1: \ifx\shlhetal\undefinedcontrolsequence\let\shlhetal\relax\fi
2: % To: shani@math.huji.ac.il, "Saharon Shelah's Office" <shlhetal@math.huji.ac.il>, Saharon Shelah <shelah@math.huji.ac.il>
3: % Subject: from scanned alex pages-revisions f703
4: % Date: Mon, 27 Nov 2006 10:42:03 -0500 (EST)
5: % From: Alice Leonhardt <leonhard@math.rutgers.edu>
6: % Mime-Version: 1.0
7: % Content-Description: revisions
8: % X-sliced-and-diced-by: 'savemail' 2.0, Oct 2005
9:
10: \input amstex
11: % % \input mathdefs
12: % *** start including mathdefs.tex ***
13: \expandafter\ifx\csname mathdefs.tex\endcsname\relax
14: \expandafter\gdef\csname mathdefs.tex\endcsname{}
15: \else \message{Hey! Apparently you were trying to
16: \string\input{mathdefs.tex} twice. This does not make sense.}
17: \errmessage{Please edit your file (probably \jobname.tex) and remove
18: any duplicate ``\string\input'' lines}\endinput\fi
19:
20:
21: %mathdefs.tex v1.3.2
22:
23: %%% Changes from v1.0: footnote macros, warning for duplicated tags,
24: %%% control sequences \( and \verbatimtags.
25: %%% From v1.2: \pretags, redefinition of \( using \ifinner, multi-part
26: %%% equation numbering, control sequences \[, \references, and
27: %%% \resetbracket.
28: %%% From v1.3: \rm in \lastpart; write root of multi-part tag to .tgs
29:
30: %See file texdefs.doc for documentation.
31:
32: \catcode`\X=12\catcode`\@=11
33:
34: %Minor control sequences:
35: \def\n@wcount{\alloc@0\count\countdef\insc@unt}
36: \def\n@wwrite{\alloc@7\write\chardef\sixt@@n}
37: \def\n@wread{\alloc@6\read\chardef\sixt@@n}
38: \def\r@s@t{\relax}\def\v@idline{\par}\def\@mputate#1/{#1}
39: \def\l@c@l#1X{\firstpart.#1}\def\gl@b@l#1X{#1}\def\t@d@l#1X{{}}
40:
41: %Creation of tag families and output of assignments and citations:
42: \def\crossrefs#1{\ifx\all#1\let\tr@ce=\all\else\def\tr@ce{#1,}\fi
43: \n@wwrite\cit@tionsout\openout\cit@tionsout=\jobname.cit
44: \write\cit@tionsout{\tr@ce}\expandafter\setfl@gs\tr@ce,}
45: \def\setfl@gs#1,{\def\@{#1}\ifx\@\empty\let\next=\relax
46: \else\let\next=\setfl@gs\expandafter\xdef
47: \csname#1tr@cetrue\endcsname{}\fi\next}
48: \def\m@ketag#1#2{\expandafter\n@wcount\csname#2tagno\endcsname
49: \csname#2tagno\endcsname=0\let\tail=\all\xdef\all{\tail#2,}
50: \ifx#1\l@c@l\let\tail=\r@s@t\xdef\r@s@t{\csname#2tagno\endcsname=0\tail}\fi
51: \expandafter\gdef\csname#2cite\endcsname##1{\expandafter
52: \ifx\csname#2tag##1\endcsname\relax?\else\csname#2tag##1\endcsname\fi
53: \expandafter\ifx\csname#2tr@cetrue\endcsname\relax\else
54: \write\cit@tionsout{#2tag ##1 cited on page \folio.}\fi}
55: \expandafter\gdef\csname#2page\endcsname##1{\expandafter
56: \ifx\csname#2page##1\endcsname\relax?\else\csname#2page##1\endcsname\fi
57: \expandafter\ifx\csname#2tr@cetrue\endcsname\relax\else
58: \write\cit@tionsout{#2tag ##1 cited on page \folio.}\fi}
59: \expandafter\gdef\csname#2tag\endcsname##1{\expandafter
60: \ifx\csname#2check##1\endcsname\relax
61: \expandafter\xdef\csname#2check##1\endcsname{}%
62: \else\immediate\write16{Warning: #2tag ##1 used more than once.}\fi
63: \multit@g{#1}{#2}##1/X%
64: \write\t@gsout{#2tag ##1 assigned number \csname#2tag##1\endcsname\space
65: on page \number\count0.}%
66: \csname#2tag##1\endcsname}}
67:
68:
69: \def\multit@g#1#2#3/#4X{\def\t@mp{#4}\ifx\t@mp\empty%
70: \global\advance\csname#2tagno\endcsname by 1
71: \expandafter\xdef\csname#2tag#3\endcsname
72: {#1\number\csname#2tagno\endcsnameX}%
73: \else\expandafter\ifx\csname#2last#3\endcsname\relax
74: \expandafter\n@wcount\csname#2last#3\endcsname
75: \global\advance\csname#2tagno\endcsname by 1
76: \expandafter\xdef\csname#2tag#3\endcsname
77: {#1\number\csname#2tagno\endcsnameX}
78: \write\t@gsout{#2tag #3 assigned number \csname#2tag#3\endcsname\space
79: on page \number\count0.}\fi
80: \global\advance\csname#2last#3\endcsname by 1
81: \def\t@mp{\expandafter\xdef\csname#2tag#3/}%
82: \expandafter\t@mp\@mputate#4\endcsname
83: {\csname#2tag#3\endcsname\lastpart{\csname#2last#3\endcsname}}\fi}
84: \def\t@gs#1{\def\all{}\m@ketag#1e\m@ketag#1s\m@ketag\t@d@l p
85: \let\realscite\scite
86: \let\realstag\stag
87: \m@ketag\gl@b@l r \n@wread\t@gsin
88: \openin\t@gsin=\jobname.tgs \re@der \closein\t@gsin
89: \n@wwrite\t@gsout\openout\t@gsout=\jobname.tgs }
90: \outer\def\localtags{\t@gs\l@c@l}
91: \outer\def\globaltags{\t@gs\gl@b@l}
92: \outer\def\newlocaltag#1{\m@ketag\l@c@l{#1}}
93: \outer\def\newglobaltag#1{\m@ketag\gl@b@l{#1}}
94:
95: %Reading in tag information:
96: \newif\ifpr@
97: \def\m@kecs #1tag #2 assigned number #3 on page #4.%
98: {\expandafter\gdef\csname#1tag#2\endcsname{#3}
99: \expandafter\gdef\csname#1page#2\endcsname{#4}
100: \ifpr@\expandafter\xdef\csname#1check#2\endcsname{}\fi}
101: \def\re@der{\ifeof\t@gsin\let\next=\relax\else
102: \read\t@gsin to\t@gline\ifx\t@gline\v@idline\else
103: \expandafter\m@kecs \t@gline\fi\let \next=\re@der\fi\next}
104: \def\pretags#1{\pr@true\pret@gs#1,,}
105: \def\pret@gs#1,{\def\@{#1}\ifx\@\empty\let\n@xtfile=\relax
106: \else\let\n@xtfile=\pret@gs \openin\t@gsin=#1.tgs \message{#1} \re@der
107: \closein\t@gsin\fi \n@xtfile}
108:
109: %Sections and subsections; local numbering:
110: \newcount\sectno\sectno=0\newcount\subsectno\subsectno=0
111: \newif\ifultr@local \def\ultralocal{\ultr@localtrue}
112: \def\firstpart{\number\sectno}
113: \def\lastpart#1{\ifcase#1 \or a\or b\or c\or d\or e\or f\or g\or h\or
114: i\or k\or l\or m\or n\or o\or p\or q\or r\or s\or t\or u\or v\or w\or
115: x\or y\or z \fi}
116: \def\closeup{\vskip-\bigskipamount}
117: \def\resetall{\global\advance\sectno by 1\subsectno=0
118: \gdef\firstpart{\number\sectno}\r@s@t}
119: \def\resetsub{\global\advance\subsectno by 1
120: \gdef\firstpart{\number\sectno.\number\subsectno}\r@s@t}
121: \def\newsection#1\par{\resetall\vskip0pt plus.3\vsize\penalty-250
122: \vskip0pt plus-.3\vsize\bigskip\bigskip
123: \message{#1}\leftline{\bf#1}\nobreak\bigskip}
124: \def\subsection#1\par{\ifultr@local\resetsub\fi
125: \vskip0pt plus.2\vsize\penalty-250\vskip0pt plus-.2\vsize
126: \bigskip\smallskip\message{#1}\leftline{\bf#1}\nobreak\medskip}
127:
128: %jj tags:
129: % On Andrzej's request: we want to be able
130: % to show tags as in noverbatim, with verbatim in the margin,
131: % and cites as in verbatim, with nonverbatim in the margin
132: % mg -- July 2000
133:
134: \newdimen\marginshift
135:
136: \newdimen\margindelta
137: \newdimen\marginmax
138: \newdimen\marginmin
139:
140: \def\margininit{
141: \marginmax=3 true cm % how much room, approximately
142:
143: \margindelta=0.1 true cm % distance between entries
144: \marginmin=0.1true cm % where will leftmost entry be
145: \marginshift=\marginmin
146: } % we cannot execute this right now, since
147: % there may be a \magnification coming later in the
148: % main file. So we call \margininit at the end of
149: % alice2jlem
150:
151: \def\t@gsjj#1,{\def\@{#1}\ifx\@\empty\let\next=\relax\else\let\next=\t@gsjj
152: \def\@@{p}\ifx\@\@@\else
153: \expandafter\gdef\csname#1cite\endcsname##1{\citejj{##1}}
154: \expandafter\gdef\csname#1page\endcsname##1{?}
155: \expandafter\gdef\csname#1tag\endcsname##1{\tagjj{##1}}\fi\fi\next}
156: \newif\ifshowstuffinmargin
157: \showstuffinmarginfalse
158: \def\jjtags{\ifx\shlhetal\relax
159: % so this is a public version --> no-op
160: \else
161: \ifx\shlhetal\undefinedcontrolseq
162: % again, this is a public version --> no-op
163: \else
164: \showstuffinmargintrue
165: \ifx\all\relax\else\expandafter\t@gsjj\all,\fi\fi \fi
166: }
167:
168: % \def\zeigentag#1{\def\xxy{#1}\show\xxy\show\realstag\realstag#1\mginpar{\hbox{$\langle$}#1\hbox{$\rangle$}}}
169: % \def\zeigencite#1{\hbox{$\langle$}#1\hbox{$\rangle$}\mginpar{\realscite#1}}
170:
171:
172: \def\tagjj#1{\realstag{#1}\oldmginpar{\zeigen{#1}}}
173: % \def\citejj#1{\zeigen{#1}\mginpar{\rechnen{#1}}}
174: \def\citejj#1{\rechnen{#1}\mginpar{\zeigen{#1}}} % modified Sep 02, saharon's suggestion
175:
176: % \def\rechnen#1{rechnen(\expandafter\ifx\csname stag#1\endcsname\relax ??\else
177: % \csname stag#1\endcsname\fi)endrech }
178: \def\rechnen#1{\expandafter\ifx\csname stag#1\endcsname\relax ??\else
179: \csname stag#1\endcsname\fi}
180:
181: % \def\zeigentag#1{//#1--\csname stag#1\endcsname::}
182:
183: \newdimen\theight
184:
185:
186:
187: \def\marginfont{\sevenrm}
188:
189:
190:
191: \def\trymarginbox#1{\setbox0=\hbox{\marginfont\hskip\marginshift #1}%
192: \global\marginshift\wd0
193: \global\advance\marginshift\margindelta}
194:
195:
196: \def \oldmginpar#1{%
197: \ifvmode\setbox0\hbox to \hsize{\hfill\rlap{\marginfont\quad#1}}%
198: \ht0 0cm
199: \dp0 0cm
200: \box0\vskip-\baselineskip
201: \else
202: \vadjust{\trymarginbox{#1}%
203: \ifdim\marginshift>\marginmax \global\marginshift\marginmin
204: \trymarginbox{#1}%
205: \fi
206: \theight=\ht0
207: \advance\theight by \dp0 \advance\theight by \lineskip
208: \kern -\theight \vbox to \theight{\rightline{\rlap{\box0}}%
209: \vss}}\fi}
210:
211: \newdimen\upordown
212: \global\upordown=8pt
213: \font\tinyfont=cmtt8 % scaled 700
214: \def\mginpar#1{\smash{\hbox to 0cm{\kern-10pt\raise7pt\hbox{\tinyfont #1}\hss}}}
215: % testing, october 2005, mg
216: \def\mginpar#1{{\hbox to 0cm{\kern-10pt\raise\upordown\hbox{\tinyfont #1}\hss}}\global\upordown-\upordown}
217:
218:
219: % \def\mginpar#1{mg-#1-mg }
220:
221:
222:
223: %Verbatim tags:
224: \def\t@gsoff#1,{\def\@{#1}\ifx\@\empty\let\next=\relax\else\let\next=\t@gsoff
225: \def\@@{p}\ifx\@\@@\else
226: \expandafter\gdef\csname#1cite\endcsname##1{\zeigen{##1}}
227: \expandafter\gdef\csname#1page\endcsname##1{?}
228: \expandafter\gdef\csname#1tag\endcsname##1{\zeigen{##1}}\fi\fi\next}
229: \def\verbatimtags{\showstuffinmarginfalse
230: \ifx\all\relax\else\expandafter\t@gsoff\all,\fi}
231: %%% \def\zeigen#1{zeigen(\hbox{$\langle$}#1\hbox{$\rangle$})endz }
232: \def\zeigen#1{\hbox{$\scriptstyle\langle$}#1\hbox{$\scriptstyle\rangle$}}
233:
234: % % \def\margincite#1{\ifshowstuffinmargin\mginpar{\rechnen{#1}}\fi}
235: % changed, april 2003, mg: we now have always the verbatim tag in the margin
236: \def\margincite#1{\ifshowstuffinmargin\mginpar{\zeigen{#1}}\fi}
237:
238: \def\margintag#1{\ifshowstuffinmargin\oldmginpar{\zeigen{#1}}\fi}
239:
240: \def\marginplain#1{\ifshowstuffinmargin\mginpar{{#1}}\fi}
241: \def\marginbf#1{\marginplain{{\bf \ \ #1}}}
242:
243:
244: %Equation numbering:
245: \def\(#1){\edef\dot@g{\ifmmode\ifinner(\hbox{\noexpand\etag{#1}})
246: \else\noexpand\eqno(\hbox{\noexpand\etag{#1}})\fi
247: \else(\noexpand\ecite{#1})\fi}\dot@g}
248:
249: %Reference numbering:
250: \newif\ifbr@ck
251: \def\eat#1{}
252: \def\[#1]{\br@cktrue[\br@cket#1'X]}
253: \def\br@cket#1'#2X{\def\temp{#2}\ifx\temp\empty\let\next\eat
254: \else\let\next\br@cket\fi
255: \ifbr@ck\br@ckfalse\br@ck@t#1,X\else\br@cktrue#1\fi\next#2X}
256: \def\br@ck@t#1,#2X{\def\temp{#2}\ifx\temp\empty\let\neext\eat
257: \else\let\neext\br@ck@t\def\temp{,}\fi
258: \def\teemp{#1}\ifx\teemp\empty\else\rcite{#1}\fi\temp\neext#2X}
259: \def\resetbr@cket{\gdef\[##1]{[\rtag{##1}]}}
260: \def\references{\resetbr@cket\newsection References\par}
261:
262: %Footnotes:
263: \newtoks\symb@ls\newtoks\s@mb@ls\newtoks\p@gelist\n@wcount\ftn@mber
264: \ftn@mber=1\newif\ifftn@mbers\ftn@mbersfalse\newif\ifbyp@ge\byp@gefalse
265: \def\defm@rk{\ifftn@mbers\n@mberm@rk\else\symb@lm@rk\fi}
266: \def\n@mberm@rk{\xdef\m@rk{{\the\ftn@mber}}%
267: \global\advance\ftn@mber by 1 }
268: \def\rot@te#1{\let\temp=#1\global#1=\expandafter\r@t@te\the\temp,X}
269: \def\r@t@te#1,#2X{{#2#1}\xdef\m@rk{{#1}}}
270: \def\b@@st#1{{$^{#1}$}}\def\str@p#1{#1}
271: \def\symb@lm@rk{\ifbyp@ge\rot@te\p@gelist\ifnum\expandafter\str@p\m@rk=1
272: \s@mb@ls=\symb@ls\fi\write\f@nsout{\number\count0}\fi \rot@te\s@mb@ls}
273: \def\byp@ge{\byp@getrue\n@wwrite\f@nsin\openin\f@nsin=\jobname.fns
274: \n@wcount\currentp@ge\currentp@ge=0\p@gelist={0}
275: \re@dfns\closein\f@nsin\rot@te\p@gelist
276: \n@wread\f@nsout\openout\f@nsout=\jobname.fns }
277: \def\m@kelist#1X#2{{#1,#2}}
278: \def\re@dfns{\ifeof\f@nsin\let\next=\relax\else\read\f@nsin to \f@nline
279: \ifx\f@nline\v@idline\else\let\t@mplist=\p@gelist
280: \ifnum\currentp@ge=\f@nline
281: \global\p@gelist=\expandafter\m@kelist\the\t@mplistX0
282: \else\currentp@ge=\f@nline
283: \global\p@gelist=\expandafter\m@kelist\the\t@mplistX1\fi\fi
284: \let\next=\re@dfns\fi\next}
285: \def\symbols#1{\symb@ls={#1}\s@mb@ls=\symb@ls}
286: \def\bigsymbol{\textstyle}
287: \symbols{\bigsymbol\ast,\dagger,\ddagger,\sharp,\flat,\natural,\star}
288: \def\ftnumbers{\ftn@mberstrue} \def\ftsymbols{\ftn@mbersfalse}
289: \def\paginal{\byp@ge} \def\resetftnumbers{\ftn@mber=1}
290: \def\ftnote#1{\defm@rk\expandafter\expandafter\expandafter\footnote
291: \expandafter\b@@st\m@rk{#1}}
292:
293: %Miscellaneous macros:
294: \long\def\jump#1\endjump{}
295: \def\ssum{\mathop{\lower .1em\hbox{$\textstyle\Sigma$}}\nolimits}
296: \def\down#1{_{{}_{\scriptstyle #1}}}
297: \def\qed{\nobreak\kern 1em \vrule height .5em width .5em depth 0em}
298: \def\newneq{\hbox{\rlap{\hbox to 1\wd9{\hss$=$\hss}}\raise .1em
299: \hbox to 1\wd9{\hss$\scriptscriptstyle/$\hss}}}
300: \def\subsetne{\setbox9 = \hbox{$\subset$}\mathrel{\hbox{\rlap
301: {\lower .4em \newneq}\raise .13em \hbox{$\subset$}}}}
302: \def\supsetne{\setbox9 = \hbox{$\subset$}\mathrel{\hbox{\rlap
303: {\lower .4em \newneq}\raise .13em \hbox{$\supset$}}}}
304:
305: %Blackboard bold:
306: \def\vbar{\mathchoice{\vrule height6.3ptdepth-.5ptwidth.8pt\kern-.8pt}
307: {\vrule height6.3ptdepth-.5ptwidth.8pt\kern-.8pt}
308: {\vrule height4.1ptdepth-.35ptwidth.6pt\kern-.6pt}
309: {\vrule height3.1ptdepth-.25ptwidth.5pt\kern-.5pt}}
310: \def\f@dge{\mathchoice{}{}{\mkern.5mu}{\mkern.8mu}}
311: \def\b@c#1#2{{\rm \mkern#2mu\vbar\mkern-#2mu#1}}
312: \def\b@b#1{{\rm I\mkern-3.5mu #1}}
313: \def\b@a#1#2{{\rm #1\mkern-#2mu\f@dge #1}}
314: \def\bb#1{{\count4=`#1 \advance\count4by-64 \ifcase\count4\or\b@a A{11.5}\or
315: \b@b B\or\b@c C{5}\or\b@b D\or\b@b E\or\b@b F \or\b@c G{5}\or\b@b H\or
316: \b@b I\or\b@c J{3}\or\b@b K\or\b@b L \or\b@b M\or\b@b N\or\b@c O{5} \or
317: \b@b P\or\b@c Q{5}\or\b@b R\or\b@a S{8}\or\b@a T{10.5}\or\b@c U{5}\or
318: \b@a V{12}\or\b@a W{16.5}\or\b@a X{11}\or\b@a Y{11.7}\or\b@a Z{7.5}\fi}}
319:
320: \catcode`\X=11 \catcode`\@=12
321:
322:
323: % Sep 2003, mg:
324:
325: % % \input 300stuff
326: % *** start including 300stuff.tex ***
327:
328: % definitions needed to process the 300x papers (plus 88 etc)
329:
330: \let\thischap\jobname
331:
332:
333:
334: \def\partof#1{\csname returnthe#1part\endcsname}
335: \def\chapof#1{\csname returnthe#1chap\endcsname}
336:
337: \def\setchapter#1,#2,#3;{%
338: \expandafter\def\csname returnthe#1part\endcsname{#2}%
339: \expandafter\def\csname returnthe#1chap\endcsname{#3}%
340: }
341:
342: \setchapter 300a,A,II.A;
343: \setchapter 300b,A,II.B;
344: \setchapter 300c,A,II.C;
345: \setchapter 300d,A,II.D;
346: \setchapter 300e,A,II.E;
347: \setchapter 300f,A,II.F;
348: \setchapter 300g,A,II.G;
349: \setchapter E53,B,N;
350: \setchapter 88r,B,I;
351: \setchapter 600,B,III;
352: \setchapter 705,B,IV;
353: \setchapter 734,B,V;
354:
355:
356: % \def\cprefix#1{% \cprefix{300b} generates A.II. or II. or "nothing"
357: % %\leavevmode\vrule width 2cm height 1cm depth 0cm
358: % \edef\theotherpart{\partof{#1}}\edef\theotherchap{\chapof{#1}}%
359: % \ifx\theotherpart\thispart
360: % \ifx\theotherchap\thischap % nothing
361: % \else % same part, different chap
362: % \theotherchap%
363: % \fi
364: % \else % different part
365: % \theotherpart.\theotherchap\fi}
366:
367:
368: \def\cprefix#1{% \cprefix{300b} generates II. or "nothing"
369: %\leavevmode\vrule width 2cm height 1cm depth 0cm
370: \edef\theotherpart{\partof{#1}}\edef\theotherchap{\chapof{#1}}%
371: \ifx\theotherpart\thispart
372: \ifx\theotherchap\thischap % nothing
373: \else % same part, different chap
374: \theotherchap%
375: \fi
376: \else % different part
377: \theotherchap\fi}
378:
379:
380: % \sectioncite[\S4]{300b} -> II.\S4 or A.II.\S4
381: \def\sectioncite[#1]#2{%
382: \cprefix{#2}#1}
383:
384: \def\chaptercite#1{Chapter \cprefix{#1}}
385:
386: % now define \thispart
387: \edef\thispart{\partof{\thischap}}
388: \edef\thischap{\chapof{\thischap}}
389:
390: \def\lastpage of '#1' is #2.{\expandafter\def\csname lastpage#1\endcsname{#2}}
391:
392: % *** end including 300stuff.tex ***
393:
394: \def\spuriousreset{}
395:
396: % *** end including mathdefs.tex ***
397: % % \input citeadd
398: % *** start including citeadd.tex ***
399: % citeadd -- a few additions for
400: % files from alice that were procesed with "citealice"
401:
402: \expandafter\ifx\csname citeadd.tex\endcsname\relax
403: \expandafter\gdef\csname citeadd.tex\endcsname{}
404: \else \message{Hey! Apparently you were trying to
405: \string\input{citeadd.tex} twice. This does not make sense.}
406: \errmessage{Please edit your file (probably \jobname.tex) and remove
407: any duplicate ``\string\input'' lines}\endinput\fi
408:
409:
410:
411:
412:
413: \def\sciteu{\sciteerror{undefined}}
414: \def\sciteuphantom{\complainaboutcitation{undefined}}
415:
416: \def\scitet{\sciteerror{ambiguous}}
417: \def\scitetphantom{\complainaboutcitation{ambiguous}}
418:
419:
420:
421:
422: \def\sciteerror#1#2{{\mathortextbf{\scite{#2}}}\complainaboutcitation{#1}{#2}}
423: \def\mathortextbf#1{\hbox{\bf #1}}
424: \def\complainaboutcitation#1#2{%
425: \vadjust{\line{\llap{---$\!\!>$ }\qquad scite$\{$#2$\}$ #1\hfil}}}
426:
427:
428:
429:
430:
431: % *** end including citeadd.tex ***
432: \sectno=-1 % start with sect 0
433: \localtags
434: \jjtags
435: \NoBlackBoxes
436: \define\mr{\medskip\roster}
437: \define\sn{\smallskip\noindent}
438: \define\mn{\medskip\noindent}
439: \define\bn{\bigskip\noindent}
440: \define\ub{\underbar}
441: \define\wilog{\text{without loss of generality}}
442: \define\ermn{\endroster\medskip\noindent}
443: \define\dbca{\dsize\bigcap}
444: \define\dbcu{\dsize\bigcup}
445: \define \nl{\newline}
446: % (citealice) \define\red{RED STARTS HERE}
447: % (citealice) \define\endred{RED STOPS HERE}
448: % (citealice) \define\blue{BLUE STARTS HERE}
449: % (citealice) \define\endblue{BLUE STOPS HERE}
450: % (citealice) \define\green{GREEN STARTS HERE}
451: % (citealice) \define\endgreen{GREEN STOPS HERE}
452:
453: \magnification=\magstep 1
454: \documentstyle{amsppt}
455: % % \input alice2000
456: % *** start including alice2000.tex ***
457: % This file should be inputted whenever we use amsppt.sty and
458: % old tex.
459: % Here we redefine \subjclass (use 1991 instead of 2000, otherwise
460: % the following definition comes directly from
461: %%
462: %% `amsppt.sty', generated
463: %% on <1997/2/2> with the docstrip utility (2.2i).
464: %%
465: %% The original source files were:
466: %%
467: %% amsppt.doc
468: %%% ====================================================================
469: %%% @AMSTeX-style-file{
470: %%% filename = "amsppt.sty",
471: %%% version = "2.1h",
472: %%% date = "1997/02/02",
473: %%% time = "09:27:44 EST",
474: %%% checksum = "56844 3264 16617 137829",
475: %%% author = "American Mathematical Society",
476: %%% address = "PO Box 6248, Providence, RI 02940-6248, USA",
477: %%% telephone = "401-455-4080 or (in the USA) 800-321-4AMS",
478:
479: { % the braces make the catcode-change local.
480: \catcode`@11
481:
482: \ifx\alicetwothousandloaded@\relax
483: \endinput\else\global\let\alicetwothousandloaded@\relax\fi
484:
485:
486: \gdef\subjclass{\let\savedef@\subjclass
487: \def\subjclass##1\endsubjclass{\let\subjclass\savedef@
488: \toks@{\def\usualspace{{\rm\enspace}}\eightpoint}%
489: \toks@@{##1\unskip.}%
490: \edef\thesubjclass@{\the\toks@
491: \frills@{{\noexpand\rm2000 {\noexpand\it Mathematics Subject
492: Classification}.\noexpand\enspace}}%
493: \the\toks@@}}%
494: \nofrillscheck\subjclass}
495: }
496:
497: % *** end including alice2000.tex ***
498: % % \input alice2jlem
499: % *** start including alice2jlem.tex ***
500: %% # Keywords Input file to be used for texing Alice's files
501:
502: \expandafter\ifx\csname alice2jlem.tex\endcsname\relax
503: \expandafter\xdef\csname alice2jlem.tex\endcsname{\the\catcode`@}
504: \else \message{Hey! Apparently you were trying to
505: \string\input{alice2jlem.tex} twice. This does not make sense.}
506: \errmessage{Please edit your file (probably \jobname.tex) and remove
507: any duplicate ``\string\input'' lines}\endinput\fi
508:
509:
510:
511: % % \input bib4plain
512: % *** start including bib4plain.tex ***
513: \expandafter\ifx\csname bib4plain.tex\endcsname\relax
514: \expandafter\gdef\csname bib4plain.tex\endcsname{}
515: \else \message{Hey! Apparently you were trying to \string\input
516: bib4plain.tex twice. This does not make sense.}
517: \errmessage{Please edit your file (probably \jobname.tex) and remove
518: any duplicate ``\string\input'' lines}\endinput\fi
519:
520:
521: % This file should be inputted if you want to use
522: % bibtex fom within plain TeX.
523: \def\makeatletter{\catcode`\@11 } % Not really need for standard
524: \def\makeatother{\catcode`\@12 } % bibtex files, but these commands
525: \def\renewcommand{\newcommand} % are used in our literal-unsrt.bst
526: % and other files.
527: % The above lines were written by Martin.Goldstern@tuwien.ac.at
528: % The rest is just the file btxmac.tex.
529: % (except for one small change below, regarding undefined citations)
530:
531: %% @texfile{
532: %% author = "Karl Berry and Oren Patashnik",
533: %% version = "0.99j",
534: %% date = "14 Mar 1992",
535: %% filename = "btxmac.tex",
536: %% address = "Please use electronic mail",
537: %% checksum = "834 4503 33061",
538: %% email = "opbibtex@cs.stanford.edu",
539: %% codetable = "ISO/ASCII",
540: %% supported = "yes",
541: %% docstring = "Defines macros that make BibTeX work with plain TeX",
542: %% }
543: % BibTeX-for-TeX macros, version 0.99j, for BibTeX 0.99c, TeX 3.0 or later.
544: % Copyright (C) 1990--92 by Karl Berry and Oren Patashnik; all rights reserved.
545: % You may copy this file provided: that it's accompanied by the
546: % "BibTeXing" document, whose text is contained in the file `btxdoc.tex';
547: % that any documentation you write for these macros also gives a
548: % reference for "BibTeXing"; and that either you make absolutely no
549: % changes to your copy, or if you do make changes, (1) you name the file
550: % something other than `btxmac.tex' and you remove all occurrences of
551: % `btxmac.tex' from the file, (2) you put, somewhere in the first twenty
552: % lines of the file, your name, along with an electronic address at which
553: % others who might use the file may reach you, and (3) you remove each
554: % occurrence of Oren's name and electronic address from this file. These
555: % restrictions help ensure that all standard versions of these macros are
556: % identical, and that Oren doesn't get deluged with inappropriate e-mail.
557: %
558: % This file, btxmac.tex, contains TeX macros that allow BibTeX, a
559: % bibliography program that was originally designed for use with LaTeX,
560: % to work with plain TeX. Please report any bugs (outright goofs,
561: % improvable macros, misfeatures, or unclear documentation) to Oren
562: % Patashnik (opbibtex@cs.stanford.edu). These macros will become frozen
563: % shortly after BibTeX version 1.00 is released.
564: %
565: % AMS-TEX WARNING: We tried very hard, for version .99i of these macros,
566: % to make them compatible with AmS-TeX. We succeeded to the extent
567: % that, if you use one of the standard bibliography styles, you probably
568: % won't notice any problems with version 0.99i of btxmac.tex. But
569: % ultimately we failed, in that the inherent incompatibilities between
570: % plain TeX and AmS-TeX kept making these macros break, for certain
571: % inputs or certain styles. Examples: (1) AmS-TeX treats at-signs as
572: % special, in ways that plain TeX and LaTeX don't, so that, for example,
573: % you can't have any `@' characters in an argument to the \cite command,
574: % the way you can in TeX or LaTeX; (2) AmS-TeX decided that plain TeX's
575: % and LaTeX's macron-accent control sequence `\=' should be undefined;
576: % so you'll need to define `\=' to be `\B' to get the xampl.bib example
577: % suggested below to work with AmS-TeX; (3) AmS-TeX redefines the tie
578: % character `~' of plain TeX, and AmS-TeX's `amsppt' style redefines
579: % plain TeX's `\nobreak' macro, so that if you use an author-date style
580: % like `apalike' and you have a multiple-author reference for which the
581: % author-date style automatically produces a citation in the text like
582: % `(Jones et~al., 1992)' you will throw AmS-TeX's `amsppt' style into
583: % an infinite loop, exceeding its input stack size. In practice, such
584: % incompatibilities surface infrequently; but it is now clear to us that
585: % it's not worth the effort (perhaps it's not even possible) to make the
586: % btxmac.tex macros robust when used both with plain TeX and Ams-TeX.
587: % If the BibTeX/AmS-TeX results attainable with the current btxmac.tex
588: % macros are sufficient, fine. But if there's a demand for more robust
589: % BibTeX/AmS-TeX behavior, then someone who's very familiar with the
590: % AmS-TeX package should probably make an amsbtxmc.tex version of the
591: % macros (remembering to follow the copyright restrictions above).
592: % Until then, if you're an AmS-TeX user, or a LaTeX or plain TeX user
593: % sharing files with an AmS-TeX user, beware.
594: % END OF AMS-TEX WARNING.
595: %
596: % To use these macros you should be familiar with how BibTeX interacts
597: % with LaTeX, since BibTeX's interaction with TeX is very similar; that
598: % interaction is explained in the LaTeX manual. It also helps to
599: % have read "BibTeXing", the documentation that accompanies BibTeX.
600: % Then, if you want, you should redefine any of the macros that begin
601: % with `\bbl' or `\biblabel' or `\print' that you need to get formatting
602: % different from the default (the default settings are designed to
603: % accompany a bibliography style like BibTeX's standard style `plain').
604: % The macros you might want to change are described briefly a few
605: % paragraphs hence. [To get started without reading any documentation,
606: % try running the nine-line .tex file below through TeX and BibTeX.
607: % Remember the general scheme: Running (La)TeX writes information on
608: % the .aux (auxiliary) file; then running BibTeX reads information from
609: % the .aux, .bst (style), and .bib (database) files, and writes
610: % information (the bibliography) on a .bbl file; then running (La)TeX
611: % incorporates the bibliography; then running (La)TeX once more fixes
612: % the remaining forward references into the bibliography. Thus, to get
613: % everything incorporated into your output, you'll have to run (La)TeX,
614: % BibTeX, (La)TeX, (La)TeX. (Standup, sitdown, fight, fight, fight.)]
615: %
616: % These macros can stand alone or they can be \input into a macro
617: % package, like Eplain, that is sufficiently compatible with plain TeX.
618: % To use these macros to format the 0.99 version of the xampl.bib file
619: % that's distributed with BibTeX (that version of the file has no
620: % self-identification), you'll need to define \mbox, which is a LaTeX
621: % command, to be \hbox, as in the example below.
622: %
623: % Here's a nine-line plain TeX file for trying out btxmac.tex; of course
624: % you'll have to remove the comment characters at the beginning of each
625: % line, and, depending on your system, you might have to take steps so
626: % that BibTeX can "see" the files xampl.bib and plain.bst (BibTeX will
627: % give you two empty-field warning messages that you should ignore).
628: %
629: % \def\mbox#1{\leavevmode\hbox{#1}}
630: % \input btxmac
631: % \noindent This cites Aamport's gnominious article~\cite{article-full}.
632: % \medskip
633: % \leftline{\bf References}
634: % \nocite{*} % put all database entries into the reference list
635: % \bibliography{xampl} % specify the database files; here, just xampl.bib
636: % \bibliographystyle{plain} % specify plain.bst as the style file
637: % \bye
638: %
639: %
640: % HISTORY
641: %
642: % Karl Berry wrote the original version of these macros in 1989 and
643: % 1990, for use in his `Eplain' package. Oren Patashnik modified them
644: % slightly in July 1990, as part of the official BibTeX distribution.
645: %
646: % 1-Aug-90 Version 0.99a, not released to the general public.
647: % 14-Aug-90 0.99b, first general release.
648: % 26-Aug-90 0.99c, made \@undefinedmessage work with other macro packages.
649: % 6-Sep-90 0.99d, allowed for general formatting of bibliography labels,
650: % for general formatting of (in-text) citations, and for
651: % changing certain catcodes while reading the .aux file.
652: % 14-Nov-90 0.99e, changed the way \@setletters works, made some \new...'s
653: % non-outer, and changed the way Eplain reads this file.
654: % 12-Dec-90 0.99f, made \@resetnumerals change the `,' and `.' catcodes; and
655: % added \biblabelextrahang, \@getoptionalarg, and \bblsc.
656: % 11-Mar-91 0.99g, made a few minor changes required by the way Eplain reads
657: % this file, but no functional changes.
658: % 24-Apr-91 0.99h, inhibited the reading and writing of the .aux file if it
659: % isn't used or if the \noauxfile macro is defined, and
660: % removed some .aux-file-opening detritus; printed the
661: % cite-key of undefined citations in \tt font; changed the
662: % catcode of `_' inside \cite; and called \@resetnumerals
663: % from inside a group.
664: % 29-Feb-92 0.99i, made these macros semi-compatible with AmS-TeX; removed
665: % \@resetnumerals, \@setletters, \@tokstostring, and
666: % friends; changed the way \cite handles catcodes; changed
667: % \@getoptionalarg, and had \bibitem and \newcommand use
668: % it; added \@futurenonspacelet and (to facilitate the use
669: % of multiple reference lists) \bblfilebasename; changed
670: % \biblabelprint to use the new macros \biblabelprecontents
671: % and \biblabelpostcontents, and to, by default, right-
672: % justify numeric labels; and renamed \biblabelextrahang to
673: % the more descriptive \biblabelextraspace.
674: % 14-Mar-92 0.99j, made 0.99i's use of `\\' local to btxmac.tex.
675: %
676: %
677: % The LaTeX-related commands defined in this file include (a) the four
678: % commands that a user types (\bibliography, \bibliographystyle, \cite,
679: % and \nocite); (b) the three commands that BibTeX looks for in the .aux
680: % file (\bibdata, \bibstyle, and \citation---there is a fourth command
681: % that BibTeX looks for, but that command is related to LaTeX's \include
682: % facility, so these macros ignore that command); and (c) a LaTeX
683: % command (\newcommand) that's written by one of the four standard
684: % bibliography styles (alpha). The definitions here are not exactly the
685: % same as the corresponding LaTeX definitions (those eight LaTeX
686: % definitions depend on a significant fraction of LaTeX itself). But
687: % the only substantial differences are with \newcommand, which here,
688: % without complaining, lets you redefine a preexisting control sequence
689: % (in LaTeX, \newcommand won't let you redefine a preexisting command),
690: % and which here doesn't make the control sequences it defines \long (in
691: % LaTeX, that happens automatically); there may also be other minor
692: % differences. To summarize: Unless you know what you're doing, you
693: % shouldn't define any control sequences with these eight names:
694: %
695: % \bibdata
696: % \bibliography
697: % \bibliographystyle
698: % \bibstyle
699: % \citation
700: % \cite
701: % \newcommand
702: % \nocite
703: %
704: % There are three other commands written by one or more of the four
705: % standard (plain, abbrv, alpha, unsrt) or four semistandard (acm,
706: % apalike, ieeetr, siam) bibliography styles; they take effect only
707: % within the bibliography, and are redefinable, as explained later:
708: %
709: % \em
710: % \newblock
711: % \sc
712: %
713: % There's one control sequence you might want to use (but not redefine)
714: % in redefining \biblabelprint:
715: %
716: % \biblabelwidth
717: %
718: % There are fifteen other control sequences (explained later in more detail)
719: % that the macros of this file will use if you define them---you should
720: % define them after the \input btxmac command but before the \bibliography
721: % command. The first six begin with `\bbl' and affect fonts, spacing,
722: % perhaps other characteristics of the bibliography, and which .bbl files
723: % get read; the next five begin with `\biblabel' and determine how labels
724: % are formatted in the bibliography; and the last four begin with `\print'
725: % and determine how the in-text citations are formatted:
726: %
727: % \bblem
728: % \bblfilebasename
729: % \bblhook
730: % \bblnewblock
731: % \bblrm
732: % \bblsc
733: % \biblabelcontents
734: % \biblabelprecontents
735: % \biblabelprint
736: % \biblabelpostcontents
737: % \biblabelextraspace
738: % \printbetweencitations
739: % \printcitefinish
740: % \printcitenote
741: % \printcitestart
742: %
743: % If it's defined before the \input btxmac command, the control sequence
744: % below inhibits the reading and writing of the .aux file(s), and the
745: % issuing of related warning messages. Any definition will do. This
746: % feature might help when you're working on draft stages of a document:
747: %
748: % \noauxfile
749: %
750: %
751: % Here's another control sequence (it's described later) that you
752: % probably won't want to redefine unless you are writing another macro
753: % package; if you do redefine it, however, do it before the \input btxmac
754: % command (and notice that it has an `@' in its name):
755: %
756: % \@undefinedmessage
757: %
758: % Any other control sequence in this file that might conflict with
759: % something you've defined will have an `@' in its name, so such conflicts
760: % are unlikely; but if you're worried about a specific control sequence
761: % name, do a text search of this file to look for it.
762: %
763: %
764: % So to start things off we turn `@' into a letter (category code 11),
765: % keeping track of the old category code for future restoration.
766: % (Simply resetting it to 12 when we leave these macros is
767: % insufficient.) The use of `\cite' as a temporary control sequence is
768: % a kludge, but it's a reasonably simple way to accomplish what we need
769: % without possibly overwriting something (without an `@' in its name)
770: % that might already be defined.
771: %
772: \edef\cite{\the\catcode`@}%
773: \catcode`@ = 11
774: \let\@oldatcatcode = \cite
775: \chardef\@letter = 11
776: \chardef\@other = 12
777: %
778: %
779: % Next come some things that will be useful later.
780: %
781: % Make an outer definition into an inner one (due to Chris Thompson).
782: % The arguments should be the control sequence to be defined, and the
783: % new of the \outer control sequence, as characters; the control
784: % sequence #1 is defined to be just the same as \csname#2\endcsname, but
785: % not \outer. For example, \@innerdef\innernewcount{newcount} would
786: % define \innernewcount to be a non-outer version of \newcount.
787: %
788: \def\@innerdef#1#2{\edef#1{\expandafter\noexpand\csname #2\endcsname}}%
789: %
790: % We use \@innerdef to make some of our allocations local, because
791: % Eplain includes our code inside a conditional. We put @'s in the
792: % names to minimize the (already small) chance of conflicts.
793: %
794: \@innerdef\@innernewcount{newcount}%
795: \@innerdef\@innernewdimen{newdimen}%
796: \@innerdef\@innernewif{newif}%
797: \@innerdef\@innernewwrite{newwrite}%
798: %
799: %
800: % Swallow one parameter.
801: %
802: \def\@gobble#1{}%
803: %
804: %
805: % Use TeX 3.0's \inputlineno to get the line number, for better error
806: % messages, but if we're using an old version of TeX, don't do anything.
807: %
808: \ifx\inputlineno\@undefined
809: \let\@linenumber = \empty % Pre-3.0.
810: \else
811: \def\@linenumber{\the\inputlineno:\space}%
812: \fi
813: %
814: %
815: % The following macro \@futurenonspacelet (from the TeXbook) behaves
816: % essentially like \futurelet except that it discards any implicit or
817: % explicit space tokens that intervene before a nonspace is scanned:
818: %
819: \def\@futurenonspacelet#1{\def\cs{#1}%
820: \afterassignment\@stepone\let\@nexttoken=
821: }%
822: \begingroup % The grouping here avoids stepping on an outside use of `\\'.
823: \def\\{\global\let\@stoken= }%
824: \\ % now \@stoken is a space token (\\ is a control symbol, so that
825: % space after it is seen).
826: \endgroup
827: \def\@stepone{\expandafter\futurelet\cs\@steptwo}%
828: \def\@steptwo{\expandafter\ifx\cs\@stoken\let\@@next=\@stepthree
829: \else\let\@@next=\@nexttoken\fi \@@next}%
830: \def\@stepthree{\afterassignment\@stepone\let\@@next= }%
831: %
832: %
833: % \@getoptionalarg\CS gets an optional argument from the input, enclosed
834: % in brackets, then expands \CS. We set \@optionalarg to \empty if we
835: % don't find one, otherwise to the text of the argument. This assumes
836: % the brackets don't have a funny category code.
837: %
838: \def\@getoptionalarg#1{%
839: \let\@optionaltemp = #1%
840: \let\@optionalnext = \relax
841: \@futurenonspacelet\@optionalnext\@bracketcheck
842: }%
843: %
844: % The \expandafter's in this macro let us avoid the use of \aftergroup,
845: % which is somewhat more expensive.
846: %
847: \def\@bracketcheck{%
848: \ifx [\@optionalnext
849: \expandafter\@@getoptionalarg
850: \else
851: \let\@optionalarg = \empty
852: % We can't do the \temp after the \fi, because then the \temp gets
853: % in the way of reading the optional argument from the input, if
854: % we do have one.
855: \expandafter\@optionaltemp
856: \fi
857: }%
858: %
859: \def\@@getoptionalarg[#1]{%
860: \def\@optionalarg{#1}%
861: \@optionaltemp
862: }%
863: %
864: %
865: % From LaTeX.
866: %
867: \def\@nnil{\@nil}%
868: \def\@fornoop#1\@@#2#3{}%
869: %
870: \def\@for#1:=#2\do#3{%
871: \edef\@fortmp{#2}%
872: \ifx\@fortmp\empty \else
873: \expandafter\@forloop#2,\@nil,\@nil\@@#1{#3}%
874: \fi
875: }%
876: %
877: \def\@forloop#1,#2,#3\@@#4#5{\def#4{#1}\ifx #4\@nnil \else
878: #5\def#4{#2}\ifx #4\@nnil \else#5\@iforloop #3\@@#4{#5}\fi\fi
879: }%
880: %
881: \def\@iforloop#1,#2\@@#3#4{\def#3{#1}\ifx #3\@nnil
882: \let\@nextwhile=\@fornoop \else
883: #4\relax\let\@nextwhile=\@iforloop\fi\@nextwhile#2\@@#3{#4}%
884: }%
885: %
886: %
887: % This macro tests if a file \jobname.#1 exists, and sets \if@fileexists
888: % appropriately. If an optional argument is given, it is used as the
889: % root part of the filename instead of \jobname.
890: %
891: \@innernewif\if@fileexists
892: %
893: \def\@testfileexistence{\@getoptionalarg\@finishtestfileexistence}%
894: \def\@finishtestfileexistence#1{%
895: \begingroup
896: \def\extension{#1}%
897: \immediate\openin0 =
898: \ifx\@optionalarg\empty\jobname\else\@optionalarg\fi
899: \ifx\extension\empty \else .#1\fi
900: \space
901: \ifeof 0
902: \global\@fileexistsfalse
903: \else
904: \global\@fileexiststrue
905: \fi
906: \immediate\closein0
907: \endgroup
908: }%
909: %
910: %
911: %% [[[start of BibTeX-specific stuff]]]
912: %
913: % Now come the four main LaTeX commands and their associated .aux
914: % commands. Just as in LaTeX, \bibliographystyle defines the BibTeX
915: % style name (.bst file, that is), and \bibliography defines the
916: % database (.bib) file(s). The corresponding .aux-file commands are
917: % \bibstyle and \bibdata, which are there only for BibTeX's (but not
918: % LaTeX's) use.
919: %
920: \def\bibliographystyle#1{%
921: \@readauxfile
922: \@writeaux{\string\bibstyle{#1}}%
923: }%
924: \let\bibstyle = \@gobble
925: %
926: % As well as writing the \bibdata command to tell BibTeX which .bib
927: % files to read, we read the .bbl file that BibTeX (or a person,
928: % conceivably) has produced. We use \bblfilebasename as the root of the
929: % filename to read; this defaults to \jobname.
930: %
931: \let\bblfilebasename = \jobname
932: \def\bibliography#1{%
933: \@readauxfile
934: \@writeaux{\string\bibdata{#1}}%
935: \@testfileexistence[\bblfilebasename]{bbl}%
936: \if@fileexists
937: % We just output a non-discardable item (the `whatsit' with the
938: % \bibdata command). This means that the glue that will be
939: % inserted next (\parskip or \baselineskip, most likely) will be a
940: % legal breakpoint. Most likely, this is after some kind of
941: % heading, where we don't want to allow a page break. So:
942: \nobreak
943: \@readbblfile
944: \fi
945: }%
946: \let\bibdata = \@gobble
947: %
948: % The \nocite{label,label,...} command writes its argument to \@auxfile,
949: % unless instructed not to, but produces no text in the document. Both
950: % the \nocite and \cite commands produce \citation commands in the .aux file.
951: %
952: \def\nocite#1{%
953: \@readauxfile
954: \@writeaux{\string\citation{#1}}%
955: }%
956: %
957: \@innernewif\if@notfirstcitation
958: %
959: % \cite[note]{label,label,...} produces the citations for the labels as
960: % well. If the optional argument `note' is present, it's added after
961: % the labels. Since \cite calls \nocite to do its .aux-file writing,
962: % \cite doesn't need to call \@readauxfile (\nocite does).
963: %
964: \def\cite{\@getoptionalarg\@cite}%
965: %
966: % Typeset the citations for the labels in #1, followed by the note, if
967: % it exists. To change the citation's format in the text, redefine one
968: % or more `\print...' macros, whose defaults appear later in this file.
969: %
970: \def\@cite#1{%
971: % Remember the optional argument, in case one of the macros we call
972: % below ends up looking for an optional argument itself. For
973: % example, if a \cite[note] triggers reading the .aux file, then the
974: % [note] would be clobbered, since \@testfileexistence looks for an
975: % optional arg.
976: \let\@citenotetext = \@optionalarg
977: % Start printing the text, beginning with a left bracket by default.
978: \printcitestart
979: % It's complicated, but because \nocite puts a `whatsit' onto the list,
980: % \nocite should follow \printcitestart. It's conceivable, but very
981: % unlikely, that this `whatsit' will cause a problem (glue that doesn't
982: % disappear when you want it to is the most likely symptom), requiring
983: % a change either to \printcitestart or to the label that the .bst file
984: % produces.
985: \nocite{#1}%
986: \@notfirstcitationfalse
987: \@for \@citation :=#1\do
988: {%
989: \expandafter\@onecitation\@citation\@@
990: }%
991: \ifx\empty\@citenotetext\else
992: \printcitenote{\@citenotetext}%
993: \fi
994: \printcitefinish
995: }%
996: %
997: \newif\ifweareinprivate
998: \weareinprivatetrue
999: \ifx\shlhetal\undefinedcontrolseq\weareinprivatefalse\fi
1000: \ifx\shlhetal\relax\weareinprivatefalse\fi
1001: \def\@onecitation#1\@@{%
1002: \if@notfirstcitation
1003: \printbetweencitations
1004: \fi
1005: %
1006: \expandafter \ifx \csname\@citelabel{#1}\endcsname \relax
1007: \if@citewarning
1008: \message{\@linenumber Undefined citation `#1'.}%
1009: \fi
1010: % Give it a dummy definition:
1011: \ifweareinprivate
1012: \expandafter\gdef\csname\@citelabel{#1}\endcsname{%
1013: % Change: marginal remark added, goldstrn@math.huji.ac.il,
1014: % goldstern@tuwien.ac.at, May 1996 mg
1015: % !!! change !!!
1016: \strut
1017: \vadjust{\vskip-\dp\strutbox
1018: \vbox to 0pt{\vss\parindent0cm \leftskip=\hsize
1019: \advance\leftskip3mm
1020: \advance\hsize 4cm\strut\openup-4pt
1021: \rightskip 0cm plus 1cm minus 0.5cm ? #1 ?\strut}}
1022: {\tt
1023: \escapechar = -1
1024: \nobreak\hskip0pt\pfeilsw%\special{ps:100 0 0 setrgbcolor }%
1025: \expandafter\string\csname#1\endcsname
1026: %\special{ps:0 0 0 setrgbcolor }
1027: \pfeilso
1028: \nobreak\hskip0pt
1029: }%
1030: }%
1031: \else % ifweareinprivate = false
1032: \expandafter\gdef\csname\@citelabel{#1}\endcsname{%
1033: {\tt\expandafter\string\csname#1\endcsname}
1034: }%
1035: \fi % ifweareinprivate
1036: \fi
1037: % Now produce the text, whether it was undefined or not.
1038: \csname\@citelabel{#1}\endcsname
1039: \@notfirstcitationtrue
1040: }%
1041: %
1042: % Given a label `foo', the macro `\b@foo' is supposed to
1043: % hold the text that should be produced.
1044: %
1045: \def\@citelabel#1{b@#1}%
1046: %
1047: % So, how does a citation label get defined? When we read the .bbl file
1048: % (below), a \bibitem writes out a \@citedef command. And when we read
1049: % the \@citedef, we define \@citelabel{#1}, where #1 is the user's
1050: % label.
1051: %
1052: \def\@citedef#1#2{\expandafter\gdef\csname\@citelabel{#1}\endcsname{#2}}%
1053: %
1054: %
1055: % Reading the .bbl file also produces the typeset bibliography. Please
1056: % notice, however, that we do not produce the title for the references
1057: % (e.g., `References'), as LaTeX does. The formatting and spacing of
1058: % that title, whether it should go into the headline, and so on, are all
1059: % things determined by your format. We cannot know those things in
1060: % advance. If you wish, you can define \bblhook to produce the title.
1061: % Or just do it before the \bibliography command.
1062: %
1063: \def\@readbblfile{%
1064: % Define a counter to tell us which item number we are on, unless
1065: % we've already defined it (because the document has more than one
1066: % bibliography).
1067: \ifx\@itemnum\@undefined
1068: \@innernewcount\@itemnum
1069: \fi
1070: %
1071: \begingroup
1072: \def\begin##1##2{%
1073: % ##1 is just `thebibliography'.
1074: % ##2 is the widest label.
1075: % We set (new dimen) \biblabelwidth based on the widest label
1076: \setbox0 = \hbox{\biblabelcontents{##2}}%
1077: \biblabelwidth = \wd0
1078: }%
1079: \def\end##1{}% ##1 is `thebibliography' again.
1080: %
1081: % Here we have two possibilities:
1082: % \bibitem[typesetlabel]{citationlabel}
1083: % \bibitem{citationlabel}
1084: % If we have the second of these, the citations are numbered, starting
1085: % from one; we use our own count register \@itemnum for this.
1086: %
1087: \@itemnum = 0
1088: \def\bibitem{\@getoptionalarg\@bibitem}%
1089: \def\@bibitem{%
1090: \ifx\@optionalarg\empty
1091: \expandafter\@numberedbibitem
1092: \else
1093: \expandafter\@alphabibitem
1094: \fi
1095: }%
1096: \def\@alphabibitem##1{%
1097: % Need \xdef here for various reasons.
1098: \expandafter \xdef\csname\@citelabel{##1}\endcsname {\@optionalarg}%
1099: % Left-justify alpha labels, unless \biblabel{pre,post}contents
1100: % are already defined.
1101: \ifx\biblabelprecontents\@undefined
1102: \let\biblabelprecontents = \relax
1103: \fi
1104: \ifx\biblabelpostcontents\@undefined
1105: \let\biblabelpostcontents = \hss
1106: \fi
1107: \@finishbibitem{##1}%
1108: }%
1109: %
1110: \def\@numberedbibitem##1{%
1111: \advance\@itemnum by 1
1112: \expandafter \xdef\csname\@citelabel{##1}\endcsname{\number\@itemnum}%
1113: % Right-justify numeric labels, unless \biblabel{pre,post}contents
1114: % are already defined.
1115: \ifx\biblabelprecontents\@undefined
1116: \let\biblabelprecontents = \hss
1117: \fi
1118: \ifx\biblabelpostcontents\@undefined
1119: \let\biblabelpostcontents = \relax
1120: \fi
1121: \@finishbibitem{##1}%
1122: }%
1123: %
1124: \def\@finishbibitem##1{%
1125: \biblabelprint{\csname\@citelabel{##1}\endcsname}%
1126: \@writeaux{\string\@citedef{##1}{\csname\@citelabel{##1}\endcsname}}%
1127: \ignorespaces
1128: }%
1129: %
1130: % Do the printing (we're producing the bibliography, remember).
1131: %
1132: \let\em = \bblem
1133: \let\newblock = \bblnewblock
1134: \let\sc = \bblsc
1135: % Punctuation won't affect spacing;
1136: \frenchspacing
1137: % the penalties below are from LaTeX's [article,book,report].sty;
1138: \clubpenalty = 4000 \widowpenalty = 4000
1139: % the next two values come from LaTeX's \sloppy command;
1140: \tolerance = 10000 \hfuzz = .5pt
1141: \everypar = {\hangindent = \biblabelwidth
1142: \advance\hangindent by \biblabelextraspace}%
1143: \bblrm
1144: % the \parskip is a guess at what looks good;
1145: \parskip = 1.5ex plus .5ex minus .5ex
1146: % and the space between label and text comes from LaTeX's \labelsep.
1147: \biblabelextraspace = .5em
1148: \bblhook
1149: %
1150: \input \bblfilebasename.bbl
1151: \endgroup
1152: }%
1153: %
1154: % The widest label's width is useful for redefining \biblabelprint;
1155: % you redefine \biblabelwidth, in effect, by redefining the
1156: % \biblabelcontents macro that appears below. And \biblabelextraspace,
1157: % which is redefinable inside \bblhook, is added to \biblabelwidth to
1158: % determine the amount of hanging indentation.
1159: %
1160: \@innernewdimen\biblabelwidth
1161: \@innernewdimen\biblabelextraspace
1162: %
1163: % Now come the main macros that are related to the printing of the
1164: % bibliography. Since you might want to redefine them, they are given
1165: % default definitions outside of \@readbblfile.
1166: %
1167: % The first one controls the printing of a bibliography entry's label.
1168: % If you change it, make sure that it starts with something like
1169: % \noindent or \indent or \leavevmode that puts TeX into horizontal mode
1170: % (even if the label itself is empty); otherwise, the hanging
1171: % indentation will get messed up in certain circumstances.
1172: %
1173: \def\biblabelprint#1{%
1174: \noindent
1175: \hbox to \biblabelwidth{%
1176: \biblabelprecontents
1177: \biblabelcontents{#1}%
1178: \biblabelpostcontents
1179: }%
1180: \kern\biblabelextraspace
1181: }%
1182: %
1183: % If you are using numeric labels, and you want them left-justified
1184: % (numeric labels by default are right-justified), do something like:
1185: % \def\biblabelprecontents{\relax}
1186: % \def\biblabelpostcontents{\hss}
1187: %
1188: % By default the labels are typeset in \bblrm, and enclosed in brackets.
1189: %
1190: \def\biblabelcontents#1{{\bblrm [#1]}}%
1191: %
1192: % The main text, too, is typeset using \bblrm, which is \rm by default.
1193: %
1194: \def\bblrm{\rm}%
1195: %
1196: % Emphasis for producing, e.g., titles, is done with \it by default.
1197: %
1198: \def\bblem{\it}%
1199: %
1200: % Some styles use a caps-and-small-caps font for author names. LaTeX
1201: % defines an \sc command but plain TeX doesn't, so we need one here.
1202: % The definition below doesn't load the font unless it's needed, but it
1203: % tries to load only the 10pt version, because it might not exist at
1204: % other point sizes.
1205: %
1206: \def\bblsc{\ifx\@scfont\@undefined
1207: \font\@scfont = cmcsc10
1208: \fi
1209: \@scfont
1210: }%
1211: %
1212: % The major parts of an entry are separated with \bblnewblock. The
1213: % numbers below are taken from LaTeX's `article' style.
1214: %
1215: \def\bblnewblock{\hskip .11em plus .33em minus .07em }%
1216: %
1217: % Here's where you stick any other bibliography-formatting goodies, or
1218: % redefine the values above.
1219: %
1220: \let\bblhook = \empty
1221: %
1222: %
1223: % Here are the four default definitions for formatting the in-text
1224: % citations. These are what you redefine (after your \input btxmac but
1225: % before your \bibliography) to get parens instead of brackets, or
1226: % superscripts, or footnotes, or whatever.
1227: %
1228: \def\printcitestart{[}% left bracket
1229: \def\printcitefinish{]}% right bracket
1230: \def\printbetweencitations{, }% comma, space
1231: \def\printcitenote#1{, #1}% comma, space, note (if it exists)
1232: %
1233: % That scheme is pretty flexible. For example you could use
1234: % \def\printcitestart{\unskip $^\bgroup}
1235: % \def\printcitefinish{\egroup$}
1236: % \def\printbetweencitations{,}
1237: % \def\printcitenote#1{\hbox{\sevenrm\space (#1)}}
1238: % \font\eighttt = cmtt8
1239: % \scriptfont\ttfam = \eighttt
1240: % to get superscripted in-text citations. (The scriptfont stuff
1241: % exists only to print an undefined citation; it's in cmtt8 because
1242: % there is no cmtt7.) To get something radically different, however,
1243: % you'll have to define your own \cite command.
1244: %
1245: % When we read `\citation' from the .aux file, it means nothing.
1246: %
1247: \let\citation = \@gobble
1248: %
1249: %
1250: % Now comes the stuff for dealing with LaTeX's \newcommand. As
1251: % mentioned earlier, this \newcommand will redefine a preexisting
1252: % command; that's different from how LaTeX's \newcommand behaves.
1253: %
1254: \@innernewcount\@numparams
1255: %
1256: % \newcommand{\foo}[n]{text} defines the control sequence \foo to have
1257: % n parameters, and replacement text `text'.
1258: %
1259: \def\newcommand#1{%
1260: \def\@commandname{#1}%
1261: \@getoptionalarg\@continuenewcommand
1262: }%
1263: %
1264: % Figure out if this definition has parameters.
1265: %
1266: \def\@continuenewcommand{%
1267: % If no optional argument, we have zero parameters. Otherwise, we
1268: % have that many.
1269: \@numparams = \ifx\@optionalarg\empty 0\else\@optionalarg \fi \relax
1270: \@newcommand
1271: }%
1272: %
1273: % \@numparams is how many arguments this command has. The name of the
1274: % command is \@commandname. The replacement text for the new macro is #1.
1275: %
1276: \def\@newcommand#1{%
1277: \def\@startdef{\expandafter\edef\@commandname}%
1278: \ifnum\@numparams=0
1279: \let\@paramdef = \empty
1280: \else
1281: \ifnum\@numparams>9
1282: \errmessage{\the\@numparams\space is too many parameters}%
1283: \else
1284: \ifnum\@numparams<0
1285: \errmessage{\the\@numparams\space is too few parameters}%
1286: \else
1287: \edef\@paramdef{%
1288: % This is disgusting, but \loop doesn't work inside \edef,
1289: % because \body isn't defined.
1290: \ifcase\@numparams
1291: \empty No arguments.
1292: \or ####1%
1293: \or ####1####2%
1294: \or ####1####2####3%
1295: \or ####1####2####3####4%
1296: \or ####1####2####3####4####5%
1297: \or ####1####2####3####4####5####6%
1298: \or ####1####2####3####4####5####6####7%
1299: \or ####1####2####3####4####5####6####7####8%
1300: \or ####1####2####3####4####5####6####7####8####9%
1301: \fi
1302: }%
1303: \fi
1304: \fi
1305: \fi
1306: \expandafter\@startdef\@paramdef{#1}%
1307: }%
1308: %
1309: %% [[[end of BibTeX-specific stuff]]]
1310: %
1311: %
1312: % Names of references (arguments given in the \cite and \nocite
1313: % commands) and file names (arguments given in the \bibliography and
1314: % \bibliographystyle commands) are recorded in \jobname.aux, called the
1315: % \@auxfile in these macros. Here's how they get read in.
1316: %
1317: \def\@readauxfile{%
1318: \if@auxfiledone \else % remember: \@auxfiledonetrue if \noauxfile is defined
1319: \global\@auxfiledonetrue
1320: \@testfileexistence{aux}%
1321: \if@fileexists
1322: \begingroup
1323: % Because we might be in horizontal mode when \@readauxfile
1324: % is called (if it's in response to a \cite or \nocite), we
1325: % want to ignore all the would-be spaces at the ends of
1326: % lines in the aux file. Fortunately, it's highly unlikely
1327: % an end-of-line might actually be desired.
1328: % And because we don't change the category code of anything
1329: % but @, primitives like \gdef can't be used to define labels
1330: % in the aux file. The solution adopted by btxmac.tex is to
1331: % write `\@citedef{LABEL}{DEFINITION}' to the aux file, and
1332: % use \csname on LABEL.
1333: \endlinechar = -1
1334: \catcode`@ = 11
1335: \input \jobname.aux
1336: \endgroup
1337: \else
1338: \message{\@undefinedmessage}%
1339: \global\@citewarningfalse
1340: \fi
1341: \immediate\openout\@auxfile = \jobname.aux
1342: \fi
1343: }%
1344: %
1345: % The \@readauxfile macro does all that work the first time it's called.
1346: % Since it's called once for every \cite, \nocite, \bibliography, and
1347: % \bibliographystyle command that the user issues, we need to remember
1348: % whether the work's been done. It's considered done if we're not to do
1349: % it---that is, if \noauxfile is defined.
1350: %
1351: \newif\if@auxfiledone
1352: \ifx\noauxfile\@undefined \else \@auxfiledonetrue\fi
1353: %
1354: % It's conceivable you'd want to change how other characters are read;
1355: % to do that, change their category code before doing \input btxmac.
1356: %
1357: %
1358: % After reading the .aux file, \@readauxfile opens it for writing.
1359: % The \@writeaux macro does the actual writing (as long as
1360: % \noauxfile is undefined).
1361: %
1362: \@innernewwrite\@auxfile
1363: \def\@writeaux#1{\ifx\noauxfile\@undefined \write\@auxfile{#1}\fi}%
1364: %
1365: %
1366: % A macro package that uses btxmac.tex might define
1367: % \@undefinedmessage (before doing an \input btxmac).
1368: %
1369: \ifx\@undefinedmessage\@undefined
1370: \def\@undefinedmessage{No .aux file; I won't give you warnings about
1371: undefined citations.}%
1372: \fi
1373: %
1374: % Even if citations are undefined, we want to complain only if
1375: % \@citewarningtrue. The default is to set \@citewarningtrue unless
1376: % \noauxfile is defined. Again, a macro package that uses
1377: % btxmac.tex might want to redefine this.
1378: %
1379: \@innernewif\if@citewarning
1380: \ifx\noauxfile\@undefined \@citewarningtrue\fi
1381: %
1382: %
1383: % Finally, before leaving we restore @'s old category code.
1384: %
1385: \catcode`@ = \@oldatcatcode
1386:
1387:
1388:
1389: \def\pfeilso{\leavevmode
1390: \vrule width 1pt height9pt depth 0pt\relax
1391: \vrule width 1pt height8.7pt depth 0pt\relax
1392: \vrule width 1pt height8.3pt depth 0pt\relax
1393: \vrule width 1pt height8.0pt depth 0pt\relax
1394: \vrule width 1pt height7.7pt depth 0pt\relax
1395: \vrule width 1pt height7.3pt depth 0pt\relax
1396: \vrule width 1pt height7.0pt depth 0pt\relax
1397: \vrule width 1pt height6.7pt depth 0pt\relax
1398: \vrule width 1pt height6.3pt depth 0pt\relax
1399: \vrule width 1pt height6.0pt depth 0pt\relax
1400: \vrule width 1pt height5.7pt depth 0pt\relax
1401: \vrule width 1pt height5.3pt depth 0pt\relax
1402: \vrule width 1pt height5.0pt depth 0pt\relax
1403: \vrule width 1pt height4.7pt depth 0pt\relax
1404: \vrule width 1pt height4.3pt depth 0pt\relax
1405: \vrule width 1pt height4.0pt depth 0pt\relax
1406: \vrule width 1pt height3.7pt depth 0pt\relax
1407: \vrule width 1pt height3.3pt depth 0pt\relax
1408: \vrule width 1pt height3.0pt depth 0pt\relax
1409: \vrule width 1pt height2.7pt depth 0pt\relax
1410: \vrule width 1pt height2.3pt depth 0pt\relax
1411: \vrule width 1pt height2.0pt depth 0pt\relax
1412: \vrule width 1pt height1.7pt depth 0pt\relax
1413: \vrule width 1pt height1.3pt depth 0pt\relax
1414: \vrule width 1pt height1.0pt depth 0pt\relax
1415: \vrule width 1pt height0.7pt depth 0pt\relax
1416: \vrule width 1pt height0.3pt depth 0pt\relax}
1417:
1418: \def\pfeilnw{\leavevmode
1419: \vrule width 1pt height 9pt depth -9pt
1420: \vrule width 1pt height 9pt depth -8.7pt
1421: \vrule width 1pt height 9pt depth -8.3pt
1422: \vrule width 1pt height 9pt depth -8.0pt
1423: \vrule width 1pt height 9pt depth -7.7pt
1424: \vrule width 1pt height 9pt depth -7.3pt
1425: \vrule width 1pt height 9pt depth -7.0pt
1426: \vrule width 1pt height 9pt depth -6.7pt
1427: \vrule width 1pt height 9pt depth -6.3pt
1428: \vrule width 1pt height 9pt depth -6.0pt
1429: \vrule width 1pt height 9pt depth -5.7pt
1430: \vrule width 1pt height 9pt depth -5.3pt
1431: \vrule width 1pt height 9pt depth -5.0pt
1432: \vrule width 1pt height 9pt depth -4.7pt
1433: \vrule width 1pt height 9pt depth -4.3pt
1434: \vrule width 1pt height 9pt depth -4.0pt
1435: \vrule width 1pt height 9pt depth -3.7pt
1436: \vrule width 1pt height 9pt depth -3.3pt
1437: \vrule width 1pt height 9pt depth -3.0pt
1438: \vrule width 1pt height 9pt depth -2.7pt
1439: \vrule width 1pt height 9pt depth -2.3pt
1440: \vrule width 1pt height 9pt depth -2.0pt
1441: \vrule width 1pt height 9pt depth -1.7pt
1442: \vrule width 1pt height 9pt depth -1.3pt
1443: \vrule width 1pt height 9pt depth -1.0pt
1444: \vrule width 1pt height 9pt depth -0.7pt
1445: \vrule width 1pt height 9pt depth -0.3pt }
1446:
1447:
1448: \def\pfeilsw{ \leavevmode
1449: \vrule width 1pt height0.3pt depth 0pt\relax
1450: \vrule width 1pt height0.7pt depth 0pt\relax
1451: \vrule width 1pt height1.0pt depth 0pt\relax
1452: \vrule width 1pt height1.3pt depth 0pt\relax
1453: \vrule width 1pt height1.7pt depth 0pt\relax
1454: \vrule width 1pt height2.0pt depth 0pt\relax
1455: \vrule width 1pt height2.3pt depth 0pt\relax
1456: \vrule width 1pt height2.7pt depth 0pt\relax
1457: \vrule width 1pt height3.0pt depth 0pt\relax
1458: \vrule width 1pt height3.3pt depth 0pt\relax
1459: \vrule width 1pt height3.7pt depth 0pt\relax
1460: \vrule width 1pt height4.0pt depth 0pt\relax
1461: \vrule width 1pt height4.3pt depth 0pt\relax
1462: \vrule width 1pt height4.7pt depth 0pt\relax
1463: \vrule width 1pt height5.0pt depth 0pt\relax
1464: \vrule width 1pt height5.3pt depth 0pt\relax
1465: \vrule width 1pt height5.7pt depth 0pt\relax
1466: \vrule width 1pt height6.0pt depth 0pt\relax
1467: \vrule width 1pt height6.3pt depth 0pt\relax
1468: \vrule width 1pt height6.7pt depth 0pt\relax
1469: \vrule width 1pt height7.0pt depth 0pt\relax
1470: \vrule width 1pt height7.3pt depth 0pt\relax
1471: \vrule width 1pt height7.7pt depth 0pt\relax
1472: \vrule width 1pt height8.0pt depth 0pt\relax
1473: \vrule width 1pt height8.3pt depth 0pt\relax
1474: \vrule width 1pt height8.7pt depth 0pt\relax
1475: \vrule width 1pt height9pt depth 0pt\relax
1476: }
1477:
1478: \def\pfeilno{ \leavevmode
1479: \vrule width 1pt height 9pt depth -0.3pt
1480: \vrule width 1pt height 9pt depth -0.7pt
1481: \vrule width 1pt height 9pt depth -1.0pt
1482: \vrule width 1pt height 9pt depth -1.3pt
1483: \vrule width 1pt height 9pt depth -1.7pt
1484: \vrule width 1pt height 9pt depth -2.0pt
1485: \vrule width 1pt height 9pt depth -2.3pt
1486: \vrule width 1pt height 9pt depth -2.7pt
1487: \vrule width 1pt height 9pt depth -3.0pt
1488: \vrule width 1pt height 9pt depth -3.3pt
1489: \vrule width 1pt height 9pt depth -3.7pt
1490: \vrule width 1pt height 9pt depth -4.0pt
1491: \vrule width 1pt height 9pt depth -4.3pt
1492: \vrule width 1pt height 9pt depth -4.7pt
1493: \vrule width 1pt height 9pt depth -5.0pt
1494: \vrule width 1pt height 9pt depth -5.3pt
1495: \vrule width 1pt height 9pt depth -5.7pt
1496: \vrule width 1pt height 9pt depth -6.0pt
1497: \vrule width 1pt height 9pt depth -6.3pt
1498: \vrule width 1pt height 9pt depth -6.7pt
1499: \vrule width 1pt height 9pt depth -7.0pt
1500: \vrule width 1pt height 9pt depth -7.3pt
1501: \vrule width 1pt height 9pt depth -7.7pt
1502: \vrule width 1pt height 9pt depth -8.0pt
1503: \vrule width 1pt height 9pt depth -8.3pt
1504: \vrule width 1pt height 9pt depth -8.7pt
1505: \vrule width 1pt height 9pt depth -9pt
1506: }
1507:
1508:
1509: % *** end including bib4plain.tex ***
1510: % This will define \cite and make sure it works as in latex
1511:
1512:
1513: \def\widestnumber#1#2{}
1514: % Our amstex-ppt style does not know about \widestnumber
1515:
1516:
1517:
1518:
1519: \def\citewarning#1{\ifx\shlhetal\relax
1520: % normal mode, do not write anything
1521: \else
1522: % private mode
1523: \par{#1}\par
1524: \fi
1525: }
1526:
1527:
1528:
1529:
1530: \def\rm{\fam0 \tenrm}
1531:
1532:
1533: \def\fakesubhead#1\endsubhead{\bigskip\noindent{\bf#1}\par}
1534:
1535:
1536: % % \input rsfs
1537: % *** start including rsfs.tex ***
1538:
1539: % # Keywords: Script or Calligraphic (Caligraphic) letters with the RSFS Font
1540:
1541: % The story so far: July 1998 -- Saharon would like to have a
1542: % ``nicer'' calligraphic font. In particualr, the leters S and P in
1543: % the usual calligraphic font do not look ``special'' enough.
1544: %
1545: % I found out that ``rsfs'' (``Ralph Smith Formal Script'') may be
1546: % what he wants. I installed the mf file, the .tfm file, as well as
1547: % a few pk files in ~/TeX/rsfs. Let's hope that this is enough.
1548: % Using amstex, all you have to do is to \input rsfs.tex
1549: % Files prepared with citealice willdothis automatically.
1550: %
1551: % Note: for some reason xdvi calls MakeTeXpk, then Maketexpk
1552: % complains about wrong resolution, but still writes commands to
1553: % missfont.log...
1554: %
1555:
1556:
1557:
1558: % we redefine a macro inside amstex's \Cal command , so that it calls
1559: % our nice font ``rsfs'' rather than the usual calligraphic font.
1560: % Note thisworks for amstex only.
1561: % In plain tex, would have to add definitions of \Cal
1562: % in latex... we should insteaduse mathrsfs.sty
1563: %
1564:
1565:
1566:
1567: \font\textrsfs=rsfs10
1568: \font\scriptrsfs=rsfs7
1569: \font\scriptscriptrsfs=rsfs5
1570:
1571:
1572: \newfam\rsfsfam
1573: \textfont\rsfsfam=\textrsfs
1574: \scriptfont\rsfsfam=\scriptrsfs
1575: \scriptscriptfont\rsfsfam=\scriptscriptrsfs
1576:
1577:
1578: \edef\oldcatcodeofat{\the\catcode`\@}
1579: \catcode`\@11
1580:
1581: \def\Cal@@#1{\noaccents@ \fam \rsfsfam #1}
1582:
1583: \catcode`\@\oldcatcodeofat
1584:
1585: % *** end including rsfs.tex ***
1586:
1587: \expandafter\ifx \csname margininit\endcsname \relax\else\margininit\fi
1588:
1589:
1590:
1591: \long\def\red#1\endred{}
1592: \long\def\green#1\endgreen{}
1593: \long\def\blue#1\endblue{}
1594: \long\def\private#1\endprivate{}
1595:
1596: \def\endred{ \unmatched endred! }
1597: \def\endgreen{ \unmatched endgreen! }
1598: \def\endblue{ \unmatched endblue! }
1599: \def\endprivate{ \unmatched endprivate! }
1600:
1601: \ifx\latexcolors\undefinedcs\def\latexcolors{}\fi
1602:
1603: \def\emptycs{}
1604: \def\evaluatelatexcolors{%
1605: \ifx\latexcolors\emptycs\else
1606: \expandafter\xxevaluate\latexcolors\xxfertig\evaluatelatexcolors\fi}
1607: \def\xxevaluate#1,#2\xxfertig{\setupthiscolor{#1}%
1608: \def\latexcolors{#2}}
1609:
1610: \def\definedred{rgb 0.5 0 0}
1611: \def\definedgreen{rgb 0 0.5 0}
1612: \def\definedblue{rgb 0 0 0.5}
1613: % \def\definedprivate{rgb 0.5 0 0.5}
1614: \def\definedprivate{rgb 0 0 0 } % black
1615:
1616: \font\smallfont=cmsl7
1617: \def\rutgerscolor{\ifmmode\else\endgraf\fi\smallfont% \vrule width 3cm height 1cm depth 0cm
1618: \advance\leftskip0.5cm\relax}
1619: \def\setupthiscolor#1{\edef\tmptmpcs{\noexpand\bgroup\noexpand\rutgerscolor
1620: \noexpand\def\noexpand\currentcolor{#1}%
1621: \noexpand\special{color push \csname defined#1\endcsname}}%
1622: \expandafter\let\csname#1\endcsname\tmptmpcs
1623: \def\tmptmpcs{\checkColorUnmatched{#1}\popthecolor}
1624: \expandafter\let\csname end#1\endcsname\tmptmpcs}
1625:
1626: \def\checkColorUnmatched#1{\def\expectcolor{#1}%
1627: \ifx\expectcolor\currentcolor % OK!
1628: \else \edef\failhere{\noexpand\tryingToClose '\currentcolor' with end\expectcolor}\failhere\fi}
1629:
1630: \def\currentcolor{???}
1631:
1632:
1633:
1634: \def\popthecolor{\special{color pop}\ifmmode\else\endgraf\fi\egroup}
1635:
1636: \expandafter\def\csname#1\endcsname{\special{color push}}
1637:
1638: \evaluatelatexcolors
1639:
1640:
1641: \let\outerhead\head
1642: \def\head{\innerhead}
1643: \let\innerhead\outerhead
1644:
1645: \let\outersubhead\subhead
1646: \def\subhead{\innersubhead}
1647: \let\innersubhead\outersubhead
1648:
1649: \let\outersubsubhead\subsubhead
1650: \def\subsubhead{\innersubsubhead}
1651: \let\innersubsubhead\outersubsubhead
1652:
1653:
1654:
1655: \let\outerproclaim\proclaim
1656: \def\proclaim{\innerproclaim}
1657: \let\innerproclaim\outerproclaim
1658:
1659:
1660:
1661: %
1662: % \newdimen\wzero
1663: % \newdimen\hzero
1664: % \newdimen\dzero
1665: %
1666: % \def\whiteblack#1{\ifx \shlhetal\relax
1667: % \hbox{#1}% normal mode
1668: % \else\ifx\shlhetal\undefinedcontrolsequence
1669: % \hbox{#1}% normal mode
1670: % \else
1671: % % private mode
1672: % \setbox0=\hbox{#1}\leavevmode
1673: % \wzero=\wd0\hzero=\ht0\dzero=\dp0
1674: % \advance\wzero1mm
1675: % \advance\dzero1mm
1676: % \advance\hzero1mm
1677: % \hbox{\vrule width \wzero height \hzero depth\dzero\hskip-\wzero
1678: % \special{ps: currentgray 1 setgray}%
1679: % \hbox to \wzero{\hss\copy0\hss}%
1680: % \special{ps:setgray}}\fi\fi
1681: % }
1682: %
1683: %
1684:
1685: \def\demo#1{\medskip\noindent{\it #1.\/}}
1686: \def\enddemo{\smallskip}
1687:
1688: \def\remark#1{\medskip\noindent{\it #1.\/}}
1689: \def\endremark{\smallskip}
1690:
1691: \def\beginaside{\endgraf\leftskip2cm \vrule width 0pt\relax}
1692: \def\endaside{\endgraf\leftskip0cm \vrule width 0pt\relax}
1693:
1694: % *** end including alice2jlem.tex ***
1695: \pageheight{8.5truein}
1696: %\pageheight{48.5pc}
1697: \topmatter
1698: \title{Model theoretic stability and categoricity for complete metric spaces}
1699: \endtitle
1700: \rightheadtext{Model theoretic stability and categoricity}
1701: \author {Saharon Shelah and Alexander Usvyatsov
1702: \thanks {\null\newline We would like to thank
1703: Alice Leonhardt for the beautiful typing. \null\newline
1704: The first author would like to thank the Israel Science Foundation for
1705: partial support of this research (Grant No. 242/03).
1706: Publication 837. } \endthanks} \endauthor
1707:
1708: \affil{The Hebrew University of Jerusalem \\
1709: Einstein Institute of Mathematics \\
1710: Edmond J. Safra Campus, Givat Ram \\
1711: Jerusalem 91904, Israel
1712: \medskip
1713: Department of Mathematics \\
1714: Hill Center-Busch Campus \\
1715: Rutgers, The State University of New Jersey \\
1716: 110 Frelinghuysen Road \\
1717: Piscataway, NJ 08854-8019 USA
1718: \medskip
1719: UCLA \\
1720: Department of Mathematics \\
1721: P.O. Box 951555 \\
1722: Los Angeles, CA 90095-1555 USA} \endaffil
1723: \medskip
1724:
1725: \abstract
1726: We deal with the systematic development of stability for the context
1727: of approximate elementary submodels of a monster metric space,
1728: which is not far, but still very distinct from the first order
1729: case. In particular we
1730: prove the analogue of Morley's theorem for the classes of complete
1731: metric spaces. \endabstract
1732: \endtopmatter
1733: \document
1734:
1735:
1736: \newpage
1737:
1738: \head {\S1 Introduction and Preliminaries} \endhead \resetall \sectno=1
1739: \spuriousreset
1740: \bigskip
1741:
1742: We work in the context of a compact homogeneous model ${\frak C}$ which is
1743: also a complete metric space with a definable metric $\bold d(x,y)$
1744: and all the predicates and function symbols respect the metric. Such
1745: a monster model will be called a ``monster metric space" (a momspace),
1746: Definition \scite{MOM}.
1747: We investigate the class $K$ of ``almost elementary"
1748: complete submodels of ${\frak C}$, Definition \scite{MOM.2.16}.
1749:
1750: This paper is devoted to categoricty of such classes
1751: $K$ in uncountable cardinalities
1752: (generalizing Morley's theorem to this context).
1753: As we believe that isometry is too strong as a notion of
1754: isomorphism for classes of metric structures, we try to weaken the
1755: assumptions and work with $\varepsilon$-embeddings instead
1756: (Definitions \scite{ISO.1}, \scite{ISO.3}).
1757:
1758: Several suggestions for a framework suitable for model theoretic
1759: treatment of classes arising in functional analysis and dynamics have
1760: been made in the last 40 years by Chang, Keisler, Stern, Henson, and
1761: more recently by Ben-Yaacov. All these attempts were concerned with
1762: allowing a certain amount of compactness (e.g. ``capturing"
1763: ultra-products of Banach spaces introduced by Krivine), without having
1764: to deal with non-standard elements. In this paper the authors choose
1765: to work in the most general context which still allows compactness,
1766: therefore generalizing all the above frameworks. The main tools and
1767: techniques used here are borrowed from homogeneous model theory.
1768:
1769: Homogeneous model theory was introduced and first studied by Keisler,
1770: developed further by the first author, Grossberg, Leslsmann, Hyttinen
1771: and others. It investigates classes of elementary (sometimes somewhat
1772: ``saturated") submodels of a big homogeneous model (``monster"), see
1773: precise definitions later.
1774:
1775: In \cite{Sh:54}, the first author classified such ``monsters" with
1776: respect to the amount of compactness they admit. Monsters which are
1777: compact in a language closed under negation are called ``monsters of
1778: kind II". Later Hrushovski suggested the name ``Robinson
1779: Theories" for such classes, see \cite{Hrxz}. Monsters which are compact
1780: in a language not necessarily closed under negations are called
1781: ``monsters of kind III". Recently Ben-Yaacov has studied this
1782: context in great detail. He called such monsters ``compact abstract
1783: theories", in short CATS, see \cite{BY03}.
1784:
1785: A simple generalization (replacing equality by definable metric)
1786: allows us to speak of a monster model of a class of metric spaces. We
1787: call such monsters ``monster metric spaces", see Definition
1788: \scite{MOM} below. Several results are proven in this general
1789: context, but some require compactness, Definition \scite{HMC.4}(2).
1790: Therefore, our main theorem (Theorem \scite{CATEG}) holds in the
1791: metric analogue of ``monsters of kind III", ``metric cats" in
1792: Ben-Yaacov's terminology.
1793:
1794: Independently of our work (and simultaneously),
1795: Ben-Yaacov investigated categoricity for
1796: metric cats under the additional assumption of the topology on the
1797: space of types being Hausdorff, which is the metric analogue of
1798: ``monsters of kind II" (Robinson theories). In this context one
1799: can reconstruct most of classical stability theory (e.g., independence
1800: based on non-dividing), see \cite{BY05}. These methods fail in the
1801: more general context we were working at. Therefore, techniques
1802: developed and used here are very different, and rely heavily on
1803: non-splitting and Ehrenfeucht-Mostowski constructions. These tools do
1804: not make any significant use of compactness, and we believe that this
1805: assumption can be eliminated by modifying our methods slightly.
1806: At this point we decided not to make the effort, but we try to mention
1807: where exactly compactness is used.
1808:
1809: In a recent work \cite{BeUs0y}, Ben-Yaacov and the second author
1810: introduce a framework of continuous first order logic, closely
1811: related to \cite{ChKe66} and show that once modified slightly, most
1812: model-theoretic approaches to classes of metric spaces (such as
1813: Henson's logic, see \cite{HeIo02}, Hausdorff metric cats, see
1814: \cite{BY05}) are equivalent to continuous logic. Although if continuous
1815: model theory had been discovered earlier, this paper might have looked
1816: differently, we would still like to point out that equivalences shown
1817: in \cite{BeUs0y} do not include monster metric spaces, not even compact
1818: ones. The assumption of the topology on the type space being
1819: Hausdorff is absolutely crucial in \cite{BeUs0y} and \cite{BY05}; it
1820: provides us with the ability to ``approximate" negations, which makes
1821: continuous logic very similar to classical first order logic (of
1822: course, this has many advantages). Lacking Hausdorff assumption, one
1823: has to use more general methods in order to reobtain basic properties.
1824: This is why ``monsters of kind III" (general cats) have been studied
1825: and understood much less than first order on Robinson theories, even
1826: in the discrete (non-continuous, classical first order) context.
1827:
1828: Some work has been done, though: the first author proved the analogue
1829: of Morley's theorem for existentially closed models in \cite{Sh:54},
1830: classes of existentially closed models were investigated further by Pillay in
1831: \cite{Pi00}, general cats were studied by Ben-Yaacov in \cite{BY03},
1832: \cite{BY03a} and other papers.
1833: Our work continues this investigation in the more
1834: general metric context (classical model theory can be viewed as a
1835: particular case with discrete metric).
1836:
1837: Several words should be said also about the difference between the
1838: discrete and the metric context. For example, why could we have not
1839: simply modified the proof in \cite{Sh:54} slightly and obtain Theorem
1840: \scite{CATEG}? The answer is that our categoricity assumption is
1841: significantly weaker, as we assume categoricity only for complete
1842: structures. For instance, the class of infinite-dimensional Hilbert
1843: spaces is categorical in all densities, but not so is the class of
1844: inner-product sapces, not necessarily complete. Starting from a
1845: weaker assumption we aim for a weaker conclusion; but consequences of
1846: our assumption are still significantly weaker than in \cite{Sh:54}
1847: (e.g. $\aleph_0$-stability is lost, we only have a topological
1848: version), which complicates life significantly.
1849:
1850: This work was originally carried out as a Ph.D. thesis of the second
1851: author under the supervision of the first one. The paper is an
1852: expanded version of the thesis wich was written in Hebrew and
1853: submitted to the Hebrew University.
1854: \bn
1855: \centerline{$* \qquad * \qquad *$}
1856: \bn
1857:
1858: The paper is organized as follows:
1859:
1860: We introduce our context in \S2. In particular, we define the class
1861: of models which is investigated (``almost elementary" submodels of
1862: ${\frak C},M \prec_1 {\frak C}$). This notion generalizes Henson's
1863: approximate elementary submodels (see \cite{HeIo02}). It has the
1864: following important property: if $M \prec_1 {\frak C}$, then its
1865: completion (metric closure) $N$ also satisfies $N \prec_1 {\frak C}$.
1866: We will be interested mostly in complete (as metric spaces) ``almost
1867: elementary" submodels of ${\frak C}$.
1868:
1869: The next section, \S3, is devoted to different kinds of approximations to
1870: formulae and types. The importance of considering these
1871: approximations, i.e. topological neighborhoods of partial types, becomes clear
1872: later, when stability, isolation, and other central notions are
1873: discussed.
1874:
1875: In \S4 we generalize the notion of stability in a cardinal $\lambda$
1876: to our topological context. We say that ${\frak C}$ is
1877: $0^+-\lambda$-stable if for no $\varepsilon > 0$ can we find an
1878: $\varepsilon$-net of $\lambda^+$ types over a set of cardinality
1879: $\lambda$, i.e. the space of types over a set of cardinality $\lambda$
1880: has (in a sense) density $\lambda$. This is a generalization of
1881: $\lambda$-stability for Banach spaces studied by Iovino in \cite{Io99}.
1882: It is equivalent to the definitions given by Ben-Yaacov for Hausdorff
1883: cats in \cite{BY05}, and in the context of continuous theories it
1884: coincides with the definitions given in \cite{BeUs0y}.
1885:
1886: We prove equivalence of several similar definitions for
1887: $0^+-\lambda$-stability, connect $0^+-\aleph_0$-stability to
1888: non-splitting of types, classical stability in homogeneous model
1889: theory, existence of average types, saturation of a (closure of a)
1890: union of $(D,\aleph_1)$-homogeneous models. Notions of isolation are
1891: developed and density of strictly isolated types is proved under the
1892: assumption of $0^+-\aleph_0$-stability.
1893:
1894: In \S5 we develop the theory of Ehrenfeucht-Mostowski models in our
1895: context. As we lack forking calculus, some basic facts
1896: (e.g. existence of $(D,\aleph_1)$-homogeneous models in all
1897: uncountable cardinalities) require a different approach, which is
1898: provided by the EM-models techniques.
1899: Also in the proof of the main theorem (\S8) we take
1900: advantage of the representation of the homogeneous model as an EM-model
1901: in order to find inside it a converging sequence which is ``close" to
1902: a subsequence of a given uncountable sequence.
1903:
1904: \S6 is devoted to notions of $\varepsilon$-embeddings and
1905: $\varepsilon$-isomorphisms. We try to weaken our assumptions as much as
1906: we can, and choose to work with the following notion of ``weakly
1907: uncountably categorical" classes: for every $\varepsilon > 0$ there
1908: exists $\lambda > \aleph_0$ in which every two models and
1909: $\varepsilon$-isomorphic to each other. This property seems a priori
1910: weaker than uncountable categoricity in some $\lambda > \aleph_0$,
1911: and even than ``there exists $\lambda > \aleph_0$ such that for every
1912: $\varepsilon > 0$ any two models of density $\lambda$ are
1913: $\varepsilon$-isomorphic". But the main theorem (\scite{CATEG})
1914: states the following: assume ${\frak C}$ is weakly uncountably
1915: categorical. Then every model of density $\lambda > \aleph_0$ is
1916: $(D,\lambda)$-homogeneous (in particular, unique up to an isometry).
1917: So all the above notions turn out to be equivalent.
1918:
1919: In \S7 we prepare the ground for the proof of \scite{CATEG}, showing that
1920: any wu-categorical (weakly uncountably categorical)
1921: momspace is uni-dimensional (in the sense of
1922: \cite{Sh:3}), i.e., any $(D,\aleph_1)$-homogeneous model of density
1923: character $\lambda$ is, in fact, $(D,\lambda)$-homogeneous. The last
1924: section, \S8, contains the proof of the main theorem, Theorem \scite{CATEG}.
1925:
1926: \bn
1927: \centerline {$* \qquad * \qquad *$}
1928: \bn
1929: We recall now basic definitions considering homogeneous model theory
1930: (see \cite{Sh:3},\cite{Sh:54} and \cite{GrLe02}):
1931: \definition{\stag{0.1} Definition} 1) Let $\tau$ be a vocabulary,
1932: $D$ a set of complete $\tau$-types over $\emptyset$ in finitely many
1933: variables. A $\tau$-model $M$ is called a \ub{$D$-model} if
1934: $D(M) \subseteq D$ (where $D(M) =$ the set of complete
1935: finite $\tau$-types over $\emptyset$ realized in $M$). $M$ is called
1936: \ub{$(D,\lambda)$-homogeneous} if $D(M) = D$ and
1937: $M$ is $\lambda$-homogeneous.
1938: \nl
1939: 2) We call $D$ as in (1) a \ub{finite diagram} if for some model
1940: $M,D = D(M)$.
1941: \nl
1942: 3) $A$ is a $D$-set in $M$ if $A \subseteq M$ and $\bar a \in
1943: {}^{\omega >}A \Rightarrow \text{\rm tp}(\bar a,\emptyset,M) \in D$.
1944: For $A$ a $D$-set, $p \in \bold S^m_D(A,M)$ if there are $N,\bar a$ such
1945: that $M \prec N,\bar a \in N$ realizes $p$ and $A \cup \bar a$ is a
1946: $D$-set in $N$.
1947: \nl
1948: \enddefinition
1949: \bigskip
1950:
1951: \remark{\stag{0.2} Remark} Let $D$ be a finite diagram. $M$ is
1952: $(D,\lambda)$-homogeneous \ub{iff} $M$ is universal for
1953: $D$-models of cardinality $\le \lambda$ (or just $\lambda$-sets when
1954: $\lambda < |\tau| + \aleph_0$) and $\lambda$-homogeneous
1955: \ub{iff} $D(M)=D$ and $M$ is
1956: $(D,\lambda)$-saturated (i.e. every $D$-type over a
1957: subset of cardinality $< \lambda$ is realized in $M$).
1958: \endremark
1959: \bigskip
1960:
1961: \demo{Proof} Easy (or see \cite{GrLe02}(2.3),(2.4)).
1962: \enddemo
1963: \bn
1964: This motivates the following
1965: \definition{\stag{0.3} Definition} Let $\lambda^*$ be big enough. A
1966: $(D,\lambda^*)$-homogeneous model ${\frak C}$ will be called a
1967: \ub{$D$-monster} (or a homomogeneous monster for $D$) .
1968: We usually assume $\|{\frak C}\| = \lambda^*$.
1969: \enddefinition
1970: \bn
1971: Recall:
1972: \definition{\stag{0.4} Definition} 1) A finite diagram $D$ is
1973: called $\lambda$-\ub{good} if there is a
1974: $(D,\lambda)$-homogeneous model $M$ of cardinality $\ge \lambda$.
1975: \nl
1976: 2) $D$ is good if it is $\lambda$-good for every $\lambda$.
1977: \enddefinition
1978: \bigskip
1979:
1980: \remark{\stag{0.5} Remark} So $D$ is good iff it has a monster.
1981: \endremark
1982: \bigskip
1983:
1984: \demo{\stag{0.6} Convention} In this paper we will fix a
1985: good finite diagram $D$ and a $D$-monster model ${\frak C}$.
1986: \enddemo
1987: \bn
1988: \bigskip
1989:
1990: \demo{\stag{0.7} Observation} 1) If ($D$ is good, ${\frak C}$ a
1991: $D$-monster model) $p \in \bold S_D(A),A \subseteq B$ \ub{then}
1992: there is $q,p \subseteq q \in \bold S_D(B)$.
1993: \enddemo
1994: \bn
1995: \ub{Question}: Why can we use ``good $D$"? There are several
1996: answers.
1997:
1998: Basically, Claim \scite{0.7S} says that every $D$ which is the
1999: finite diagram of a \ub{compact} momspace (see Definitions
2000: \scite{MOM}, \scite{MOM.1}) is good. Claim \scite{0.8S} says that
2001: even without compactness, categoricity implies stability, which
2002: implies $D$ is good by \cite{Sh:3}. The reader can easily omit these
2003: claims in the first reading.
2004: \bigskip
2005:
2006: \proclaim{\stag{0.7S} Claim} Assume
2007: \mr
2008: \item "{$(a)$}" $D$ is a finite diagram, i.e., a set of complete
2009: $n$-types in $\Bbb L(\tau_{\frak C})$
2010: \sn
2011: \item "{$(b)$}" ${\frak C}$ is $(D,\aleph_0)$-homogeneous
2012: \sn
2013: \item "{$(c)$}" $\Delta$ is full for ${\frak C}$ (see Definition
2014: \scite{HMC.4})
2015: \sn
2016: \item "{$(d)$}" ${\frak C}$ is $\Delta$-compact.
2017: \ermn
2018: \ub{Then}
2019: \mr
2020: \item "{$(*)$}" $(\alpha) \quad D$ is good
2021: \sn
2022: \item "{${{}}$}" $(\beta) \quad$ if $\kappa(D) < \infty$ then $\kappa(D) \le
2023: (|\tau_{\frak C}| + \aleph_0)^+$ ($\kappa(D)$ as in \cite{Sh:54}).
2024: \endroster
2025: \endproclaim
2026: \bigskip
2027:
2028: \demo{Proof} See \cite{Sh:54}, the discussion of monsters of kind III or
2029: \cite{BY03}, existence of a universal domain.
2030: \enddemo
2031: \bigskip
2032:
2033: \proclaim{\stag{0.8S} Claim} 1) Assume that
2034: \mr
2035: \item "{$(a)$}" $\tau$ a countable metric vocabulary
2036: and let $\delta(*) = (2^{\aleph_0})^+$
2037: \sn
2038: \item "{$(b)$}" $\tau = \tau_D$
2039: \sn
2040: \item "{$(c)$}" $D$ is a finite diagram
2041: \sn
2042: \item "{$(d)$}" there is a $(D,\aleph_1)$-homogeneous model $M$
2043: \sn
2044: \item "{$(e)$}" there is a $D$-model of cardinality $\ge
2045: \beth_{\delta(*)}$ (probably less is enough) and
2046: \sn
2047: \item "{$(f)$}" $K^1_D = \{M:M \in K^c_1\}$ (see Definition \scite{MOM.2.16})
2048: is categorical in $\lambda > \aleph_0$
2049: or is wu-categorical (see Definition \scite{ISO.3}).
2050: \ermn
2051: \ub{Then} $D$ is stable, hence is good.
2052: \endproclaim
2053: \bigskip
2054:
2055: \demo{Proof} By \scite{OS}, \scite{STAB.2s.0} hence we get stability,
2056: $D$ is good follows by \cite{Sh:3}.
2057: \enddemo
2058: \bigskip
2059:
2060: \demo{\stag{0.9} Notations}
2061:
2062: $\lambda,\mu,\chi \qquad \qquad$ infinite cardinals
2063:
2064: $\alpha,\beta,\gamma \qquad \qquad$ infinite ordinals
2065:
2066: $\delta \qquad \qquad \qquad$ limit ordinals
2067:
2068: $\nu,\eta \qquad \qquad \qquad$ sequences of ordinals
2069:
2070: $\varphi,\psi,\vartheta \qquad \qquad$ formulae
2071:
2072: ${\frak C} \qquad \qquad \quad$ the monster model
2073:
2074: $M,N \qquad \qquad$ models (in the monster)
2075:
2076: $A,B,C \qquad \qquad$ sets (in the monster)
2077:
2078: $\varepsilon,\zeta,\xi \qquad \qquad$ positive reals
2079:
2080: $I,J \qquad \qquad$ order types
2081:
2082: $\bold I,\bold J \qquad \qquad$ indiscernible sequences
2083:
2084: $\bold d \qquad \qquad \qquad$ a metric
2085: \enddemo
2086: \bigskip
2087:
2088: \head {\S2 Main context - monster metric spaces} \endhead \resetall \sectno=2
2089: \spuriousreset
2090: \bigskip
2091:
2092: In this section we discuss our main context. We start with some
2093: notations.
2094: \definition{\stag{1n.1} Definition} Let $(X,\bold d)$ be a
2095: metric space.
2096: \nl
2097: 1) We extend the metric to
2098: $n$-tuples: for $\langle a_i:i < n \rangle,\langle b_i:i < n\rangle$
2099: we define $\bold d(\bar a,\bar b) = \text{\rm max}\{\bold
2100: d(a_i,b_i):i<n\}$.
2101: \nl
2102: 2) For $\bar a$ an $n$-tuple, $A$ a set of $n$-tuples,
2103: $\bold d(\bar a,A) = \bold d(A,\bar a) = \text{\rm inf}\{\bold d(\bar
2104: a,\bar b):\bar b \in A\}$.
2105: \nl
2106: 3) For sets $A,B \subseteq X^n$, we define two versions
2107: of distances:
2108:
2109: $$
2110: \bold d_1(A,B) = \text{\rm inf}\{\bold d(\bar a,\bar b):\bar a \in
2111: A,\bar b \in B\}
2112: $$
2113:
2114: $$
2115: \bold d_2(A,B) = \text{\rm sup}\{\{\bold d_1(\bar a,B):\bar a \in A\}
2116: \cup \{\bold d_1(A,\bar b):\bar b \in B\}\}.
2117: $$
2118: \mn
2119: 4) We denote the density (the density character) of $X$ by Ch$(X)$.
2120: So Ch$(X)$ is the minimal cardinality of a dense subset of $X$.
2121: \nl
2122: 5) We denote the topological (metric) closure of a set $A$ by $\bar A$
2123: or mcl$(A)$.
2124: \enddefinition
2125: \bigskip
2126:
2127: \definition{\stag{1.1} Definition} 1) We call a vocabulary $\tau$
2128: \ub{metric} if it contains predicates $P_{q_1,q_2}(x,y)$ for all $0
2129: \le q_1 \le q_2$ rationals. We call the collection of these
2130: predicates a \ub{metric scheme}.
2131: \nl
2132: 2) Given a metric vocabulary $\tau$, we call a $\tau$-structure
2133: $M$ \ub{semi-metric} if for some (unique) $\bold d$:
2134: \mr
2135: \item "{$(i)$}" $M$ is a metric space with the metric $\bold d$
2136: \sn
2137: \item "{$(ii)$}" $\bold d$ is definable in $M$ by the $\tau$-metric
2138: scheme, i.e., $\bold d(a,b) \in [q_1,q_2]$ iff $M \models
2139: P_{q_1,q_2}(a,b)$ for all rationals $q_1,q_2$ and $a,b \in M$.
2140: \ermn
2141: 3) We call a semi-metric structure \ub{complete} if $(M,\bold d)$ is
2142: complete as a metric space.
2143: \enddefinition
2144: \bigskip
2145:
2146: \remark{\stag{1.2} Remark} Given a semi-metric $\tau$-structure $M$, we
2147: will usually write $M \models ``\bold d(a,b) \le q"$, etc., forgetting
2148: the $\tau$-metric scheme.
2149:
2150: Note that in a pseudo-metric structure for each $r_1,r_2 \in \Bbb R$, the
2151: property ``$r_1 \le \bold d(x,z) \le r_2$" is $0$-type-definable by a
2152: set of atomic formulas.
2153: \endremark
2154: \bigskip
2155:
2156: \definition{\stag{HMC.sO} Definition} 1) $(M,\bold d)$ is a
2157: \ub{metric} structure (or model) \ub{when}:
2158: \mr
2159: \item "{$(a)$}" $\tau({\frak C})$ is a metric vocabulary and
2160: $(M,\bold d)$ is a semi-metric structure
2161: \sn
2162: \item "{$(b)$}" $P^M \subseteq {}^{\text{arity}(P)}M$ is closed
2163: (with respect to $\bold d$) for every predicate $P \in \tau_M$
2164: \sn
2165: \item "{$(c)$}" $F^M:{}^{\text{arity}(F)}M \rightarrow N$ is a
2166: continuous function for every function symbol $F \in \tau_M$.
2167: \ermn
2168: 2) $({\frak C},\bold d)$ is a (homogeneous) metric monster
2169: if: $({\frak C},\bold d)$
2170: is a metric model, ${\frak C}$ is a $D$-monster for some finite
2171: diagram $D$, constant here.
2172: \enddefinition
2173: \bigskip
2174:
2175: \remark{\stag{1.3} Remark} Each homogeneous monster admits the
2176: discrete definable metric, i.e., $\bold d(a,b)=1$ for all $a \ne b \in
2177: {\frak C}$, and it is definable by the equality and inequality. So
2178: each homogeneous monster is a metric monster with the discrete metric.
2179: \endremark
2180: \bigskip
2181:
2182: \demo{\stag{1.4} Notations} We will often identify $\varphi =
2183: \varphi(\bar x,\bar a)$ (for $\varphi$ a formula,
2184: $\bar a \in {\frak C}$) with the set
2185: of the realizations in ${\frak C}$ of $\varphi(\bar x,\bar a)$, i.e.,
2186: $\varphi({\frak C},\bar a) = \varphi^{\frak C} =: \{\bar b \in {}^{\ell g(\bar
2187: x)}{\frak C}:{\frak C} \models \varphi[\bar b,\bar a]\}$.
2188: So $\bold d_1(\varphi,\psi)$ in
2189: fact means $\bold d_1(\varphi^{\frak C},\psi^{\frak C})$, etc.
2190: \enddemo
2191: \bn
2192: \centerline {$* \qquad * \qquad *$}
2193: \bn
2194: The following definition is the analogue of abstract elementary classes
2195: (see \cite{Sh:88r}) in our context.
2196: \definition{\stag{AMC} Definition} Let $({\frak K},\le_{\frak K})$ be
2197: an ordered class of $\tau = \tau({\frak K})$ complete metric
2198: structures ($\tau$ is a metric vocabulary), ${\frak K}$ closed under
2199: $\tau$-isomorhisms. We call \nl
2200: $({\frak K},\le_{\frak K})$ an
2201: \ub{abstract metric class} (a.m.c.) if
2202: \mr
2203: \item $\langle M_i:i < \delta \rangle$ is a
2204: $\le_{\frak K}$-increasing sequence (from ${\frak K}$, of
2205: course), then $M = \text{\rm mcl}(\dbcu_{i<\delta} M_i)
2206: \in {\frak K}$. Moreover, $[i < \delta \Rightarrow M_i \le_{\frak K}
2207: M]$ and $[M_i \le_{\frak K} N$ for all $i<\delta
2208: \Rightarrow M \le_{\frak K} N]$
2209: \sn
2210: \item for $M_1 \subseteq M_2 \subseteq M_3$ from ${\frak K},[M_1,M_2
2211: \le_{\frak K} M_3] \Rightarrow M_1 \le_{\frak K} M_2$
2212: \sn
2213: \item for some cardinal LS$({\frak K})$ we have the ``downward
2214: L\"owenheim-Skolem theorem", i.e., for each $M \in {\frak K},A
2215: \subseteq M$, there exists $N \in {\frak K},N \le_{\frak K} M,A
2216: \subseteq N$, Ch$(N) \le |A| + \text{\rm LS}({\frak K})$.
2217: \endroster
2218: \enddefinition
2219: \bigskip
2220:
2221: \definition{\stag{HMC.2} Definition} We say that a set of formula
2222: $\Delta$ for a homogeneous metric monster ${\frak C},\bold d$ is
2223: \ub{admissible} if for each $\varphi(\bar x) \in \Delta$, the set
2224: $\varphi^{\frak C} = \{\bar a \in {\frak C},{\frak C} \models
2225: \varphi(\bar a)\}$ is closed with respect to the metric $\bold d$
2226: (topology induced by it).
2227: \enddefinition
2228: \bigskip
2229:
2230: \definition{\stag{HMC.2S} Definition} In ${\frak C},\varphi,\psi$ are
2231: contradictory if $\bold d_1(\varphi^{\frak C},\psi^{\frak C}) > 0$.
2232: \enddefinition
2233: \bigskip
2234:
2235: \remark{\stag{HMC.2T} Remark} Later (see \scite{MOM.3A}) we show that
2236: in our context this is equivalent to $\varphi^{\frak C} \cap
2237: \psi^{\frak C} = \emptyset$.
2238: \endremark
2239: \bigskip
2240:
2241: \demo{\stag{HMC.3} Example} If $\bold d$ is discrete, then each subset
2242: of ${\frak C}$ is closed, so the set of all formulas $\Delta = \Bbb
2243: L(\tau_{\frak C})$ is admissible.
2244: \enddemo
2245: \bn
2246: In order to give a non-trivial example, we define
2247: \definition{\stag{POS.1} Definition} For a metric model $(M,\bold d)$
2248: and a formula $\varphi(x,\bar a)$ with parameters $\bar a \in M$, we
2249: say that $M \models \exists^* x \varphi(x,\bar a)$ (there almost
2250: exists $x$ such that $\varphi(x,\bar a)$) if for every $\varepsilon > 0$ there
2251: exist $b,\bar a'$ such that $M \models \varphi(b,\bar a')$ and $\bold
2252: d(\bar a,\bar a') \le \varepsilon$.
2253: \enddefinition
2254: \bigskip
2255:
2256: \definition{\stag{POS} Definition} 1) We define \ub{positive} formulae
2257: by induction: each atomic formula is positive, for $\varphi,\psi$
2258: positive, $\varphi \wedge \psi,\varphi \vee \psi,\exists^* x
2259: \varphi,\forall x \varphi$ are positive. So negation and implication
2260: are not allowed.
2261: \nl
2262: 2) \ub{Positive existential} formulae are defined similarly without
2263: allowing $\forall x \varphi$.
2264: \enddefinition
2265: \bigskip
2266:
2267: \demo{\stag{HMC.so.2} Observation} For $(M,\bold d)$ a metric model
2268: and positive $\varphi(\bar x) \in \Bbb L(\tau_M),\varphi^M = \{\bar a
2269: \in {}^{\ell g(\bar x)}M:M \models \varphi[\bar a]\}$ is a closed
2270: subset of ${}^{\ell g(\bar x)}M$ (under $\bold d$), i.e., $\Delta =$
2271: ``the positive formulae" is admissible.
2272: \enddemo
2273: \bigskip
2274:
2275: \definition{\stag{HMC.4} Definition} 1) For a homogeneous monster
2276: ${\frak C}$, a set of formulae $\Delta$ is called \ub{full} if
2277: \mr
2278: \item "{$(i)$}" for all $\bar a \in {\frak C},A \subseteq {\frak C}$,
2279: tp$(\bar a,A,{\frak C})$ is determined by the $\Delta$-type
2280: tp$_\Delta(a,A,{\frak C})$
2281: \sn
2282: \item "{$(ii)$}" if $\varphi \in \Delta,\bar a \in {\frak C}$ and
2283: ${\frak C} \models \neg \varphi(\bar a)$, then there exists $\psi \in
2284: \Delta$ such that $(\varphi,\psi)$ is contradictory (see \scite{HMC.2S})
2285: and ${\frak C} \models \psi(\bar a)$.
2286: \ermn
2287: 2) We call ${\frak C}$ \ub{$\Delta$-compact} (where $\Delta$ is a set
2288: of formulae, i.e., $\subseteq \Bbb L(\tau_{\frak C})$)
2289: if each set of $\Delta$-formulae with parameters from
2290: ${\frak C}$ of cardinality $<|{\frak C}|$ which is finitely
2291: satisfiable in ${\frak C}$, is realized in ${\frak C}$. We omit
2292: $\Delta$ if constant, and abusing notation write $\Delta = \Delta({\frak C})$.
2293: \enddefinition
2294: \bigskip
2295:
2296: \definition{\stag{HMC.4s.0} Definition} We say that $\Delta$ is
2297: full$^+$ if: as in \scite{HMC.4} but in (ii) the quantifier depth of
2298: $\psi$ is $\le$ the quantifier depth of $\varphi$.
2299: \enddefinition
2300: \bn
2301: Now we make the main definition of this section, introducing the
2302: context of this paper.
2303: \definition{\stag{MOM} Definition} 1) A metric
2304: homogeneous class $({\frak K},
2305: \le_{\frak K})$ (equivalently: its metric homogeneous monster
2306: ${\frak C}$) is called $\Delta$-\ub{momspace} (monster metric space)
2307: if $\Delta = \Delta({\frak C})$ is a set of formulae
2308: containing the metric scheme, such that
2309: \mr
2310: \item "{$(a)$}" $\Delta \subseteq \Bbb L(\tau_{\frak C})$ is closed
2311: under $\wedge,\exists^*$ and subformulae
2312: \sn
2313: \item "{$(b)$}" $\Delta$ is admissible
2314: \sn
2315: \item "{$(c)$}" $\Delta$ is full for ${\frak C}$.
2316: \ermn
2317: 2) Let ``${\frak C}$ is momspace" mean ``for some $\Delta$" and choose
2318: such $\Delta = \Delta({\frak C})$ (well, it is not necessarily unique
2319: but we ignore this).
2320: \nl
2321: 3) The metric $\bold d = \bold d_{\frak C}$ can be defined from
2322: ${\frak C}$ so we can ``forget" to mention $\bold d$, but still will
2323: usually say $({\frak C},\bold d)$, e.g. to distinguish from
2324: ${\frak C}$ when we use the
2325: $(D,\lambda)$-homogeneous context.
2326: \enddefinition
2327: \bigskip
2328:
2329: \demo{\stag{MOM.2.15} Convention} $({\frak C},\bold d)$ is a fixed momspace.
2330: \enddemo
2331: \bn
2332: The class of models we are interested in is defined below.
2333: \definition{\stag{MOM.2.16} Definition} 1) $K_1 = K^1_{\frak C}$ is
2334: the class of $M$ such that:
2335: \mr
2336: \item "{$(a)$}" $M \subseteq {\frak C}$
2337: \sn
2338: \item "{$(b)$}" $M \prec^1_{\Delta} {\frak C}$ which means:
2339: \nl
2340: \ub{if} $\varepsilon > 0,{\frak C} \models \varphi[\bar b,\bar
2341: a],\varphi(\bar y,\bar x) \in \Delta,\bar a \in {}^{\ell g(\bar x)}M$
2342: \ub{then} for some $\bar b' \in {}^{\ell g(\bar y)} M$ we have
2343: ${\frak C} \models (\exists \bar x,\bar y)(\varphi(\bar y,\bar x) \wedge
2344: \bold d(\bar y \char 94 \bar x,\bar b' \char 94 \bar a)$.
2345: \ermn
2346: 2) $K^c_1$ is the class of members of $K_1$ which are \ub{complete}.
2347: \enddefinition
2348: \bigskip
2349:
2350: \proclaim{\stag{MOM.2.19} Claim} 1) Assume $M \subseteq {\frak C}$ and
2351: $N = \text{\rm mcl}(M)$. Then
2352: \mr
2353: \item "{$(a)$}" $M \in K_1$ iff $N \in K_1$ iff $N \in K^c_1$.
2354: \ermn
2355: 2) $(K_1,\subseteq)$ is an abstract metric class
2356: \endproclaim
2357: \bigskip
2358:
2359: \definition{\stag{MOM.1} Definition} A momspace $({\frak C},\bold d)$
2360: is called \ub{compact} if ${\frak C}$ is $\Delta({\frak C})$-compact.
2361: \enddefinition
2362: \bigskip
2363:
2364: \demo{\stag{MON.1s.2} Convention} We work in a compact
2365: $\Delta$-momspace $({\frak C},\bold d),M,N$ denote submodels which
2366: are from $K_1$; though really interested in the closed (complete) ones,
2367: i.e. versions of categoricity are defined using complete models.
2368: \ub{BUT} we try to mention when we use compactness.
2369: \enddemo
2370: \bigskip
2371:
2372: \demo{\stag{MOM.1s.3} Observation} [${\frak C}$ compact] For
2373: $\varphi(x,\bar y) \in \Delta,\bar a \in {\frak C}$, we have ${\frak
2374: C} \models \exists^* x \varphi(x,\bar a) \Leftrightarrow {\frak C}
2375: \models \exists x \varphi(x,\bar a)$ [so we can ``forget" about the
2376: new existential quantifier and use the classical one].
2377: \enddemo
2378: \bigskip
2379:
2380: \demo{Proof} Assume ${\frak C} \models \exists^* x \varphi(x,\bar
2381: a)$. Then the set $\{\varphi(x,\bar y),\bold d(\bar y,\bar a) \le
2382: \frac 1n:n < \omega\}$ is consistent in ${\frak C}$ (by compactness),
2383: so ${\frak C} \models \exists x \varphi(x,\bar a)$. \hfill$\square$
2384: \enddemo
2385: \bigskip
2386:
2387: \demo{\stag{MOM.1.2.19} Examples} 1) Let $T$ be first order, ${\frak C}$ a
2388: big saturated model of $T$, then ${\frak C}$ is a compact momspace
2389: (with discrete metric), $\Delta({\frak C}) = {\Cal L}$ (all formulae).
2390: \nl
2391: 2) Let $T$ be a Robinson theory (see \cite{Hrxz}), ${\frak C}$ its
2392: universal domain. Then it is a compact momspace.
2393: \nl
2394: 3) Consider the unit ball of a monster Banach space as in
2395: \cite{HeIo02}, Chapter 12.
2396: Then $({\frak C},\|\cdot\|$) is a compact momspace (we write the norm
2397: here instead of the metric) where $\Delta({\frak C}) =$ positive formulae (more
2398: precisely positive bounded formulae, see \cite{HeIo02}, Chapter 5),
2399: ${\frak K} = K^c_1 = \{M:M \prec_A {\frak C}\}$, see
2400: \cite{HeIo02}(6.2) for the definition of $\prec_A$.
2401: Compactness is proven in Chapter 9 there. One can easily show that
2402: $\prec_A = \prec^1_\Delta$, see also \scite{MOM.suB}, \cite{HeIo02}(6.6).
2403: \nl
2404: 4) Metric Hausdorff CATs, see \cite{BY03} for a definition of a Hausdorff
2405: CAT. A CAT is metric if its monster is metric. See \cite{BY05} in
2406: Hausdorff metric cats.
2407: \nl
2408: 5) The main example we have in mind: $({\frak C},\bold d)$ is a
2409: compact momspace for $\Delta({\frak C}) =$ positive existential
2410: formulae, ${\frak K} = K^c_1$.
2411: So in particular it is a metric cat (not necessarily Hausdorff).
2412: \nl
2413: \enddemo
2414: \bigskip
2415:
2416: \head {\S3 Approximations to formulas and types} \endhead \resetall \sectno=3
2417: \spuriousreset
2418: \bigskip
2419:
2420: The following notions of $\varepsilon$-approximations of formulae is of
2421: major importance. We give two different definitions and will use both
2422: for different purposes. Note that \scite{MOM.2} simply defines topological
2423: neighborhoods, while in \scite{MOM.2s.13} by moving the parameters we
2424: allow the formula to change
2425: a little, see also \scite{MOM.suB}. Let $\varepsilon$ denote a
2426: non-negative rational number.
2427: \bigskip
2428:
2429: \definition{\stag{MOM.2} Definition} 1) For a formula $\varphi(\bar
2430: x)$ possibly with parameters, we define
2431: $\varphi^{[\varepsilon]}(\bar x) = \exists \bar x'(\varphi(\bar x') \wedge
2432: \bold d(\bar x,\bar x') \le \varepsilon)$. So
2433: $\varphi^{[\varepsilon]}(\bar x,\bar a) = (\exists \bar x')[\bold
2434: d(\bar x,\bar x') \le \varepsilon \wedge \varphi(\bar x',\bar a)]$.
2435: \nl
2436: 2) For a (partial) type $p$, define $p^{[\varepsilon]} =
2437: \{(\dsize \bigwedge_{\ell < n} \varphi_\ell)^{[\varepsilon]}:n < \omega$ and
2438: $\varphi_0,\dotsc,\varphi_{n-1} \in p\}$.
2439: \nl
2440: \enddefinition
2441: \bigskip
2442:
2443: \definition{\stag{MOM.2s.13} Definition} 1) For a formula
2444: $\varphi(\bar x,\bar a)$ let $\varphi^{<\varepsilon>}(\bar
2445: x,\bar a) = (\exists \bar x',\bar y')(\varphi(\bar x',\bar y')
2446: \wedge \bold d(\bar x',\bar x) \le \varepsilon \wedge
2447: \bold d(\bar y',\bar a) \le \varepsilon)$.
2448: \nl
2449: 2) For a (partial) type $p,p^{<\varepsilon>}(\bar x) = \{(\dsize \bigwedge_\ell
2450: \varphi_\ell(\bar x,\bar a_\ell))^{<\varepsilon>}:n < \omega$ and
2451: $\varphi_\ell(\bar x,\bar a_\ell) \in p$ for $\ell <n\}$.
2452: \enddefinition
2453: \bigskip
2454:
2455: \definition{\stag{MOM.2T} Definition} 1) For a (partial) type (maybe
2456: with parameters) $p$, or set $B$ and $\varepsilon > 0$, we say that
2457: $\bar c \in B$ realizes $p^{[\varepsilon]}$ if ${\frak C} \models
2458: p^{[\varepsilon]}(\bar c)$. We say that $p^{[\varepsilon]}$ is
2459: realized in $B$ if some $\bar c \in B$ realizes it.
2460: \nl
2461: 2) Similar for $p^{<\varepsilon>}$.
2462: \enddefinition
2463: \bigskip
2464:
2465: \remark{\stag{MOM.2s.3} Remark} 1) Note that
2466: $\varphi^{[\varepsilon]}(\bar x,\bar a)$ is equivalent to
2467: $\varphi^{[\varepsilon]}_1(\bar x)$, when we expand ${\frak C}$ by
2468: individual constants for $\bar a$ and let $\varphi_1(\bar x)
2469: = \varphi(\bar x,\bar a)$.
2470: \nl
2471: 2) $\varphi^{[\varepsilon]}\models \varphi^{<\varepsilon>}$ for all
2472: $\varphi$.
2473: \nl
2474: 3) For a formula without parameters $\varphi,\varphi^{[\varepsilon]} =
2475: \varphi^{<\varepsilon>}$.
2476: \nl
2477: 4) For $\zeta \ge \varepsilon \ge 0,\varphi^{[\varepsilon]} \models
2478: \varphi^{[\zeta]},\varphi^{<\varepsilon>} \models
2479: \varphi^{<\zeta>}$.
2480: \nl
2481: 5) For $\varepsilon=0,\varphi^{[\varepsilon]} \equiv \varphi^{<\varepsilon>}
2482: \equiv \varphi$.
2483: \endremark
2484: \bigskip
2485:
2486: \demo{\stag{MOM.3} Observation} 1) If $\varphi = \varphi(\bar x,\bar b)$
2487: (as usual is admissible), then
2488: $\varphi \equiv \dsize \bigwedge_{\varepsilon >0}
2489: \varphi^{[\varepsilon]}$.
2490: \nl
2491: 2) If $\varphi = \varphi(\bar x,\bar a)$ then $\varphi
2492: \equiv \dsize \bigwedge_{\varepsilon > 0} \,\,
2493: \dsize \bigvee_{\zeta \in (0,\varepsilon)} \varphi^{[\zeta]}$.
2494: \nl
2495: 3) Similar for $\varphi^{<\varepsilon>}$.
2496: \enddemo
2497: \bigskip
2498:
2499: \demo{Proof} 1) $[a \models \varphi \Rightarrow a \models \bigwedge
2500: \{\varphi^{[\varepsilon]}:\varepsilon > 0\}]$ is obvious.
2501:
2502: Suppose $\bar a \models \varphi^{[\varepsilon]}$ for all $\varepsilon$, so
2503: for each $n$ there exists $\bar a_n$ such that ${\frak C}
2504: \models \varphi[\bar a_n]$ and $\bold d(\bar a,\bar a_n)
2505: \le \frac 1n$. So $\langle \bar a_n:n \in \omega
2506: \rangle$ converges to $\bar a$,
2507: therefore $\bar a \models \varphi$, as $\varphi^{\frak C}$ is a closed
2508: set.
2509: \nl
2510: 2),3) Similar.
2511: \enddemo
2512: \bigskip
2513:
2514: \proclaim{\stag{4n.10} Claim} 1) $(\theta^{<\zeta_1>}(\bar
2515: x))^{<\zeta_2>} \equiv \theta^{<\zeta_1+\zeta_2>}(\bar x)$.
2516: \nl
2517: 2) $(p^{<\zeta_1>}(\bar x))^{<\zeta_2>} \equiv
2518: p^{<\zeta_1+\zeta_2>}$.
2519: \nl
2520: 3) Assume $\bold d(\bar b,\bar b') \le \zeta_1 - \zeta_2$ then
2521: $\theta^{<\zeta_2>}(x,b') \models \theta^{<\zeta_1>}(x,b)$.
2522: \endproclaim
2523: \bigskip
2524:
2525: \proclaim{\stag{MOM.3A} Claim} 1) If ${\frak C}$ is $\Delta$-compact
2526: \ub{then} a pair (of admissible $\Delta$-formulas) $\varphi
2527: = \varphi(\bar x,\bar a),\psi
2528: = \psi(\bar x,\bar b)$ is contradictory (see definition \scite{HMC.2S})
2529: iff $\varphi^{\frak C} \cap \psi^{\frak C} = \emptyset$.
2530: \nl
2531: 2) [${\frak C}$ not necessarily compact] A pair as in (1)
2532: $\varphi,\psi$ is contradictory iff for some $\varepsilon >0,
2533: \varphi^{[\varepsilon]} \cap \psi^{[\varepsilon]} = \emptyset$.
2534: \nl
2535: 3) If $\varphi,\psi$ in (2) are without parameters, we can add ``iff
2536: $\varphi^{<\varepsilon>} \cap \psi^{<\varepsilon>} = \emptyset$ for
2537: some $\varepsilon > 0$".
2538: \nl
2539: 4) If ${\frak C}$ is compact, (3) holds for formulae with parameters.
2540: \endproclaim
2541: \bigskip
2542:
2543: \demo{Proof} 1) Obviously if $(\varphi,\psi)$ is a contradictory pair
2544: then $\varphi^{\frak C} \cap \psi^{\frak C} = \emptyset$. Now assume
2545: $\bold d_1(\varphi,\psi) = 0$. So for each $n > 0$ there exists
2546: $\bar a_n,\bar b_n \in {\frak C}$ such that $\bold d(\bar a_n,\bar
2547: b_n) < \frac 1n$ and
2548: $\varphi(\bar a_n,\bar a),\psi(\bar b_n,\bar b)$ hold. Therefore the set
2549: $\{\varphi(\bar x,\bar a),\psi(\bar y,\bar b)\} \cup \{\bold d(\bar
2550: x,\bar y) \le \frac 1n:n=1,2,\ldots\}$ is finitely consistent. By
2551: compactness we obtain $\bar a^*,\bar b^*$ realizing it, but
2552: necessarily $\bold d(\bar a^*,\bar b^*) = 0$, so $\bar a^* = \bar b^*$
2553: realizes both $\varphi$ and $\psi$ and we are done.
2554: \nl
2555: 2) Trivial.
2556: \nl
2557: 3) Follows from (2) and \scite{MOM.2s.3}(3).
2558: \nl
2559: 4) Assume $\varphi^{<\varepsilon>} \cap \psi^{<\varepsilon>} \ne
2560: \emptyset$ for all $\varepsilon > 0$, so the set
2561: $\{\varphi^{<\varepsilon>}(\bar x,\bar a) \wedge
2562: \psi^{<\varepsilon>}(\bar x,\bar b):\varepsilon >0\}$ is finitely
2563: satisfiable (by \scite{MOM.2s.3}(4)) and therefore realized in ${\frak
2564: C}$, now by \scite{MOM.3}(3) $\varphi \cap \psi \ne \emptyset$,
2565: therefore $\varphi$ and $\psi$ are not contradictory. The other
2566: direction is trivial. \hfill$\square_{\scite{MOM.3A}}$
2567: \enddemo
2568: \bigskip
2569:
2570: \proclaim{\stag{MOM.4} Claim} [Let $({\frak C},\bold d)$ be a compact
2571: momspace], $p$ a type over a set $A$.
2572: \nl
2573: Then $\bar a \in {\frak C}$ realizes $p^{[\varepsilon]}$ iff there
2574: exists $\bar a' \in {\frak C}$ realizing $p,\bold d(\bar a,\bar a')
2575: \le \varepsilon$ (so $p^{[\varepsilon]}$ is just the
2576: ``$\varepsilon$-neighborhood" of $p$).
2577: \endproclaim
2578: \bigskip
2579:
2580: \demo{Proof} The if direction is obvious. Now suppose $\bar a
2581: \models p^{[\varepsilon]}$, so $[\varphi(\bar a)]^{[\varepsilon]}$
2582: holds for all $\varphi \in p$, but there is no $\bar a' \models p$
2583: such that $\bold d(\bar a,\bar a') \le \varepsilon$. Then the set
2584:
2585: $$
2586: \{\bold d(\bar x',a') \le \varepsilon\} \cup \{\varphi(\bar x'):
2587: \varphi \in p,\varphi \text{ is a } \Delta({\frak C})\text{-formula}\}
2588: $$
2589: \mn
2590: is inconsistent (as $\Delta$ is full for ${\frak C}$), and by
2591: compactness we get a contradiction (because in the definition of
2592: $p^{[\varepsilon]}$ we essentially close $p$ under conjunctions).
2593: \enddemo
2594: \bn
2595: We connect the relevant notion of submodel defined in \scite{MOM.2.16}
2596: with the ``logical" notion of approximation:
2597: \demo{\stag{MOM.suB} Observation} $M \prec^1_\Delta N$ ($M,N$
2598: structures, see \scite{MOM.2.16} for the definition of
2599: $\prec^1_\Delta$) iff for every $\varphi(\bar y,\bar x) \in
2600: \Delta,\varepsilon > 0$, if $N \models \exists \bar y \varphi(\bar
2601: y,\bar a),\bar a \in M$, then $N \models \varphi^{<\varepsilon>}(\bar
2602: b,\bar a)$ for some $\bar b \in M$.
2603: \enddemo
2604: \bigskip
2605:
2606: Note that the assumption above of $M$ being complete in fact
2607: follows from $(D,\aleph_1)$-homogeneous:
2608: \bigskip
2609:
2610: \demo{\stag{MOM.sub.2} Observation} Let $M \in K_1$ be
2611: $(D,\aleph_1)$-homogeneous, then $M \in K^c_1$, i.e., is complete.
2612: \enddemo
2613: \bigskip
2614:
2615: \demo{Proof} Let $\langle a_n:n < \omega\rangle$
2616: converge to $a \in {\frak C},a_n \in M$. So the set
2617: $\{\bold d(x,a_n) \le {\bold d} (a,a_n):n \in \Bbb
2618: N\}$ is realized in $M$, so $\langle a_n:n <\omega\rangle$ converges
2619: in $M$ (to the same limit, of course). \hfill$\square_{\scite{MOM.sub.2}}$
2620: \enddemo
2621: \bigskip
2622:
2623: \definition{\stag{MOM.4A} Definition} We call $M \in K^c_1$ pseudo
2624: $(D,\lambda)$-homogeneous if for every $A \subseteq M,|A| < \lambda$,
2625: for every $p \in S^m_D(A)$, for every $\varepsilon >
2626: 0,p^{<\varepsilon >}(\lambda)$ is realized in $M$ (see Definition
2627: \scite{MOM.2T}).
2628: \nl
2629: Note that replacing $p^{<\varepsilon>}$ above by $p^{[\varepsilon]}$,
2630: we get $(D,\lambda)$-homogeneous, see \scite{MOM.4B} below.
2631: \enddefinition
2632: \bigskip
2633:
2634: \definition{\stag{MOM.4A.1} Definition} We say that a type $p \in
2635: S(A)$ is \ub{almost realized} in $B$ (or $B$ almost realizes $p$) if
2636: for all $\varepsilon > 0$ there exists $b_\varepsilon \in
2637: B,b_\varepsilon \models p^{[\varepsilon]}$.
2638: \enddefinition
2639: \bigskip
2640:
2641: \proclaim{\stag{MOM.4B} Claim} [$({\frak C},\bold d)$ compact] Let $M \in
2642: K^c_1$ be such that every 1-type over a subset of $M$ of cardinality
2643: less than $\lambda$ is almost realized in $M$ ($\lambda$ infinite).
2644: Then $M$ is $(D,\lambda)$-homogeneous.
2645: \endproclaim
2646: \bigskip
2647:
2648: \demo{Proof} Let $p \in S^1_D(A),A \subseteq M,|A| < \lambda$.
2649: Choose by induction $b_n \models p^{[\frac{1}{2^n}]},b_n \in M,\bold
2650: d(b_{n+1},b_n) \le \frac{1}{2^{n-1}}$.
2651:
2652: Why is this possible? Let $q_n = p(x)
2653: \cup\{\bold d(x,b_n) \le \frac{1}{2^n}\}$.
2654: As $b_n \models p^{[\frac{1}{2^n}]}$, some $b^* \in {\frak C}$
2655: realizes $q_n$ (see \scite{MOM.4}) and let $q^*_n = \text{ tp}_D(b^*/A
2656: \cup \{b_n\})$, so $q^*_n \in S^1_D(A \cup \{b_n\})$, and by the
2657: assumption there exists $b_{n+1} \in M$ realizing
2658: $(q^*_n)^{[\frac{1}{2^{n+1}}]}$. So $b_{n+1} \models
2659: p^{[\frac{1}{2^{n+1}}]},\bold d(b_{n+1},b_n) \le \frac{1}{2^n} +
2660: \frac{1}{2^{n+1}} \le \frac{1}{2^{n-1}}$, as required.
2661:
2662: Now, $\langle b_n:n < \omega\rangle$ is a Cauchy sequence, let $b \in
2663: M$ be its limit ($M \in K^c_1$, so is complete). Now $b \models
2664: p^{[\varepsilon]}$ for all $\varepsilon > 0$, so $b \models p$ (see
2665: \scite{MOM.3}(1)), and we are done.
2666: \enddemo
2667: \bigskip
2668:
2669: \proclaim{\stag{MOM.4C} Corollary} Let $M \in K^c_1$ be
2670: non-$(D,\lambda)$-homogeneous. Then there exists $A \subseteq M,|A| <
2671: \lambda,p \in S^1_D(A)$ and $\varepsilon > 0$ such that
2672: $p^{[\varepsilon]}$ is omitted in $M$.
2673: \endproclaim
2674: \bigskip
2675:
2676: \demo{Proof} This is just restating \scite{MOM.4B}; but we prefer
2677: this form for later use.
2678: \enddemo
2679: \bigskip
2680:
2681: \remark{\stag{MOM.4D} Remark} We will not use the notion of pseudo
2682: homogeneity (Definition \scite{MOM.4A}), in this paper (as a
2683: postoriori all the models will turn out to be
2684: $(D,\lambda)$-homogeneous), but it is interesting to point out what
2685: non-categorical classes can look like. See \scite{4n.23} later.
2686: \endremark
2687: \bigskip
2688:
2689: \proclaim{\stag{MON.2s.1} Claim} If $A \subseteq {\frak C},p \in
2690: \bold S_D(\text{\rm mcl}(A))$ and $c \in {\frak C}$ realizes $p
2691: \restriction A$ \ub{then} it realizes $p$.
2692: \endproclaim
2693: \bigskip
2694:
2695: \demo{Proof} If $\varphi(x,a_1,\dotsc,a_m) \in p$ ($\varphi \in
2696: \Delta$ of course, parameters not suppressed) then for each $n <
2697: \omega$ we can choose $b^n_\ell \in A$ (for $\ell =
2698: 1,\dotsc,m)$ such that $\bold d(a_\ell,b^n_\ell) < \frac{1}{n+2}$.
2699: Hence $\varphi(x,b^n_1,\dotsc,b^n_m)^{<1/n+2>}$ belongs to $p$ as
2700: $\varphi(x,a_1,\dotsc,a_m)$ implies it, hence it belongs to $p \restriction A$.
2701:
2702: So if $c \in {\frak C}$ realizes $p \restriction A$ then $\models
2703: \varphi(c,b^n_1,\dotsc,b^n_m)^{<1/n+2>}$ for each $n$. If ${\frak C} \models
2704: \neg \varphi[c,a_1,\dotsc,a_m]$ then some $\psi,\psi =
2705: \psi(y,x_1,\dotsc,x_m) \in \Delta$ and $\varphi(y,x_1,\dotsc,x_n)$ are
2706: contradictory and ${\frak C} \models \psi[c,a_1,\dotsc,a_n]$. So for
2707: all $n$, the tuples $cb^n_1,\dotsc,b^n_m$ lies in both
2708: $(\varphi^{<\frac{1}{n+2}>})^{\frak C}$
2709: and $(\psi^{<\frac{1}{n+2}>})^{\frak C}$, therefore
2710: $\varphi^{<\varepsilon>} \cap
2711: \psi^{<\varepsilon>} \ne \emptyset$ for all $\varepsilon > 0$, which
2712: contradicts $\varphi$ and $\psi$ being a contradictory pair, see
2713: \scite{MOM.3A}(4).
2714: \enddemo
2715: \bigskip
2716:
2717: \head {\S4 Stability in momspaces} \endhead \resetall \sectno=4
2718: \spuriousreset
2719: \bigskip
2720:
2721: We define a topological version of stability. The intuition behind
2722: the definition is that there may be many types, but the density of the
2723: space of types is small. Our definition generalizes
2724: Iovino's stability for Banach spaces, see \cite{Io99} for Hausdorff
2725: cats and continuous theories, it coincides with definitions given in
2726: \cite{BY05}, \cite{BeUs0y} respectively. Note that for
2727: elementary homogeneous class (i.e. discrete metric), this definition
2728: coincides with the usual one (so certainly for an elementary class,
2729: i.e., ${\frak C}$ saturated). For a non-discrete metric the classical
2730: $\lambda$-stability of $D$ (counting types, as in \cite{Sh:3}) is
2731: stronger than the topological relative we define here, and is equivalent
2732: if and only if $\lambda = \lambda^{\aleph_0}$. Stability in our sense
2733: (i.e., $\lambda$-stable for some $\lambda$) is equivalent to stability
2734: for $D$, but for a specific $\lambda$ (e.g. $\lambda = \aleph_0$) the
2735: notions differ.
2736: \bigskip
2737:
2738: \demo{\stag{STAB.0} Hypothesis} $({\frak C},\bold d)$ is a momspace for
2739: $\Delta = \Delta_{\frak C}$.
2740: \enddemo
2741: \bigskip
2742:
2743: \definition{\stag{STAB.1} Definition} 1) A momspace $({\frak C},\bold
2744: d)$ is called \ub{$0^+-\lambda$-stable} if for all $A \subseteq {\frak
2745: C}$ of cardinality $\lambda$ there exists $B \subseteq {\frak C}$ of
2746: cardinality $\lambda$ such that each $p \in \bold S_D(A)$ is realized
2747: in mcl$(B) = \bar B$ (i.e., the topological closure of $B$).
2748: \enddefinition
2749: \bigskip
2750:
2751: \proclaim{\stag{STAB.2} Claim} [$({\frak C},\bold d)$ compact]
2752: For a compact momspace ${\frak C}$, the following are equivalent:
2753: \mr
2754: \item "{$(A)$}" ${\frak C}$ is $0^+-\lambda$-stable
2755: \sn
2756: \item "{$(B)$}" for each $A \subseteq {\frak C},|A| \le \lambda$, there exists
2757: $B,|B| \le \lambda$ such that for each $\varepsilon >0$ and $p \in
2758: \bold S_D(A),p^{[\varepsilon]}$ is realized in $B$ ($B$ almost realizes
2759: all types over $A$, see Definition \scite{MOM.4A.1})
2760: \sn
2761: \item "{$(C)$}" there are no
2762: $A,|A| \le \lambda,\langle p_i:i < \lambda^+\rangle$ a
2763: sequence of members of $\bold S_D(A)$, i.e., complete
2764: $D$-types over $A$ and $\varepsilon > 0$ such that
2765: $p^{[\varepsilon]}_i \cap p^{[\varepsilon]}_j = \emptyset$ for all
2766: $i,j < \lambda^+$.
2767: \endroster
2768: \endproclaim
2769: \bigskip
2770:
2771: \demo{Proof} $(A) \Rightarrow (C)$.
2772:
2773: Assume (A). Suppose (C) fails,
2774: so we have $\langle p_i:i < \lambda^+\rangle$ over $A$ as there. By
2775: (A) we can find $B$ of cardinality $\le \lambda$
2776: such that each $p_i$ is realized in mcl$(B) = \bar B$, by,
2777: say, $a_i$. As $\{a_i:i < \lambda^+\}$ form an $\varepsilon$-net (by
2778: the nature of the $p_i$'s), it is obvious that density of $\bar B$ is
2779: at least $\lambda^+$, contradiction ($|B|=\lambda$).
2780: \mn
2781: $(C) \Rightarrow (B)$.
2782:
2783: Let $A$ be given. We construct $B_n$ by induction:
2784:
2785: $-B_0 = A$
2786:
2787: $-B_{n+1}$ realizes all complete $D$-types over $B_n$ up to $\frac{1}{n+1}$,
2788: i.e.
2789: \nl
2790:
2791: \hskip20pt if $p \in S_D(B_n)$ then for some $a \in B_{n+1}$,
2792: tp$(a,B_n)^{[\frac{1}{n+1}]}({\frak C}) \cap p^{[\frac{1}{n+1}]}
2793: ({\frak C}) \ne \emptyset$
2794:
2795: $-|B_n| = \lambda$
2796:
2797: $B = \cup B_n$ is obviously as required in (B).
2798:
2799: How is $B_{n+1}$ constructed? Let $\varepsilon = 1/2(n+1)$ and let
2800: $\langle p^k_i:i < \lambda
2801: \rangle$ be a maximal $\varepsilon$-disjoint set of complete types of
2802: $\kappa$-tuples over $B_n$, (I.e. $(p^k_i)^{[\varepsilon]} \cap
2803: (p^k_j)^{[\varepsilon]} = \emptyset$ for all $i,j$)
2804:
2805: $B_{n+1} = B_n \cup \{\bar a^k_i:\bar a^k_i$ realizes $p^k_i$ for some
2806: $i < \lambda,\kappa < \omega\}$.
2807:
2808: Now each complete $k$-type $p$ over $B_n$ is $\frac{1}{n+1}$-realized in
2809: $B_{n+1}$, as there is $i < \lambda$ such that
2810: $(p^k_i)^{[\varepsilon]}({\frak C}) \cap p^{[\varepsilon]}({\frak C})$
2811: is non-empty, let $\bar b$ be in the intersection. We can replace
2812: $\bar b$ by any $\bar b'$ realizing tp$(b,B_n)$ hence \wilog \, $\bold
2813: d(\bar b,a^k_i) < \varepsilon$. Also we can find $\bar c$ realizing
2814: $p$ such that $\bold d(\bar b,\bar c) < \varepsilon$. So $\bar a^k_i$
2815: witnesses $p$ is $\frac{1}{n+1}$-realized in $B_{n+1}$.
2816: \mn
2817: (B) $\Rightarrow$ (A).
2818:
2819: Let $A$ be given. Define $B_0 = A,B_{n+1}$ almost realizes all types
2820: over $B_n,B_n \subseteq B_{n+1},|B_n| = \lambda$. Let $B = \dbcu_{n < \omega}
2821: B_n$. Let $p \in \bold S_D(A)$. Pick by induction on $n \ge 1,a_n
2822: \in B_n$ such that $a_n \models p^{[\frac 1n]}$ and
2823: $\bold d(a_n,a_{n-1}) \le \frac {1}{2^n}$ if $n>1$
2824: (possible by \scite{MOM.4} as $B_n$
2825: almost realizes all types over $\dbcu_{i < n} B_n$). $\langle
2826: a_n:n <\omega \rangle$ is obviously a Cauchy sequence, let $a$ be its
2827: limit in $\bar B$ and we are done by \scite{MOM.3}(1)
2828: as $a \models p^{[\frac 1n]}$ for all $n$.
2829: \enddemo
2830: \bigskip
2831:
2832: \demo{\stag{STAB.2A} Observation} 1) If ${\frak C}$ is
2833: $\lambda$-stable (in the sense of $D$, see \cite{Sh:3}), then it
2834: is $0^+-\lambda$-stable.
2835: \nl
2836: 2) If ${\frak C}$ is $0^+-\lambda$-stable and $\lambda =
2837: \lambda^{\aleph_0}$ then ${\frak C}$ is $\lambda$-stable.
2838: \enddemo
2839: \bigskip
2840:
2841: \demo{Proof} 1) Trivial.
2842: \nl
2843: 2) Recall $|\bar B| = |B|^{\aleph_0}$.
2844: \enddemo
2845: \bn
2846: Recall
2847: \definition{\stag{STAB.2s.2} Definition} A (partial) type $q$ \ub{splits}
2848: over a set $A$ if there are $\varphi(\bar x,\bar a),\psi(\bar x,\bar
2849: b) \in q,\varphi(\bar x,\bar y),\psi(\bar x,\bar y)$ contradictory,
2850: and tp$(\bar a/A) = \text{\rm tp}(\bar b/A)$.
2851: \enddefinition
2852: \bigskip
2853:
2854: \demo{\stag{STAB.3} Observation} For a complete type $p,p^{[\varepsilon]}$
2855: splits over $A \Leftrightarrow$ there exist $\varphi(\bar x,\bar
2856: a),\psi(\bar x,\bar b) \in p,\bold d_1(\varphi(\bar x,\bar
2857: y),\psi(\bar x,\bar y)) > 2 \varepsilon$,
2858: tp$(\bar a/A) = \text{\rm tp}(\bar b/A)$.
2859: \enddemo
2860: \bigskip
2861:
2862: \demo{Proof} $p^{[\varepsilon]}$ splits over $A \Leftrightarrow$
2863: there exist $\varphi^{[\varepsilon]}(\bar x,\bar
2864: a),\psi^{[\varepsilon]}(\bar x,\bar b) \in p^{[\varepsilon]}$ (where
2865: $\varphi(\bar x,\bar a),\psi(\bar x,\bar b) \in p)$ such that $\bar a
2866: \equiv_A \bar b$ and $(\varphi^{[\varepsilon]}(\bar x,\bar
2867: y),\psi^{[\varepsilon]}(\bar x,\bar y))$ is a contradictory pair [so
2868: $(\varphi,\psi)$ is a $2\varepsilon$-contradictory pair].
2869: \enddemo
2870: \bigskip
2871:
2872: \demo{\stag{STAB.3s.1} Observation} 1) The type $p$ splits over $A$
2873: \ub{iff} for some $\varepsilon >0$, the type $p^{[\varepsilon]}$
2874: splits over $A$.
2875: \enddemo
2876: \bigskip
2877:
2878: \demo{Proof} $p$ splits over $A$ iff some contradictory pair
2879: $(\varphi,\psi)$ witnesses this, now pick $\varepsilon = \frac{\bold
2880: d_1(\varphi,\psi)}{3}$ and use \scite{STAB.3}.
2881: \enddemo
2882: \bigskip
2883:
2884: \demo{\stag{STAB.3.4} Notation} We write $\bar a \equiv_A \bar b$ for
2885: tp$(\bar a/A) = \text{\rm tp}(\bar b/A)$.
2886: \enddemo
2887: \bigskip
2888:
2889: \proclaim{\stag{STAB.4} Lemma} Let ${\frak C}$ be $0^+-\aleph_0$-stable.
2890: \ub{Then} for any
2891: $B \subseteq {\frak C},p \in \bold S_D(B)$ and $\varepsilon > 0,
2892: p^{[\varepsilon]}$ does not split over a finite subset of $B$.
2893: \endproclaim
2894: \bigskip
2895:
2896: \demo{Proof} Suppose $p^{[\varepsilon]}$ splits over every finite subset
2897: of its domain. We construct finite sets $A_n$ for $n < \omega$ and
2898: elementary maps $F_\eta$ for $\eta \in {}^{\omega >}2$ as follows:
2899: \mr
2900: \item $A_0 = \emptyset$
2901: \sn
2902: \item $A_{n+1} = A_n \cup \{\bar a_n,\bar b_n\}$, where there
2903: are $\varphi_n(\bar x,\bar y),\psi_n(\bar x,\bar y)$
2904: such that $\varphi_n(\bar x,\bar a_n),\psi_n(\bar x,\bar b_n)$ exemplify
2905: $p^{[\varepsilon]}$ splits over $A_n$, i.e. $\bar a_n \equiv_{A_n}
2906: \bar b_n,\varphi_n(\bar x,\bar a_n),
2907: \psi_n(\bar x,\bar b_n) \in p^{[\varepsilon]}$ and
2908: $\varphi^{[\varepsilon]}_n({\frak C},\bar a_n) \cap
2909: \psi^{[\varepsilon]}_n({\frak C},\bar b_n) = \emptyset$
2910: \sn
2911: \item for $\eta \in {}^n 2,F_\eta:A_n \rightarrow {\frak C}$ is an
2912: elementary mapping
2913: \sn
2914: \item for $\eta \in {}^n 2,F_{\eta {}^\frown \,\langle 0\rangle}
2915: (\bar a_n) = F_{\eta {}^\frown \,\langle 1 \rangle} (\bar b_n)$ and
2916: $F_{\eta {}^\frown \,\langle 0 \rangle},F_{\eta {}^\frown\,\langle 1 \rangle}$
2917: extend $F_\eta$.
2918: \ermn
2919: The construction is straightforward. Now denote for $\eta
2920: \in {}^\omega 2,F_\eta = \dbcu_{n < \omega} F_{\eta \upharpoonright n},
2921: p^*_\eta = F_\eta(p),A = \cup\{\text{\rm Rang}(F_\eta):\eta
2922: \in {}^{\omega >} 2\}$ (so it is countable) and choose \footnote{this uses
2923: ``$D$ is good"} $p_\eta,p^*_\eta \subseteq p_\eta \in \bold S_D(A)$.
2924: Obviously, $\eta \ne \nu \in
2925: {}^\omega 2 \Rightarrow p_\eta^{[\varepsilon]} \cap
2926: p^{[\varepsilon]}_\nu = \emptyset$, contradicting
2927: $0^+-\aleph_0$-stability by \scite{STAB.2} by an implication not using
2928: compactness.
2929: \enddemo
2930: \bigskip
2931:
2932: \proclaim{\stag{STAB.2s.0} Claim} Assume
2933: $({\frak C},\bold d)$ is $0^+-\aleph_0$-stable. \ub{Then}
2934: \mr
2935: \item "{$(a)$}" if $p \in \bold S_D(B)$ then $p$ does not split over some
2936: countable $A \subseteq B$
2937: \ermn
2938: \mr
2939: \item "{$(b)$}" $D$ is stable (in the sense of \cite{Sh:3}), $\kappa(D)
2940: \le \aleph_1$.
2941: \endroster
2942: \endproclaim
2943: \bigskip
2944:
2945: \remark{Remark} Note that it is close but not as in first order;
2946: there may be $\aleph_0$ exceptions.
2947: \endremark
2948: \bigskip
2949:
2950: \demo{Proof}
2951: \mn
2952: \ub{Clause (a)}:
2953:
2954: By the previous claim for every $\varepsilon >0$ for some finite
2955: $B_\varepsilon \subseteq A,p$ does not
2956: $\varepsilon$-split over $B_\varepsilon$. Let $B = \cup\{B_{1/(n+1)}:n
2957: < \omega\}$, so $B$ is a countable subset of $A$ and by
2958: \scite{STAB.3s.1} and the obvious monotonicity of
2959: non-$\varepsilon$-splitting, $p$ does not split over $B$.
2960: \mn
2961:
2962: \ub{Clause (b)}: Follows from (a).
2963: \bigskip
2964:
2965: \definition{\stag{ST.1} Definition} Given an uncountable indiscernible
2966: set $\bold I \subseteq {}^m {\frak C}$ and a set $A \subseteq {\frak
2967: C}$, define the average type of $\bold I$ over $A$, Av$(A,\bold I)$ as
2968: follows:
2969:
2970: $$
2971: \align
2972: \text{Av}(A,\bold I) = \{\varphi(\bar x,\bar a):&\varphi(\bar x,\bar y)
2973: \in \Delta,\ell g(\bar x) = m,\bar a \in A, \text{ and for infinitely many} \\
2974: &\bar c \in \bold I,\varphi(\bar c,\bar a) \text{ holds}\}.
2975: \endalign
2976: $$
2977: \enddefinition
2978: \bigskip
2979:
2980: \demo{\stag{ST.2} Fact} If ${\frak C}$ is stable, then any
2981: indiscernible sequence is an indiscernible set.
2982: \enddemo
2983: \bigskip
2984:
2985: \demo{Proof} Standard.
2986:
2987: We often say ``$\bold I$ is indiscernible" meaning an indiscernible
2988: sequence, which is the same as indiscernible set.
2989: \enddemo
2990: \bigskip
2991:
2992: \proclaim{\stag{ST.3} Claim} Let ${\frak C}$ be $0^+-\aleph_0$-stable.
2993: \nl
2994: 1) If $\bold I \subseteq {}^m {\frak C}$ is indiscernible uncountable,
2995: $A \subseteq {\frak C}$ a set, then {\rm Av}$(A,\bold I) \in
2996: S^m_D(A)$.
2997: \nl
2998: 2) If $A,\bold I$ are as in (1), then
2999:
3000: $$
3001: \align
3002: \text{Av}(A,\bold I) = \{\varphi(\bar x,\bar a):&\varphi(\bar x,\bar y)
3003: \in \Delta,\ell g(\bar x) = m,\bar a \in M \text{ and for all but} \\
3004: &\text{ countably many } \bar c \in \bold I \text{ does }
3005: \varphi(\bar c,\bar a) \text{ hold}\}.
3006: \endalign
3007: $$
3008: \endproclaim
3009: \bigskip
3010:
3011: \demo{Proof} Using the standard argument,
3012: one shows that for a given contradictory pair $(\varphi,\psi) =
3013: (\varphi(\bar x,\bar z),\psi(\bar x,\bar z))$ and $\bar a \in M$, one
3014: of the sets $\{\bar c \in \bold I:\models \varphi(\bar c,\bar
3015: a)\},\{\bar c \in \bold I:\models \psi(\bar c,\bar a)\}$ is finite
3016: (otherwise, let $\varepsilon = \bold d_1(\varphi,\psi)$, and construct
3017: $2^{\aleph_0} \, \varepsilon$-distant types over a countable set,
3018: contradictory $0^+-\aleph_0$-stability; in fact, one constructs
3019: $2^{\aleph_0}$ pairwise distinct $(\varphi,\psi)$-types, i.e., types
3020: mentioning only $\varphi$ and $\psi$).
3021:
3022: Now given a formula $\varphi(\bar x,\bar a)$ over $M$, if $\{\bar c
3023: \in \bold I:\varphi(\bar c,\bar a)$ holds$\}$ is infinite, then for
3024: each $\psi(\bar x,\bar z)$ such that $(\varphi,\psi)$ is
3025: contradictory, the set $J_\psi = \{\bar c \in \bold I:\psi(\bar c,\bar
3026: a)$ holds$\}$ is necessarily finite, so taking the union of $J_\psi$
3027: over all such $\psi$, we obtain a countable set of exceptions
3028:
3029: $$
3030: J = \cup J_\psi = \{\bar c \in \bold I:\models \neg \varphi(\bar
3031: c,\bar a)\}.
3032: $$
3033: \mn
3034: This completes the proof of clause (2). For clause (1), let
3035: $\varphi(\bar x,\bar a)$ be a formula over $A,\ell g(\bar x) = m$. If
3036: for uncountably many $\bar c \in \bold I,\varphi(\bar c,\bar a)$ holds,
3037: then $\varphi(\bar x,\bar a) \in \text{ Av}(A,\bold I)$, otherwise for
3038: some $\psi(\bar x,\bar a)$ such that $(\varphi,\psi)$ is a
3039: contradictory pair, $\psi(\bar c,\bar a)$ holds for uncountably many
3040: $\bar c \in \bold I$, so $\psi(\bar x,\bar a) \in \text{ Av}(A,\bold
3041: I)$. Clearly, only one of the two options above is possible, so (1) follows.
3042: \enddemo
3043: \bn
3044: \ub{Discussion}: Why countable and not finite in the definition of
3045: averages? Even if the majority satisfies
3046: $\varphi(\bar x,\bar a)$, for each $\varepsilon$ there can be finitely many
3047: $\bar c \in \bold I$ such that
3048: ${\frak C} \models \psi(\bar c,\bar a)$ and $(\varphi,\psi)$ are
3049: $\varepsilon$-contradictory, and this finite number can increase when
3050: $\varepsilon$ goes to 0.
3051: \enddemo
3052: \bigskip
3053:
3054: \proclaim{\stag{ST.4} Lemma} If $M$ is
3055: $(D,\aleph_1)$-homogeneous, $p \in \bold S^m(M)$
3056: \ub{then} for some uncountable $\bold I \subseteq {}^m M$, we have $p
3057: = \text{\rm Av}(M,\bold I)$.
3058: \endproclaim
3059: \bigskip
3060:
3061: \demo{Proof} Let $B$ be as in \scite{STAB.2s.0},
3062: clause (a). Let $m=1$ for simplicity. Choose $a_\alpha \in M$
3063: realizing $p \restriction B \cup\{a_\beta:\beta < \alpha\}$ by
3064: induction on $\alpha < \omega_1$.
3065: Now $\bold I = \langle a_\alpha:\alpha < \omega_1\rangle$ is
3066: indiscernible over $B$ by \cite[I,\S2]{Sh:c}.
3067:
3068: If $q = \text{Av}(M,\bold I) \ne p$ still $q \in \bold S_D(M)$ (by
3069: \scite{ST.3}(1)) and
3070: we can find $\varphi(x,\bar b) \in q,\psi(x,\bar b) \in p$ and they
3071: are contradictory. So $u = \{\alpha < \omega_1:{\frak C} \models
3072: \varphi[a_\alpha,\bar b]\}$ is infinite let $v \subseteq u$ be of
3073: cardinality $\aleph_0$, let $v \subseteq \alpha(*) < \omega_1$ and
3074: choose $a'_\beta \in M(\beta \in [\alpha(*),\omega_1)$ realizing $p
3075: \restriction (B \cup \{a_i:i < \alpha(*)\} \cup \bar b \cup
3076: \{a'_\gamma:\gamma \in (\alpha(*),\beta)\}$.
3077: \nl
3078: Easy contradiction: for a given contradictory pair $(\varphi,\psi)$
3079: all but finitely many elements of $\bold I$ have to ``make a choice", see
3080: also \scite{ST.3}(1).
3081: \enddemo
3082: \bigskip
3083:
3084: \definition{\stag{STAB.2s.1} Definition} 1) A momspace
3085: $({\frak C},\bold d)$ is called $(\mu,*)$-superstable if
3086: given $\langle M_i:i < \omega \rangle$ an
3087: increasing chain of $(D,\mu)$-homogeneous models $(M_i \in K^c_1$,
3088: of course), mcl$(\dbcu_{i < \omega} M_i)$ is $(D,\mu)$-homogeneous.
3089: \nl
3090: 2) We omit $\mu$ if this holds for every $\mu$ large enough.
3091: \nl
3092: 3) Fully $*$-superstable means for every $\mu > |\tau_{\frak C}|
3093: + \aleph_0$.
3094: \enddefinition
3095: \bigskip
3096:
3097: \remark{Remark} This definition generalizes superstability for
3098: ${\frak C}$ a saturated model of a first order theory, and
3099: ${\frak C}$ a homogeneous monster.
3100: \endremark
3101: \bn
3102: The following claim will be mainly of interest for us when $\mu = \aleph_0$:
3103: \proclaim{\stag{STAB.5} Claim} Let $({\frak C},\bold d)$ be
3104: $0^+-\aleph_0$-stable and compact. Then $({\frak C},\bold d)$ is
3105: $(\mu^+,*)$-superstable for every $\mu \ge \aleph_0$.
3106: \endproclaim
3107: \bigskip
3108:
3109: \demo{Proof} Let $\langle M_n:n < \omega \rangle$ be an increasing
3110: sequence of $(D,\mu^+)$-homogeneous models, and assume $p \in
3111: \bold S(A),A \subseteq \overline{\dbcu_{n < \omega} M_n}$ of cardinality
3112: $\le \mu$ is not realized in $M_\omega = \overline{\bigcup M_n}$.
3113:
3114: By increasing $A$ and the $0^+-\aleph_0$-stability
3115: (i.e., for every $\varepsilon > 0$ trying to build a tree
3116: $\langle p_\eta:\eta \in {}^{\omega >} 2\rangle$ of
3117: $\varepsilon$-contradictory types) \wilog \, $p$ has a unique
3118: extension in $\bold S_D(M_\omega)$, call it $q$.
3119: By \scite{MOM.4C} as $({\frak C},\bold d)$ is compact we can add
3120: that for some
3121: $\varepsilon >0,p^{[\varepsilon]}$ is not realized in $M_\omega$.
3122: Without loss of generality $A \subseteq \text{\rm mcl}
3123: (A \cap \dbcu_{n < \omega} M_n)$ hence by \scite{MON.2s.1} \wilog \,
3124: $A \subseteq \dbcu_{n < \omega} M_n$ (as
3125: $p$ is determined by its restriction to $A \cap (\dbcu_{n < \omega} M_n))$.
3126:
3127: Now by \scite{STAB.4} there is a finite $B
3128: \subseteq A$ over which $q \restriction \cup\{M_n:n < \omega\}$
3129: does not $(\varepsilon/5)$-split. Let $n
3130: < \omega$ be such that $B \subseteq M_n$ \wilog \,
3131: $q \restriction M_n$ does not split over $A \cap M_n$ (by increasing
3132: $A$ and \scite{STAB.2s.0}(a)).
3133: Let $q_n =q \restriction M_n$ and let $A_n =
3134: M_n \cap A$. As $M_n$ is $(D,\mu^+)$-homogeneous, by Lemma
3135: \scite{ST.4}, there is an
3136: uncountable indiscernible sequence $\bold I_n$ in $M_n$ with
3137: Av$(M_n,\bold I_n) = q_n$; \wilog \,
3138: $|\bold I_n| = \mu^+$, and $\bold I_n$ is indiscernible over $A$ (not just
3139: $A_n$!) (as for each finite type $\bar b$ from $A$, tp$(\bar b/\bold
3140: I_n)$ does not split over a countable subset, so we can remove a
3141: subset of $\bold I_n$ of cardinality $\mu$). Now
3142: as elements of $\bold I_n$ do not realize $p^{[\varepsilon]}$, for
3143: some formula $\varphi(\bar x,\bar a) \in p$ (really $\bar x$ is a singleton)
3144: there exists $\psi(\bar x,\bar z)$ such that $\bold d_1(\varphi,\psi)
3145: > \varepsilon$ and $\forall \bar c \in \bold I_n,\psi(\bar c,\bar a)$.
3146:
3147: For $k < \omega$ let
3148: $\bar c_k \in \bold I_n$ be pairwise distinct. We can find $\bar a'
3149: \in M_n$ which realizes tp$(\bar a,A_n \cup B \cup \cup\{\bar c_k:k
3150: <\omega\})$. So by clause $(c)$ of \scite{STAB.2s.0} we know that for
3151: all but countably many $\bar c \in \bold I_n$ we have ${\frak C}
3152: \models \psi[\bar c,\bar a']$ (as this happens for $\{\bar c_k:k <
3153: \omega\} \subseteq I_n$), hence $\psi(\bar x,\bar a') \in
3154: q_n \subseteq q$.
3155: \enddemo
3156: \bn
3157: Now we obtain: for some $m> n,\bar a \in M_m$ hence
3158: $(\varphi(\bar x,\bar a),\psi(\bar x,\bar a'))$ witness that
3159: $q \restriction \cup\{M_\ell:\ell < \omega\}$ does $\varepsilon$-split
3160: over $B$, hence $q$ does, which is a contradiction to the choice of
3161: $B$. \hfill$\square$
3162: \bn
3163: We shall now proceed to proving an analogue of density of isolated
3164: types.
3165: As the ``right" notion of a type in our context seems to
3166: be an $\varepsilon$-neighborhood of a complete type, the assumption of
3167: ``non-isolated" will not be enough for us.
3168:
3169:
3170: \definition{\stag{B.10.1} Definition} We say that $M \in K_1$
3171: $(<\varepsilon)$-omits $p(\bar x)$, a type over $A \subseteq M$, when for
3172: no $\zeta \in [0,\varepsilon)_{\Bbb R}$ and $\bar b \in M$, does
3173: $\bar b$ realize $p^{[\zeta]}$.
3174: \nl
3175: \enddefinition
3176: \bigskip
3177:
3178: \definition{\stag{4n.9} Definition} 1) We say that a formula
3179: $\psi(\bar x,\bar b)$ pseudo $(\varepsilon,\zeta)$ isolates a type $p(\bar
3180: x)$ if $\psi^{<\zeta>}(\bar x,\bar b) \models p^{[\varepsilon]}(\bar x)$
3181: (note that the roles of $\varepsilon,\zeta$ are not symmetric and the
3182: different notions of approximation!). In other words, if
3183: ${\frak C} \models \psi^{<\zeta>}[\bar a,\bar b]$ then for
3184: some $\bar a' \in p({\frak C})$ we have $\bold d(\bar a',\bar a)
3185: \le \varepsilon$ (so $\bar a'$ realizes $p,\bar a$ realizes
3186: $\psi^{<\zeta>}(\bar x,\bar b)$).
3187: \nl
3188: 2) We say that $A$ is a pseudo $(< \varepsilon)$-support for
3189: $p(\bar x)$ or $A$ pseudo $(< \varepsilon)$-supports $p$
3190: when there is a consistent
3191: $\psi(\bar x,\bar b),\bar b \subseteq A$ and positive $\zeta_1,\zeta_2$ such
3192: that $\psi^{<\zeta_1>}(\bar x,\bar b) \models
3193: p^{[\varepsilon-\zeta_2]}(\bar x)$. So $\psi(\bar x,\bar b)$ psuedo
3194: $(\varepsilon-\zeta_2,\zeta_1)$-isolates $p(\bar x)$.
3195: \nl
3196: 3) We say that $A$ really $(< \varepsilon)$-omits $p(\bar x)$ if it
3197: does not pseudo $(< \varepsilon)$-support $p(\bar x)$.
3198: \enddefinition
3199: \bigskip
3200:
3201: \proclaim{\stag{4n.9.1} Claim} 1) If $M \in K^c_1 \, (< \varepsilon)$-omits
3202: $p(\bar x),p(\bar x) \in \bold S^m(A),A \subseteq M$
3203: \ub{then} $M$ really $(< \varepsilon)$-omits $p(\bar x)$.
3204: \nl
3205: 2) If $p(\bar x)$ is a type over $M$ and $M$ really
3206: $(< \varepsilon)$-omits $p(\bar x)$, then $M(< \varepsilon)$-omits $p$.
3207: \nl
3208: \endproclaim
3209: \bigskip
3210:
3211: \demo{Proof} 1) Assume $M$ is a pseudo $(< \varepsilon)$-support for $p(\bar
3212: x)$, i.e., there exist $\psi(\bar x,\bar b)$ over $M$ and
3213: $\zeta_1,\zeta_2 > 0$ such that $\psi^{<\zeta_1>}(\bar x,\bar b)
3214: \models p^{[\varepsilon - \zeta_2]}(\bar x)$. ${\frak C} \models
3215: \exists x \psi(\bar x,\bar b),M \in K^c_1$, so for some $\bar a \in
3216: M,{\frak C} \models \psi^{<\zeta_1>}(\bar a,\bar b)$ (see
3217: \scite{MOM.suB}), therefore
3218: $p^{[\varepsilon-\zeta_2]}(\bar a)$ holds, and we have
3219: $p^{(<\varepsilon)}(\bar x)$ is realized in $M$.
3220: \nl
3221: 2) Easier: assume $\bar a \models p^{[\varepsilon-\zeta]}(\bar x)$ for
3222: $\zeta > 0,\bar a \in M$. Then the formula ``$\bar x = \bar a$" is
3223: over $M$ and pseudo $(\varepsilon - \zeta_2,\zeta_1)$-isolates $p(\bar
3224: x)$ for $\zeta_2 = \zeta_1 = \frac{\zeta}{3}$.
3225: \enddemo
3226: \bigskip
3227:
3228: \definition{\stag{4n.13} Definition} 1) We say
3229: $\varphi(\bar x,\bar b)$ strictly
3230: $(\varepsilon,\zeta)$-isolates a type $p$ if $\varphi(\bar x,\bar b) \in p$ and
3231: $\varphi(\bar x,\bar b)$ pseudo $(\varepsilon,\zeta)$-isolates $p$.
3232: \nl
3233: 2) We say $\varphi(\bar x,\bar
3234: b)$ is strictly $(\varepsilon,\zeta)$-isolating over $A$ when
3235: \mr
3236: \item "{$(a)$}" $\bar b \subseteq A$
3237: \sn
3238: \item "{$(b)$}" if $\varphi(\bar x,\bar b) \in p \in \bold S^{\ell g(\bar
3239: x)}_D(A)$ then $\varphi(\bar x,\bar b)$ strictly isolates $p$.
3240: \ermn
3241: 3) We say that $p \in S^m_D(A)$ is strictly
3242: $(\varepsilon,\zeta)$-isolated if some $\varphi$ strictly
3243: $(\varepsilon,\zeta)$-isolates it.
3244: \nl
3245: 3A) ``Strictly $\varepsilon$-isolate" means ``for some $\zeta >
3246: 0$, strictly $(\varepsilon,\zeta)$-isolate".
3247: \nl
3248: 4) We say that $p \in \bold S^m_D(A)$ is strictly isolated when for every
3249: $\varepsilon > 0$ for some $\varphi(\bar x,\bar a) \in p$ and some
3250: $\zeta > 0$ the formula $\varphi(\bar x,\bar a)$
3251: strictly $(\varepsilon,\zeta)$-isolates the type $p$ (i.e., $p$ is
3252: $\varepsilon$-strictly isolated for all $\varepsilon > 0$).
3253: \enddefinition
3254: \bigskip
3255:
3256: \proclaim{\stag{4n.14} Claim} [$({\frak C},\bold d)$ is compact]
3257:
3258: Assume that
3259: \mr
3260: \item "{$(a)$}" $p \in \bold S^m_D(A)$ or just $p$ is an $m$-type closed
3261: under conjunctions
3262: \sn
3263: \item "{$(b)$}" $\psi(\bar x,\bar b) \in p$
3264: \sn
3265: \item "{$(c)$}" $\varepsilon > 0$ and $\zeta \ge 0$.
3266: \ermn
3267: \ub{Then} one of the following occurs
3268: \mr
3269: \item "{$(\alpha)$}" there is a pair $(\psi_1(\bar x,\bar
3270: y),\psi_2(\bar x,\bar y))$ of formulas and a sequence $\bar b^*$ from $A$ such
3271: that $\bar b \triangleleft \bar b^*,
3272: \ell g(\bar b^*) = \ell g(\bar y)$ such that
3273: $\psi(\bar x,\bar b) \wedge \psi_1(x,\bar b^*),\psi^{<\zeta>}
3274: (\bar x,\bar b) \wedge \psi_2(\bar x,\bar b^*)$ are
3275: $\varepsilon$-contradictory (and both consistent, of course),
3276: hence for no $\bar a,{\frak C}
3277: \models (\exists \bar x)(\bold d(\bar x,\bar a) \le \varepsilon/2
3278: \wedge \psi(\bar x,\bar b) \wedge \psi_1(\bar x,\bar b^*))$ and
3279: ${\frak C} \models (\exists \bar x)[\bold d(\bar x,\bar a) \le
3280: \varepsilon/2 \wedge \psi^{<\zeta>}(\bar x,\bar b) \wedge \psi_2(\bar
3281: x,\bar b^*)]$
3282: \sn
3283: \item "{$(\beta)$}" for every $\bar a'$ such that ${\frak C}
3284: \models \psi^{<\zeta>}[\bar a',\bar b]$ there is a sequence $\bar a''$
3285: realizing $p$ such that $\bold d(\bar a',\bar a'') \le \varepsilon$
3286: (so $\psi(x,\bar b)$ strictly $(\varepsilon,\zeta)$-isolates $p$, see
3287: \scite{4n.13}).
3288: \endroster
3289: \endproclaim
3290: \bigskip
3291:
3292: \demo{Proof} We can assume that clause $(\beta)$ fails and let $\bar a'$
3293: exemplify it. So for every $\bar a''$ realizing $p$ we have
3294: $\bold d(\bar a'',\bar a') > \varepsilon$.
3295:
3296: Let $q(\bar y) = \text{\rm tp}(\bar a',A)$ so $\psi^{<\zeta>}(\bar
3297: y,\bar b) \in q$ as ${\frak C} \models \psi^{<\zeta>}(\bar a',\bar b)$,
3298: and let $r(\bar x,\bar y) =
3299: p(\bar x) \cup q(\bar y) \cup \{\bold d(\bar x,\bar y) \le
3300: \varepsilon\}$. If $r(\bar x,\bar y)$ is consistent, so is $r(\bar
3301: x,\bar a')$, and any $\bar a''$ realizing $r(\bar x,\bar a')$ is as
3302: required in clause $(\beta)$, contradicting our assumption.
3303: So $r(\bar x,\bar y)$ is inconsistent.
3304: As $({\frak C},\bold d)$ is compact and $p(\bar x),q(\bar y)$ are
3305: closed under conjunctions,
3306: $\psi(\bar x,\bar b) \in p(\bar x),\psi^{<\zeta>}(\bar y,\bar b) \in q(\bar y)$
3307: and as we can add dummy
3308: variants, there are $\bar b^* \subseteq A,\bar b \triangleleft \bar
3309: b^*$ and $\psi_1(\bar x,\bar b^*) \in p(\bar x),\psi_2(\bar
3310: y,\bar b^*) \in q(\bar y)$ such that $\{\psi(\bar x,\bar b) \wedge
3311: \psi_1(\bar x,\bar b^*),\psi^{<\zeta>}(\bar x,\bar b) \wedge
3312: \psi_2(\bar y,\bar b^*),\bold d(\bar x,\bar y) \le
3313: \varepsilon\}$ is contradictory. \hfill$\square$
3314:
3315: So we get clause $(\alpha)$.
3316: \enddemo
3317: \bigskip
3318:
3319: \proclaim{\stag{4n.12} An isolation Claim} [$({\frak C},\bold d)$ is
3320: $0^+-\aleph_0$-stable and compact]
3321: \nl
3322: 1) If $\bar a \subseteq A$ and $\varphi(\bar x,\bar a)$ is consistent
3323: and $\varepsilon > 0$ \ub{then} we can find $\varphi_1(\bar x,\bar
3324: a_1)$ and $\zeta >0$ such that $\bar a_1 \subseteq A$ and $\varphi(\bar
3325: x,\bar a) \wedge \varphi_1(\bar x,\bar a_1)$ is strictly
3326: $(\varepsilon,\zeta)$-isolating over $A$, see Definition \scite{4n.13}.
3327: \nl
3328: 1A) Similarly omitting $\zeta$ getting strictly $\varepsilon$-isolating.
3329: \nl
3330: 2) The set of strictly isolated $p \in S^m_D(A)$ is dense, i.e. for
3331: every $\varphi(\bar x,\bar a)$ with $\bar a \in A$, there exists $p
3332: \in S^m_D(A),\varphi(\bar x,\bar a) \in p,p$ is strictly isolated.
3333: \endproclaim
3334: \bigskip
3335:
3336: \demo{Proof of \scite{4n.12}} Part (1A) is restating Part (1).
3337: Also part (2) follows from part (1) by choosing $\varphi_n(\bar x,\bar
3338: a_n)$ such that $\varphi(\bar x,\bar a) \wedge \varphi_1(\bar x,\bar
3339: a_1) \wedge \ldots \wedge \varphi_n(\bar x,\bar a_n)$ is
3340: $\frac{1}{n+1}$-isolating over $A$ (iterating 1A) and applying
3341: compactness of ${\frak C}$. So we concentrate on proving part (1) .
3342:
3343: We can choose $\zeta_n > 0$ for $n < \omega$ such that,
3344: e.g. $\Sigma\{\zeta_n:n < \omega\} \le \varepsilon/5$
3345: \nl
3346:
3347: Assume that $\psi(\bar x,\bar b),\bar b \subseteq A$ is a
3348: counterexample. Now we choose $\langle \psi_\eta(\bar x,\bar
3349: a_\eta):\eta \in {}^n 2 \rangle$ by induction on $n$ such that
3350: \mr
3351: \item "{$\boxtimes$}" $(a) \quad \bar a_\eta \subseteq A$
3352: \sn
3353: \item "{${{}}$}" $(b) \quad \psi_\eta(\bar x,\bar a_\eta)$ is
3354: consistent
3355: \sn
3356: \item "{${{}}$}" $(c) \quad \psi_{<>}(\bar x,\bar a_{<>}) = \psi(\bar
3357: x,\bar b)$
3358: \sn
3359: \item "{${{}}$}" $(d) \quad$ if $\nu {}^\frown \langle 0 \rangle,\nu
3360: {}^\frown \langle 1 \rangle \in {}^n 2$ then $\psi_{\nu {}^\frown
3361: \langle 0 \rangle}(\bar x,\bar a_{\nu {}^\frown \langle 0 \rangle}),
3362: \psi_{\nu {}^\frown \langle 1 \rangle}(\bar x,\bar a_{\nu
3363: {}^\frown \langle 1 \rangle})$ are
3364: \nl
3365:
3366: \hskip20pt $\varepsilon$-contradictory
3367: \sn
3368: \item "{${{}}$}" $(e) \quad$ if $\eta = \nu {}^\frown \langle 0
3369: \rangle \in {}^n 2$ then $\psi_\eta(\bar x,\bar a_\eta) \models
3370: \psi_\nu(\bar x,\bar a_\nu)$
3371: \sn
3372: \item "{${{}}$}" $(f) \quad$ if $\eta = \nu {}^\frown \langle 1
3373: \rangle \in {}^n 2$ then $\psi_\eta(\bar x,\bar a_\eta) \models
3374: \psi^{<\zeta_n>}_\nu(\bar x,\bar a_\nu)$.
3375: \ermn
3376: By \scite{4n.14} there is no problem to carry the definition, i.e.,
3377: having $\psi_\nu(\bar x,\bar a_\nu)$, clause $(\beta)$ of
3378: \scite{4n.14} cannot hold (with $\psi_\nu,\bar a_\nu$ here standing
3379: for $\psi,\bar b$ there) as ``$\psi(\bar x,\bar a)$ is a
3380: counterexample". Hence clause $(\alpha)$ there holds, let us
3381: choose $\psi_{\nu {}^\frown <\ell>}(\bar x,\bar a_{\nu {}^\frown
3382: <\ell>})$ for $\ell=0,1$.
3383:
3384: Now let $\xi_n = \Sigma\{\zeta_m:m \in [n,\omega)\}$, so clearly
3385: \mr
3386: \item "{$(*)$}" if $n(1) < n(2) < \omega$ and
3387: $\eta_\ell \in {}^{n(\ell)}2$ for $\ell=1,2$ and $\eta_1
3388: \triangleleft \eta_2$ then $\psi^{<\xi_{n(2)}>}_{\eta_2}(\bar
3389: x,\bar a_{\eta_2}) \models \psi^{<\xi_{n(1)}>}_{\eta_1}(\bar x,
3390: \bar a_{\eta_1})$
3391: \nl
3392: [Why? By \scite{4n.10} using clauses (e) + (f) of $\boxtimes$ we get
3393: $\psi_{\eta_2} \models \psi^{<\xi_{n(1),n(2)}>}_{\eta_1}$, where
3394: $\xi_{n(1),n(2)} = \Sigma\{\zeta_m:m \in [n(1),n(2))\}$. Now use
3395: \scite{4n.10} again.]
3396: \ermn
3397: Now let $C = \cup\{\bar a_\eta:\eta \in {}^{\omega >}2\}$.
3398: \nl
3399: Hence
3400: \mr
3401: \item "{$(*)$}" for $\eta \in {}^\omega 2$ the set
3402: $\{\psi^{<\xi_n>}_{\eta \restriction n}(\bar x,\bar a_{\eta
3403: \restriction n}):n < \omega\}$ is consistent, hence is included in some
3404: $p_\eta \in \bold S(C)$
3405: \sn
3406: \item "{$(*)$}" if $\nu {}^\frown \langle \ell \rangle \triangleleft
3407: \eta_\ell \in {}^\omega 2$ for $\ell=0,1$ then
3408: $p^{[\varepsilon/5]}_{\eta_1}(\bar x) \cup
3409: p^{[\varepsilon/5]}_{\eta_2}(\bar x)$ is inconsistent.
3410: \nl
3411: [Why? By $\boxtimes(d)$ and the choice of $\zeta_n$], a contradiction
3412: to $0^+-\aleph_0$-stability.
3413: \nl
3414: ${{}}$ \hfill$\square_{\scite{4n.12}}$
3415: \endroster
3416: \enddemo
3417: \bn
3418: \centerline {$* \qquad * \qquad *$}
3419: \bn
3420: The following is not used at present but clarifies non-categoricity.
3421: Recall Definition \scite{MOM.4A} and Claim \scite{MOM.4B}. Note that
3422: \scite{MOM.4C} says that a non-$(D,\lambda)$-homogeneuos model omits
3423: some $p^{[\varepsilon]}$. Here we clarify what happens in the case of
3424: pseudo $(D,\lambda)$-homogeneous non-$(D,\lambda)$-homogeneous model.
3425: \proclaim{\stag{4n.23} Claim} Assume $M \in K^c_1,\lambda > \aleph_0 +
3426: |\tau_{\frak C}|$ and $M$ is not $(D,\lambda)$-homogeneous. Then $(*)$ or
3427: $(**)$
3428: \mr
3429: \item "{$(*)$}" $(a) \quad N \prec^1_\Delta M,|N| < \lambda$
3430: \sn
3431: \item "{${{}}$}" $(b) \quad p \in \bold S_D(N)$
3432: \sn
3433: \item "{${{}}$}" $(c) \quad$ if ${\frak C}$ is $0^+-\mu$-stable for
3434: some $\mu \in [\aleph_0 + |\tau_{\frak C}|,\lambda)$ then $p$ has a
3435: \nl
3436:
3437: \hskip25pt unique extension in $\bold S_D(M)$
3438: \sn
3439: \item "{${{}}$}" $(d) \quad \varepsilon > 0$ and $M$ omits
3440: $p^{<\varepsilon>}$
3441: \sn
3442: \item "{$(**)$}" $(a) \quad$ for every $B \subseteq M,|B| < \lambda$
3443: and $q \in \bold S_D(B)$ and $\zeta > 0$, the type $q^{<\zeta>}$
3444: \nl
3445:
3446: \hskip25pt is realized in $M$ (so $M$ is pseudo $\lambda$-saturated)
3447: \sn
3448: \item "{${{}}$}" $(b) \quad N \prec^1_\Delta M,|\tau_{\frak C}| +
3449: \aleph_0 \le \|N\| < \lambda$
3450: \sn
3451: \item "{${{}}$}" $(c) \quad p \in \bold S_D(N)$
3452: \sn
3453: \item "{${{}}$}" $(d) \quad$ if ${\frak C}$ is $0^+-\mu$-stable for
3454: some $\mu \in [|\tau_{\frak C}| + \aleph_0,\lambda)$ then $p$ has a
3455: unique
3456: \nl
3457:
3458: \hskip25pt extension in $\bold S_D(M)$
3459: \sn
3460: \item "{${{}}$}" $(e) \quad \varepsilon > 0,p^{[\varepsilon]}$ is
3461: omitted by $M$
3462: \sn
3463: \item "{${{}}$}" $(f) \quad$ for every $n$, for some
3464: $c_n,\varepsilon_n,\zeta_n$ we have $1/(n+1) > \varepsilon_n >
3465: \zeta_n$
3466: \nl
3467:
3468: \hskip25pt $> 0,c_n \in M$ realizes $p^{<\varepsilon_n>}$ and $\bold
3469: d(c_n,p^{<\zeta_n>}({\frak C})) \ge 10 \times \varepsilon_n - \zeta_n$.
3470: \endroster
3471: \endproclaim
3472: \bigskip
3473:
3474: \demo{Proof} Clearly there is $A \subseteq M$ such that
3475: $|A| < \lambda$ and $p \in \bold S^1_D(A)$ is omitted.
3476: Let $\mu = |A| + |\tau_{\frak C}| + \aleph_0$ so $\mu < \lambda$. If
3477: ${\frak C}$ is $0^+-\mu'$-stable for some $\mu' \in [|\tau_{\frak C}|
3478: + \aleph_0,\lambda)$, easily \wilog \, $p$ has a unique extension in
3479: $\bold S_D(M)$ and $A = |N|,N \prec^1_\Delta M$. If for some
3480: $\varepsilon > 0,p^{<\varepsilon>}$ is also omitted by $M$, the case $(*)$
3481: holds. So we may assume $(*)$ fails, in other words, clause (a) of
3482: $(**)$ holds.
3483:
3484: Let $n^* < \omega,\varepsilon^* = \frac 12$ (any $\varepsilon^* > 0$
3485: works). We are going to find $c_{n^*},\varepsilon_{n^*},\zeta_{n^*}$
3486: as required in clause (f) above. First, we try to choose $b_n
3487: \in M$ by induction on $n < \omega$ such that
3488: \mr
3489: \item "{$\circledast_n$}" $(a) \quad b_n \in M$
3490: \sn
3491: \item "{${{}}$}" $(b) \quad b_n$ realizes $p^{<\varepsilon^*/(n+1)>}$
3492: \sn
3493: \item "{${{}}$}" $(c) \quad$ if $n=m+1$ then $\bold d(b_n,b_m) \le 10
3494: \times \varepsilon^*/2^n$.
3495: \ermn
3496: For $n=0$ there is $b_0 \in M$ realizing $p^{<\varepsilon^*/1>} =
3497: p^{<\varepsilon^*>}$ by clause (a) of $(**)$. So we can begin.
3498:
3499: \bn
3500: \ub{Point 1}: We cannot succeed to choose $\langle b_n:n < \omega
3501: \rangle$.
3502: \nl
3503: Why? Suppose we have succeeded. Then
3504: $\langle b_n:n < \omega \rangle$ is a Cauchy sequence and therefore
3505: converges to some $b^* \in M$. We will show that $b^* \models p$.
3506: If $\varphi(x,\bar a) \in p$, then for each $n,{\frak C} \models
3507: \varphi^{<\varepsilon^*/(n+1)>}(b_n,\bar a)$ so there is $\bar a_n
3508: \in {}^{\omega >}{\frak C},b'_n \in {\frak C}$ such that
3509: \mr
3510: \item "{$\boxtimes$}" $(a) \quad \bold d(b'_n,b_n) \le
3511: \varepsilon^*/(n+1)$
3512: \sn
3513: \item "{${{}}$}" $(b) \quad \bold d(\bar a_n,\bar a) \le
3514: \varepsilon^*/(n+1)$
3515: \sn
3516: \item "{${{}}$}" $(c) \quad \models \varphi[b'_n,\bar a_n]$.
3517: \ermn
3518: Now
3519:
3520: $$
3521: \langle b'_n:n < \omega \rangle \text{ converges to } b^*
3522: $$
3523:
3524: $$
3525: \langle \bar a_n:n < \omega \rangle \text{ converges to } \bar a
3526: $$
3527: \mn
3528: hence $\models \varphi[b^*,\bar a]$. So $b^* \models p,b* \in M$, a
3529: contradiction.
3530: \enddemo
3531: \bn
3532: \ub{Point 2}:
3533:
3534: So we are stuck in some $n=m+1$ so let $c_{n^*} := b_m,
3535: \varepsilon_{n^*} = \varepsilon^*/2^n,\zeta_{n^*} =
3536: \varepsilon^*/2^{n+2}$. If the demand in (f) of $(**)$ fails, then
3537: there is $b'_m \in {\frak C}$ realizing $p^{<\zeta_{n^*}>}(\bar x) \cup
3538: \{\bold d(x,b_m) \le 10 \times \varepsilon_{n^*} - \zeta_{n^*}\}$ hence
3539: $p^{<\zeta_{n^*}>}(x) \cup \{\bold d(x,b_m) \le 10 \times
3540: \varepsilon_{n^*} - \zeta_{n^*}\}$
3541: is consistent hence it is contained in some $q_n \in \bold S(N \cup\{b_m\})$.
3542:
3543: So for every $\zeta >0$ (we use $\zeta$ small enough) there is $b_n
3544: \in M$ realizing $q^{<\zeta>}(x)$ (recall we are assuming
3545: $(**)(a)!$). So $b_n$ realizes $(p^{<\zeta_{n^*}>})^{<\zeta>}$
3546: hence $p^{<\zeta_{n^*}+ \zeta>}$ hence if
3547: $\zeta$ is small enough, $p^{<\varepsilon^*/(n+1)>}$. Also $\bold
3548: d(b_n,b_m) \le (10 \times \varepsilon_{n^*} - \zeta_{n^*}) + \zeta + \zeta
3549: < 10 \times \varepsilon_{n^*} = 10 \times \varepsilon^*/2^n$
3550: (because if $(a',b')$ realized $\bold d^{<\zeta>}(x,y) \le \xi$
3551: then $\bold d(a',b') \le \xi + \zeta +
3552: \zeta$).
3553:
3554: So $b_n$ is as required in $\circledast_n (a)-(c)$ above, so we could
3555: have continued choosing the $b_n$. \hfill$\square_{\scite{4n.12}}$
3556: \bigskip
3557:
3558: \head {\S5 Ehrenfeucht-Mostowski models} \endhead \resetall \sectno=5
3559: \spuriousreset
3560: \bigskip
3561:
3562: In this section we adapt the technique of constructing
3563: Ehrenfeucht-Mostowski models to our context. The reader should have a
3564: look at chapter 7 of \cite{Sh:c} for the basic definitions ($\Phi$
3565: proper, etc.). The basic idea is the following: we start with ${\frak
3566: C}$ in vocabulary $\tau$. Adding skolem functions, we obtain
3567: vocabulary $\tau'$. Choosing an indiscernible sequence and taking its
3568: type (its EM - ``blueprint") $\Phi$, for each order type $J$ we can
3569: construct EM$(J,\Phi)$ (like in chapter 7 of \cite{Sh:c}), which will
3570: be an elementary submodel of ${\frak C}$ expanded to $\tau'$, therefore
3571: its restriction to $\tau$, EM$_\tau(J,\Phi)$ is an elementary submodel of
3572: ${\frak C}$, although not necessarily complete. Taking the
3573: completion, we obtain a model in $K^c_1$.
3574: Adding more structure to the language we can make it
3575: $(D,\lambda)$-homogeneous, and more, see below.
3576:
3577: Given a vocabulary $\tau^*$ with skolem functions, and a
3578: $\tau^*$-diagram of indiscernibles $\Phi$ (EM-blueprint), we denote
3579: for each order-type $I$, the EM-model (the $\tau^*$-skolem hull of a
3580: sequence $\langle a_i:i \in I \rangle$) by EM$_{\tau^*}(I,\Phi)$ or
3581: EM$(I,\Phi)$ if $\tau^*$ is clear from the context. We denote by
3582: EM$_{\tau_0}(I,\Phi)$ the restriction of EM$(I,\Phi)$ to the vocabulary
3583: $\tau_0 \subseteq \tau^*$.
3584:
3585: Let ${\frak C}$ be a momspace. Let $\tau$ be the vocabulary of ${\frak C}$.
3586: It is clear that for any $\tau^*$ (with skolem functions) expanding
3587: $\tau$, a $\tau^*$-diagram of indiscernibles $\Phi$ (in ${\frak C}$ expanded
3588: to $\tau^*$), $I$ an order, we can think of
3589: EM$_\tau(I,\Phi)$ as an elementary submodel of ${\frak C}$, so
3590: EM$_\tau(I,\Phi) \prec_{\Bbb L(\tau({\frak C}))} {\frak C}$. This is
3591: not necessarily true for the completion, but
3592: $\overline{\text{EM}_\tau(I,\Phi)} \prec^1_\Delta {\frak C}$ by
3593: \scite{MOM.2.19}.
3594: \bigskip
3595:
3596: \proclaim{\stag{EM.1} Claim} Let $({\frak C},\bold d)$
3597: be a momspace, $\tau$ the
3598: vocabulary of ${\frak C},|\tau| \le \aleph_0,\tau' \supseteq
3599: \tau,\tau'$ with Skolem functions, $\Phi'$ a $\tau'$-blueprint.
3600: \nl
3601: 0A) For every linear order $J$, {\rm EM}$_\tau(J,\Phi') \in K_1$
3602: \nl
3603: and {\rm mcl}$(\text{\rm EM}_\tau(J,\Phi')) \in K^c_1$.
3604:
3605: 0B) If $J_1 \subseteq J_2$ then {\rm EM}$_\tau(J_1,\Phi')
3606: \prec^1_\Delta \text{\rm EM}_\tau(J_2,\Phi')$; moreover {\rm
3607: EM}$_\tau(J_1,\Phi') \prec \text{\rm EM}_\tau(J_1,\Phi')$ and
3608: {\rm mcl}$(\text{\rm EM}_\tau(J_1,\Phi')) \prec^1_\Delta \text{\rm
3609: mcl}(\text{\rm EM}_\tau(J_2,\Phi'))$.
3610: \nl
3611:
3612: 1) There exists $\tau^*$ expanding $\tau',|\tau^*| = 2^{\aleph_0}$ and
3613: a $\tau^*$-diagram $\Phi^*$ such that for each finite order $J$,
3614: EM$_\tau(J,\Phi^*)$ is $(D,\aleph_1)$-homogeneous.
3615: \nl
3616: 2) If ${\frak C}$ is $(\aleph_1,*)$-superstable, then $\Phi^*$ as in (1) works
3617: for all orders $J$, but we have to take the closure,
3618: i.e., mcl(EM$_\tau(J,\Phi^*))$ which $\in K^c_1$ is
3619: $(D,\aleph_1)$-homogeneous for all $J$.
3620: \nl
3621: 3) If ${\frak C}$ is $0^+-\omega$-stable, then $\tau^*$ in (1) and (2)
3622: can be
3623: chosen of cardinality $\aleph_1$.
3624: \endproclaim
3625: \bigskip
3626:
3627: \demo{Proof} 0) (A),(B) straight.
3628: \nl
3629: 1) Choose for $i < \aleph_1,\tau_i,\Phi_i,|\tau_i|
3630: = 2^{\aleph_0}$ with skolem functions expanding $\tau'$
3631: increasing continuous such that for each
3632: $\tau_i$-type $p$ over a finite subset of the skeleton of
3633: EM$(I,\Phi_i)$, say $a_1,\dotsc,a_n$, there exists a function symbol
3634: $f_p$ in $\tau_{i+1}$ such that $f_p(a_1,\dotsc,a_n)$ realizes $p$.
3635:
3636: More precisely, we do the following: for any consistent set $p$ of
3637: formulas of the form $\varphi = \varphi(x,y_1,\dotsc,y_n) \in \tau_i$
3638: ($y_1,\ldots,y_n$ are the parameters; some of the $y_i$'s may be
3639: dummy variables) such that $p$ is closed under conjunctions and for every
3640: $\varphi \in p,\exists x \varphi(x,y_1,\dotsc,y_n) \in \Phi_i$, we add
3641: a function symbol $f_p$ to $\tau_{i+1}$ such that for every $\varphi \in
3642: p$ the following formula is in $\Phi_{i+1}:\exists x
3643: \varphi(x,y_1,\dotsc,y_n) \rightarrow
3644: \varphi(f_p(y_1,\dotsc,y_n),y_1,\dotsc,y_n)$.
3645: \nl
3646: Now let $\Phi^* = \dbcu_{i < \aleph_1} \Phi_i$ and let $M = \text{\rm
3647: EM}_\tau(J,\Phi^*)$ for some finite $J$. Choose $A \subseteq M$
3648: countable, $p \in S(A)$. As $A$ is countable, it can be viewed as a
3649: countable subset of EM$(J,\Phi_i)$ for some $i,p$ is a type over the finite
3650: skeleton, so realized in EM$(J,\Phi_{i+1})$, therefore in $M$, as
3651: required.
3652: \nl
3653: 2) By induction on $|J|$.
3654: We just need to show that for an increasing sequence of linear orders
3655: $J_i,\dbcu_i \text{\rm EM}(J_i,\Phi) =
3656: \text{\rm EM}(\dbcu_i J_i,\Phi)$ and this is clear as
3657: elements of EM$(J,\Phi)$ have finite character, i.e., use only
3658: finitely many elements of the skeleton $J$. Of course, we then have
3659: to take metric closure.
3660: \nl
3661: 3) Let $J_n$ be a linear order with $n$ elements.
3662: Similarly to (1), we choose by induction on $i < \omega_1$
3663: countable $\tau_i \subseteq \tau^*$ increasing continuous, closed
3664: under skolem functions, and $\Phi_i$ such that each type over
3665: EM$(J_n,\Phi_i)$ is almost realized in EM$(J_n,\Phi_{i+1})$: we choose a
3666: countable set $B$ which almost realizes all types over EM$(J_n,\Phi_i)$,
3667: and for each such $p$ and for each $\kappa$ we have
3668: $f_{p,\kappa} \in \Phi_{i+1}$ such
3669: that $f_{p,\kappa}(\bar a)$ is $\frac 1k$-close to a realization of
3670: $\varphi$ for each $\varphi \in p$.
3671: \enddemo
3672: \bigskip
3673:
3674: \proclaim{\stag{EM.2} Corollary} If ${\frak C}$ is
3675: $0^+-\omega$-stable, it has a $(D,\aleph_1)$-homogeneous model
3676: in all uncountable density characters.
3677: \endproclaim
3678: \bigskip
3679:
3680: \demo{Proof} Let $\lambda > \aleph_0$.
3681: Consider mcl(EM$(\lambda,\Phi^*)),\Phi^*$ as in
3682: \scite{EM.1}(3) (note that $|\Phi^*| \le \aleph_1$) and use
3683: \scite{EM.1}(2) + \scite{STAB.5}.
3684: \enddemo
3685: \bn
3686: \margintag{EM.3}\ub{\stag{EM.3} Discussion}: If $({\frak C},\bold d)$ is
3687: $0^+-\aleph_0$-stable, does it have a $(D,\lambda)$ homogeneous model in
3688: every $\lambda$? By \scite{UNI.1} this follows from categoricity,
3689: which is good enough for our purposes.
3690: \bigskip
3691:
3692: \head {\S6 Embeddings, isomorphisms and categoricity} \endhead \resetall \sectno=6
3693: \spuriousreset
3694: \bigskip
3695:
3696: In this section we introduce notions of $\varepsilon$-embedding and
3697: $\varepsilon$-isomorphism which are weaker than isometry.
3698: This will lead us to the notion of weak uncountable categoricity that
3699: we investigate in \S8.
3700: \bigskip
3701:
3702: \demo{\stag{ISO.0S} Convention} Models are from $K=K_1$.
3703: \enddemo
3704: \bigskip
3705:
3706: \definition{\stag{ISO.1} Definition} For two metric structures in the
3707: same vocabulary $\tau$ and $\varepsilon \ge 0$ we say
3708: \mr
3709: \item $f:M_1 \rightarrow M_2$ is an $\varepsilon$-embedding if for every
3710: $\Delta$-formula $\varphi,\bar a \in M_1,M_1 \models \varphi(\bar a)
3711: \Rightarrow M_2 \models \varphi^{[\varepsilon]}(\bar a)$
3712: \sn
3713: \item $f:M_1 \rightarrow M_2$ is an $\varepsilon$-isomorphism if it is
3714: an $\varepsilon$-embedding which is one-to-one and onto
3715: \sn
3716: \item $M_1,M_2$ are $\varepsilon^+$-isomorphic if there exists a
3717: $\zeta$-isomorphism $f_\zeta:M_1 \rightarrow M_2$ for all $\zeta >
3718: \varepsilon$
3719: \sn
3720: \endroster
3721: \enddefinition
3722: \bigskip
3723:
3724: \demo{\stag{ISO.2} Observation} 1) $0$-embedding is a regular notion of
3725: (isometric) embedding, $0$-isomorphisms is regular isomorphism (in
3726: particular isometry).
3727: \nl
3728:
3729: 2) If there exists a $\zeta$-isomorphism $f_\zeta:M_1 \rightarrow
3730: M_2$ for all $\zeta > \varepsilon$, then there exists a
3731: $\zeta$-isomorphism $g_\zeta:M_2 \rightarrow M_1$ for all $\zeta >
3732: \varepsilon$ (so clause (3) of the definition above makes sense).
3733: \enddemo
3734: \bigskip
3735:
3736: \demo{Proof} Clear.
3737: \enddemo
3738: \bigskip
3739:
3740: The following definition is the central one.
3741: \definition{\stag{ISO.3} Definition} Let $({\frak C},\bold d)$ be a
3742: momspace, $\varepsilon \ge 0,\lambda$ a cardinal.
3743: \nl
3744: 1) We say ${\frak C}$ is $\varepsilon^+$-categorical in $\lambda$ if
3745: every two complete $M_1,M_2 \in K^c_1$ of density $\lambda$ are
3746: $\varepsilon^+$-isomorphic.
3747: \nl
3748: 2) We say ${\frak C}$ is categorical in $\lambda$ if every two complete
3749: $M_1,M_2 \in K^c_1$ of density $\lambda$ are isomorphic.
3750: \nl
3751: 3) We say that ${\frak C}$ is possibly categorical
3752: ($\varepsilon^+$-categorical) if it is categorical
3753: ($\varepsilon^+$-categorical) in some $\lambda > \aleph_0$.
3754: \nl
3755: 4) We say that ${\frak C}$ is weakly uncountably categorical
3756: (wu-categorical) if the following holds: for each $\varepsilon >0$
3757: there exists a cardinal $\lambda$ such that ${\frak C}$ is
3758: $\varepsilon^+$-categorical in $\lambda$.
3759: \enddefinition
3760: \bigskip
3761:
3762: \demo{\stag{ISO.4} Observation} 1) If $\varepsilon \ge \zeta$ then
3763: $\zeta^+$-categoricity implies $\varepsilon^+$-categoricity (for a
3764: specific $\lambda$).
3765: \nl
3766: 2) Possible $0^+$-categoricity implies weak uncountable
3767: categoricity.
3768: \nl
3769: 3) Categoricity implies all the other notions (for a specific $\lambda$).
3770: \enddemo
3771: \bigskip
3772:
3773: \proclaim{\stag{OS} Theorem} [$({\frak C},\bold d)$ compact]
3774: \nl
3775: Let $K = K^c_1({\frak C})$ be a wu-categorical
3776: momspace with countable language. Then ${\frak C}$ is $0^+-\aleph_0$-stable.
3777: \endproclaim
3778: \bigskip
3779:
3780: \demo{Proof} Let $\tau$ be $\tau(K),\tau'$ is $\tau$ expanded with skolem
3781: functions, $\Phi$-proper for $K$. So $\tau'$ is countable.
3782: \enddemo
3783: \bigskip
3784:
3785: \demo{\stag{OS.1} Subclaim} Under these assumptions, let $I$ be a
3786: well-ordered set, $M_0 = \overline{\text{EM}_\tau(I,\Phi)}$, $A
3787: \subseteq M_0$ is countable $\varepsilon > 0$. Then
3788: each $\varepsilon$-disjoint set ${\Cal P}$ of types from $\bold S_D(A),
3789: \frac{\varepsilon}{2}$-realized in $M_0$ (so
3790: $p_1,p_2 \in {\Cal P} \Rightarrow p^{[\varepsilon]}_1 \cup
3791: p^{[\varepsilon]}_2$ is contradictory and for each $p \in
3792: {\Cal P},p^{[\frac{\varepsilon}{2}]}$ is realized in $M_0$) is countable.
3793: \enddemo
3794: \bigskip
3795:
3796: \demo{Proof of the Subclaim} If not, let $\langle p_i:i < \omega_1
3797: \rangle$ be $\varepsilon$-disjoint types over
3798: $A,p^{[\frac{\varepsilon}{2}]}_i$ realized in $M_0$ by $\bar b_i$.
3799: Pick $\bar b^0_i \in$ {\rm EM}$(I,\Phi),\bold d(\bar b^0_i,\bar b_i) \le
3800: \frac{\varepsilon}{100}$. Without loss of generality $A = \text{\rm
3801: EM}(J,\Phi)$ for $J \subseteq I,|J| \le \aleph_0$. As $J$ is well
3802: ordered, by the standard argument, there are uncountably many
3803: $b^0_i$'s satisfying the same type over $A$, but $b^0_i \models
3804: p^{[\varepsilon]}_i$ and $p^{[\varepsilon]}_i,p^{[\varepsilon]}_j$ are
3805: contradictory for $i \ne j$, a contradiction. \hfill$\square_{\scite{OS.1}}$
3806: \enddemo
3807: \bn
3808: Now we prove the theorem. Assuming ${\frak C}$ is not
3809: $0^+-\omega$-stable, we get $A \subseteq {\frak C}$ countable and
3810: $\langle p_i:i < \omega_1 \rangle \, \varepsilon$-disjoint types over
3811: $A$ for some $\varepsilon > 0$ (remember \scite{STAB.2}). Let $\lambda$
3812: be such that ${\frak C}$ is $\delta^+$-categorical in $\lambda$ for
3813: $\delta << \varepsilon$. Now apply the usual argument: choose $M_1
3814: \in K$ of density $\lambda$ which includes $A$ and $\langle b_i:i <
3815: \omega_1\rangle$ realizations of $\langle p_i:i < \omega_1 \rangle$,
3816: and on the other hand consider $M_0 =
3817: \overline{\text{EM}(\lambda,\Phi)}$. Applying $f:M_1 \rightarrow
3818: M_0$ which is a $\delta_1$-embedding $\delta_1 <
3819: \frac{\varepsilon}{2}$, we get that $\langle f(b_i):i
3820: < \omega_1\rangle$ contradict \scite{OS.1}. \hfill$\square_{\scite{OS}}$
3821: \bigskip
3822:
3823: \proclaim{\stag{MOM.EXIST} Corollary} Let ${\frak C}$ and $K$ be as in
3824: \scite{OS}, then $K$ has a $(D,\aleph_1)$-homogeneous model
3825: in $\lambda$ (recall this means of density $\lambda$)
3826: for all $\lambda > \aleph_0$.
3827: \endproclaim
3828: \bigskip
3829:
3830: \demo{Proof} By \scite{OS} and \scite{EM.2}.
3831: \enddemo
3832: \bigskip
3833:
3834: \head {\S7 Uni-dimensionality} \endhead \resetall \sectno=7
3835: \spuriousreset
3836: \bn
3837:
3838: The following notion was explored in \cite{Sh:3} but not defined there:
3839: \definition{\stag{UD.1} Definition} A (good) finite diagram
3840: $D$ is \ub{uni-dimensional} if for some regular $\lambda$ there is no
3841: $(D,\lambda)$-homogeneous model of $K$ which is not
3842: $\lambda^+$-homogeneous in cardinality $\ge \lambda^+$.
3843: \enddefinition
3844: \bn
3845: In \cite{Sh:3} it is essentially proven (see \cite[\S6]{Sh:3}) that:
3846: \proclaim{\stag{UD.2} Theorem} Assume $D$ is stable. Then the
3847: following are equivalent:
3848: \mr
3849: \item $D$ is not uni-dimensional
3850: \sn
3851: \item there is some regular $\lambda$ such that there are maximally
3852: $(D,\lambda)$-homogeneous models of arbitrary large cardinalities
3853: \sn
3854: \item for all large enough regular $\lambda < \mu$, there is a
3855: $(D,\lambda)$-homogeneous model $M$ and $\langle a_i:i < \mu
3856: \rangle,\langle b_i:i < \lambda \rangle$ mutually indiscernible
3857: sequences in $M$ such that $\langle b_i:i < \lambda \rangle$ is a
3858: maximal indiscernible sequence in $M$, i.e., can not be extended in
3859: $M$
3860: \sn
3861: \item there is a cardinal $\lambda$ and a model $M$ of cardinality
3862: $\lambda$ which is $(D,\aleph_1)$-homogeneous, but not
3863: $(D,\lambda)$-homogeneous.
3864: \endroster
3865: \endproclaim
3866: \bigskip
3867:
3868: \remark{\stag{UD.2A} Remark} In our context we will say ``${\frak C}$
3869: is uni-dimensional" or ``$K$ is uni-dimensional", meaning that $D$ is.
3870: \endremark
3871: \bigskip
3872:
3873: \demo{\stag{UD.3} Reminder} 1) For an indiscernible set $\bold I \subseteq
3874: {\frak C}$ of cardinality $> |\tau_{\frak C}| + \aleph_0$
3875: and a set $A \subseteq {\frak C}$, we define
3876:
3877: $$
3878: \align
3879: \text{Av}(\bold I,A) = \{\varphi(\bar x,\bar a):&\bar a \in A,
3880: \text{ infinitely many elements} \\
3881: &\text{ of } \bold I \text{ satisfy } \varphi(\bar x,\bar a)\}.
3882: \endalign
3883: $$
3884: \mn
3885: We call this set the \ub{average type} of $\bold I$ over $A$. (See
3886: \scite{STAB.2s.0} ``all but finitely many" is wrong, see
3887: \cite{Sh:863} on ``cutting indiscernibles", the weaker version fits here.)
3888: \nl
3889: 2) For stable ${\frak C}$, for any indiscernible $\bold I,|\bold I|
3890: > |\tau_{\frak C}| + \aleph_0$ and set $A$, Av$(\bold I,A)$ is a
3891: complete type, see \scite{ST.3}(1).
3892: \nl
3893: 3) Let $\bold I = \langle \bar a_i:i < \alpha \rangle$, where $\langle \bar
3894: a_i:i \le \alpha\rangle$ is indiscernible. Then $\bar a_\alpha
3895: \models \text{\rm Av}(\bold I,\cup \bold I)$.
3896: \nl
3897: 4) Let $\bold I$ be indiscernible, $\bold I =
3898: \langle \bar a_i:i < \alpha \rangle$
3899: and let $\bar a_\alpha \models \text{\rm Av}(\bold I,\cup \bold I)$.
3900: Then $\langle \bar a_i:i \le \alpha \rangle$ is indiscernible.
3901: \nl
3902: 5) It follows from (3) + (4) that $\bold I = \langle \bar a_i:i < \alpha
3903: \rangle \subseteq M$ is a maximal indiscernible sequence (set) in $M$
3904: iff Av$(\bold I,\cup \bold I)$ is omitted in $M$.
3905: \nl
3906: 6) $\varphi(x,\bar a_{i_1},\dotsc,\bar a_{i_n}) \in \text{\rm
3907: Av}(\bold I,\cup \bold I)$
3908: for $\bold I = \langle \bar a_i:i < \delta\rangle$ ($\delta$-limit ordinal) iff
3909: $\bar a_j \models \varphi(x,\bar a_{i_1},\dotsc,\bar a_{i_n})$
3910: for some/every $j \notin \{i_1,\dotsc,i_n\}$.
3911: \enddemo
3912: \bigskip
3913:
3914: \proclaim{\stag{UNI} Theorem} 1) Let $({\frak C},\bold d)$ be a
3915: momspace, $\tau({\frak C})$ countable, $0^+$-categorical in
3916: $\lambda,\lambda > \aleph_0$. Then ${\frak C}$ is uni-dimensional.
3917: \nl
3918: 2) The same is true if ${\frak C}$ is wu-categorical.
3919: \endproclaim
3920: \bigskip
3921:
3922: \demo{Proof} 1) If not, choose $0 < \theta_1 << \theta_2$ and let $M$
3923: be a model $M \in K,|M| = \theta_2,M$ is $\theta_1$-homogeneous,
3924: $\langle a_i:i < \theta_2 \rangle,\langle b_i:i < \theta_1 \rangle$
3925: mutually indiscernible, $\langle a_i:i < \theta_1 \rangle$ cannot be
3926: extended in $M$, so (denoting $\bold I = \langle a_i:i < \theta_2
3927: \rangle,\bold J = \langle b_i:i < \theta_1 \rangle$)
3928: Av$(\bold J,\cup \bold J)$ is omitted in $M$.
3929:
3930: We now expand the language by a predicate $P$ for $J$ and skolem
3931: functions, call the new vocabulary $\tau'$. Let $T' = \text{\rm
3932: Th}_{\tau'}(M')$ (where $M'$ is $M$ in the expanded language).
3933:
3934: Note that
3935: \mr
3936: \item "{$\circledast_0$}" $T' \models ``P$ is a $\tau$-indiscernible
3937: set",i.e., for every $\tau$-formula, $T'$ implies that any two tuples
3938: from $P$ behave the same.
3939: \ermn
3940: The type $p = \text{\rm Av}_\tau(\bold J, \cup \bold J)$ is omitted
3941: in $M$, therefore \wilog \, by \scite{MOM.4C},
3942: for some $\varepsilon,p^{[\varepsilon]}$ is omitted. If we choose
3943: $\theta_1,\theta_2$ carefully enough $(\theta_2 >> \theta_1)$ in the
3944: beginning, then by the Erd\"os-Rado theorem (as in the proof of
3945: \cite[VIII,5.3]{Sh:c}, or using \cite{BY03a}(1.2)) we can choose a diagram
3946: of indiscernibles (EM-blueprint) $\Phi$ in vocabulary $\tau'$ such
3947: that for any $\mu$, denoting the skeleton of $M'_0 :=
3948: \,\text{\rm EM}(\mu,\Phi)$ by $I' = \langle a'_i:i < \mu \rangle$, we have
3949: \mr
3950: \item "{$\circledast_1$}" $\bold I'$ is a $\tau'$-indiscernible sequence
3951: (set), moreover, it is $\tau'$-indiscernible over $P^{M'_0}$
3952: \sn
3953: \item "{$\circledast_2$}" for each $n < \omega$, for some
3954: $i_1,\dotsc,i_n < \theta_2,a'_0,\dotsc,a'_{n-1}$ has the same
3955: $\tau'$-type as $a_{i_1},\dotsc,a_{i_n}$.
3956: \ermn
3957: Now:
3958: \mr
3959: \item "{$\circledast_3$}" let $M_0 = \text{ EM}_\tau(\mu,\Phi)$, then
3960: $M_0 \prec {\frak C}$, so mcl$(M_0) \in K = K^c_1$ (as on the one hand
3961: $\tau'$ has skolem functions, and on the other hand $M'_0$ does not
3962: realize $\tau$-types over $\emptyset$ that were not realized in $M$,
3963: so $M_0$ is a $D$-model)
3964: \sn
3965: \item "{$\circledast_4$}" $M'_0 \models T'$ (skolem functions, so
3966: $M'_0 \equiv M'$)
3967: \sn
3968: \item "{$\circledast_5$}" $P^{M'_0}$ is a $\tau$-indiscernible set
3969: (by $\circledast_4$ and $\circledast_0$ above)
3970: \sn
3971: \item "{$\circledast_6$}" $P^{M'_0}$ is countable. Why? Each $b \in
3972: P^{M'_0}$ is of the form $\sigma(a'_{i_1},\dotsc,a'_{i_n})$ for some
3973: $\tau'$-term $\sigma$. But as $\bold I'$ is
3974: $\tau'$-indiscernible over $P^{M'_0}$,
3975: each such $b$ depends only on $\sigma$, and there are countably many
3976: $\tau'$-terms ($\tau$ is countable, and so is $\tau'$).
3977: \ermn
3978: Denote $\bold J' = P^{M'_0}$, a $\tau$-indiscernible set.
3979: Denote $p' = \text{\rm Av}_\tau(\bold J', \cup \bold J')$. Then
3980: \nl
3981: $[p']^{[\varepsilon]}$ is omitted in $M_0$.
3982: \nl
3983: Why? Pick $\sigma(a'_{i_j},\dotsc,a'_{i_n}) \in M'_0$. Let
3984: $j_1,\dotsc,j_n$ be such that $a'_{i_1},\dotsc,a'_{i_n} \equiv
3985: a_{j_1},\dotsc,a_{j_n},\sigma(a_{j_1},\dotsc,a_{j_n}) \in M'$ does not
3986: realize $p^{[\varepsilon]}$, so for some $\varphi(x) \in p,M' \models
3987: \bold d_1(\sigma(\bar a_j),\varphi) \ge \varepsilon$.
3988: \nl
3989: Note: $\varphi(x) = \varphi(x,\bar c),\bar c \in P^{M'}$. Call
3990: that $\varphi(x,\bar c) \in p \Leftrightarrow M' \models
3991: \varphi(d,\bar c)$ for some/all $d \in P^{M'},d \cap \bar c =
3992: \emptyset$ (as $P^{M'} = J$ is an indiscernible set, see
3993: \scite{UD.3}(6)). So $M' \models
3994: ``\exists \bar c \in P$ such that $[\forall d \in P \backslash \bar
3995: c,\varphi(d,\bar c)] \and [\bold d_1(\sigma(\bar a_j),\varphi(x,\bar c)) \ge
3996: \varepsilon]"$. Therefore, $M'_0$ satisfies the same formula with
3997: $\sigma(\bar a'_i)$, which obviously means that $\sigma(\bar a'_i)$ does not
3998: satisfy $[p']^{[\varepsilon]}$, as required.
3999:
4000: We have finished now: let $\mu = \lambda$, so in $\bar M_0 =
4001: \overline{\text{EM}_\tau(\lambda,\Phi)}$ we have a countable
4002: indiscernible set whose average is omitted (as $(p')^{[\varepsilon]}$
4003: is omitted in $M_0$), so $\bar M_0$ is a
4004: non-$(D,\aleph_1)$-homogeneous model in ${\frak K}$ of density
4005: $\lambda$, but by \scite{MOM.EXIST} we have a
4006: $(D,\aleph_1)$-homogenous model of density $\lambda > \aleph_0$. So
4007: categoricity in $\lambda$ fails, moreover, $0^+$-categoricity fails,
4008: as $\bar M_0$ is at least $\frac{\varepsilon}{2}$-distant from any
4009: $(D,\aleph_1)$-homogeneous model.
4010: \nl
4011: 2) Repeat the proof of (1), and at the end choose $\lambda$ in which,
4012: say, $(\frac{\varepsilon}{2})$-categoricity holds, and get the same
4013: contradiction.
4014: \enddemo
4015: \bigskip
4016:
4017: \proclaim{\stag{UNI.1} Claim} Let ${\frak C}$ be a wu-categorical
4018: momspace, $K = K^c_1({\frak C})$.
4019: \nl
4020: 1) There exists a $(D,\lambda)$-homogeneous model in $K$ for all
4021: $\lambda > \aleph_0$.
4022: \nl
4023: 2) Each $(D,\aleph_1)$-homogeneous model in $K$ of density $\lambda$
4024: is $(D,\lambda)$-homogeneous.
4025: \nl
4026: 3) If $K$ is $0^+$-categorical in $\lambda > \aleph_0$, then each
4027: $K$-model of density $\ge \lambda$ is $(D,\lambda)$-homogeneous.
4028: \endproclaim
4029: \bigskip
4030:
4031: \demo{Proof} 1) By \scite{MOM.EXIST} and uni-dimensionality.
4032: \nl
4033: 2) By the equivalence \scite{UD.2}.
4034: \nl
4035: 3) Otherwise by a L\"owenheim-Skolem argument we will get a
4036: non-$(D,\lambda)$-homogeneous model of density $\lambda$, and
4037: together with (1) this will lead to a contradiction.
4038: \enddemo
4039: \bigskip
4040:
4041: \head {\S8 The main theorem} \endhead \resetall \sectno=8
4042: \spuriousreset
4043: \bigskip
4044:
4045: \demo{\stag{CATEG.0} Hypothesis} ${\frak C}$ is a compact momspace
4046: with countable vocabulary.
4047: \enddemo
4048: \bigskip
4049:
4050: \proclaim{\stag{CATEG} Theorem} Assume $K = K^c_1({\frak C})$ wu-categorical.
4051: \ub{Then} $K$ is $\lambda$-categorical for
4052: all $\lambda > \aleph_0$, moreover, any model of $K$ of
4053: density $\lambda > \aleph_0$ is $(D,\lambda)$-homogeneous.
4054: \endproclaim
4055: \bigskip
4056:
4057: \demo{Proof} Suppose not, so ${\frak C}$ is $0^+-\aleph_0$-stable,
4058: $(\aleph_1,*)$-superstable, uni-dimensional by \scite{OS},
4059: \scite{STAB.5}, \scite{UNI}, and
4060: there are $\lambda > \aleph_0,M^* \in K^c_1$,
4061: ch$(M^*) = \lambda > \aleph_0,M^*$ is not $(D,\lambda)$-homogeneous.
4062: By \scite{UNI.1}, $M^*$ is not $(D,\aleph_1)$-homogeneous.
4063: \enddemo
4064: \bn
4065: Then there exists $B \subseteq M^*, \; |B| \le \aleph_0, \; p \in
4066: \bold S^1_D(B),p$ is omitted in $M^*$, and in fact by \scite{MOM.4C},
4067: $p^{[2 \varepsilon]}$ is omitted in $M^*$ for some $\varepsilon
4068: > 0$. Therefore (by \scite{4n.9.1}) $p$ has no pseudo
4069: $(<\varepsilon)$-support in $M^*$, see Definition \scite{4n.9}.
4070:
4071: Let $\mu > \lambda$ be large enough.
4072: \nl
4073:
4074: We choose $a_\alpha$ by induction on $\alpha < \mu$ such that
4075: \mr
4076: \item "{$\circledast$}" $A_\alpha = M^* \cup \{\bar a_\beta:\beta <
4077: \alpha\}$ really $(<\varepsilon)$-omits $p(\bar x)$.
4078: \endroster
4079: \bn
4080: \ub{Case (a)}: If $A_\alpha =: M^* \cup \{\bar a_\beta:\beta <
4081: \alpha\}$ is not in $K_1$. Then first choose $\varphi(\bar x,\bar y) \in
4082: \Delta,\bar a \subseteq A_\alpha$ such that $\varphi(\bar x,\bar a)$
4083: witnesses $A_\alpha \notin
4084: K_1$, and second choose $\bar a_\alpha$ realize some strictly isolated $q \in
4085: \bold S^{\ell g(\bar x)}_D(A_\alpha)$ which contains $\varphi(\bar
4086: x,\bar a)$.
4087:
4088: By \scite{8n.2} below, i.e., next claim, this is possible and
4089: $\circledast$ is preserved.
4090: \bn
4091: \ub{Case (b)}: Not (a), then $\bar a_\alpha \notin A_\alpha$ and
4092: $\circledast$ holds, using \scite{8n.3} below.
4093: \nl
4094: Having carried out the construction, let $M
4095: = \text{\rm mcl}(A_\mu)$. On the one hand, $M$ belongs to $K^c_1$
4096: as $A_\mu \in K_1$ (in fact, $A_\mu \prec {\frak C}$) by case (a) of
4097: the construction. On the other hand, note that $M$ is not
4098: $(D,\aleph_1)$-homogeneous, and moreover,
4099: cannot be $\varepsilon$-isomorphic to a $(D,\aleph_1)$-homogeneous
4100: model, as $p(x)$ is $(< \varepsilon)$-omitted by it ($p$ is really $(<
4101: \varepsilon)$-omitted by $\circledast$, and recall
4102: \scite{4n.9.1}(2)). So we can construct such a model of arbitrarily
4103: large density character, and by a L\"owenheim-Skolem argument,of
4104: \ub{any} density character large enough.
4105:
4106: But for each $\mu > \aleph_0$ there is $M \in K^c_1$,
4107: Ch$(M) = \mu$, $M$ is $(D,\aleph_1)$-homogeneous (by \scite{EM.1}(2)), a
4108: contradiction to wu-categoricity.
4109: \bn
4110: In order to complete the proof of the main theorem, we only need to
4111: show that the construction above is possible, which is done in the
4112: following two claims.
4113: \proclaim{\stag{8n.2} Claim} [$({\frak C},\bold d)$ is
4114: $0^+-\aleph_0$-stable, compact]
4115:
4116: Let $p \in \bold S^1_D(B),B$ \ub{countable}, $A \supseteq B,A$ is not a
4117: pseudo $(< \varepsilon)$-support for $p$, see Definition \scite{4n.9}.
4118: Let $\varphi(\bar x,\bar a)$ be a consistent formula over $A$. \ub{Then} there
4119: exists $\bar b \in {\frak C}$ such that ${\frak C} \models \varphi
4120: (\bar b,\bar a)$ and $A \cup \bar b$ is not a pseudo
4121: $(< \varepsilon)$-support for $p$. In fact, it is enough to choose
4122: $\bar b$ such that {\rm tp}$(\bar b,A)$ is strictly
4123: isolated and ${\frak C} \models \varphi(\bar b,\bar a)$.
4124: \endproclaim
4125: \bigskip
4126:
4127: \demo{Proof} By \scite{4n.12}(2) for some $\bar b \in \varphi({\frak
4128: C},\bar a)$, tp$(\bar b,A)$ is strictly isolated.
4129: So assume toward contradiction
4130: \mr
4131: \item "{$(*)_1$}" $A \cup \bar b$ is a pseudo $(<
4132: \varepsilon)$-support for $p(\bar x)$.
4133: \ermn
4134: Hence (by Definition \scite{4n.9}) there are
4135: $\zeta(1),\zeta(2),\vartheta_1(\bar x,\bar b,\bar c)$
4136: \mr
4137: \item "{$(*)_2$}" $\vartheta^{<\zeta(1)>}_1(\bar x,\bar b,\bar c_1)
4138: \models p^{[\varepsilon-\zeta(2)]}(\bar x)$ and $\bar c_1 \subseteq
4139: A$ and $\vartheta_1(\bar x,\bar b,\bar c_1)$ is consistent.
4140: \ermn
4141: As tp$(\bar b,A)$ is strictly isolated (see Definition \scite{4n.13}),
4142: there are $\zeta(3) > 0$ and
4143: $\psi(\bar y,\bar c_2)$ such that
4144: \mr
4145: \item "{$(*)_3$}" $(i) \quad \psi(\bar y,\bar c_2) \in \text{\rm
4146: tp}(\bar b,A)$ so $\bar c_2 \subseteq A$
4147: \sn
4148: \item "{${{}}$}" $(ii) \quad \psi(\bar y,\bar c_2)$ pseudo
4149: $(\zeta(1),\zeta(3))$-isolates tp$(\bar b,A)$,
4150: i.e. $\psi^{<\zeta(3)>}(\bar y,c_2] \models \text{ tp}(\bar b,A)^{[\zeta(1)]}$.
4151: \ermn
4152: Let
4153: \mr
4154: \item "{$(*)_4$}" $\vartheta_2(\bar x,\bar c_1,\bar c_2) = (\exists \bar
4155: y)[\psi(\bar y,\bar c_2) \wedge \vartheta_1(\bar x,\bar y,c_1)]$.
4156: \ermn
4157: Clearly
4158: \mr
4159: \item "{$(*)_5$}" $\vartheta_2(\bar x,\bar c_1,\bar c_2)$ is consistent.
4160: \ermn
4161: Choose $\zeta(4)$ such that
4162: \mr
4163: \item "{$(*)_6$}" $0 < \zeta(4) < \zeta(3)$ and $\zeta(4) < \zeta(1)$.
4164: \ermn
4165: We shall now show that $\vartheta_2(x,\bar c_1,\bar c_2)$
4166: pseudo $(\varepsilon -\zeta(2),\zeta(4))$-isolates $p$
4167: (and is over $A$), i.e.
4168: \mr
4169: \item "{$\boxtimes$}" $\vartheta^{<\zeta(4)>}_2(\bar x,\bar c_1,\bar
4170: c_2) \models p^{[\varepsilon-\zeta(2)]}(\bar x)$.
4171: \ermn
4172: This will give a contradiction to the assumption that $A$ is not a
4173: pseudo $(< \varepsilon)$-support for $p$.
4174: So assume
4175: \mr
4176: \item "{$(*)_7$}" ${\frak C} \models \vartheta^{<\zeta(4)>}_2[\bar
4177: a,\bar c_1,\bar c_2]$.
4178: \ermn
4179: By the definition of $\vartheta^{<\zeta(4)>}_2$, there are $\bar a',\bar
4180: c'_1,\bar c'_2$ such that
4181: \mr
4182: \item "{$(*)_8$}" $(i) \quad {\frak C} \models \vartheta_2[\bar a',
4183: \bar c'_1,\bar c'_2]$
4184: \sn
4185: \item "{${{}}$}" $(ii) \quad \bold d(\bar a',\bar a) \le \zeta(4) \le
4186: \zeta(1)$
4187: \sn
4188: \item "{${{}}$}" $(iii) \quad \bold d(\bar c'_1,\bar c_1) \le
4189: \zeta(4) \le \zeta(1)$
4190: \sn
4191: \item "{${{}}$}" $(iv) \quad \bold d(\bar c'_2,\bar c_2) \le
4192: \zeta(4) \le \zeta(3)$.
4193: \ermn
4194: By the choice of $\vartheta_2$, i.e. $(*)_4$, for some $\bar b'$
4195: \mr
4196: \item "{$(*)_9$}" $(i) \quad {\frak C} \models \psi[\bar b',\bar c'_2]$
4197: \sn
4198: \item "{${{}}$}" $(ii) \quad {\frak C} \models
4199: \vartheta_1[\bar a',\bar b',\bar c'_1]$.
4200: \ermn
4201: By the choice of $\psi(\bar y,\bar c_2)$, i.e., $(*)_3(ii)$ (note that
4202: as $0 < \zeta(4) \le \zeta(3)$, we have $\psi^{<\zeta(3)>}(\bar
4203: b',\bar c_2))$ there is $\bar b'' \in {\frak C}$ such that
4204: \mr
4205: \item "{$(*)_{10}$}" $(i) \quad \bar b''$ realizes tp$(\bar b,A)$
4206: \sn
4207: \item "{${{}}$}" $(ii) \quad \bold d(\bar b'',\bar b') \le \zeta(1)$.
4208: \ermn
4209: So $\bold d(\bar a,\bar a') \le \zeta(4) \le \zeta(1),\bold d(\bar
4210: b',b'') \le \zeta(1)$ and $\bold d(\bar c_1,\bar c'_1)
4211: \le \zeta(4) \le \zeta(1)$, therefore by $(*)_9(ii)$ we get
4212: \mr
4213: \item "{$(*)_{11}$}" ${\frak C} \models \vartheta^{<\zeta(1)>}_1
4214: [\bar a,\bar b'',\bar c_1]$.
4215: \ermn
4216: So by $(*)_2$, replacing $\bar b$ with $\bar b''$ which has the same
4217: type over $A$
4218: \mr
4219: \item "{$(*)_{12}$}" $\bar a$ realizes $p^{[\varepsilon-\zeta(2)]}$
4220: \ermn
4221: so we have finished proving $\boxtimes$, hence getting the desired
4222: contradiction. \hfill$\square_{\scite{8n.2}}$
4223: \enddemo
4224: \bigskip
4225:
4226: \proclaim{\stag{8n.3} Claim} [$({\frak C},\bold d)$ is
4227: $0^+-\aleph_0$-stable,uni-dimensional]
4228: \nl
4229: 1) Assume
4230: \mr
4231: \item "{$(a)$}" $M = \text{\rm mcl}(M) \subseteq {\frak C}$
4232: \sn
4233: \item "{$(b)$}" {\rm Ch}$(M) > \aleph_0$
4234: \sn
4235: \item "{$(c)$}" $B \subseteq M$ is countable
4236: \sn
4237: \item "{$(d)$}" $p \in \bold S^m_D(B)$
4238: \sn
4239: \item "{$(e)$}" $M$ really $(< \varepsilon)$-omits $p(\bar x)$, see
4240: Definition \scite{4n.9}.
4241: \ermn
4242: \ub{Then} for some $\bar b \in {\frak C} \backslash M$, also $M \cup
4243: \{\bar b\}$ pseudo $(< \varepsilon)$-omits $p$.
4244: \nl
4245: 2) Assume clauses (a)-(e) and $\bar b \in {\frak C},\bar b
4246: \notin M$, then $M \cup \bar b$ pseudo $(< \varepsilon)$-omits
4247: $p(x)$ when: for every $\bar c \in {}^{\omega >}M$ and $\zeta >0$
4248: there is $\bar b'$ realizing {\rm tp}$(\bar b,B \cup \bar c)$ such that
4249: $\bold d(\bar b',{}^{\ell g(\bar b)}M) < \zeta$.
4250: \nl
4251: 3) In clause (2) it is enough to assume that for every $\bar c \in
4252: {}^{\omega >}M$ and $\zeta > 0$ there exist $\bar b',\bar b^-$ such
4253: that $\bold d(\bar b,\bar b^-) < \zeta,\bold d(\bar b',M) < \zeta$
4254: and $\bar b'$ realizes tp$(\bar b^-,B \cup \bar c)$.
4255: \endproclaim
4256: \bigskip
4257:
4258: \remark{Remark} Assuming just $|\tau_{\frak C}| + \aleph_0 + |B| < \text{\rm
4259: Ch}(M)$ suffices. Assuming Ch$(M) > |B| + 2^{\aleph_0} +
4260: |\tau_{\frak C}|$, we can waive the assumption on $0^+-\aleph_0$-stability.
4261: \endremark
4262: \bigskip
4263:
4264: \demo{Proof} 1) Let $\Phi^*$ be as in \scite{EM.1}(3), so
4265: $\tau(\Phi^*)$ of cardinality $\aleph_1$. Recall
4266: \mr
4267: \item "{$(*)$}" for every uncountable (large enough if we do not
4268: assume $\tau$ countable) linear order $I,M^*_I = \text{\rm mcl}
4269: (\text{\rm EM}_\tau(I,\Phi^*))$ is $(D,|I|)$-homogeneous.
4270: \ermn
4271: [It is $(D,\aleph_1)$-homogeneous and use uni-dimensionality.]
4272: Choose $I=\lambda$ for $\lambda$ regular,
4273: $\lambda > \text{ Ch}(M)$, and let $M^* = M^*_\lambda$.
4274:
4275: So \wilog \, $M \subseteq M^*$. As Ch$(M) > \aleph_0$, we can by
4276: \scite{C.2.B} below
4277: find $\varepsilon > 0$ and $a_\alpha \in M$ for $\alpha < \omega_1$
4278: such that $\alpha < \beta < \omega_1 \Rightarrow \bold
4279: d(a_\alpha,a_\beta) > \varepsilon$. Let $\{a_{\omega_1+n}:n <
4280: \omega\} \subseteq M$ include $B$. Now for each $\alpha < \omega_1 +
4281: \omega$ and $n <\omega$ we can find $b_{\alpha,n} \in \text{\rm
4282: EM}_\tau(I,\Phi^*)$ such that $\bold d(a_\alpha,b_{\alpha,n}) <
4283: 1/(n+1)$. Let $b_{\alpha,n} =
4284: \sigma_{\alpha,n}(a_{t_{\alpha,n,0}},\dotsc,a_{t_{\alpha,n,k(\alpha,n)-1}})$.
4285:
4286: Let (for $\alpha < \omega_1$)
4287:
4288: $$
4289: S_\alpha = \{t_{\beta,n,\ell}:(\beta < \alpha) \vee \beta \in
4290: [\omega_1,\omega_1 +\omega) \text{ and }
4291: n < \omega,\ell < k(\beta,n)\} \subseteq \lambda.
4292: $$
4293: \mn
4294: Let (for $\alpha < \omega_1,n < \omega,\ell < k(\alpha,n))$
4295:
4296: $$
4297: \gamma_{\alpha,n,\ell} = \text{\rm Min}\{\gamma \in S_\alpha \cup
4298: \{\lambda\}:t_{\alpha,n,\ell} \le \gamma\}.
4299: $$
4300: \enddemo
4301: \bigskip
4302:
4303: \demo{\stag{8n.3.1} Subclaim} Under these assumptions, there exists
4304: $C$, a club of $\omega_1$, such that
4305: \mr
4306: \item "{$(*)$}" if $\delta \in C$ and $m < \omega$ then the following
4307: set is stationary
4308: $$
4309: \align
4310: W_{\delta,m} = \{\alpha \in C:&\text{for every } n \le
4311: m,\sigma_{\alpha,n} = \sigma_{\delta,n}, \text{ hence } k(\alpha,n) =
4312: k(\delta,n) \text{ and} \\
4313: &(\gamma_{\alpha,n,\ell} = \gamma_{\delta,n,\ell}) \wedge
4314: (\gamma_{\alpha,n,\ell} \in S_\alpha \equiv \gamma_{\delta,n,\ell}
4315: \in S_\delta) \text{ for } \ell < k(\delta,n)\}.
4316: \endalign
4317: $$
4318: \endroster
4319: \enddemo
4320: \bigskip
4321:
4322: \demo{Proof} By a standard coding argument, there exist
4323: functions $f_n:\omega_1 \rightarrow \omega_1$ (for $n < \omega$) such
4324: that for $\alpha < \omega_1,f_n(\alpha)$ ``encodes" the finite sequence
4325:
4326: $$
4327: \langle \sigma_{\alpha,m}:m \le n\rangle \frown
4328: \langle(\beta^*_{\alpha,m,\ell},m^*_{\alpha,m,\ell},\ell^*_{\alpha,m,\ell}):m
4329: \le n,\ell < k(\alpha,m)\rangle
4330: $$
4331: \mn
4332: where $\beta^*_{\alpha,m,\ell} < \omega_1 +
4333: \omega,m^*_{\alpha,m,\ell}$ and $\ell^*_{\alpha,m,\ell}$ are natural
4334: numbers, satisfying:
4335: \mr
4336: \item "{$\otimes$}"
4337: $(\beta^*_{\alpha,m,\ell},m^*_{\alpha,m,\ell},\ell^*_{\alpha,m,\ell})$
4338: is the minimal (lexicographically) triple $(\beta^*,m^*,\ell^*)$ such
4339: that $t_{\beta^*,m^*,\ell^*} = \gamma_{\alpha,m,\ell}$.
4340: \ermn
4341: In fact, there exists such coding $f_n:\omega_1 \rightarrow \omega_1$
4342: such that on a club $C'_n,f_n$ is regressive. Now by F\"odor's lemma,
4343: the following set contains a club (as its $\omega_1$ complement cannot
4344: contain a stationary set):
4345:
4346: $$
4347: C''_n = \{\delta \in C'_n:\{\alpha \in C'_n:f_n(\alpha) =
4348: f_n(\delta)\} \text{ is stationary}\}.
4349: $$
4350: \mn
4351: We call this club $C_n$ and let $C = \dbca_{n < \omega} C_n$, obviously
4352: $C$ is as required. \hfill$\square_{\scite{8n.3.1}}$
4353: \enddemo
4354: \bigskip
4355:
4356: \demo{\stag{8n.3A} Subclaim} Under these assumptions,
4357: there exist $\delta(*) \in C$
4358: such that $(\forall m < \omega)(\exists \delta \in
4359: \delta(*) \cap C)(\delta(*) \in W_{\delta,m})$.
4360: \enddemo
4361: \bigskip
4362:
4363: \demo{Proof} Note that $\delta(*) \in W_{\delta,m}
4364: \Leftrightarrow \delta \in W_{\delta(*),m} \Leftrightarrow
4365: W_{\delta,m} = W_{\delta(*),m}$, so all we are looking for is
4366: $\delta(*)$ satisfying $\delta(*) > \text{ min }W_{\delta(*),m}$ for
4367: all $m$; now if for all $\delta < \omega_1 \exists m < \omega$ such
4368: that $\delta = \text{ min } W_{\delta,m}$ we set an easy
4369: contradiction (e.g. by F\"odor's lemma, although it is an overkill here).
4370: \hfill$\square_{\scite{8n.3A}}$
4371:
4372: By a similar argument we can find $\delta(*) \in C$ such that there
4373: exists a sequence $\langle \delta_n:n < \omega\rangle,\delta_n \in
4374: C,\delta_n < \delta_{n+1} < \delta(*)$ for all $n$ and
4375: $W_{\delta(*),n} = W_{\delta_n,n}$ for all $n$, so in particular $m
4376: \le n \Rightarrow W_{\delta_m,m} = W_{\delta_n,m} = W_{\delta(*),m}$.
4377: \nl
4378: We obtain (by the choice of $\delta_n,C,W_{\delta_n,m}$, etc.):
4379: \mr
4380: \item "{$\circledast_1$}" $\langle b_{\delta_n,m}:n \ge m\rangle$ is
4381: an indiscernible sequence over $[B_m = B \cup\{b_{\alpha,i}:\alpha <
4382: \delta_m,i < \omega\}]$ for each $m$.
4383: \ermn
4384: In fact, we can say more:
4385: \mr
4386: \item "{$\circledast_2$}" for each $\bar c \in \text{ EM}(I,\Phi^*)$
4387: finite and for each $m < \omega$, there exists $n^* < \omega$ such that
4388: $\langle b_{\delta_n,m},n \ge n^*\rangle$ is indiscernible
4389: over $B_m \cup \bar c$.
4390: \ermn
4391: We would like now to continue the indiscernible sequences above in a
4392: proper extension of $M^+$.
4393:
4394: Let $J = I \times \Bbb Q$ and let us identify $I$ with $I \times
4395: \{0\}$, so we think of $I$ as a subset of $J$. Now
4396: for each $m < \omega,\ell < \kappa(
4397: \delta_m,m)$, look at the sequence $\langle t_{\delta_n,m,\ell}:n \ge
4398: m\rangle$. It is either constant or strictly increasing (recall
4399: $\circledast_1$), in the first case define $t^*_{m,\ell} = t_{\delta_m,m,\ell}
4400: \in I \subseteq J$, otherwise choose $t^*_{m,\ell} \in J$ such that
4401: $t_{\delta_n,m,\ell} < t^*_{m,\ell} < \text{\rm
4402: sup}\{t_{\delta_n,m,\ell}:n \ge m\}$.
4403:
4404: Note:
4405: \mr
4406: \item "{$(**)$}" for each $m < \omega$, the (quantifier free) type
4407: of the sequence $\langle t^*_{n,\ell}:n < m,\ell < k(\delta(*),n
4408: \rangle$ in the language $\{<\}$ (order) is the same as the type of
4409: $\langle t_{\delta(*),n,\ell}:n < m,\ell < k(\delta(*),n)\rangle$.
4410: \ermn
4411: Let $M^+ = \text{\rm mcl}(\text{\rm EM}_{\tau({\frak C})}(J,\Phi^*))
4412: \prec^1_\Delta {\frak C},M^+$ extends $M^*$. Let $b^*_n =
4413: \sigma_{\delta(*),n}(t^*_{n,0},\dotsc,t^*_{n,k(\delta(*),n)-1})$ so $b^*_n
4414: \in M^+,\langle b^*_n:n < \omega \rangle$ is a Cauchy sequence
4415: (as $\langle b_{\delta^*,n}:n < \omega \rangle$ is a Cauchy sequence,
4416: use the indiscernibility of $J$) with limit $b^* \in M^+$.
4417:
4418: Note (by $\circledast_2$ and the choice of $b^*_n$)
4419: \mr
4420: \item "{$\circledast_3$}" for each $\bar c \in \text{ EM}(I,\Phi^*)$
4421: finite and for each $m < \omega$ there exists $n^* < \omega$ such that
4422: $\langle b_{\delta_n,m}:n \ge n^*\rangle \frown b^*_m$ is
4423: indiscernible over $B \cup \bar c$. In fact, by \scite{MON.2s.1}, the
4424: same is true for each $\bar c \in M^*$ finite.
4425: \ermn
4426: We also observe
4427: \mr
4428: \item "{$\circledast_4$}" $b^* \in M^+ \backslash M^*$.
4429: \ermn
4430: Why? Recall that $\langle a_\alpha:\alpha <
4431: \omega_1\rangle$ form an $\varepsilon$-net, so for $m$ big enough,
4432: $b_{\delta_{n_1},m},b_{\delta_{n_2},m}$ can not be too close for $n_1
4433: \ne n_2$. Combining this with $\circledast_3$ we see that $b^*_m$ can
4434: not be $\frac{\varepsilon}{3}$ close to any $\bar c \in M^*$.
4435: \mr
4436: \item "{$\circledast_5$}" $b^*$ is as required in (3).
4437: \ermn
4438: Why? Given $\bar c \in M$ finite and $\zeta >0$ we choose $m$ and $n$
4439: big enough such that $b^*_m$ is close enough to $b^*,b_{\delta_n,m}$
4440: is close enough to $a_{\delta_n} \in M_1$, and tp$(b_{\delta_1,m},B
4441: \cup \bar c) = \text{ tp}(b^*_m,B \cup \bar c)$ [possible by $\circledast_3$].
4442: \nl
4443:
4444: 2) Assume this fails, so there are $\zeta > 0,\bar c \subseteq M$ and a
4445: formula $\vartheta(\bar x,\bar b,\bar c)$
4446: such that $\vartheta^{<\zeta>]}(x,\bar
4447: b,\bar c) \models p^{[\varepsilon-\zeta]}(\bar x)$.
4448:
4449: Choose $\zeta_1,\zeta_2 > 0$ such that $\zeta_1 < \zeta_2 < \zeta$ and
4450: $\zeta_1 < \zeta - \zeta_2$. By the assumption there are $\bar
4451: b',\bar b''$ such that
4452: \mr
4453: \item "{$(*)$}" $(a) \quad \bar b''$ realizes tp$(\bar b,B \cup \bar c)$
4454: \sn
4455: \item "{${{}}$}" $(b) \quad \bar b' \subseteq M$ of length $\ell
4456: g(\bar b)$
4457: \sn
4458: \item "{${{}}$}" $(c) \quad \bold d(\bar b',\bar b'') < \zeta_1 <
4459: \zeta - \zeta_2$.
4460: \ermn
4461: By clause (a) of $(*)$ as $\theta^{<\zeta>}(\bar x,\bar b,\bar c)
4462: \models p^{[\varepsilon-\zeta]}(\bar x)$ also $\theta^{<\zeta>}(\bar
4463: x,\bar b'',\bar c) \models p^{[\varepsilon-\zeta]}(\bar x)$. Hence
4464: (by $*(c)$ and \scite{4n.10}(3)) $\theta^{<\zeta_2>}(\bar x,\bar
4465: b',\bar c) \models p^{[\varepsilon-\zeta]}(\bar x)$, a contradiction.
4466: \nl
4467: 3) Similar proof to (2), first using \scite{4n.10}(3) to show that
4468: $\theta^{<\zeta'>}(\bar x,\bar b^-,\bar c) \models
4469: p^{[\varepsilon-\zeta]}(\bar x)$ for some $\zeta'$.
4470: \hfill$\square_{\scite{8n.3}}$
4471: \enddemo
4472: \bigskip
4473:
4474: \demo{\stag{C.2.B} Observation} Let
4475: $(X,\bold d)$ be a non-separable metric space. Then there exist
4476:
4477: $$
4478: \langle a_i:i < \omega_1\rangle \subseteq X \text{ and } \varepsilon^*
4479: > 0 \text{ such that } \bold d(a_i,a_j) \ge \varepsilon^* \, \forall i,j <
4480: \omega_1.
4481: $$
4482: \enddemo
4483: \bigskip
4484:
4485: \demo{Proof}
4486: Choose by induction on $i < \omega_1,a_i$ such that
4487: $a_i \notin \{a_j:j < i\}$. Choose
4488: $0 < \varepsilon_i \le d(a_i,\{a_j:j < i\}),\varepsilon_i \in \Bbb Q$.
4489: Without loss of generality $\varepsilon_i = \varepsilon^*$ for all
4490: $i$, and we are done. \hfill$\square_{\scite{CATEG}}$
4491: \enddemo
4492: \bigskip
4493:
4494: \nocite{ignore-this-bibtex-warning}
4495: %% you may want to move the following lines up a bit
4496: \newpage
4497:
4498: REFERENCES.
4499: \bibliographystyle{lit-plain}
4500: \bibliography{lista,listb,listx,listf,liste}
4501:
4502: \enddocument %%
4503:
4504:
4505:
4506:
4507:
4508: