvvv}=3$, \cr
1622: \n^{(c)}_{m_{\ell}} , & if $s_{\vvvv}=1$. \cr}
1623: \Eq(3.3) $$
1624: %
1625: In the latter case $(n_{\ell},m_{\ell})$
1626: is the momentum of the line coming out from $\vvv$,
1627: and one has $c=a$ if the momentum of the entering line is
1628: $(n_{\ell},m_{\ell})$ and
1629: $c=b$ if the momentum of the entering line is
1630: $(n_{\ell},-m_{\ell})$; we shall call $\n$-vertices
1631: the nodes $\vvv$ of $w$-type with $s_{\vvvv}=1$.
1632: In order to unify notations we can associate also
1633: to the nodes $\vvv$ of $w$-type two mode labels, by setting
1634: $(n_{\vvvv}',m_{\vvvv}')=(0,0)$ and $(n_{\vvvv},m_{\vvvv})=(0,0)$.
1635:
1636: \*
1637:
1638: \0(7) To the end-points $\vvv$ we associate only a first mode
1639: label $(n_{\vvvv}',m_{\vvvv}')$, with $|m_{\vvvv}'|=|n_{\vvvv}'|$,
1640: and an {\it end-point factor}
1641: %
1642: $$ V_{\vvvv} = (-1)^{1+\d_{n_{\vvvv}',m_{\vvvv}'}} a_{0,n_{\vvvv}'}
1643: = a_{0,m_{\vvvv}} .
1644: \Eq(3.4) $$
1645: %
1646: The line coming out from an end-point has to be a $v$-line.
1647:
1648: \*
1649:
1650: \0(8) The momentum $(n_{\ell},m_{\ell})$ of a line $\ell$ is
1651: related to the mode labels of the nodes preceding $\ell$;
1652: if a line $\ell$ comes out from a node $v$ one has
1653: %
1654: $$ \eqalign{
1655: n_{\ell} & = n_{\vvvv} + \sum_{\scriptstyle \wwww \in V(\th)
1656: \atop \scriptstyle \wwww \prec \vvvv} \left( n_{\wwww}' + n_{\wwww} \right)
1657: + \sum_{\scriptstyle \wwww \in E(\th) \atop
1658: \scriptstyle \wwww \prec \vvvv} n_{\wwww}' , \cr
1659: m_{\ell} & = m_{\vvvv} + \sum_{\scriptstyle \wwww \in V(\th)
1660: \atop \scriptstyle \wwww \prec \vvvv}
1661: \left( m_{\wwww}' + m_{\wwww} \right) +
1662: \sum_{\scriptstyle \wwww \in E(\th)
1663: \atop \scriptstyle \wwww \prec \vvvv} m_{\wwww}' , \cr}
1664: \Eq(3.5) $$
1665: %
1666: where some of the mode labels can be vanishing according
1667: to the notations introduced above. If $\ell$ comes out
1668: from an end-point we set $(n_{\ell},m_{\ell})=(0,0)$.
1669:
1670: \*
1671:
1672: We define $\Theta^{*(k)}_{n,m}$ as the {\it set of
1673: unequivalent labeled trees}, formed by following the rules
1674: (1) to (8) given above, and with the further following constraints:
1675: \\
1676: (i) if $(n_{\ell_{0}},m_{\ell_{0}})$ denotes the momentum
1677: flowing through the root line $\ell_{0}$ and
1678: $(n_{\vvvv_{0}}',m_{\vvvv_{0}}')$ is the first mode label associated
1679: to the node $\vvv_{0}$ which $\ell_{0}$ comes out from
1680: (special vertex), then one has
1681: $n=n_{\ell_{0}} + n_{\vvvv_{0}}'$ and $m=m_{\ell_{0}} + m_{\vvvv_{0}}'$;
1682: \\
1683: (ii) one has
1684: %
1685: $$ k = \left| V_{w}(\th) \right| +
1686: \sum_{\vvvv \in V_{v}^{1}(\th)} k_{\vvvv} ,
1687: \Eq(3.6) $$
1688: %
1689: with $k$ called the {\it order} of the tree.
1690:
1691: An example of tree is given in Figure 3.1, where only the labels $v,w$
1692: of the nodes have been explicitly written.
1693:
1694: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1695: % FIGURA 1
1696: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1697: \midinsert
1698: \*
1699: \insertplotttt{256pt}{151pt}{
1700: \ins{30pt}{69pt}{$v$}
1701: \ins{67pt}{52pt}{$w$}
1702: \ins{67pt}{87pt}{$w$}
1703: \ins{104pt}{33pt}{$w$}
1704: \ins{104pt}{105pt}{$w$}
1705: \ins{140pt}{15pt}{$v$}
1706: \ins{140pt}{52pt}{$w$}
1707: \ins{140pt}{105pt}{$v$}
1708: \ins{140pt}{124pt}{$v$}
1709: \ins{176pt}{15pt}{$w$}
1710: \ins{176pt}{69pt}{$w$}
1711: \ins{212pt}{33pt}{$v$}
1712: \ins{212pt}{15pt}{$w$}
1713: }
1714: {gmpfig1}
1715: \*
1716: \line{\vtop{\line{\hskip1.5truecm\vbox{\advance\hsize by -3.1 truecm
1717: \centerline{\nota
1718: \figura(3.1)}
1719: } \hfill} }}
1720: \*
1721: \endinsert
1722: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1723:
1724: \*
1725:
1726: \0{\bf Definition 3.}
1727: {\it For all $\th\in \Theta^{*(k)}_{n,m}$, we call
1728: %
1729: $$ \Val(\th) = \Big( \prod_{\ell \in L(\th)} g_{\ell} \Big)
1730: \Big( \prod_{\vvvv \in V(\th)} \h_{\vvvv} \Big)
1731: \Big( \prod_{\vvvv \in E(\th) } V_{\vvvv} \Big) ,
1732: \Eq(3.8) $$
1733: %
1734: the {\rm value} of the tree $\th$.}
1735:
1736: \*
1737:
1738: Then the main result about the formal expansion of the
1739: solution is provided by the following result.
1740:
1741: \*
1742:
1743: \0{\bf Lemma 2.} {\it We can write
1744: %
1745: $$ u^{(k)}_{n,m} =
1746: \sum_{\th\in \Theta^{*(k)}_{n,m}} \Val(\th) ,
1747: \Eq(3.9) $$
1748: %
1749: and if the root line $\ell_{0}$ is a $v$-line the tree value
1750: is a contribution to $v^{(k)}_{n,\pm n}$, while if
1751: $\ell_{0}$ is a $w$-line the tree value is
1752: a contribution to $w^{(k)}_{n,m}$.
1753: The factors $C^{(k)}$ are defined as
1754: %
1755: $$ C^{(k)} = r_{0}^{-1}
1756: {\mathop{\sum}_{\th\in \Theta^{*(k)}_{n,n}}}^{*} a_{0,-n} \Val(\th) ,
1757: \Eq(3.10) $$
1758: %
1759: where the $*$ in the sum means the extra constraint
1760: $s_{\vvvv_{0}}=3$ for the node immediately preceding the root
1761: (which is the special vertex of the rooted tree).}
1762:
1763: \*
1764:
1765: \0{\it Proof.}
1766: The proof is done by induction in $k$.
1767: Imagine to represent graphically $a_{0,n}$ as a (small) white bullet
1768: with a line coming out from it, as in Figure 3.2a, and
1769: $u^{(k)}_{n,m}$, $k\ge 1$, as a (big) black bullet
1770: with a line coming out from it, as in Figure 3.2b.
1771:
1772: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1773: % FIGURA 2
1774: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1775: \midinsert
1776: \*
1777: \insertplotttt{145pt}{21pt}{
1778: \ins{20pt}{-5pt}{a}
1779: \ins{110pt}{-5pt}{b}
1780: }
1781: {gmpfig2}
1782: \*
1783: \line{\vtop{\line{\hskip1.5truecm\vbox{\advance\hsize by -3.1 truecm
1784: \centerline{\nota
1785: \figura(3.2)}
1786: } \hfill} }}
1787: \*
1788: \endinsert
1789: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1790:
1791: One should imagine that labels $k,n,m$ are associated
1792: to the black bullet representing $u^{(k)}_{n,m}$,
1793: while a white bullet representing $a_{0,n}$ carries the
1794: labels $n,m=\pm n$.
1795:
1796: For $k=1$ the proof of \equ(3.9) and \equ(3.10) is just a check
1797: from the diagrammatic rules and the recursive
1798: definitions \equ(2.25) and \equ(2.27), and it can be performed as follows.
1799:
1800: Consider first the case $|n|\neq |m|$, so that $u^{(1)}_{n,m}=
1801: w^{(1)}_{n,m}$. By taking into account only the badge
1802: labels of the lines, by item (4) there is only one tree whose root line
1803: is a $w$-line, and it has one node $\vvv_{0}$ (the special vertex
1804: of the tree) with $s_{\vvvv_{0}}=3$,
1805: hence three end-points $\vvv_{1}$, $\vvv_{2}$ and $\vvv_{3}$.
1806: By applying the rules listed above one obtains, for $|n|\neq |m|$,
1807: %
1808: $$ w^{(1)}_{n,m}= {1\over -\o^2 n^2+\tilde\o^2_{m}}
1809: \sum_{n_1+n_2+n_3=n \atop m_1+m_2+m_3=m} v^{(0)}_{n_1,m_1}
1810: v^{(0)}_{n_2,m_2} v^{(0)}_{n_3,m_3} =
1811: \sum_{\th \in \Theta^{*(1)}_{n,m} } \Val(\th) ,
1812: \Eq(3.11) $$
1813: %
1814: where the sum is over all trees $\th$ which can be obtained
1815: from the tree appearing in Figure 3.3 by summing over
1816: all labels which are not explicitly written.
1817:
1818: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1819: % FIGURA 3
1820: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1821: \midinsert
1822: \*
1823: \insertplotttt{76pt}{79pt}{
1824: \ins{30pt}{30pt}{$w$}
1825: }
1826: {gmpfig3}
1827: \*
1828: \line{\vtop{\line{\hskip1.5truecm\vbox{\advance\hsize by -3.1 truecm
1829: \centerline{\nota
1830: \figura(3.3)}
1831: } \hfill} }}
1832: \*
1833: \endinsert
1834: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1835:
1836: It is easy to realize that \equ(3.11)
1837: corresponds to \equ(2.29) for $k=1$.
1838: Each end-point $\vvv_{i}$ is graphically a white bullet with
1839: first mode labels $(n_{i},m_{i})$ and second mode labels $(0,0)$,
1840: and has associated an end-point factor
1841: $(-1)^{1+\d_{n_{i},m_{i}}} a_{0,n_{i}}$ (see \equ(3.4) in item (7)).
1842: The node $\vvv_{0}$ is represented as a (small) gray bullet,
1843: with mode labels $(0,0)$ and $(0,0)$,
1844: and the factor associated to it is $\h_{\vvvv_{0}}=\e$
1845: (see \equ(3.3) in item (6)).
1846: We associate to the line $\ell$ coming out from the node $\vvv_{0}$
1847: a momentum $(n_{\ell},n_{\ell})$, with $n_{\ell}=n$, and a propagator
1848: $g_{\ell}=1/(-\o^{2}n_{\ell}^{2}+\tilde\o_{m_{\ell}}^{2})$
1849: (see \equ(3.1) in item (3)).
1850:
1851: Now we consider the case $|n| =|m|$, so that $u^{(1)}_{n,m}=
1852: \pm A^{(1)}_{n}$ (see \equ(2.14)).
1853: By taking into account only the badge labels of the lines,
1854: there are four trees contributing to $A^{(1)}_{n}$:
1855: they are represented by the four trees in Figure 3.4
1856: (the tree b and c are simply obtained from the tree a
1857: by a different choice of the $w$-line entering the last node).
1858:
1859: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1860: % FIGURA 4
1861: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1862: \midinsert
1863: \*
1864: \insertplotttt{407pt}{121pt}{
1865: \ins{26pt}{55pt}{$v$}
1866: \ins{54pt}{83pt}{$w$}
1867: \ins{140pt}{55pt}{$v$}
1868: \ins{166pt}{55pt}{$w$}
1869: \ins{256pt}{55pt}{$v$}
1870: \ins{286pt}{25pt}{$w$}
1871: \ins{372pt}{55pt}{$v$}
1872: \ins{40pt}{15pt}{a}
1873: \ins{168pt}{15pt}{b}
1874: \ins{275pt}{15pt}{c}
1875: \ins{370pt}{15pt}{d}
1876: }
1877: {gmpfig4}
1878: \*
1879: \line{\vtop{\line{\hskip1.5truecm\vbox{\advance\hsize by -3.1 truecm
1880: \centerline{\nota
1881: \figura(3.4)}
1882: } \hfill} }}
1883: \*
1884: \endinsert
1885: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1886:
1887: In the trees of Figures 3.4a, 3.4b and 3.4c the root line comes out from
1888: a node $\vvv_{0}$ (the special vertex of the tree) with $s_{\vvvv_{0}}=3$,
1889: and two of the entering lines come out from end-points:
1890: then the other line has to be a $w$-line (by item (4)),
1891: and \equ(3.6) requires that the subtree which has such a line as
1892: root line is exactly the tree represented in Figure 3.2.
1893: In the tree of Figure 4.4d the root line comes out from a node $\vvv_{0}$
1894: with $s_{\vvvv_{0}}=1$, hence the line entering $\vvv_{0}$ is a $v$-line
1895: coming out from an end-point (again see item (4)).
1896:
1897: By defining $\Theta^{*(1)}_{n,n}$ as the set of all labeled
1898: trees which can be obtained by assigning to the trees in Figure 3.4
1899: the labels which are not explicitly written, one finds
1900: %
1901: $$ A^{(1)}_{n} = \sum_{\th \in \Theta^{*(1)}_{n,n} } \Val(\th) ,
1902: \Eq(3.12) $$
1903: %
1904: which corresponds to the sum of two contributions. The first one
1905: arises from the trees of Figures 3.4a, 3.4b and 3.4c, and it is given by
1906: %
1907: $$ 3 \sum_{n'\in\zzz} \LLL_{n,n'}
1908: \sum_{\scriptstyle n'_1+n'_2+n'_3=n' \atop
1909: \scriptstyle m'_1+m'_2+m'_3=n'}
1910: v^{(0)}_{n'_1,m'_1} v^{(0)}_{n'_2,m'_2} w^{(1)}_{n'_3,m'_3} ,
1911: \Eq(3.13) $$
1912: %
1913: where one has
1914: %
1915: $$ \eqalign{
1916: \LLL_{n,n'} & =
1917: B_{\mmmm} \O_{\mmmm}^{-2} D_{\mmmm}^{2}
1918: {\mathop{\sum}_{\scriptstyle n_{1}+n_{2}=n-n' \atop
1919: \scriptstyle n_{2}=-n'}}^{*}
1920: s_{n_{1}} s_{n_{2}} +
1921: B_{\mmmm} {\mathop{\sum}_{n_{1}+n_{2}=n-n'}}^{*}
1922: {1 \over i^{2} (n_{2}+n')^{2}} cd_{n_{1}} cd_{n_{2}} \cr
1923: & \qquad \qquad +
1924: B_{\mmmm} \O_{\mmmm}^{-1} D_{\mmmm}
1925: {\mathop{\sum}_{n_{1}+n_{2}=n-n'}}^{*}
1926: {1 \over i (n_{2}+n')} s_{n_{1}} cd_{n_{2}} \cr
1927: & \qquad \qquad -
1928: B_{\mmmm} \O_{\mmmm}^{-1} D_{\mmmm}
1929: {\mathop{\sum}_{n_{1}+n_{2}=n-n'}}^{*}
1930: {1 \over i (n_{2}+n')} cd_{n_{1}} s_{n_{2}} , \cr}
1931: \Eq(3.14) $$
1932: %
1933: with the $*$ denoting the constraint $n_{2}+n'\neq 0$.
1934: The first and second mode labels associated to the node $\vvv_{0}$ are,
1935: respectively, $(m_{\vvvv_{0}}',n_{\vvvv_{0}}')=(n_{1},n_{1})$
1936: and $(m_{\vvvv_{0}},n_{\vvvv_{0}})=(n_{2},n_{2})$, while
1937: the momentum flowing through the root line is given
1938: by $(n_{\ell},m_{\ell})$, with $|m_{\ell}|=|n_{\ell}|$ expressed
1939: according to the definition \equ(3.5) in item (8):
1940: the corresponding propagator is $(n_{\ell})^{\d_{\ell}}$ for
1941: $n_{\ell}\neq 0$ and $1$ for $n_{\ell}=0$, as in \equ(3.1) in item (3).
1942:
1943: The second contribution corresponds to the tree of Figure 3.4d,
1944: and it is given by
1945: %
1946: $$ \sum_{n'\in\zzz} \LLL_{n,n'} C^{(1)} a_{0,n'} ,
1947: \Eq(3.15) $$
1948: %
1949: with the same expression \equ(3.14) for $\LLL_{n,n'}$ and
1950: $C^{(1)}$ still undetermined.
1951: The mode labels of the node $\vvv_{0}$ and the momentum
1952: of the root line are as before.
1953:
1954: Then one immediately realize that the sum of \equ(3.13)
1955: and \equ(3.15) corresponds to \equ(2.25) for $k=1$.
1956:
1957: Finally that $C^{(1)}$ is given by \equ(3.9)
1958: follows from \equ(2.10). This completes the check of the case $k=1$.
1959:
1960: In general from \equ(2.29) one gets,
1961: for $\th\in \Theta^{*(k)}_{n,m}$ contributing
1962: to $w^{(k)}_{n,m}$, that the tree value $\Val(\th)$
1963: is obtained by summing all contribution either of the form
1964: %
1965: $$ \eqalign{
1966: & {1\over -\o^2 n^2 + \tilde\o_{m}^2} \, \e
1967: \sum_{\scriptstyle n_{1}+n_{2}+n_{3}=n \atop
1968: \scriptstyle m_{1}+m_{2}+m_{3}=m}
1969: \sum_{k_{1}+k_{2}+k_{3}=k-1} \cr& \qquad
1970: \sum_{\th_{1} \in \Theta^{*(k_{1})}_{n_{1},m_{1}}}
1971: \sum_{\th_{2} \in \Theta^{*(k_{2})}_{n_{2},m_{2}}}
1972: \sum_{\th_{3} \in \Theta^{*(k_{3})}_{n_{3},m_{3}}}
1973: \Val(\th_1)\,\Val(\th_2) \, \Val(\th_3) , \cr}
1974: \Eq(3.16) $$
1975: %
1976: or of the form
1977: %
1978: $$ {1\over -\o^2 n^2 + \tilde\o_{m}^2} \sum_{c=a,b} \n_{m}^{(c)}
1979: \sum_{\th_{1} \in \Theta^{*(k,1)}_{n,m^{(c)}}} \Val(\th_1) ,
1980: \Eq(3.17) $$
1981: %
1982: with $m^{(a)}=m$ and $m^{(b)}=-m$; the corresponding
1983: graphical representations are as in Figure 3.5.
1984:
1985: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1986: % FIGURA 5
1987: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1988: \midinsert
1989: \*
1990: \insertplotttt{375pt}{183pt}{
1991: \ins{25pt}{34pt}{$w$}
1992: \ins{25pt}{135pt}{$w$}
1993: \ins{126pt}{34pt}{$w$}
1994: \ins{126pt}{135pt}{$w$}
1995: \ins{228pt}{34pt}{$w$}
1996: \ins{228pt}{135pt}{$w$}
1997: \ins{328pt}{34pt}{$w$}
1998: \ins{328pt}{135pt}{$w$}
1999: \ins{30pt}{110pt}{a}
2000: \ins{130pt}{110pt}{b}
2001: \ins{230pt}{110pt}{c}
2002: \ins{330pt}{110pt}{d}
2003: \ins{30pt}{10pt}{e}
2004: \ins{130pt}{10pt}{f}
2005: \ins{230pt}{10pt}{g}
2006: \ins{330pt}{10pt}{h}
2007: }
2008: {gmpfig5}
2009: \*
2010: \line{\vtop{\line{\hskip1.5truecm\vbox{\advance\hsize by -3.1 truecm
2011: \centerline{\nota
2012: \figura(3.5)}} \hfill} }}
2013: \*
2014: \endinsert
2015: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2016:
2017: Therefore, by simply applying the diagrammatic rules given above,
2018: we see that by summing together the contribution \equ(3.16)
2019: and \equ(3.17) we obtain \equ(3.9) for $|n| \neq |m|$.
2020:
2021: A similar discussion applies to $A^{(k)}_{n}$,
2022: and one finds that $A^{(k)}_{n}$ can be written as
2023: sum of contribution either of the form
2024: %
2025: $$ \eqalign{
2026: & \sum_{n'\in\zzz} \LLL_{n,n'}
2027: \sum_{\scriptstyle n'_1+n'_2+n'_3=n' \atop
2028: \scriptstyle m'_1+m'_2+m'_3=n'}
2029: \sum_{k_{1}+k_{2}+k_{3}=k} \cr & \qquad
2030: \sum_{\th_{1} \in \Theta^{*(k_{1})}_{n_{1},m_{1}}}
2031: \sum_{\th_{2} \in \Theta^{*(k_{2})}_{n_{2},m_{2}}}
2032: \sum_{\th_{3} \in \Theta^{*(k_{3})}_{n_{3},m_{3}}}
2033: \Val(\th_1)\,\Val(\th_2) \, \Val(\th_3) , \cr}
2034: \Eq(3.18) $$
2035: %
2036: or of the form
2037: %
2038: $$ \sum_{n'\in\zzz} \LLL_{n,n'} C^{(k)} a_{0,n'} ,
2039: \Eq(3.19) $$
2040: %
2041: with $C^{(k)}$ still undetermined. Both
2042: \equ(3.18) and \equ(3.19) are of the form $\Val(\th)$,
2043: for $\th\in\Theta^{*(k)}_{n,m}$.
2044: A graphical representation is in Figure 3.6.
2045:
2046: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2047: % FIGURA 6
2048: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2049: \midinsert
2050: \*
2051: \insertplotttt{429pt}{74pt}{
2052: \ins{23pt}{30pt}{$v$}
2053: \ins{112pt}{30pt}{$v$}
2054: \ins{203pt}{30pt}{$v$}
2055: \ins{294pt}{30pt}{$v$}
2056: \ins{386pt}{30pt}{$v$}
2057: \ins{25pt}{10pt}{a}
2058: \ins{118pt}{10pt}{b}
2059: \ins{209pt}{10pt}{c}
2060: \ins{300pt}{10pt}{d}
2061: \ins{390pt}{10pt}{e}
2062: }
2063: {gmpfig6}
2064: \*
2065: \line{\vtop{\line{\hskip1.5truecm\vbox{\advance\hsize by -3.1 truecm
2066: \centerline{\nota
2067: \figura(3.6)}} \hfill} }}
2068: \*
2069: \endinsert
2070: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2071:
2072: Analogously to the case $k=1$ the coefficients $C^{(k)}$
2073: are found to be expressed by \equ(3.10).
2074: Then the lemma is proved. \qed
2075:
2076: \*
2077:
2078: \0{\bf Lemma 3.}
2079: {\it For any rooted tree $\th$ one has
2080: $|V_{v}^{3}(\th)| \le 2|V_{w}^{3}(\th)|+2|V_{v}^{1}(\th)|$
2081: and $|E(\th)|\le 2(|V_{v}^{3}(\th)|+|V_{w}^{3}(\th)|)+1$.}
2082:
2083: \*
2084:
2085: \0{\it Proof.} First of all note that $|V_{w}^{3}(\th)|=0$
2086: requires $|V_{v}^{1}(\th)| \ge 1$, so that one has
2087: $|V_{w}^{3}(\th)|+|V_{v}^{1}(\th)|\ge 1$ for all trees $\th$.
2088:
2089: We prove by induction on the number $N$ of nodes the bound
2090: %
2091: $$ \left| V_{v}^{3}(\th) \right| \le \cases{
2092: 2|V_{w}^{3}(\th)|+2|V_{v}^{1}(\th)|- 1, &
2093: if the root line is a $v$-line , \cr
2094: 2|V_{w}^{3}(\th)|+2|V_{v}^{1}(\th)| - 2 &
2095: if the root line is a $w$-line , \cr}
2096: \Eq(3.20) $$
2097: %
2098: which will immediately imply the first assertion.
2099:
2100: For $N=1$ the bound is trivially satisfied, as Figures 3.3 and 3.4 show.
2101:
2102: Then assume that \equ(3.20) holds for the trees with $N'$ nodes,
2103: for all $N'<N$, and consider a tree $\th$ with $V(\th)=N$.
2104:
2105: If the special vertex $\vvv_{0}$ of $\th$ is not in $V_{v}^{3}(\th)$
2106: (hence it is in $V_{w}(\th)$) the bound \equ(3.20) follows
2107: trivially by the inductive hypothesis.
2108:
2109: If $\vvv_{0}\in V_{v}^{3}(\th)$ then we can write
2110: %
2111: $$ |V_{v}^{3}(\th)| = 1 + \sum_{i=1}^{s} |V_{v}^{3}(\th_{i})| ,
2112: \Eq(3.21) $$
2113: %
2114: where $\th_{1},\ldots,\th_{s}$ are the subtrees whose
2115: root line is one of the lines entering $\vvv_{0}$.
2116: One must have $s\ge1$, as $s=0$ would correspond to have
2117: all the entering lines of $\vvv_{0}$ coming out from end-points,
2118: hence to have $N=1$.
2119:
2120: If $s\ge 2$ one has from \equ(3.21) and from the inductive hypothesis
2121: %
2122: $$ |V_{v}^{3}(\th)| \le 1 + \sum_{i=1}^{s}
2123: \left( 2 |V_{w}^{3}(\th_{i})| + 2 |V_{v}^{1}(\th_{i})| - 1 \right)
2124: \le 1 + 2|V_{w}^{3}(\th)| + 2|V_{v}^{1}(\th)| - 2 ,
2125: \Eq(3.22) $$
2126: %
2127: and the bound \equ(3.20) follows.
2128:
2129: If $s=1$ then the root line of $\th_{1}$ has to be a $w$-line
2130: by item (4), so that one has
2131: %
2132: $$ |V_{v}^{3}(\th)| \le 1 + \left( 2 |V_{w}^{3}(\th_{1})|
2133: + 2 |V_{v}^{1}(\th)| - 2 \right)
2134: \Eq(3.23) $$
2135: %
2136: which again yields \equ(3.20).
2137:
2138: Finally the second assertion follows from the
2139: standard (trivial) property of trees
2140: %
2141: $$ \sum_{\vvvv\in V(\th)} \left( s_{\vvvv}-1 \right) = |E(\th)| - 1 ,
2142: \Eq(3.24) $$
2143: %
2144: and the observation that in our case one has $s_{\vvvv}\le 3$. \qed
2145:
2146:
2147: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2148: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2149: \*\*
2150: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2151: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2152: \section(4,Tree expansion: the multiscale decomposition)
2153: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2154: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2155:
2156: \0We assume the Diophantine conditions \equ(2.31). We introduce a
2157: multiscale decomposition of the propagators of the $w$-lines.
2158: Let $\chi(x)$ be a $C^\io$ non-increasing function such that
2159: $\chi(x)=0$ if $|x|\ge C_{0}$ and $\chi(x)=1$ if $|x|\le 2C_{0}$
2160: ($C_0$ is the same constant appearing in \equ(2.31)),
2161: and let $\chi_h(x)=\chi(2^{h}x)-\chi(2^{h+1}x)$ for
2162: $h \ge 0$, and $\chi_{-1}(x)=1-\chi(x)$; such functions
2163: realize a smooth partition of the unity as
2164: %
2165: $$ 1=\chi_{-1}(x)+\sum_{h=0}^{\io} \chi_h(x)
2166: = \sum_{h=-1}^{\io} \chi_h(x) .
2167: \Eq(4.1) $$
2168: %
2169: If $\chi_{h}(x) \neq 0$ for $h\ge 0$ one has
2170: $2^{-h-1}C_{0}\le|x|\le2^{-h+1}C_{0}$, while if $\chi_{-1}(x)\neq 0$
2171: one has $|x| \ge C_{0}$.
2172:
2173: We write the propagator of any $w$-line
2174: as sum of propagators on single scales in the following way:
2175: %
2176: $$ g (\o n,m) =
2177: \sum_{h=-1}^{\io}
2178: { \chi_h( | \o n|-\tilde\o_{m})
2179: \over - \o^{2} n^2+\tilde\o_{m}^2 }
2180: = \sum_{h=-1}^{\io} g^{(h)}(\o n,m) .
2181: \Eq(4.2)$$
2182: %
2183: Note that we can bound $|g^{(h)}(\o n,m)|\le 2^{-h+1}C_{0}$.
2184:
2185: This means that we can attach to each $w$-line $\ell$ in $L(\th)$
2186: a scale label $h_{\ell}\ge -1$, which is the scale of the propagator
2187: which is associated to $\ell$. We can denote with
2188: $\Theta^{(k)}_{n,m}$ the set of trees
2189: which differ from the previous ones simply because
2190: the lines carry also the scale labels.
2191: The set $\Theta^{(k)}_{n,m}$ is defined according to the
2192: rules (1) to (8) of Section \secc(3), by changing item (3) into
2193: the following one.
2194:
2195: \*
2196:
2197: \0(3$^{\prime}$) To each line $\ell$ coming out from nodes of $w$-type
2198: we associate a {\it scale label} $h_{\ell}\ge -1$.
2199: For notational convenience we associate a scale label
2200: $h=-1$ to the lines coming out from the nodes
2201: of $v$-type and to the lines coming out from the end-points.
2202: To each line $\ell$ we associate a {\it propagator}
2203: %
2204: $$ g^{(h_{\ell})}_{\ell} \= g^{(h_{\ell})} (\o n_{\ell},m_{\ell}) =
2205: \cases{
2206: \displaystyle
2207: { \chi_{h_{\ell}} ( | \o n_{\ell}|-\tilde\o_{m_{\ell}})
2208: \over - \o^{2} n_{\ell}^2 +\tilde\o_{m_{\ell}}^2} , &
2209: if $\g_{\ell}=w$, \cr
2210: \displaystyle {1 \over (i n_{\ell})^{\d_{\ell}} } , &
2211: if $\g_{\ell}=v$, $n_{\ell}\neq 0$ , \cr
2212: & \cr
2213: \displaystyle 1 , &
2214: if $\g_{\ell}=v$, $n_{\ell} = 0$ , \cr}
2215: \Eq(4.3) $$
2216: %
2217: with momentum $(n_{\ell},m_{\ell})$.
2218:
2219: \*
2220:
2221: \0{\bf Definition 4.}
2222: {\it For all $\th\in \Theta^{(k)}_{n,m}$, we define
2223: %
2224: $$ \Val(\th) =
2225: \Big( \prod_{\ell \in L(\th)} g_{\ell}^{(h_{\ell})} \Big)
2226: \Big( \prod_{\vvvv \in V(\th)} \h_{{\vvvv}} \Big)
2227: \Big( \prod_{\vvvv \in E(\th) } V_{\vvvv} \Big) ,
2228: \Eq(4.4) $$
2229: %
2230: the {\rm value of the tree} $\th$.}
2231:
2232: \*
2233:
2234: Then \equ(3.9) and \equ(3.10) are replaced, respectively, with
2235: %
2236: $$ u^{(k)}_{n,m} =
2237: \sum_{\th\in\Th^{(k)}_{ n,m}} \Val(\th) ,
2238: \Eq(4.5) $$
2239: %
2240: and
2241: %
2242: $$ C^{(k)} = r_{0}^{-1}
2243: {\mathop{\sum}_{\th\in \Th^{(k)}_{n,n}}}^{*}
2244: a_{0,-n} \Val(\th) ,
2245: \Eq(4.6) $$
2246: %
2247: with the new definition for the tree value $\Val(\th)$
2248: and with $*$ meaning the same constraint as in \equ(3.10).
2249:
2250: \*
2251:
2252: \0{\bf Definition 5.} {\it
2253: A {\rm cluster} $T$ is a connected set of nodes which are linked
2254: by a continuous path of lines with the same scale label $h_{T}$ or
2255: a lower one and which are maximal; we shall say that the cluster
2256: has scale $h_{T}$. We shall denote with $V(T)$ and $E(T)$
2257: the set of nodes and the set of end-points, respectively, which are
2258: contained inside the cluster $T$, and with $L(T)$
2259: the set of lines connecting them. As for trees
2260: we call $V_{v}(T)$ and $V_{w}(T)$ the sets
2261: of nodes $\vvv\in V(T)$ which are of $v$-type and
2262: of $w$-type respectively.}
2263:
2264: \*
2265:
2266: We define the {\rm order} $k_{T}$ of a cluster $T$
2267: as the order of a tree (see item (ii) before Definition 3),
2268: with the sums restricted to the nodes internal to the cluster.
2269:
2270: An inclusion relation is established between clusters,
2271: in such a way that the innermost clusters are the clusters
2272: with lowest scale, and so on. Each cluster $T$ can have an arbitrary
2273: number of lines entering it (incoming lines), but only one or zero line
2274: coming from it (outcoming line); we shall denote the latter
2275: (when it exists) with $\ell_{T}^{1}$.
2276: We shall call {\it external lines} of the cluster $T$
2277: the lines which either enter or come out from $T$, and
2278: we shall denote by $h_{T}^{(e)}$
2279: the minimum among the scales of the external lines of $T$.
2280: Define also
2281: %
2282: $$ \eqalign{
2283: K(\th) & = \sum_{\vvvv \in V(\th)}
2284: \left( |n_{\vvvv}'|+|n_{\vvvv}| \right) +
2285: \sum_{\vvvv \in E(\th) } |n_{\vvvv}'| , \cr
2286: K(T) & = \sum_{\vvvv \in V(T)}
2287: \left( |n_{\vvvv}'|+|n_{\vvvv}| \right) +
2288: \sum_{\vvvv \in E(\th) } |n_{\vvvv}'| , \cr}
2289: \Eq(4.7) $$
2290: %
2291: where we recall that one has $(n_{\vvvv}',m_{\vvvv}')=
2292: (n_{\vvvv},m_{\vvvv})=0$ if $\vvv\in V(\th)$ is of $w$-type.
2293:
2294: \*
2295:
2296: If a cluster has only one entering line $\ell_{T}^{2}$
2297: and $(n,m)$ is the momentum of such a line,
2298: for any line $\ell\in L(T)$ one can write
2299: $ (n_{\ell},m_{\ell}) = (n^{0}_{\ell},m^{0}_{\ell}) + \h_{\ell}(n,m)$,
2300: where $\h_{\ell}=1$ if the line $\ell$ is along the path connecting
2301: the external lines of $T$ and $\h_{\ell}=0$ otherwise.
2302:
2303: \*
2304:
2305: \0{\bf Definition 6.} {\it A cluster $T$ with only one incoming
2306: line $\ell_{T}^{2}$ such that one has
2307: %
2308: $$ n_{\ell_{T}^{1}} = n_{\ell_{T}^{2}} \quad \hbox{ and } \quad
2309: m_{\ell_{T}^{1}}= \pm m_{\ell_{T}^{2}}
2310: \Eq(4.8) $$
2311: %
2312: will be called a {\rm self-energy graph} or a {\rm resonance}.
2313: In such a case we shall call a {\rm resonant line}
2314: the line $\ell_{T}^{1}$, and we shall refer to its momentum
2315: as the momentum of the self-energy graph.}
2316:
2317: \*
2318:
2319: Examples of self-energy graphs $T$ with $k_{T}=1$ are represented
2320: in Figure 4.1. The lines crossing the encircling bubbles
2321: are the external lines, and in the figure are assumed to be on
2322: scales higher than the lines internal to the bubbles. There
2323: are 9 self-energy graphs with $k_{T}=1$: they are all obtained
2324: by the three drawn in Figure 4.1, simply by considering
2325: all possible unequivalent trees.
2326:
2327: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2328: % FIGURA 7
2329: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2330: \midinsert
2331: \*
2332: \insertplotttt{221pt}{113pt}{
2333: \ins{24pt}{64pt}{$w$}
2334: \ins{140pt}{64pt}{$w$}
2335: \ins{168pt}{36pt}{$v$}
2336: \ins{24pt}{12pt}{a}
2337: \ins{142pt}{12pt}{b}
2338: }
2339: {gmpfig7}
2340: \*
2341: \line{\vtop{\line{\hskip1.5truecm\vbox{\advance\hsize by -3.1 truecm
2342: \centerline{\nota
2343: \figura(4.1)}} \hfill} }}
2344: \*
2345: \endinsert
2346: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2347:
2348: \*
2349:
2350: \0{\bf Definition 7.}
2351: {\it The {\rm value of the self-energy graph $T$} with momentum $(n,m)$
2352: associated to the line $\ell_{T}^{2}$ is defined as
2353: %
2354: $$\VV_{T}^{h}(\o n,m)=
2355: \Big( \prod_{\ell \in T} g^{(h_{\ell})}_{\ell} \Big)
2356: \Big( \prod_{\vvv \in V(T)} \h_{\vvvv} \Big)
2357: \Big( \prod_{\vvv \in E(T) } V_{\vvvv} \Big) .
2358: \Eq(4.9) $$
2359: %
2360: where $h=h^{(e)}_T$ is the minimum between the scales of the
2361: two external lines of $T$ (they can differ at most by a unit),
2362: and one has
2363: %
2364: $$ \eqalign{
2365: n(T) & \= \sum_{\vvvv \in V(T)} \left( n_{\vvvv}' + n_{\vvvv} \right) +
2366: \sum_{\vvvv\in E(T) } n_{\vvvv}' = 0 , \cr
2367: m(T) & \= \sum_{\vvvv \in V(T)} \left( m_{\vvvv}' + m_{\vvvv} \right) +
2368: \sum_{\vvvv\in E(T) } m_{\vvvv}' \in \{0,2m\} , \cr}
2369: \Eq(4.10) $$
2370: %
2371: by definition of self-energy graph; one says that $T$ is
2372: a resonance of type $c=a$ when $m(T)=0$ and a resonance
2373: of type $c=b$ when $m(T)=2m$.}
2374:
2375: \*
2376:
2377: \0{\bf Definition 8.}
2378: {\it Given a tree $\th$, we shall denote by
2379: $N_{h}(\th)$ the number of lines with scale $h$, and by
2380: $C_{h}(\th)$ the number of clusters with scale $h$.}
2381:
2382: \*
2383:
2384: Then the product of propagators appearing in \equ(4.4)
2385: can be bounded as
2386: %
2387: $$ \Big| \prod_{\ell \in L(\th)} g_{\ell}^{(h_{\ell})} \Big| \le
2388: \Big( \prod_{h=0}^{\io} 2^{h N_{h}(\th)} \Big)
2389: \Big( \prod_{\scriptstyle \ell\in L(\th) \atop \scriptstyle \g_{\ell} =w}
2390: {\chi_{h_{\ell}}(|\o n_{\ell}|-
2391: \tilde \o_{m_{\ell}}) \over |\o n_{\ell}|+\tilde\o_{m_{\ell}} } \Big)
2392: \Big( \prod_{\scriptstyle \ell\in L(\th) \atop
2393: \scriptstyle \g_{\ell} =v, \, n_{\ell}\neq 0}
2394: {1 \over |\o n_{\ell}|} \Big) ,
2395: \Eq(4.11) $$
2396: %
2397: and this will be used later.
2398:
2399: \*
2400:
2401:
2402: \0{\bf Lemma 4.} {\it Assume $0<C_{0}<1/2$ and that there is a
2403: constant $C_{1}$ such that one has $| \tilde\o_{m}-|m| |\le C_{1}\e/|m|$.
2404: If $\e$ is small enough for any tree $\th \in \Th_{n,m}^{(k)}$
2405: and for any line $\ell$ on a scale $h_{\ell} \ge 0$ one has
2406: $\min\{m_{\ell},n_{\ell}\} \ge 1/2\e$.}
2407:
2408: \*
2409:
2410: \0{\it Proof.} If a line $\ell$ with momentum $(n,m)$ is on scale
2411: $h \ge 0$ then one has
2412: %
2413: $$ \eqalign{
2414: {1 \over 2} > C_{0} \ge \left| |\o n| - \tilde\o_{m} \right| & \ge
2415: \left| \left( \sqrt{1-\e} - 1 \right) |n| + \left( |n| - |m| \right)
2416: - C_{1} \e/|m| \right| \cr
2417: & \ge \left| \left| {\e \, |n| \over 1 + \sqrt{1-\e}} -
2418: \left(|n|-|m|\right) \right| - C_{1}\e/|m| \right| , \cr}
2419: \Eq(4.12) $$
2420: %
2421: with $|n|\neq|m|$, hence $|n-m|\ge 1$, so that $|n|\ge 1/2\e$.
2422: Moreover one has $||\o n|-\tilde \o_{m}| \le 1/2$ and
2423: $\tilde\o_{m}-|m|=O(\e)$, and one obtains also $|m|>1/2\e$. \qed
2424:
2425: \*
2426:
2427: \0{\bf Lemma 5.} {\it
2428: Define $h_{0}$ such that $2^{h_{0}} \le 16 C_{0}/\e < 2^{h_{0}+1}$,
2429: and assume that there is a constant $C_{1}$ such that
2430: one has $| \tilde\o_{m}-|m| |\le C_{1}\e/|m|$.
2431: If $\e$ is small enough for any tree $\th \in \Th_{n,m}^{(k)}$
2432: and for all $h\ge h_{0}$ one has
2433: %
2434: $$ N_h(\th) \le 4 K(\th) 2^{(2-h)/\t} -
2435: C_{h}(\th) + S_{h}(\th) + M_{h}^{\n}(\th) ,
2436: \Eq(4.13) $$
2437: %
2438: where $K(\th)$ is defined in \equ(4.7),
2439: while $S_{h}(\th)$ is the number of self-energy graphs $T$
2440: in $\th$ with $h_{T}^{(e)}=h$
2441: and $M_{h}^{\n}(\th)$ is the number of $\n$-vertices in $\th$
2442: such that the maximum scale of the two external lines is $h$.}
2443:
2444: \*
2445:
2446: \0{\it Proof.}
2447: We prove inductively the bound
2448: %
2449: $$ N^{*}_{h}(\th) \le \max\{ 0 , 2 K(\th) 2^{(2-h)/\t} - 1 \} ,
2450: \Eq(4.14) $$
2451: %
2452: where $N_{h}^{*}(\th)$ is the number of non-resonant lines in $L(\th)$
2453: on scale $h'\ge h$.
2454:
2455: First of all note that for a tree $\th$ to have a line on scale
2456: $h$ the condition $K(\th) > 2^{(h-1)/\t}$ is necessary,
2457: by the first Diophantine conditions in \equ(2.31).
2458: This means that one can have $N_{h}^{*}(\th) \ge 1$ only
2459: if $K=K(\th)$ is such that $K > k_{0}\=2^{(h-1)/\t}$:
2460: therefore for values $K \le k_{0}$ the bound \equ(4.14) is satisfied.
2461:
2462: If $K=K(\th)>k_{0}$, we assume that
2463: the bound holds for all trees $\th'$ with $K(\th')<K$.
2464: Define $E_{h}=2^{-1}(2^{(2-h)/\t})^{-1}$: so we have
2465: to prove that $N^{*}_{h}(\th)\le \max\{0,K(\th) E_{h}^{-1}-1\}$.
2466:
2467: Call $\ell$ the root line of $\th$ and
2468: $\ell_{1},\ldots,\ell_{m}$ the $m\ge 0$ lines on scale $\ge h$
2469: which are the closest to $\ell$ (i.e. such that no other line
2470: along the paths connecting the lines $\ell_{1},\ldots,\ell_{m}$
2471: to the root line is on scale $\ge h$).
2472:
2473: If the root line $\ell$ of $\th$ is either on scale $< h$
2474: or on scale $\ge h$ and resonant, then
2475: %
2476: $$ N_{h}^{*}(\th) = \sum_{i=1}^{m} N_{h}^{*}(\th_{i}) ,
2477: \Eq(4.15) $$
2478: %
2479: where $\th_{i}$ is the subtree with $\ell_{i}$ as root line,
2480: hence the bound follows by the inductive hypothesis.
2481:
2482: If the root line $\ell$ has scale $\ge h$ and is non-resonant,
2483: then $\ell_{1},\ldots,\ell_{m}$ are the entering line of a cluster $T$.
2484:
2485: By denoting again with $\th_{i}$ the subtree
2486: having $\ell_{i}$ as root line, one has
2487: %
2488: $$ N_{h}^{*}(\th) = 1 + \sum_{i=1}^{m} N_{h}^{*}(\th_{i}) ,
2489: \Eq(4.16) $$
2490: %
2491: so that the bound becomes trivial if either $m=0$ or $m\ge 2$.
2492:
2493: If $m=1$ then one has a cluster $T$ with two external lines
2494: $\ell$ and $\ell_{1}$, which are both with scales $\ge h$; then
2495: %
2496: $$ \left| |\o n_{\ell}| - \tilde\o_{m_{\ell}} \right|
2497: \le 2^{-h+1} C_{0} , \qquad
2498: \left| |\o n_{\ell_{1}}| - \tilde\o_{m_{\ell_{1}}} \right|
2499: \le 2^{-h+1} C_{0} ,
2500: \Eq(4.17) $$
2501: %
2502: and recall that $T$ is not a self-energy graph.
2503:
2504: Note that the validity of both inequalities in \equ(4.17)
2505: for $h \ge h_{0}$ imply that one has $|n_{\ell}-n_{\ell_{1}}|\neq
2506: |m_{\ell}\pm m_{\ell_{1}}|$, as we are going to show.
2507:
2508: By Lemma 4 we know that one has $\min\{m_{\ell},n_{\ell}\} \ge 1/2\e$.
2509: Then from \equ(4.17) we have,
2510: for some $\h_{\ell},\h_{\ell_{1}}\in\{\pm 1\}$,
2511: %
2512: $$ 2^{-h+2} C_{0} \ge \big| \o
2513: ( n_{\ell}-n_{\ell_{1}}) + \h_{\ell} \tilde\o_{m_{\ell}}+
2514: \h_{\ell_{1}} \tilde\o_{m_{\ell_{1}}} \big|,
2515: \Eq(4.18) $$
2516: %
2517: so that if one had $|n_{\ell}-n_{\ell_{1}}|
2518: = |m_{\ell}\pm m_{\ell_{1}}|$ we would obtain for $\e$ small enough
2519: %
2520: $$ 2^{-h+2} C_{0} \ge {\e \over 1 + \sqrt{1-\e}} \left|
2521: n_{\ell} - n_{\ell_{1}} \right| - {C_{1} \e \over |m_{\ell}|} -
2522: {C_{1} \e \over |m_{\ell_{1}}|} \ge {\e \over 2} - 4 C_{1} \e^{2} >
2523: {\e \over 4} ,
2524: \Eq(4.19) $$
2525: %
2526: which is contradictory with $h \le h_{0}$; hence one has
2527: $|n_{\ell}-n_{\ell_{1}}| \neq |m_{\ell}\pm m_{\ell_{1}}|$.
2528:
2529: Then, by \equ(4.17) and for $|n_{\ell}-n_{\ell_{1}}|
2530: \neq |m_{\ell}\pm m_{\ell_{1}}|$,
2531: one has, for suitable $\h_{\ell},\h_{\ell_{1}}\in\{+,-\}$,
2532: %
2533: $$ 2^{-h+2} C_{0} \ge \big| \o
2534: ( n_{\ell}-n_{\ell_{1}}) + \h_{\ell} \tilde\o_{m_{\ell}}+
2535: \h_{\ell_{1}} \tilde\o_{m_{\ell_{1}}} \big| \ge
2536: C_{0} |n_{\ell}-n_{\ell_{1}}|^{-\t} ,
2537: \Eq(4.20) $$
2538: %
2539: where the second Diophantine conditions in \equ(2.31) have been used.
2540: Hence $K(\th)-K(\th_{1})> E_{h}$,
2541: which, inserted into \equ(4.16) with $m=1$, gives,
2542: by using the inductive hypothesis,
2543: %
2544: $$ \eqalign{
2545: N_{h}^{*}(\th) & = 1 + N_{h}^{*}(\th_{1})
2546: \le 1 + K(\th_{1}) E_{h}^{-1} - 1 \cr
2547: & \le 1 + \Big( K(\th) - E_{h}\Big) E_{h}^{-1} - 1
2548: \le K(\th) E_{h}^{-1} - 1 , \cr}
2549: \Eq(4.21) $$
2550: %
2551: hence the bound is proved also
2552: if the root line is on scale $\ge h$.
2553:
2554: In the same way one proves that, if we denote with $C_{h}(\th)$
2555: the number of clusters on scale $h$, one has
2556: %
2557: $$ C_{h}(\th) \le \max\{ 0 , 2K(\th) 2^{(2-h)/\t} - 1 \} ;
2558: \Eq(4.22) $$
2559: %
2560: see \cita{GM} for details. \qed
2561:
2562: \*
2563:
2564: Note that the argument above is very close to \cita{GM}:
2565: this is due to the fact that the external lines of any self-energy graph
2566: $T$ are both $w$-lines, so that the only effect of the presence
2567: of the $v$-lines and of the nodes of $v$-type is in the
2568: contribution to $K(T)$.
2569:
2570: \*
2571:
2572: The following lemma deals with the lines on scale $h <h_{0}$.
2573:
2574: \*
2575:
2576: \0{\bf Lemma 6.} {\it Let $h_{0}$ be defined as in Lemma 2
2577: and $C_{0}<1/2$, and assume that there is a constant $C_{1}$
2578: such that one has $|\tilde\o_{m}-|m| |\le C_{1}\e$.
2579: If $\e$ is small enough for $h<h_{0}$ one has $|g^{(h)}_{\ell}|\le 32$.}
2580:
2581: \*
2582:
2583: \0{\it Proof.}
2584: Either if $h\neq h_{\ell}$ of $h=h_{\ell}=-1$ the bound is trivial.
2585: If $h=h_{\ell} \ge 0$ one has
2586: %
2587: $$ g^{(h)}_{\ell} = { \chi_{h}(|\o n_{\ell}|-\tilde\o_{m_{\ell}})
2588: \over -|\o n_{\ell}|+\tilde\o_{m} } \,
2589: {1 \over |\o n_{\ell}|+\tilde\o_{m} } ,
2590: \Eq(4.23) $$
2591: %
2592: where $|\o n_{\ell}|+\tilde\o_{m} \ge 1/2\e$ by Lemma 4.
2593: Then one has
2594: %
2595: $$ { 1 \over |\o n_{\ell}|+\tilde\o_{m} }\le 2 \e ,
2596: \Eq(4.24) $$
2597: %
2598: which, inserted in \equ(4.23), gives $|g^{(h)}_{\ell}| \le
2599: 2^{h+2} \e/C_{0} \le 32$, so that the lemma is proved. \qed
2600:
2601:
2602: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2603: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2604: \*\*
2605: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2606: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2607: \section(5,The renormalized expansion)
2608: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2609: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2610:
2611: \0It is an immediate consequence of Lemma 5 and Lemma 6
2612: that all the trees $\th$ with no self-energy graphs or
2613: $\n$-vertices admit a bound $O(C^k\e^k)$,
2614: where $C$ is a constant. However the generic tree
2615: $\th$ with
2616: $S_h(\th)\not=0$ admits a much worse bound, namely $O(C^k\e^k k!^\a)$,
2617: for some constant $\a$, and the presence of factorials
2618: prevent us to prove the convergence of the series;
2619: in KAM theory this is called {\it accumulation
2620: of small divisors}.
2621: It is convenient then to consider another expansion
2622: for $u_{n,m}$, which is essentially a resummation
2623: of the one introduced in Sections \secc(3) and \secc(4).
2624:
2625: We define the set $\Th^{(k)\RR}_{n,m}$
2626: of {\it renormalized trees}, which are defined
2627: as $\Th^{(k)}_{n,m}$
2628: except that the following rules are added.
2629:
2630: \*
2631:
2632: \0(9) To each self-energy graph (with $|m|\ge 1$) the
2633: $\RR=\openone-\LL$ operation is applied, where $\LL$ acts on
2634: the self-energy graphs in the following way,
2635: for $h\ge 0$ and $|m|\ge 1$,
2636: %
2637: $$ \LL \VV_T^{h}(\o n,m) =
2638: \VV_T^{h}({\rm sgn}(n)\,\tilde\o_{m},m) ,
2639: \Eq(5.1) $$
2640: %
2641: $\RR$ is called {\it regularization operator}; its action
2642: simply means that each self-energy graph
2643: $\VV_T^{h}(\o n,m)$ must be replaced by $\RR
2644: \VV_T^{h}(\o n,m)$.
2645:
2646: \*
2647:
2648: \0(10) To the nodes $\vvv$ of $w$-type with $s_{\vvvv}=1$
2649: (which we still call $\n$-vertices) and with
2650: $h\ge 0$ the minimal scale among the lines entering or exiting $\vvv$,
2651: we associate a factor $2^{-h} \n^{(c)}_{h,m}$, $c=a,b$,
2652: where $(n,m)$ and $(n,\pm m)$, with $|m|\ge 1$,
2653: are the momenta of the lines
2654: and $a$ corresponds to the sign $+$ and $b$ to the sign $-$ in $\pm m$.
2655:
2656: \*
2657:
2658: \0(11) The set $\{h_\ell\}$ of the scales
2659: associated to the lines $\ell\in L(\th)$ must
2660: satisfy the following constraint
2661: (which we call {\it compatibility}):
2662: fixed $(n_\ell,m_\ell)$ for any $\ell\in L(\th)$ and replaced
2663: $\RR$ with $\openone$ at each self-energy graph, one must have
2664: $\chi_{h_{\ell}}(|\o n_{\ell}|-\tilde\o_{m_{\ell}}) \neq 0$.
2665:
2666: \*
2667:
2668: \0(12) The factors $C^{(k_{\vvvv})}$
2669: in \equ(3.2) are replaced with $\G$, to be considered a parameter.
2670:
2671: \*
2672:
2673: The set $\Th^{(k)\RR}_{n,m}$ is defined as $\Th^{(k)}_{n,m}$ with
2674: the new rules and with the constrain that the order $k$ is given by
2675: $k=|V_{w}(\th)|+|V_{v}^{1}(\th)|$.
2676:
2677: \*
2678:
2679: We consider the following expansion
2680: %
2681: $$ \tilde u_{n,m} = \sum_{k=1}^{\io}\m^k
2682: \sum_{\th \in \Th^{(k)\RR}_{n,m}} \Val(\th) ,
2683: \Eq(5.4) $$
2684: %
2685: where, for $|m|\ge 1$ and $h\ge 0$, $\n^{(c)}_{h,m}$ is given by
2686: %
2687: $$ 2^{-h} \m \n^{(c)}_{h,m}=\m \n^{(c)}_{m} + {1 \over 2}
2688: \sum_{\s=\pm} \sum_{T \in \TT^{(c)}_{< h}} \m^{k_{T}}
2689: \VV_T^{h}(\s\tilde \o_{m},m) ,
2690: \Eq(5.3) $$
2691: %
2692: with $c=a,b$, and $\TT^{(c)}_{< h}$ denoting the set of
2693: self-energy graphs $T$ of type $c$
2694: (see item (6) in Section \secc(3)) with $h_{T}<h$,
2695: and $\G$ is determined by the self-consistence equation
2696: %
2697: $$ \G = r_{0}^{-1}
2698: \sum_{k=1}^{\io} \m^{k-1}
2699: {\mathop{\sum}_{\th \in \Th^{(k)\RR}_{n,n}}}^{*}
2700: a_{0,-n} \Val(\th) ,
2701: \Eq(5.3a) $$
2702: %
2703: with $*$ denoting the same constraint as in \equ(3.10).
2704:
2705: We shall set also $\n_{m}^{(c)}=\n_{-1,m}^{(c)}$.
2706: Note that $\VV_T^{h}(\s\tilde \o_{m},m)$
2707: is indipendent from $\s$.
2708:
2709: Calling $L_0(\th), V_0(\th), E_0(\th)$
2710: the set of lines, node and end-points
2711: not contained in any self-energy graph, and $S_{0}(\th)$
2712: the {\it maximal self-energy graphs}, \ie the self-energy
2713: graphs which are not contained in any self-energy graphs,
2714: we can write $\Val(\th)$ in \equ(5.4) as
2715: %
2716: $$ \Val(\th) =
2717: \Big( \prod_{\ell \in L_0(\th)} g_{\ell}^{(h_{\ell})} \Big)
2718: \Big( \prod_{\vvv \in V_0(\th)} \h_{{\vvvv}} \Big)
2719: \Big( \prod_{\vvv \in E_0(\th) } V_{\vvvv}\Big)
2720: \Big(\prod_{T\in S_0(\th)} \RR \VV^{h_T^{e}}_T(\o n_{\ell_T},m_{\ell_T})
2721: \Big) ,
2722: \Eq(5.4aa) $$
2723: %
2724: and by definition
2725: %
2726: $$ \RR \VV^{h_T^{e}}_T(\o n_{\ell_T},m_{\ell_T})=
2727: \VV^{h_T^{e}}_T(\o n_{\ell_T},m_{\ell_T})-
2728: \VV_T^{h_T^{e}}({\rm sgn}(n_{\ell_T})\,\tilde\o_{m_{\ell_T}},m_{\ell_T}) ,
2729: \Eq(5.1aa) $$
2730: %
2731: and $\VV^{h_T^{e}}_T(\o n_{\ell_T},m_{\ell_T})$
2732: is given by
2733: %
2734: $$\VV^{h_T^{e}}_T(\o n_{\ell_T},m_{\ell_T})=
2735: \Big( \prod_{\ell \in L_0(T)} g^{(h_{\ell})}_{\ell} \Big)
2736: \Big( \prod_{\vvv \in V_0(T)} \h_{\vvvv} \Big)
2737: \Big( \prod_{\vvv \in E_0(T) } V_{\vvvv} \Big)
2738: \Big(\prod_{T'\in S_0(T)} \RR \VV^{h_{T'}^{e}}_{T'}
2739: (\o n_{\ell_{T'}},m_{\ell_{T'}}) \Big) .
2740: \Eq(5.9aa) $$
2741:
2742: First (Lemma 7) we will show that the expansion \equ(5.4)
2743: is well defined, for $\n_{h,m}, \G=O(\e)$;
2744: then (Lemma 8 and 9) we show that under the same conditions also
2745: the r.h.s. of \equ(5.3) is well defined; moreover (Lemma 10)
2746: we prove by using \equ(5.3) that it is indeed
2747: possible to choose $\n_{m}^{(c)}$
2748: such that $\n_{h,m}=O(\e)$ for any $h$; then (Lemma 11 and 12)
2749: we show that \equ(5.4) admit a solution $\G=O(\e)$;
2750: finally (Lemma 13) we show that
2751: indeed \equ(5.4) solves the last of \equ(1.13) and \equ(2.2);
2752: this completes the proof of Proposition 1. In the next
2753: section we will solve the implicit function
2754: problem of \equ(2.1) so completing the proof of Theorem 1.
2755:
2756: We start from the following lemma stating that, if the
2757: $\n_{h,m}^{(c)}$ and $\G$ functions are bounded, then the expansion
2758: \equ(5.4) is well defined.
2759:
2760: \*
2761:
2762: \0{\bf Lemma 7.} {\it Assume that there exist a constant $C$ such that
2763: one has $|\G|\le C\e $ and $|\n_{h,m}^{(c)}| \le C\e$, with $c=a,b$,
2764: for all $|m|\ge 1$ and all $h\ge 0$.
2765: Then for all $\m_0>0$ there exists $\e_0>0$
2766: such that for all $|\m|\le\m_0$ and
2767: for all $0<\e< \e_{0}$ and for all $(n,m)\in \ZZZ^{2}$ one has
2768: %
2769: $$ \left| \tilde u_{n,m} \right| \le D_{0} \e \m \,
2770: e^{-\k(|n|+|m|)/4} ,
2771: \Eq(5.6) $$
2772: %
2773: where $D_{0}$ is a positive constant.
2774: Moreover $u_{n,m}$ is analytic in $\m$ and in
2775: the parameters $\n_{m',h'}^{(c)}$, with $c=a,b$ and $|m'| \ge 1$.}
2776:
2777: \*
2778:
2779: \0{\it Proof.} In order to take into account the $\RR$
2780: operation we write \equ(5.1aa) as
2781: %
2782: $$\RR \VV^{h_T^{e}}_T(\o n_{\ell_T},m_{\ell_T})=
2783: \left(\o n_{\ell_T}-\tilde\o_{m_{\ell_T}} \right)
2784: \int_{0}^{1} {\rm d}t \dpr
2785: \VV^{h_T^{e}}_T(\o n_{\ell_T}+t(\o n_{\ell_T}-\tilde\o_{m_{\ell_T}}) ,
2786: m_{\ell_T}) ,
2787: \Eq(5.1aaa) $$
2788: %
2789: where $\dpr$ denotes the derivative with respect to the argument
2790: $\o n_{\ell_T}+t(\o n_{\ell_T}-\tilde\o_{m_{\ell_T}})$.
2791:
2792: By \equ(5.9aa) we see that the derivatives can be applied
2793: either on the propagators in $L_0(T)$, or
2794: on the $\RR \VV^{h_{T'}^{e}}_{T'}$.
2795: In the first case there is an extra factor
2796: $2^{-h^{(e)}_T+h_{T}}$ with respect to the bound
2797: \equ(4.11): $2^{-h^{(e)}_T}$ is obtained from
2798: $\o n_{\ell_T}-\tilde\o_{m_{\ell_T}}$
2799: while $\partial g^{(h_T)}$ is bounded proportionally to $2^{2 h_T}$;
2800: in the second case note that $\dpr_t \RR \VV^{h_{T'}^{e}}_{T'}=
2801: \dpr_t \VV^{h_{T'}^{e}}_{T'}$
2802: as $\LL\VV_{T'}^{h^{(e)}_{T'}}$ is independent of
2803: $t$; if the derivative acts on the propagator
2804: of a line $\ell\in L(T)$, we get a gain factor
2805: %
2806: $$ 2^{-h^{(e)}_T+h_{T'}} \le 2^{-h^{(e)}_T+h_{T}}
2807: 2^{-h^{(e)}_{T'}+h_{T'}} ,
2808: \Eq(5.7) $$
2809: %
2810: as $h^{(e)}_{T'}\le h_{T}$. We can iterate
2811: this procedure until all the $\RR$
2812: operations are applied on propagators;
2813: at the end (i) the propagators are derived at most one time;
2814: (ii) the number of terms so generated is $\le k$;
2815: (iii) to each self-energy graph $T$ a factor
2816: $2^{-h^{(e)}_T+h_{T}}$ is associated.
2817:
2818: Assuming that $|\n^{(c)}_{h,m}|\le C\e$ and $|\G|\le C\e$,
2819: for any $\th$ one obtains,
2820: for suitable constants $\overline{D}$
2821: %
2822: $$ \eqalign{
2823: & \left| \Val(\th) \right| \le
2824: \e^{|V_w(\th)|+|V_v^{(1)}(\th)|}
2825: \overline{D}^{|V(\th)|} \cr
2826: & \qquad \qquad
2827: \Big( \prod_{h=h_0}^{\io} \exp \Big[ h \log 2 \Big( 4 K(\th)
2828: 2^{-(h-2)/\t} - C_{h}(\th) + S_{h}(\th) +
2829: M_{h}^{\n}(\th) \Big) \Big] \Big) \cr
2830: & \qquad \qquad
2831: \Big( \prod_{T \in \SS(\th)\atop h^{(e)}_T \ge h_0 }
2832: 2^{-h^{(e)}_T+h_T} \Big)
2833: \Big( \prod_{h=h_{0}}^{\io} 2^{-h M_{h}^{\n}(\th)} \Big)
2834: \cr & \qquad \qquad
2835: \Big( \prod_{\vvv\in V(\th) \cup E(\th)}
2836: e^{-\k(|n_{\vvvv}|+|n_{\vvvv}'|)} \Big)
2837: \Big( \prod_{\vvv\in V(\th) \cup E(\th)}
2838: e^{-\k(|m_{\vvvv}|+|m_{\vvvv}'|)} \Big) , \cr}
2839: \Eq(5.8) $$
2840: %
2841: where the second line
2842: is a bound for $\prod_{h\ge h_0} 2^{h N_h(\th)}$
2843: and we have used that by item (12) $N_h(\th)$ can be
2844: bounded through Lemma 5), and Lemma 4 has been used
2845: for the lines on scales $h< h_{0}$;
2846: moreover $ \prod_{h=h_{0}}^\io 2^{-h M_\n^h(\th)}$
2847: takes into account the factors $2^{-h}$ arising from
2848: the running coupling constants $\n^{(c)}_{h,m}$
2849: and the action of $\RR$ produces, as discussed above,
2850: the factor $\prod_{T \in \SS(\th)} 2^{-h^{(e)}_T+h_T}$.
2851: Then one has
2852: %
2853: $$ \eqalign{
2854: & \Big( \prod_{h=h_0}^{\io} 2^{h S_{h}(\th)} \Big)
2855: \Big( \prod_{T \in \SS(\th) } 2^{-h^{(e)}_T} \Big) = 1 , \cr
2856: %
2857: & \Big( \prod_{h=h_0}^{\io} 2^{-h C_{h}(\th)} \Big)
2858: \Big(\prod_{T \in \SS(\th)}2^{h_T} \Big) \le 1 . \cr}
2859: \Eq(5.11) $$
2860: %
2861: We have to sum the values of all trees,
2862: so we have to worry about the sum of the labels.
2863: Recall that a labeled tree is obtained from an unlabeled tree
2864: by assigning all the labels to the points and the lines:
2865: so the sum over all possible labeled trees can be written
2866: as sum over all unlabeled trees and of labels.
2867: For a fixed unlabeled tree $\th$ with a given number of nodes,
2868: say $N$, we can assign first the mode labels
2869: $\{(n_{\vvvv}',m_{\vvvv}'),
2870: (n_{\vvvv},m_{\vvvv})\}_{v\in V(\th)\cup E(\th)}$,
2871: and we sum over all the other labels, which
2872: gives $4^{|V_{v}(\th)|}$ (for the labels $j_{\vvvv}$)
2873: times $2^{|L(\th)|}$ (for the scale labels):
2874: then all the other labels are uniquely fixed.
2875: Then we can perform the sum over the mode labels by
2876: using the exponential decay arising from the node factors
2877: \equ(3.2) and end-point factors \equ(3.4).
2878: Finally we have to sum over the unlabeled trees, and this
2879: gives a factor $4^{N}$ \cita{HP}. By Lemma 3,
2880: one has $|V(\th)|= |V_w(\th)|+|V_{v}^{(1)}(\th)|+|V_{v}^{(3)}(\th)|\le
2881: 3(|V_{w}^{(3)}(\th)|+|V_v^{(1)}(\th)|)$, hence $N \le 3k$,
2882: so that $\sum_{\th\in\Th^{(k)\RR}_{n,m}} |\Val(\th)|\le D^k\e^{k}$,
2883: for some positive constant $D$.
2884:
2885: Therefore, for fixed $(n,m)$ one has
2886: %
2887: $$ \sum_{k=1}^{\io} \sum_{\th\in\Th^{(k)}_{n,m}}\m^k
2888: \left| \Val(\th) \right| \le
2889: D_{0} \m\e \, e^{-\k(|n|+|m|)/4} ,
2890: \Eq(5.13) $$
2891: %
2892: for some positive constant $D_{0}$,
2893: so that \equ(5.6) is proved. \qed
2894:
2895: \*
2896:
2897: From \equ(5.3) we know that
2898: the quantities $\n_{h,m}^{(c)}$, for $h\ge0$ and $|m|\ge 1$,
2899: verify the recursive relations
2900: %
2901: $$ \m \n^{(c)}_{h+1,m} = 2 \m \n^{(c)}_{h,m} + \b^{(c)}_{h,m}
2902: (\tilde\o,\e,\{\n_{h',m'}^{(c')}\}) ,
2903: \Eq(5.14) $$
2904: %
2905: where, by defining $\TT^{(c)}_h$ as the set of self-energy graphs
2906: in $\TT^{(c)}_{<h+1}$ which are on scale $h$, the {\it beta function}
2907: %
2908: $$ \b^{(c)}_{h,m} \=
2909: \b^{(c)}_{h,m}(\tilde\o,\e,\{\n_{h',m'}^{(c')}\}) = 2^{h+1}
2910: {1 \over 2} \sum_{\s=\pm} \sum_{T\in\TT_h^{(c)}} \m^{k_{T}}
2911: \VV_T^{h+1}(\s\tilde\o_{m},m) ,
2912: \Eq(5.15) $$
2913: %
2914: depends only on the scales $h'\le h$.
2915:
2916: In order to obtain a bound on the beta function, hence
2917: on the running coupling constants, we need to bound
2918: $\VV_{T}^{h+1}(\pm \tilde\o_{m},m)$ for $T\in\TT_h^{(c)}$.
2919: We define $\wt \Theta^{(k)\RR}_{n,m}$
2920: as the set $\Theta^{(k)\RR}_{n,m}$
2921: introduced before,
2922: but by changing item (7) into the following one:
2923:
2924: \*
2925:
2926: \0(7$^{\prime}$) We divide the set $\wt E(\th)$
2927: of end-points into two sets $E(\th)$ and $E_{0}(\th)$.
2928: To each end-point $\vvv \in E(\th)$ we associate a first mode
2929: label $(n_{\vvvv}',m_{\vvvv}')$, with $|m_{\vvvv}'|=|n_{\vvvv}'|$,
2930: a second mode label $(0,0)$
2931: and an {\it end-point factor} $ V_{\vvvv} =
2932: (-1)^{1+\d_{n_{\vvvv}',m_{\vvvv}'}} a_{0,n_{\vvvv}'}$,
2933: while $E_{0}(\th)$ is either the empty set or
2934: a single end-point $\vvvv_{0}$, and, in the latter case,
2935: to the end-point $\vvv \in E_{0}(\th)$
2936: we associate a first mode
2937: label $(\lis\o_{m_{\vvvv}},n_{\vvvv})$,
2938: where $\lis\o_{m_{\vvvv}}=\tilde\o_{m_{\vvvv}}/\o$,
2939: a second mode label $(0,0)$
2940: and an {\it end-point factor} $V_{\vvvv} =1$.
2941:
2942: \*
2943:
2944: Then we have the following generalization of Lemma 4.
2945:
2946: \*
2947:
2948: \0{\bf Lemma 8.} {\it
2949: If $\e$ is small enough for any tree $\th \in \wt\Th_{n,m}^{(k)\RR}$
2950: one has
2951: %
2952: $$ N_{h}(\th) \le 4 K(\th) 2^{(2-h)/\t} -
2953: C_{h}(\th) + S_{h}(\th) + M_{h}^{\n}(\th) ,
2954: \Eq(5.20) $$
2955: %
2956: where the notations are as in lemma 4.}
2957:
2958: \*
2959:
2960: \0{\it Proof.}
2961: Lemma 4 holds for $E_0(\th)=0$;
2962: we mimic the proof of Lemma 4 proving that
2963: %
2964: $$ N^{*}_{h}(\th) \le \max\{ 0 , 2 K(\th) 2^{(2-h)/\t} \} ,
2965: \Eq(5.22) $$
2966: %
2967: for all trees $\th$ with $E_{0}(\th) \neq \emptyset$,
2968: again by induction on $K(\th)$.
2969:
2970: For any line $\ell \in L(\th)$ set $\h_{\ell}=1$ if the line
2971: is along the path connecting $\vvvv_{0}$ to the root and
2972: $\h_{\ell}=0$ otherwise, and write
2973: %
2974: $$ n_{\ell} = n_{\ell}^{0} + \h_{\ell} \lis\o_{m} ,
2975: \qquad m_{\ell} = m_{\ell}^{0} + \h_{\ell} m ,
2976: \Eq(5.23) $$
2977: %
2978: which implicitly defines $n_{\ell}^{0}$ and $m_{\ell}^{0}$.
2979:
2980: Define $k_{0}=2^{(h-1)/\t}$. One has $N_{h}^{*}(\th)=0$ for
2981: $K(\th)<k_{0}$, because if a line $\bar\ell\in L(\th)$ is indeed on scale
2982: $h$ then $|\o n_{\bar\ell}-\tilde \o_{m_{\bar\ell}}|<C_{0}2^{1-h}$,
2983: so that \equ(5.23) and the Diophantine conditions imply
2984: %
2985: $$ K(\th) \ge \left| n_{\bar\ell}^{0} \right|
2986: > 2^{(h-1)/\t} \= k_{0} .
2987: \Eq(5.24) $$
2988: %
2989: Then, for $K \ge k_{0}$, we assume that the bound \equ(5.22)
2990: holds for all $K(\th)=K'<K$,
2991: and we show that it follows also for $K(\th)=K$.
2992:
2993: If the root line $\ell$ of $\th$ is either on scale $<h$
2994: or on scale $\ge h$ and resonant,
2995: the bound \equ(5.22) follows immediately from the bound \equ(4.13)
2996: and from the inductive hypothesis.
2997:
2998: The same occurs if the root line is on scale $\ge h$ and non-resonant,
2999: and, by calling $\ell_{1},\ldots,\ell_{m}$ the lines on scale
3000: $\ge h$ which are the closest to $\ell$, one has $m \ge 2$:
3001: in fact in such a case at least $m-1$ among the subtrees
3002: $\th_{1},\ldots,\th_{m}$ having $\ell_{1},\ldots,\ell_{m}$,
3003: respectively, as root lines have $E(\th_{i})=\emptyset$, so that
3004: we can write, by \equ(4.13) and by the inductive hypothesis,
3005: %
3006: $$ N_{h}^{*}(\th) = 1 + \sum_{i=1}^{m} N_{h}^{*}(\th_{i})
3007: \le 1 + E_{h}^{-1} \sum_{i=1}^{m} K(\th_{i}) - (m-1)
3008: \le E_{h} K(\th) ,
3009: \Eq(5.24a) $$
3010: %
3011: so that \equ(5.22) follows.
3012:
3013: If $m=0$ then $N_{h}^{*}(\th)=1$ and $K(\th) 2^{(2-h)/\t} \ge 1$
3014: because one must have $K(\th) \ge k_{0}$.
3015:
3016: So the only non-trivial case is when
3017: one has $m=1$. If this happens $\ell_{1}$ is, by construction,
3018: the root line of a tree $\th_{1}$ such that
3019: $K(\th) = K(T) + K(\th_{1})$, where $T$ is the cluster
3020: which has $\ell$ and $\ell_{1}$ as external lines and
3021: $K(T)$, defined in \equ(4.7), satisfies the bound
3022: $K(T) \ge |n_{\ell_{1}} - n_{\ell}|$.
3023:
3024: Moreover, if $E_{0}(\th_{1})\neq \emptyset$, one has
3025: %
3026: $$ \eqalign{
3027: & \left| |\o n_{\ell}^{0}+
3028: \tilde\o_{m}| - \tilde\o_{m_{\ell}} \right|
3029: \le 2^{-h+1} C_{0} , \cr
3030: & \left| |\o n_{\ell_{1}}^{0} +
3031: \tilde\o_{m}| - \tilde\o_{m_{\ell_{1}}} \right|
3032: \le 2^{-h+1} C_{0} , \cr}
3033: \Eq(5.26a) $$
3034: %
3035: so that, for suitable $\h_{\ell},\h_{\ell_{1}}\in \{-,+\}$, we obtain
3036: %
3037: $$ 2^{-h+2} C_{0} \ge \big| \o
3038: ( n^{0}_{\ell}- n^{0}_{\ell_{1}}) +\h_{\ell} \tilde\o_{m_{\ell}}+
3039: \h_{\ell_{1}} \tilde\o_{m_{\ell_{1}}} \big|
3040: \ge C_{0} |n^{0}_{\ell}-n^{0}_{\ell_{1}}|^{-\t} \=
3041: C_{0} |n_{\ell}-n_{\ell_{1}}|^{-\t} ,
3042: \Eq(5.27) $$
3043: %
3044: by the second Diophantine conditions in \equ(2.31),
3045: as the quantities $\tilde \o_{m}$
3046: appearing in \equ(5.26a) cancel out.
3047: Therefore one obtains by the inductive hypothesis
3048: %
3049: $$ N_{h}^{*}(\th) \le 1 + K(\th_{1}) E_{h}^{-1}
3050: \le 1 + K(\th) E_{h}^{-1} -
3051: K(T) E_{h}^{-1} \le K(\th) E_{h}^{-1} ,
3052: \Eq(5.28) $$
3053: %
3054: hence the first bound in \equ(5.22) is proved.
3055:
3056: If $E_{0}(\th_{1})= \emptyset$, one has
3057: %
3058: $$ N_{h}^{*}(\th) \le 1 + K(\th_{1}) E_{h}^{-1} - 1
3059: \le 1 + K(\th) E_{h}^{-1} - 1 \le K(\th) E_{h}^{-1} ,
3060: \Eq(5.28a) $$
3061: %
3062: and \equ(5.22) follows also in such a case. \qed
3063:
3064: \*
3065:
3066: The following bound for $\VV_T^{h+1}(\pm\tilde\o_{m},m)$,
3067: $h \ge h_{0}$, can then be obtained.
3068:
3069: \*
3070:
3071: \0{\bf Lemma 9.} {\it
3072: Assume that there exist a constant $C$ such that
3073: one has $|\G|\le C\e $ and $|\n_{h,m}^{(c)}| \le C\e$, with $c=a,b$,
3074: for all $|m|\ge 1$ and all $h\ge 0$.
3075: Then if $\e$ is small enough
3076: for all $h\ge 0$ and for all $T\in\TT_{h}^{(c)}$ one has
3077: %
3078: $$|\VV_T^{h+1}(\pm\tilde\o_{m},m)|\le
3079: B^{|V(T)|} e^{-\k 2^{(h-1)/\t}/4} e^{-\k K(T)/4}
3080: \e^{|V_v^{(1)}(T)|+|V_w(T)|} ,
3081: \Eq(5.17) $$
3082: %
3083: where $B$ is a constant and $K(T)$ is defined in \equ(4.7).}
3084:
3085: \*
3086:
3087: \0{\it Proof.}
3088: By using lemma 7
3089: we obtain for all $T\in\TT_{h}^{(c)}$ and
3090: assuming $h\ge h_0$ we get the bound
3091: %
3092: $$ \eqalignno{
3093: & \left| \VV_{T}^{h+1}(\pm\tilde\o_{m},m) \right|
3094: \le \lis B^{|V(T)|} \e^{|V_v^{(1)}(T)|+|V_w(T)|} \cr
3095: & \qquad \qquad \qquad
3096: \prod_{h'=h_0}^{h} \exp \Big[ 4 K(T) \log 2 h' 2^{(2-h')/\t} -
3097: C_{h'}(T) + S_{h'}(T) + M_{h'}^{\n}(T) \Big] \cr
3098: & \qquad \qquad \qquad
3099: \Big( \prod_{T'\subset T\atop h_T^{(\e)}\ge h_0} 2^{-h^{(e)}_T+h_T} \Big)
3100: \Big( \prod_{h'=h_0}^{h} 2^{-h' M_{h'}^{\n}(T)} \Big) e^{-\k|K(T)|/2} ,
3101: & \eq(5.36a) \cr} $$
3102: %
3103: where $\lis B$ is a suitable constant.
3104: If $h< h_0$ the bound trivializes as the r.h.s. reduces simply
3105: to $C^{|V(T)|} |\e|^{|V_v^{(1)}(T)|+|V_w(T)|}
3106: e^{-{\k\over 2}|K(T)|}$.
3107: The main difference with
3108: respect to Lemma 6 is that, given
3109: a self-energy graph $T\in\TT_{h}^{(c)}$, there is at least a
3110: line $\ell\in L(T)$ on scale $h_{\ell}=h$ and with
3111: propagator
3112: %
3113: $$ {1 \over - \o^{2} (n^{0}_{\ell} + \h_{\ell}
3114: \lis \o_{m} )^{2} + \tilde \o^{2}_{m^{0}_{\ell} + \h_{\ell} m} } ,
3115: \Eq(5.37) $$
3116: %
3117: where $\h_{\ell}=1$ if the line $\ell$ belongs to the path of
3118: lines connecting the entering line (carrying a momentum $(n,m)$)
3119: of $T$ with the line coming out of $T$, and $\h_{\ell}=0$ otherwise.
3120: Then one has by the Mel'nikov conditions
3121: %
3122: $$ C_{0} |n^{0}_{\ell}|^{-\t}\le \left| \o n^{0}_{\ell}
3123: + \h_{\ell} \tilde \o_m \pm \tilde \o_{m^{0}_{\ell} +\h_{\ell} m} \right|
3124: \le C_{0} 2^{-h+1} ,
3125: \Eq(5.38) $$
3126: %
3127: so that $|n^{0}_{\ell}|\ge 2^{(h-1)/\t}$.
3128: On the other hand one has $|n^{0}_{\ell}| \le K(T)$,
3129: hence $K(T)\ge 2^{(h-1)/\t}$; so we get the bound \equ(5.17).\qed
3130:
3131: \*
3132:
3133: It is an immediate consequence of the above Lemma
3134: that for all $\m_0>0$ there exists $\e_0>0$
3135: such that for all $|\m|\le\m_0$ and $0<\e<\e_0$ one has
3136: $|\b^{(c)}_{h,m}|\le B_1\e|\m|$, with $B_1$
3137: a suitable constant.
3138:
3139: We have then proved convergence {\it assuming that}
3140: the parameters $\n_{h,m}$ and $\G$ are bounded; we have
3141: to show that this is actually the case, if the $\n_m^{(c)}$
3142: in \equ(2.32) are chosen in a proper way.
3143:
3144: We start proving that it is possible to choose
3145: $\n^{(c)}=\{\n^{(c)}_{m}\}_{|m| \ge 1}$ such that,
3146: for a suitable positive constant $C$, one has
3147: $|\n^{(c)}_{h,m}|\le C\e$ for all $h\ge 0$ and for all $|m|\ge 1$.
3148:
3149: For any sequence $a\=\{a_{m}\}_{|m|\ge 1}$ we introduce the norm
3150: %
3151: $$ \left\| a \right\|_{\io} =\sup_{|m|\ge 1} |a_{m}| .
3152: \Eq(5.20a) $$
3153: %
3154: Then we have the following result.
3155:
3156: \*
3157:
3158: \0{\bf Lemma 10.}
3159: {\it Assume that there exists a constant $C$ such that $|\G|\le C\e$.
3160: Then for all $\m_0>0$ there exists $\e_0>$
3161: such that, for all $|\m|\le\m_0$ and
3162: for all $0< \e <\e_{0}$
3163: there is
3164: a family of intervals $I^{(\bar h)}_{c,m}$, $\bar h\ge 0$, $|m|\ge 1$,
3165: $c=a,b$, such that $I^{(\bar h+1)}_{c,m}\subset I^{(\bar h)}_{c,m}$,
3166: $|I^{\bar h}_{c,m}|\le 2\e (\sqrt{2})^{-(\bar h+1)}$ and,
3167: if $\n^{(c)}_{m}\in I^{(\bar h)}_{c,m}$, then
3168: %
3169: $$ \| \n^{(c)}_{h} \|_{\io}\le D \e, \qquad \bar h\ge h\ge 0 ,
3170: \Eq(5.21) $$
3171: %
3172: for some positive constant $D$. Finally one has
3173: $\n_{h,-m}^{(c)}=\n_{h,m}^{(c)}$, $c=a,b$,
3174: for all $\bar h \ge h\ge 0$ and for all $|m|\ge 1$.
3175: Therefore one has $\|\n_{h}\|_{\io}\le C\e$ for all $h \ge 0$,
3176: for some positive constant $D$;
3177: in particular $|\n_{m}|\le D\e$ for all $m\ge 1$.}
3178:
3179: \*
3180:
3181: \0{\it Proof.}
3182: \0The proof is done by induction on $\bar h$.
3183: Let us define $J^{(h)}_{c,m}=[-\e,\e]$ and
3184: call $J^{(h)}=\times_{|m|\ge 1,c=a,b} J^{(h)}_{c,m}$
3185: and $I^{(h)}=\times_{|m|\ge 1,c=a,b} I^{(h)}_{c,m}$.
3186:
3187: We suppose that there exists $I^{(\bar h)}$ such that,
3188: if $\n$ spans $I^{(\bar h)}$ then $\n_{\bar h}$
3189: spans $J^{(\bar h)}$ and $|\n^{(c)}_{h,m}|\le D\e$ for
3190: $\bar h\ge h\ge 0$; we want to show that the same holds for $\bar h+1$.
3191: Let us call $\tilde J^{(\bar h+1)}$ the interval spanned by
3192: $\{\n^{(c)}_{\bar h+1,m}\}_{|m|\ge 1,c=a,b}$
3193: when $\{\n^{(c)}_{m}\}_{|m|\ge 1,c=a,b}$ span $I^{(\bar h)}$.
3194: For any $\{\n^{(c)}_{m}\}_{|m|\ge 1,c=a,b}\in I^{(\bar h)}$ one has
3195: $\{\n_{\bar h+1,m}\}_{|m|\ge 1,c=a,b} \in [-2\e-D\e^2, 2\e+D\e^{2}]$,
3196: where the bound \equ(5.17) has been used.
3197: This means that $J^{(\bar h+1)}$
3198: is strictly contained in $\tilde J^{(\bar h+1)}$.
3199: On the other hand it is obvious that there
3200: is a one-to-one correspondence between
3201: $\{\n^{(c)}_{m}\}_{|m|>1,c=a,b}$
3202: and the sequence $\{\n^{(c)}_{h,m}\}_{|m|\ge 1,c=a,b}$,
3203: $\bar h+1\ge h\ge 0$. Hence there is a set
3204: $I^{(\bar h+1)}\subset I^{(\bar h)}$
3205: such that, if $\{\n^{(c)}_m\}_{|m|\ge 1,c=a,b}$
3206: spans $I^{(\bar h+1)}$, then $\{\n^{(c)}_{\bar h+1,m}\}_{|m|\ge 1,c=a,b}$
3207: spans the interval $J^{(\bar h)}$ and, for $\e$ small enough,
3208: $|\n_{h}|_{\io}\le C\e$ for $\bar h+1\ge h\ge 0$.
3209:
3210: The previous computations also
3211: show that the inductive hypothesis is verified also for $\bar h=0$
3212: so that we have proved that there exists a decreasing sets
3213: of intervals $I^{(\bar h)}$ such that if
3214:
3215: $\{\n^{(c)}_{m}\}_{|m|>1,c=a,b} \in I^{(\bar h)}$
3216: then the sequence $\{\n^{(c)}_{h,m}\}_{|m|\ge 1,c=a,b}$
3217: is well defined for $h\le\bar h$ and it verifies
3218: $|\n_{h,m}^{(c)}|\le C|\e|$. In order to prove the bound on the size of
3219: $I^{(\bar h)}_{c,m}$ let us denote by $\{\n^{(c)}_{h,m}\}_{|m|\ge 1,c=a,b}$
3220: and $\{\n^{\prime(c)}_{h,m}\}_{|m|\ge 1,c=a,b}$,
3221: $0\le h\le \bar h$, the sequences corresponding to
3222: $\{\n^{(c)}_{m}\}_{|m|\ge 1,c=a,b}$ and
3223:
3224: $\{ \n^{\prime(c)}_{m} \}_{|m|\ge 1,c=a,b}$ in
3225: $I^{(\bar h)}$. We have
3226: %
3227: $$ \m\n^{(c)}_{h+1,m} - \m\n^{\prime(c)}_{h+1,m}=
3228: 2 \left( \m\n^{(c)}_{h,m}-\m\n^{\prime(c)}_{h,m} \right) +
3229: \b^{(c)}_{h,m} - \b^{\prime(c)}_{h,m} ,
3230: \Eq(5.64a)$$
3231: %
3232: where $\b^{(c)}_{h,m}$ and $\b^{\prime(c)}_{h,m}$ are
3233: shorthands for the beta functions. Then, as
3234: $|\n_k-\n'_k|_{\io} \le |\n_h-\n'_h|_{\io}$ for all $k\le h$, we have
3235: %
3236: $$ | \n_h-\n'_h |_{\io} \le {1 \over 2} |\n_{h+1}-\n'_{h+1} |_{\io}
3237: + D \e^{2} |\n_h-\n'_h |_{\io} .
3238: \Eq(5.65) $$
3239: %
3240: Hence if $\e$ is small enough then one has
3241: %
3242: $$ \|\n-\n'\|_{\io} \le (\sqrt{2})^{-(\bar h+1)}
3243: \|\n_{\bar h}-\n'_{\bar h}\|_{\io} .
3244: \Eq(5.66a) $$
3245: %
3246: Since, by definition, if $\n$ spans $I^{(\bar h)}$,
3247: then $\n_{\bar h}$ spans the interval $J^{(\bar h)}$,
3248: of size $2 |\e|$, the size of $I^{(\bar h)}$ is
3249: bounded by $2|\e| (\sqrt{2})^{(-\bar h-1)}$.
3250:
3251: Finally note that one can choose
3252: $\n^{(c)}_{m}=\n^{(c)}_{-m}$ and then
3253: $\n^{(c)}_{h,m}=\n^{(c)}_{h,-m}$ for any $|m|\ge 1$ and any
3254: $\bar h\ge h\ge 0$; this follows from the fact
3255: that the function $\b_{k,m}^{(c)}$ in \equ(5.15) is even
3256: under the exchange $m\to-m$; it depends on $m$ through $\tilde \o_m$
3257: (which is an even function of $m$), through the end-points
3258: $v\in E(\th)$ (which are odd under the exchange $m\to-m$; but their
3259: number must be even) and finally through
3260: $\nu^{(q-1)}_{k,m}$ which are assumed inductively to be even.\qed
3261: \*
3262: It will be useful to explicitly
3263: construct the $\n^{(c)}_{h,m}$ by a contraction method.
3264: By iterating \equ(5.14) we find, for $|m|\ge 1$
3265: %
3266: $$ \m\n^{(c)}_{h,m}=2^{h+1} \left( \m\n^{(c)}_m+\sum_{k=-1}^{h-1}
3267: 2^{-k-2} \b^{(c)}_{k,m}(\tilde\o,\e,\{ \n^{(c')}_{k',m'} \} ) \right) ,
3268: \Eq(5.22a)$$
3269: %
3270: where $\b^{(c)}_{k,m}(\tilde\o,\e,\{ \n^{(c')}_{k',m'} \} )$ depends on
3271: $\n_{k',m'}$ with $k'\le k-1$. If we put $h=\bar h$
3272: in \equ(5.22a) we get
3273: %
3274: $$ \m\n_{m}^{(c)}=-\sum_{k=-1}^{\bar h-1}
3275: 2^{-k-2} \b^{(c)}_{k,m}(\tilde\o,\e,\{ \n^{(c')}_{k',m'} \} )
3276: +2^{-\bar h-1}\m\n_{\bar h,m}^{(c)}
3277: \Eq(5.23a) $$
3278: %
3279: and, combining \equ(5.22a) with \equ(5.23a), we find,
3280: for $\bar h> h\ge 0$,
3281: %
3282: $$ \m\n^{(c)}_{h,m}=-2^{h+1} \left( \sum_{k=h}^{\bar h-1}
3283: 2^{-k-2} \b^{(c)}_{k,m}(\tilde\o,\e, \{ \n^{(c')}_{k',m'} \})
3284: \right) + 2^{h-\bar h} \m\n_{\bar h,m}^{(c)} .
3285: \Eq(5.24b) $$
3286: %
3287: The sequences $\{\n^{(c)}_{h,m}\}_{|m|>1}$, $\bar h> h\ge h_{0}$,
3288: parametrized by $\{\n^{(c)}_{\bar h,m}\}_{|m| \ge 2}$
3289: such that $\|\n^{(c)}_{\bar h}\|_\io\le C\e$,
3290: can be obtained as the limit as $q\to\io$
3291: of the sequences $\{\n^{(c)(q)}_{h,m}\}$, $q\ge 0$,
3292: defined recursively as
3293: %
3294: $$ \eqalign{
3295: \m\n_{h,m}^{(c)(0)} & = 0 , \cr
3296: \m\n_{h,m}^{(c)(q)} & = - 2^{h+1}\left(\sum_{k=h}^{\bar h-1}
3297: 2^{-k-2} \b^{(c)}_{k,m}(\tilde\o,\e,
3298: \{\n^{(c')(q-1)}_{k',m'} \})\right)+2^{h-\bar h} \m\n_{\bar h,m}^{(c)} . \cr}
3299: \Eq(5.25) $$
3300: %
3301: In fact, it is easy to show inductively that, if
3302: $\e$ is small enough, $\|\n_h^{(q)}\|_{\io}\le C\e$,
3303: so that \equ(5.25) is meaningful, and
3304: %
3305: $$ \max_{0\le h\le \bar h} \|\n_h^{(q)}-\n_h^{(q-1)} \|_{\io}
3306: \le (C\e)^{q} .
3307: \Eq(5.26) $$
3308: %
3309: For $q=1$ this is true as $\n_h^{(c)(0)}=0$;
3310: for $q>1$ it follows by the fact that
3311: $\b^{(c)}_k(\tilde\o,\e, \{\n^{(c')(q-1)}_{k',m'} \})-
3312: \b^{(c)}_k(\tilde\o,\e, \{\n^{(c')(q-2)}_{k',m'} \})$
3313: can be written as a sum of terms
3314: in which there are at least one $\n$-vertex,
3315: with a difference $\n_{h'}^{(c')(q-1)}-\n_{h'}^{(c')(q-2)}$,
3316: with $h'\ge k$, in place of the corresponding
3317: $\nu_{h'}^{(c')}$, and one node carrying an $\e$.
3318: Then $\n_h^{(q)}$ converges as $q\to\io$,
3319: for $\bar h<h\le 1$, to a limit $\n_h$, satisfying
3320: the bound $\|\n_h\|_{\io}\le C\e$.
3321: Since the solution is unique, it must coincide with one in
3322: Lemma 10.
3323:
3324: \*
3325:
3326: We have then constructed a sequence of $\n_{h,m}^{(c)}$
3327: solving \equ(5.24b) for any $\bar h>1$ and any $\n_{\bar h,m}^{(c)}$;
3328: we shall call $\n_{h,m}^{(c)}(\G)$ the solution of \equ(5.24b)
3329: with $\bar h=\io$ and $\n^{(c)}_{\io,m}=0$,
3330: to stress the dependence on $\G$.
3331:
3332: We will prove the following Lemma.
3333:
3334: \*
3335:
3336: \0{\bf Lemma 11.} {\it Under the
3337: the same conditions of Lemma 10
3338: it holds that for any $h\ge 0$
3339: %
3340: $$ \|\nu_h(\G^1)-\nu_h(\G^2)\|_{\io}\le D \e |\G^1-\G^2| ,
3341: \Eq(5.100)$$
3342: %
3343: for a suitable constant $D$.}
3344:
3345: \*
3346:
3347: \0{\it Proof.} Calling
3348: $\n_h^{(q)}(\G)$ the l.h.s. of \equ(5.25)
3349: with $\bar h=\io$ and $\n_{\io,m}=0$, we can
3350: show by induction on $q$ that
3351: %
3352: $$ \|\nu_h^{(q)}(\G^1)-\nu_h^{(q)}(\G^2)\|_{\io}\le D \e |\G^1-\G^2| ,
3353: \Eq(5.101) $$
3354: %
3355: We find convenient to write explicitly the dependence
3356: of the function $\b^{(c)}_{h,m}$ from the parameter
3357: $\G$, so that we rewrite
3358: $\b^{(c)}_{k,m}(\tilde\o,\e,
3359: \{\n^{(c')(q-1)}_{k',m'} \}$
3360: in the r.h.s. of \equ(5.25) as
3361: $\b^{(c)}_{k,m}(\tilde\o,\e,\G,
3362: \{\n^{(c')(q-1)}_{k',m'}(\G) \}$.
3363: %
3364: Then from \equ(5.25) we get
3365: %
3366: $$\eqalign{
3367: & \m\n_{m}^{(c)(q)}(\G^1)-\m\n_{m}^{(c)(q)}(\G^2)
3368: = \cr & \sum_{k=h}^{\io}
3369: 2^{h-k-1}
3370: [\b^{(c)}_{k,m}(\tilde\o,\e,\G^1,
3371: \{\n^{(c')(q-1)}_{k',m'}(\G^1) \})-
3372: \b^{(c)}_{k,m}(\tilde\o,\e,\G^2,
3373: \{\n^{(c')(q-1)}_{k',m'}(\G^2) \})]\cr}
3374: \Eq(5.25aa) $$
3375: %
3376: When $q=1$ we have that $\b^{(c)}_{k,m}(\tilde\o,\e,\G^1,
3377: \{\n^{(c')(q-1)}_{k',m'}(\G^1) \})-
3378: \b^{(c)}_{k,m}(\tilde\o,\e,\G^2,
3379: \{\n^{(c')(q-1)}_{k',m'}(\G^2) \})$
3380: is given by a sum of self energy graphs with
3381: one node $\vvv$ with a factor $\h_{\vvvv}$
3382: with $\G$ replaced by $\G^1-\G^2$; as there is at least
3383: a vertex $\vvv$ of w-type by the definition of the self energy graphs
3384: we obtain
3385: %
3386: $$ \|\nu_h^{(1)}(\G^1)-\nu_h^{(1)}(\G^2)\|_{\io}\le
3387: \left( D_{1} \e+\tilde D_1\e^2 \right) |\G^1-\G^2| ,
3388: \Eq(5.102) $$
3389: %
3390: for positive constants $D_{1}<2 D$ and $\tilde D_2$, where
3391: $D_{1} \e|\G^1-\G^2|$ is a bound for the self-energy first order contribution.
3392:
3393: For $q>1$ we can write the difference
3394: in \equ(5.25aa) as
3395: %
3396:
3397: $$\eqalign{
3398: & \left( \b^{(c)}_{k,m}(\tilde\o,\e,\G^1,
3399: \{\n^{(c')(q-1)}_{k',m'}(\G^1) \})-
3400: \b^{(c)}_{k,m}(\tilde\o,\e,\G^2,
3401: \{\n^{(c)(q-1)}_{k',m'}(\G^1) \}) \right) \cr
3402: & + \qquad \left( \b^{(c)}_{k,m}(\tilde\o,\e,\G^2,
3403: \{\n^{(c)(q-1)}_{k',m'}(\G^1) \})-
3404: \b^{(c)}_{k,m}(\tilde\o,\e,\G^2,
3405: \{\n^{(c)(q-1)}_{k',m'}(\G^2) \}) \right) . \cr}
3406: \Eq(5.677)$$
3407: %
3408: The first factor is given by a sum over self-energy graphs
3409: with one node $\vvv$ with a factor $\h_{\vvvv}$
3410: with $\G$ replaced by $\G^1-\G^2$; the other difference
3411: is given by a sum over self energy graphs with a $\n$-vertex
3412: to which is associated a factor
3413: $\n^{(c)(q-1)}_{k',m'}(\G^1)-\n^{(c)(q-1)}_{k',m'}(\G^2)$;
3414: hence we find
3415: %
3416: $$ \|\nu_h^{(q)}(\G^1)-\nu_h^{(q)}(\G^2)\|_{\io} \le
3417: \left( D_{1} \e + D_3\e^2 \right)
3418: |\G^1-\G^2| + \e D_2 \sup_{h \ge 0}
3419: \|\n_h^{(q-1)}(\G^1)- \n_h^{(q-1)}(\G^2) \|_\io ,
3420: \Eq(5.103) $$
3421: %
3422: where $D_1\e |\G^1-\G^2|$ is a bound for
3423: the first order contribution coming from the first line in \equ(5.677),
3424: while the last summand in \equ(5.103)
3425: is a bound from the terms from the last line of \equ(5.677).
3426: Then \equ(5.101) follows with $D=2D_{1}$, for $\e$ small enough. \qed
3427:
3428: \*
3429:
3430: By using Lemma 11 we can show that the self consistence equation
3431: for $\G$ \equ(5.3a) has a unique solution $\G=O(\e)$.
3432:
3433: \*
3434:
3435: \0{\bf Lemma 12.} {\it
3436: For all $\m_0>0$ there exists $\e_0>0$
3437: such that, for all $|\m| \le \m_0$ and
3438: for all $0<\e <\e_{0}$, given $\n_{m}^{(c)}(\G)$
3439: chosen as in Lemma 9
3440: (with $\bar h=\io$ and $\n_{\io,m}^{(c)}=0$)
3441: it holds that \equ(5.3a) has a solution $|\G|\le C\e$
3442: where $C$ is a suitable constant.}
3443:
3444: \*
3445:
3446: \0{\it Proof.}
3447: The solution of \equ(5.3a) can be
3448: obtained as the limit as $q\to\io$ of the sequence $\G^{(q)}$,
3449: $q\ge 0$, defined recursively as
3450: %
3451: $$ \eqalign{
3452: \G^{(0)} & = 0 , \cr
3453: \G^{(q)} & = r_{0}^{-1}
3454: \sum_{k=1}^{\io} {\mathop{\sum}_{\th \in \Th^{(k)(q-1))\RR}_{n,n}}}^{*}
3455: a_{0,-n} \Val(\th) , \cr}
3456: \Eq(5.256) $$
3457: %
3458: where we define $\Th^{(k)(q)\RR}_{n,m}$ as the set of trees
3459: identical to $\Th^{(k)\RR}_{n,m}$
3460: except that $\G$ in $\h_{\vvvv}$
3461: is replaced by $\G^{(q)}$ and each $\n_{h,m}$ is replaced by
3462: $\n_{h,m}(\G^{(q)})$, for all $h\ge 0, |m|\ge 1$.
3463: \equ(5.256) is a contraction
3464: defined on the set $|\G|\le C\e$, for $\e$ small.
3465: In fact if $|\G^{(q-1)}|\le C\e$, then by \equ(5.256)
3466: $|\G^{(q)}|\le C_1\e+C_2 C\e^2$, where we have used
3467: that the first order contribution to the r.h.s.
3468: of \equ(5.256) is $\G$-indipendent (see Section \secc(3)),
3469: and $C_1\e<\e C /$ is a bound for it;
3470: hence for $\e$ small enough \equ(5.256)
3471: send the interval $|\G|\le C\e$ to itself.
3472:
3473: Moreover we can show inductively that
3474: %
3475: $$\|\G^{(q)}-\G^{(q-1)} \|_{\io}
3476: \le (C\e)^{q} .
3477: \Eq(5.266) $$
3478: %\
3479: For $q=1$ this is true;
3480: for $q>1$ $\G^{(q)}-\G^{(q-1)}$
3481: can be written as sum of trees
3482: in which a)or to a node $\vvv'$
3483: is associated a factor proportional to $\G^{(q-1)}-\G^{(q-2)}$;
3484: or b) to a $\nu$ vertex is associated $\nu_{h',m'}(\G^{(q-1)})-
3485: \nu_{h',m'}(\G^{(q-2)})$ for some $h',m'$.
3486: In the first case we note that the constraint in
3487: the sum in the r.h.s. of \equ(5.256)
3488: implies that $s_{v_0}=3$ for the special vertex of $\th$;
3489: hence, item (4) in Section \secc(3), says that $\vvv'\not= v_0$
3490: so that such terms are bounded by $D_1 \e|\G^{(1)}-\G^{(2)}|$
3491: (a term order $O(\G^{1}-\G^2)$ should have three
3492: $v$ lines entering $v_0$ and two of them coming from
3493: end points, which is impossible).
3494: In the second case we use
3495: \equ(5.100), and we bound such terms by $D_2 \e|\G^{(1)}-\G^{(2)}|$ .
3496: Hence by induction \equ(5.266) is found, if $C\ge D_1/4,
3497: D_2/4, C_1/4$). \qed
3498:
3499: \*
3500:
3501: We have finally to prove that $\tilde u_{n,m}$
3502: solves the last of \equ(1.13) and \equ(2.2).
3503:
3504: \*
3505:
3506: \0{\bf Lemma 13.} {\it
3507: For all $\m_0>0$ there exists $\e_0>$
3508: such that, for all $|\m|\le\m_0$ and
3509: for all $0<\e < \e_{0}$, given $\n_{m}^{(c)}(\G)$
3510: chosen as in Lemma 9 and $\G$ chosen as in lemma 12
3511: then $\tilde u_{n,m}$ solves the last of \equ(1.13) and \equ(2.2).}
3512:
3513: \*
3514:
3515: \0{\it Proof.} Let us consider first the case
3516: in which $|n|\not=|m|$ and we call $\Th_{n,m}^{\RR}=\bigcup_{k}
3517: \Th^{(k)\RR}_{n,m}$; assume also
3518: (what of course is not restrictive) that $n,m$
3519: is such that $\chi_{h_0}(|\o n|-\tilde\o_m)+
3520: \chi_{h_0+1}(|\o n|-\tilde\o_m)=1$.
3521: We call $\Th_{n,m,}^{\RR}$
3522: the set of trees $\th\in \Th_{n,m}^{\RR}$
3523: with root line at scale $h_0$, so that
3524: %
3525: $$ \tilde u_{n,m}=\sum_{\th\in \Th_{n,m}^{\RR}}\Val(\th)=
3526: \sum_{\th\in \Th_{n,m,h_0}^{\RR}}\Val(\th)+
3527: \sum_{\th\in \Th_{n,m,h_0+1}^{\RR}}\Val(\th) ,
3528: \Eq(5.cvv)$$
3529: %
3530: and we write
3531: $\Th_{n,m,h_0}^{\RR}=\Th_{n,m,h_0}^{\a,\RR}\bigcup
3532: \Th_{n,m,h_0}^{\b,\RR}$, where $\Th_{n,m,h_0}^{\a,\RR}$
3533: are the trees with $s_{\vvvv_0}=1$, while $\Th_{n,m,\bar h}^{\b,\RR}$
3534: are the trees with $s_{\vvvv_0}=3$ and $\vvv_0$ is the special
3535: vertex (see Definition 2). Then
3536: %
3537: $$ \eqalign{
3538: & \sum_{\th\in \Th_{n,m,h_0}^{\a,\RR}}\Val(\th)
3539: = \sum_{c=a,b} \Big( g^{(\bar h)}(n,m) 2^{-h_0}\n_{h_0,m}^{(c)}
3540: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0}}
3541: \Val(\th) \cr
3542: & \qquad + g^{(\bar h)}(n,m) 2^{-\bar h}\n_{h_0,m}^{(c)}
3543: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0+1}}
3544: \Val(\th) \cr
3545: & \qquad + g^{(h_0+1)}(n,m) 2^{-h_0}\n^{(c)}_{h_0,m}
3546: \sum_{\th \in \Th^{\RR}_{n,m,h_0}}
3547: \Val(\th) \cr
3548: & \qquad + g^{(h_0+1)}(n,m) 2^{-h_0-1}\n^{(c)}_{h_0+1,m}
3549: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0+1}}
3550: \Val(\th) \Big) , \cr}
3551: \Eq(5.1aaaa)
3552: $$
3553: where $m_c$ is such that $m_a=m$ and $m_b=-m$.
3554: %
3555: On the other hand we can write
3556: $\Th_{n,m,h_0 }^{\b,\RR}=\Th_{n,m,h_0}^{\b1,\RR}\bigcup
3557: \Th_{n,m,h_0}^{\b2,\RR}$, where
3558: $\Th_{n,m,h_0}^{\b1,\RR}$ are the trees
3559: such that the root line $\ell_0$
3560: is the external line of a self-energy graph,
3561: and $\Th_{n,m,h_0}^{\b2,\RR}$ is the complement.
3562: Then we can write
3563: %
3564: $$\eqalign{
3565: &\sum_{\th\in\Th_{n,m,h_0}^{\b1,\RR}}
3566: \Val(\th)=\sum_{c=a,b}
3567: g^{(h_0)}(n,m) \sum_{T \in \tilde\TT^{(c)}_{< h_0}}
3568: \RR\VV_T^{\bar h}(\o n,m)
3569: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0}}
3570: \Val(\th)\cr
3571: & \qquad +g^{(h_0)}(n,m) \sum_{T \in \tilde\TT^{(c)}_{< h_0}}
3572: \RR\VV_T^{h_0}(\o n,m)
3573: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0+1}}
3574: \Val(\th) \cr
3575: & \qquad + g^{(h_0+1)}(n,m) \sum_{T \in
3576: \tilde\TT^{(c)}_{< h_0}}
3577: \RR\VV_T^{h_0}(\o n,m)
3578: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0}}
3579: \Val(\th) \cr
3580: & \qquad + g^{(h_0+1)}(n,m) \sum_{T \in \tilde\TT^{(c)}_{< h_0+1}}
3581: \RR\VV_T^{h_0+1}(\o n,m)
3582: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0+1}}
3583: \Val(\th) , \cr}
3584: \Eq(5.bbb) $$
3585: %
3586: where $\tilde\TT^{(c)}_{<h_0}$
3587: is the set of self-energy graphs such that if $\vvv$
3588: is the vertex to which the external line $\ell_0$
3589: is attached, than $s_{\vvvv}=3$. Note that if $T$ belongs
3590: to the complementary of $\tilde\TT^{(c)}_{< \bar h}$,
3591: then $\LL\VV_T^h=0$ by the compact support properties of $g^{(h)}$.
3592:
3593: Summing \equ(5.1aa) and \equ(5.bbb)
3594: and using \equ(5.1), \equ(5.3) we get
3595: %
3596: $$\eqalign{
3597: & \sum_{\th\in\Th_{n,m,h_0}^{\b1,\RR}}
3598: \Val(\th)=\sum_{c=a,b}
3599: g^{(h_0)}(n,m) \sum_{T \in \TT^{(c)}_{< h_0}}
3600: \VV_T^{h_0}(\o n,m)
3601: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0}}
3602: \Val(\th)\cr
3603: & \qquad +g^{(h_0)}(n,m) \sum_{T \in \TT^{(c)}_{< h_0}}
3604: \VV_T^{h_0}(\o n,m)
3605: \sum_{\th \in \Th^{\RR}_{n,m,h_0+1}}
3606: \Val(\th) \cr
3607: & \qquad + g^{(h_0+1)}(n,m) \sum_{T \in \TT^{(c)}_{< h_0}}
3608: \VV_T^{h_0}(\o n,m)
3609: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0}}
3610: \Val(\th) \cr
3611: & \qquad + g^{(h_0+1)}(n,m) \sum_{T \in \TT^{(c)}_{< h_0+1}}
3612: \VV_T^{h_0+1}(\o n,m)
3613: \sum_{\th \in \Th^{\RR}_{n,m_c,h_0+1}}
3614: \Val(\th) \cr
3615: & \qquad +\n^{(c)}_m \left( g^{(h_0)}(n,m)+g^{(h_0+1)}(n,m) \right)
3616: \Big( \sum_{\th \in \Th^{\RR}_{n,m_c,h_0}}
3617: \Val(\th)+\sum_{\th \in \Th^{\RR}_{n,m_c,h_0+1}}
3618: \Val(\th) \Big) . \cr}
3619: \Eq(5.bbbb) $$
3620: %
3621: The last line is equal to
3622: %
3623: $$ {1 \over -\o^2 n^2 + \tilde\o_{m}^2}
3624: [\n_{m}^{(a)} \tilde u_{n,m}+ \n_{m}^{(b)} \tilde u_{n,-m}] ,
3625: \Eq(5.ss)$$
3626: %
3627: while adding the first three lines in \equ(5.bbb)
3628: to $\sum_{\th\in\Th_{n,m,h_0}^{\b1,\RR}}
3629: \Val(\th)$ we get
3630: %
3631: $$\e\sum_{n_1+n_2+n_3=n\atop m_1+m_2+m_3=m}
3632: \tilde u_{n_1,m_1}\tilde u_{n_2,m_2}\tilde u_{n_3,m_3} ,
3633: \Eq(5.sss)$$
3634: %
3635: from which we get that
3636: $\sum_{\th\in \Th_{n,m}^{\RR}}\Val(\th)$, for $\m=1$,
3637: is a formal solution of \equ(2.2).
3638: A similar result holds for $|n|=|m|$. \qed
3639:
3640:
3641: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3642: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3643: \*\*
3644: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3645: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3646: \section(6,Construction of the perturbed frequencies)
3647: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3648: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3649:
3650: \0In the following it will be convenient to set
3651: $\tilde\o=\{\o_{m}\}_{|m|\ge 2}$.
3652: By the analysis of the previous sections we have
3653: found the counterterms $\{\n_m(\tilde\o,\e)\}_{|m|\ge 2}$
3654: as functions of $\e$ and $\tilde\o$.
3655: We have now to invert the relations
3656: %
3657: $$ \tilde\o^2_m-\n_{m}(\tilde\o,\e) = m^{2} ,
3658: \Eq(6.1)$$
3659: %
3660: in order to prove Proposition 2.
3661:
3662: We shall show that there exists a sequence of sets
3663: $\{\EE^{(p)}\}_{p=0}^{\io}$ in $[0,\e_{0}]$, such that
3664: $\EE^{(p+1)} \subset \EE^{(p)}$, and a sequence of functions
3665: $\{\tilde\o^{(p)}(\e)\}_{p=0}^{\io}$, with each
3666: $\tilde\o^{(p)} \= \tilde\o^{(p)}(\e)$ defined
3667: for $\e\in\EE^{(p)}$, such that for all $\e\in\EE$, with
3668: %
3669: $$ \EE = \bigcap_{p=0}^{\io} \EE^{(p)} =
3670: \lim_{p\to\io} \EE^{(p)} ,
3671: \Eq(6.2) $$
3672: %
3673: there exists the limit
3674: %
3675: $$ \tilde\o^{(\io)}(\e) = \lim_{p\to\io} \tilde\o^{(p)}(\e) ,
3676: \Eq(6.3) $$
3677: %
3678: and it solves \equ(6.1).
3679:
3680: To fulfill the program above we shall define since the beginning
3681: $\o=\o_{\e}=\sqrt{1-\e}$, and we shall follow an
3682: iterative scheme by setting, for $|m|\ge 1$,
3683: %
3684: $$ \eqalign{
3685: & \tilde \o_{m}^{(0)2} = m^2 , \cr
3686: & \tilde \o_{m}^{(p)2} \= \tilde \o_m^{(p)2}(\e) =
3687: m^{2} + \n_{m}(\tilde\o^{(p-1)},\e) , \qquad p\ge 1 , \cr}
3688: \Eq(6.4) $$
3689: %
3690: and reducing recursively the set of admissible values of $\e$.
3691:
3692: We start by imposing on $\e$ the Diophantine conditions
3693: %
3694: $$ | \o n \pm m |
3695: \ge 2 C_{0} | n|^{-\t_{0}}
3696: \qquad \forall n\in\ZZZ \setminus \{ 0 \} \hbox{ and }
3697: \forall m \in \ZZZ \setminus \{0\} \hbox{ such that } |m| \neq |n| ,
3698: \Eq(6.5) $$
3699: %
3700: where $C_{0}$ and $\t_{0}$ are two positive constants.
3701:
3702: This will imply some restrictions on the admissible
3703: values of $\e$, as the following result shows.
3704:
3705: \*
3706:
3707: \0{\bf Lemma 14.}
3708: {\it For all $0 < C_{0} \le 1/2$ there exist $\e_{0}>0$
3709: and $\g_{0},\d_{0}>0$ such that the set $\EE^{(0)}$ of
3710: values $\e\in[0,\e_{0}]$ for which \equ(6.5) are satisfied has
3711: Lebesgue measure $|\EE^{(0)}|\ge \e_{0}(1-\g_{0}\e_{0}^{\d_{0}})$
3712: provided that one has $\t_{0}>1$.}
3713:
3714: \*
3715:
3716: \0{\it Proof.}
3717: For $(n,m)$ such that $|\o n \pm m|\ge 2C_0$
3718: the Diophantine conditions in \equ(6.5) are trivially satisfied.
3719: We consider then $(n,m)$ such that $|\o n \pm m|<2C_0$
3720: and we write, if $0 < C_0 \le 1/2$
3721: %
3722: $$ 1 > 2 C_0 > |\o n\pm m| \ge
3723: \left| {\e n\over 1 + \sqrt{1-\e}} - n \pm m \right| ,
3724: \Eq(6.6)$$
3725: %
3726: and as $|n\pm m|\ge 1$ one gets $|n| \ge 1/2\e \ge 1/2\e_{0}$.
3727: Moreover, for fixed $n$, the set $\MM$ of $m$'s such that
3728: $|\o n \pm m|<1$ contains at most $2+\e_{0}|n|$ values. By writing
3729: %
3730: $$ f(\e(t)) = n \sqrt{1-\e(t)} \pm m = t {2C_{0} \over |n|^{\t_{0}}} ,
3731: \qquad t\in[0,1] ,
3732: \Eq(6.7) $$
3733: %
3734: and, calling $\III^{(0)}$ the set of $\e$ such that $| \o n \pm m |
3735: < 2C_{0} | n|^{-\t_{0}}$ is verified for some $(n,m)$,
3736: one finds for the Lebesgue measure of $\III^{(0)}$
3737: %
3738: $$ {\rm meas}(\III^{(0)}) = \int_{\III^{(0)}} {\rm d}\e =
3739: \sum_{|n|\ge 1/2\e_{0}} \sum_{|m| \in \MM}
3740: \int_{-1}^{1} {\rm d}t \left|
3741: {{\rm d}\e(t) \over {\rm d}t} \right| .
3742: \Eq(6.8) $$
3743: %
3744: We have from \equ(6.7)
3745: %
3746: $$ {{\rm d} f \over {\rm d}t} = {{\rm d} f \over {\rm d} \e}
3747: {{\rm d} \e \over {\rm d} t} = { 2 C_{0} \over |n|^{\t_0} } ,
3748: \Eq(6.9) $$
3749: %
3750: so that, noting that one has $|\dpr f/\dpr \e| \ge |n|/4$,
3751: we have to exclude a set of measure
3752: %
3753: $$ \sum_{|n| \ge N} { 8 C_{0} \over |n|^{\t_{0}+1}}
3754: \left( 2 + \e_{0} |n| \right) \le
3755: {\rm const.} \, \e_{0}^{\t_{0}} ,
3756: \Eq(6.10) $$
3757: %
3758: and one has to impose $\t_{0}=1+\d_{0}$, with $\d_{0}>0$. \qed
3759:
3760: \*
3761:
3762: For $p\ge 1$ the sets $\EE^{(p)}$ will be defined recursively as
3763: %
3764: $$ \eqalign{
3765: \EE^{(p)} & = \Big\{ \e \in \EE^{(p-1)} \;:\;
3766: | \o n \pm \tilde\o_{m}^{(p)} | >
3767: C_{0} | n|^{-\t} \qquad \forall |m| \neq |n|, \cr
3768: & \qquad \qquad
3769: | \o n \pm ( \tilde\o_{m}^{(p)} \pm
3770: \tilde \o_{m'}^{(p)} ) | > C_{0} | n|^{-\t} \qquad
3771: |n| \neq |m \pm m'| \Big\} ,
3772: \qquad p \ge 1 , \cr}
3773: \Eq(6.11) $$
3774: %
3775: for $\t>\t_{0}$ to be fixed.
3776:
3777: In Appendix \secc(A4) we prove the following result.
3778:
3779: \*
3780:
3781: \0{\bf Lemma 15.} {\it For all $p\ge 1$ one has
3782: %
3783: $$ \left\| \tilde\o^{(p)}(\e)-\tilde\o^{(p-1)}(\e) \right\|_{\io}
3784: \le C \e_{0}^{p} \qquad \forall \e\in \EE^{(p)} ,
3785: \Eq(6.12) $$
3786: %
3787: for some constant $C$.}
3788:
3789: \*
3790:
3791: Therefore we can conclude that there exist a sequence
3792: $\{\tilde\o^{(p)}(\e)\}_{p=0}^{\io}$
3793: converging to $\tilde\o^{(\io)}(\e)$ for $\e\in\EE$.
3794: We have now to show that the set $\EE$ has positive (large) measure.
3795:
3796: It is convenient to introduce a set of
3797: variables $\m(\tilde\o,\e)$ such that
3798: %
3799: $$ \tilde\o_{m} + \m_{m}(\tilde\o,\e) = \o_{m} \= |m| ;
3800: \Eq(6.13) $$
3801: %
3802: the variables $\m(\tilde\o,\e)$ and the counterterms are
3803: trivially related by
3804: %
3805: $$ - \n_{m}(\tilde\o,\e) = \m_{m}^{2}(\tilde\o,\e) +
3806: 2\tilde\o_{m}\m_{m}(\tilde\o,\e) .
3807: \Eq(6.14) $$
3808: %
3809:
3810: One can write $\tilde\o^{(p)}_{m}=\o_{m}-\m_{m}(\tilde\o^{(p-1)},\e)$,
3811: according to \equ(6.4). We shall impose the Diophantine conditions
3812: %
3813: $$ \eqalignno{
3814: & | \o n \pm ( \o_{m} - \m_{m}(\tilde\o^{(p-1)},\e))| > C_{0} | n|^{-\t} ,
3815: & \eq(6.15) \cr
3816: & | \o n \pm ( ( \o_{m} \pm \o_{m'}) -
3817: (\m_{m}(\tilde\o^{(p-1)},\e) \pm \m_{m'}(\tilde\o^{(p-1)},\e) )) | >
3818: C_{0} | n|^{-\t} . \cr} $$
3819: %
3820: Suppose that for $\e\in\EE^{(p-1)}$ the functions
3821: $\m_{m}(\tilde\o^{(p-1)},\e)$ are well defined; then define
3822: $\III^{(p)}=\III^{(p)}_{1}\cup \III^{(p)}_{2}\cup\III^{(p)}_{3}$,
3823: where $\III^{(p)}_{1}$ is the set of values $\e \in \EE^{(p-1)}$
3824: verifying the conditions
3825: %
3826: $$ \left| \o n \pm \left( \o_{m} - \m_{m}(\tilde\o^{(p-1)},\e)
3827: \right) \right| \le C_{0} \left| n \right|^{-\t} ,
3828: \Eq(6.16) $$
3829: %
3830: $\III^{(p)}_{2}$ is the set of values $\e$ verifying the conditions
3831: %
3832: $$ \left| \o n \pm \left( ( \o_{m} - \o_{m'}) -
3833: (\m_{m}(\tilde\o^{(p-1)},\e) - \m_{m'}(\tilde\o^{(p-1)},\e) )
3834: \right) \right| \le C_{0} | n|^{-\t} ,
3835: \Eq(6.17) $$
3836: %
3837: and $\III^{(p)}_{3}$ is the set of values $\e$ verifying the conditions
3838: %
3839: $$ \left| \o n \pm \left( ( \o_{m} + \o_{m'}) -
3840: (\m_{m}(\tilde\o^{(p-1)},\e) + \m_{m'}(\tilde\o^{(p-1)},\e) ) \right)
3841: \right| \le C_{0} | n|^{-\t} .
3842: \Eq(6.18) $$
3843: %
3844: For future convenience we shall call, for $i=1,2,3$, $\III^{(p)}_{i}(n)$
3845: the subsets of $\III^{(p)}_{i}$ which verify the Diophantine conditions
3846: \equ(6.16), \equ(6.17) and \equ(6.18), respectively, for fixed $n$.
3847:
3848: We want to bound the measure of the set $\III^{(p)}$.
3849: First we need to know a little better the dependence on $\e$ and
3850: $\tilde \o$ of the counterterms: this is provided by the following result.
3851:
3852: \*
3853:
3854: \0{\bf Lemma 16.} {\it For all $p\ge 1$ and for
3855: all $\e\in\EE^{(p)}$ there exists a positive constant $C$ such that
3856: %
3857: $$ \eqalign{
3858: & \left| \n_{m}(\tilde\o^{(p)},\e) \right| \le C \e , \qquad
3859: \left| \dpr_{\tilde\o^{(p)}_{m'}} \n_{m}(\tilde\o^{(p)},\e) \right|
3860: \le C \e , \qquad m\ge 1 , \cr
3861: & \left| \m_{m}(\tilde\o^{(p)},\e) \right| \le {C \e \over m} , \qquad
3862: \left| \dpr_{\tilde\o^{(p)}_{m'}} \m_{m}(\tilde\o^{(p)},\e) \right|
3863: \le {C \e \over m} , \qquad m\ge 1 , \cr
3864: & \left| \dpr_{\e} \m^{(p)}_{m}(\o,\e) \right| \le C ,
3865: \qquad m\ge 1 , \cr}
3866: \Eq(6.19) $$
3867: %
3868: where the derivatives are in the sense of Whitney {\rm \cita{Wh}}.}
3869:
3870: \*
3871:
3872: \0{\it Proof.}
3873: The bound for $\n_{m}(\tilde\o^{(p)},\e)$ is obvious by construction.
3874:
3875: In order to prove the bound for
3876: $\dpr_{\o^{(p)}_{m'}} \n_{m}(\tilde\o^{(p)},\e)$ note that one has
3877: %
3878: $$ \left| g^{(h_{\ell_{j}})}_{\ell_{j}}(\tilde \o') -
3879: g^{(h_{\ell_{j}})}_{\ell_{j}}(\tilde \o)- (\tilde\o-\tilde\o') \,
3880: \dpr_{\tilde\o} g^{(h_{\ell_{j}})}_{\ell_{j}}(\tilde \o)\right|
3881: \le C \, n_{\ell_{j}}^{2} 2^{-3h_{\ell_{j}}}
3882: \left\| \tilde \o' - \tilde \o \right\|^2_{\io} ,
3883: \Eq(6.20) $$
3884: %
3885: from the compact support properties of the propagator.
3886:
3887: Let us consider the quantity $\nu(\tilde\o',\e)-
3888: \nu(\tilde\o,\e) - (\o-\o') \, \dpr_{\tilde\o} \nu(\tilde\o,\e)$,
3889: where $\dpr_{\tilde\o} \nu(\tilde\o,\e)$ denotes the
3890: derivative in the sense of Whitney,
3891: and note that it can be expressed as a sum over
3892: trees each one containing a line with propagator
3893: $g^{(h_{\ell_{j}})}_{\ell_{j}}(\tilde \o') -
3894: g^{(h_{\ell_{j}})}_{\ell_{j}}(\tilde \o)- (\tilde\o-\tilde\o') \,
3895: \dpr_{\tilde\o} g^{(h_{\ell_{j}})}_{\ell_{j}}(\tilde \o)$,
3896: by proceeding as in the proof of lemma 9 of \cita{GM}. Then we find
3897: %
3898: $$ \left\| \nu(\tilde\o',\e)-\nu(\tilde\o,\e) - (\tilde\o-\tilde\o')
3899: \, \dpr_{\tilde\o} \n(\tilde\o,\e) \right\|_{\io} \le
3900: C \e \left\| \tilde \o' - \tilde \o \right\|^2_{\io} ,
3901: \Eq(6.21) $$
3902: %
3903: and the second bound in \equ(6.19) follows.
3904:
3905: The bounds for $\m_{m}(\tilde\o^{(p)},\e)$ and
3906: $\dpr_{\o^{(p)}_{m'}} \m_{m}(\tilde\o^{(p)},\e)$ simply follow
3907: from \equ(6.14) which gives
3908: %
3909: $$ \m_{m} (\tilde\o^{(p)},\e) =
3910: { \n_{m}(\tilde\o^{(p)},\e) \over |m| } \,
3911: {- 1 \over 1 + \sqrt{1 - \n_{m}(\tilde\o^{(p)},\e)/m} } .
3912: \Eq(6.22) $$
3913: %
3914:
3915: In order to prove the last bound in \equ(6.19) we prove
3916: by induction the bound
3917: %
3918: $$ \left| \dpr_{\e} \tilde\o^{(p)}_{m}(\o,\e) \right| \le C ,
3919: \qquad m\ge 1 ,
3920: \Eq(6.23) $$
3921: %
3922: by assuming that it holds for $\tilde \o^{(p-1)}$;
3923: then from \equ(6.4) we have
3924: %
3925: $$ \eqalignno{
3926: & 2\tilde\o^{(p)}\dpr_{\e} \tilde\o_m^{(p)}(\e) = -
3927: \dpr_{\e} \n_m(\tilde\o^{(p-1)}(\e),\e)
3928: & \eq(6.24) \cr
3929: & \qquad = - \sum_{m'\in\zzz} \dpr_{\tilde\o^{(p-1)}_{m'}}
3930: \n_m(\tilde\o^{(p-1)}(\e),\e) \, \dpr_{\e} \tilde\o_{m'}^{(p-1)}(\e)
3931: - \left. \dpr_{\h} \n_m(\tilde\o_m^{(p-1)}(\e),\h)
3932: \right|_{\h=\e}, \cr} $$
3933: %
3934: as it is easy to realize by noting that $\m_{m}$ can
3935: depend on $\tilde \o_{m'}$ when $|m'|>|m|$ only if
3936: the sum of the absolute values of the mode labels $m_{v}$
3937: is greater than $|m'-m|$, while we can bound the sum of the
3938: contributions with $|m'|<|m|$ by a constant times $\e$,
3939: simply by using the second line in \equ(6.19).
3940:
3941: Hence from the inductive hypothesis and the proved bounds
3942: in \equ(6.19), we obtain
3943: %
3944: $$ \left\| \tilde\o^{(p)}(\e')-\tilde\o^{(p)}(\e)-
3945: (\e-\e') \, \dpr_\e \tilde\o^{(p)}(\e) \right\|_{\io}
3946: \le C \left| \e' - \e \right| ,
3947: \Eq(6.25) $$
3948: %
3949: so that also the bound \equ(6.23) and hence the last
3950: bound in \equ(6.16) follow. \qed
3951:
3952: \*
3953:
3954: Now we can bound the measure of the set we have to exclude:
3955: this will conclude the proof of Proposition 2.
3956:
3957: We start with the estimate of the measure of the set $\III^{(p)}_{1}$.
3958: When \equ(6.16) is satisfied one must have
3959: %
3960: $$ \eqalign{
3961: & C_{2} |m| \le | \o_{m} - \m_{m}(\tilde\o^{(p-1)},\e) |
3962: \le \o|n| + C_{0} | n|^{-\t} \le C_{2}'| n| , \cr
3963: & C_{1} |m| \ge | \o_{m} - \m_{m}(\tilde\o^{(p-1)},\e) |
3964: \ge \o|n| - C_{0} | n|^{-\t} \ge C_{1}'| n| , \cr}
3965: \Eq(6.26) $$
3966: %
3967: which implies
3968: %
3969: $$ M_{1} |n| \le |m| \le M_{2} |n| ,
3970: \qquad M_{1} = {C_{1} \over C_{1}'} ,
3971: \qquad M_{2} = {C_{2}' \over C_{2}} ,
3972: \Eq(6.27) $$
3973: %
3974: and it is easy to see that, for fixed $n$, the set $\MM_{0}(n)$ of $m$'s
3975: such that \equ(6.16) are satisfied contains at most $2+\e_{0}|n|$ values.
3976:
3977: Furthermore (by using also \equ(6.5)) from \equ(6.16)
3978: one obtains also
3979: %
3980: $$ \eqalign{
3981: 2 C_{0} | n|^{-\t_{0}} & \le | \o |n| - \o_{m} |
3982: \le | \o |n| - \o_{m} + \m_{m}(\tilde \o^{(p-1)},\e) | +
3983: | \m_{m}(\tilde \o^{(p-1)},\e) | \cr
3984: & \le C_{0} | n|^{-\t} + C\e_{0}/|m| , \cr}
3985: \Eq(6.28) $$
3986: %
3987: which implies, together with \equ(6.36), for $\t \ge \t_{0}$,
3988: %
3989: $$ |n| \ge N_{0} \= \left( { C_{0} M_{1} \over C \e_{0} }
3990: \right)^{1/(\t_{0}-1)} .
3991: \Eq(6.29) $$
3992: %
3993: Let us write $\o(\e)=\o_{\e}=\sqrt{1-\e}$ and
3994: consider the function $\m(\tilde\o^{(p-1)},\e)$:
3995: we can define a map $t\to \e(t)$ such that
3996: %
3997: $$ f(\e(t)) \= \o(\e(t)) |n| - \o_{m} + \m_{m}(\tilde\o^{(p-1)},\e(t)) =
3998: t {C_{0} \over | n|^{\t} } , \qquad t \in [-1,1] ,
3999: \Eq(6.30)$$
4000: %
4001: describes the interval defined by \equ(6.16); then the Lebesgue measure
4002: of $\III_{1}^{(p)}$ is
4003: %
4004: $$ {\rm meas}(\III^{(p)}_{1}) = \int_{\III^{(p)}_1} {\rm d}\e =
4005: \sum_{|n|\ge N_{0}} \sum_{m \in \MM_{0}(n)}
4006: \int_{-1}^{1} {\rm d}t \left|
4007: {{\rm d}\e(t) \over {\rm d}t} \right| .
4008: \Eq(6.31) $$
4009: %
4010: We have from \equ(6.30)
4011: %
4012: $$ {{\rm d} f \over {\rm d}t} = {{\rm d} f \over {\rm d} \e}
4013: {{\rm d} \e \over {\rm d} t} = { C_{0} \over |n|^{\t} } ,
4014: \Eq(6.32) $$
4015: %
4016: hence
4017: %
4018: $$ {\rm meas}(\III^{(p)}_{1}) =
4019: \sum_{|n|\ge N_{0}} \sum_{m \in \MM_{0}(n)}
4020: { C_{0} \over | n|^{\t}} \int_{-1}^{1} {\rm d} t
4021: \left| {{\rm d} f(\e(t))\over {\rm d} \e(t)} \right|^{-1} .
4022: \Eq(6.33) $$
4023: %
4024: In order to perform the derivative in \equ(6.30) we write
4025: %
4026: $$ {{\rm d} \m_{m}\over {\rm d}\e} = \dpr_\e \m_{m} +
4027: \sum_{|m'|\ge 1} \dpr_{\o^{(p-1)}_{m'}}\m_{m}
4028: \dpr_\e \tilde\o_{m'}^{(p-1)} ,
4029: \Eq(6.34) $$
4030: %
4031: where $\dpr_\e \m_{m}$ and $\dpr_\e \tilde\o_{m'}^{(p-1)}$
4032: are bounded through lemma 11. Moreover one has
4033: %
4034: $$ \left| \dpr_{\tilde\o^{(p-1)}_{m'}} \m_{m} \right| \le
4035: C \e\, e^{-\k|m'-m|/2} , \qquad |m'| > |m| ,
4036: \Eq(6.35)$$
4037: %
4038: and the sum over $m'$ can be dealt with as in \equ(6.24).
4039:
4040: At the end we get that the sum in \equ(6.34)
4041: is $O(\e)$, and from \equ(6.29) and \equ(6.30) we obtain,
4042: if $\e_{0}$ is small enough,
4043: %
4044: $$ \left| {\dpr f(\e(t)) \over \dpr \e(t)} \right|
4045: \ge { \left| n \right| \over 4} ,
4046: \Eq(6.36) $$
4047: %
4048: so that one has
4049: %
4050: $$ \eqalign{
4051: \int_{\III^{(p)}_1} {\rm d}\e
4052: & \le \hbox{const.} \sum_{|n|\ge N_{0}}
4053: { C_{0} \over | n|^{\t+1}}
4054: \left( 2 + \e_{0} |n| \right) \cr
4055: & \le {\rm const.} \, \e_{0}
4056: \left( \e_{0}^{(\t-1)/(\t_{0}-1)} + \e_{0}^{(\t-\t_{0}+1)/(\t_{0}-1)}
4057: \right) . \cr}
4058: \Eq(6.37) $$
4059: %
4060: Therefore for $\t \ge \max\{1,\t_{0}-1\}$ the Lebesgue measure of the set
4061: $\III^{(p)}_{1}$ is bounded by $\e_{0}^{1+\d_{1}}$, with $\d_{1}>0$.
4062:
4063: Now we discuss how to bound the measure of the set
4064: $\III^{(p)}_{2}$. We start by noting that from \equ(6.30) we obtain,
4065: if $m,\ell>0$,
4066: %
4067: $$ \left| \tilde\o^{(p)}_{m+\ell} - \tilde\o^{(p)}_{m} -
4068: \ell \right| \le {C \e \over m} + {C \e \over m + \ell} ,
4069: \Eq(6.38) $$
4070: %
4071: for all $p\ge 1$.
4072:
4073: By the parity properties of
4074: $\tilde\o_{m}$ without loss of generality we can confine
4075: ourselves to the case $n>0$, $m'>m \ge 2$,
4076: and $|\o n-(\tilde\o_{m'}^{(p)}-\tilde\o_{m}^{(p)})|<1$.
4077: Then the discussion proceeds as follows.
4078:
4079: When the conditions \equ(6.17) are satisfied, one has
4080: %
4081: $$ \eqalign{
4082: 2 C_{0} | n|^{-\t_{0}} & \le
4083: | \o n - (\o_{m'}-\o_{m}) | \cr
4084: & \le | \o n -
4085: (\o_{m'} - \m^{(p-1)}_{m'}(\tilde\o^{(p-1)},\e) ) \cr
4086: & \qquad \qquad + (\o_{m} - \m^{(p-1)}_{m}(\tilde\o^{(p-1)},\e) ) |
4087: \cr & \qquad \qquad + | \m_{m'}(\tilde \o^{(p-1)},\e) -
4088: \m_{m}(\tilde \o^{(p-1)},\e) | \cr
4089: & \le C_{0} | n|^{-\t} + {C \e \over m} + {C \e \over m'}
4090: \le C_{0}|n|^{-\t} + {2C\e_{0} \over m} , \cr}
4091: \Eq(6.39) $$
4092: %
4093: which implies for $\t \ge \t_{0}$
4094: %
4095: $$ | n| \ge N_{1} \= \left( { C_{0} \over 2 C \e_{0} }
4096: \right)^{1/\t_{0}} .
4097: \Eq(6.40) $$
4098: %
4099:
4100: We can bound the Lebesgue measure
4101: of the set $\III^{(p)}_{2}$ by distinguishing,
4102: for fixed $(n,\ell)$, with $\ell=m'-m>0$,
4103: the values $m\le m_{0}$ and $m>m_{0}$,
4104: where $m_{0}$ is determined by the request that
4105: one has for $m>m_{0}$
4106: %
4107: $$ {2 C \e_{0} \over m} \le {C_{0} \over |n|^{\t_{0}}} ,
4108: \Eq(6.41) $$
4109: %
4110: which gives
4111: %
4112: $$ m_{0} = m_{0}(n) = \left( {2 C |n|^{\t_{0}} \e_{0} \over C_{0}} \right) .
4113: \Eq(6.42) $$
4114: %
4115: Therefore for $m>m_{0}$ and $\t \ge \t_{0}$
4116: one has, from \equ(6.5), \equ(6.38) and \equ(6.41),
4117: %
4118: $$ \left| \o n -
4119: (\tilde \o_{m'}^{(p)} - \tilde \o_{m}^{(p)} ) \right|
4120: \ge \left| \o n - \ell \right|
4121: - {2C\e \over m } \ge {2C_{0} \over |n|^{\t_{0}}} -
4122: {C_{0} \over |n|^{\t_{0}} } \ge {C_{0} \over |n|^{\t_{0}} }
4123: \ge {C_{0} \over |n|^{\t} } ,
4124: \Eq(6.43) $$
4125: %
4126: so that one has to exclude no further value from
4127: $\EE^{(p-1)}$, provided one takes $\t \ge \t_{0}$.
4128:
4129: For $m<m_{0}$ define $L_{0}$ such that
4130: %
4131: $$ C_{3} \ell \le \left| \tilde \o_{m+\ell}^{(p)} -
4132: \tilde \o_{m}^{(p)} \right| < \o n + 1 < C_{3}' |n| , \qquad
4133: L_{0} = { C_{3}' \over C_{3} } ,
4134: \Eq(6.44) $$
4135: %
4136: where \equ(6.26) has been used. Again, for fixed $n$,
4137: the set $\LL_{0}(n)$ of $\ell$'s such that \equ(6.26) is satisfied
4138: with $m'-m=\ell$ contains at most $2+\e_{0}|n|$ values.
4139:
4140: Therefore, by reasoning as in obtaining \equ(6.37),
4141: one finds that for $m<m_{0}$ one has to exclude from $\EE^{(p-1)}$
4142: a set of measure bounded by a constant times
4143: %
4144: $$ \sum_{|n| \ge N_{1}} \sum_{\ell \in \LL_{0}(n)}
4145: \sum_{m < m_{0}(n)} {C_{0} \over |n|^{\t+1}} \le
4146: {\rm const.} \, \e_{0} \left(
4147: \e_{0}^{(\t-\t_{0})/\t_{0}} +
4148: \e_{0}^{1+ (\t-\t_{0}-1)/\t_{0}} \right),
4149: \Eq(6.45) $$
4150: %
4151: provided that one has $\t>\t_{0}+1>2$ the Lebesgue measure of the set
4152: $\III^{(p)}_{2}$ is bounded by $\e_{0}^{1+\d_{2}}$, with $\d_{2}>0$.
4153:
4154: Finally we study the measure of the set $\III^{(p)}_{3}$.
4155: If $n>0$, $|m'|>|m|$ and $|(\tilde\o_{m}^{(p)}+\tilde\o_{m'})-
4156: \o n| < 1$ (which again is the only case we can confine ourselves
4157: to study), then one has to sum over $|n|\le N_{1}$,
4158: with $N_{1}$ given by \equ(6.40). For such values of $n$ one has
4159: %
4160: $$ \eqalign{
4161: \left| \o n - (\tilde \o_{m'}^{(p)} + \tilde \o_{m}^{(p)} ) \right|
4162: & \ge \left| \o n - ( |m|+|m'|) \right| - {2C\e \over |m| } \cr
4163: & \ge {2C_{0} \over |n|^{\t_{0}}} -
4164: {C_{0} \over |n|^{\t_{0}} } \ge {C_{0} \over |n|^{\t_{0}} }
4165: \ge {C_{0} \over |n|^{\t} } , \cr}
4166: \Eq(6.46) $$
4167: %
4168: as soon as $|m|>m_{0}$, with $m_{0}$ given by \equ(6.42).
4169: Therefore we have to take into account only the values
4170: of $m$ such that $|m|<m_{0}$, and we can also note that $|m'|$
4171: is uniquely determined by the values of $n$ and $m$.
4172: Then one can proceed as in the previous case and in the end one
4173: excludes a further subset of $\EE^{(p-1)}$
4174: whose Lebesgue measure is bounded by a constant times
4175: %
4176: $$ \sum_{|n| \ge N_{1}}
4177: \sum_{m < m_{0}(n)} {C_{0} \over |n|^{\t+1}} \le
4178: {\rm const.} \, \e_{0}^{1+ (\t-\t_{0})/\t_{0}} ,
4179: \Eq(6.47) $$
4180: %
4181: so that the Lebesgue measure of the set
4182: $\III^{(p)}_{3}$ is bounded by $\e_{0}^{1+\d_{3}}$, with $\d_{3}>0$,
4183: provided that one takes $\t>\t_{0}$.
4184:
4185: By summing together the bounds for $\III^{(p)}_{1}$,
4186: for $\III^{(p)}_{2}$ and for $\III^{(p)}_{3}$, then the bound
4187: %
4188: $$ {\rm meas}(\III^{(p)}) \le b \, \e_{0}^{\d+1}
4189: \Eq(6.48) $$
4190: %
4191: with $\d>0$, follows for all $p\ge 1$, if $\t$ is chosen to be
4192: $\t>\t_{0}+1>2$.
4193:
4194: We can conclude the proof of Proposition 2 through the
4195: following result, which shows that the bound \equ(6.42)
4196: essentially extends to the union of all $\III^{(p)}$
4197: (at the cost of taking a larger constant $B$ instead of $b$).
4198:
4199: \*
4200:
4201: \0{\bf Lemma 17.} {\it Define $\III^{(p)}$ as the set of values
4202: in $\EE^{(p)}$ verifying \equ(6.26) to \equ(6.28) for $\t>1$.
4203: Then one has, for two suitable positive constants $B$ and $\d$,
4204: %
4205: $$ {\rm meas}\left( \cup_{p=0}^{\io} \III^{(p)} \right)
4206: \le B \e_{0}^{\d+1} ,
4207: \Eq(6.49) $$
4208: %
4209: where ${\rm meas}$ denotes the Lebesgue measure.}
4210:
4211: \*
4212:
4213: \0{\it Proof.}
4214: First of all we check that, if we call $\e^{(p)}_{j}(n)$
4215: the centers of the intervals $\III^{(p)}_{j}(n)$, with $j=1,2,3$,
4216: then one has
4217: %
4218: $$ \left| \e^{(p+1)}_{j}(n) - \e^{(p)}_{j}(n) \right|
4219: \le D \e_{0}^{p} ,
4220: \Eq(6.50) $$
4221: %
4222: for a suitable constant $D$.
4223:
4224: The center $\e^{(p)}_{1}(n)$ is defined by the condition
4225: %
4226: $$ \o n \pm \left( \o_{m} - \m_{m}(\tilde\o^{(p-1)}(\e^{(p)}_{1}(n)),
4227: \e^{(p)}_{1}( n)) \right) = 0 ,
4228: \Eq(6.51) $$
4229: %
4230: where Whitney extensions are considered outside $\EE^{(p-1)}$;
4231: then, by subtracting \equ(6.51) from the equivalent
4232: expression for $p+1$, we have
4233: %
4234: $$ \left(
4235: \m_{m}(\tilde\o^{(p)}(\e^{(p+1)}_{1}( n)),\e^{(p+1)}_{1}( n)) -
4236: \m_{m}(\tilde\o^{(p-1)}(\e^{(p)}_{1}( n)),\e^{(p)}_{1}( n))
4237: \right) = 0 .
4238: \Eq(6.52) $$
4239: %
4240: In \equ(6.48) one has
4241: %
4242: $$ \eqalign{
4243: & \m_{m}(\tilde\o'(\e'),\e') - \m_{m}(\tilde\o(\e),\e) =
4244: \m_{m}(\tilde\o'(\e'),\e') - \m_{m}(\tilde\o(\e'),\e') \cr
4245: & \qquad +
4246: \m_{m}(\tilde\o(\e'),\e') - \m_{m}(\tilde\o(\e),\e')
4247: + (\m_{m}(\tilde\o(\e),\e') -
4248: \m_{m}(\tilde\o(\e),\e)) , \cr}
4249: \Eq(6.53) $$
4250: %
4251: and, from Lemma 13,
4252: %
4253: $$ \eqalign{
4254: & \left| \m_{m}(\tilde \o'(\e'),\e') - \m_{m}(\tilde \o(\e'),\e') \right|
4255: \le C \e \left\|\o'-\o\right\|_{\io} , \cr
4256: & \left| \m_{m}(\tilde \o(\e'),\e') - \m_{m}(\tilde \o(\e),\e') \right|
4257: \le C \e \left| \e'-\e \right| , \cr}
4258: \Eq(6.54) $$
4259: %
4260: so that we get, by lemma 10,
4261: %
4262: $$ \left| \e^{(p+1)}_{1}( n) - \e^{(p)}_{1}( n) \right|
4263: \le C \e_{0}^{p} ,
4264: \Eq(6.55) $$
4265: %
4266: for a suitable positive constant $C$.
4267: This proves the bound \equ(6.50) for $j=1$.
4268: Analogously one can consider the cases $j=2$ and $j=3$,
4269: and a similar result is found.
4270:
4271: %Define $p_{0}\=p_{0}( n,j)$ such that for all $p>p_0$
4272: %the distance of $\e^{p}_j(n)$ from $\e^{p_0}_j(n)$
4273: %is smaller that $C_0 |n|^{-\t}$.
4274: By
4275: \equ(6.10), \equ(6.51), \equ(6.54) and \equ(6.12) it follows
4276: for $p>p_0$
4277: %
4278: $$|\e^{(p)}_{j}( n) - \e^{(p_{0})}_{j}( n)|
4279: \le C\sum_{k=p_0}^\io \e_0^{k}=C {\e_0^{p_0}\over 1-\e_0}
4280: \Eq(6.56) $$
4281: %
4282: %$$C|n|\e_0^{p_0}\le C_{0} | n|^{-\t} .
4283: %\Eq(6.65) $$
4284: %$$ |\e^{(p_{0}+1)}_{j}( n) - \e^{(p_{0})}_{j}( n)|
4285: %\le C_{0} | n|^{-\t} .
4286: %\Eq(6.65) $$
4287: %
4288: so that one can ensure that $ |\e^{(p)}_{j}( n) - \e^{(p_{0})}_{j}( n)|
4289: \le C_{0} | n|^{-\t}$ for $p\ge p_0$ by choosing
4290: %
4291: $$ p_{0} = p_{0}( n,j) \le \hbox{const.} \log | n| .
4292: \Eq(6.57) $$
4293: %
4294: For all $p\le p_{0}$ define $\JJ^{(p)}_{j}( n)$ as the set
4295: of values $\e$ such that \equ(6.16), \equ(6.17) and \equ(6.18)
4296: are satisfied with $C_{0}$ replaced with $2C_{0}$.
4297: By the definition of $p_{0}$ all the intervals $\III^{(p)}_{j}( n)$
4298: fall inside the union of the intervals
4299: $\JJ^{(0)}_{j}( n),\ldots,\JJ^{(p_{0})}_{j}( n)$
4300: as soon as $p>p_{0}$. This means that, by \equ(6.31) to \equ(6.37)
4301: %
4302: $$ \eqalign{
4303: {\rm meas}\left( \cup_{p=0}^{\io} \III_1^{(p)} \right) &
4304: \le {\rm meas}\left( \cup_{p=0}^{p_0} \JJ_1^{(p)} \right)
4305: \le \sum_{|n|\ge N_{0}}
4306: {\rm meas} (\III^{(p)}_{1}(n)) \cr
4307: & \le \hbox{const.}
4308: \sum_{|n|\ge N_{0}} \sum_{p=0}^{p_{0}(n,1)}
4309: {2 C_{0} \over | n|^{\t+1} }
4310: \le {\rm const.} \sum_{n=N_{0}}^{\io} n^{-(\t+1)} \log n \le
4311: B \e_{0}^{1+\d} , \cr}
4312: \Eq(6.58) $$
4313: %
4314: with suitable $B$ and $\d$, in order to take into account
4315: the logarithmic corrections due to \equ(6.55).
4316: Analogously one obtains the bounds
4317: $ {\rm meas}(\III_{2}) \le B \e_{0}^{1+\d}$ and
4318: ${\rm meas}(\III_{3}) \le B \e_{0}^{1+\d}$
4319: for the the Lebesgue measures of the sets $\III_{2}$
4320: and $\III_{3}$
4321: (possibly redefining the constant $B$).
4322: This completes the proof of the bound \equ(6.48). \qed
4323:
4324:
4325: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4326: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4327: \*\*
4328: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4329: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4330: \section(7,The case 1$\minore\tt\leq$2)
4331: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4332: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4333:
4334: \0Proposition 1 was proved assuming that $\tilde\o$ verifies
4335: the first and the second Diophantine condition \equ(2.31) with $\t>2$.
4336: Her we wnat to prove that it is possible to obtain a result similar
4337: to Proposition 1 assuming only the first Diophantine
4338: condition and $1<\t\le 2$, and that
4339: also in such a case the set of allowed values of $\e$
4340: have large Lebesgue measure, so that a result analogous
4341: to Proposition 2 holds.
4342:
4343: \*
4344:
4345: The proof of the analogue of Proposition 1 is an immediate adaptation
4346: of the analysis in Sectiones \secc(4) and \secc(5).
4347: First we consider a slight different multiscale
4348: decomposition of the propagator; instead of \equ(4.2)
4349: we write
4350: %
4351: $$ {1\over - \o^{2} n_{\ell}^2 +\tilde\o_{m_{\ell}}^2}=
4352: {\chi( \sqrt{6|\o^{2} n_{\ell}^2 -\tilde\o_{m_{\ell}}^2|})
4353: \over - \o^{2} n_{\ell}^2 +\tilde\o_{m_{\ell}}^2} +
4354: {\chi_{-1}( \sqrt{6|\o^{2} n_{\ell}^2 -\tilde\o_{m_{\ell}}^2|})
4355: \over - \o^{2} n_{\ell}^2 +\tilde\o_{m_{\ell}}^2} ,
4356: \Eq(7.1aa) $$
4357: %
4358: and the denominator in the first addend of the r.h.s.
4359: of \equ(7.1aa) is smaller than $C_{0}/6$; as in Section \secc(4)
4360: we assume $C_{0}\le 1/2$. If $|n|\ne |m|$ and
4361: $|\o^2 n^2 -\tilde\o_{m}^2| < C_{0}/6$, then,
4362: by reasoning as in \equ(4.12), we obtain
4363: %
4364: $$|n|> |m|> {3\over 4} |n| , \qquad
4365: \min\{ |m| , |n| \} > {1 \over 2\e} ,
4366: \Eq(7.fon)$$
4367: %
4368: and
4369: %
4370: $$ |\o |n|- \tilde\o_m|< {C_{0} \over 6 (\o |n|+ \tilde\o_{m})} <
4371: {C_{0}\over 6|n|} .
4372: \Eq(7.10e)$$
4373: %
4374:
4375: We can decompose the first summand in \equ(7.1aa), obtaining
4376: %
4377: $$\chi( \sqrt{6|\o^{2} n_{\ell}^2 -\tilde\o_{m_{\ell}}^2|})
4378: \sum_{h=-1}^{\io}
4379: { \chi_h( | \o n|-\tilde\o_{m})
4380: \over - \o^{2} n^2+\tilde\o_{m}^2 }
4381: \equiv \sum_{h=-1}^{\io} g^{(h)}(\o n,m) ,
4382: \Eq(7.2aa)$$
4383: %
4384: and the scales from $-1$ to $h_{0}$, with $h_{0}$ given in the statement
4385: of Lemma 5, can be bounded as in Lemma 6.
4386:
4387: If $N_{h}(\th)$ is the number of lines on scale $h\ge h_{0}$,
4388: and assuming only the first Melnikov condition in \equ(2.31),
4389: the following reult holds.
4390:
4391: \*
4392:
4393: \0{\bf Lemma 18.} {\it
4394: Assume that there is a constant $C_{1}$ such that
4395: one has $| \tilde\o_{m}-|m| |\le C_{1}\e/|m|$ and $\tilde\o$
4396: verifies the first Melnikov condition in \equ(2.31) with $\t>1$.
4397: If $\e$ is small enough for any tree $\th \in \Th_{n,m}^{(k)}$ and
4398: for all $h\ge h_{0}$ one has
4399: %
4400: $$ \bar N_h(\th) \le 4 K(\th) 2^{(2-h)/\t^2} -
4401: C_{h}(\th) + M_{h}^{\n}(\th) + S_{h}(\th) ,
4402: \Eq(7.11) $$
4403: %
4404: where $M_{h}^{\n}(\th)$ and $S_{h}(\th)$ are defined as
4405: the number of $\n$-vertices in $\th$ such that the maximum scale
4406: of the two external lines is $h$ and, respectively, the number of
4407: self-energy graphs in $\th$ with $h_{T}^{e}=h$.}
4408:
4409: \*
4410:
4411: \0{\it Proof.}
4412: We prove inductively the bound
4413: %
4414: $$ N^{*}_{h}(\th) \le \max\{ 0 , 2 K(\th) 2^{(1-h)/\t^2} - 1 \} ,
4415: \Eq(7.12) $$
4416: %
4417: where $N_{h}^{*}(\th)$ is the number of non-resonant lines in $L(\th)$
4418: on scale $h'\ge h$.
4419: We proceed exactly as in the proof of Lemma 5.
4420: First of all note that for a tree $\th$ to have a line on scale
4421: $h$ the condition $K(\th) > 2^{(h-1)/\t}> 2^{(h-1)/\t^2} $ is necessary,
4422: by the first Diophantine conditions in \equ(2.31).
4423: This means that one can have $N_{h}^{*}(\th) \ge 1$ only
4424: if $K=K(\th)$ is such that $K > k_{0}\=2^{(h-1)/\t}$:
4425: therefore for values $K \le k_{0}$ the bound \equ(7.12) is satisfied.
4426: If $K=K(\th)>k_{0}$, we assume that
4427: the bound holds for all trees $\th'$ with $K(\th')<K$.
4428: Define $E_{h}=2^{-1}(2^{(1-h)/\t^2})^{-1}$, we have
4429: to prove that $N^{*}_{h}(\th)\le \max\{0,K(\th) E_{h}^{-1}-1\}$.
4430:
4431: The dangerous case is if $m=1$ then one has a cluster $T$
4432: with two external lines $\ell$ and $\ell_{1}$,
4433: which are both with scales $\ge h$; then
4434: %
4435: $$ \left| |\o n_{\ell}| - \tilde\o_{m_{\ell}} \right|
4436: \le 2^{-h+1} C_{0} , \qquad
4437: \left| |\o n_{\ell_{1}}| - \tilde\o_{m_{\ell_{1}}} \right|
4438: \le 2^{-h+1} C_{0} ,
4439: \Eq(7.15a) $$
4440: %
4441: and recall that $T$ is not a self-energy graph,
4442: so that $n_{\ell}\not=n_{\ell_1}$.
4443: Note that the validity of both inequalities in \equ(7.15a)
4444: for $h \ge h_{0}$ imply, by Lemma 5, that one has
4445: %
4446: $$|n_{\ell}-n_{\ell_{1}}|\neq
4447: |m_{\ell}\pm m_{\ell_{1}}| .
4448: \Eq(7.10c)$$
4449: %
4450: Moreover from \equ(7.10e), \equ(7.10c) and \equ(7.fon) one obtains
4451: %
4452: $$ {C_0\over |n_{\ell}-n_{\ell_{1}}|^{\t}}\le
4453: |\o|n_{\ell}-n_{\ell_1}| -| m_{\ell_1}\pm m_{\ell_1}|| <
4454: {C_{0} \over 3 \min \{ |n_{\ell}| , |n_{\ell_{1}}| \} } ,
4455: \Eq(7.100) $$
4456: %
4457: which implies
4458: %
4459: $$ \left| n_{\ell}-n_{\ell_1} \right| \ge 3^{1/\t}
4460: \min \{ |n_{\ell}| , |n_{\ell_{1}}| \}^{1/\t} ,
4461: \Eq(7.10d)$$
4462: %
4463:
4464: Finally if $C_{0}|n|^{-\t} \le | |\o n| - \tilde\o_{m} | <
4465: C_{0} 2^{-h+1 } $ we have $|n| \ge 2^{(h-1)/\t}$, so that
4466: from \equ(7.10c) we obtain $|n_{\ell}-n_{\ell_{1}}| \ge
4467: 3^{1/\t} 2^{(h-1)/\t^2}$.
4468: Then $K(\th)-K(\th_{1})> E_{h}$, which gives \equ(7.12),
4469: by using the inductive hypothesis. \qed
4470:
4471: \*
4472:
4473: The analysis in Section \secc(5) can be repeated with the following
4474: modifications. The {\it renormalized trees} are defined as in
4475: Section \secc(5), but the rule (9) is replaced with
4476:
4477: \*
4478:
4479: \0(9$^{\prime}$) To each self-energy graph $T$ the $\LL'$ operation
4480: is applied, where
4481: %
4482: $$\LL'\VV_T^{h}(\o n,m)= \VV_T^{h}(\o n,m) ,
4483: \Eq(7.1aaaaa)$$
4484: %
4485: if $T$ is such that its two external lines are attached
4486: to the same node $\vvv_{0}$ of $T$ (see as an example
4487: the first graph of Figure 4.1), and $\LL'\VV_T^{h}(\o n,m)=0$ otherwise.
4488:
4489: \*
4490:
4491: With this definition we have the expansion \equ(5.4) to \equ(5.3a),
4492: with $\n_{h,m}$ replaced with $\n_{h}$ (as $\LL'\VV_T^{h}(\o n,m)$
4493: is independent of $n,m$).
4494:
4495: We have that the analogue of Lemma 7 still holds also with $\LL',\RR'$
4496: replacing $\LL,\RR$; indeed by construction
4497: the dependence on $(n,m)$ in $\RR' \VV_T^{h_T}(\o n,m)$
4498: is due to the propagators of lines $\ell$ along the path
4499: connecting the external lines of the self-energy graph.
4500: The propagators of such lines have the form
4501: %
4502: $$ { \chi_h(|\o n_{\ell}|-\tilde \o_{m_{\ell}})
4503: \over - \left( |\o n_{\ell}| +\tilde\o_{m_{\ell}} \right)
4504: \left( |\o n_{\ell}| - \tilde\o_{m_{\ell}} \right)} ,
4505: \Eq(7.71a) $$
4506: %
4507: and the second factor in the denominator is
4508: bounded proportionally to $2^{-h_{\ell}}$,
4509: while the first is bounded proportionally
4510: to $|m_{\ell}^{0}+m_{\ell_{T}^{e}}|$.
4511: Then $\RR' \VV_T^{h_T}(\o n,m)$ is bounded by a constant
4512: times $e^{-\k |m_{\ell_{0}}|}/|m_{\ell}^{0}+m_{\ell_{T}^{e}}|$,
4513: hence by a constant times $1/m_{\ell_{T}^{e}}$.
4514:
4515: We can also bound the propagator associated to one of the
4516: external lines of $T$ with $|n_{\ell^{e}_{T}}|^{\t-1}$, by using
4517: the first Diophantine condition, so that
4518: then bound \equ(5.8) is replaced, using also Lemma 18, by
4519: %
4520: $$ \eqalign{
4521: & \left| \Val(\th) \right| \le \e^{k}
4522: \overline{D}^{k}
4523: \Big( \prod_{h=h_0}^{\io} \exp \Big[ h \log 2 \Big( 4 K(\th)
4524: 2^{-(h-2)/\t^{2}} - C_{h}(\th) +
4525: M_{h}^{\n}(\th) \Big) \Big] \Big) \cr
4526: & \qquad \qquad
4527: \Big( \prod_{T \in \SS(\th)}
4528: {|n_{\ell^e_T}^{\t-1}|\over|m_{\ell^e_T}|}
4529: \Big)
4530: \Big( \prod_{h=h_{0}}^{\io} 2^{-h M_{h}^{\n}(\th)} \Big)
4531: \cr & \qquad \qquad
4532: \Big( \prod_{\vvv\in V(\th) \cup E(\th)} e^{-\k(|n_{v}|+|n_{v}'|)} \Big)
4533: \Big( \prod_{\vvv\in V(\th) \cup E(\th)} e^{-\k(|m_{v}|+|m_{v}'|)} \Big)
4534: , \cr}
4535: \Eq(6.8a) $$
4536: %
4537: where $|n_{\ell^e_T}^{\t-1}|/|m_{\ell^{e}_{T}}|$ is less than a contant
4538: provided one takes $\t\le 2$. Hence \equ(5.6) follows.
4539: There is no need of Lemma 8, while the analogues of Lemma 9 and Lemma 10
4540: can be proved with some obvious changes.
4541: For instance in \equ(5.38) one has $\h_{\ell}=0$, which shows that
4542: the first Mel'nikov is required, and $\n_{m}$ is replaced with $\n$,
4543: i.e. one needs only one counterterm.
4544:
4545: Finally also the analogue of Proposition 2 can be proved by
4546: reasoning as in Section \secc(6), simply by observing that
4547: in order to impose the first Mel'nikov conditions \equ(6.16)
4548: one can take $\t_{0}=\t$; note indeed that the condition $\t>2$
4549: was made necessary to obtain the second Mel'nikov conditions
4550: \equ(6.17) and \equ(6.18).
4551:
4552:
4553: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4554: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4555: \*\*
4556: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4557: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4558: \section(8,Generalizations of the results)
4559: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4560: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4561:
4562: \0So far we have considered only the case $\f(u)=u^{3}$ in \equ(1.1).
4563: Now we consider the case in which the function $\f(u)$ in \equ(1.1)
4564: is replaced with any odd analytic function
4565: %
4566: $$ \f(u)= \F u^{3} + O(u^{5}) , \qquad \F \neq 0 .
4567: \Eq(8.1) $$
4568: %
4569: Define $\o=\sqrt{1-\l\e}$ as in Section \secc(1). Then by choosing
4570: $\l=\s$ we obtain, instead of \equ(1.13),
4571: %
4572: $$ \eqalign{
4573: \hbox{Q} \qquad & \cases{
4574: n^{2} a_{n} = \left[ (v+w)^{3} + O(\e) \right]_{n,n} , & \cr
4575: - n^{2} a_{n} = \left[ (v+w)^{3} + O(\e) \right]_{n,-n} , & \cr} \cr
4576: \hbox{P} \qquad & \left( - \o^2 n^2 + m^2 \right) w_{n,m} = \e
4577: \left[ (v+w)^{3} + O(\e) \right]_{n,m} , \qquad |m| \neq |n| , \cr}
4578: \Eq(8.2) $$
4579: %
4580: where $O(\e)$ denotes analytic functions in $u$ and $\e$
4581: of order at least one in $\e$. Then we can introduce an auxiliary
4582: parameter $\m$, by replacing $\e$ with $\e\m$ in \equ(8.2)
4583: (recall that at the end one has to set $\m=1$).
4584: Then we can proceed as in the previous sections.
4585: The equation for $a_{0}$ is the same as before, and only the
4586: diagrammatic rules for the coefficients $u^{(k)}_{n,m}$
4587: have to be slightly modified.
4588: The nodes can have any odd number of entering lines,
4589: that is $s_{\vvvv}=1,3,5,\ldots$.
4590: For nodes $\vvv$ of $w$-type with $s_{\vvvv} \ge 3$ the node factor
4591: is given by $\h_{\vvvv}=\e^{(s_{\vvvv}-1)/2}$, which has to be added
4592: to the list of node factors \equ(3.3), while for nodes
4593: $\vvv$ of $v$-type one has to add to the list in \equ(3.2)
4594: other (obvious) contributions, arising from the fact that
4595: the function $f^{(k)}_{n}$ in \equ(2.12)
4596: has to be replaced with
4597: %
4598: $$ f^{(k)}_{n} = \sum_{k'=1}^{k} f^{(k,k')}_{n} ,
4599: \Eq(8.3) $$
4600: %
4601: where the function $f^{(k,1)}_{n}$ is given by \equ(2.12),
4602: while for $k\ge 2$ and $1< k' \le k$ one has
4603: %
4604: $$ f^{(k,k')}_{n} =
4605: - \sum_{\scriptstyle k_{1}+\ldots+k_{2k'+1}=k-k'}
4606: \sum_{\scriptstyle n_{1}+\ldots+n_{2k'+1} = n \atop
4607: \scriptstyle m_{1}+\ldots+m_{2k'+1} = m}
4608: u^{(k_{1})} _{n_{1},m_{1}} \ldots u^{(k_{2k'+1})} _{n_{2k'+1},m_{2k'+1}} ,
4609: \Eq(8.4) $$
4610: %
4611: where the symbols have to be interpreted according to \equ(2.13)
4612: and \equ(2.14); note that $s=2k'+1$ is the number of lines
4613: entering the node in the corresponding graphical representation.
4614:
4615: In the same way one has to modify \equ(2.29) by replacing
4616: the last term in the r.h.s. with
4617: %
4618: $$ [\f(v+w)]^{(k)}_{n,m} = \sum_{k'=1}^{k}
4619: [\f(v+w)]^{(k,k')}_{n,m} ,
4620: \Eq(8.5) $$
4621: %
4622: where $[\f(v+w)]^{(k,1)}_{n,m}$ is given by the old \equ(2.30),
4623: while all the other terms are given by expressions analogous
4624: to \equ(8.4).
4625:
4626: The discussion then proceeds exactly as in the previous cases.
4627: Of course we have to use that, by writing
4628: %
4629: $$ \f(u) = \sum_{k=1}^{\io} \F_{k} u^{2k+1} ,
4630: \Eq(8.6) $$
4631: %
4632: with $\F_{1}=\F$, the constants $\F_{k}$ can be bounded
4633: for all $k \ge 1$ by a constant to the power $k$
4634: (which follows from the analyticity assumption).
4635:
4636: By taking into account the new diagrammatic rules,
4637: one change in the proper way the definition of the tree value,
4638: and a result analogous to Lemma 2 is easily obtained.
4639: The second statement of Lemma 3 has to be changed into
4640: $|E(\th)|\le \sum_{\vvvv\in V(\th)} (s_{\vvvv}-1)+1$
4641: (see \equ(3.24)), while the bound on $|V_{v}^{3}(\th)|$
4642: still holds. For $V_{v}^{s}(\th)$, with $s\ge 5$,
4643: one has to use the factors $\e^{(s-1)/2}$.
4644:
4645: Nothing changes in the following sections, except that
4646: the bound \equ(5.11) has to be suitably modified
4647: in order to take into account the presence of the new kinds of
4648: nodes (that is the nodes with branching number more than three).
4649:
4650: At the end we obtain the proof of Theorem 1 for general
4651: odd nonlinearities starting from the third order.
4652: Until now we are still confining ourselves to the case $j=1$.
4653:
4654: If we choose $j>1$ we have perform a preliminary rescaling
4655: $u(t,x) \to j u (jt,jx) $, and we write down the equation
4656: for $U(t,x)=u(jt,jx)$. If $\f(u)=Fu^{3}$ we see immediately
4657: that the function $U$ solves the same equation as before,
4658: so that the same conditions on $\e$ has to be imposed
4659: in order to find a solution. In the general case
4660: \equ(8.2) holds with $j^{2}\e$ replacing $\e$.
4661: This completes the proof of Theorem 1 in all cases.
4662:
4663:
4664: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4665: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4666: \*\*
4667: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4668: \appendix(A1,The solution of {\equ(1.5)})
4669: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4670: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4671:
4672: \0The odd $2\pi$-periodic solutions of \equ(1.5) can be found in the
4673: following way \cita{LS}, \cita{PBC}.
4674: First we consider \equ(1.5), with $\langle a_{0}^{2} \rangle$
4675: replaced by a parameter $c_{0}$, and we see that
4676: %
4677: $$ a_{0}(\x) = V \, \sn (\O\x,\mm) ,
4678: \Eqa(A1.1) $$
4679: %
4680: where $\sn(\O\x,\mm)$ is the sine-amplitude function
4681: with modulus $\sqrt{m}$ \cita{GrR}, \cita{AS}, is an odd solution of
4682: %
4683: $$ \ddot a_{0} = - 3 c_{0} a_{0} - a_{0}^{3} ,
4684: \Eqa(A1.2) $$
4685: %
4686: if the following relations are verified by $V,\O,c_0,\mm$
4687: %
4688: $$ V = \sqrt{ -2\mm} \O , \qquad \qquad
4689: {V^{2} \over 6 c_{0} + V^{2}} = - \mm ,
4690: \Eqa(A1.3) $$
4691: %
4692: Of course both $V$ and $\mm$ can be written as function of $c_0$
4693: and $\O$ as
4694: %
4695: $$ \mm = {{3 c_0\over\O^2}}-1 , \qquad \qquad
4696: V = \sqrt{2-{6 c_0\over\O^2}} \, \O .
4697: \Eqa(A1.4) $$
4698: %
4699: In particular one finds
4700: %
4701: $$ \dpr_\O V = \sqrt{-2 \mm} + {1\over \sqrt{-2\mm}} \, {6 c_0\over\O^2}
4702: = {\sqrt{-2 \mm}\over \mm} \left( \mm -{3 c_0\over\O^2} \right) =
4703: {\sqrt{2\over -\mm}} ,
4704: \Eqa(A1.5)$$
4705: %
4706: so that one has
4707: %
4708: $$ {\O{\dpr_{\O}}V \over V} = {1\over -\mm } .
4709: \Eqa(A1.6) $$
4710: %
4711:
4712: If we impose also that the solution \equ(A1.1) is $2\p$-periodic (and we
4713: recall that $4 K(\mm)$ is the natural period of the sine-amplitude
4714: $\sn(\x,\mm)$, \cita{AS}), we obtain $\O = \O_{\mmmm} = 2K(\mm)/\p$,
4715: and $V$ is fixed to the value $V_{\mmmm}=\sqrt{-2\mm}\O_{\mmmm}$.
4716:
4717: Finally imposing that $c_{0}$ equals the average of $a_{0}^{2}$
4718: fixes $\mm$ to be the solution of \cita{PBC}
4719: %
4720: $$ E(\mm) = K(\mm) \, {7+\mm \over 6} ,
4721: \Eqa(A1.7) $$
4722: %
4723: where
4724: %
4725: $$ K(\mm) = \int_{0}^{\p/2} {\rm d}\th \, {1 \over
4726: \sqrt{1 - \mm \sin^{2}\th} } , \qquad
4727: E(\mm) = \int_{0}^{\p/2} {\rm d}\th \,
4728: \sqrt{ 1 - \mm \sin^{2} \th}
4729: \Eqa(A1.8) $$
4730: %
4731: are, respectively, the complete elliptic integral of first kind and
4732: the complete elliptic integral of second kind \cita{AS}, and we have
4733: used that the average of $\sn^{2}(\O\x,\mm)$ is $(K(\mm)-E(\mm))/K(\mm)$.
4734: This gives $\mm\approx -0.2554$.
4735:
4736: One can find $2\p/j$-periodic solutions by noticing that
4737: equations \equ(1.5) are invariant by the simmetry $a_{0}(\x)
4738: \rightarrow \a \, a_{0}(\a \x)$, so that a complete set of odd
4739: $2\p$-periodic solutions is provided by $a_0(\x,j)= j \, a_0(j\x)$.
4740:
4741:
4742: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4743: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4744: \*\*
4745: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4746: \appendix(A2, Proof of Lemma 1)
4747: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4748: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4749:
4750: \0The solution of \equ(2.19) is found by variation of constants.
4751: We write $\langle a_{0}^{2} \rangle = c_{0}$, and consider $c_{0}$
4752: as a parameter. Let $a_{0}(\x)$ be the solution \equ(A1.1)
4753: of equation \equ(A1.2), and define the Wronskian matrix
4754: %
4755: $$ W(\x) = \left( \matrix{
4756: w_{11}(\x) & w_{12}(\x) \cr
4757: w_{21}(\x) & w_{22}(\x) \cr} \right) ,
4758: \Eqa(A2.1) $$
4759: %
4760: which solves the linearized equation
4761: %
4762: $$ \dot W = M(\x) W , \qquad
4763: M(\x) = \left( \matrix{
4764: 0 & 1 \cr -3a_{0}^{2}(\x)-3 c_{0} & 0 \cr} \right) ,
4765: \Eqa(A2.2) $$
4766: %
4767: and is such that $W(0)=\openone$ and $\det W(\x)=1$ $\forall\x$.
4768: We need two independent solutions of the the linearized
4769: equation. We can take one as $\dot a_0$, the other as
4770: $\partial_\O a_0$ (we recall that $\mm$ and $V$ are function of
4771: $\O$ and $c_0$ through \equ(A1.3)); then one has
4772: %
4773: $$ \eqalignno{
4774: w_{11}(\x) & = {1\over \O V} \dot a_{0}(\x) =
4775: \cn(\O\x,\mm)\,\dn(\O\x,\mm) , \cr
4776: w_{21}(\x) & = \dot w_{11}(\x) =
4777: -\sn(\O\x,\mm) \left( \dn^{2}(\O\x,m) +
4778: \mm \, \cn^{2}(\O\x,\mm) \right) , \cr
4779: w_{12}(\x) & = {1 \over V (1+D_{\mmmm})} \dpr_{\O} a_{0}(\x)
4780: & \eqa(A2.3) \cr
4781: & = B_{\mmmm} \left( \x \, \cn(\O\x,\mm) \, \dn(\O\x,\mm) +
4782: \O^{-1} D_{\mmmm} \sn(\O\x,\mm) \right) , \cr w_{22}(\x)
4783: & = \dot w_{12}(\x) \cr
4784: & = \cn(\O\x,\mm) \, \dn(\O\x,\mm) - \O B_{\mmmm} \x \, \sn(\O\x,\mm)
4785: \left( \dn^{2}(\O\x,\mm) + \mm \, \cn^{2}(\O\x,\mm) \right) , \cr} $$
4786: %
4787: where we have used \equ(A1.3) to define the dimensionless constants
4788: %
4789: $$ D_{\mmmm} = {\O \dpr_{\O} V \over V } = {1\over -\mm} , \qquad\qquad
4790: B_{\mmmm} = {1 \over \left( 1 + D_{\mmmm} \right) } = {- \mm \over 1 - \mm} .
4791: \Eqa(A2.4) $$
4792: %
4793:
4794: As we are interested in the case $c_{0}=\langle a_{0}^{2} \rangle$
4795: we set $\O=\O_{\mmmm}=2K(\mm)/\p$ (in order to have the period equal to
4796: $2\p$) and we fix $\mm$ as in Appendix \secc(A1).
4797:
4798: Then, by defining $X=(y,\dot y)$ and $F=(0,h)$,
4799: we can write the solution of \equ(2.19) as the first component of
4800: %
4801: $$ X(\x) = W(\x)\,\lis X +
4802: W(\x) \int_{0}^{\x} {\rm d}\x' \, W^{-1}(\x') \, F(\x') ,
4803: \Eqa(A2.5) $$
4804: %
4805: where $\lis X = (0,\dot y(0))$ denote the corrections
4806: to the initial conditions $(a_{0}(0),\dot a_{0}(0))$,
4807: and $y(0)=0$ as we are looking for an odd solution.
4808:
4809: Shorten $c(\x)\=\cn(\O_{\mmmm}\x,\mm)$,
4810: $s(\x)\=\sn(\O_{\mmmm}\x,\mm)$, and $d(\x)\=\dn(\O_{\mmmm}\x,\mm)$,
4811: and define $cd(\x)=\cn(\O_{\mmmm}\x,\mm)\,\dn(\O_{\mmmm}\x,\mm)$.
4812: One can write the first component of \equ(A2.5) as
4813: %
4814: $$ \eqalign{
4815: y(\x) & = w_{12}(\x) \, \dot y(0) + \int_{0}^{\x} {\rm d} \x'
4816: \left( w_{12}(\x)\,w_{11}(\x') -
4817: w_{11}(\x)\,w_{12}(\x') \right) h(\x') \cr
4818: & = B_{\mmmm} \left( \x \, cd(\x) +
4819: \O_{\mmmm}^{-1}D_{\mmmm} s(\x) \right)\dot y(0) \cr
4820: & \qquad + B_{\mmmm} \left( cd(\x) \int_{0}^{\x} {\rm d}\x'
4821: \int_{0}^{\x'} {\rm d}\x'' \, cd(\x'')\, h (\x'') \right. \cr
4822: & \qquad \left. + \O_{\mmmm}^{-1} D_{\mmmm}
4823: \left( s(\x) \int_{0}^{\x} {\rm d}\x' \, cd(\x')\,h(\x') -
4824: cd(\x) \int_{0}^{\x} {\rm d}\x'\, s(\x') \, h(\x')
4825: \right) \right) , \cr}
4826: \Eqa(A2.6) $$
4827: %
4828: as we have explictly written
4829: %
4830: $$ \eqalignno{
4831: & w_{12}(\x)\,w_{11}(\x')-w_{11}(\x)\,w_{12}(\x')
4832: & \eqa(A2.7) \cr
4833: & \quad = B_{\mmmm} \left( \x\, cd(\x)\,cd(\x')
4834: + \O_{\mmmm}^{-1}D_{\mmmm} s(\x)\,cd(\x') - cd(\x)\,\x'\, cd(\x') -
4835: \O_{\mmmm}^{-1}D_{\mmmm} cd(\x)\,s(\x') \right) , \cr} $$
4836: %
4837: and integrated by parts
4838: %
4839: $$ \eqalign{
4840: & \int_{0}^{\x} {\rm d}\x' \left(
4841: \x \, cd(\x) \, cd(\x') - cd(\x)\,\x'\,cd(\x') \right) h(\x') \cr
4842: & \qquad \qquad \qquad \qquad =
4843: cd(\x) \int_{0}^{\x} {\rm d}\x' \int_{0}^{\x'}
4844: {\rm d}\x'' cd(\x'')\,h(\x'') . \cr}
4845: \Eqa(A2.8) $$
4846: %
4847: By using that if $\PP[F]=0$ then $\PP[\II[F]]=0$
4848: and that $\II$ switches parity we can rewrite \equ(A2.6) as
4849: %
4850: $$ \eqalign{
4851: y(\x) & = B_m \Big( \x \, cd(\x) \Big(\dot y(0)-
4852: \II[ cd\, h](0) - \O_{\mmmm}^{-1}D_{\mmmm}\la s\, h\ra\Big) +
4853: \O_{\mmmm}^{-1}D_{\mmmm} s(\x) \Big( \dot y(0)-\II[cd\,h](0)\Big) \cr
4854: & \qquad + \O_{\mmmm}^{-1}D_m\big (s(\x) \, \II[cd\, h](\x) - cd(\x) \,
4855: \II[\PP[s \, h]](\x) \big) + cd(\x) \, \II[\II[cd\,h]](\x) \Big) . \cr}
4856: \Eqa(A2.9) $$
4857: %
4858: This is an odd $2\p$-periodic analytic function provided that we choose
4859: %
4860: $$ \dot y(0) - \II[cd \, h](0) -
4861: \O_{\mmmm}^{-1}D_{\mmmm}\la s \, h \ra = 0 ,
4862: \Eqa(A2.10) $$
4863: %a
4864: which fixes the parameter $\dot y(0)$; hence \equ(2.20) is found
4865:
4866:
4867: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4868: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4869: \*\*
4870: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4871: \appendix(A3, Proof of {\equ(2.25)})
4872: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4873: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4874:
4875: \0We have to compute $\langle a_0 \LLL[a_0] \rangle$, which is
4876: given by (recall that one has $a_0(\x)=V_{\mmmm}\,s(\x)$
4877: and $cd(\x)=\O_{\mmmm}^{-1} \dot s(\x)$)
4878: %
4879: $$ \la a_0 \LLL[a_0]\ra = B_{\mmmm} \O_{\mmmm}^{-2} V_{\mmmm}^{2}
4880: \left( D_{\mmmm}^{2} \la s^{2} \ra^{2} +
4881: D_{\mmmm} \la s^{2} \, \II [s \dot s ] \ra -
4882: D_{\mmmm} \la s \dot s \,\II [\PP[ s^{2}] ] \ra +
4883: \la s\dot s \II[\II [s \dot s] ] \ra \right) .
4884: \Eqa(A3.1)$$
4885: %
4886: Using that one has $\II[s \dot s] = (s^2- \langle s^2 \rangle)/2$,
4887: we obtain
4888: %
4889: $$ \la s^2 \II [s \dot s ] \ra = {1\over 2}
4890: \left( \la s^4\ra-\la s^2\ra^2 \right) ;
4891: \Eqa(A3.2) $$
4892: %
4893: moreover, integrating by parts, we find
4894: %
4895: $$ \eqalign{
4896: & \la s \dot s \,\II[\PP[ s^2]] \ra = -
4897: {1\over 2} \la s^4\ra+{1\over 2}\la s^2\ra^2, \cr
4898: & \la s\dot s \II[\II[s \dot s]] \ra = -
4899: {1\over 4} \la s^4 \ra+{1\over 4}\la s^2\ra^2 , \cr}
4900: \Eqa(A3.3) $$
4901: %
4902: so that we finally get
4903: %
4904: $$ \la a_0 \LLL[a_0] \ra= {1 \over 2} V_{\mmmm}^{2}
4905: \O_{\mmmm}^{-2} B_{\mmmm}
4906: \left( \left( 2D_{\mmmm} - {1\over2} \right)
4907: \langle s^{4} \rangle + \left( 2D_{\mmmm} \left( D_{\mmmm}-1 \right)
4908: + {1\over 2} \right) \langle s^{2} \rangle^{2} \right) ,
4909: \Eqa(A3.4) $$
4910: %
4911: which is strictly positive by \equ(A2.4) and by the choice of $\mm$
4912: according to Appendix \secc(A1); then \equ(2.25) follows.
4913:
4914:
4915: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4916: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4917: \*\*
4918: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4919: \appendix(A4, Proof of Lemma 15)
4920: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4921: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4922:
4923: \0We shall prove inductively on $p$ the bounds \equ(6.12).
4924: From \equ(6.4) we have
4925: %
4926: $$ | \tilde\o^{(p)}_{m}(\e) - \tilde\o^{(p-1)}_{m}(\e) |
4927: \le C | \n_{m} (\tilde \o^{(p-1)}(\e),\e) -
4928: \n_{m} (\tilde \o^{(p-2)}(\e),\e) | ,
4929: \Eqa(A4.1) $$
4930: %
4931: as we can bound $|\tilde\o^{(p)}_{m}(\e)+\tilde\o^{(p-1)}_{m}(\e)|
4932: \ge 1$ for $\e \in \EE^{(p)}$.
4933:
4934: We set, for $|m| \ge 1$
4935: %
4936: $$ \D \n_{h,m} \= \n_{h,m} (\tilde \o^{(p-1)}(\e),\e) -
4937: \n_{h,m} (\tilde \o^{(p-2)}(\e),\e)
4938: = \lim_{q \to \io} \D \n^{(q)}_{h,m} ,
4939: \Eqa(A4.2) $$
4940: %
4941: where we have used the notations \equ(5.25) to define
4942: %
4943: $$ \D \n^{(q)}_{h,m} = \n^{(q)}_{h,m} (\tilde\o^{(p-1)}(\e),\e) -
4944: \n^{(q)}_{h,m} (\tilde \o^{(p-2)}(\e),\e) .
4945: \Eqa(A4.3) $$
4946: %
4947: We want to prove inductively on $q$ the bound
4948: %
4949: $$ \left| \D \n^{(q)}_{h,m} \right| \le C \e
4950: \| \tilde \o^{(p-1)}(\e) - \tilde \o^{(p-2)}(\e) \|_{\io} ,
4951: \Eqa(A4.4) $$
4952: %
4953: for some constant $C$, uniformly in $q$, $h$ and $m$.
4954:
4955: For $q=0$ the bound \equ(A4.4) is trivially satisfied.
4956: Then assume that \equ(A4.4) hold for all $q'<q$.
4957:
4958: For simplicity we set $\tilde \o=\tilde \o^{(p-1)}(\e)$ and
4959: $\tilde \o'=\tilde \o^{(p-2)}(\e)$.
4960: We can write, from \equ(6.25), for $|m|\ge 1$
4961: and for $h\ge 0$,
4962: %
4963: $$ \eqalign{
4964: & \D \n^{(q)}_{h,m} = - \sum_{k = h}^{\bar h - 1} 2^{-k-2}
4965: \Big( \b^{(q)}_{k,m}(\tilde\o,\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\})
4966: \cr & \qquad \qquad -
4967: \b^{(q)}_{k,m}(\tilde \o',\e,\{\n^{(q-1)}_{k'}(\tilde\o',\e)\})
4968: \Big) , \cr}
4969: \Eqa(A4.5) $$
4970: %
4971: where we recall that $\b^{(q)}_{k,m}
4972: (\tilde\o,\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\})$
4973: depend only on $\n^{(q-1)}_{k'}(\tilde\o,\e)$ with $k'\le k-1$, and
4974: we can set
4975: %
4976: $$ \eqalign{
4977: & \b^{(q)}_{k,m}(\tilde\o,\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\}) \cr
4978: & \qquad \qquad
4979: = \b^{(a)(q)}_{k,m}(\tilde\o,\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\}) -
4980: \b^{(b)(q)}_{k,m}(\tilde\o,\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\}) , \cr}
4981: \Eqa(A4.6) $$
4982: %
4983: according to the settings in \equ(2.3).
4984: Then we can split the differences in \equ(A4.6) into
4985: %
4986: $$ \eqalign{
4987: & \b^{(q)}_{k,m}(\tilde\o,\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\}) -
4988: \b^{(q)}_{k,m}(\tilde\o',\e,\{\n^{(q-1)}_{k'}(\tilde\o',\e)\}) \cr
4989: & \qquad =
4990: \left( \b^{(q)}_{k,m}(\tilde\o,\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\}) -
4991: \b^{(q)}_{k,m}(\tilde\o',\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\}) \right) \cr
4992: & \qquad \qquad +
4993: \left( \b^{(q)}_{k,m}(\tilde\o',\e,\{\n^{(q-1)}_{k'}(\tilde\o,\e)\}) -
4994: \b^{(q)}_{k,m}(\tilde\o',\e,\{\n^{(q-1)}_{k'}(\tilde\o',\e)\})
4995: \right) , \cr}
4996: \Eqa(A4.7) $$
4997: %
4998: and we bound separately the two terms.
4999:
5000: The second term can be expressed as sum of trees
5001: $\th$ which differ from the previously considered ones as,
5002: among the nodes $v$ of $w$-type with only one entering line,
5003: there are some with $\n^{(c_{\vvvv})(q-1)}_{k_{\vvvv}}(\tilde\o,\e)$,
5004: some with $\n^{(c_{\vvvv})(q-1)}_{k_{\vvvv}}(\tilde\o',\e)$ and one with
5005: $\n^{(c_{\vvvv})(q-1)}_{k_{\vvvv}}(\tilde\o,\e)-
5006: \n^{(c_{\vvvv})(q-1)}_{k_{\vvvv}}(\tilde\o',\e)$. Then we can bound
5007: %
5008: $$ \eqalign{
5009: & \left| \b^{(q)}_{k,m}(\tilde\o',\{\n^{(q-1)}_{k'}(\tilde\o,\e)\}) -
5010: \b^{(q)}_{k,m}(\tilde\o',\{\n^{(q-1)}_{k'}(\tilde\o',\e)\}) \right| \cr
5011: & \qquad \qquad
5012: \le D_{1} \e\, \sup_{h'\ge 0} \sup_{|m'| \ge 2}
5013: \D \n^{(q-1)}_{h',m'}
5014: \le D_{1} C \e^{2} \left\| \tilde\o' - \tilde\o \right\|_{\io} , \cr}
5015: \Eqa(A4.8) $$
5016: %
5017: by the inductive hypothesis.
5018:
5019: We are left with the first term in \equ(A4.8).
5020: We can reason as in \cita{GM} (which we refer to for details),
5021: and at the end, instead of (A9.14) of \cita{GM}, we obtain
5022: %
5023: $$ \eqalignno{
5024: & \left| \Val'(\th_{0})
5025: \left( g^{(h_{\ell_{j}})}_{\ell_{j}}(\tilde\o') -
5026: g^{(h_{\ell_{j}})}_{\ell_{j}}(\tilde\o) \right)
5027: \Val'(\th_{1})
5028: \Big( \prod_{i=2}^{s} \Val(\th_{i}) \Big) \right|
5029: & \eqa(A4.9) \cr
5030: & \qquad \qquad
5031: \le C^{|V(T)|} e^{-\k K(T)/2}
5032: \e^{|V_{w}(T)|+|V_{v}^{1}(T)|}
5033: \left\| \tilde \o' - \tilde\o \right\|_{\io} \cr
5034: & \qquad \qquad
5035: \le C^{|V(T)|} \e^{|V_{w}(T)|+|V_{v}^{1}(T)|}
5036: e^{-\k K(T)/4} e^{-\k 2^{(k-1)/\t}/4}
5037: \left\| \tilde \o' - \tilde\o \right\|_{\io} , \cr} $$
5038: %
5039: where $K(T)$ is defined in \equ(4.7).
5040:
5041: Therefore can bound $\left\| \tilde\o^{(p)}(\e)-
5042: \tilde\o^{(p-1)}(\e) \right\|_{\io}$
5043: with a constant times $\e$ times the same expression with
5044: $p$ replaced with $p-1$, \ie
5045: $\left\| \tilde\o^{(p-1)}(\e)-\tilde\o^{(p-2)}(\e)
5046: \right\|_{\io}$, so that, by the inductive hypothesis,
5047: the bound \equ(6.12) follows.
5048:
5049:
5050: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5051: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5052: %\*\*
5053: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5054: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5055: % Acknowledgments
5056: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5057: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5058:
5059: %\0{\bf Acknowledgments.} Ack nowled gments ack nowled gments
5060: %ack nowled gments ack nowled gments ack nowled gments ack nowled gments
5061: %ack nowled gments ack nowled gments ack nowled gments ack nowled gments
5062: %ack nowled gments ack nowled gments ack nowled gments ack nowled gments.
5063:
5064: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5065: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5066: %\*\*
5067: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5068: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5069: % References
5070: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5071: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5072:
5073: \rife{AS}{1}{
5074: M. Abramowitz, I.A. Stegun:
5075: {\it Handbook of mathematical functions},
5076: Do\-ver, New York, 1965. }
5077: %
5078: \rife{Ba}{2}{
5079: D. Bambusi,
5080: {\it Lyapunov center theorem for some nonlinear PDE's: a simple proof},
5081: Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4)
5082: {\bf 29} (2000), no. 4, 823--837. }
5083: %
5084: \rife{BP}{3}{
5085: D. Bambusi, S. Paleari,
5086: {\it Families of periodic solutions of resonant PDEs},
5087: J. Nonlinear Sci.
5088: {\bf 11} (2001), no. 1, 69--87. }
5089: %
5090: \rife{BB1}{4}{
5091: M. Berti, Ph. Bolle,
5092: {\it Periodic solutions of nonlinear wave
5093: equations with general nonlinearities},
5094: Comm. Math. Phys.
5095: {\bf 243} (2003), no. 2, 315--328. }
5096: %
5097: \rife{BB2}{5}{
5098: M. Berti, Ph. Bolle,
5099: {\it Multiplicity of periodic solutions of nonlinear
5100: wave equations},
5101: Nonlinear Anal.,
5102: to appear.}
5103: %
5104: \rife{BL}{6}{
5105: G.D. Birkhoff, D.C. Lewis,
5106: {\it On the periodic motions near a given periodic motion
5107: of a dynamical system},
5108: Ann. Math.
5109: {\bf 12} (1933), 117--133. }
5110: %
5111: \rife{Bo1}{7}{
5112: J. Bourgain,
5113: {\it Construction of quasi-periodic solutions for Hamiltonian
5114: perturbations of linear equations and applications to nonlinear PDE},
5115: Internat. Math. Res. Notices
5116: {\bf 1994}, no. 11, 475ff., approx. 21 pp. (electronic). }
5117: %
5118: \rife{Bo2}{8}{
5119: J. Bourgain,
5120: {\it Construction of periodic solutions of nonlinear
5121: wave equations in higher dimension},
5122: Geom. Funct. Anal.
5123: {\bf 5} (1995), 629--639. }
5124: %
5125: \rife{Bo3}{9}{
5126: J. Bourgain,
5127: {\it Quasi-periodic solutions of Hamiltonian perturbations of 2D
5128: linear Schr\"o\-din\-ger equations},
5129: Ann. of Math. (2)
5130: {\bf 148} (1998), no. 2, 363--439. }
5131: %
5132: \rife{BCN}{10}{
5133: H. Br\'ezis, J.-M. Coron, L. Nirenberg,
5134: {\it Free vibrations for a nonlinear wave equation and a theorem
5135: of P. Rabinowitz},
5136: Comm. Pure Appl. Math.
5137: {\bf 33} (1980), no. 5, 667--684. }
5138: %
5139: \rife{BN}{11}{
5140: H. Br\'ezis, L. Nirenberg,
5141: {\it Forced vibrations for a nonlinear wave equation},
5142: Comm. Pure Appl. Math.
5143: {\bf 31} (1978), no. 1, 1--30. }
5144: %
5145: \rife{Cr}{12}{
5146: W. Craig,
5147: {\it Probl\`emes de petits diviseurs dans les \'equations aux
5148: d\'eriv\'ees partielles},
5149: Pano\-ra\-mas and Syntheses 9. Soci\'et\'e Math\'ematique de France,
5150: Paris, 2000. }
5151: %
5152: \rife{CW1}{13}{
5153: W. Craig and C.E. Wayne,
5154: {\it Newton's method and periodic solutions of nonlinear
5155: wave equations},
5156: Comm. Pure Appl. Math.
5157: {\bf 46} (1993), 1409--1498. }
5158: %
5159: \rife{CW2}{14}{
5160: W. Craig and C.E. Wayne,
5161: {\it Nonlinear waves and the $1:1:2$ resonance},
5162: ``Singular limits of dispersive waves'' (Lyon, 1991), 297--313,
5163: NATO Adv. Sci. Inst. Ser. B Phys., 320, Plenum, New York, 1994. } %
5164: \rife{FR}{15}{
5165: E.R. Fadell, P.H. Rabinowitz,
5166: {\it Generalized cohomological index theories for Lie group
5167: actions with an application to bifurcation questions
5168: for Hamiltonian systems},
5169: Invent. Math.
5170: {\bf 45} (1978), no. 2, 139--174. }
5171: %
5172: \rife{GG}{16}{
5173: G. Gallavotti, G. Gentile,
5174: {\it Hyperbolic low-dimensional invariant tori and
5175: summations of divergent series},
5176: Comm. Math. Phys.
5177: {\bf 227} (2002), no. 3, 421--460. }
5178: %
5179: \rife{Ge}{17}{
5180: G. Gentile,
5181: {\it Whiskered tori with prefixed frequencies and Lyapunov spectrum},
5182: Dynam. Stability Systems
5183: {\bf 10} (1995), no. 3, 269--308. }
5184: %
5185: \rife{GGM}{18}{
5186: G. Gallavotti, G. Gentile, V. Mastropietro,
5187: {\it A field theory approach to Lindstedt series for hyperbolic
5188: tori in three time scales problems},
5189: J. Math. Phys.
5190: {\bf 40} (1999), no. 12, 6430--6472. }
5191: %
5192: \rife{GM}{19}{
5193: G. Gentile and V. Mastropietro,
5194: {\it Construction of periodic solutions of the nonlinear wave equation
5195: with Dirichlet boundary conditions by the Linsdtedt series method},
5196: J. Math. Pures Appl. (9),
5197: to appear. }
5198: %
5199: \rife{GoR}{20}{
5200: C. Godsil, G. Royle,
5201: {\it Algebraic graph theory},
5202: Graduate Texts in Mathematics 207.
5203: Spring\-er, New York, 2001. }
5204: %
5205: \rife{GrR}{21}{
5206: I.S. Gradshteyn, I.M. Ryzhik,
5207: {\it Table of integrals, series, and products},
5208: Sixth edition, Academic Press, Inc., San Diego, 2000. }
5209: %
5210: \rife{HP}{22}{
5211: F. Harary, E.M. Palmer,
5212: {\it Graphical enumeration},
5213: Academic Press, New York-London, 1973. }
5214: %
5215: \rife{K1}{23}{
5216: S.B. Kuksin,
5217: {\it Nearly integrable infinite-dimensional Hamiltonian systems},
5218: Lecture Notes in Mathematics 1556, Springer, Berlin, 1994. }
5219: %
5220: \rife{K2}{24}{
5221: S.B. Kuksin,
5222: {\it Fifteen years of KAM for PDE},
5223: Preprint, 2003. }
5224: %
5225: \rife{KP}{25}{
5226: S.B. Kuksin, J.P\"oschel,
5227: {Invariant Cantor manifolds of quasi-periodic oscillations
5228: for a nonlinear Schr\"odinger equation},
5229: Ann. of Math. (2)
5230: {\bf 143} (1996), no. 1, 149--179. }
5231: %
5232: \rife{LS}{26}{
5233: B.V. Lidski\u\i, E.I. Shul$^{\prime}$man,
5234: {\it Periodic solutions of the equation $u_{tt}-u_{xx}+u^{3}=0$},
5235: Funct. Anal. Appl.
5236: {\bf 22} (1988), no. 4, 332--333. }
5237: %
5238: \rife{L}{27}{
5239: A. M. Lyapunov,
5240: {\it Probl\`eme g\'en\'eral de la stabilit\'e du mouvement},
5241: Ann. Sc. Fac. Toulouse
5242: {\bf 2} (1907), 203--474. }
5243: %
5244: \rife{PBC}{28}{
5245: S. Paleari, D. Bambusi, S. Cacciatori,
5246: {\it Normal form and exponential stability for some
5247: nonlinear string equations},
5248: Z. Angew. Math. Phys.
5249: {\bf 52} (2001), no. 6, 1033--1052. }
5250: %
5251: \rife{P}{29}{
5252: J. P\"oschel,
5253: {\it Quasi-periodic solutions for a nonlinear wave equation},
5254: Comment. Math. Helv.
5255: {\bf 71} (1996), no. 2, 269--296. }
5256: %
5257: \rife{R1}{30}{
5258: H.P. Rabinowitz,
5259: {\it Periodic solutions of nonlinear hyperbolic
5260: partial differential equations},
5261: Comm. Pure Appl. Math.
5262: {\bf 20} (1967), 145--205. }
5263: %
5264: \rife{R2}{31}{
5265: P.H. Rabinowitz,
5266: {\it Free vibrations for a semilinear wave equation},
5267: Comm. Pure Appl. Math.
5268: {\bf 31} (1978), no. 1, 31--68. }
5269: %
5270: \rife{Wa}{32}{
5271: C.E. Wayne,
5272: {\it Periodic and quasi-periodic solutions of nonlinear wave
5273: equations via KAM theory},
5274: Comm. Math. Phys.
5275: {\bf 127} (1990), no. 3, 479--528. }
5276: %
5277: \rife{We}{33}{
5278: A. Weinstein,
5279: {\it Normal modes for nonlinear Hamiltonian systems}, Invent. Math.
5280: {\bf 20} (1973), 47--57. }
5281: %
5282: \rife{Wh}{34}{
5283: H. Whitney,
5284: {\it Analytic extensions of differential
5285: functions defined in closed sets},
5286: Trans. Amer. Math. Soc.
5287: {\bf 36} (1934), no. 1, 63--89. }
5288: \biblio
5289: \ciao
5290:
5291:
5292: