1: % 600, processed by citealice (2006-07-10) on Thu Aug 9 11:03:38 IDT 2007
2:
3: \ifx\shlhetal\undefinedcontrolsequence\let\shlhetal\relax\fi
4: % To: "Saharon Shelah's Office" <shlhetal@math.huji.ac.il>, Saharon Shelah <shelah@math.huji.ac.il>, <shani@math.huji.ac.il>
5: % Subject: sh600
6: % Date: Wed, 8 Aug 2007 10:53:33 -0400 (EDT)
7: % From: Alice Leonhardt <leonhard@math.rutgers.edu>
8: % Mime-Version: 1.0
9: % Content-Description: uniformity
10: % X-sliced-and-diced-by: 'savemail' 2.0, Oct 2005
11:
12: % begin file 600A.tex
13: % Martin defined a few things
14: % sent on Thur, Jun 2005 16:14:41 -0400 (EDT)
15:
16: \input amstex
17: % % \input mathdefs
18: % *** start including mathdefs.tex ***
19: \expandafter\ifx\csname mathdefs.tex\endcsname\relax
20: \expandafter\gdef\csname mathdefs.tex\endcsname{}
21: \else \message{Hey! Apparently you were trying to
22: \string\input{mathdefs.tex} twice. This does not make sense.}
23: \errmessage{Please edit your file (probably \jobname.tex) and remove
24: any duplicate ``\string\input'' lines}\endinput\fi
25:
26:
27: %mathdefs.tex v1.3.2
28:
29: %%% Changes from v1.0: footnote macros, warning for duplicated tags,
30: %%% control sequences \( and \verbatimtags.
31: %%% From v1.2: \pretags, redefinition of \( using \ifinner, multi-part
32: %%% equation numbering, control sequences \[, \references, and
33: %%% \resetbracket.
34: %%% From v1.3: \rm in \lastpart; write root of multi-part tag to .tgs
35:
36: %See file texdefs.doc for documentation.
37:
38: \catcode`\X=12\catcode`\@=11
39:
40: %Minor control sequences:
41: \def\n@wcount{\alloc@0\count\countdef\insc@unt}
42: \def\n@wwrite{\alloc@7\write\chardef\sixt@@n}
43: \def\n@wread{\alloc@6\read\chardef\sixt@@n}
44: \def\r@s@t{\relax}\def\v@idline{\par}\def\@mputate#1/{#1}
45: \def\l@c@l#1X{\firstpart.#1}\def\gl@b@l#1X{#1}\def\t@d@l#1X{{}}
46:
47: %Creation of tag families and output of assignments and citations:
48: \def\crossrefs#1{\ifx\all#1\let\tr@ce=\all\else\def\tr@ce{#1,}\fi
49: \n@wwrite\cit@tionsout\openout\cit@tionsout=\jobname.cit
50: \write\cit@tionsout{\tr@ce}\expandafter\setfl@gs\tr@ce,}
51: \def\setfl@gs#1,{\def\@{#1}\ifx\@\empty\let\next=\relax
52: \else\let\next=\setfl@gs\expandafter\xdef
53: \csname#1tr@cetrue\endcsname{}\fi\next}
54: \def\m@ketag#1#2{\expandafter\n@wcount\csname#2tagno\endcsname
55: \csname#2tagno\endcsname=0\let\tail=\all\xdef\all{\tail#2,}
56: \ifx#1\l@c@l\let\tail=\r@s@t\xdef\r@s@t{\csname#2tagno\endcsname=0\tail}\fi
57: \expandafter\gdef\csname#2cite\endcsname##1{\expandafter
58: \ifx\csname#2tag##1\endcsname\relax?\else\csname#2tag##1\endcsname\fi
59: \expandafter\ifx\csname#2tr@cetrue\endcsname\relax\else
60: \write\cit@tionsout{#2tag ##1 cited on page \folio.}\fi}
61: \expandafter\gdef\csname#2page\endcsname##1{\expandafter
62: \ifx\csname#2page##1\endcsname\relax?\else\csname#2page##1\endcsname\fi
63: \expandafter\ifx\csname#2tr@cetrue\endcsname\relax\else
64: \write\cit@tionsout{#2tag ##1 cited on page \folio.}\fi}
65: \expandafter\gdef\csname#2tag\endcsname##1{\expandafter
66: \ifx\csname#2check##1\endcsname\relax
67: \expandafter\xdef\csname#2check##1\endcsname{}%
68: \else\immediate\write16{Warning: #2tag ##1 used more than once.}\fi
69: \multit@g{#1}{#2}##1/X%
70: \write\t@gsout{#2tag ##1 assigned number \csname#2tag##1\endcsname\space
71: on page \number\count0.}%
72: \csname#2tag##1\endcsname}}
73:
74:
75: \def\multit@g#1#2#3/#4X{\def\t@mp{#4}\ifx\t@mp\empty%
76: \global\advance\csname#2tagno\endcsname by 1
77: \expandafter\xdef\csname#2tag#3\endcsname
78: {#1\number\csname#2tagno\endcsnameX}%
79: \else\expandafter\ifx\csname#2last#3\endcsname\relax
80: \expandafter\n@wcount\csname#2last#3\endcsname
81: \global\advance\csname#2tagno\endcsname by 1
82: \expandafter\xdef\csname#2tag#3\endcsname
83: {#1\number\csname#2tagno\endcsnameX}
84: \write\t@gsout{#2tag #3 assigned number \csname#2tag#3\endcsname\space
85: on page \number\count0.}\fi
86: \global\advance\csname#2last#3\endcsname by 1
87: \def\t@mp{\expandafter\xdef\csname#2tag#3/}%
88: \expandafter\t@mp\@mputate#4\endcsname
89: {\csname#2tag#3\endcsname\lastpart{\csname#2last#3\endcsname}}\fi}
90: \def\t@gs#1{\def\all{}\m@ketag#1e\m@ketag#1s\m@ketag\t@d@l p
91: \let\realscite\scite
92: \let\realstag\stag
93: \m@ketag\gl@b@l r \n@wread\t@gsin
94: \openin\t@gsin=\jobname.tgs \re@der \closein\t@gsin
95: \n@wwrite\t@gsout\openout\t@gsout=\jobname.tgs }
96: \outer\def\localtags{\t@gs\l@c@l}
97: \outer\def\globaltags{\t@gs\gl@b@l}
98: \outer\def\newlocaltag#1{\m@ketag\l@c@l{#1}}
99: \outer\def\newglobaltag#1{\m@ketag\gl@b@l{#1}}
100:
101: %Reading in tag information:
102: \newif\ifpr@
103: \def\m@kecs #1tag #2 assigned number #3 on page #4.%
104: {\expandafter\gdef\csname#1tag#2\endcsname{#3}
105: \expandafter\gdef\csname#1page#2\endcsname{#4}
106: \ifpr@\expandafter\xdef\csname#1check#2\endcsname{}\fi}
107: \def\re@der{\ifeof\t@gsin\let\next=\relax\else
108: \read\t@gsin to\t@gline\ifx\t@gline\v@idline\else
109: \expandafter\m@kecs \t@gline\fi\let \next=\re@der\fi\next}
110: \def\pretags#1{\pr@true\pret@gs#1,,}
111: \def\pret@gs#1,{\def\@{#1}\ifx\@\empty\let\n@xtfile=\relax
112: \else\let\n@xtfile=\pret@gs \openin\t@gsin=#1.tgs \message{#1} \re@der
113: \closein\t@gsin\fi \n@xtfile}
114:
115: %Sections and subsections; local numbering:
116: \newcount\sectno\sectno=0\newcount\subsectno\subsectno=0
117: \newif\ifultr@local \def\ultralocal{\ultr@localtrue}
118: \def\firstpart{\number\sectno}
119: \def\lastpart#1{\ifcase#1 \or a\or b\or c\or d\or e\or f\or g\or h\or
120: 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
121: x\or y\or z \fi}
122: \def\closeup{\vskip-\bigskipamount}
123: \def\resetall{\global\advance\sectno by 1\subsectno=0
124: \gdef\firstpart{\number\sectno}\r@s@t}
125: \def\resetsub{\global\advance\subsectno by 1
126: \gdef\firstpart{\number\sectno.\number\subsectno}\r@s@t}
127: \def\newsection#1\par{\resetall\vskip0pt plus.3\vsize\penalty-250
128: \vskip0pt plus-.3\vsize\bigskip\bigskip
129: \message{#1}\leftline{\bf#1}\nobreak\bigskip}
130: \def\subsection#1\par{\ifultr@local\resetsub\fi
131: \vskip0pt plus.2\vsize\penalty-250\vskip0pt plus-.2\vsize
132: \bigskip\smallskip\message{#1}\leftline{\bf#1}\nobreak\medskip}
133:
134: %jj tags:
135: % On Andrzej's request: we want to be able
136: % to show tags as in noverbatim, with verbatim in the margin,
137: % and cites as in verbatim, with nonverbatim in the margin
138: % mg -- July 2000
139:
140: \newdimen\marginshift
141:
142: \newdimen\margindelta
143: \newdimen\marginmax
144: \newdimen\marginmin
145:
146: \def\margininit{
147: \marginmax=3 true cm % how much room, approximately
148:
149: \margindelta=0.1 true cm % distance between entries
150: \marginmin=0.1true cm % where will leftmost entry be
151: \marginshift=\marginmin
152: } % we cannot execute this right now, since
153: % there may be a \magnification coming later in the
154: % main file. So we call \margininit at the end of
155: % alice2jlem
156:
157: \def\t@gsjj#1,{\def\@{#1}\ifx\@\empty\let\next=\relax\else\let\next=\t@gsjj
158: \def\@@{p}\ifx\@\@@\else
159: \expandafter\gdef\csname#1cite\endcsname##1{\citejj{##1}}
160: \expandafter\gdef\csname#1page\endcsname##1{?}
161: \expandafter\gdef\csname#1tag\endcsname##1{\tagjj{##1}}\fi\fi\next}
162: \newif\ifshowstuffinmargin
163: \showstuffinmarginfalse
164: \def\jjtags{\ifx\shlhetal\relax
165: % so this is a public version --> no-op
166: \else
167: \ifx\shlhetal\undefinedcontrolseq
168: % again, this is a public version --> no-op
169: \else
170: \showstuffinmargintrue
171: \ifx\all\relax\else\expandafter\t@gsjj\all,\fi\fi \fi
172: }
173:
174: % \def\zeigentag#1{\def\xxy{#1}\show\xxy\show\realstag\realstag#1\mginpar{\hbox{$\langle$}#1\hbox{$\rangle$}}}
175: % \def\zeigencite#1{\hbox{$\langle$}#1\hbox{$\rangle$}\mginpar{\realscite#1}}
176:
177:
178: \def\tagjj#1{\realstag{#1}\oldmginpar{\zeigen{#1}}}
179: % \def\citejj#1{\zeigen{#1}\mginpar{\rechnen{#1}}}
180: \def\citejj#1{\rechnen{#1}\mginpar{\zeigen{#1}}} % modified Sep 02, saharon's suggestion
181:
182: % \def\rechnen#1{rechnen(\expandafter\ifx\csname stag#1\endcsname\relax ??\else
183: % \csname stag#1\endcsname\fi)endrech }
184: \def\rechnen#1{\expandafter\ifx\csname stag#1\endcsname\relax ??\else
185: \csname stag#1\endcsname\fi}
186:
187: % \def\zeigentag#1{//#1--\csname stag#1\endcsname::}
188:
189: \newdimen\theight
190:
191:
192:
193: \def\marginfont{\sevenrm}
194:
195:
196:
197: \def\trymarginbox#1{\setbox0=\hbox{\marginfont\hskip\marginshift #1}%
198: \global\marginshift\wd0
199: \global\advance\marginshift\margindelta}
200:
201:
202: \def \oldmginpar#1{%
203: \ifvmode\setbox0\hbox to \hsize{\hfill\rlap{\marginfont\quad#1}}%
204: \ht0 0cm
205: \dp0 0cm
206: \box0\vskip-\baselineskip
207: \else
208: \vadjust{\trymarginbox{#1}%
209: \ifdim\marginshift>\marginmax \global\marginshift\marginmin
210: \trymarginbox{#1}%
211: \fi
212: \theight=\ht0
213: \advance\theight by \dp0 \advance\theight by \lineskip
214: \kern -\theight \vbox to \theight{\rightline{\rlap{\box0}}%
215: \vss}}\fi}
216:
217: \newdimen\upordown
218: \global\upordown=8pt
219: \font\tinyfont=cmtt8 % scaled 700
220: \def\mginpar#1{\smash{\hbox to 0cm{\kern-10pt\raise7pt\hbox{\tinyfont #1}\hss}}}
221: % testing, october 2005, mg
222: \def\mginpar#1{{\hbox to 0cm{\kern-10pt\raise\upordown\hbox{\tinyfont #1}\hss}}\global\upordown-\upordown}
223:
224:
225: % \def\mginpar#1{mg-#1-mg }
226:
227:
228:
229: %Verbatim tags:
230: \def\t@gsoff#1,{\def\@{#1}\ifx\@\empty\let\next=\relax\else\let\next=\t@gsoff
231: \def\@@{p}\ifx\@\@@\else
232: \expandafter\gdef\csname#1cite\endcsname##1{\zeigen{##1}}
233: \expandafter\gdef\csname#1page\endcsname##1{?}
234: \expandafter\gdef\csname#1tag\endcsname##1{\zeigen{##1}}\fi\fi\next}
235: \def\verbatimtags{\showstuffinmarginfalse
236: \ifx\all\relax\else\expandafter\t@gsoff\all,\fi}
237: %%% \def\zeigen#1{zeigen(\hbox{$\langle$}#1\hbox{$\rangle$})endz }
238: \def\zeigen#1{\hbox{$\scriptstyle\langle$}#1\hbox{$\scriptstyle\rangle$}}
239:
240: % % \def\margincite#1{\ifshowstuffinmargin\mginpar{\rechnen{#1}}\fi}
241: % changed, april 2003, mg: we now have always the verbatim tag in the margin
242: \def\margincite#1{\ifshowstuffinmargin\mginpar{\zeigen{#1}}\fi}
243:
244: \def\margintag#1{\ifshowstuffinmargin\oldmginpar{\zeigen{#1}}\fi}
245:
246: \def\marginplain#1{\ifshowstuffinmargin\mginpar{{#1}}\fi}
247: \def\marginbf#1{\marginplain{{\bf \ \ #1}}}
248:
249:
250: %Equation numbering:
251: \def\(#1){\edef\dot@g{\ifmmode\ifinner(\hbox{\noexpand\etag{#1}})
252: \else\noexpand\eqno(\hbox{\noexpand\etag{#1}})\fi
253: \else(\noexpand\ecite{#1})\fi}\dot@g}
254:
255: %Reference numbering:
256: \newif\ifbr@ck
257: \def\eat#1{}
258: \def\[#1]{\br@cktrue[\br@cket#1'X]}
259: \def\br@cket#1'#2X{\def\temp{#2}\ifx\temp\empty\let\next\eat
260: \else\let\next\br@cket\fi
261: \ifbr@ck\br@ckfalse\br@ck@t#1,X\else\br@cktrue#1\fi\next#2X}
262: \def\br@ck@t#1,#2X{\def\temp{#2}\ifx\temp\empty\let\neext\eat
263: \else\let\neext\br@ck@t\def\temp{,}\fi
264: \def\teemp{#1}\ifx\teemp\empty\else\rcite{#1}\fi\temp\neext#2X}
265: \def\resetbr@cket{\gdef\[##1]{[\rtag{##1}]}}
266: \def\references{\resetbr@cket\newsection References\par}
267:
268: %Footnotes:
269: \newtoks\symb@ls\newtoks\s@mb@ls\newtoks\p@gelist\n@wcount\ftn@mber
270: \ftn@mber=1\newif\ifftn@mbers\ftn@mbersfalse\newif\ifbyp@ge\byp@gefalse
271: \def\defm@rk{\ifftn@mbers\n@mberm@rk\else\symb@lm@rk\fi}
272: \def\n@mberm@rk{\xdef\m@rk{{\the\ftn@mber}}%
273: \global\advance\ftn@mber by 1 }
274: \def\rot@te#1{\let\temp=#1\global#1=\expandafter\r@t@te\the\temp,X}
275: \def\r@t@te#1,#2X{{#2#1}\xdef\m@rk{{#1}}}
276: \def\b@@st#1{{$^{#1}$}}\def\str@p#1{#1}
277: \def\symb@lm@rk{\ifbyp@ge\rot@te\p@gelist\ifnum\expandafter\str@p\m@rk=1
278: \s@mb@ls=\symb@ls\fi\write\f@nsout{\number\count0}\fi \rot@te\s@mb@ls}
279: \def\byp@ge{\byp@getrue\n@wwrite\f@nsin\openin\f@nsin=\jobname.fns
280: \n@wcount\currentp@ge\currentp@ge=0\p@gelist={0}
281: \re@dfns\closein\f@nsin\rot@te\p@gelist
282: \n@wread\f@nsout\openout\f@nsout=\jobname.fns }
283: \def\m@kelist#1X#2{{#1,#2}}
284: \def\re@dfns{\ifeof\f@nsin\let\next=\relax\else\read\f@nsin to \f@nline
285: \ifx\f@nline\v@idline\else\let\t@mplist=\p@gelist
286: \ifnum\currentp@ge=\f@nline
287: \global\p@gelist=\expandafter\m@kelist\the\t@mplistX0
288: \else\currentp@ge=\f@nline
289: \global\p@gelist=\expandafter\m@kelist\the\t@mplistX1\fi\fi
290: \let\next=\re@dfns\fi\next}
291: \def\symbols#1{\symb@ls={#1}\s@mb@ls=\symb@ls}
292: \def\bigsymbol{\textstyle}
293: \symbols{\bigsymbol\ast,\dagger,\ddagger,\sharp,\flat,\natural,\star}
294: \def\ftnumbers{\ftn@mberstrue} \def\ftsymbols{\ftn@mbersfalse}
295: \def\paginal{\byp@ge} \def\resetftnumbers{\ftn@mber=1}
296: \def\ftnote#1{\defm@rk\expandafter\expandafter\expandafter\footnote
297: \expandafter\b@@st\m@rk{#1}}
298:
299: %Miscellaneous macros:
300: \long\def\jump#1\endjump{}
301: \def\ssum{\mathop{\lower .1em\hbox{$\textstyle\Sigma$}}\nolimits}
302: \def\down#1{_{{}_{\scriptstyle #1}}}
303: \def\qed{\nobreak\kern 1em \vrule height .5em width .5em depth 0em}
304: \def\newneq{\hbox{\rlap{\hbox to 1\wd9{\hss$=$\hss}}\raise .1em
305: \hbox to 1\wd9{\hss$\scriptscriptstyle/$\hss}}}
306: \def\subsetne{\setbox9 = \hbox{$\subset$}\mathrel{\hbox{\rlap
307: {\lower .4em \newneq}\raise .13em \hbox{$\subset$}}}}
308: \def\supsetne{\setbox9 = \hbox{$\subset$}\mathrel{\hbox{\rlap
309: {\lower .4em \newneq}\raise .13em \hbox{$\supset$}}}}
310:
311: %Blackboard bold:
312: \def\vbar{\mathchoice{\vrule height6.3ptdepth-.5ptwidth.8pt\kern-.8pt}
313: {\vrule height6.3ptdepth-.5ptwidth.8pt\kern-.8pt}
314: {\vrule height4.1ptdepth-.35ptwidth.6pt\kern-.6pt}
315: {\vrule height3.1ptdepth-.25ptwidth.5pt\kern-.5pt}}
316: \def\f@dge{\mathchoice{}{}{\mkern.5mu}{\mkern.8mu}}
317: \def\b@c#1#2{{\rm \mkern#2mu\vbar\mkern-#2mu#1}}
318: \def\b@b#1{{\rm I\mkern-3.5mu #1}}
319: \def\b@a#1#2{{\rm #1\mkern-#2mu\f@dge #1}}
320: \def\bb#1{{\count4=`#1 \advance\count4by-64 \ifcase\count4\or\b@a A{11.5}\or
321: \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
322: \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
323: \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
324: \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}}
325:
326: \catcode`\X=11 \catcode`\@=12
327:
328:
329: % Sep 2003, mg:
330:
331: % % \input 300stuff
332: % *** start including 300stuff.tex ***
333:
334: % definitions needed to process the 300x papers (plus 88 etc)
335:
336: \let\thischap\jobname
337:
338:
339:
340: \def\partof#1{\csname returnthe#1part\endcsname}
341: \def\chapof#1{\csname returnthe#1chap\endcsname}
342:
343: \def\setchapter#1,#2,#3;{%
344: \expandafter\def\csname returnthe#1part\endcsname{#2}%
345: \expandafter\def\csname returnthe#1chap\endcsname{#3}%
346: }
347:
348: \setchapter 300a,A,II.A;
349: \setchapter 300b,A,II.B;
350: \setchapter 300c,A,II.C;
351: \setchapter 300d,A,II.D;
352: \setchapter 300e,A,II.E;
353: \setchapter 300f,A,II.F;
354: \setchapter 300g,A,II.G;
355: \setchapter E53,B,N;
356: \setchapter 88r,B,I;
357: \setchapter 600,B,III;
358: \setchapter 705,B,IV;
359: \setchapter 734,B,V;
360:
361:
362: % \def\cprefix#1{% \cprefix{300b} generates A.II. or II. or "nothing"
363: % %\leavevmode\vrule width 2cm height 1cm depth 0cm
364: % \edef\theotherpart{\partof{#1}}\edef\theotherchap{\chapof{#1}}%
365: % \ifx\theotherpart\thispart
366: % \ifx\theotherchap\thischap % nothing
367: % \else % same part, different chap
368: % \theotherchap%
369: % \fi
370: % \else % different part
371: % \theotherpart.\theotherchap\fi}
372:
373:
374: \def\cprefix#1{% \cprefix{300b} generates II. or "nothing"
375: %\leavevmode\vrule width 2cm height 1cm depth 0cm
376: \edef\theotherpart{\partof{#1}}\edef\theotherchap{\chapof{#1}}%
377: \ifx\theotherpart\thispart
378: \ifx\theotherchap\thischap % nothing
379: \else % same part, different chap
380: \theotherchap%
381: \fi
382: \else % different part
383: \theotherchap\fi}
384:
385:
386: % \sectioncite[\S4]{300b} -> II.\S4 or A.II.\S4
387: \def\sectioncite[#1]#2{%
388: \cprefix{#2}#1}
389:
390: \def\chaptercite#1{Chapter \cprefix{#1}}
391:
392: % now define \thispart
393: \edef\thispart{\partof{\thischap}}
394: \edef\thischap{\chapof{\thischap}}
395:
396: \def\lastpage of '#1' is #2.{\expandafter\def\csname lastpage#1\endcsname{#2}}
397:
398: % *** end including 300stuff.tex ***
399:
400: \def\spuriousreset{}
401:
402: % *** end including mathdefs.tex ***
403: % % \input citeadd
404: % *** start including citeadd.tex ***
405: % citeadd -- a few additions for
406: % files from alice that were procesed with "citealice"
407:
408: \expandafter\ifx\csname citeadd.tex\endcsname\relax
409: \expandafter\gdef\csname citeadd.tex\endcsname{}
410: \else \message{Hey! Apparently you were trying to
411: \string\input{citeadd.tex} twice. This does not make sense.}
412: \errmessage{Please edit your file (probably \jobname.tex) and remove
413: any duplicate ``\string\input'' lines}\endinput\fi
414:
415:
416:
417:
418:
419: \def\sciteu{\sciteerror{undefined}}
420: \def\sciteuphantom{\complainaboutcitation{undefined}}
421:
422: \def\scitet{\sciteerror{ambiguous}}
423: \def\scitetphantom{\complainaboutcitation{ambiguous}}
424:
425:
426:
427:
428: \def\sciteerror#1#2{{\mathortextbf{\scite{#2}}}\complainaboutcitation{#1}{#2}}
429: \def\mathortextbf#1{\hbox{\bf #1}}
430: \def\complainaboutcitation#1#2{%
431: \vadjust{\line{\llap{---$\!\!>$ }\qquad scite$\{$#2$\}$ #1\hfil}}}
432:
433:
434:
435:
436:
437: % *** end including citeadd.tex ***
438: \sectno=-1 % start with sect 0
439: % % \input epsf
440: % *** start including epsf.tex ***
441: %%% ====================================================================
442: %%% This file is freely redistributable and placed into the
443: %%% public domain by Tomas Rokicki.
444: %%% @TeX-file{
445: %%% author = "Tom Rokicki",
446: %%% version = "2.7k",
447: %%% date = "19 July 1997",
448: %%% time = "10:00:05 MDT",
449: %%% filename = "epsf.tex",
450: %%% address = "Tom Rokicki
451: %%% Box 2081
452: %%% Stanford, CA 94309
453: %%% USA",
454: %%% telephone = "+1 415 855 9989",
455: %%% email = "rokicki@cs.stanford.edu (Internet)",
456: %%% codetable = "ISO/ASCII",
457: %%% keywords = "PostScript, TeX",
458: %%% supported = "yes",
459: %%% abstract = "This file contains macros to support the inclusion
460: %%% of Encapsulated PostScript files in TeX documents.",
461: %%% docstring = "This file contains TeX macros to include an
462: %%% Encapsulated PostScript graphic. It works
463: %%% by finding the bounding box comment,
464: %%% calculating the correct scale values, and
465: %%% inserting a vbox of the appropriate size at
466: %%% the current position in the TeX document.
467: %%%
468: %%% To use, simply say
469: %%%
470: %%% \input epsf % somewhere early on in your TeX file
471: %%%
472: %%% % then where you want to insert a vbox for a figure:
473: %%% \epsfbox{filename.ps}
474: %%%
475: %%% Alternatively, you can supply your own
476: %%% bounding box by
477: %%%
478: %%% \epsfbox[0 0 30 50]{filename.ps}
479: %%%
480: %%% This will not read in the file, and will
481: %%% instead use the bounding box you specify.
482: %%%
483: %%% The effect will be to typeset the figure as
484: %%% a TeX box, at the point of your \epsfbox
485: %%% command. By default, the graphic will have
486: %%% its `natural' width (namely the width of
487: %%% its bounding box, as described in
488: %%% filename.ps). The TeX box will have depth
489: %%% zero.
490: %%%
491: %%% You can enlarge or reduce the figure by
492: %%% saying
493: %%%
494: %%% \epsfxsize=<dimen> \epsfbox{filename.ps}
495: %%% or
496: %%% \epsfysize=<dimen> \epsfbox{filename.ps}
497: %%%
498: %%% instead. Then the width of the TeX box will
499: %%% be \epsfxsize and its height will be scaled
500: %%% proportionately (or the height will be
501: %%% \epsfysize and its width will be scaled
502: %%% proportionately).
503: %%%
504: %%% The width (and height) is restored to zero
505: %%% after each use, so \epsfxsize or \epsfysize
506: %%% must be specified before EACH use of
507: %%% \epsfbox.
508: %%%
509: %%% A more general facility for sizing is
510: %%% available by defining the \epsfsize macro.
511: %%% Normally you can redefine this macro to do
512: %%% almost anything. The first parameter is
513: %%% the natural x size of the PostScript
514: %%% graphic, the second parameter is the
515: %%% natural y size of the PostScript graphic.
516: %%% It must return the xsize to use, or 0 if
517: %%% natural scaling is to be used. Common uses
518: %%% include:
519: %%%
520: %%% \epsfxsize % just leave the old value alone
521: %%% 0pt % use the natural sizes
522: %%% #1 % use the natural sizes
523: %%% \hsize % scale to full width
524: %%% 0.5#1 % scale to 50% of natural size
525: %%% \ifnum #1>\hsize\hsize\else#1\fi
526: %%% % smaller of natural, hsize
527: %%%
528: %%% If you want TeX to report the size of the
529: %%% figure (as a message on your terminal when
530: %%% it processes each figure), say
531: %%% `\epsfverbosetrue'.
532: %%%
533: %%% If you only want to get the bounding box
534: %%% extents, without producing any output boxes
535: %%% or \special{}, then say
536: %%% \epsfgetbb{filename}. The extents will be
537: %%% saved in the macros \epsfllx \epsflly
538: %%% \epsfurx \epsfury in PostScript units of
539: %%% big points.
540: %%%
541: %%% Revision history:
542: %%%
543: %%% ---------------------------------------------
544: %%% epsf.tex macro file:
545: %%% Originally written by Tomas Rokicki of
546: %%% Radical Eye Software, 29 Mar 1989.
547: %%%
548: %%% ---------------------------------------------
549: %%% Revised by Don Knuth, 3 Jan 1990.
550: %%%
551: %%% ---------------------------------------------
552: %%% Revised by Tomas Rokicki, 18 Jul 1990.
553: %%% Accept bounding boxes with no space after
554: %%% the colon.
555: %%%
556: %%% ---------------------------------------------
557: %%% Revised by Nelson H. F. Beebe
558: %%% <beebe@math.utah.edu>, 03 Dec 1991 [2.0].
559: %%% Add version number and date typeout.
560: %%%
561: %%% Use \immediate\write16 instead of \message
562: %%% to ensure output on new line.
563: %%%
564: %%% Handle nested EPS files.
565: %%%
566: %%% Handle %%BoundingBox: (atend) lines.
567: %%%
568: %%% Do not quit when blank lines are found.
569: %%%
570: %%% Add a few percents to remove generation of
571: %%% spurious blank space.
572: %%%
573: %%% Move \special output to
574: %%% \epsfspecial{filename} so that other macro
575: %%% packages can input this one, then change
576: %%% the definition of \epsfspecial to match
577: %%% another DVI driver.
578: %%%
579: %%% Move size computation to \epsfsetsize which
580: %%% can be called by the user; the verbose
581: %%% output of the bounding box and scaled width
582: %%% and height happens here.
583: %%%
584: %%% ---------------------------------------------
585: %%% Revised by Nelson H. F. Beebe
586: %%% <beebe@math.utah.edu>, 05 May 1992 [2.1].
587: %%% Wrap \leavevmode\hbox{} around \vbox{} with
588: %%% the \special so that \epsffile{} can be
589: %%% used inside \begin{center}...\end{center}
590: %%%
591: %%% ---------------------------------------------
592: %%% Revised by Nelson H. F. Beebe
593: %%% <beebe@math.utah.edu>, 09 Dec 1992 [2.2].
594: %%% Introduce \epsfshow{true,false} and
595: %%% \epsfframe{true,false} macros; the latter
596: %%% suppresses the insertion of the PostScript,
597: %%% and instead just creates an empty box,
598: %%% which may be handy for rapid prototyping.
599: %%%
600: %%% ---------------------------------------------
601: %%% Revised by Nelson H. F. Beebe
602: %%% <beebe@math.utah.edu>, 14 Dec 1992 [2.3].
603: %%% Add \epsfshowfilename{true,false}. When
604: %%% true, and \epsfshowfalse is specified, the
605: %%% PostScript file name will be displayed
606: %%% centered in the figure box.
607: %%%
608: %%% ---------------------------------------------
609: %%% Revised by Nelson H. F. Beebe
610: %%% <beebe@math.utah.edu>, 20 June 1993 [2.4].
611: %%% Remove non-zero debug setting of \epsfframemargin,
612: %%% and change margin handling to preserve EPS image
613: %%% size and aspect ratio, so that the actual
614: %%% box is \epsfxsize+\epsfframemargin wide by
615: %%% \epsfysize+\epsfframemargin high.
616: %%% Reduce output of \epsfshowfilenametrue to
617: %%% just the bare file name.
618: %%%
619: %%% ---------------------------------------------
620: %%% Revised by Nelson H. F. Beebe
621: %%% <beebe@math.utah.edu>, 13 July 1993 [2.5].
622: %%% Add \epsfframethickness for control of
623: %%% \epsfframe frame lines.
624: %%%
625: %%% ---------------------------------------------
626: %%% Revised by Nelson H. F. Beebe
627: %%% <beebe@math.utah.edu>, 02 July 1996 [2.6]
628: %%% Add missing initialization \epsfatendfalse;
629: %%% the lack of this resulted in the wrong
630: %%% BoundingBox being picked up, mea culpa, sigh...
631: %%% ---------------------------------------------
632: %%%
633: %%% ---------------------------------------------
634: %%% Revised by Nelson H. F. Beebe
635: %%% <beebe@math.utah.edu>, 25 October 1996 [2.7]
636: %%% Update to match changes in from dvips 5-600
637: %%% distribution: new user-accessible macros:
638: %%% \epsfclipon, \epsfclipoff, \epsfdrafton,
639: %%% \epsfdraftoff, change \empty to \epsfempty.
640: %%% ---------------------------------------------
641: %%%
642: %%% Modified to avoid verbosity, give help.
643: %%% --kb@cs.umb.edu, for Texinfo.
644: %%% }
645: %%% ====================================================================
646: %
647: \ifx\epsfannounce\undefined \def\epsfannounce{\immediate\write16}\fi
648: \epsfannounce{This is `epsf.tex' v2.7k <10 July 1997>}%
649: %
650: \newread\epsffilein % file to \read
651: \newif\ifepsfatend % need to scan to LAST %%BoundingBox comment?
652: \newif\ifepsfbbfound % success?
653: \newif\ifepsfdraft % use draft mode?
654: \newif\ifepsffileok % continue looking for the bounding box?
655: \newif\ifepsfframe % frame the bounding box?
656: \newif\ifepsfshow % show PostScript file, or just bounding box?
657: \epsfshowtrue % default is to display PostScript file
658: \newif\ifepsfshowfilename % show the file name if \epsfshowfalse specified?
659: \newif\ifepsfverbose % report what you're making?
660: \newdimen\epsfframemargin % margin between box and frame
661: \newdimen\epsfframethickness % thickness of frame rules
662: \newdimen\epsfrsize % vertical size before scaling
663: \newdimen\epsftmp % register for arithmetic manipulation
664: \newdimen\epsftsize % horizontal size before scaling
665: \newdimen\epsfxsize % horizontal size after scaling
666: \newdimen\epsfysize % vertical size after scaling
667: \newdimen\pspoints % conversion factor
668: %
669: \pspoints = 1bp % Adobe points are `big'
670: \epsfxsize = 0pt % default value, means `use natural size'
671: \epsfysize = 0pt % ditto
672: \epsfframemargin = 0pt % default value: frame box flush around picture
673: \epsfframethickness = 0.4pt % TeX's default rule thickness
674: %
675: \def\epsfbox#1{\global\def\epsfllx{72}\global\def\epsflly{72}%
676: \global\def\epsfurx{540}\global\def\epsfury{720}%
677: \def\lbracket{[}\def\testit{#1}\ifx\testit\lbracket
678: \let\next=\epsfgetlitbb\else\let\next=\epsfnormal\fi\next{#1}}%
679: %
680: % We use \epsfgetlitbb if the user specified an explicit bounding box,
681: % and \epsfnormal otherwise. Because \epsfgetbb can be called
682: % separately to retrieve the bounding box, we move the verbose
683: % printing the bounding box extents and size on the terminal to
684: % \epsfstatus. Therefore, when the user provided the bounding box,
685: % \epsfgetbb will not be called, so we must call \epsfsetsize and
686: % \epsfstatus ourselves.
687: %
688: \def\epsfgetlitbb#1#2 #3 #4 #5]#6{%
689: \epsfgrab #2 #3 #4 #5 .\\%
690: \epsfsetsize
691: \epsfstatus{#6}%
692: \epsfsetgraph{#6}%
693: }%
694: %
695: \def\epsfnormal#1{%
696: \epsfgetbb{#1}%
697: \epsfsetgraph{#1}%
698: }%
699: %
700: \newhelp\epsfnoopenhelp{The PostScript image file must be findable by
701: TeX, i.e., somewhere in the TEXINPUTS (or equivalent) path.}%
702: %
703: \def\epsfgetbb#1{%
704: %
705: % The first thing we need to do is to open the
706: % PostScript file, if possible.
707: %
708: \openin\epsffilein=#1
709: \ifeof\epsffilein
710: \errhelp = \epsfnoopenhelp
711: \errmessage{Could not open file #1, ignoring it}%
712: \else %process the file
713: {% %start a group to contain catcode changes
714: % Make all special characters, except space, to be of type
715: % `other' so we process the file in almost verbatim mode
716: % (TeXbook, p. 344).
717: \chardef\other=12
718: \def\do##1{\catcode`##1=\other}%
719: \dospecials
720: \catcode`\ =10
721: \epsffileoktrue %true while we are looping
722: \epsfatendfalse %[02-Jul-1996]: add forgotten initialization
723: \loop %reading lines from the EPS file
724: \read\epsffilein to \epsffileline
725: \ifeof\epsffilein %then no more input
726: \epsffileokfalse %so set completion flag
727: \else %otherwise process one line
728: \expandafter\epsfaux\epsffileline:. \\%
729: \fi
730: \ifepsffileok
731: \repeat
732: \ifepsfbbfound
733: \else
734: \ifepsfverbose
735: \immediate\write16{No BoundingBox comment found in %
736: file #1; using defaults}%
737: \fi
738: \fi
739: }% %end catcode changes
740: \closein\epsffilein
741: \fi %end of file processing
742: \epsfsetsize %compute size parameters
743: \epsfstatus{#1}%
744: }%
745: %
746: % Clipping control:
747: \def\epsfclipon{\def\epsfclipstring{ clip}}%
748: \def\epsfclipoff{\def\epsfclipstring{\ifepsfdraft\space clip\fi}}%
749: \epsfclipoff % default for dvips is OFF
750: %
751: % The special that is emitted by \epsfsetgraph comes from this macro.
752: % It is defined separately to allow easy customization by other
753: % packages that first \input epsf.tex, then redefine \epsfspecial.
754: % This macro is invoked in the lower-left corner of a box of the
755: % width and height determined from the arguments to \epsffile, or
756: % from the %%BoundingBox in the EPS file itself.
757: %
758: % This version is for dvips:
759: \def\epsfspecial#1{%
760: \epsftmp=10\epsfxsize
761: \divide\epsftmp\pspoints
762: \ifnum\epsfrsize=0\relax
763: \special{PSfile=\ifepsfdraft psdraft.ps\else#1\fi\space
764: llx=\epsfllx\space
765: lly=\epsflly\space
766: urx=\epsfurx\space
767: ury=\epsfury\space
768: rwi=\number\epsftmp
769: \epsfclipstring
770: }%
771: \else
772: \epsfrsize=10\epsfysize
773: \divide\epsfrsize\pspoints
774: \special{PSfile=\ifepsfdraft psdraft.ps\else#1\fi\space
775: llx=\epsfllx\space
776: lly=\epsflly\space
777: urx=\epsfurx\space
778: ury=\epsfury\space
779: rwi=\number\epsftmp\space
780: rhi=\number\epsfrsize
781: \epsfclipstring
782: }%
783: \fi
784: }%
785: %
786: % \epsfframe macro adapted from the TeXbook, exercise 21.3, p. 223, 331.
787: % but modified to set the box width to the natural width, rather
788: % than the line width, and to include space for margins and rules
789: \def\epsfframe#1%
790: {%
791: \leavevmode % so we can put this inside
792: % a centered environment
793: \setbox0 = \hbox{#1}%
794: \dimen0 = \wd0 % natural width of argument
795: \advance \dimen0 by 2\epsfframemargin % plus width of 2 margins
796: \advance \dimen0 by 2\epsfframethickness % plus width of 2 rule lines
797: \vbox
798: {%
799: \hrule height \epsfframethickness depth 0pt
800: \hbox to \dimen0
801: {%
802: \hss
803: \vrule width \epsfframethickness
804: \kern \epsfframemargin
805: \vbox {\kern \epsfframemargin \box0 \kern \epsfframemargin }%
806: \kern \epsfframemargin
807: \vrule width \epsfframethickness
808: \hss
809: }% end hbox
810: \hrule height 0pt depth \epsfframethickness
811: }% end vbox
812: }%
813: %
814: \def\epsfsetgraph#1%
815: {%
816: %
817: % Make the vbox and stick in a \special that the DVI driver can
818: % parse. \vfil and \hfil are used to place the \special origin at
819: % the lower-left corner of the vbox. \epsfspecial can be redefined
820: % to produce alternate \special syntaxes.
821: %
822: \leavevmode
823: \hbox{% so we can put this in \begin{center}...\end{center}
824: \ifepsfframe\expandafter\epsfframe\fi
825: {\vbox to\epsfysize
826: {%
827: \ifepsfshow
828: % output \special{} at lower-left corner of figure box
829: \vfil
830: \hbox to \epsfxsize{\epsfspecial{#1}\hfil}%
831: \else
832: \vfil
833: \hbox to\epsfxsize{%
834: \hss
835: \ifepsfshowfilename
836: {%
837: \epsfframemargin=3pt % local change of margin
838: \epsfframe{{\tt #1}}%
839: }%
840: \fi
841: \hss
842: }%
843: \vfil
844: \fi
845: }%
846: }}%
847: %
848: % Reset \epsfxsize and \epsfysize, as documented above.
849: %
850: \global\epsfxsize=0pt
851: \global\epsfysize=0pt
852: }%
853: %
854: % Now we have to calculate the scale and offset values to use.
855: % First we compute the natural sizes.
856: %
857: \def\epsfsetsize
858: {%
859: \epsfrsize=\epsfury\pspoints
860: \advance\epsfrsize by-\epsflly\pspoints
861: \epsftsize=\epsfurx\pspoints
862: \advance\epsftsize by-\epsfllx\pspoints
863: %
864: % If `epsfxsize' is 0, we default to the natural size of the picture.
865: % Otherwise we scale the graph to be \epsfxsize wide.
866: %
867: \epsfxsize=\epsfsize{\epsftsize}{\epsfrsize}%
868: \ifnum \epsfxsize=0
869: \ifnum \epsfysize=0
870: \epsfxsize=\epsftsize
871: \epsfysize=\epsfrsize
872: \epsfrsize=0pt
873: %
874: % We have a sticky problem here: TeX doesn't do floating point arithmetic!
875: % Our goal is to compute y = rx/t. The following loop does this reasonably
876: % fast, with an error of at most about 16 sp (about 1/4000 pt).
877: %
878: \else
879: \epsftmp=\epsftsize \divide\epsftmp\epsfrsize
880: \epsfxsize=\epsfysize \multiply\epsfxsize\epsftmp
881: \multiply\epsftmp\epsfrsize \advance\epsftsize-\epsftmp
882: \epsftmp=\epsfysize
883: \loop \advance\epsftsize\epsftsize \divide\epsftmp 2
884: \ifnum \epsftmp>0
885: \ifnum \epsftsize<\epsfrsize
886: \else
887: \advance\epsftsize-\epsfrsize \advance\epsfxsize\epsftmp
888: \fi
889: \repeat
890: \epsfrsize=0pt
891: \fi
892: \else
893: \ifnum \epsfysize=0
894: \epsftmp=\epsfrsize \divide\epsftmp\epsftsize
895: \epsfysize=\epsfxsize \multiply\epsfysize\epsftmp
896: \multiply\epsftmp\epsftsize \advance\epsfrsize-\epsftmp
897: \epsftmp=\epsfxsize
898: \loop \advance\epsfrsize\epsfrsize \divide\epsftmp 2
899: \ifnum \epsftmp>0
900: \ifnum \epsfrsize<\epsftsize
901: \else
902: \advance\epsfrsize-\epsftsize \advance\epsfysize\epsftmp
903: \fi
904: \repeat
905: \epsfrsize=0pt
906: \else
907: \epsfrsize=\epsfysize
908: \fi
909: \fi
910: }%
911: %
912: % Issue some status messages if the user requested them
913: %
914: \def\epsfstatus#1{% arg = filename
915: \ifepsfverbose
916: \immediate\write16{#1: BoundingBox:
917: llx = \epsfllx\space lly = \epsflly\space
918: urx = \epsfurx\space ury = \epsfury\space}%
919: \immediate\write16{#1: scaled width = \the\epsfxsize\space
920: scaled height = \the\epsfysize}%
921: \fi
922: }%
923: %
924: % We still need to define the tricky \epsfaux macro. This requires
925: % a couple of magic constants for comparison purposes.
926: %
927: {\catcode`\%=12 \global\let\epsfpercent=%\global\def\epsfbblit{%BoundingBox}}%
928: \global\def\epsfatend{(atend)}%
929: %
930: % So we're ready to check for `%BoundingBox:' and to grab the
931: % values if they are found.
932: %
933: % If we find a line
934: %
935: % %%BoundingBox: (atend)
936: %
937: % then we ignore it, but set a flag to force parsing all of the
938: % file, so the last %%BoundingBox parsed will be the one used. This
939: % is necessary, because EPS files can themselves contain other EPS
940: % files with their own %%BoundingBox comments.
941: %
942: % If we find a line
943: %
944: % %%BoundingBox: llx lly urx ury
945: %
946: % then we save the 4 values in \epsfllx, \epsflly, \epsfurx, \epsfury.
947: % Then, if we have not previously parsed an (atend), we flag completion
948: % and can stop reading the file. Otherwise, we must keep on reading
949: % to end of file so that we find the values on the LAST %%BoundingBox.
950: \long\def\epsfaux#1#2:#3\\%
951: {%
952: \def\testit{#2}% % save second character up to just before colon
953: \ifx#1\epsfpercent % then first char is percent (quick test)
954: \ifx\testit\epsfbblit % then (slow test) we have %%BoundingBox
955: \epsfgrab #3 . . . \\%
956: \ifx\epsfllx\epsfatend % then ignore %%BoundingBox: (atend)
957: \global\epsfatendtrue
958: \else % else found %%BoundingBox: llx lly urx ury
959: \ifepsfatend % then keep parsing ALL %%BoundingBox lines
960: \else % else stop after first one parsed
961: \epsffileokfalse
962: \fi
963: \global\epsfbbfoundtrue
964: \fi
965: \fi
966: \fi
967: }%
968: %
969: % Here we grab the values and stuff them in the appropriate definitions.
970: %
971: \def\epsfempty{}%
972: \def\epsfgrab #1 #2 #3 #4 #5\\{%
973: \global\def\epsfllx{#1}\ifx\epsfllx\epsfempty
974: \epsfgrab #2 #3 #4 #5 .\\\else
975: \global\def\epsflly{#2}%
976: \global\def\epsfurx{#3}\global\def\epsfury{#4}\fi
977: }%
978: %
979: % We default the epsfsize macro.
980: %
981: \def\epsfsize#1#2{\epsfxsize}%
982: %
983: % Finally, another definition for compatibility with older macros.
984: %
985: \let\epsffile=\epsfbox
986:
987: % *** end including epsf.tex ***
988: \localtags
989: \jjtags
990: \newbox\noforkbox \newdimen\forklinewidth
991: \forklinewidth=0.3pt %% maybe 0.6? 0.7 ??
992: \setbox0\hbox{$\textstyle\bigcup$}
993: \setbox1\hbox to \wd0{\hfil\vrule width \forklinewidth depth \dp0
994: height \ht0 \hfil}
995: \wd1=0 cm
996: \setbox\noforkbox\hbox{\box1\box0\relax}
997: \def\unionstick{\mathop{\copy\noforkbox}\limits}
998: \def\nonfork#1#2_#3{#1\unionstick_{\textstyle #3}#2}
999: \def\nonforkin#1#2_#3^#4{#1\unionstick_{\textstyle #3}^{\textstyle #4}#2}
1000: %
1001: %%% and the ``does fork'' symbol:
1002: \setbox0\hbox{$\textstyle\bigcup$}
1003: %%%%%\setbox1\hbox to \wd0{\hfil$\nmid$\hfil}
1004: \setbox1\hbox to \wd0{\hfil{\sl /\/}\hfil}
1005: \setbox2\hbox to \wd0{\hfil\vrule height \ht0 depth \dp0 width
1006: \forklinewidth\hfil}
1007: \wd1=0cm
1008: \wd2=0cm
1009: \newbox\doesforkbox
1010: \setbox\doesforkbox\hbox{\box1\box0\relax}
1011: \def\nunionstick{\mathop{\copy\doesforkbox}\limits}
1012:
1013: \def\fork#1#2_#3{#1\nunionstick_{\textstyle #3}#2}
1014: \def\forkin#1#2_#3^#4{#1\nunionstick_{\textstyle #3}^{\textstyle #4}#2}
1015: % cite redefined??no! \def\cite #1{\rm[#1]}
1016: \NoBlackBoxes
1017: \define\ortp{\text{\bf tp}}
1018: \define\motp{\text{\rm TP}}
1019: \define\sftp{\text{\rm tp}}
1020:
1021: \define\mr{\medskip\roster}
1022: \define\sn{\smallskip\noindent}
1023: \define\mn{\medskip\noindent}
1024: \define\bn{\bigskip\noindent}
1025: \define\ub{\underbar}
1026: \define\wilog{\text{without loss of generality}}
1027: \define\ermn{\endroster\medskip\noindent}
1028: \define\dbca{\dsize\bigcap}
1029: \define\dbcu{\dsize\bigcup}
1030: \define \nl{\newline}
1031: % (citealice) \define\blue{BLUE STARTS HERE}
1032: % (citealice) \define\endblue{BLUE STOPS HERE}
1033: \magnification=\magstep 1
1034: \documentstyle{amsppt}
1035: % % \input alice2000
1036: % *** start including alice2000.tex ***
1037: % This file should be inputted whenever we use amsppt.sty and
1038: % old tex.
1039: % Here we redefine \subjclass (use 1991 instead of 2000, otherwise
1040: % the following definition comes directly from
1041: %%
1042: %% `amsppt.sty', generated
1043: %% on <1997/2/2> with the docstrip utility (2.2i).
1044: %%
1045: %% The original source files were:
1046: %%
1047: %% amsppt.doc
1048: %%% ====================================================================
1049: %%% @AMSTeX-style-file{
1050: %%% filename = "amsppt.sty",
1051: %%% version = "2.1h",
1052: %%% date = "1997/02/02",
1053: %%% time = "09:27:44 EST",
1054: %%% checksum = "56844 3264 16617 137829",
1055: %%% author = "American Mathematical Society",
1056: %%% address = "PO Box 6248, Providence, RI 02940-6248, USA",
1057: %%% telephone = "401-455-4080 or (in the USA) 800-321-4AMS",
1058:
1059: { % the braces make the catcode-change local.
1060: \catcode`@11
1061:
1062: \ifx\alicetwothousandloaded@\relax
1063: \endinput\else\global\let\alicetwothousandloaded@\relax\fi
1064:
1065:
1066: \gdef\subjclass{\let\savedef@\subjclass
1067: \def\subjclass##1\endsubjclass{\let\subjclass\savedef@
1068: \toks@{\def\usualspace{{\rm\enspace}}\eightpoint}%
1069: \toks@@{##1\unskip.}%
1070: \edef\thesubjclass@{\the\toks@
1071: \frills@{{\noexpand\rm2000 {\noexpand\it Mathematics Subject
1072: Classification}.\noexpand\enspace}}%
1073: \the\toks@@}}%
1074: \nofrillscheck\subjclass}
1075: }
1076:
1077: % *** end including alice2000.tex ***
1078: % % \input alice2jlem
1079: % *** start including alice2jlem.tex ***
1080: %% # Keywords Input file to be used for texing Alice's files
1081:
1082: \expandafter\ifx\csname alice2jlem.tex\endcsname\relax
1083: \expandafter\xdef\csname alice2jlem.tex\endcsname{\the\catcode`@}
1084: \else \message{Hey! Apparently you were trying to
1085: \string\input{alice2jlem.tex} twice. This does not make sense.}
1086: \errmessage{Please edit your file (probably \jobname.tex) and remove
1087: any duplicate ``\string\input'' lines}\endinput\fi
1088:
1089:
1090:
1091: % % \input bib4plain
1092: % *** start including bib4plain.tex ***
1093: \expandafter\ifx\csname bib4plain.tex\endcsname\relax
1094: \expandafter\gdef\csname bib4plain.tex\endcsname{}
1095: \else \message{Hey! Apparently you were trying to \string\input
1096: bib4plain.tex twice. This does not make sense.}
1097: \errmessage{Please edit your file (probably \jobname.tex) and remove
1098: any duplicate ``\string\input'' lines}\endinput\fi
1099:
1100:
1101: % This file should be inputted if you want to use
1102: % bibtex fom within plain TeX.
1103: \def\makeatletter{\catcode`\@11 } % Not really need for standard
1104: \def\makeatother{\catcode`\@12 } % bibtex files, but these commands
1105: \def\renewcommand{\newcommand} % are used in our literal-unsrt.bst
1106: % and other files.
1107: % The above lines were written by Martin.Goldstern@tuwien.ac.at
1108: % The rest is just the file btxmac.tex.
1109: % (except for one small change below, regarding undefined citations)
1110:
1111: %% @texfile{
1112: %% author = "Karl Berry and Oren Patashnik",
1113: %% version = "0.99j",
1114: %% date = "14 Mar 1992",
1115: %% filename = "btxmac.tex",
1116: %% address = "Please use electronic mail",
1117: %% checksum = "834 4503 33061",
1118: %% email = "opbibtex@cs.stanford.edu",
1119: %% codetable = "ISO/ASCII",
1120: %% supported = "yes",
1121: %% docstring = "Defines macros that make BibTeX work with plain TeX",
1122: %% }
1123: % BibTeX-for-TeX macros, version 0.99j, for BibTeX 0.99c, TeX 3.0 or later.
1124: % Copyright (C) 1990--92 by Karl Berry and Oren Patashnik; all rights reserved.
1125: % You may copy this file provided: that it's accompanied by the
1126: % "BibTeXing" document, whose text is contained in the file `btxdoc.tex';
1127: % that any documentation you write for these macros also gives a
1128: % reference for "BibTeXing"; and that either you make absolutely no
1129: % changes to your copy, or if you do make changes, (1) you name the file
1130: % something other than `btxmac.tex' and you remove all occurrences of
1131: % `btxmac.tex' from the file, (2) you put, somewhere in the first twenty
1132: % lines of the file, your name, along with an electronic address at which
1133: % others who might use the file may reach you, and (3) you remove each
1134: % occurrence of Oren's name and electronic address from this file. These
1135: % restrictions help ensure that all standard versions of these macros are
1136: % identical, and that Oren doesn't get deluged with inappropriate e-mail.
1137: %
1138: % This file, btxmac.tex, contains TeX macros that allow BibTeX, a
1139: % bibliography program that was originally designed for use with LaTeX,
1140: % to work with plain TeX. Please report any bugs (outright goofs,
1141: % improvable macros, misfeatures, or unclear documentation) to Oren
1142: % Patashnik (opbibtex@cs.stanford.edu). These macros will become frozen
1143: % shortly after BibTeX version 1.00 is released.
1144: %
1145: % AMS-TEX WARNING: We tried very hard, for version .99i of these macros,
1146: % to make them compatible with AmS-TeX. We succeeded to the extent
1147: % that, if you use one of the standard bibliography styles, you probably
1148: % won't notice any problems with version 0.99i of btxmac.tex. But
1149: % ultimately we failed, in that the inherent incompatibilities between
1150: % plain TeX and AmS-TeX kept making these macros break, for certain
1151: % inputs or certain styles. Examples: (1) AmS-TeX treats at-signs as
1152: % special, in ways that plain TeX and LaTeX don't, so that, for example,
1153: % you can't have any `@' characters in an argument to the \cite command,
1154: % the way you can in TeX or LaTeX; (2) AmS-TeX decided that plain TeX's
1155: % and LaTeX's macron-accent control sequence `\=' should be undefined;
1156: % so you'll need to define `\=' to be `\B' to get the xampl.bib example
1157: % suggested below to work with AmS-TeX; (3) AmS-TeX redefines the tie
1158: % character `~' of plain TeX, and AmS-TeX's `amsppt' style redefines
1159: % plain TeX's `\nobreak' macro, so that if you use an author-date style
1160: % like `apalike' and you have a multiple-author reference for which the
1161: % author-date style automatically produces a citation in the text like
1162: % `(Jones et~al., 1992)' you will throw AmS-TeX's `amsppt' style into
1163: % an infinite loop, exceeding its input stack size. In practice, such
1164: % incompatibilities surface infrequently; but it is now clear to us that
1165: % it's not worth the effort (perhaps it's not even possible) to make the
1166: % btxmac.tex macros robust when used both with plain TeX and Ams-TeX.
1167: % If the BibTeX/AmS-TeX results attainable with the current btxmac.tex
1168: % macros are sufficient, fine. But if there's a demand for more robust
1169: % BibTeX/AmS-TeX behavior, then someone who's very familiar with the
1170: % AmS-TeX package should probably make an amsbtxmc.tex version of the
1171: % macros (remembering to follow the copyright restrictions above).
1172: % Until then, if you're an AmS-TeX user, or a LaTeX or plain TeX user
1173: % sharing files with an AmS-TeX user, beware.
1174: % END OF AMS-TEX WARNING.
1175: %
1176: % To use these macros you should be familiar with how BibTeX interacts
1177: % with LaTeX, since BibTeX's interaction with TeX is very similar; that
1178: % interaction is explained in the LaTeX manual. It also helps to
1179: % have read "BibTeXing", the documentation that accompanies BibTeX.
1180: % Then, if you want, you should redefine any of the macros that begin
1181: % with `\bbl' or `\biblabel' or `\print' that you need to get formatting
1182: % different from the default (the default settings are designed to
1183: % accompany a bibliography style like BibTeX's standard style `plain').
1184: % The macros you might want to change are described briefly a few
1185: % paragraphs hence. [To get started without reading any documentation,
1186: % try running the nine-line .tex file below through TeX and BibTeX.
1187: % Remember the general scheme: Running (La)TeX writes information on
1188: % the .aux (auxiliary) file; then running BibTeX reads information from
1189: % the .aux, .bst (style), and .bib (database) files, and writes
1190: % information (the bibliography) on a .bbl file; then running (La)TeX
1191: % incorporates the bibliography; then running (La)TeX once more fixes
1192: % the remaining forward references into the bibliography. Thus, to get
1193: % everything incorporated into your output, you'll have to run (La)TeX,
1194: % BibTeX, (La)TeX, (La)TeX. (Standup, sitdown, fight, fight, fight.)]
1195: %
1196: % These macros can stand alone or they can be \input into a macro
1197: % package, like Eplain, that is sufficiently compatible with plain TeX.
1198: % To use these macros to format the 0.99 version of the xampl.bib file
1199: % that's distributed with BibTeX (that version of the file has no
1200: % self-identification), you'll need to define \mbox, which is a LaTeX
1201: % command, to be \hbox, as in the example below.
1202: %
1203: % Here's a nine-line plain TeX file for trying out btxmac.tex; of course
1204: % you'll have to remove the comment characters at the beginning of each
1205: % line, and, depending on your system, you might have to take steps so
1206: % that BibTeX can "see" the files xampl.bib and plain.bst (BibTeX will
1207: % give you two empty-field warning messages that you should ignore).
1208: %
1209: % \def\mbox#1{\leavevmode\hbox{#1}}
1210: % \input btxmac
1211: % \noindent This cites Aamport's gnominious article~\cite{article-full}.
1212: % \medskip
1213: % \leftline{\bf References}
1214: % \nocite{*} % put all database entries into the reference list
1215: % \bibliography{xampl} % specify the database files; here, just xampl.bib
1216: % \bibliographystyle{plain} % specify plain.bst as the style file
1217: % \bye
1218: %
1219: %
1220: % HISTORY
1221: %
1222: % Karl Berry wrote the original version of these macros in 1989 and
1223: % 1990, for use in his `Eplain' package. Oren Patashnik modified them
1224: % slightly in July 1990, as part of the official BibTeX distribution.
1225: %
1226: % 1-Aug-90 Version 0.99a, not released to the general public.
1227: % 14-Aug-90 0.99b, first general release.
1228: % 26-Aug-90 0.99c, made \@undefinedmessage work with other macro packages.
1229: % 6-Sep-90 0.99d, allowed for general formatting of bibliography labels,
1230: % for general formatting of (in-text) citations, and for
1231: % changing certain catcodes while reading the .aux file.
1232: % 14-Nov-90 0.99e, changed the way \@setletters works, made some \new...'s
1233: % non-outer, and changed the way Eplain reads this file.
1234: % 12-Dec-90 0.99f, made \@resetnumerals change the `,' and `.' catcodes; and
1235: % added \biblabelextrahang, \@getoptionalarg, and \bblsc.
1236: % 11-Mar-91 0.99g, made a few minor changes required by the way Eplain reads
1237: % this file, but no functional changes.
1238: % 24-Apr-91 0.99h, inhibited the reading and writing of the .aux file if it
1239: % isn't used or if the \noauxfile macro is defined, and
1240: % removed some .aux-file-opening detritus; printed the
1241: % cite-key of undefined citations in \tt font; changed the
1242: % catcode of `_' inside \cite; and called \@resetnumerals
1243: % from inside a group.
1244: % 29-Feb-92 0.99i, made these macros semi-compatible with AmS-TeX; removed
1245: % \@resetnumerals, \@setletters, \@tokstostring, and
1246: % friends; changed the way \cite handles catcodes; changed
1247: % \@getoptionalarg, and had \bibitem and \newcommand use
1248: % it; added \@futurenonspacelet and (to facilitate the use
1249: % of multiple reference lists) \bblfilebasename; changed
1250: % \biblabelprint to use the new macros \biblabelprecontents
1251: % and \biblabelpostcontents, and to, by default, right-
1252: % justify numeric labels; and renamed \biblabelextrahang to
1253: % the more descriptive \biblabelextraspace.
1254: % 14-Mar-92 0.99j, made 0.99i's use of `\\' local to btxmac.tex.
1255: %
1256: %
1257: % The LaTeX-related commands defined in this file include (a) the four
1258: % commands that a user types (\bibliography, \bibliographystyle, \cite,
1259: % and \nocite); (b) the three commands that BibTeX looks for in the .aux
1260: % file (\bibdata, \bibstyle, and \citation---there is a fourth command
1261: % that BibTeX looks for, but that command is related to LaTeX's \include
1262: % facility, so these macros ignore that command); and (c) a LaTeX
1263: % command (\newcommand) that's written by one of the four standard
1264: % bibliography styles (alpha). The definitions here are not exactly the
1265: % same as the corresponding LaTeX definitions (those eight LaTeX
1266: % definitions depend on a significant fraction of LaTeX itself). But
1267: % the only substantial differences are with \newcommand, which here,
1268: % without complaining, lets you redefine a preexisting control sequence
1269: % (in LaTeX, \newcommand won't let you redefine a preexisting command),
1270: % and which here doesn't make the control sequences it defines \long (in
1271: % LaTeX, that happens automatically); there may also be other minor
1272: % differences. To summarize: Unless you know what you're doing, you
1273: % shouldn't define any control sequences with these eight names:
1274: %
1275: % \bibdata
1276: % \bibliography
1277: % \bibliographystyle
1278: % \bibstyle
1279: % \citation
1280: % \cite
1281: % \newcommand
1282: % \nocite
1283: %
1284: % There are three other commands written by one or more of the four
1285: % standard (plain, abbrv, alpha, unsrt) or four semistandard (acm,
1286: % apalike, ieeetr, siam) bibliography styles; they take effect only
1287: % within the bibliography, and are redefinable, as explained later:
1288: %
1289: % \em
1290: % \newblock
1291: % \sc
1292: %
1293: % There's one control sequence you might want to use (but not redefine)
1294: % in redefining \biblabelprint:
1295: %
1296: % \biblabelwidth
1297: %
1298: % There are fifteen other control sequences (explained later in more detail)
1299: % that the macros of this file will use if you define them---you should
1300: % define them after the \input btxmac command but before the \bibliography
1301: % command. The first six begin with `\bbl' and affect fonts, spacing,
1302: % perhaps other characteristics of the bibliography, and which .bbl files
1303: % get read; the next five begin with `\biblabel' and determine how labels
1304: % are formatted in the bibliography; and the last four begin with `\print'
1305: % and determine how the in-text citations are formatted:
1306: %
1307: % \bblem
1308: % \bblfilebasename
1309: % \bblhook
1310: % \bblnewblock
1311: % \bblrm
1312: % \bblsc
1313: % \biblabelcontents
1314: % \biblabelprecontents
1315: % \biblabelprint
1316: % \biblabelpostcontents
1317: % \biblabelextraspace
1318: % \printbetweencitations
1319: % \printcitefinish
1320: % \printcitenote
1321: % \printcitestart
1322: %
1323: % If it's defined before the \input btxmac command, the control sequence
1324: % below inhibits the reading and writing of the .aux file(s), and the
1325: % issuing of related warning messages. Any definition will do. This
1326: % feature might help when you're working on draft stages of a document:
1327: %
1328: % \noauxfile
1329: %
1330: %
1331: % Here's another control sequence (it's described later) that you
1332: % probably won't want to redefine unless you are writing another macro
1333: % package; if you do redefine it, however, do it before the \input btxmac
1334: % command (and notice that it has an `@' in its name):
1335: %
1336: % \@undefinedmessage
1337: %
1338: % Any other control sequence in this file that might conflict with
1339: % something you've defined will have an `@' in its name, so such conflicts
1340: % are unlikely; but if you're worried about a specific control sequence
1341: % name, do a text search of this file to look for it.
1342: %
1343: %
1344: % So to start things off we turn `@' into a letter (category code 11),
1345: % keeping track of the old category code for future restoration.
1346: % (Simply resetting it to 12 when we leave these macros is
1347: % insufficient.) The use of `\cite' as a temporary control sequence is
1348: % a kludge, but it's a reasonably simple way to accomplish what we need
1349: % without possibly overwriting something (without an `@' in its name)
1350: % that might already be defined.
1351: %
1352: \edef\cite{\the\catcode`@}%
1353: \catcode`@ = 11
1354: \let\@oldatcatcode = \cite
1355: \chardef\@letter = 11
1356: \chardef\@other = 12
1357: %
1358: %
1359: % Next come some things that will be useful later.
1360: %
1361: % Make an outer definition into an inner one (due to Chris Thompson).
1362: % The arguments should be the control sequence to be defined, and the
1363: % new of the \outer control sequence, as characters; the control
1364: % sequence #1 is defined to be just the same as \csname#2\endcsname, but
1365: % not \outer. For example, \@innerdef\innernewcount{newcount} would
1366: % define \innernewcount to be a non-outer version of \newcount.
1367: %
1368: \def\@innerdef#1#2{\edef#1{\expandafter\noexpand\csname #2\endcsname}}%
1369: %
1370: % We use \@innerdef to make some of our allocations local, because
1371: % Eplain includes our code inside a conditional. We put @'s in the
1372: % names to minimize the (already small) chance of conflicts.
1373: %
1374: \@innerdef\@innernewcount{newcount}%
1375: \@innerdef\@innernewdimen{newdimen}%
1376: \@innerdef\@innernewif{newif}%
1377: \@innerdef\@innernewwrite{newwrite}%
1378: %
1379: %
1380: % Swallow one parameter.
1381: %
1382: \def\@gobble#1{}%
1383: %
1384: %
1385: % Use TeX 3.0's \inputlineno to get the line number, for better error
1386: % messages, but if we're using an old version of TeX, don't do anything.
1387: %
1388: \ifx\inputlineno\@undefined
1389: \let\@linenumber = \empty % Pre-3.0.
1390: \else
1391: \def\@linenumber{\the\inputlineno:\space}%
1392: \fi
1393: %
1394: %
1395: % The following macro \@futurenonspacelet (from the TeXbook) behaves
1396: % essentially like \futurelet except that it discards any implicit or
1397: % explicit space tokens that intervene before a nonspace is scanned:
1398: %
1399: \def\@futurenonspacelet#1{\def\cs{#1}%
1400: \afterassignment\@stepone\let\@nexttoken=
1401: }%
1402: \begingroup % The grouping here avoids stepping on an outside use of `\\'.
1403: \def\\{\global\let\@stoken= }%
1404: \\ % now \@stoken is a space token (\\ is a control symbol, so that
1405: % space after it is seen).
1406: \endgroup
1407: \def\@stepone{\expandafter\futurelet\cs\@steptwo}%
1408: \def\@steptwo{\expandafter\ifx\cs\@stoken\let\@@next=\@stepthree
1409: \else\let\@@next=\@nexttoken\fi \@@next}%
1410: \def\@stepthree{\afterassignment\@stepone\let\@@next= }%
1411: %
1412: %
1413: % \@getoptionalarg\CS gets an optional argument from the input, enclosed
1414: % in brackets, then expands \CS. We set \@optionalarg to \empty if we
1415: % don't find one, otherwise to the text of the argument. This assumes
1416: % the brackets don't have a funny category code.
1417: %
1418: \def\@getoptionalarg#1{%
1419: \let\@optionaltemp = #1%
1420: \let\@optionalnext = \relax
1421: \@futurenonspacelet\@optionalnext\@bracketcheck
1422: }%
1423: %
1424: % The \expandafter's in this macro let us avoid the use of \aftergroup,
1425: % which is somewhat more expensive.
1426: %
1427: \def\@bracketcheck{%
1428: \ifx [\@optionalnext
1429: \expandafter\@@getoptionalarg
1430: \else
1431: \let\@optionalarg = \empty
1432: % We can't do the \temp after the \fi, because then the \temp gets
1433: % in the way of reading the optional argument from the input, if
1434: % we do have one.
1435: \expandafter\@optionaltemp
1436: \fi
1437: }%
1438: %
1439: \def\@@getoptionalarg[#1]{%
1440: \def\@optionalarg{#1}%
1441: \@optionaltemp
1442: }%
1443: %
1444: %
1445: % From LaTeX.
1446: %
1447: \def\@nnil{\@nil}%
1448: \def\@fornoop#1\@@#2#3{}%
1449: %
1450: \def\@for#1:=#2\do#3{%
1451: \edef\@fortmp{#2}%
1452: \ifx\@fortmp\empty \else
1453: \expandafter\@forloop#2,\@nil,\@nil\@@#1{#3}%
1454: \fi
1455: }%
1456: %
1457: \def\@forloop#1,#2,#3\@@#4#5{\def#4{#1}\ifx #4\@nnil \else
1458: #5\def#4{#2}\ifx #4\@nnil \else#5\@iforloop #3\@@#4{#5}\fi\fi
1459: }%
1460: %
1461: \def\@iforloop#1,#2\@@#3#4{\def#3{#1}\ifx #3\@nnil
1462: \let\@nextwhile=\@fornoop \else
1463: #4\relax\let\@nextwhile=\@iforloop\fi\@nextwhile#2\@@#3{#4}%
1464: }%
1465: %
1466: %
1467: % This macro tests if a file \jobname.#1 exists, and sets \if@fileexists
1468: % appropriately. If an optional argument is given, it is used as the
1469: % root part of the filename instead of \jobname.
1470: %
1471: \@innernewif\if@fileexists
1472: %
1473: \def\@testfileexistence{\@getoptionalarg\@finishtestfileexistence}%
1474: \def\@finishtestfileexistence#1{%
1475: \begingroup
1476: \def\extension{#1}%
1477: \immediate\openin0 =
1478: \ifx\@optionalarg\empty\jobname\else\@optionalarg\fi
1479: \ifx\extension\empty \else .#1\fi
1480: \space
1481: \ifeof 0
1482: \global\@fileexistsfalse
1483: \else
1484: \global\@fileexiststrue
1485: \fi
1486: \immediate\closein0
1487: \endgroup
1488: }%
1489: %
1490: %
1491: %% [[[start of BibTeX-specific stuff]]]
1492: %
1493: % Now come the four main LaTeX commands and their associated .aux
1494: % commands. Just as in LaTeX, \bibliographystyle defines the BibTeX
1495: % style name (.bst file, that is), and \bibliography defines the
1496: % database (.bib) file(s). The corresponding .aux-file commands are
1497: % \bibstyle and \bibdata, which are there only for BibTeX's (but not
1498: % LaTeX's) use.
1499: %
1500: \def\bibliographystyle#1{%
1501: \@readauxfile
1502: \@writeaux{\string\bibstyle{#1}}%
1503: }%
1504: \let\bibstyle = \@gobble
1505: %
1506: % As well as writing the \bibdata command to tell BibTeX which .bib
1507: % files to read, we read the .bbl file that BibTeX (or a person,
1508: % conceivably) has produced. We use \bblfilebasename as the root of the
1509: % filename to read; this defaults to \jobname.
1510: %
1511: \let\bblfilebasename = \jobname
1512: \def\bibliography#1{%
1513: \@readauxfile
1514: \@writeaux{\string\bibdata{#1}}%
1515: \@testfileexistence[\bblfilebasename]{bbl}%
1516: \if@fileexists
1517: % We just output a non-discardable item (the `whatsit' with the
1518: % \bibdata command). This means that the glue that will be
1519: % inserted next (\parskip or \baselineskip, most likely) will be a
1520: % legal breakpoint. Most likely, this is after some kind of
1521: % heading, where we don't want to allow a page break. So:
1522: \nobreak
1523: \@readbblfile
1524: \fi
1525: }%
1526: \let\bibdata = \@gobble
1527: %
1528: % The \nocite{label,label,...} command writes its argument to \@auxfile,
1529: % unless instructed not to, but produces no text in the document. Both
1530: % the \nocite and \cite commands produce \citation commands in the .aux file.
1531: %
1532: \def\nocite#1{%
1533: \@readauxfile
1534: \@writeaux{\string\citation{#1}}%
1535: }%
1536: %
1537: \@innernewif\if@notfirstcitation
1538: %
1539: % \cite[note]{label,label,...} produces the citations for the labels as
1540: % well. If the optional argument `note' is present, it's added after
1541: % the labels. Since \cite calls \nocite to do its .aux-file writing,
1542: % \cite doesn't need to call \@readauxfile (\nocite does).
1543: %
1544: \def\cite{\@getoptionalarg\@cite}%
1545: %
1546: % Typeset the citations for the labels in #1, followed by the note, if
1547: % it exists. To change the citation's format in the text, redefine one
1548: % or more `\print...' macros, whose defaults appear later in this file.
1549: %
1550: \def\@cite#1{%
1551: % Remember the optional argument, in case one of the macros we call
1552: % below ends up looking for an optional argument itself. For
1553: % example, if a \cite[note] triggers reading the .aux file, then the
1554: % [note] would be clobbered, since \@testfileexistence looks for an
1555: % optional arg.
1556: \let\@citenotetext = \@optionalarg
1557: % Start printing the text, beginning with a left bracket by default.
1558: \printcitestart
1559: % It's complicated, but because \nocite puts a `whatsit' onto the list,
1560: % \nocite should follow \printcitestart. It's conceivable, but very
1561: % unlikely, that this `whatsit' will cause a problem (glue that doesn't
1562: % disappear when you want it to is the most likely symptom), requiring
1563: % a change either to \printcitestart or to the label that the .bst file
1564: % produces.
1565: \nocite{#1}%
1566: \@notfirstcitationfalse
1567: \@for \@citation :=#1\do
1568: {%
1569: \expandafter\@onecitation\@citation\@@
1570: }%
1571: \ifx\empty\@citenotetext\else
1572: \printcitenote{\@citenotetext}%
1573: \fi
1574: \printcitefinish
1575: }%
1576: %
1577: \newif\ifweareinprivate
1578: \weareinprivatetrue
1579: \ifx\shlhetal\undefinedcontrolseq\weareinprivatefalse\fi
1580: \ifx\shlhetal\relax\weareinprivatefalse\fi
1581: \def\@onecitation#1\@@{%
1582: \if@notfirstcitation
1583: \printbetweencitations
1584: \fi
1585: %
1586: \expandafter \ifx \csname\@citelabel{#1}\endcsname \relax
1587: \if@citewarning
1588: \message{\@linenumber Undefined citation `#1'.}%
1589: \fi
1590: % Give it a dummy definition:
1591: \ifweareinprivate
1592: \expandafter\gdef\csname\@citelabel{#1}\endcsname{%
1593: % Change: marginal remark added, goldstrn@math.huji.ac.il,
1594: % goldstern@tuwien.ac.at, May 1996 mg
1595: % !!! change !!!
1596: \strut
1597: \vadjust{\vskip-\dp\strutbox
1598: \vbox to 0pt{\vss\parindent0cm \leftskip=\hsize
1599: \advance\leftskip3mm
1600: \advance\hsize 4cm\strut\openup-4pt
1601: \rightskip 0cm plus 1cm minus 0.5cm ? #1 ?\strut}}
1602: {\tt
1603: \escapechar = -1
1604: \nobreak\hskip0pt\pfeilsw%\special{ps:100 0 0 setrgbcolor }%
1605: \expandafter\string\csname#1\endcsname
1606: %\special{ps:0 0 0 setrgbcolor }
1607: \pfeilso
1608: \nobreak\hskip0pt
1609: }%
1610: }%
1611: \else % ifweareinprivate = false
1612: \expandafter\gdef\csname\@citelabel{#1}\endcsname{%
1613: {\tt\expandafter\string\csname#1\endcsname}
1614: }%
1615: \fi % ifweareinprivate
1616: \fi
1617: % Now produce the text, whether it was undefined or not.
1618: \csname\@citelabel{#1}\endcsname
1619: \@notfirstcitationtrue
1620: }%
1621: %
1622: % Given a label `foo', the macro `\b@foo' is supposed to
1623: % hold the text that should be produced.
1624: %
1625: \def\@citelabel#1{b@#1}%
1626: %
1627: % So, how does a citation label get defined? When we read the .bbl file
1628: % (below), a \bibitem writes out a \@citedef command. And when we read
1629: % the \@citedef, we define \@citelabel{#1}, where #1 is the user's
1630: % label.
1631: %
1632: \def\@citedef#1#2{\expandafter\gdef\csname\@citelabel{#1}\endcsname{#2}}%
1633: %
1634: %
1635: % Reading the .bbl file also produces the typeset bibliography. Please
1636: % notice, however, that we do not produce the title for the references
1637: % (e.g., `References'), as LaTeX does. The formatting and spacing of
1638: % that title, whether it should go into the headline, and so on, are all
1639: % things determined by your format. We cannot know those things in
1640: % advance. If you wish, you can define \bblhook to produce the title.
1641: % Or just do it before the \bibliography command.
1642: %
1643: \def\@readbblfile{%
1644: % Define a counter to tell us which item number we are on, unless
1645: % we've already defined it (because the document has more than one
1646: % bibliography).
1647: \ifx\@itemnum\@undefined
1648: \@innernewcount\@itemnum
1649: \fi
1650: %
1651: \begingroup
1652: \def\begin##1##2{%
1653: % ##1 is just `thebibliography'.
1654: % ##2 is the widest label.
1655: % We set (new dimen) \biblabelwidth based on the widest label
1656: \setbox0 = \hbox{\biblabelcontents{##2}}%
1657: \biblabelwidth = \wd0
1658: }%
1659: \def\end##1{}% ##1 is `thebibliography' again.
1660: %
1661: % Here we have two possibilities:
1662: % \bibitem[typesetlabel]{citationlabel}
1663: % \bibitem{citationlabel}
1664: % If we have the second of these, the citations are numbered, starting
1665: % from one; we use our own count register \@itemnum for this.
1666: %
1667: \@itemnum = 0
1668: \def\bibitem{\@getoptionalarg\@bibitem}%
1669: \def\@bibitem{%
1670: \ifx\@optionalarg\empty
1671: \expandafter\@numberedbibitem
1672: \else
1673: \expandafter\@alphabibitem
1674: \fi
1675: }%
1676: \def\@alphabibitem##1{%
1677: % Need \xdef here for various reasons.
1678: \expandafter \xdef\csname\@citelabel{##1}\endcsname {\@optionalarg}%
1679: % Left-justify alpha labels, unless \biblabel{pre,post}contents
1680: % are already defined.
1681: \ifx\biblabelprecontents\@undefined
1682: \let\biblabelprecontents = \relax
1683: \fi
1684: \ifx\biblabelpostcontents\@undefined
1685: \let\biblabelpostcontents = \hss
1686: \fi
1687: \@finishbibitem{##1}%
1688: }%
1689: %
1690: \def\@numberedbibitem##1{%
1691: \advance\@itemnum by 1
1692: \expandafter \xdef\csname\@citelabel{##1}\endcsname{\number\@itemnum}%
1693: % Right-justify numeric labels, unless \biblabel{pre,post}contents
1694: % are already defined.
1695: \ifx\biblabelprecontents\@undefined
1696: \let\biblabelprecontents = \hss
1697: \fi
1698: \ifx\biblabelpostcontents\@undefined
1699: \let\biblabelpostcontents = \relax
1700: \fi
1701: \@finishbibitem{##1}%
1702: }%
1703: %
1704: \def\@finishbibitem##1{%
1705: \biblabelprint{\csname\@citelabel{##1}\endcsname}%
1706: \@writeaux{\string\@citedef{##1}{\csname\@citelabel{##1}\endcsname}}%
1707: \ignorespaces
1708: }%
1709: %
1710: % Do the printing (we're producing the bibliography, remember).
1711: %
1712: \let\em = \bblem
1713: \let\newblock = \bblnewblock
1714: \let\sc = \bblsc
1715: % Punctuation won't affect spacing;
1716: \frenchspacing
1717: % the penalties below are from LaTeX's [article,book,report].sty;
1718: \clubpenalty = 4000 \widowpenalty = 4000
1719: % the next two values come from LaTeX's \sloppy command;
1720: \tolerance = 10000 \hfuzz = .5pt
1721: \everypar = {\hangindent = \biblabelwidth
1722: \advance\hangindent by \biblabelextraspace}%
1723: \bblrm
1724: % the \parskip is a guess at what looks good;
1725: \parskip = 1.5ex plus .5ex minus .5ex
1726: % and the space between label and text comes from LaTeX's \labelsep.
1727: \biblabelextraspace = .5em
1728: \bblhook
1729: %
1730: \input \bblfilebasename.bbl
1731: \endgroup
1732: }%
1733: %
1734: % The widest label's width is useful for redefining \biblabelprint;
1735: % you redefine \biblabelwidth, in effect, by redefining the
1736: % \biblabelcontents macro that appears below. And \biblabelextraspace,
1737: % which is redefinable inside \bblhook, is added to \biblabelwidth to
1738: % determine the amount of hanging indentation.
1739: %
1740: \@innernewdimen\biblabelwidth
1741: \@innernewdimen\biblabelextraspace
1742: %
1743: % Now come the main macros that are related to the printing of the
1744: % bibliography. Since you might want to redefine them, they are given
1745: % default definitions outside of \@readbblfile.
1746: %
1747: % The first one controls the printing of a bibliography entry's label.
1748: % If you change it, make sure that it starts with something like
1749: % \noindent or \indent or \leavevmode that puts TeX into horizontal mode
1750: % (even if the label itself is empty); otherwise, the hanging
1751: % indentation will get messed up in certain circumstances.
1752: %
1753: \def\biblabelprint#1{%
1754: \noindent
1755: \hbox to \biblabelwidth{%
1756: \biblabelprecontents
1757: \biblabelcontents{#1}%
1758: \biblabelpostcontents
1759: }%
1760: \kern\biblabelextraspace
1761: }%
1762: %
1763: % If you are using numeric labels, and you want them left-justified
1764: % (numeric labels by default are right-justified), do something like:
1765: % \def\biblabelprecontents{\relax}
1766: % \def\biblabelpostcontents{\hss}
1767: %
1768: % By default the labels are typeset in \bblrm, and enclosed in brackets.
1769: %
1770: \def\biblabelcontents#1{{\bblrm [#1]}}%
1771: %
1772: % The main text, too, is typeset using \bblrm, which is \rm by default.
1773: %
1774: \def\bblrm{\rm}%
1775: %
1776: % Emphasis for producing, e.g., titles, is done with \it by default.
1777: %
1778: \def\bblem{\it}%
1779: %
1780: % Some styles use a caps-and-small-caps font for author names. LaTeX
1781: % defines an \sc command but plain TeX doesn't, so we need one here.
1782: % The definition below doesn't load the font unless it's needed, but it
1783: % tries to load only the 10pt version, because it might not exist at
1784: % other point sizes.
1785: %
1786: \def\bblsc{\ifx\@scfont\@undefined
1787: \font\@scfont = cmcsc10
1788: \fi
1789: \@scfont
1790: }%
1791: %
1792: % The major parts of an entry are separated with \bblnewblock. The
1793: % numbers below are taken from LaTeX's `article' style.
1794: %
1795: \def\bblnewblock{\hskip .11em plus .33em minus .07em }%
1796: %
1797: % Here's where you stick any other bibliography-formatting goodies, or
1798: % redefine the values above.
1799: %
1800: \let\bblhook = \empty
1801: %
1802: %
1803: % Here are the four default definitions for formatting the in-text
1804: % citations. These are what you redefine (after your \input btxmac but
1805: % before your \bibliography) to get parens instead of brackets, or
1806: % superscripts, or footnotes, or whatever.
1807: %
1808: \def\printcitestart{[}% left bracket
1809: \def\printcitefinish{]}% right bracket
1810: \def\printbetweencitations{, }% comma, space
1811: \def\printcitenote#1{, #1}% comma, space, note (if it exists)
1812: %
1813: % That scheme is pretty flexible. For example you could use
1814: % \def\printcitestart{\unskip $^\bgroup}
1815: % \def\printcitefinish{\egroup$}
1816: % \def\printbetweencitations{,}
1817: % \def\printcitenote#1{\hbox{\sevenrm\space (#1)}}
1818: % \font\eighttt = cmtt8
1819: % \scriptfont\ttfam = \eighttt
1820: % to get superscripted in-text citations. (The scriptfont stuff
1821: % exists only to print an undefined citation; it's in cmtt8 because
1822: % there is no cmtt7.) To get something radically different, however,
1823: % you'll have to define your own \cite command.
1824: %
1825: % When we read `\citation' from the .aux file, it means nothing.
1826: %
1827: \let\citation = \@gobble
1828: %
1829: %
1830: % Now comes the stuff for dealing with LaTeX's \newcommand. As
1831: % mentioned earlier, this \newcommand will redefine a preexisting
1832: % command; that's different from how LaTeX's \newcommand behaves.
1833: %
1834: \@innernewcount\@numparams
1835: %
1836: % \newcommand{\foo}[n]{text} defines the control sequence \foo to have
1837: % n parameters, and replacement text `text'.
1838: %
1839: \def\newcommand#1{%
1840: \def\@commandname{#1}%
1841: \@getoptionalarg\@continuenewcommand
1842: }%
1843: %
1844: % Figure out if this definition has parameters.
1845: %
1846: \def\@continuenewcommand{%
1847: % If no optional argument, we have zero parameters. Otherwise, we
1848: % have that many.
1849: \@numparams = \ifx\@optionalarg\empty 0\else\@optionalarg \fi \relax
1850: \@newcommand
1851: }%
1852: %
1853: % \@numparams is how many arguments this command has. The name of the
1854: % command is \@commandname. The replacement text for the new macro is #1.
1855: %
1856: \def\@newcommand#1{%
1857: \def\@startdef{\expandafter\edef\@commandname}%
1858: \ifnum\@numparams=0
1859: \let\@paramdef = \empty
1860: \else
1861: \ifnum\@numparams>9
1862: \errmessage{\the\@numparams\space is too many parameters}%
1863: \else
1864: \ifnum\@numparams<0
1865: \errmessage{\the\@numparams\space is too few parameters}%
1866: \else
1867: \edef\@paramdef{%
1868: % This is disgusting, but \loop doesn't work inside \edef,
1869: % because \body isn't defined.
1870: \ifcase\@numparams
1871: \empty No arguments.
1872: \or ####1%
1873: \or ####1####2%
1874: \or ####1####2####3%
1875: \or ####1####2####3####4%
1876: \or ####1####2####3####4####5%
1877: \or ####1####2####3####4####5####6%
1878: \or ####1####2####3####4####5####6####7%
1879: \or ####1####2####3####4####5####6####7####8%
1880: \or ####1####2####3####4####5####6####7####8####9%
1881: \fi
1882: }%
1883: \fi
1884: \fi
1885: \fi
1886: \expandafter\@startdef\@paramdef{#1}%
1887: }%
1888: %
1889: %% [[[end of BibTeX-specific stuff]]]
1890: %
1891: %
1892: % Names of references (arguments given in the \cite and \nocite
1893: % commands) and file names (arguments given in the \bibliography and
1894: % \bibliographystyle commands) are recorded in \jobname.aux, called the
1895: % \@auxfile in these macros. Here's how they get read in.
1896: %
1897: \def\@readauxfile{%
1898: \if@auxfiledone \else % remember: \@auxfiledonetrue if \noauxfile is defined
1899: \global\@auxfiledonetrue
1900: \@testfileexistence{aux}%
1901: \if@fileexists
1902: \begingroup
1903: % Because we might be in horizontal mode when \@readauxfile
1904: % is called (if it's in response to a \cite or \nocite), we
1905: % want to ignore all the would-be spaces at the ends of
1906: % lines in the aux file. Fortunately, it's highly unlikely
1907: % an end-of-line might actually be desired.
1908: % And because we don't change the category code of anything
1909: % but @, primitives like \gdef can't be used to define labels
1910: % in the aux file. The solution adopted by btxmac.tex is to
1911: % write `\@citedef{LABEL}{DEFINITION}' to the aux file, and
1912: % use \csname on LABEL.
1913: \endlinechar = -1
1914: \catcode`@ = 11
1915: \input \jobname.aux
1916: \endgroup
1917: \else
1918: \message{\@undefinedmessage}%
1919: \global\@citewarningfalse
1920: \fi
1921: \immediate\openout\@auxfile = \jobname.aux
1922: \fi
1923: }%
1924: %
1925: % The \@readauxfile macro does all that work the first time it's called.
1926: % Since it's called once for every \cite, \nocite, \bibliography, and
1927: % \bibliographystyle command that the user issues, we need to remember
1928: % whether the work's been done. It's considered done if we're not to do
1929: % it---that is, if \noauxfile is defined.
1930: %
1931: \newif\if@auxfiledone
1932: \ifx\noauxfile\@undefined \else \@auxfiledonetrue\fi
1933: %
1934: % It's conceivable you'd want to change how other characters are read;
1935: % to do that, change their category code before doing \input btxmac.
1936: %
1937: %
1938: % After reading the .aux file, \@readauxfile opens it for writing.
1939: % The \@writeaux macro does the actual writing (as long as
1940: % \noauxfile is undefined).
1941: %
1942: \@innernewwrite\@auxfile
1943: \def\@writeaux#1{\ifx\noauxfile\@undefined \write\@auxfile{#1}\fi}%
1944: %
1945: %
1946: % A macro package that uses btxmac.tex might define
1947: % \@undefinedmessage (before doing an \input btxmac).
1948: %
1949: \ifx\@undefinedmessage\@undefined
1950: \def\@undefinedmessage{No .aux file; I won't give you warnings about
1951: undefined citations.}%
1952: \fi
1953: %
1954: % Even if citations are undefined, we want to complain only if
1955: % \@citewarningtrue. The default is to set \@citewarningtrue unless
1956: % \noauxfile is defined. Again, a macro package that uses
1957: % btxmac.tex might want to redefine this.
1958: %
1959: \@innernewif\if@citewarning
1960: \ifx\noauxfile\@undefined \@citewarningtrue\fi
1961: %
1962: %
1963: % Finally, before leaving we restore @'s old category code.
1964: %
1965: \catcode`@ = \@oldatcatcode
1966:
1967:
1968:
1969: \def\pfeilso{\leavevmode
1970: \vrule width 1pt height9pt depth 0pt\relax
1971: \vrule width 1pt height8.7pt depth 0pt\relax
1972: \vrule width 1pt height8.3pt depth 0pt\relax
1973: \vrule width 1pt height8.0pt depth 0pt\relax
1974: \vrule width 1pt height7.7pt depth 0pt\relax
1975: \vrule width 1pt height7.3pt depth 0pt\relax
1976: \vrule width 1pt height7.0pt depth 0pt\relax
1977: \vrule width 1pt height6.7pt depth 0pt\relax
1978: \vrule width 1pt height6.3pt depth 0pt\relax
1979: \vrule width 1pt height6.0pt depth 0pt\relax
1980: \vrule width 1pt height5.7pt depth 0pt\relax
1981: \vrule width 1pt height5.3pt depth 0pt\relax
1982: \vrule width 1pt height5.0pt depth 0pt\relax
1983: \vrule width 1pt height4.7pt depth 0pt\relax
1984: \vrule width 1pt height4.3pt depth 0pt\relax
1985: \vrule width 1pt height4.0pt depth 0pt\relax
1986: \vrule width 1pt height3.7pt depth 0pt\relax
1987: \vrule width 1pt height3.3pt depth 0pt\relax
1988: \vrule width 1pt height3.0pt depth 0pt\relax
1989: \vrule width 1pt height2.7pt depth 0pt\relax
1990: \vrule width 1pt height2.3pt depth 0pt\relax
1991: \vrule width 1pt height2.0pt depth 0pt\relax
1992: \vrule width 1pt height1.7pt depth 0pt\relax
1993: \vrule width 1pt height1.3pt depth 0pt\relax
1994: \vrule width 1pt height1.0pt depth 0pt\relax
1995: \vrule width 1pt height0.7pt depth 0pt\relax
1996: \vrule width 1pt height0.3pt depth 0pt\relax}
1997:
1998: \def\pfeilnw{\leavevmode
1999: \vrule width 1pt height 9pt depth -9pt
2000: \vrule width 1pt height 9pt depth -8.7pt
2001: \vrule width 1pt height 9pt depth -8.3pt
2002: \vrule width 1pt height 9pt depth -8.0pt
2003: \vrule width 1pt height 9pt depth -7.7pt
2004: \vrule width 1pt height 9pt depth -7.3pt
2005: \vrule width 1pt height 9pt depth -7.0pt
2006: \vrule width 1pt height 9pt depth -6.7pt
2007: \vrule width 1pt height 9pt depth -6.3pt
2008: \vrule width 1pt height 9pt depth -6.0pt
2009: \vrule width 1pt height 9pt depth -5.7pt
2010: \vrule width 1pt height 9pt depth -5.3pt
2011: \vrule width 1pt height 9pt depth -5.0pt
2012: \vrule width 1pt height 9pt depth -4.7pt
2013: \vrule width 1pt height 9pt depth -4.3pt
2014: \vrule width 1pt height 9pt depth -4.0pt
2015: \vrule width 1pt height 9pt depth -3.7pt
2016: \vrule width 1pt height 9pt depth -3.3pt
2017: \vrule width 1pt height 9pt depth -3.0pt
2018: \vrule width 1pt height 9pt depth -2.7pt
2019: \vrule width 1pt height 9pt depth -2.3pt
2020: \vrule width 1pt height 9pt depth -2.0pt
2021: \vrule width 1pt height 9pt depth -1.7pt
2022: \vrule width 1pt height 9pt depth -1.3pt
2023: \vrule width 1pt height 9pt depth -1.0pt
2024: \vrule width 1pt height 9pt depth -0.7pt
2025: \vrule width 1pt height 9pt depth -0.3pt }
2026:
2027:
2028: \def\pfeilsw{ \leavevmode
2029: \vrule width 1pt height0.3pt depth 0pt\relax
2030: \vrule width 1pt height0.7pt depth 0pt\relax
2031: \vrule width 1pt height1.0pt depth 0pt\relax
2032: \vrule width 1pt height1.3pt depth 0pt\relax
2033: \vrule width 1pt height1.7pt depth 0pt\relax
2034: \vrule width 1pt height2.0pt depth 0pt\relax
2035: \vrule width 1pt height2.3pt depth 0pt\relax
2036: \vrule width 1pt height2.7pt depth 0pt\relax
2037: \vrule width 1pt height3.0pt depth 0pt\relax
2038: \vrule width 1pt height3.3pt depth 0pt\relax
2039: \vrule width 1pt height3.7pt depth 0pt\relax
2040: \vrule width 1pt height4.0pt depth 0pt\relax
2041: \vrule width 1pt height4.3pt depth 0pt\relax
2042: \vrule width 1pt height4.7pt depth 0pt\relax
2043: \vrule width 1pt height5.0pt depth 0pt\relax
2044: \vrule width 1pt height5.3pt depth 0pt\relax
2045: \vrule width 1pt height5.7pt depth 0pt\relax
2046: \vrule width 1pt height6.0pt depth 0pt\relax
2047: \vrule width 1pt height6.3pt depth 0pt\relax
2048: \vrule width 1pt height6.7pt depth 0pt\relax
2049: \vrule width 1pt height7.0pt depth 0pt\relax
2050: \vrule width 1pt height7.3pt depth 0pt\relax
2051: \vrule width 1pt height7.7pt depth 0pt\relax
2052: \vrule width 1pt height8.0pt depth 0pt\relax
2053: \vrule width 1pt height8.3pt depth 0pt\relax
2054: \vrule width 1pt height8.7pt depth 0pt\relax
2055: \vrule width 1pt height9pt depth 0pt\relax
2056: }
2057:
2058: \def\pfeilno{ \leavevmode
2059: \vrule width 1pt height 9pt depth -0.3pt
2060: \vrule width 1pt height 9pt depth -0.7pt
2061: \vrule width 1pt height 9pt depth -1.0pt
2062: \vrule width 1pt height 9pt depth -1.3pt
2063: \vrule width 1pt height 9pt depth -1.7pt
2064: \vrule width 1pt height 9pt depth -2.0pt
2065: \vrule width 1pt height 9pt depth -2.3pt
2066: \vrule width 1pt height 9pt depth -2.7pt
2067: \vrule width 1pt height 9pt depth -3.0pt
2068: \vrule width 1pt height 9pt depth -3.3pt
2069: \vrule width 1pt height 9pt depth -3.7pt
2070: \vrule width 1pt height 9pt depth -4.0pt
2071: \vrule width 1pt height 9pt depth -4.3pt
2072: \vrule width 1pt height 9pt depth -4.7pt
2073: \vrule width 1pt height 9pt depth -5.0pt
2074: \vrule width 1pt height 9pt depth -5.3pt
2075: \vrule width 1pt height 9pt depth -5.7pt
2076: \vrule width 1pt height 9pt depth -6.0pt
2077: \vrule width 1pt height 9pt depth -6.3pt
2078: \vrule width 1pt height 9pt depth -6.7pt
2079: \vrule width 1pt height 9pt depth -7.0pt
2080: \vrule width 1pt height 9pt depth -7.3pt
2081: \vrule width 1pt height 9pt depth -7.7pt
2082: \vrule width 1pt height 9pt depth -8.0pt
2083: \vrule width 1pt height 9pt depth -8.3pt
2084: \vrule width 1pt height 9pt depth -8.7pt
2085: \vrule width 1pt height 9pt depth -9pt
2086: }
2087:
2088:
2089: % *** end including bib4plain.tex ***
2090: % This will define \cite and make sure it works as in latex
2091:
2092:
2093: \def\widestnumber#1#2{}
2094: % Our amstex-ppt style does not know about \widestnumber
2095:
2096:
2097:
2098:
2099: \def\citewarning#1{\ifx\shlhetal\relax
2100: % normal mode, do not write anything
2101: \else
2102: % private mode
2103: \par{#1}\par
2104: \fi
2105: }
2106:
2107:
2108:
2109:
2110: \def\rm{\fam0 \tenrm}
2111:
2112:
2113: \def\fakesubhead#1\endsubhead{\bigskip\noindent{\bf#1}\par}
2114:
2115:
2116: % % \input rsfs
2117: % *** start including rsfs.tex ***
2118:
2119: % # Keywords: Script or Calligraphic (Caligraphic) letters with the RSFS Font
2120:
2121: % The story so far: July 1998 -- Saharon would like to have a
2122: % ``nicer'' calligraphic font. In particualr, the leters S and P in
2123: % the usual calligraphic font do not look ``special'' enough.
2124: %
2125: % I found out that ``rsfs'' (``Ralph Smith Formal Script'') may be
2126: % what he wants. I installed the mf file, the .tfm file, as well as
2127: % a few pk files in ~/TeX/rsfs. Let's hope that this is enough.
2128: % Using amstex, all you have to do is to \input rsfs.tex
2129: % Files prepared with citealice willdothis automatically.
2130: %
2131: % Note: for some reason xdvi calls MakeTeXpk, then Maketexpk
2132: % complains about wrong resolution, but still writes commands to
2133: % missfont.log...
2134: %
2135:
2136:
2137:
2138: % we redefine a macro inside amstex's \Cal command , so that it calls
2139: % our nice font ``rsfs'' rather than the usual calligraphic font.
2140: % Note thisworks for amstex only.
2141: % In plain tex, would have to add definitions of \Cal
2142: % in latex... we should insteaduse mathrsfs.sty
2143: %
2144:
2145:
2146:
2147: \font\textrsfs=rsfs10
2148: \font\scriptrsfs=rsfs7
2149: \font\scriptscriptrsfs=rsfs5
2150:
2151:
2152: \newfam\rsfsfam
2153: \textfont\rsfsfam=\textrsfs
2154: \scriptfont\rsfsfam=\scriptrsfs
2155: \scriptscriptfont\rsfsfam=\scriptscriptrsfs
2156:
2157:
2158: \edef\oldcatcodeofat{\the\catcode`\@}
2159: \catcode`\@11
2160:
2161: \def\Cal@@#1{\noaccents@ \fam \rsfsfam #1}
2162:
2163: \catcode`\@\oldcatcodeofat
2164:
2165: % *** end including rsfs.tex ***
2166:
2167: \expandafter\ifx \csname margininit\endcsname \relax\else\margininit\fi
2168:
2169:
2170:
2171: \long\def\red#1\endred{}
2172: \long\def\green#1\endgreen{}
2173: \long\def\blue#1\endblue{}
2174: \long\def\private#1\endprivate{}
2175:
2176: \def\endred{ \unmatched endred! }
2177: \def\endgreen{ \unmatched endgreen! }
2178: \def\endblue{ \unmatched endblue! }
2179: \def\endprivate{ \unmatched endprivate! }
2180:
2181: \ifx\latexcolors\undefinedcs\def\latexcolors{}\fi
2182:
2183: \def\emptycs{}
2184: \def\evaluatelatexcolors{%
2185: \ifx\latexcolors\emptycs\else
2186: \expandafter\xxevaluate\latexcolors\xxfertig\evaluatelatexcolors\fi}
2187: \def\xxevaluate#1,#2\xxfertig{\setupthiscolor{#1}%
2188: \def\latexcolors{#2}}
2189:
2190: \def\definedred{rgb 0.5 0 0}
2191: \def\definedgreen{rgb 0 0.5 0}
2192: \def\definedblue{rgb 0 0 0.5}
2193: % \def\definedprivate{rgb 0.5 0 0.5}
2194: \def\definedprivate{rgb 0 0 0 } % black
2195:
2196: \font\smallfont=cmsl7
2197: \def\rutgerscolor{\ifmmode\else\endgraf\fi\smallfont% \vrule width 3cm height 1cm depth 0cm
2198: \advance\leftskip0.5cm\relax}
2199: \def\setupthiscolor#1{\edef\tmptmpcs{\noexpand\bgroup\noexpand\rutgerscolor
2200: \noexpand\def\noexpand\currentcolor{#1}%
2201: \noexpand\special{color push \csname defined#1\endcsname}}%
2202: \expandafter\let\csname#1\endcsname\tmptmpcs
2203: \def\tmptmpcs{\checkColorUnmatched{#1}\popthecolor}
2204: \expandafter\let\csname end#1\endcsname\tmptmpcs}
2205:
2206: \def\checkColorUnmatched#1{\def\expectcolor{#1}%
2207: \ifx\expectcolor\currentcolor % OK!
2208: \else \edef\failhere{\noexpand\tryingToClose '\currentcolor' with end\expectcolor}\failhere\fi}
2209:
2210: \def\currentcolor{???}
2211:
2212:
2213:
2214: \def\popthecolor{\special{color pop}\ifmmode\else\endgraf\fi\egroup}
2215:
2216: \expandafter\def\csname#1\endcsname{\special{color push}}
2217:
2218: \evaluatelatexcolors
2219:
2220:
2221: \let\outerhead\head
2222: \def\head{\innerhead}
2223: \let\innerhead\outerhead
2224:
2225: \let\outersubhead\subhead
2226: \def\subhead{\innersubhead}
2227: \let\innersubhead\outersubhead
2228:
2229: \let\outersubsubhead\subsubhead
2230: \def\subsubhead{\innersubsubhead}
2231: \let\innersubsubhead\outersubsubhead
2232:
2233:
2234:
2235: \let\outerproclaim\proclaim
2236: \def\proclaim{\innerproclaim}
2237: \let\innerproclaim\outerproclaim
2238:
2239:
2240:
2241: %
2242: % \newdimen\wzero
2243: % \newdimen\hzero
2244: % \newdimen\dzero
2245: %
2246: % \def\whiteblack#1{\ifx \shlhetal\relax
2247: % \hbox{#1}% normal mode
2248: % \else\ifx\shlhetal\undefinedcontrolsequence
2249: % \hbox{#1}% normal mode
2250: % \else
2251: % % private mode
2252: % \setbox0=\hbox{#1}\leavevmode
2253: % \wzero=\wd0\hzero=\ht0\dzero=\dp0
2254: % \advance\wzero1mm
2255: % \advance\dzero1mm
2256: % \advance\hzero1mm
2257: % \hbox{\vrule width \wzero height \hzero depth\dzero\hskip-\wzero
2258: % \special{ps: currentgray 1 setgray}%
2259: % \hbox to \wzero{\hss\copy0\hss}%
2260: % \special{ps:setgray}}\fi\fi
2261: % }
2262: %
2263: %
2264:
2265: \def\demo#1{\medskip\noindent{\it #1.\/}}
2266: \def\enddemo{\smallskip}
2267:
2268: \def\remark#1{\medskip\noindent{\it #1.\/}}
2269: \def\endremark{\smallskip}
2270:
2271: \def\beginaside{\endgraf\leftskip2cm \vrule width 0pt\relax}
2272: \def\endaside{\endgraf\leftskip0cm \vrule width 0pt\relax}
2273:
2274: % *** end including alice2jlem.tex ***
2275: \pageheight{8.5truein}
2276: \topmatter
2277: \title {Categoricity in abstract elementary classes: going up
2278: inductive step} \endtitle
2279: \rightheadtext{Categoricity in abstract elementary classes}
2280: \author {Saharon Shelah \thanks{\null\newline
2281: I thank Alice Leonhardt for the beautiful typing \null\newline
2282: This research was supported by the United States-Israel Binational
2283: Science Foundation and in its final stages also by the Israel Science
2284: Foundation (Grant no. 242/03). Publication 600. } \endthanks} \endauthor
2285:
2286: \affil{The Hebrew University of Jerusalem \\
2287: Einstein Institute of Mathematics \\
2288: Edmond J. Safra Campus, Givat Ram
2289: Jerusalem 91904, Israel
2290: \medskip
2291: Department of Mathematics \\
2292: Hill Center-Busch Campus \\
2293: Rutgers, The State University of New Jersey\\
2294: 110 Frelinghuysen Road
2295: Piscataway, NJ 08854-8019 USA} \endaffil
2296:
2297: \keywords model theory,
2298: abstract elementary classes, classification theory,
2299: categoricity, non-structure theory \endkeywords
2300: \subjclass 03C45, 03C75, 03C95, 03C50 \endsubjclass
2301: \mn
2302:
2303: \abstract We deal with beginning stability theory for ``reasonable"
2304: non-elementary classes without
2305: any remnants of compactness like dealing with models
2306: above Hanf number or by the class being definable by
2307: $\Bbb L_{\omega_1,\omega}$. \endabstract
2308: \endtopmatter
2309: \document
2310:
2311: \pretags{300a,300b,300c,300d,300e,300f,300g,300x,300y,300z,88r,705,E53,734}
2312: \newpage
2313:
2314: \head {\S0 Introduction} \endhead \resetall \sectno=0
2315: \spuriousreset
2316: \bn
2317: The paper's main explicit result is proving Theorem \scite{600-0.A} below.
2318: It is done axiomatically, in a ``superstable" abstract framework with the
2319: set of ``axioms" of the frame, verified by applying earlier works, so
2320: it suggests this frame as the, or at least a major,
2321: non-elementary parallel of superstable. \nl
2322: A major case to which this is applied, is the one from
2323: \cite{Sh:576}; we continue this work in several ways but
2324: the use of \cite{Sh:576}
2325: is only in verifying the basic framework; we refer the reader
2326: to the book's introduction or
2327: \cite[\S0]{Sh:576} for background and some further claims but all
2328: the definitions and basic properties appear here.
2329: Otherwise, the heavy use of earlier works is in
2330: proving that our abstract framework applies in those contexts.
2331: If $\lambda = \aleph_0$ is O.K. for you, you may use \chaptercite{88r} or
2332: \cite{Sh:48} instead of \cite{Sh:576}.
2333:
2334: Naturally, our deeper aim is to develop stability theory (actually a parallel
2335: of the theory of superstable elementary classes) for non-elementary
2336: classes. We use the number of non-isomorphic models as test problem.
2337: Our main conclusion is \scite{600-0.A} below.
2338: As a concession to supposedly general opinion, we restrict ourselves here
2339: to the $\lambda$-good framework and delay dealing with weak relatives
2340: (see \cite{Sh:849}, \cite{Sh:838}, Jarden-Shelah \cite{JrSh:875},
2341: \cite{Sh:E46}; in
2342: the first we start from \cite[\S1-\S5]{Sh:576}, i.e., before proving that
2343: $K^{3,\text{uq}}_\lambda$ is dense).
2344: Also, we assume the (normal) weak-diamond ideal on the $\lambda^{+
2345: \ell}$ is not saturated (for $\ell=1,\dotsc,n-1$) and delay dealing with the
2346: elimination of this extra assumption (to \cite{Sh:838} there we waive
2347: the ``not $\lambda^{+\ell+1}$-saturated$^1$ (ideal)", the price is
2348: that we replace most $< 2^{\lambda^{+\ell+1}}$ by $<
2349: \mu_{\text{unif}}(\lambda^{+\ell+1},2^{\lambda^{+ \ell}})$, see on it
2350: \marginbf{!!}{\cprefix{88r}.\scite{88r-0.wD}}(3)).
2351: \bigskip
2352:
2353: \proclaim{\stag{600-0.A} Theorem} Assume $2^\lambda < 2^{\lambda^{+1}} < \dots
2354: < 2^{\lambda^{+n+1}}$ and the (so called weak diamond)
2355: normal \footnote{recall that as $2^{\lambda_{\ell -1}} <
2356: 2^{\lambda_\ell}$ this ideal is not trivial, i.e., $\lambda^{+ \ell}$
2357: is not in the ideal}
2358: ideal ${\text{\rm WDmId\/}}(\lambda^{+ \ell})$ is
2359: not $\lambda^{+ \ell +1}$-saturated
2360: \footnote{actually the statement ``some normal ideal on $\mu^+$ is
2361: $\mu^{++}$-saturated" is ``expensive", i.e., of large consistency
2362: strength, etc., so it is ``hard" for this assumption to fail}
2363: for $\ell = 1,\dotsc,n$. \nl
2364: 1) Let ${\frak K}$ be an abstract elementary class (see \S1 below)
2365: categorical in $\lambda$ and $\lambda^+$ with
2366: ${\text{\rm LS\/}}({\frak K}) \le \lambda$ (e.g. the class of models of
2367: $\psi \in
2368: \Bbb L_{\lambda^+,\omega}$ with $\le_{\frak K}$ defined naturally).
2369: If $1 \le \dot I(\lambda^{+2},{\frak K})$
2370: and $2 \le \ell \le n \Rightarrow \dot I(\lambda^{+ \ell},
2371: {\frak K}) < 2^{\lambda^{+ \ell}}$, \ub{then}
2372: ${\frak K}$ has a model of cardinality $\lambda^{+n+1}$. \nl
2373: 2) Assume $\lambda = \aleph_0$, and $\psi \in
2374: \Bbb L_{\omega_1,\omega}(\bold Q)$.
2375: If $1 \le \dot I(\lambda^{+ \ell},\psi) < 2^{\lambda^{+ \ell}}$
2376: for $\ell = 1, \dotsc,n-1$ \ub{then} $\psi$ has a model in
2377: $\lambda^{+ n}$ (see \cite{Sh:48}).
2378: \endproclaim
2379: \bn
2380: Note that if $n = 3$, then \scite{600-0.A}(1) is already
2381: proved in \cite{Sh:576}. If ${\frak K}$ is the class
2382: of models of some $\psi \in \Bbb L_{\omega_1,\omega}$ this is proved in
2383: \cite{Sh:87a}, \cite{Sh:87b}, but the proof here \ub{does not generalize} the
2384: proofs there. It is a different one (of course, they are related).
2385: There, for proving the theorem for $n$, we have to consider
2386: a few statements on $(\aleph_m,{\Cal P}^-(n-m))$-systems for all $m \le n$,
2387: (going up and down).
2388: A major point (there) is
2389: that for $n=0$, as $\lambda = \aleph_0$ we have the omitting
2390: type theorem and the types are ``classical", that is, are
2391: sets of formulas. This helps in
2392: proving strong dichotomies; so the analysis of what occurs in $\lambda^{+n}
2393: = \aleph_n$ is helped by those dichotomies. Whereas here we deal with
2394: $\lambda,\lambda^+,\lambda^{+2},
2395: \lambda^{+3}$ and then ``forget" $\lambda$ and deal with $\lambda^+,
2396: \lambda^{+2},\lambda^{+3},\lambda^{+4}$, etc. However, there are
2397: some further theorems proved in \cite{Sh:87a}, \cite{Sh:87b},
2398: whose parallels are not proved
2399: here, mainly that if for every $n$, in $\lambda^{+n}$ we get the
2400: ``structure" side, then the class has models in every $\mu \ge
2401: \lambda$, and theorems about categoricity.
2402: We shall deal with them in subsequent works, mainly \chaptercite{705}.
2403: Also in \cite{Sh:48}, \cite{Sh:88} = \chaptercite{88r}
2404: we started to deal with $\psi \in
2405: \Bbb L_{\omega_1,\omega}(\bold Q)$
2406: dealing with $\aleph_1,\aleph_2$; we put it in our present
2407: framework. Of course, also the framework
2408: of \chaptercite{88r} is integrated into our present context.
2409: In the axiomatic framework (introduced in \S2) we are able to
2410: present a lemma, speaking
2411: on only 4 cardinals, and which implies the theorem \scite{600-0.A}. (Why?
2412: Because in \S3 by \cite{Sh:576} we can get a so-called good
2413: $\lambda^+$-frame ${\frak s}$ with $K^{\frak s} \subseteq {\frak K}$,
2414: and then we prove a similar theorem on good frames by induction on
2415: $n$, with the induction step done by the lemma mentioned above).
2416: For this, parts of the proof are a generalization of the proof
2417: of \cite[\S8,\S9,\S10]{Sh:576}.
2418: \bn
2419: A major theme here (and even more so in \chaptercite{705}) is \nl
2420: \margintag{600-0.B}\ub{\stag{600-0.B} Thesis}: It is worthwhile to develop model theory (and
2421: superstability in particular) in the context of ${\frak K}_\lambda$ or
2422: $K_{\lambda^{+ \ell}},\ell \in \{0,\dotsc,n\}$, i.e., restrict
2423: ourselves to one, few, an interval of cardinals. We may have
2424: good understanding of the class in this context, while in general
2425: cardinals we are lost.
2426: \bn
2427: As in \cite{Sh:c} for first order classes \nl
2428: \margintag{600-0.C}\ub{\stag{600-0.C} Thesis}: It is reasonable first to develop the theory for the
2429: class of (quite) saturated enough models as it is smoother and even if you
2430: prefer to investigate the non-restricted case, the saturated case will
2431: clarify it. In our case this will mean investigating ${\frak s}^{+n}$
2432: for each $n$ and then $\cap\{{\frak K}^{{\frak s}^{+n}}:n < \omega\}$.
2433: \bn
2434: \margintag{600-0.D}\ub{\stag{600-0.D} The Better to be poor Thesis}: Better to know what
2435: is essential. e.g., you may have better closure properties (here a
2436: major point of poverty is having no formulas, this is even more
2437: noticeable in \chaptercite{705}).
2438: \sn
2439: I thank John Baldwin, Alex Usvyatsov, Andres Villaveces and Adi Yarden
2440: for many complaints and corrections.
2441:
2442: \S1 gives a self-contained introduction to a.e.c. (abstract elementary
2443: classes), including definitions of types, $M_2$ is
2444: $(\lambda,\kappa)$-brimmed over $M_1$ and saturativity = universality
2445: + model homogeneity. An interesting point is observing that any
2446: $\lambda$-a.e.c. ${\frak K}_\lambda$ can be lifted to ${\frak K}_{\ge
2447: \lambda}$, uniquely; so it does not matter if we deal with ${\frak
2448: K}_\lambda$ or ${\frak K}_{\ge \lambda}$ (unlike the situation for
2449: good $\lambda$-frames, which if we lift, we in general, lose some
2450: essential properties).
2451:
2452: The good $\lambda$-frames introduced in \S2 are a very central notion
2453: here. It concentrates on one cardinal $\lambda$, in ${\frak K}_\lambda$
2454: we have amalgamation and more, hence types, in the orbital,
2455: not in the classical sense of set of formulas, for models of cardinality
2456: $\lambda$ can be reasonably defined (we concentrate on so-called basic
2457: types) and we axiomatically have a non-forking relation for them.
2458:
2459: In \S3 we show that starting with classes belonging to reasonably large
2460: families, from assumptions on categoricity (or few models), good
2461: $\lambda$-frames arise. In \S4 we deduce some things on good
2462: $\lambda$-frames; mainly: stability in $\lambda$, existence and (full)
2463: uniqueness of $(\lambda,*)$-brimmed extensions of $M \in K_\lambda$.
2464:
2465: Concerning \S5 we know that if $M \in K_\lambda$ and $p \in {\Cal
2466: S}^{\text{bs}}(M)$ then there is $(M,N,a) \in K^{3,\text{bs}}_\lambda$
2467: such that \ortp$(a,M,N)=p$. But can we find a special (``minimal" or
2468: ``prime") triple in some sense? Note that if $(M_1,N_1,a)
2469: \le_{\text{bs}} (M_2,N_2,a)$ then $N_2$ is an amalgamation of
2470: $N_1,M_2$ over $M_1$ (restricting ourselves to the case
2471: ``\ortp$(a,M_2,N_2)$ does not fork over $M_1$") and we may wonder is this
2472: amalgamation unique (i.e., allowing to increase or decrease $N_2$).
2473: If this holds for any such $(M_2,N_2,a)$ we say $(M_1,N_1,a)$ has
2474: uniqueness (= belongs to $K^{3,\text{uq}}_\lambda =
2475: K^{3,\text{uq}}_{\frak s}$). Specifically we ask: is
2476: $K^{3,\text{uq}}_\lambda$ dense in
2477: $(K^{3,\text{bs}}_\lambda,\le_{\text{bs}})$? If no, we get a
2478: non-structure result; if yes, we shall (assuming categoricity) deduce
2479: the ``existence for $K^{3,\text{uq}}_{\frak s}$" and this is used
2480: later as a building block for non-forking amalgamation of models.
2481:
2482: So our next aim is to find ``non-forking" amalgamation of models (in
2483: \S6). We first note that there is at most one such notion which
2484: fulfills our expectations (and ``respect" ${\frak s}$). Now if
2485: $\nonfork{}{}_{}(M_0,M_1,a,M_3),M_0 \le_{\frak K} M_2 \le_{\frak K}
2486: M_3$ and $(M_0,M_2,a) \in K^{3,\text{uq}}_\lambda$ by our demands we
2487: have to say that $M_1,M_2$ are in non-forking amalgamation over $M_0$
2488: inside $M_3$. Closing this family under the closure demands we expect
2489: to arrive to a notion NF$_\lambda = \text{\rm NF}_{\frak s}$
2490: which should be the
2491: right one (if a solution exists at all). But then we have to work on
2492: proving that it has all the properties it hopefully has.
2493:
2494: A major aim in advancing to $\lambda^+$ is having a superlimit model.
2495: So in \S7 we find out who he should be: the saturated model of ${\frak
2496: K}_{\lambda^+}$, but is it superlimit? We use our NF$_\lambda$ to
2497: define a ``nice" order $\le^*_{\lambda^+}$ on ${\frak K}_{\lambda^+}$,
2498: investigate it and prove the existence of a superlimit model under
2499: this partial order. To
2500: complete the move to $\lambda^+$ we would like to have that the class
2501: of $\lambda^+$-saturated model with the partial order $\le^*_{\lambda^+}$ is a
2502: $\lambda^+$-a.e.c. Well, we do not prove it but rather use it as a
2503: dividing line: if it fails we eventually get many models in ${\frak
2504: K}_{\lambda^{++}}$ (coding a stationary subset of $\lambda^{++}$
2505: (really any $S \subseteq \{\delta < \lambda^{++}:\text{cf}(\delta) =
2506: \lambda^+\}$)), see \S8.
2507:
2508: Lastly, we pay our debts: prove the theorems which were the motivation
2509: of this work, in \S9.
2510: \bn
2511: \centerline {$* \qquad * \qquad *$}
2512: \bn
2513: \centerline {Reading Instructions}
2514: \bn
2515: As usual these are instructions on what you can avoid reading.
2516:
2517: Note that \S3 contains the examples, i.e., it shows how ``good
2518: $\lambda$-frame", our main object of study here, arise in previous
2519: works. This, on the one hand, may help the reader to understand what
2520: is a good frame and, on the other hand, helps us in the end to draw
2521: conclusions continuing those works. However, it is \ub{not} necessary
2522: here otherwise, so you may ignore it.
2523:
2524: Note that we treat the subject axiomatically, in a
2525: general enough way to treat the
2526: cases which exist without trying too much to eliminate axioms as long as the
2527: cases are covered (and probably most potential readers will feel they are
2528: more than general enough). \nl
2529: We shall assume
2530: \mr
2531: \item "{$(*)_0$}" $2^\lambda < 2^{\lambda^+} < 2^{\lambda^{+2}} < \ldots
2532: < 2^{\lambda^{+n}}$ and $n \ge 2$.
2533: \ermn
2534: In the end of \S1 there are some basic definitions.
2535: \bn
2536: \ub{Reading Plan 0}: We accept the good frames as interesting per se
2537: so ignore \S3 (which gives ``examples") and: \S1 tells you all you
2538: need to know on abstract elementary classes; \S2 presents frames, etc.
2539: \bn
2540: \ub{Reading Plan 1}: The reader decides to understand why we reprove
2541: the main theorem of \cite{Sh:87a}, \cite{Sh:87b} so
2542: \mr
2543: \item "{$(*)_1$}" $K$ is the class of models of some $\psi \in
2544: \Bbb L_{\lambda^+,\omega}$ (with a natural notion of elementary embedding
2545: $\prec_{\Cal L}$ for ${\Cal L}$ a fragment of $\Bbb L_{\lambda^+,\omega}$
2546: of cardinality $\le \lambda$ to which $\psi$ belongs).
2547: \ermn
2548: So in fact (as we can replace, for this result, $K$ by any class with
2549: fewer models still satisfying the assumptions) \wilog \,
2550: \mr
2551: \item "{$(*)'_1$}" if $\lambda = \aleph_0$ then $K$
2552: is the class of atomic models of some complete first order theory,
2553: $\le_{\frak K}$ is being elementary submodel.
2554: \ermn
2555: The theorems we are seeking are of the form
2556: \mr
2557: \item "{$(*)_2$}" if $K$ has few models in $\lambda + \aleph_1,\lambda^+,
2558: \dotsc,\lambda^{+n}$ then it has a model in $\lambda^{+n+1}$. \nl
2559: [Why ``$\lambda + \aleph_1$"? If $\lambda > \aleph_0$ this means
2560: $\lambda$ whereas if $\lambda = \aleph_0$ this means that we do not
2561: require ``few model in $\lambda = \aleph_0$". The reason is that for
2562: the class or models of $\psi \in \Bbb L_{\omega_1,\omega}$ (or $\in
2563: \Bbb L_{\omega_1,\omega}(\bold Q)$ or an a.e.c. which is
2564: PC$_{\aleph_0}$, see Definition \scite{600-0.4}) we
2565: have considerable knowledge of general methods of
2566: building models of cardinality $\aleph_1$, for general $\lambda$ we
2567: are very poor in such knowedge (probably as there is much less).]
2568: \ermn
2569: But, of course, what we would really like to have are rudiments of stability
2570: theory (non-forking amalgamation, superlimit models, etc.). Now
2571: reading plan 1 is to follow reading plan 2 below \ub{but}
2572: replacing the use of Claim \scite{600-Ex.4} and \cite{Sh:576} by the
2573: use of a simplified version of \scite{600-Ex.1} and \cite{Sh:87a}.
2574: \bn
2575: \ub{Reading Plan 2}: The reader wants to understand the proof of
2576: $(*)_2$ for arbitrary ${\frak K}$ and $\lambda$.
2577: The reader
2578: \mr
2579: \item "{$(a)$}" knows at least the main definitions and results
2580: of \cite{Sh:576},
2581: \nl
2582: or just
2583: \sn
2584: \item "{$(b)$}" reads the main definitions of \S1 here (in
2585: \scite{600-0.1} - \scite{600-0.6}) and is
2586: willing to believe some quotations of results of \cite{Sh:576}.
2587: \ermn
2588: We start assuming
2589: ${\frak K}$ is an abstract elementary class, LS$({\frak K}) \le
2590: \lambda$ (or read \S1 here until \scite{600-0.22}) and
2591: ${\frak K}$ is categorical in $\lambda,
2592: \lambda^+$ and $1 \le \dot I(\lambda^{++},K)
2593: < 2^{\lambda^{++}}$ and moreover, $1 \le \dot I(\lambda^{++},K) <
2594: 2^{\lambda^{++}}$. As an appetizer and to understand types and the
2595: definition of types and saturated (in the present context) and
2596: brimmed, read from \S1 until \scite{600-4a.10}.
2597: \nl
2598: He should read in \S2 Definition \scite{600-1.1} of $\lambda$-good frame, an
2599: axiomatic framework and then read the following two Definitions
2600: \scite{600-1.6}, \scite{600-1.7} and Claim \scite{600-1.8}. In \S3, \scite{600-Ex.4}
2601: show how by \cite{Sh:576} the context there gives a $\lambda^+$-good frame;
2602: of course the reader may just believe instead of reading proofs, and
2603: he may remember that our basic types are minimal in this case. \nl
2604: In \S4 he should read some consequences of the axioms proved with weaker
2605: axioms, understanding here and later $<^*$ as $\le_{\frak K} \restriction
2606: K_\lambda$. \nl
2607: Then in \S5 we show some
2608: amount of unique amalgamation. Then \S6,\S7,\S8 do a parallel to
2609: \cite[\S8,\S9,\S10]{Sh:576} in our context; still there are
2610: differences, in particular our context is not necessarily
2611: uni-dimensional which complicates matters. But if we restrict ourselves to
2612: continuing \cite{Sh:576}, our frame is ``uni-dimensional", we could have
2613: simplified the proofs
2614: by using ${\Cal S}^{\text{bs}}(M)$ as the set of minimal types.
2615: \bn
2616: \ub{Reading Plan 3}: $\psi \in \Bbb L_{\omega_1,\omega}(\bold Q)$ so
2617: $\lambda= \aleph_0,1 \le \dot I(\aleph_1,\psi) < 2^{\aleph_1}$.
2618:
2619: For this,
2620: \cite{Sh:576} is irrelevant (except quoting the ``\ub{black box}" use of
2621: the combinatorial section \S3
2622: of \cite{Sh:576} when using the weak diamond to get many non-isomorphic
2623: models in \S5).
2624:
2625: Now reading plan 3 is to follow reading plan 2 but
2626: \scite{600-Ex.4} is replaced by \scite{600-Ex.1A} which relies on
2627: \cite{Sh:48}, i.e., it proves that we get an $\aleph_1$-good frame
2628: investigating $\psi \in \Bbb L_{\omega_1,\omega}(\bold Q)$.
2629:
2630: Note that our class may well be such that ${\frak K}$ is the parallel
2631: of ``superstable non-multidimension complete first order theory";
2632: e.g., $\psi_1 = (\bold Q x)(P(x)) \wedge(\bold Q x)(\neg P(x)),
2633: \tau_\psi = \{P\},P$ a unary predicate; this is
2634: categorical in $\aleph_1$ and has no models in $\aleph_0$ and $\psi_1$
2635: has 3 models in $\aleph_2$. But if we
2636: use $\psi_0 = (\forall x)(P(x) \equiv P(x))$ we have
2637: $\dot I(\aleph_1,\psi_0) = \aleph_0$; however, even starting with $\psi_1$, the
2638: derived a.e.c. ${\frak K}$ has exactly three non-isomorphic models in
2639: $\aleph_1$. In general we derived an a.e.c. ${\frak K}$ from $\psi$
2640: such that: ${\frak K}$ is an a.e.c. with LS number $\aleph_0$,
2641: categorical in $\aleph_0$, and the number of somewhat ``saturated"
2642: models of ${\frak K}$ in $\lambda$ is $\le \dot I(\lambda,\psi)$ for
2643: $\lambda \ge \aleph_1$. The relationship of $\psi$ and ${\frak K}$ is
2644: not comfortable; as it means that, for general results to be applied,
2645: they have to be somewhat stronger, e.g. ``${\frak K}$ has
2646: $2^{\lambda^{++}}$ non-isomorphic \ub{$\lambda^+$-saturated} models of
2647: cardinality $\lambda^{++}$". The reason is
2648: LS$({\frak K}) = \lambda = \aleph_0$; we have to find many
2649: somewhat $\lambda^+$-saturated models as we have first in a sense
2650: eliminate the quantifier $\bold Q =
2651: \exists^{\ge \aleph_1}$, (i.e., the choice of
2652: the class of models and of the order guaranteed that what has to be
2653: countable is countable, and $\lambda^+$-saturation guarantees that what
2654: should be uncountable is uncountable).
2655: \bn
2656: \ub{Reading Plan 4}: ${\frak K}$ an abstract elementary class which is
2657: PC$_\omega$ ($= \aleph_0$-presentable, see Definition \scite{600-0.4});
2658: see \chaptercite{88r} or \cite{Mw85a} which includes a
2659: friendly presentation of \cite[\S1-\S3]{Sh:88} so of \sectioncite[\S1-\S3]{88r}).
2660:
2661: Like plan 3 but we have to use \scite{600-Ex.1} instead of
2662: \scite{600-Ex.1A} and fortunately
2663: the reader is encouraged to read \sectioncite[\S4,\S5]{88r} to understand why we
2664: get a $\lambda$-good quadruple.
2665: \newpage
2666:
2667: \head {\S1 Abstract elementary classes} \endhead \resetall \sectno=1
2668: \spuriousreset
2669: \bn
2670:
2671: First we present the basic material on a.e.c. ${\frak K}$, so types,
2672: saturativity and brimmness (so most is repeating some things from
2673: \sectioncite[\S1]{88r} and from \chaptercite{300b}).
2674: \nl
2675: Second we show that the situation in $\lambda = \text{ LS}({\frak K})$
2676: determine the situation above $\lambda$, moreover such lifting always
2677: exists; so a $\lambda$-a.e.c. can be lifted to a $(\ge
2678: \lambda)$-a.e.c. in one and only one way.
2679: \bigskip
2680:
2681: \demo{\stag{600-0.1} Conventions} Here ${\frak K} = (K,\le_{\frak K})$, where
2682: $K$ is a class of $\tau$-models for a fixed vocabulary $\tau = \tau_K =
2683: \tau_{\frak K}$ and $\le_{\frak K}$ is a two-place relation on the models in
2684: $K$. We do not always
2685: strictly distinguish between
2686: ${\frak K},K$ and $(K,\le_{\frak K})$. We shall
2687: assume
2688: that $K,\le_{\frak K}$ are fixed, and $M \le_{\frak K} N \Rightarrow
2689: M,N \in K$; and we assume that it is an abstract elementary class, see
2690: Definition \scite{600-0.2} below. When we use $<_{\frak K}$ in the $\prec$
2691: sense (elementary submodel for first order logic), we write
2692: $\prec_{\Bbb L}$.
2693: \enddemo
2694: \bigskip
2695:
2696: \definition{\stag{600-0.1A} Definition} For a class of $\tau_K$-models we
2697: let $\dot I(\lambda,K) = |\{M/\cong:M \in K,
2698: \|M\| = \lambda\}|$.
2699: \enddefinition
2700: \bigskip
2701:
2702: \definition{\stag{600-0.1B} Definition} 1) We say $\bar M = \langle M_i:i <
2703: \mu \rangle$ is a representation or filtration of a model $M$ of
2704: cardinality $\mu$ \ub{if} $\tau_{M_i} = \tau_M,M_i$
2705: is $\subseteq$-increasing continuous, $\|M_i\| < \|M\|$ and $M =
2706: \cup\{M_i:i < \mu\}$ and
2707: $\mu = \chi^+ \Rightarrow \|M_i\| = \chi$. \nl
2708: 2) We say $\bar M$ is a $\le_{\frak K}$-representation or
2709: $\le_{\frak K}$-filtration
2710: \ub{if} in addition $M_i \le_{\frak K} M$ for $i < \|M\|$
2711: (hence $M_i,M \in K$ and $\langle M_i:i < \mu \rangle$ is $\le_{\frak
2712: K}$-increasing continuous, by Av V from Definition \scite{600-0.2}).
2713: \enddefinition
2714: \bigskip
2715:
2716: \definition{\stag{600-0.2} Definition} We say ${\frak K} =
2717: (K,\le_{\frak K})$ is an
2718: abstract elementary class, a.e.c. in short, if
2719: ($\tau$ is as in \scite{600-0.1}, $Ax 0$ holds and) AxI-VI hold where: \nl
2720: $Ax 0$: The holding of
2721: $M \in K,N \le_{\frak K} M$ depends on $N,M$ only up to isomorphism, i.e.,
2722: $[M \in K,M \cong N \Rightarrow N \in K]$, and [if $N \le_{\frak K} M$ and
2723: $f$ is an isomorphism from $M$ onto the $\tau$-model $M'$ mapping $N$ onto
2724: $N'$ \underbar{then} $N' \le_{\frak K} M'$].
2725: \medskip
2726:
2727: $Ax I$: If $M \le_{\frak K} N$ then $M \subseteq N$ (i.e. $M$ is a submodel
2728: of $N$).
2729: \medskip
2730:
2731: $Ax II$: $M_0 \le_{\frak K} M_1 \le_{\frak K} M_2$ implies $M_0 \le_{\frak K}
2732: M_2$ and $M \le_{\frak K} M$ for $M \in K$.
2733: \medskip
2734:
2735: $Ax III$: If $\lambda$ is a regular cardinal, $M_i$ (for $i < \lambda$) is
2736: $\le_{\frak K}$-increasing (i.e. $i < j < \lambda$ implies $M_i \le_{\frak K}
2737: M_j$) and continuous (i.e. for limit ordinal $\delta < \lambda$ we have
2738: \newline
2739: $M_\delta = \dsize \bigcup_{i < \delta} M_i$) \underbar{then}
2740: $M_0 \le_{\frak K} \dsize \bigcup_{i < \lambda} M_i$.
2741: \medskip
2742:
2743: $Ax IV$: If $\lambda$ is a regular cardinal, $M_i$ (for $i < \lambda)$ is
2744: $\le_{\frak K}$-increasing continuous and
2745: $M_i \le_{\frak K} N$ for $i < \lambda$ \underbar{then}
2746: $\dsize \bigcup_{i < \lambda} M_i \le_{\frak K} N$.
2747: \medskip
2748:
2749: $Ax V$: If $M_0 \subseteq M_1$ and $M_\ell \le_{\frak K} N$ for $\ell =
2750: 0,1$, \ub{then} $M_0 \le_{\frak K} M_1$.
2751: \medskip
2752:
2753: $Ax VI$: LS$({\frak K})$ exists
2754: \footnote{We normally assume $M \in {\frak K}
2755: \Rightarrow \|M\| \ge \text{ LS}({\frak K})$ so may forget to write
2756: $\|M\| ``+\text{ LS}({\frak K})"$ instead $\|M\|$, here
2757: there is no loss in it. It is also
2758: natural to assume $|\tau({\frak K})| \le \text{ LS}
2759: ({\frak K})$ which means just
2760: increasing LS$({\frak K})$, but no real need here; dealing with Hanf
2761: numbers it is natural.}, where LS$({\frak K})$ is
2762: the minimal cardinal $\lambda$ such that: if \newline
2763: $A \subseteq N$ and $|A| \le \lambda$ \underbar{then} for some
2764: $M \le_{\frak K} N$ we have $A \subseteq |M|$ and $\|M\| \le \lambda$.
2765: \enddefinition
2766: \bn
2767: \margintag{600-0.2B}\underbar{\stag{600-0.2B} Notation}: 1) $K_\lambda =
2768: \{ M \in K:\|M\| = \lambda\}$ and
2769: $K_{< \lambda} = \dsize \bigcup_{\mu < \lambda} K_\mu$.
2770: \bigskip
2771:
2772: \definition{\stag{600-0.3} Definition} 1) The embedding $f:N \rightarrow M$
2773: is $\le_{\frak K}$-embedding when its range is
2774: the universe of a model $N' \le_{\frak K} M$,
2775: (so $f:N \rightarrow N'$ is an isomorphism onto).
2776: \nl
2777: 2) We say $f$ is a $\le_{\frak K}$-embedding of $M_1$ into $M_2$ over
2778: $M_0$ when for some $M'_1$ we have: $M_0 \le_{\frak K} M_1,M_0
2779: \le_{\frak K} M'_1 \le_{\frak K} M_2$ and $f$ is an isomorphism from
2780: $M_1$ onto $M'_1$ extending the mapping id$_{M_0}$.
2781: \enddefinition
2782: \bn
2783: Recall
2784: \demo{\stag{600-0.6} Observation} Let $I$ be a directed set (i.e., $I$ is
2785: partially ordered by $\le = \le^I$, such that
2786: any two elements have a common upper bound). \newline
2787: 1) If $M_t$ is defined for $t \in I$, and $t \le s \in I$ implies $M_t
2788: \le_{\frak K} M_s$ \underbar{then} for every $t \in I$ we have
2789: $M_t \le_{\frak K} \dsize \bigcup_{s \in I} M_s$. \newline
2790: 2) If in addition $t \in I$ implies $M_t \le_{\frak K} N$ \underbar{then}
2791: $\dsize \bigcup_{s \in I} M_s \le_{\frak K} N$.
2792: \enddemo
2793: \bigskip
2794:
2795: \demo{Proof} Easy or see \marginbf{!!}{\cprefix{88r}.\scite{88r-1.6}} which does not rely on
2796: anything else. \hfill$\square_{\scite{600-0.6}}$
2797: \enddemo
2798: \bigskip
2799:
2800: \proclaim{\stag{600-0.7} Claim} 1) For every $N \in K$ there is a directed
2801: partial order $I$ of cardinality $\le \|N\|$ and sequence
2802: $\bar M = \langle M_t:t \in
2803: I \rangle$ such that $t \in I \Rightarrow M_t \le_{\frak K} N,\|M_t\| \le
2804: { \text{\rm LS\/}}({\frak K}),
2805: I \models s < t \Rightarrow M_s \le_{\frak K} M_t$ and $N =
2806: \dbcu_{t \in I} M_t$. If $\|N\| \ge \text{\rm LS}({\frak K})$ we can add
2807: $\|M_t\| = \text{\rm LS}({\frak K})$ for $t \in I$.
2808: \nl
2809: 2) For every $N_1 \le_{\frak K} N_2$ we can find $\langle M^\ell_t:t \in
2810: I_\ell \rangle$ as in part (1) for $\ell=1,2$
2811: such that $I_1 \subseteq I_2$ and
2812: $t \in I_1 \Rightarrow M^2_t = M^1_t$. \nl
2813: 3) Any
2814: $\lambda \ge {\text{\rm LS\/}}({\frak K})$
2815: satisfies the requirement in the definition of ${\text{\rm LS\/}}
2816: ({\frak K})$.
2817: \endproclaim
2818: \bigskip
2819:
2820: \demo{Proof} Easy or see \marginbf{!!}{\cprefix{88r}.\scite{88r-1.7}} which does not require
2821: anything else. \hfill$\square_{\scite{600-0.7}}$
2822: \enddemo
2823: \bn
2824: We now (in \scite{600-0.12}) recall the (non-classical) definition of type
2825: (note that it is natural to look at types only over models which are
2826: amalgamation bases, see part (4) of \scite{600-0.12} below and
2827: consider only extensions of
2828: the models of the same cardinality).
2829: Note that though the choice of the name indicates that they are
2830: supposed to behave like complete types over models as in classical
2831: model theory (on which we are not relying), this does not
2832: guarantee most of the basic properties. E.g., when cf$(\delta) =
2833: \aleph_0$, uniqueness of $p_\delta \in {\Cal S}(M_\delta)$ such that
2834: $i < \delta \Rightarrow p_\delta \restriction M_i = p_i$ is not
2835: guaranteed even if $p_i \in {\Cal S}(M_i),M_i$ is
2836: $\le_{\frak K}$-increasing continuous for $i \le \delta$ and $i < j
2837: < \delta \Rightarrow p_i = p_j \restriction M_i$. Still
2838: we have existence: if for
2839: $i < \delta,p_i \in {\Cal S}(M_i)$
2840: increasing with $i$, then there is $p_\delta \in {\Cal S}(\cup \{M_i:i
2841: < \delta\})$ such that $i < \delta \Rightarrow p_i = p_\delta
2842: \restriction M_i$. But when cf$(\delta) > \aleph_0$ even existence
2843: is not guaranteed.
2844: \definition{\stag{600-0.12} Definition} 1) For $M \in K_\mu$ we
2845: define ${\Cal S}(M) = {\Cal S}_{\frak K}(M)$ as
2846: $\{\text{\ortp}(a,M,N):M \le_{\frak K} N \in K_\mu
2847: \text{ and } a \in N\}$ where
2848: $\text{\ortp}(a,M,N) = \text{ \ortp}_{\frak K}(a,M,N) =
2849: (M,N,a)/{\Cal E}_M$ where ${\Cal E}_M$ is the transitive closure
2850: of ${\Cal E}^{\text{at}}_M$, and the two-place relation
2851: ${\Cal E}^{\text{at}}_M$ is defined by:
2852:
2853: $$
2854: \align
2855: (M,N_1,a_1){\Cal E}^{\text{at}}_M &(M,N_2,a_2)
2856: \text{ \ub{iff} } M \le_{\frak K}
2857: N_\ell,a_\ell \in N_\ell,\|N_\ell\| = \mu = \|M\| \\
2858: &\text{ for } \ell=1,2 \text{ and } \text{ there is }
2859: N \in K_\mu \text{ and } \le_{\frak K} \text{-embeddings} \\
2860: &f_\ell:N_\ell \rightarrow N \text{ for } \ell = 1,2 \text{ such that:} \\
2861: &f_1 \restriction M = \text{ id}_M = f_2 \restriction M \text{ and }
2862: f_1(a_1) = f_2(a_2).
2863: \endalign
2864: $$
2865: \mn
2866: We may say $p = \text{ \ortp}(a,M,N)$ is the type which $a$ realizes over
2867: $M$ in $N$. Of course, all those notions depend on ${\frak K}$ so we
2868: may write \ortp$_{\frak K}(a,M,N)$ and ${\Cal E}_M[{\frak K}],
2869: {\Cal E}^{\text{at}}_M[{\frak K}]$.
2870: \nl
2871: (If in Definition \scite{600-0.2} we do not require $M \in K \Rightarrow
2872: \|M\| \ge \text{\rm LS}({\frak K})$, here we should allow any $N$ such
2873: that $\|M\| \le \|N\| \le M + \text{\rm LS}({\frak K})$.)
2874: \nl
2875: 1A) For $M \in {\frak K}_\mu$ let\footnote{we can insist that $N \in
2876: K_\mu$, the difference is not serious} ${\Cal S}_{\frak K}(M) =
2877: \{\text{\rm \ortp}(a,M,N):M \le_{\frak K} N$ and $N \in K_\mu$ or just
2878: $N \in K_{\le(\mu +\text{LS}({\frak K}))}$ and $a \in N\}$ and ${\Cal
2879: S}^{\text{na}}_{\frak K}(M) = \{\ortp(a,M,N):M \le_{\frak K} N$ and
2880: $N \in K_{\le(\mu + \text{LS}({\frak K}))}$ and $a \in N
2881: \backslash M\}$, na stands for non-algebraic. We may write ${\Cal
2882: S}^{\text{na}}(M)$ omitting ${\frak K}$ when ${\frak K}$ is clear from
2883: the context; so omitting na means $a \in N$ rather than $a \in N
2884: \backslash M$.
2885: \nl
2886: 2) Let $M \in K_\mu$ and $M \le_{\frak K} N$.
2887: We say ``$a$ realizes $p$ in $N$" and ``$p = \text{ \ortp}(a,M,N)$" when:
2888: if $a \in N,p \in {\Cal S}(M)$ and
2889: $N' \in K_{\le(\mu+\text{LS}({\frak K}))}$ satisfies
2890: $M \le_{\frak K} N' \le_{\frak K} N$ and $a \in N'$ then
2891: $p = \text{ \ortp}(a,M,N')$ and there is at least one such $N'$;
2892: so $M,N' \in K_\mu$ (or just
2893: $M \le \|N'\| \le \mu + \text{\rm LS}({\frak K}))$
2894: but possibly $N \notin K_\mu$.
2895: \nl
2896: 3) We say ``$a_2$ strongly \footnote{note that
2897: ${\Cal E}^{\text{at}}_M$ is not an
2898: equivalence relation and hence in general is not ${\Cal E}_M$}
2899: realizes $(M,N^1,a_1)/{\Cal E}^{\text{at}}_M$ in $N$" \ub{when}
2900: for some $N^2$ of cardinality $\le \|M\| + \text{\rm LS}({\frak K})$
2901: we have $M \le_{\frak K} N^2 \le_{\frak K} N$ and
2902: $a_2 \in N^2$ and $(M,N^1,a_1)\,{\Cal E}^{\text{at}}_M\,(M,N^2,a_2)$.
2903: \nl
2904: 4) We say $M_0 \in K_\lambda$ is an amalgamation base (in ${\frak K}$,
2905: but normally ${\frak K}$ is understood from the context) \ub{if}: for every
2906: $M_1,M_2 \in K_\lambda$ and $\le_{\frak K}$-embeddings
2907: $f_\ell:M_0 \rightarrow M_\ell$ (for $\ell = 1,2$) there is $M_3 \in
2908: K_\lambda$ and
2909: $\le_{\frak K}$-embeddings $g_\ell:M_\ell \rightarrow
2910: M_3$ (for $\ell=1,2$) such that $g_1 \circ f_1 = g_2 \circ f_2$.
2911: Similarly for ${\frak K}_{\le \lambda}$.
2912: \nl
2913: 4A) ${\frak K}$ has amalgamation in $\lambda$ (or
2914: $\lambda$-amalgamation or ${\frak K}_\lambda$ has amalgamation) \ub{when}
2915: every $M \in K_\lambda$ is an amalgamation base.
2916: \nl
2917: 4B) ${\frak K}$ has the $\lambda$-JEP or JEP$_\lambda$ (or ${\frak
2918: K}_\lambda$ has the JEP) when any $M_1,M_2 \in K_\lambda$ can be
2919: $\le_{\frak K}$-embedded into some $M \in K_\lambda$.
2920: \nl
2921: 5) We say ${\frak K}$ is stable in $\lambda$ \underbar{if} (LS$({\frak K})
2922: \le \lambda$ and) $M \in K_\lambda \Rightarrow |{\Cal S}(M)| \le \lambda$.
2923: \nl
2924: 6) We say $p=q \restriction M$ if $p \in {\Cal S}(M),q \in {\Cal S}(N),
2925: M \le_{\frak K} N$ and for some $N^+,N \le_{\frak K} N^+$ and $a \in N^+$ we
2926: have $p = \text{ \ortp}(a,M,N^+)$ and $q = \text{ \ortp}(a,N,N^+)$; see
2927: \scite{600-0.12A}(1),(2).
2928: We may express this also as ``$q$ extends $p$ or $p$ is
2929: the restriction of $q$ to $M$".
2930: \nl
2931: 7) For finite $m$, for $M \le_{\frak K} N,\bar a \in {}^m N$ we can define
2932: \ortp$(\bar a,M,N)$ and ${\Cal S}^m_{\frak K}(M)$
2933: similarly and ${\Cal S}^{< \omega}_{\frak K}(M) =
2934: \dbcu_{m < \omega} {\Cal S}^m_{\frak K}(M)$; similarly for
2935: ${\Cal S}^\alpha(M)$ (but we shall not use this in any
2936: essential way, so we agree ${\Cal S}(M) = {\Cal S}^1(M)$.) Again we
2937: may omit ${\frak K}$ when clear from the context.
2938: \nl
2939: 8) We say that $p \in {\Cal S}_{\frak K}(M)$ is algebraic when some $a
2940: \in M$ realizes it.
2941: \nl
2942: 9) We say that $p \in {\Cal S}_{\frak K}(M)$ is minimal \ub{when}
2943: it is not algebraic
2944: and for every $N \in K$ of cardinality $\le \|M\| + \text{ LS}({\frak K})$
2945: which $\le_{\frak K}$-extend $M$, the type $p$ has at most one
2946: non-algebraic extension in ${\Cal S}_{\frak K}(M)$.
2947: \enddefinition
2948: \bigskip
2949:
2950: \remark{\stag{600-0.12Y} Remark} 1) Note that here ``amalgamation base"
2951: means only for extensions of the same cardinality!
2952: \nl
2953: 2) The notion ``minimal type" is important (for categoricity)
2954: but not used much in this chapter.
2955: \endremark
2956: \bigskip
2957:
2958: \demo{\stag{600-0.12A} Observation}
2959: 0) Assume
2960: $M \in K_\mu$ and $M \le_{\frak K} N,a \in N$ \ub{then} \ortp$(a,M,N)$ is
2961: well defined and is $p$ \ub{if} for some $M' \in K_\mu$ we have $M
2962: \cup \{a\} \subseteq M' \le_{\frak K} N$ and
2963: $p = \text{ \ortp}(a,M,M')$.
2964: \nl
2965: 1) If $M \le_{\frak K} N_1 \le_{\frak K}
2966: N_2,M \in K_\mu$ and $a \in N_1$ \ub{then}
2967: \ortp$(a,M,N_1)$ is well defined and equal to \ortp$(a,M,N_2)$, (more
2968: transparent if ${\frak K}$ has the $\mu$-amalgamation which is the
2969: real case anyhow).
2970: \nl
2971: 2) If $M \le_{\frak K} N$ and $q \in {\Cal S}(N)$ \ub{then} for one and
2972: only one $p$ we have $p = q \restriction M$.
2973: \nl
2974: 3) If $M_0 \le_{\frak K} M_1 \le_{\frak K} M_2$ and $p \in {\Cal S}
2975: (M_2)$ then $p \restriction M_0 = (p \restriction M_1) \restriction
2976: M_0$.
2977: \nl
2978: 4) If $M \in {\frak K}_\mu$ is an amalgamation base then
2979: ${\Cal E}^{\text{at}}_M$ is a transitive relation hence is equal to
2980: ${\Cal E}_M$.
2981: \nl
2982: 5) If $M \le_{\frak K} N$ are from ${\frak K}_\lambda,M$ is an
2983: amalgamation base and $p \in {\Cal S}(M)$ then there is $q \in {\Cal
2984: S}(N)$ extending $p$, so the mapping $q \mapsto q \restriction M$ is a
2985: function from ${\Cal S}(N)$ onto ${\Cal S}(M)$.
2986: \enddemo
2987: \bigskip
2988:
2989: \demo{Proof} Easy. \hfill$\square_{\scite{600-0.12A}}$
2990: \enddemo
2991: \bigskip
2992:
2993: \definition{\stag{600-0.13} Definition} 1) We say
2994: \ub{$N$ is $\lambda$-universal over $M$} when $\lambda \ge \|N\|$ and
2995: for every $M',M
2996: \le_{\frak K} M' \in K_\lambda$, there is a $\le_{\frak K}$-embedding of $M'$
2997: into $N$ over $M$. If we omit $\lambda$ we mean $\|N\|$;
2998: clearly if $N$ is universal over $M$ then $M$ is an amalgamation base. \nl
2999: 2) $K^3_\lambda = \{(M,N,a):M \le_{\frak K} N,a \in N \backslash M$ and
3000: $M,N \in {\frak K}_\lambda\}$, with the partial order $\le$ defined by
3001: $(M,N,a) \le (M',N',a')$ iff $a = a',M \le_{\frak K} M'$ and $N
3002: \le_{\frak K} N'$. \nl
3003: 3) We say $(M,N,a) \in K^3_\lambda$ is minimal \ub{when}: if
3004: $(M,N,a) \le (M',N_\ell,a)
3005: \in K^3_\lambda$ for $\ell =1,2$ implies \ortp$(a,M',N_1) =
3006: \text{ \ortp}(a,M',N_2)$ moreover, $(M',N_1,a)
3007: {\Cal E}^{\text{at}}_\lambda (M',N_2,a)$
3008: (this strengthening is not needed
3009: if every $M' \in K_\lambda$ is an amalgamation bases).
3010: \nl
3011: 4) $N \in {\frak K}$ is $\lambda$-universal if every $M \in
3012: {\frak K}_\lambda$ can be $\le_{\frak K}$-embedded into it.
3013: \enddefinition
3014: \bigskip
3015:
3016: \remark{Remark} Why do we use
3017: $\le$ on $K^3_\lambda$? Because those triples serve us
3018: as a representation of types for which direct limit exists.
3019: \endremark
3020: \bigskip
3021:
3022: \definition{\stag{600-0.15} Definition} 1) $M^* \in K_\lambda$ is
3023: \ub{superlimit} if: clauses (a) + (b) + (c) below hold (and locally
3024: superlimit if clauses (a)$^- + (b) + (c)$ below hold and is pseudo
3025: superlimit if clauses (b) + (c) below hold) where:
3026: \mr
3027: \item "{$(a)$}" it is universal, (i.e. every $M \in K_\lambda$
3028: can be $\le_{\frak K}$-embedded into $M^*$), and
3029: \sn
3030: \item "{$(b)$}" if $\langle M_i:i \le \delta \rangle$ is
3031: $\le_{\frak K}$-increasing continuous,
3032: $\delta < \lambda^+$ and $i < \delta \Rightarrow M_i \cong M^*$ then
3033: $M_\delta \cong M^*$
3034: \sn
3035: \item "{$(a)^-$}" if $M^* \le M_1 \in K_\lambda$ then there is $M_2
3036: \in K_2$ which $\le_{\frak K}$-extend $M_1$ and is isomorphic to $M^*$
3037: \sn
3038: \item "{$(c)$}" there is $M^{**}$ isomorphic to $M^*$ such that $M^*
3039: <_{\frak K} M^{**}$.
3040: \ermn
3041: 2) $M$ is $\lambda$-saturated above $\mu$ \ub{when}
3042: $\|M\| \ge \lambda > \mu \ge
3043: \text{ LS}({\frak K})$ and: $N \le_{\frak K} M,\mu \le \|N\| <
3044: \lambda,N \le_{\frak K} N_1,\|N_1\| \le \|N\| + \text{ LS}({\frak K})$
3045: and $a \in N_1$ then some $b \in M$ strongly realizes $(N,N_1,a)/{\Cal
3046: E}^{\text{at}}_N$ in $M$, see Definition \scite{600-0.12}(3).
3047: Omitting ``above $\mu$" means ``for some $\mu < \lambda$" hence
3048: ``$M$ is $\lambda^+$-saturated" mean that ``$M$ is $\lambda^+$-saturated above
3049: $\lambda$" and $K(\lambda^+$-saturated)
3050: $= \{M \in K:M$ is $\lambda^+$-saturated$\}$ and ``$M$ is saturated"
3051: means ``$M$ is $\|M\|$-saturated".
3052: \enddefinition
3053: \bigskip
3054:
3055: \proclaim{\stag{600-0.19} The Model-homogeneity = Saturativity Lemma} Let
3056: $\lambda > \mu + { \text{\rm LS\/}}({\frak K})$ and $M \in K$. \newline
3057: 1) $M$ is $\lambda$-saturated above $\mu$ \underbar{iff} $M$ is
3058: $({\Bbb D}_{{\frak K}_{\ge\mu}},\lambda)$-homogeneous above $\mu$, which
3059: means: for every
3060: $N_1 \le_{\frak K} N_2 \in K$ such that $\mu \le \|N_1\| \le \|N_2\| <
3061: \lambda$ and $N_1 \le_{\frak K} M$, there is a $\le_{\frak K}$-embedding $f$
3062: of $N_2$ into $M$ over $N_1$. \newline
3063: 2) If $M_1,M_2 \in K_\lambda$ are $\lambda$-saturated above $\mu < \lambda$
3064: and for some $N_1 \le_{\frak K} M_1,N_2 \le_{\frak K} M_2$, both
3065: of cardinality $\in [\mu,\lambda)$, we have $N_1 \cong N_2$ \ub{then}
3066: $M_1 \cong M_2$; in fact, any isomorphism $f$ from $N_1$ onto $N_2$
3067: can be extended to an isomorphism from $M_1$ onto $M_2$. \newline
3068: 3) If in (2) we demand only ``$M_2$ is $\lambda$-saturated" and $M_1 \in
3069: K_{\le \lambda}$ \underbar{then} $f$ can be extended to a
3070: $\le_{\frak K}$-embedding from $M_1$ into $M_2$. \newline
3071: 4) In part (2) instead of $N_1 \cong N_2$ it suffices to assume that
3072: $N_1$ and $N_2$ can be $\le_{\frak K}$-embedded into some $N \in K$,
3073: which holds if ${\frak K}$ has the {\rm JEP} or just {\rm JEP}$_\mu$.
3074: \nl
3075: 5) If $N$ is $\lambda$-universal over $M \in K_\mu$ and ${\frak K}$
3076: has $\mu$-JEP then $N$ is $\lambda$-universal (where $\lambda \ge
3077: \text{\rm LS}({\frak K})$ for simplicity).
3078: \nl
3079: 6) Assume $M$ is $\lambda$-saturated above $\mu$. If $N \le_{\frak K}
3080: M$ and $\mu \le \|N\| < \lambda$ \ub{then} $N$ is an amalgamation base
3081: (in $K_{\le(\|N\|+\text{LS}({\frak K}))}$ and
3082: even in ${\frak K}_{\le \lambda}$) and $|{\Cal S}(N)| \le \|M\|$.
3083: So if every $N \in K_\mu$ can be $\le_{\frak K}$-embedded into $M$
3084: then ${\frak K}$ has $\mu$-amalgamation.
3085: \endproclaim
3086: \bigskip
3087:
3088: \demo{Proof} 1) The ``if" direction is easy as $\lambda > \mu \ge
3089: \text{ LS}({\frak K})$.
3090: Let us prove the other direction.
3091:
3092: We prove this by induction on $\|N_2\|$. Now first consider the case
3093: $\|N_2\| > \|N_1\| +
3094: \text{\rm LS}({\frak K})$ then we can find a $\le_{\frak
3095: K}$-increasing continuous sequence $\langle
3096: N_{1,\varepsilon}:\varepsilon < \|N_2\|\rangle$ with union $N_2$ with
3097: $N_{1,0} = N_1$ and $\|N_{1,\varepsilon}\| \le \|N_1\| +
3098: |\varepsilon|$. Now we choose $f_\varepsilon$, a $\le_{\frak
3099: K}$-embedding of $N_{1,\varepsilon}$ into $M$, increasing continuous
3100: with $\varepsilon$ such that $f_0 = \text{\rm id}_{N_1}$. For
3101: $\varepsilon = 0$ this is trivial for $\varepsilon$ limit take unions
3102: and for $\varepsilon$ successor use the induction hypothesis. So
3103: \wilog \, $\|N_2\| \le \|N_1\| + \text{\rm LS}({\frak K})$.
3104:
3105: Let $|N_2| = \{ a_i:i < \kappa \}$, and we know $\mu \le \kappa'' :=
3106: \|N_1\| \le \kappa := \|N_2\| \le \kappa' :=
3107: \|N_1\| + \text{\rm LS}({\frak K})
3108: < \lambda$; so if, as usual, $\|N_1\| \ge \text{\rm LS}({\frak K})$
3109: then $\kappa' = \kappa$. We define by induction on $i \le \kappa,
3110: N^i_1,N^i_2,f_i$ such that:
3111: \mr
3112: \item "{(a)}" $N^i_1 \le_{\frak K} N^i_2$ and $\|N^i_1\| \le
3113: \|N^i_2\| \le \kappa'$
3114: \sn
3115: \item "{(b)}" $N^i_1$ is $\le_{\frak K}$-increasing continuous with $i$
3116: \sn
3117: \item "{(c)}" $N^i_2$ is $\le_{\frak K}$-increasing continuous with $i$
3118: \sn
3119: \item "{(d)}" $f_i$ is a $\le_{\frak K}$-embedding of $N^i_1$ into $M$
3120: \sn
3121: \item "{(e)}" $f_i$ is increasing continuous with $i$
3122: \sn
3123: \item "{(f)}" $a_i \in N^{i+1}_1$
3124: \sn
3125: \item "{(g)}" $N^0_1 = N_1,N^0_2 = N_2,f_0 = \text{\rm id}_{N_1}$.
3126: \ermn
3127: For $i = 0$, clause $(g)$ gives the definition. For $i$ limit let: \newline
3128: \sn
3129: $N^i_1 = \dsize \bigcup_{j < i} N^j_1$ and
3130: \sn
3131: $N^i_2 = \dsize \bigcup_{j < i} N^j_2$ and
3132: \sn
3133: $f_i = \dsize \bigcup_{j < i} f_j$. \newline
3134: \sn
3135: Now (a)-(f) continues to hold by continuity (and $\|N^i_2\| \le
3136: \kappa'$ easily).
3137: \medskip
3138:
3139: For $i$ successor we use our assumption; more elaborately, let
3140: $M^{i-1}_1 \le_{\frak K} M$ be $f_{i-1}(N^{i-1}_1)$ and let $M^{i-1}_2,
3141: g_{i-1}$ be such that $g_{i-1}$ is an isomorphism from $N^{i-1}_2$ \newline
3142: onto $M^{i-1}_2$ extending
3143: $f_{i-1}$, so $M^{i-1}_1 \le_{\frak K} M^{i-1}_2$ (but \wilog \nl
3144: $M^{i-1}_2 \cap M = M^{i-1}_1$). Now apply the saturation
3145: assumption with $M,M^{i-1}_1$, \nl
3146: $\text{\ortp}(g_{i-1}(a_{i-1}),M^{i-1}_1,M^{i-1}_2)$
3147: here standing for $N,M,p$ there
3148: (note: $a_{i-1} \in N_2 = N^0_2 \subseteq N^{i-1}_2$ and
3149: $\lambda > \kappa' \ge \|N^{i-1}_2\| = \|M^{i-1}_2\| \ge \|M^{i-1}_1\| =
3150: \|N^{i-1}_1\| \ge \|N^0_1\| = \|N_1\| = \kappa'' \ge \mu$ so the requirements
3151: including the requirements on the cardinalities
3152: in Definition \scite{600-0.15}(2) holds).
3153: So there is $b \in M$ such that \ortp$(b,M^{i-1}_1,M) = \text{ \ortp}(g_{i-1}
3154: (a_{i-1}),M^{i-1}_1,M^{i-1}_2)$. Moreover (if ${\frak K}$ has amalgamation
3155: in $\mu$ the proof is slightly shorter) remembering the end of the
3156: first sentence in
3157: \scite{600-0.15}(2) which speaks about ``strongly realizes" and recalling
3158: Definition \scite{600-0.12}(3) there is $b \in M$
3159: such that $b$ strongly realizes $(M^{i-1}_1,M^{i-1}_3,g_{i-1}(a_{i-1}))/
3160: {\Cal E}^{\text{at}}_{M^{i-1}_1}$ in $M$.
3161: This means (see Definition \scite{600-0.12}(3)) that for some
3162: $M^{i,*}_1$ we have $b \in M^{i,*}_1$ and
3163: $M^{i-1}_1 \le_{\frak K} M^{i,*}_1
3164: \le_{\frak K} M$ and $(M^{i-1}_1,M^{i-1}_2,g_{i-1}(a_{i-1}))
3165: {\Cal E}^{\text{at}}_{M^{i-1}_1}(M^{i-1}_1,M^{i,*}_1,b)$. This means (see
3166: Definition \scite{600-0.12}(1)) that $M^{i,*}_1$ too has cardinality
3167: $\le \kappa'$ and there is $M^{i,*}_2 \in K_{\le \kappa'}$
3168: such that $M^{i-1}_1 \le_{\frak K} M^{i,*}_2$ and
3169: there are $\le_{\frak K}$-embeddings $h^i_2,h^i_1$ of $M^{i-1}_2,M^{i,*}_1$
3170: into $M^{i,*}_2$ over $M^{i-1}_1$ respectively, such that $h^i_2(g_{i-1}
3171: (a_{i-1})) = h^i_1(b)$. \nl
3172: Now changing names, \wilog \, $h^i_1$ is the identity.
3173: \nl
3174: Let $N^i_2,h_i$ be such that $N^{i-1}_2 \le_{\frak K} N^i_2$ and $h_i$ an
3175: isomorphism from $N^i_2$ onto $M^{i,*}_2$ extending $g_{i-1}$. Let
3176: $N^i_1 = h^{-1}_i (M^{i,*}_1)$ and $f_i = (h_i \restriction N^i_1)$.
3177: \medskip
3178:
3179: We have carried the induction. Now $f_\kappa$ is a
3180: $\le_{\frak K}$-embedding of $N^\kappa_1$ into $M$ over $N_1$,
3181: but $|N_2| = \{ a_i:i <
3182: \kappa \} \subseteq N^\kappa_1$ hence by AxV of Definition
3183: \scite{600-0.2}, $N_2 \le_{\frak K} N^\kappa_1$, so
3184: $f_\kappa \restriction N_2:N_2 \rightarrow M$ is as required.
3185: \newline
3186: 2), 3) By the hence and forth argument (or see \marginbf{!!}{\cprefix{88r}.\scite{88r-2.3}},\marginbf{!!}{\cprefix{88r}.\scite{88r-2.4}}
3187: or see \cite[II,\S3]{Sh:300} = \sectioncite[\S3]{300b}). \nl
3188: 4),5),6) Easy, too. \hfill$\square_{\scite{600-0.19}}$
3189: \enddemo
3190: \bigskip
3191:
3192: \definition{\stag{600-0.21} Definition} 1) For $\sigma = \text{
3193: cf}(\sigma) \le \lambda^+$, we say \ub{$N$ is
3194: $(\lambda,\sigma)$-brimmed over $M$} if ($M \le_{\frak K} N$ are in
3195: $K_\lambda$ and) we can find a sequence $\langle
3196: M_i:i < \sigma \rangle$ which is $\le_{\frak K}$-increasing
3197: \footnote{we have not asked continuity; because in the direction we are
3198: going, it makes no difference if we add ``continuous". Then we have in
3199: general fewer cases of existence, uniqueness (of being
3200: $(\lambda,\sigma)$-brimmed over $M \in K_\lambda$) does not need extra
3201: assumptions and existence is harder}, $M_i \in
3202: K_\lambda,M_0 = M,M_{i+1}$ is $\le_{\frak K}$-universal
3203: \footnote{hence $M_i$ is an amalgamation base} over $M_i$ and
3204: $\dbcu_{i < \sigma} M_i = N$. We say $N$ is $(\lambda,\sigma)$-brimmed
3205: over $A$ if $A \subseteq N \in K_\lambda$ and we can find $\langle M_i:i <
3206: \sigma \rangle$ as above such that $A \subseteq M_0$ but $M_0
3207: \restriction A \le_{\frak K} M_0 \Rightarrow M_0 = A$; if $A =
3208: \emptyset$ we may omit ``over $A$". We say continuously
3209: $(\lambda,\sigma)$-brimmed (over $M$) \ub{when} the sequence $\langle M_i:i
3210: < \sigma\rangle$ is $\le_{\frak K}$-increasing continuous; if ${\frak
3211: K}_\lambda$ has amalgamation, the two notions coincide.
3212: \nl
3213: 2) We say
3214: $N$ is $(\lambda,*)$-brimmed over $M$ \ub{if} for some $\sigma \le
3215: \lambda,N$ is $(\lambda,\sigma)$-brimmed over $M$. We say $N$ is
3216: $(\lambda,*)$-brimmed if for some $M,N$ is $(\lambda,*)$-brimmed
3217: over $M$.
3218: \nl
3219: 3) If $\alpha < \lambda^+$ let ``$N$ is $(\lambda,\alpha)$-brimmed over
3220: $M$" mean $M \le_{\frak K} N$ are from $K_\lambda$ and cf$(\alpha) \ge
3221: \aleph_0 \Rightarrow N$ is $(\lambda,\text{cf}(\alpha))$-brimmed over $M$.
3222: \enddefinition
3223: \bn
3224: On the meaning of $(\lambda,\sigma)$-brimmed for elementary classes,
3225: see \scite{600-Ex.0}(2) below. Recall
3226: \proclaim{\stag{600-0.22} Claim} Assume $\lambda \ge \text{\rm LS}({\frak K})$.
3227: \nl
3228: 1) If ${\frak K}$ has amalgamation in $\lambda$, is stable in $\lambda$
3229: and $\sigma = {\text{\rm cf\/}}(\sigma) \le \lambda$, \ub{then}
3230: \mr
3231: \item "{$(a)$}" for every $M \in {\frak K}_\lambda$ there is $N,M
3232: \le_{\frak K} N \in K_\lambda$, universal over $M$
3233: \sn
3234: \item "{$(b)$}" for every $M \in {\frak K}_\lambda$ there is
3235: $N \in {\frak K}_\lambda$ which is $(\lambda,\sigma)$-brimmed
3236: over $M$
3237: \sn
3238: \item "{$(c)$}" if $N$ is $(\lambda,\sigma)$-brimmed over $M$
3239: \ub{then} $N$ is universal over $M$.
3240: \ermn
3241: 2) If $N_\ell$ is $(\lambda,\aleph_0)$-brimmed over $M$ for $\ell =1,2$,
3242: \ub{then} $N_1,N_2$ are isomorphic over $M$.
3243: \nl
3244: 3) Assume $\sigma = { \text{\rm cf\/}}(\sigma) \le \lambda^+$, and for
3245: every $\aleph_0 \le \theta = { \text{\rm cf\/}}(\theta)
3246: < \sigma$ any $(\lambda,\theta)$-brimmed model
3247: is an amalgamation base (in ${\frak K}$). \ub{Then}:
3248: \mr
3249: \item "{$(a)$}" if $N_\ell$ is $(\lambda,\sigma)$-brimmed
3250: over $M$ for $\ell=1,2$ \ub{then} $N_1,N_2$ are isomorphic over $M$
3251: \sn
3252: \item "{$(b)$}" if ${\frak K}$ has JEP$_\lambda$ (i.e., the
3253: joint embedding property in $\lambda$) and
3254: $N_1,N_2$ are $(\lambda,\sigma)$-brimmed \ub{then} $N_1,N_2$ are isomorphic.
3255: \ermn
3256: 3A) There is a $(\lambda,\sigma)$-brimmed model $N$ over $M \in
3257: K_\lambda$ \ub{when}: for every $\le_{{\frak K}_\lambda}$-extension
3258: $M_1$ of $M$ there is a $\le_{{\frak K}_\lambda}$-extension $M_2$ of
3259: $M_1$ which is an amalgamation base and there is a $\lambda$-universal
3260: extension $M_3 \in K_\lambda$ of $M_2$.
3261: \nl
3262: 4) Assume ${\frak K}$ has $\lambda$-amalgamation and the $\lambda$-JEP
3263: and $\bar M = \langle M_i:i \le \lambda \rangle$ is $\le_{\frak K}$-increasing
3264: continuous and $M_i \in K_\lambda$ for $i \le \lambda$.
3265: \mr
3266: \item "{$(a)$}" If $\lambda$ is regular and for every $i < \lambda,p
3267: \in {\Cal S}(M_i)$ for some $j \in (i,\lambda)$, some $a \in M_j$
3268: realizes $p$, \ub{then} $M_\lambda$ is universal over $M_0$ and is
3269: $(\lambda,\lambda)$-brimmed over $M_0$
3270: \sn
3271: \item "{$(b)$}" if for every $i < \lambda$ every $p \in {\Cal
3272: S}(M_i)$ is realized in $M_{i+1}$ \ub{then} $M_\lambda$ is
3273: $(\lambda,\text{\rm cf}(\lambda))$-brimmed over $M_0$.
3274: \ermn
3275: 5) Assume $\sigma = {\text{\rm cf\/}}(\sigma) \le \lambda$ and $M \in
3276: {\frak K}$ is continuous $(\lambda,\sigma)$-brimmed. \ub{Then} $M$ is
3277: locally a
3278: $(\lambda,\{\sigma\})$-strongly limit model in ${\frak K}_\lambda$
3279: (see Definition \marginbf{!!}{\cprefix{88r}.\scite{88r-3.1}}(2),(7), not used).
3280: \endproclaim
3281: \bigskip
3282:
3283: \demo{Proof} 1) Clause (c) holds by Definition \scite{600-0.21}.
3284:
3285: As for clause (a), for any given $M \in K_\lambda$, easily there
3286: is an $\le_{\frak K}$-increasing continuous sequence
3287: $\langle M_i:i \le \delta \rangle$ of models from $K_\lambda,M_0 = M$
3288: such that $p \in {\Cal S}(M_i) \Rightarrow p$ is realized in $M_{i+1}$,
3289: this by stability + amalgamation. So $\langle M_i:i \le \lambda
3290: \rangle$ is as in part (4) hence by clause (b) of part (4) we get that
3291: $M_\delta$ is $\le_{\frak K}$-universal over $M_0=M$ so we are done.
3292: Clause (b) follows by (a) as ${\frak K}_\lambda$ has $\lambda$-amalgamation.
3293: \nl
3294: 2) By (3)(a).
3295: \nl
3296: 3) \ub{Clause (a)} holds by the hence and forth argument, that is
3297: assume $\langle N_{\ell,i}:i \le \sigma \rangle$ be $\le_{\frak K}$-increasing
3298: continuous, $N_{\ell,0} =M,N_{\ell,i+1}$ is universal over
3299: $N_{\ell,i}$ and $N_\ell = N_{\ell,\sigma}$ so $N_{\ell,i} \in
3300: {\frak K}_\lambda$. We now choose $f_i$ by induction on $i \le
3301: \sigma$ such that:
3302: \mr
3303: \widestnumber\item{$(iii)$}
3304: \item "{$(i)$}" if $i$ is odd, $f_i$ is a $\le_{\frak K}$-embedding
3305: of $N_{1,i}$ into $N_{2,i}$
3306: \sn
3307: \item "{$(ii)$}" if $i$ is even, $f^{-1}_i$ is a
3308: $\le_{\frak K}$-embedding of $N_{2,i}$ into $N_{1,i}$
3309: \sn
3310: \item "{$(iii)$}" if $i$ is limit then $f_i$ is an
3311: isomorphism from $N_{1,i}$ onto $N_{2,i}$
3312: \sn
3313: \item "{$(iv)$}" $f_i$ is increasing continuous with $i$
3314: \sn
3315: \item "{$(v)$}" if $i=0$ then $f_0 = \text{\rm id}_M$.
3316: \ermn
3317: For $i=0$ let $f_0 = \text{ id}_M$. If $i=2j+2$ use ``$N_{1,i}$ is
3318: a universal extension of $N_{1,2j+1}$ (in ${\frak K}_\lambda$) and
3319: $f_{2j+1}$ is a $\le_{\frak K}$-embedding of $N_{1,2j+1}$ into
3320: $N_{2,2j+1}$ (by clause (i) applied to $2j+1$)
3321: and $N_{1,2j+1}$ is an amalgamation base". That is, $N_{2,i}$ is a
3322: $\le_{\frak K}$-extension of $f_{2j+1}(N_{2j+1})$ which is an
3323: amalgamation base so $f^{-1}_{2j+1}$ can be extended to a $\le_{\frak
3324: K}$-embedding of $f^{-1}_i$ of $N_{2,i}$ into $N_{1,i}$.
3325: For $i=2j+1$ use ``$N_{2,i}$ is a universal extension
3326: (in ${\frak K}_\lambda$) of
3327: $N_{2,2j}$ and $f^{-1}_{2j}$ is a $\le_{\frak K}$-embedding of
3328: $N_{2,2j}$ into $N_{1,2j}$ and $N_{2,2j}$ is an amalgamation base (in
3329: ${\frak K}_\lambda$)".
3330: \nl
3331: For $i$ limit let $f_i = \cup\{f_j:j < i\}$. Clearly $f_\sigma$ is an
3332: isomorphism from $N_1 = N_{1,\sigma}$ onto $N_{2,\sigma} = N_2$ so we
3333: are done.
3334:
3335: As for clause (b), we can assume that $\langle N_{\ell,i}:i \le \sigma
3336: \rangle$ exemplifies ``$N_\ell$ is $(\lambda,\sigma)$-brimmed" for
3337: $\ell=1,2$. By the JEP$_\lambda$ there is a pair $(g_1,N)$ such that
3338: $N_{1,0} \le_{\frak K} N \in K_\lambda$ and $g_1$ is a $\le_{\frak
3339: K}$-embedding of $N_{2,0}$ into $N$. As above there is a $\le_{\frak
3340: K}$-embedding $g_2$ of $N$ into $N_{1,1}$ over $N_{1,0}$. Let $f_0 =
3341: g_2$ and continue as in the proof of clause (a).
3342: \nl
3343: 3A) Easy, too.
3344: \nl
3345: 4) We first proved weaker versions of (a) and of (b) called
3346: (a)$^-$,(b)$^-$ respectively.
3347: \bn
3348: \ub{Clause $(a)^-$}: Like (a) but we conclude only: $M_\lambda$ is
3349: universal over $M_0$.
3350:
3351: Let $M_0 \le_{\frak K} N \in K_\lambda$ and we
3352: let $\langle S_i:i < \lambda \rangle$ be a partition of $\lambda$ to
3353: $\lambda$ sets each with $\lambda$ members, $i \le \text{\rm
3354: Min}(S_i)$. Let $M_{1,i} = M_i$ for $i \le \lambda$ and we choose
3355: $\langle M_{2,i}:i \le \delta \rangle$ which is $\le_{\frak
3356: K}$-increasing such that $M_{2,i} \in {\frak K},
3357: M_{2,0} = M_{1,0},N \le_{\frak K} M_{2,1}$ and every type
3358: $p \in {\Cal S}(M_{2,i})$ is realized in $M_{2,i+1}$. We shall prove
3359: that $M_{1,\lambda},M_{2,\lambda}$ are isomorphic over
3360: $M_0 = M_{1,0}$, this clearly suffices. We choose a quintuple
3361: $(j_i,M_{3,i},f_{1,i},f_{2,i},\bar{\bold a}_i)$ by induction
3362: on $i < \lambda$ such that
3363: \mr
3364: \item "{$\circledast$}" $(a) \quad j_i < \lambda$ is increasing
3365: continuous
3366: \sn
3367: \item "{${{}}$}" $(b) \quad M_{3,i} \in K_\lambda$ is $\le_{\frak
3368: K}$-increasing continuous
3369: \sn
3370: \item "{${{}}$}" $(c) \quad f_{\ell,i}$ is a
3371: $\le_{\frak K}$-embedding of $M_{\ell,j_i}$ into $M$ for $\ell=1,2$
3372: \sn
3373: \item "{${{}}$}" $(d) \quad f_{\ell,i}$ is increasing continuous with
3374: $i$ for $\ell=1,2$
3375: \sn
3376: \item "{${{}}$}" $(e) \quad \bar{\bold a}_i = \langle
3377: a^i_\varepsilon:\varepsilon \in S_i\rangle$ lists the members of
3378: $M_{3,i}$
3379: \sn
3380: \item "{${{}}$}" $(f) \quad$ if $\varepsilon \in S_i$ then
3381: $a^i_\varepsilon \in \text{\rm Rang}(f_{1,2 \varepsilon +1})$ and
3382: $a^i_\varepsilon \in \text{\rm Rang}(f_{2,2 \varepsilon +2})$.
3383: \ermn
3384: If we succeed then $f_\ell := \cup\{f_{\ell,i}:i < \lambda\}$ is a
3385: $\le_{\frak K}$-embedding of $M_{1,\lambda}$ into $M_{3,\lambda} :=
3386: M_3 := \cup\{M_{3,i}:i < \lambda\}$ and this embedding is onto because
3387: $a \in M_3 \Rightarrow$ for some $i < \lambda,a \in M_{3,i}
3388: \Rightarrow$ for some $i < \lambda$ and $\varepsilon \in S_i,a =
3389: a^i_\varepsilon \Rightarrow a = a^i_\varepsilon \in \text{\rm
3390: Rang}(f_{\ell,\varepsilon +1}) \Rightarrow a \in \text{\rm
3391: Rang}(f_\ell)$. So $f^{-1}_1 \circ f_2$ is an isomorphism from
3392: $M_{2,\lambda}$ onto $M_{1,\lambda} = M_\lambda$ so as said above we are
3393: done.
3394:
3395: Carrying the induction; for $i=0$ use ``${\frak K}$ has the
3396: $\lambda$-JEP" for $M_{1,0},M_{2,0}$.
3397:
3398: For $i$ limit take unions.
3399:
3400: For $i= 2 \varepsilon +1$ let $j_i = \text{ min}\{j < \lambda_i:j >
3401: j_{2 \varepsilon}$ and $(f^1_{2 \varepsilon})^{-1} \,
3402: (\text{tp}(a^i_\varepsilon,f^1_{2 \varepsilon}(M_{1,i}),M_{3,i}) \in
3403: {\Cal S}_{\frak K}(M_{1,i})$ is realized in $M_j$ and continue as in
3404: the proof of \scite{600-0.19}(1), so can avoid using ``$(f^1_i)^{-2}$ of a
3405: type.
3406:
3407: For $i = 2 \varepsilon +2$, the proof is similar.
3408: \bn
3409: \ub{Clause $(b)^-$}: Like clause (b) but we conclude only:
3410: $M_\lambda$ is universal over $M_0$.
3411:
3412: Similar to the proof of $(a)^-$ except that we weaken clause (f) to
3413: \mr
3414: \item "{$(f)^-$}" if $\varepsilon \in S_i$ then $a^i_\varepsilon \in
3415: \text{ Rang}(f_{1,2 \varepsilon +1})$.
3416: \endroster
3417: \bn
3418: \ub{Clauses $(a),(b)$}:
3419:
3420: As above but now in the choice of $\langle M_{2,i}:i \le \lambda
3421: \rangle$ we demand that ``$M_{2,i+1}$ is $\le_{\frak K}$-universal
3422: over $M_{2,i}$" for $i < \lambda$. This is permissible as by $(b)^-$
3423: which we have already proved every member of $K_\lambda$ has a
3424: $<_{{\frak K}_\lambda}$-extension
3425: from ${\frak K}_\lambda$ which is universal over it. So
3426: $M_{2,\lambda}$ is $(\lambda,\text{\rm cf}(\lambda))$-brimmed over
3427: $M_{2,0} = M_0$ hence also $M_\lambda$ being isomorphic to
3428: $M_{2,\lambda}$ over $M_0$ is $(\lambda,\text{\rm
3429: cf}(\lambda))$-brimmed over $M_0$, as required.
3430: \nl
3431: 5) Easy and not used. (Let $\langle M_i:i \le \sigma\rangle$ witness
3432: ``$M$ is $(\lambda,\sigma)$-brimmed", so $M$ can be $\le_{\frak
3433: K}$-embedded into $M_i$, hence \wilog \, $M_0 \cong M_1$.) Now use
3434: $\bold F$ such that $\bold F(M')$ is a
3435: $\le_{{\frak K}_\lambda}$-extension of $M'$ which is
3436: $\le_{{\frak K}_\lambda}$-universal over it and is an amalgamation base.
3437: \hfill$\square_{\scite{600-0.22}}$
3438: \enddemo
3439: \bigskip
3440:
3441: \proclaim{\stag{600-4a.10} Claim} 1) Assume that ${\frak K}$ is an {\rm
3442: a.e.c., LS}$({\frak K}) \le \lambda$ and ${\frak K}$ has
3443: $\lambda$-amalgamation and is stable in $\lambda$. \ub{Then} there is a
3444: saturated $N \in K_{\lambda^+}$. Also for every saturated $N \in
3445: K_{\lambda^+}$ (in ${\frak K}$, above $\lambda$ of course) we can find
3446: a $\le_{\frak K}$-representation $\bar N = \langle N_i:i < \lambda^+
3447: \rangle$, with $N_{i+1}$ being $(\lambda,{\text{\rm
3448: cf\/}}(\lambda))$-brimmed over $N_i$ and $N_0$ being
3449: $(\lambda,\lambda)$-brimmed.
3450: \nl
3451: 2) If for $\ell=1,2$ we have $\bar N^\ell = \langle N^\ell_i:i < \lambda^+
3452: \rangle$ as in part (1), \ub{then} there is an isomorphism $f$ from
3453: $N^1$ onto $N^2$ mapping $N^1_i$ onto $N^2_i$ for each $i <
3454: \lambda^+$. Moreover, for any $i < \lambda^+$ and isomorphism $g$
3455: from $N^1_i$ onto $N^2_i$ we can find an isomorphism $f$ from $N^1$
3456: onto $N^2$ extending $g$ and mappng $N^1_j$ onto $N^2_j$ for each $j
3457: \in [i,\lambda^+)$.
3458: \nl
3459: 3) If $N^0 \le_{\frak K} N^1$ are both saturated (above $\lambda$) and
3460: are in $K_{\lambda^+}$ (hence {\rm LS}$({\frak K}) \le \lambda$),
3461: \ub{then} we can find $\le_{\frak K}$-representation $\bar
3462: N^\ell$ of $N^\ell$ as in (1) for $\ell=1,2$
3463: with $N^0_i = N^0 \cap N^1_i$, (so $N^0_i \le_{\frak K} N^1_i$) for $i
3464: < \lambda^+$.
3465: \nl
3466: 4) If $M \in K_{\lambda^+}$ and ${\frak K}$ has $\lambda$-amalgamation
3467: and is stable in $\lambda$ (and {\rm LS}$({\frak K}) \le \lambda$),
3468: \ub{then} for some $N \in K_{\lambda^+}$ saturated (above
3469: $\lambda$) we have $M \le_{\frak K} N$.
3470: \endproclaim
3471: \bigskip
3472:
3473: \demo{Proof} Easy (for (2),(3) using \scite{600-0.19}(6)), e.g.
3474: \nl
3475: 4) There is a $\le_{\frak K}$-increasing continuous sequence
3476: $\langle M_i:i < \lambda^+\rangle$ with union $M$ such that $M_i \in
3477: K_\lambda$. Now we choose $N_i$ by induction on $i < \lambda$
3478: \mr
3479: \item "{$(*)$}" $(a) \quad N_i \in K_\lambda$ is $\le_{\frak
3480: K}$-increasing continuous
3481: \sn
3482: \item "{${{}}$}" $(b) \quad N_{i+1}$ is
3483: $(\lambda,\text{\rm cf}(\lambda))$-brimmed over $N_i$
3484: \sn
3485: \item "{${{}}$}" $(c) \quad N_0 = M_0$.
3486: \ermn
3487: This is possible by \scite{600-0.22}(1). Then by induction on $i \le
3488: \lambda^+$ we choose a $\le_{\frak K}$-embedding $f_i$ of $M_i$ into $N_i$,
3489: increasing continuous with $i$. For $i=0$ let $f_i = \text{\rm
3490: id}_{M_0}$. For $i$ limit use union.
3491:
3492: Lastly, for $i=j+1$ use ``${\frak K}$ has $\lambda$-amalgamation" and
3493: ``$N_j$ is universal over $N_i$". Now by renaming \wilog \,
3494: $f_{\lambda^+} = \text{\rm id}_{N_{\lambda^+}}$ and we are done. (Of
3495: course, we hae assumed less). \hfill$\square_{\scite{600-4a.10}}$
3496: \enddemo
3497: \bn
3498: You may wonder why in this work we have not restricted our ${\frak K}$ to
3499: ``abstract elementary class in $\lambda$" say in \S2 below (or in
3500: \cite{Sh:576}); by the following facts (mainly \scite{600-0.31})
3501: this is immaterial.
3502: \bigskip
3503:
3504: \definition{\stag{600-0.30A} Definition} 1) We say that ${\frak K}_\lambda$ is
3505: a $\lambda$-abstract elementary class or $\lambda$-a.e.c.
3506: in short, \ub{when}:
3507: \mr
3508: \item "{$(a)$}" ${\frak K}_\lambda = (K_\lambda,\le_{{\frak K}_\lambda})$,
3509: \sn
3510: \item "{$(b)$}" $K_\lambda$ is a class of $\tau$-models of cardinality
3511: $\lambda$ closed under isomorphism for some vocabulary $\tau =
3512: \tau_{{\frak K}_\lambda}$,
3513: \sn
3514: \item "{$(c)$}" $\le_{{\frak K}_\lambda}$ a partial order of $K_\lambda$,
3515: closed under isomorphisms
3516: \sn
3517: \item "{$(d)$}" axioms (0 and) I,II,III,IV,V of abstract elementary classes
3518: (see \scite{600-0.2}) hold except that in AxIII we demand $\delta <
3519: \lambda^+$ (you can demand this also in AxIV).
3520: \ermn
3521: 2) For an abstract elementary class ${\frak K}$ let ${\frak K}_\lambda =
3522: (K_\lambda,\le_{\frak K} \restriction K_\lambda)$ and similarly
3523: ${\frak K}_{\ge \lambda},{\frak K}_{\le \lambda},{\frak
3524: K}_{[\lambda,\mu]}$ and define $(\le \lambda)$-a.e.c. and
3525: $[\lambda,\mu]$-a.e.c., etc.
3526: \nl
3527: 3) Definitions \scite{600-0.12}, \scite{600-0.13}, \scite{600-0.15}, \scite{600-0.21}
3528: apply to $\lambda$-a.e.c. ${\frak K}_\lambda$.
3529: \enddefinition
3530: \bigskip
3531:
3532: \demo{\stag{600-0.30B} Observation} If ${\frak K}^1$ is an a.e.c. with
3533: $K^1_\lambda \ne \emptyset$ then
3534: \mr
3535: \item "{$(a)$}" ${\frak K}^1_\lambda$ is a $\lambda$-a.e.c.
3536: \sn
3537: \item "{$(b)$}" if ${\frak K}^2_\lambda$ is a $\lambda$-a.e.c., and
3538: ${\frak K}^1_\lambda = {\frak K}^2_\lambda$ \ub{then} Definitions
3539: \scite{600-0.12}, \scite{600-0.13}, \scite{600-0.15}, \scite{600-0.21} when applied to
3540: ${\frak K}^1$ but restricting ourselves to models of cardinality
3541: $\lambda$ and when applied to ${\frak K}^2_\lambda$ are equivalent.
3542: \endroster
3543: \enddemo
3544: \bigskip
3545:
3546: \demo{Proof} Just read the definitions.
3547: \enddemo
3548: \bn
3549: We may wonder
3550: \nl
3551: \margintag{600-0.30C}\ub{\stag{600-0.30C} Problem}: Suppose ${\frak K}^1,{\frak K}^2$ are
3552: a.e.c. such that for some $\lambda > \mu \ge \text{ LS}({\frak K}^1)$,
3553: $\text{LS}({\frak K}^2),{\frak K}^1_\lambda = {\frak K}^2_\lambda$.
3554: Can we bound the first such $\lambda$ above $\mu$? (Well, better
3555: bound than the Lowenheim number of $\Bbb L_{\mu^+,\mu^+}$(second order)).
3556: \bigskip
3557:
3558: \demo{\stag{600-0.30D} Observation} 1) Let ${\frak K}$ be an a.e.c. with
3559: $\lambda = \text{ LS}({\frak K})$ and $\mu \ge \lambda$ and we define
3560: ${\frak K}_{\ge \mu}$ by: $M \in {\frak K}_{\ge \mu}$ iff $M \in K
3561: \and \|M\| \ge\mu$ and $M \le_{{\frak K}_{\ge \mu}} N$ if $M
3562: \le_{\frak K} N$ and $\|M\|,\|N\| \ge \mu$. \ub{Then} ${\frak K}_{\ge
3563: \mu}$ is an a.e.c. with LS$({\frak K}_{\ge \mu}) = \mu$.
3564: \nl
3565: 2) If ${\frak K}_\lambda$ is a $\lambda$-a.e.c. then observation
3566: \scite{600-0.6} holds when $|I| \le \lambda$.
3567: \nl
3568: 3) Claims \scite{600-0.12A}(2)-(5), \scite{600-0.22} apply to $\lambda$-a.e.c.
3569: \enddemo
3570: \bigskip
3571:
3572: \remark{\stag{600-0.30D.1} Remark} If
3573: ${\frak K}$ is an a.e.c. with Lowenheim-Skolem number $\lambda$, then
3574: every model of ${\frak K}$ can be written as a direct limit
3575: (by $\le_{\frak K}$) of members of ${\frak K}_\lambda$ (see
3576: \scite{600-0.7}(1)). Alternating we prove below that
3577: given a $\lambda$-abstract elementary class ${\frak K}_\lambda$, the
3578: class of direct limits of members of ${\frak K}_\lambda$ is an
3579: a.e.c. ${\frak K}^{\text{up}}$.
3580: We show below $({\frak K}_\lambda)^{\text{up}} = {\frak K}$, hence
3581: ${\frak K}_\lambda$ determines ${\frak K}_{\ge \lambda}$.
3582: \endremark
3583: \bigskip
3584:
3585: \proclaim{\stag{600-0.31} Lemma} Suppose
3586: ${\frak K}_\lambda$ is a $\lambda$-abstract elementary class. \nl
3587: 1) The pair $(K',\le_{{\frak K}'})$ is an abstract elementary class with
3588: Lowenheim-Skolem number $\lambda$ which we denote also by
3589: ${\frak K}^{\text{up}}$ where we define
3590:
3591: $$
3592: \align
3593: K' = \biggl\{ M:&M \text{ is a } \tau_{{\frak K}_\lambda} \text{-model, and
3594: for some directed partial order} \tag"{$(a)$}" \\
3595: &I \text{ and } \bar M = \langle M_s:s \in I \rangle \text{ we have} \\
3596: &\qquad \qquad M = \dsize \bigcup_{s \in I} M_s \\
3597: &\qquad \qquad s \in I \Rightarrow M_s \in K_\lambda \\
3598: &I \models s < t \Rightarrow M_s \le_{{\frak K}_\lambda} M_t \biggr\}.
3599: \endalign
3600: $$
3601: \mn
3602: We call such $\langle M_s:s \in I \rangle$ a witness for $M \in K'$,
3603: we call it reasonable if $|I| \le \|M\|$
3604:
3605: $$
3606: \align
3607: M \le_{{\frak K}'} N \text{ \underbar{iff} } &\text{for some directed partial
3608: order } J,\text{ and} \tag"{$(b)$}" \\
3609: &\text{directed } I \subseteq J \text{ and } \langle M_s:s \in J \rangle
3610: \text{ we have} \\
3611: &M = \dsize \bigcup_{s \in I} M_s,N = \dsize \bigcup_{t \in J} M_t,
3612: M_s \in K_\lambda \text{ and} \\
3613: &J \models s < t \Rightarrow M_s \le_{{\frak K}_\lambda} M_t.
3614: \endalign
3615: $$
3616: \mn
3617: We call such $I, \langle M_s:s \in J \rangle$ witnesses for $M
3618: \le_{{\frak K}'} N$ or say $(I,J,\langle M_s:s \in J \rangle)$ witness
3619: $M \le_{{\frak K}'} N$. \nl
3620: 2) Moreover, $K'_\lambda = K_\lambda$ and $\le_{{\frak K}'_\lambda}$
3621: (which means $\le_{{\frak K}'} \restriction K'_\lambda$) is equal to
3622: $\le_{{\frak K}_\lambda}$ so $({\frak K}')_\lambda = {\frak K}_\lambda$.
3623: \nl
3624: 3) If ${\frak K}''$ is an abstract elementary class satisfying
3625: (see \scite{600-0.30D})
3626: $K''_\lambda = K_\lambda,<_{{\frak K}''} \restriction K_\lambda =
3627: \le_{{\frak K}_\lambda}$ and ${\text{\rm LS\/}}({\frak K}'') \le
3628: \lambda$ \ub{then}
3629: \footnote{if we assume in addition that
3630: $M \in {\frak K}'' \Rightarrow \|M\| \ge \lambda$ then we can show that
3631: equality holds} ${\frak K}''_{\ge \lambda} = {\frak K}'$. \nl
3632: 4) If ${\frak K}''$ is an a.e.c., $K_\lambda \subseteq K''_\lambda$
3633: and $\le_{{\frak K}_\lambda} = \le_{{\frak K}''} \restriction
3634: K_\lambda$,
3635: \ub{then} $K' \subseteq K''$ and $\le_{{\frak K}'} = \le_{{\frak K}''}
3636: \restriction K'$.
3637: \endproclaim
3638: \bigskip
3639:
3640: \demo{Proof} The proof of part (2) is straightforward (recalling
3641: \scite{600-0.6}) and part (3) follows
3642: from \scite{600-0.7} and part (4) is easier than part (3) (both statements
3643: being proved by induction on the cardinality of the relevant models)
3644: hence we concentrate on part (1). So let us check the axioms one by one.
3645: \enddemo
3646: \bigskip
3647: \noindent
3648: \underbar{Ax 0}: $K'$ is a class of $\tau$ models, $\le_{{\frak K}'}$
3649: a two-place relation on $K'$, both closed under isomorphisms. \newline
3650: [Why? trivially.]
3651: \bn
3652: \underbar{Ax I}: If $M \le_{{\frak K}'} N$ then $M \subseteq N$. \newline
3653: [Why? trivial.]
3654: \bn
3655: \underbar{Ax II}: $M_0 \le_{{\frak K}'} M_1 \le_{{\frak K}'} M_2$
3656: implies $M_0 \le_{{\frak K}'} M_2$ and $M \in K' \Rightarrow M
3657: \le_{{\frak K}'} M$. \newline
3658: [Why? The second phrase is trivial (as if $\bar M = \langle M_t:t \in
3659: I\rangle$ witness $M \in K'$ then $(I,I,\bar M)$ witness $M
3660: \le_{{\frak K}'} M$ above).
3661: For the first phrase let for $\ell \in
3662: \{1,2\}$ the directed partial orders $I_\ell \subseteq J_\ell$ and
3663: $\bar M^\ell = \langle M^\ell_s:s \in J_\ell \rangle$ witness
3664: $M_{\ell-1} \le_{{\frak K}'}M_\ell$ and let $\bar M^0 =
3665: \langle M^0_s:s \in I_0 \rangle$ witness $M_0 \in K'$.
3666: Now \wilog \, $\bar M^0$ is reasonable, i.e.
3667: $|I_0| \le \|M_0\|$, why? by
3668: \mr
3669: \item "{$\boxtimes_1$}" every $M \in K'$ has a reasonable witness,
3670: in fact, if $\bar M = \langle M_t:t \in I \rangle$ is a witness for
3671: $M$ then for some $I' \subseteq I$ of cardinality $\le \|M\|$ we have
3672: $\bar M \restriction I'$ is a reasonable witness for $M$.
3673: \nl
3674: [Why? If $\bar M = \langle M_t:t \in I \rangle$ is a witness, for each
3675: $a \in M$ choose $t_a \in I$ such that $a \in M_{t_a}$ and let
3676: $F:[I]^{< \aleph_0} \rightarrow I$ be such that $F(\{t_1,\dotsc,t_n\})$ is an
3677: upper bound of $\{t_1,\dotsc,t_n\}$ and let $J$ be the closure of
3678: $\{t_a:a \in M\}$ under $F$; now $\bar M \restriction J$ is a reasonable
3679: witness of $M \in K'$.]
3680: \ermn
3681: Similarly
3682: \mr
3683: \item "{$\boxtimes_2$}" if $(I,J,\langle M_s:s \in J\rangle$ witness
3684: $M \le_{{\frak K}'} N$ then for some directed $I' \subseteq I,|I'| \le
3685: \|M\|$ we have $(I',J,\langle M_s:s \in J\rangle)$ witness $M \le_{K'} N$
3686: \sn
3687: \item "{$\boxtimes_3$}" if $I,\bar M = \langle M_t:t \in J \rangle$
3688: witness $M \le_{{\frak K}'} N$ \ub{then} for some directed
3689: $J' \subseteq J$ we have $\|J'\| \le |I| + \|N\|,I \subseteq J'$ and
3690: $I,\bar M \restriction J'$ witness $M \le_{{\frak K}'} N$.
3691: \ermn
3692: Clearly $\boxtimes_1$ (and $\boxtimes_2,\boxtimes_3$) are
3693: cases of the LS-argument.
3694: We shall find a witness $(I,J,\langle M_s:s \in J \rangle$) for $M_0
3695: \le_{{\frak K}'} M_2$ such that $\langle M_s:s \in I \rangle = \langle M^0_s:
3696: s \in I_0 \rangle$ so $I=I_0$ and $|J| \le \|M_2\|$.
3697: This is needed for the proof of Ax III below.
3698: Without loss of generality $I_1,I_2$ has cardinality $\le \|M_0\|,
3699: \|M_1\|$ respectively, by the proof of $\boxtimes_2$. Also \wilog
3700: \, $\bar M^1,\bar M^1 \restriction I_1,\bar M^2,\bar M^2 \restriction
3701: I_2$ are reasonable as by the same argument we can have
3702: $|J_1| \le \|M_1\|,|J_2| \le \|M_2\|$ by $\boxtimes_3$.
3703:
3704: As $\langle M^0_s:s \in I_0 \rangle$
3705: is reasonable, there is a one-to-one
3706: function $h$ from $I_0$ into $M_2$ (and even $M_0$); the function $h$
3707: will be used to get that $J$ defined below is directed.
3708: We
3709: choose by induction on $m < \omega$, for every $\bar c \in {}^m(M_2)$, sets
3710: $I_{0,\bar c},I_{1,\bar c},I_{2,\bar c},J_{1,\bar c},J_{2,\bar c}$ such
3711: that:
3712: \mr
3713: \item "{$\otimes_1(a)$}" $I_{\ell,\bar c}$
3714: is a directed subset of $I_\ell$ of
3715: cardinality $\le \lambda$ for $\ell \in \{0,1,2\}$
3716: \sn
3717: \item "{$(b)$}" $J_{\ell,\bar c}$ is a directed subset of $J_\ell$ of
3718: cardinality $\le \lambda$ for $\ell \in \{1,2\}$
3719: \sn
3720: \item "{$(c)$}" $\dsize \bigcup_{s \in I_{\ell +1,\bar c}} M^{\ell + 1}_s =
3721: \bigl( \dsize \bigcup_{s \in J_{\ell + 1,\bar c}} M^{\ell + 1}_s \bigr) \cap
3722: M_\ell$ for $\ell = 0,1$
3723: \sn
3724: \item "{$(d)$}" $\dsize \bigcup_{s \in I_{0,\bar c}} M^0_s =
3725: (\dsize \bigcup_{s \in I_{1,\bar c}} M^1_s) \cap M_0$
3726: \sn
3727: \item "{$(e)$}" $\dsize \bigcup_{s \in J_{1,\bar c}} M^1_s =
3728: \dsize \bigcup_{s \in I_{2,\bar c}} M^2_s$
3729: \sn
3730: \item "{$(f)$}" $\bar c \subseteq \dsize \bigcup_{s \in J_{2,\bar c}}
3731: M^2_s$
3732: \sn
3733: \item "{$(g)$}" if $\bar d$ is a permutation of $\bar c$
3734: (i.e., letting $m = \ell g(\bar c)$ for some one
3735: to one $g:\{0,\dotsc,m-1\} \rightarrow \{0,\dotsc,m-1\}$ we have
3736: $d_\ell = c_{g(\ell)}$) \underbar{then} $I_{\ell,\bar c} =
3737: I_{\ell,\bar d},J_{m,\bar c} = J_{m,\bar d}$ \nl
3738: (for $\ell \in \{0,1,2\},m \in \{1,2\}$)
3739: \sn
3740: \item "{$(h)$}" if $\bar d$ is a subsequence of $\bar c$ (equivalently:
3741: an initial segment of some permutation of $\bar c$)
3742: \underbar{then} $I_{\ell,\bar d} \subseteq I_{\ell,\bar c},
3743: J_{m,\bar d} \subseteq J_{m,\bar c}$ for $\ell \in \{0,1,2\},m \in \{1,2\}$
3744: \sn
3745: \item "{$(i)$}" if $h(s) = c$ so $s \in I_0$ then $s \in I_{0,<c>}$.
3746: \ermn
3747: There is no problem to carry the definition by LS-argument recalling
3748: clauses (a) + (b) and $\|M^\ell_s\| = \lambda$ when $\ell=0 \wedge s \in
3749: I_0$ or $\ell=1 \wedge s \in J_1$ or $\ell=2 \wedge s \in J_2$.
3750: Without loss of generality
3751: $I_\ell \cap {}^{\omega >}(M_2) = \emptyset$.
3752:
3753: Now let $J$ have as set of elements $I_0 \cup \{\bar c:\bar c \text{ a finite
3754: sequence from }M_2\}$ ordered by: $J \models x \le y$ iff $I_0 \models x \le
3755: y$ \underbar{or} $x \in I_0,y \in J \backslash I_0,\exists z \in
3756: I_{0,y}[x \le_{I_0} z]$ \ub{or}
3757: $x,y \in J \backslash I_0$ and $x$ is an initial segment of a
3758: permutation of $y$ (or you may identify $\bar c$ with its set of
3759: permutations).
3760: \mn
3761: Let $I = I_0$. \newline
3762: Let $M_x$ be $M^0_x$ if $x \in I_0$ and $\dbcu_{s \in J_{2,x}}
3763: M^2_s$ if $x \in J \backslash I_0$. \nl
3764: Now
3765: \mr
3766: \item "{$(*)_1$}" $J$ is a partial order
3767: \nl
3768: [Clearly $x \le_J y \le_J x \Rightarrow x=y$, hence it is enough to
3769: prove transitivity.
3770: Assume $x \le_J y \le_J z$; if all three are in $I_0$ use ``$I_0$ is a
3771: partial order", if all three are not in $J \backslash I_0$, use
3772: the definition of
3773: the order. As $x' \le_J y' \in I_0 \Rightarrow x' \in I_0$ \wilog \,
3774: $x \in I_0,z \in J \backslash I_0$. If $y \in I_0$ then (as $y \le_J
3775: z$) for some $y',y \le_{I_0} y' \in I_{0,z}$ but $x \le_{I_0} y$ (as
3776: $x,y \in I_0,x \le_J y$) hence $x \le_{I_0} y' \in I_{0,z}$ so $x \le_J
3777: z$. If $y \notin I_0$ then $I_{0,y} \subseteq I_{0,z}$ (by clause
3778: (h)) so we can finish similarly. So we have covered all cases.]
3779: \sn
3780: \item "{$(*)_2$}" $J$ is directed
3781: \nl
3782: [Let $x,y \in J$ and we shall find a common upper bound. If $x,y
3783: \notin I_0$ their concatanation $x \char 94 y$ can serve. If $x,y
3784: \in I_0$ use ``$I_0$ is directed". If $x \in I_0,y \in J \backslash
3785: I_0$, then $\langle h(x) \rangle \in J \backslash I_0$ and $z = y
3786: \char 94 \langle h(x) \rangle \in J \backslash I_0$ is $<_J$ above $y$
3787: (by the choice of $\le_J$) and is $\le_J$-above $x$ as $x \in
3788: I_{0,\langle h(x)\rangle} \subseteq I_{0,z}$ by clause (i) of
3789: $\otimes_1$ so we are done. If $x \in J \backslash
3790: I_0,y \in J_0$ the dual proof works.]
3791: \sn
3792: \item "{$(*)_3$}" if $x \in J \backslash I_0$ then $M_x \cap M_\ell
3793: \le_{{\frak K}_x} M_x$ for $\ell=0,1$
3794: \nl
3795: [Why? Clearly $M_x \cap M_0 = \cup\{M^1_t:t \in I_{1,x}\}$ by clause
3796: (c) and $M_x \cap M_1 = \cup\{M^1_t:t \in J_{1,x}\}$ by clause (c),
3797: too. Now the sets $I_{1,x} \subseteq J_{1,x} (\subseteq J_1)$
3798: are directed by $\le_{J_1}$ so by the assumption on $\langle
3799: M^1_t:t \in J_1 \rangle$ and Lemma \scite{600-0.6} we have $M_x \cap M_0
3800: \le_{{\frak K}_\lambda} M_x \cap M_1$. Using $J_2$ we can similarly
3801: prove $M_x \cap M_1 \le_{{\frak K}_\lambda} M_x \cap M_2$ and trivially
3802: $M_x \cap M_2 = M_x$. As $\le_{{\frak K}_\lambda}$ is transitive
3803: we are done.]
3804: \sn
3805: \item "{$(*)_4$}" if $x \le_J y$ then $M_x \le_{{\frak K}_\lambda} M_y$
3806: \nl
3807: [Why? If $x,y \in I_0$ use the choice of $\langle M^0_s:s \in I_0
3808: \rangle$. If $x,y \in J \backslash I_0$ the proof is similar to that
3809: of $(*)_3$ using $J_2$. If $x \in I_0,y \in J \backslash I_0$ there
3810: is $s \in I_{0,y}$ such that $x \le_{I_0} s$, hence $M_x = M^0_x
3811: \le_{{\frak K}_\lambda} M^0_s$ and as $\langle M^0_t:t \in
3812: I_{0,y}\rangle$ is $\le_{{\frak K}_\lambda}$-directed clearly
3813: $M^0_s \le_{{\frak K}_\lambda} \cup
3814: \{M^0_t:t \in I_{0,y}\} = M_y \cap M_0$ and
3815: $M_y \cap M_0 \le_{{\frak K}_\lambda} M_y$ by $(*)_3$. By the
3816: transitivity of $\le_{{\frak K}_\lambda}$ we are done.]
3817: \sn
3818: \item "{$(*)_5$}" $\cup\{M_x:x \in I\} = \cup\{M^0_x:x \in I_0\} = M_0$
3819: \nl
3820: [Why? Trivially recalling $I_0 = I$.]
3821: \sn
3822: \item "{$(*)_6$}" $M_2 = \cup\{M_x:x \in J\}$
3823: \nl
3824: [Why? Trivially.]
3825: \ermn
3826: As $I_0 \subseteq J$ is directed by $(*)_1 + (*)_2 + (*)_4 + (*)_5 + (*)_6$ we
3827: have checked that $I,\langle M_x:x \in J \rangle$ witness $M_0
3828: \le_{{\frak K}'} M_2$. This completes the proof of AxII, but we also
3829: have proved
3830: \mr
3831: \item "{$\otimes_2$}" if $\bar M = \langle M_t:t \in I \rangle$ is a
3832: reasonable witness to $M \in K'$ and $M \le_{{\frak K}'} N \in K'$, \ub{then}
3833: there is a witness $I',\bar M' = \langle M'_t:t \in J' \rangle$ to $M
3834: \le_{{\frak K}'} N$ such that $I' = I,\bar M' \restriction I = \bar M$
3835: and $\bar M'$ is reasonable and $x \le_{J'} y \wedge y \in I'
3836: \Rightarrow x \in I'$.]
3837: \ermn
3838: \underbar{Ax III}: If $\theta$ is a regular cardinal, $M_i$ (for
3839: $i < \theta)$ is $\le_{{\frak K}'}$-increasing and continuous,
3840: \underbar{then} $M_0 \le_{{\frak K}'} \dsize \bigcup_{i < \theta} M_i$
3841: (in particular $\dsize \bigcup_{i < \theta} M_i \in {\frak K}'$). \newline
3842: [Why? Let $M_\theta = \dsize \bigcup_{i < \theta} M_i$, \wilog \, $\langle
3843: M_i:i < \theta \rangle$ is not eventually constant and so \wilog \, $i <
3844: \theta \Rightarrow M_i \ne M_{i+1}$ hence $\|M_i\| \ge |i|$; (this
3845: helps below to get ``reasonable", i.e. $|I_\ell| = \|M_i\|$ for limit $i$).
3846: We choose by
3847: induction on $i \le \theta$, a directed partial order $I_i$ and $M_s$ for
3848: $s \in I_i$ such that:
3849: \mr
3850: \item "{$\otimes_3(a)$}" $\langle M_s:
3851: s \in I_i \rangle$ witness $M_i \in K'$
3852: \smallskip
3853: \noindent
3854: \item "{$(b)$}" for $j < i,I_j \subseteq I_i$ and ($I_j,I_i,\langle M_s:s
3855: \in I_i \rangle$) witness $M_j \le_{{\frak K}'} M_i$
3856: \sn
3857: \item "{$(c)$}" $I_i$ is of cardinality $\le \|M_i\|$
3858: \sn
3859: \item "{$(d)$}" if $I_i \models s \le t$ and $j < i,t \in I_j$ then
3860: $s \in I_j$
3861: \ermn
3862: For $i=0$ use the definition of $M_0 \in K'$. \newline
3863: For $i$ limit let
3864: $I_i := \dsize \bigcup_{j < i} I_j$ (and the already defined
3865: $M_s$'s) are as required because $M_i = \dsize \bigcup_{j <i} M_j$ and the
3866: induction hypothesis (and $|I_i| \le \|M_i\|$ as we have assumed above that
3867: $j < i \Rightarrow M_j \ne M_{j+1}$) . \newline
3868: For $i=j+1$ use the proof of Ax.II above with $M_j,M_i,M_i,\langle M_s:s \in
3869: I_j \rangle$ here serving as $M_0,M_1,M_2,\langle M^0_j:s \in I_0 \rangle$
3870: there, that is, we use $\otimes_2$ from there.
3871: Now for $i = \theta,\langle M_s:s \in I_\theta \rangle$ witness
3872: $M_\theta \in K'$ and $(I_i,I_\theta,\langle M_s:s \in I_\theta \rangle)$
3873: witness $M_i \le_{{\frak K}'} M_\theta$ for each $i < \theta$.]
3874: \bn
3875: \ub{Axiom IV}: Assume $\theta$ is regular and $\langle M_i:i < \theta
3876: \rangle$ is $\le_{\frak K}$-increasingly continuous, $M \in K'$ and $i <
3877: \theta \Rightarrow M_i \le_{{\frak K}'} M$ and $M_\theta = \dbcu_{i < \theta}
3878: M_i$ (so $M_\theta \subseteq M$). \ub{Then} $M_\theta \le_{{\frak K}'} M$.
3879: \nl
3880: [Why? By the proof of Ax.III there are $\langle M_s:s \in I_i \rangle$ for
3881: $i < \theta$ satisfying clauses (a),(b),(c) and (d) of $\otimes_2$
3882: there and \wilog \, $I_i \cap \theta = \emptyset$. For each $i < \theta$
3883: as $M_i \le_{{\frak K}'} M$ there are $J_i$ and $M_s$ for $s \in J_i
3884: \backslash I_i$ such that $(I_i,J_i,\langle M_s:s \in J_i
3885: \rangle)$ witnesses it; \wilog \, with $\langle \dbcu_{i < \theta} I_i \rangle
3886: \char 94 \langle J_i \backslash I_i:i < \theta
3887: \rangle$ a sequence of pairwise disjoint sets; exist by $\otimes_2$ above.
3888: Let $I := \dbcu_{i < \theta} I_i$, let
3889: $\bold i:I \rightarrow \theta$ be $\bold i(s) = \text{ Min}\{i:s \in
3890: I_i\}$ and recall $|I| \le \|M_\theta\|$ hence by clause (d) of
3891: $\otimes_2$ we have $s \le_I t \Rightarrow \bold i(s) \le \bold i(t)$ and
3892: let $h$ be a one-to-one function from $I$ into $M_\theta$.
3893: Without loss of generality the union below is disjoint and let
3894: \mr
3895: \item "{$(*)_7$}" $J := I \cup
3896: \bigl\{(A,S):A \text{ a finite subset of } M \text{ and }
3897: S \text{ a finite } \text{ subset of } I
3898: \text{ with a maximal element} \bigr\}$.
3899: \ermn
3900: ordered by: $J \models x \le y$ \ub{iff} $x,y \in I,I \models x \le y$ or
3901: $x \in I,y = (A,S) \in J \backslash I$ and $x \in S$ or $x = (A^1,S^1) \in J
3902: \backslash I,y = (A^2,S^2) \in J \backslash I,A^1 \subseteq A^2,S^1 \subseteq
3903: S^2$.
3904: \sn
3905: We choose $N_y$ for $y \in J$ as follows:
3906: \sn
3907: If $y \in I$ we let $N_y = M_y$.
3908: \sn
3909: By induction on $n < \omega$, if
3910: $y = (A,S) \in J \backslash I$ saisfies $n = |A| + |S|$,
3911: we choose the objects $N_y,I_{y,s},J_{y,s}$ for $s \in S$ such that:
3912: \mr
3913: \item "{$\otimes_3(a)$}" $I_{y,s}$
3914: is a directed subset of $I_{\bold i(s)}$ of
3915: cardinality $\le \lambda$
3916: \sn
3917: \item "{$(b)$}" $J_{y,s}$ is a directed subset of $J_{\bold i(s)}$ of
3918: cardinality $\le \lambda$
3919: \sn
3920: \item "{$(c)$}" $s \in I_{\bold i(s)}$ for $s \in S$ (follows from
3921: the definition of $\bold i(s)$)
3922: \sn
3923: \item "{$(d)$}" $I_{y,s} \subseteq J_{y,s}$ for $s \in S$ and for $s <_I t$
3924: from $S$ we have $I_{y,s} \subseteq I_{y,t} \and J_{y,s} \subseteq J_{y,t}$
3925: \sn
3926: \item "{$(e)$}" if $y_1 = (A_1,S_1) \in J \backslash I,(A_1,S_1) <_J
3927: (A,S)$ and $s \in S_1$ then \nl
3928: $I_{y_1,s} \subseteq I_{y,s},J_{y_1,s} \subseteq J_{y,s}$
3929: \sn
3930: \item "{$(f)$}" $N_y = \dbcu_{t \in J_{y,s}} M_t$ for any $s \in S$
3931: \sn
3932: \item "{$(g)$}" $A \subseteq M_t$ for some $t \in J_{y,s}$ for any $s
3933: \in S$, hence $A \subseteq N_y$.
3934: \ermn
3935: No problem to carry the induction and check that $(I,J,\langle N_y:y \in
3936: J \rangle)$ witness $M_\theta \le_{{\frak K}'} M$.
3937: \bn
3938: \ub{Axiom V}: Assume $N_0 \le_{{\frak K}'} M$ and $N_1 \le_{{\frak K}'} M$.
3939: \nl
3940: If $N_0 \subseteq N_1$, then $N_0 \le_{{\frak K}'} N_1$. \nl
3941: [Why? Let $(I_0,J_0,\langle M^0_s:s \in J_0 \rangle)$ witness $N_0
3942: \le_{{\frak K}'} M$. Let $\langle M^1_s:s \in I_1 \rangle$ witness $N_1
3943: \in {\frak K}'$ and \wilog \, $I_1$ is isomorphic to
3944: $([N_1]^{<\aleph_0},\subseteq)$ and let $h_1$ be an isomorphism from $I_1$
3945: onto $([N_1]^{< \aleph_0},\subseteq)$.
3946: Now by induction on $n$, for $s \in I_1$ satisfying $n =
3947: |\{t:t <_{I_1} s\}|$ we choose directed
3948: subsets $F_0(s),F_1(s)$ of $I_0,I_1$ respectively, each of cardinality
3949: $\le \lambda$ such that:
3950: \mr
3951: \item "{$(i)$}" $h(s) \in F_1(s)$ and
3952: $t <_{I_1} s \Rightarrow F_0(t) \subseteq F_0(s) \and
3953: F_1(t) \subseteq F_1(s)$
3954: \sn
3955: \item "{$(ii)$}" $\bigcup\{M^0_t:t \in F_0(s)\} = \bigcup\{M^1_t:t
3956: \in F_1(s)\} \cap N_0$.
3957: \ermn
3958: Now replacing $M^1_s$ by $\cup\{M^1_t:t \in F_1(s)\}$ and letting
3959: $F=F_0$ we get:
3960: \mr
3961: \item "{$(iii)$}" $t \in I_1 \wedge s \in F(t)(\subseteq I_0)
3962: \Rightarrow M^0_s \subseteq M^1_t$
3963: \sn
3964: \item "{$(iv)$}" $F$ is a function from $I_1$ to $[I_0]^{\le \lambda}$
3965: \sn
3966: \item "{$(v)$}" for $s \in I_1,F(s)$ is a directed subset of
3967: $I_0$ of cardinality
3968: $\le \lambda,M^1_s \cap N_0 = \cup\{M^0_t:t \in F(s)\}$
3969: and $I_1 \models s \le t \Rightarrow F(s) \subseteq F(t)$.
3970: \ermn
3971: As $N_1 \le_{\frak K} M$ by the
3972: proof of Ax.II, i.e., by $\otimes_2$ above we can
3973: find $J_1$ and $M^1_s$
3974: for $s \in J_1 \backslash I_1$ such that
3975: $(I_1,J_1,\langle M^1_s:s \in J_1 \rangle)$ witnesses $N_1 \le_{{\frak K}'}
3976: M$. We now prove
3977: \mr
3978: \item "{$\boxtimes_4$}" if $r \in I_1,s \in I_0$ and $s \in F(r)$ then
3979: $M^0_s \le_{{\frak K}_\lambda} M^1_r$.
3980: \ermn
3981: [Why? As $\langle M^0_s:s \in J_0 \rangle,\langle M^1_s:s \in J_1
3982: \rangle$ are both witnesses for $M \in K'$, clearly for
3983: $r \in I_1$ we can find $J'_0(r) \subseteq J_0$
3984: directed of cardinality $\le \lambda$ and $J'_1(r) \subseteq J_1$ directed
3985: of cardinality $\le \lambda$ such that $r \in J'_1(r),F(r) \subseteq
3986: J'_0(r)$ and $\dbcu_{t \in J'_0(r)} M^0_t =
3987: \dbcu_{t \in J'_1(r)} M^1_t$, call it $M^*_r$.
3988: \nl
3989: Now $M^*_r \in K'_\lambda =
3990: K_\lambda$ (by part (2) and \scite{600-0.6}) and $t \in J'_1(r) \Rightarrow
3991: M^1_t \le_{{\frak K}_\lambda} M^*_r$ (as ${\frak K}_\lambda$ is
3992: a $\lambda$-abstract elementary class applying the parallel to
3993: observation \scite{600-0.6}, i.e., \scite{600-0.30D}(2)) and
3994: similarly $t \in J'_0(r) \Rightarrow M^0_t
3995: \le_{{\frak K}_\lambda} M^*_r$. Now the $s$ from $\boxtimes_4$ satisfied
3996: $s \in F(r) \subseteq J'_0(r)$ hence
3997: $M^0_s \subseteq M^1_r$ (why? by clause (iii) above $s \in F(r)$ is
3998: as required in $\boxtimes_4$). But above we got
3999: $M^0_s \le_{\frak K} M^*_r,M^1_r \le_{\frak K} M^*_r$, so
4000: by Ax.V for ${\frak K}_\lambda$ we have $M^0_s \le_{\frak K} M^1_r$ as
4001: required in $\boxtimes_4$.]
4002:
4003: Without loss of generality $I_0 \cap I_1 = \emptyset$ and define the
4004: partial order $J$ with set of elements
4005: $I_0 \cup I_1$ by $J \models x \le y$ iff $x,y \in I_0,I_0 \models
4006: x \le y$ or
4007: $x \in I_0,y \in I_1$ and $x \in F(y)$ or $x,y \in I_1,I_1 \models
4008: x \le y$. Recalling clause (c) above, it is a partial order.
4009: Define $M_s$ for $s \in J$ as $M^0_s$ if $s \in I_0$ and as
4010: $M^1_s$ if
4011: $s \in I_1$. Now check that $(I_0,J,\langle M_s:s \in J \rangle)$
4012: witnesses $N_0 \le_{{\frak K}'} N_1$ as required.]
4013: \bn
4014: \ub{Axiom VI}: LS$({\frak K}') = \lambda$. \nl
4015: [Why? Let $M \in K',A \subseteq M,|A| + \lambda \le \mu < \|M\|$ and let
4016: $\langle M_s:s \in J \rangle$ witness $M \in K'$. As $\|M\| > \mu$ we
4017: can choose a directed $I \subseteq J$ of cardinality $\le \mu$
4018: such that $A \subseteq M' := \dbcu_{s \in I} M_s$ and so $(I,J,
4019: \langle M_s:s \in J \rangle)$
4020: witnesses $M' \le_{{\frak K}'} M$, so as $A \subseteq M'$ and
4021: $\|M'\| \le |A| + \mu$
4022: we are done.] \hfill$\square_{\scite{600-0.31}}$
4023: \bn
4024: We may like to use ${\frak K}_{\le \lambda}$ instead of
4025: ${\frak K}_\lambda$; no need as essentially ${\frak K}$ consists of
4026: two parts ${\frak K}_{\le \lambda}$ and ${\frak K}_{\ge \lambda}$
4027: which have just to agree in $\lambda$. That is
4028: \proclaim{\stag{600-0.32} Claim} Assume
4029: \mr
4030: \item "{$(a)$}" ${\frak K}^1$ is an abstract elementary class with
4031: $\lambda = {\text{\rm LS\/}}({\frak K}^1),K^1 = K^1_{\ge \lambda}$
4032: \sn
4033: \item "{$(b)$}" ${\frak K}^2_{\le \lambda}$ is a $(\le \lambda)$-abstract
4034: elementary class (defined as in \scite{600-0.30A}(1) with the obvious changes so
4035: $M \in {\frak K}^2_{\le \lambda} \Rightarrow \|M\| \le \lambda$ and
4036: in Axiom III, $\|\dbcu_i M_i\| \le \lambda$ is required)
4037: \sn
4038: \item "{$(c)$}" $K^2_\lambda = K^1_\lambda$ and $\le_{{\frak K}^2}
4039: \restriction K^2_\lambda = \le_{{\frak K}^1} \restriction K^1_\lambda$
4040: \sn
4041: \item "{$(d)$}" we define ${\frak K}$ as follows:
4042: $K = K^1 \cup K^2,M \le_{\frak K} N$
4043: iff $M \le_{{\frak K}^1} N$ or $M \le_{{\frak K}^2} N$ or for some $M',M
4044: \le_{{\frak K}^2} M' \le_{{\frak K}^1} N$.
4045: \ermn
4046: \ub{Then} ${\frak K}$ is an abstract elementary class and ${\text{\rm
4047: LS\/}}({\frak K}) = {\text{\rm LS\/}}({\frak K}^2)$ which trivially is
4048: $\le \lambda$.
4049: \endproclaim
4050: \bigskip
4051:
4052: \demo{Proof} Straight. E.g. \nl
4053: \ub{Axiom V}: We shall use freely
4054: \mr
4055: \item "{$(*)$}" ${\frak K}_{\le \lambda} = {\frak K}^2$ and ${\frak K}
4056: _{\ge \lambda} = {\frak K}^1$.
4057: \ermn
4058: So assume $N_0 \le_{\frak K} M,N_1 \le_{\frak K} M,N_0
4059: \subseteq N_1$. \nl
4060: Now if $\|N_0\| \ge \lambda$ use assumption (a), so we can assume
4061: $\|N_0\| < \lambda$. If $\|M\| \le \lambda$ we can use assumption (b) so we
4062: can assume $\|M\| > \lambda$ and by the definition of $\le_{\frak K}$ there
4063: is $M'_0 \in K^1_\lambda = K^2_\lambda$ such that $N_0 \le_{{\frak K}^2} M'_0
4064: \le_{{\frak K}^1} M$. First assume
4065: $\|N_1\| \le \lambda$, so we can find $M'_1 \in K^1_\lambda$ such that
4066: $N_1 \le_{{\frak K}^2} M'_1 \le_{{\frak K}^1} M$
4067: (why? if $N_1 \in
4068: K_{< \lambda}$, by the definition of $\le_{\frak K}$ and if
4069: $N_1 \in K_\lambda$ just choose $M'_1 = N_1$). Now we can by assumption (a)
4070: find $M'' \in K^1_\lambda$ such that $M'_0 \cup M'_1 \subseteq M''
4071: \le_{{\frak K}^1} M$, hence by assumption (a) (i.e. AxV for ${\frak K}^1$)
4072: we have $M'_0 \le_{{\frak K}^1} M'',M'_1 \le_{{\frak K}^1} M''$, so by
4073: assumption (c) we have $M'_0 \le_{{\frak K}^2} M'',M'_1
4074: \le_{{\frak K}^2} M''$. As $N_0 \le_{{\frak K}^2} M'_0
4075: \le_{{\frak K}^2} M'' \in K_{\le \lambda}$
4076: by assumption (b) we have $N_0
4077: \le_{{\frak K}^2} M''$, and similarly we have $N_1 \le_{{\frak K}^2}
4078: M''$. So $N_0 \subseteq N_1,N_0 \le_{{\frak K}^2} M'',
4079: N_1 \le_{{\frak K}^2} M'$ so by
4080: assumption (b) we have $N_0 \le_{{\frak K}^2} N_1$ hence
4081: $N_0 \le_{\frak K} N_1$.
4082:
4083: We are left with the case $\|N_1\| > \lambda$; by assumption (a) there is
4084: $N'_1 \in K_\lambda$ such that $N_0 \subseteq N'_1 \le_{{\frak K}^1} N_1$. By
4085: assumption (a) we have $N'_1 \le_{{\frak K}^1} M$, so by the previous paragraph
4086: we get $N_0 \le_{{\frak K}^2} N'_1$,
4087: together with the previous sentence we have
4088: $N_0 \le_{{\frak K}^2} N'_1 \le_{{\frak K}^1} N_1$ so by the definition of
4089: $\le_{\frak K}$ we are done. \hfill$\square_{\scite{600-0.32}}$
4090: \enddemo
4091: \bn
4092: Recall
4093: \definition{\stag{600-0.33} Definition} If $M \in K_\lambda$ is locally
4094: superlimit or just pseudo superlimit
4095: let $K_{[M]} = K^{[M]}_\lambda = \{N \in K_\lambda:N \cong M\},{\frak
4096: K}_{[M]} = {\frak K}^{[M]}_\lambda =
4097: (K_{[M]},\le_{\frak K} \restriction K^{[M]}_\lambda)$ and let
4098: ${\frak K}^{[M]}$ be the
4099: ${\frak K}'$ we get in \scite{600-0.31}(1) for ${\frak K} =
4100: {\frak K}_{[M]} = {\frak K}^{[M]}_\lambda$.
4101: We may write ${\frak K}_\lambda[M],{\frak K}[M]$.
4102: \enddefinition
4103: \bn
4104: Trivially
4105: \proclaim{\stag{600-0.34} Claim} 1) If ${\frak K}$ is an $\lambda$-{\rm a.e.c.},
4106: $M \in K_\lambda$ is locally superlimit or just pseudo superlimit
4107: \ub{then} ${\frak K}_{[M]}$
4108: is a $\lambda$-{\rm a.e.c.} which
4109: is categorical (i.e. categorical in $\lambda$).
4110: \nl
4111: 2) Assume ${\frak K}$ is an a.e.c. and $M \in {\frak K}_\lambda$ is
4112: not $\le_{\frak K}$- maximal. $M$ is pseudo superlimit (in ${\frak
4113: K}$, i.e., in ${\frak K}_\lambda$) \ub{iff} ${\frak K}_{[M]}$
4114: is a $\lambda$-a.e.c. which is categorical \ub{iff} ${\frak K}^{[M]}$
4115: is an a.e.c., categorical in $\lambda$ and $\le_{{\frak K}^{[M]}} =
4116: \le_{\frak K} \restriction K^{[M]}$.
4117: \nl
4118: 3) In (1) and (2), {\rm LS}$({\frak K}^{[M]}) = \lambda = \text{\rm Min}
4119: \{\|N\|:N \in {\frak K}^{[M]}\}$.
4120: \endproclaim
4121: \bigskip
4122:
4123: \demo{Proof} Straightforward. \hfill$\square_{\scite{600-0.34}}$
4124: \enddemo
4125: \bn
4126: \margintag{600-0.37.3}\ub{\stag{600-0.37.3} Exercise}: Assume ${\frak K}$ is a
4127: $\lambda$-a.e.c. with amalgamation and stability in $\lambda$.
4128: \ub{Then} for every $M_1 \in K_\lambda,p_1 \in {\Cal S}_{\frak K}(M_1)$
4129: we can find $M_2 \in K$ and minimal $p_2 \in {\Cal S}_{\frak
4130: K}(M_2)$ such that $M_1 \le_{\frak K} M_2,p_1 = p_2 \restriction M_1$.
4131: \bn
4132: \margintag{600-1c.29}\ub{\stag{600-1c.29} Exercise}: 1) Any $\le_{{\frak K}_\lambda}$-embedding $f_0$ of
4133: $M^1_0$ into $M^2_0$ can be extended to an isomorphism $f$ from
4134: $M^1_\delta$ onto $M^2_\delta$ such that $f(M^1_{2 \alpha})
4135: \le_{{\frak K}_\lambda} M^2_{2 \alpha},f^{-1}(M^2_{2 \alpha +1})
4136: \le_{{\frak K}_\lambda} M^1_{2 \alpha +1}$ for every $\alpha <
4137: \delta$, \ub{provided that}
4138: \mr
4139: \item "{$\circledast$}" $(a) \quad {\frak K}_\lambda$ is a
4140: $\lambda$-a.e.c. and $\delta$ is a limit ordinal $\le \lambda^+$
4141: \sn
4142: \item "{${{}}$}" $(b) \quad \langle M^\ell_\alpha:\alpha \le
4143: \delta\rangle$ is $\le_{{\frak K}_\lambda}$-increasing continuous for
4144: $\ell=1,2$
4145: \sn
4146: \item "{${{}}$}" $(c) \quad M^\ell_\alpha$ is an amalgamation base in
4147: ${\frak K}_\lambda$ (for $\alpha < \delta$ and $\ell=1,2$)
4148: \sn
4149: \item "{${{}}$}" $(d) \quad M^\ell_{\alpha +1}$ is $\le_{{\frak
4150: K}_\lambda}$-universal extension of $M^\ell_\alpha$ for $\alpha <
4151: \delta,\ell=1,2$.
4152: \ermn
4153: 2) Write the axioms of ``a $\lambda$-a.e.c." which are used.
4154: [Hint: Should be clear, and the argument appear.]
4155: \newpage
4156:
4157: \head {\S2 Good Frames} \endhead \resetall \sectno=2
4158: \spuriousreset
4159: \bigskip
4160:
4161: We first present our central definition: good $\lambda$-frame (in
4162: Definition \scite{600-1.1}). We are given the relation
4163: ``$p \in {\Cal S}(N)$ does not fork over $M \le_{\frak K} N$ when $p$
4164: is basic" (by the basic relations
4165: and axioms) so it is natural to look at how well we can ``lift" the
4166: definition of non-forking to models of cardinality $\lambda$ and later
4167: to non-forking of models (and types over them) in
4168: cardinalities $> \lambda$. Unlike the lifting of $\lambda$-a.e.c. in
4169: Lemma \scite{600-0.31}, life is not so easy.
4170: We define in \scite{600-1.6}, \scite{600-1.7},
4171: \scite{600-1.9} and we prove basic properties in \scite{600-1.8},
4172: \scite{600-1.10}, \scite{600-1.12} and less obvious ones in \scite{600-1.11},
4173: \scite{600-1.13}, \scite{600-1.13B}. This should serve as a reasonable exercise in the
4174: meaning of good frames; however, the lifting, in general, does not
4175: give good $\mu$-frames for $\mu > \lambda$. There may be no $M \in
4176: K_\mu$ at all and/or amalgamation may fail. Also the existence and
4177: uniqueness of non-forking types is problematic. We do not give up and
4178: will return to the lifting problem, under additional assumptions in
4179: \sectioncite[\S12]{705} and \cite{Sh:842}.
4180:
4181: In \scite{600-1.14} (recalling \scite{600-0.34}) we show that the case
4182: ``${\frak K}^{\frak s}$ categorical in $\lambda$" is not so rare among
4183: good $\lambda$-frames; in fact if there is a superlimit model in
4184: $\lambda$ we can restrict ${\frak K}_\lambda$ to it. So in a sense
4185: superstability and categoricity are close.
4186: For elementary classes they are less close and note that if $T$ is a
4187: complete first order superstable theory and $\lambda \ge 2^{|T|}$,
4188: \ub{then} the class ${\frak K} = {\frak K}_{T,\lambda}$ of
4189: $\lambda$-saturated model of $T$ is in general not an elementary class
4190: (though is a PC$_\lambda$ class) but is an a.e.c. categorical in
4191: $\lambda$ and for some good $\lambda$-frame ${\frak s},K_{\frak s} =
4192: {\frak K}_{T,\lambda}$. How justified is our restriction here to
4193: something like ``the $\lambda$-saturated model"? It is O.K. for our
4194: test problems but more so it is justified as
4195: our approach is to first analyze the quite saturated models.
4196:
4197: Last but not least in \scite{600-1.15} we show that
4198: one of the axioms from \scite{600-1.1},
4199: i.e., (E)(i), follows from the rest in our present definition;
4200: additional implications are in Claims \scite{600-1.16A}, \scite{600-1.16B}. Later
4201: ``Ax(X)(y)" will mean (X)(y) from Definition \scite{600-1.1}.
4202: \bn
4203: Recall that good $\lambda$-frame is intended to be a parallel to (bare
4204: bones) superstable
4205: elementary class stable in $\lambda$; here we restrict
4206: ourselves to models of cardinality $\lambda$.
4207: \definition{\stag{600-1.1} Definition} We say ${\frak s} =
4208: ({\frak K},\nonfork{}{}_\lambda,
4209: {\Cal S}^{\text{bs}}_\lambda) = ({\frak K}^{\frak s},\nonfork{}{}_{\frak s},
4210: {\Cal S}^{\text{bs}}_{\frak s})$ is a good frame in $\lambda$ or a
4211: good $\lambda$-frame ($\lambda$ may be omitted when clear,
4212: note that $\lambda = \lambda_{\frak s} =
4213: \lambda({\frak s})$ is determined by ${\frak s}$ and we may write
4214: ${\Cal S}_{\frak s}(M)$ instead of ${\Cal S}_{{\frak K}^{\frak s}}(M)$
4215: and \ortp$_{\frak s}(a,M,N)$ instead of \ortp$_{{\frak K}^{\frak s}}(a,M,N)$
4216: when $M \in K^{\frak s}_\lambda,N \in K^{\frak s}$; we may write \ortp$(a,M,N)$
4217: for \ortp$_{{\frak K}^{\frak s}}(a,M,N)$)
4218: \ub{when} the following conditions hold:
4219: \mr
4220: \widestnumber\item{$(D)(a)$}
4221: \item "{$(A)$}" ${\frak K} = (K,\le_{\frak K})$ is an abstract elementary
4222: class also denoted by ${\frak K}[{\frak s}]$,
4223: the
4224: L\"owenheim Skolem number of ${\frak K}$, being $\le \lambda$
4225: (see Definition \scite{600-0.2});
4226: there is no harm in assuming $M \in K \Rightarrow \|M\| \ge \lambda$;
4227: let ${\frak K}_{\frak s} = {\frak K}^{\frak s}_\lambda$ and
4228: $\le_{\frak s} = \le_{\frak K} \restriction K_\lambda$, and let
4229: ${\frak K}_{\frak s} = (K_\lambda,\le_{\frak s})$ and ${\frak K}[{\frak s}] =
4230: {\frak K}^{\frak s}$ so we may write ${\frak s} = ({\frak K}_{\frak
4231: s},\nonfork{}{}_{\frak s},{\Cal S}^{\text{bs}}_{\frak s})$
4232: \sn
4233: \item "{$(B)$}" ${\frak K}$ has a superlimit model in $\lambda$ which
4234: \footnote{in fact, the ``is not $<_{\frak K}$-maximal" follows by (C)} is
4235: not $<_{\frak K}$-maximal.
4236: \sn
4237: \item "{$(C)$}" ${\frak K}_\lambda$ has the amalgamation property, the
4238: JEP (joint embedding property), and has no $\le_{\frak K}$-maximal
4239: member.
4240: \sn
4241: \item "{$(D)(a)$}" ${\Cal S}^{\text{bs}} = {\Cal S}^{\text{bs}}_\lambda$
4242: (the class of basic types for ${\frak K}_\lambda$) is included in \newline
4243: $\bigcup\{{\Cal S}(M):M \in K_\lambda\}$ and is closed under isomorphisms
4244: including automorphisms;
4245: for $M \in K_\lambda$ let ${\Cal S}^{\text{bs}}(M) =
4246: {\Cal S}^{\text{bs}} \cap {\Cal S}(M)$; no harm in
4247: allowing types of finite sequences, i.e., replacing ${\Cal S}(M)$ by
4248: ${\Cal S}^{< \omega}(M)$, (${\Cal S}^\omega(M))$ is different as being
4249: new (= non-algebraic) is not preserved under increasing unions).
4250: \sn
4251: \item "{${}(b)$}" if $p \in {\Cal S}^{\text{bs}}(M)$, \ub{then} $p$ is
4252: non-algebraic (i.e. not realized by any $a \in M$).
4253: \sn
4254: \item "{${}(c)$}" \underbar{(density)} \newline
4255: if $M \le_{\frak K} N$ are from $K_\lambda$ and $M \ne N$, \underbar{then}
4256: for some $a \in N \backslash M$ we have $\text{\ortp}(a,M,N) \in
4257: {\Cal S}^{\text{bs}}$
4258: \newline
4259: \beginaside [intention: examples are:
4260: minimal types in \cite{Sh:576}, regular types for superstable first
4261: order (= elementary) classes].
4262: \sn
4263: \endaside
4264: \item "{${}(d)$}" \ub{bs-stability} \nl
4265: ${\Cal S}^{\text{bs}}(M)$ has cardinality $\le \lambda$ for $M \in
4266: K_\lambda$.
4267: \sn
4268: \item "{$(E)(a)$}" $\nonfork{}{}_{\lambda}$ denoted also by
4269: $\nonfork{}{}_{\frak s}$ or just $\nonfork{}{}_{}$, is a four
4270: place relation
4271: \footnote{we tend to forget to write the $\lambda$, this is justified by
4272: \scite{600-1.8}(2), and see Definition \scite{600-1.7}} called non-forking with
4273: $\nonfork{}{}_{}(M_0,M_1,a,M_3)$
4274: implying $M_0 \le_{\frak K} M_1 \le_{\frak K} M_3$ are from $K_\lambda,
4275: a \in M_3 \backslash M_1$ and $\text{\ortp}(a,M_0,M_3) \in
4276: {\Cal S}^{\text{bs}}(M_0)$ and \newline
4277: $\text{\ortp}(a,M_1,M_3) \in {\Cal S}^{\text{bs}}(M_1)$.
4278: Also $\nonfork{}{}_{}$ is preserved under isomorphisms and we
4279: demand: if $M_0 = M_1
4280: \le_{\frak K} M_3$ both in $K_\lambda$ and $a \in M_3$, then: \newline
4281: $\nonfork{}{}_{}(M_0,M_1,a,M_3)$ is equivalent to
4282: ``\ortp$(a,M_0,M_3) \in {\Cal S}^{\text{bs} }(M_0)$". The assertion
4283: $\nonfork{}{}_{}(M_0,M_1,a,M_3)$ is also written as
4284: $\nonforkin{M_1}{a}_{M_0}^{M_3}$ and also
4285: as ``\ortp$(a,M_1,M_3)$ does not fork
4286: over $M_0$ (inside $M_3$)" (this is justified by clause (b) below).
4287: So \ortp$(a,M_1,M_3)$ forks over $M_0$ (where $M_0 \le_{\frak s} M_1
4288: \le_{\frak s} M_3,a \in M_3$) is just the negation
4289: \nl
4290: \beginaside
4291: [Explanation: The
4292: intention is to axiomatize non-forking of types, but we already
4293: commit ourselves to dealing with basic types only. Note that in
4294: \cite{Sh:576} we know something on minimal types but other types are
4295: something else.]
4296: \sn
4297: \endaside
4298: \item "{${}(b)$}" \underbar{(monotonicity)}: \newline
4299: if $M_0 \le_{\frak K} M'_0 \le_{\frak K} M'_1 \le_{\frak K}
4300: M_1 \le_{\frak K} M_3 \le_{\frak K} M'_3,M_1 \cup \{a\}
4301: \subseteq M''_3 \le_{\frak K} M'_3$
4302: all of them in $K_\lambda$, \ub{then}
4303: $\nonfork{}{}_{}(M_0,M_1,a,M_3) \Rightarrow
4304: \nonfork{}{}_{}(M'_0,M'_1,a,M'_3)$ and $\nonfork{}{}_{}(M'_0,M'_1,a,M'_3)
4305: \Rightarrow \nonfork{}{}_{}
4306: (M'_0,M'_1,a,M''_3)$, \ub{so} it is legitimate to just
4307: say ``$\text{\ortp}(a,M_1,M_3)$
4308: does not fork over $M_0$". \nl
4309: \beginaside
4310: [Explanation: non-forking is preserved by decreasing the type,
4311: increasing the basis (= the set over which it does not fork) and
4312: increasing or decreasing
4313: the model inside which all this occurs. The same holds for
4314: stable theories only here we restrict ourselves to ``legitimate",
4315: i.e., basic types. But note that here the ``restriction of
4316: \ortp$(a,M_1,M_3)$ to $M'_1$ is basic" is worthwhile information.]
4317: \endaside
4318: \sn
4319: \item "{${}(c)$}" \underbar{(local character)}: \nl
4320: if $\langle M_i:i \le \delta + 1 \rangle$ is $\le_{\frak K}$-increasing
4321: continuous in ${\frak K}_\lambda,a \in M_{\delta + 1}$ and \newline
4322: \ortp$(a,M_\delta,M_{\delta +1}) \in {\Cal S}^{\text{bs}}
4323: (M_\delta)$ \underbar{then}
4324: for every $i < \delta$ large enough \ortp$(a,M_\delta,M_{\delta +1})$
4325: does not fork over $M_i$. \nl
4326: \beginaside
4327: [Explanation: This is a replacement for superstability which says
4328: that: if $p \in {\Cal S}(A)$ then there is a finite
4329: $B \subseteq A$ such that $p$ does not fork over $B$.]
4330: \endaside
4331: \sn
4332: \item "{${}(d)$}" \underbar{(transitivity)}: \newline
4333: if $M_0 \le_{\frak s} M'_0 \le_{\frak s} M''_0 \le_{\frak s} M_3$ are
4334: from $K_\lambda$ and
4335: $a \in M_3$ and \ortp$(a,M''_0,M_3)$ does not fork over $M'_0$ and
4336: \ortp$(a,M'_0,M_3)$ does not fork over $M_0$ (all models are in $K_\lambda$, of
4337: course, and necessarily the three relevant types are in
4338: ${\Cal S}^{\text{bs}}$), \ub{then}
4339: \ortp$(a,M''_0,M_3)$ does not fork over $M_0$
4340: \smallskip
4341: \noindent
4342: \item "{${}(e)$}" \underbar{uniqueness}: \newline
4343: if $p,q \in {\Cal S}^{\text{bs}}(M_1)$ do not fork over
4344: $M_0 \le_{\frak K} M_1$ (all in $K_\lambda$) and \newline
4345: $p \restriction M_0 = q \restriction M_0$ \underbar{then} $p = q$
4346: \sn
4347: \item "{${}(f)$}" \underbar{symmetry}: \newline
4348: if $M_0 \le_{\frak K} M_3$ are in ${\frak K}_\lambda$ and for $\ell = 1,2$
4349: we have \newline
4350: $a_\ell \in M_3$ and $\text{\ortp}(a_\ell,M_0,M_3) \in
4351: {\Cal S}^{\text{bs}}(M_0)$, \ub{then} the following are equivalent:
4352: \smallskip
4353: \noindent
4354: {\roster
4355: \itemitem{ $(\alpha)$ } there are $M_1,M'_3$ in $K_\lambda$ such that
4356: $M_0 \le_{\frak K} M_1 \le_{\frak K} M'_3$, \newline
4357: $a_1 \in M_1,M_3 \le_{\frak K} M'_3$ and
4358: $\text{\ortp}(a_2,M_1,M'_3)$ does not fork over $M_0$
4359: \smallskip
4360: \noindent
4361: \itemitem{ $(\beta)$ } there are $M_2,M'_3$ in $K_\lambda$ such that
4362: $M_0 \le_{\frak K} M_2 \le_{\frak K} M'_3$, \newline
4363: $a_2 \in M_2,M_3 \le_{\frak K} M'_3$ and $\text{\ortp}(a_1,M_2,M'_3)$
4364: does not fork over $M_0$.
4365: \endroster}
4366: \beginaside
4367: [Explanation: this is a replacement to ``\ortp$(a_1,M_0 \cup
4368: \{a_2\},M_3)$ forks over $M_0$ iff \ortp$(a_2,M_0 \cup \{a_1\},M_3)$
4369: forks over $M_0$" which is not well defined in our context.]
4370: \endaside
4371: \sn
4372: \item "{${}(g)$}" \underbar{extension existence}: \newline
4373: if $M \le_{\frak K} N$ are from $K_\lambda$ and $p \in
4374: {\Cal S}^{\text{bs}}(M)$ \ub{then} some
4375: $q \in {\Cal S}^{\text{bs}}(N)$ does not fork over $M$ and extends
4376: $p$
4377: \sn
4378: \item "{${}(h)$}" \ub{continuity}: \newline
4379: if $\langle M_i:i \le \delta \rangle$ is $\le_{\frak K}$-increasing
4380: continuous, all in $K_\lambda$ (recall $\delta$ is always a limit
4381: ordinal), $p \in {\Cal S}(M_\delta)$ and $i < \delta
4382: \Rightarrow p \restriction M_i \in {\Cal S}^{\text{bs}}(M_i)$
4383: does not fork over
4384: $M_0$ \underbar{then} $p \in {\Cal S}^{\text{bs}}(M_\delta)$ and moreover $p$
4385: does not fork over $M_0$. \nl
4386: \beginaside
4387: [Explanation: This is a replacement to: for an increasing sequence of types
4388: which do not fork over $A$, the union does not fork over $A$;
4389: equivalently if $p$ forks over $A$ then some finite subtype does.]
4390: \endaside
4391: \sn
4392: \item "{${}(i)$}" \ub{non-forking amalgamation}: \newline
4393: if for $\ell = 1,2,M_0 \le_{\frak K} M_\ell$ are from $K_\lambda,a_\ell \in
4394: M_\ell \backslash M_0$, \ortp$(a_\ell,M_0,M_\ell) \in {\Cal S}^{bs}(M_0)$,
4395: \ub{then} we can find $f_1,f_2,M_3$ satisfying $M_0 \le_{\frak K} M_3 \in
4396: K_\lambda$ such that for $\ell =1,2$ we have $f_\ell$ is a
4397: $\le_{\frak K}$-embedding of $M_\ell$ into $M_3$ over
4398: $M_0$ and \ortp$(f_\ell(a_\ell),f_{3-\ell}(M_{3-\ell}),M_3)$
4399: does not fork over $M_0$ for $\ell=1,2$.
4400: \beginaside
4401: [Explanation: This strengthens clause (g), (existence) saying we can
4402: do it twice so close to (f), symmetry, see \scite{600-1.15}.]
4403: \endaside
4404: \endroster
4405: \enddefinition
4406: \bigskip
4407: \centerline {$* \qquad * \qquad *$}
4408: \bn
4409: \margintag{600-1.1B}\ub{\stag{600-1.1B} Discussion}: 0) On connections between the axioms see
4410: \scite{600-1.15}, \scite{600-1.16A}, \scite{600-1.16B}.
4411: \nl
4412: 1) What
4413: justifies the choice of the good $\lambda$-frame as a parallel to
4414: (bare bones) superstability? Mostly starting from assumptions
4415: on few models around $\lambda$ in the a.e.c. ${\frak K}$ and reasonable,
4416: ``semi ZFC" set theoretic assumptions (e.g. involving categoricity and
4417: weak cases of G.C.H., see \S3)
4418: we can prove that, essentially, for some $\nonfork{}{}_{},
4419: {\Cal S}$ the demands in Definition \scite{600-1.1} hold.
4420: So here we get (i.e., applying our general theorem to the case of
4421: \scite{600-Ex.1}) an alternative proof of the main theorem of \cite{Sh:87a},
4422: \cite{Sh:87b} in a local version, i.e., dealing
4423: with few cardinals rather than having to deal with all the cardinals
4424: $\lambda,\lambda^{+1},\lambda^{+2},\dotsc,\lambda^{+n}$ as in
4425: \cite{Sh:87a}, \cite{Sh:87b} in an inductive proof.
4426: That is, in \cite{Sh:87b}, we get dichotomies by the omitting type
4427: theorem for countable models (and theories). So problems on
4428: $\aleph_n$ are ``translated" down to $\aleph_{n-1}$ (increasing the
4429: complexity) till we arrive to $\aleph_0$ and then ``translated" back.
4430: Hence it is important there to deal with $\aleph_0,\dotsc,\aleph_n$
4431: together. Here our $\lambda$ may not have special helpful properties,
4432: so if we succeed to prove the relevant claims then they apply to
4433: $\lambda^+$, too. There are advantages to being poor.
4434: \nl
4435: 2) Of course, we may just point out that the axioms seem reasonable
4436: and that eventually we can say much more.
4437: \nl
4438: 3) We may consider weakening bs-stability
4439: (i.e., Ax$(D)(d)$ in Definition \scite{600-1.1}) to
4440: $M \in K_\lambda \Rightarrow |{\Cal S}^{\text{bs}}(M)| \le
4441: \lambda^+$, we just have not looked into it here; Jarden-Shelah
4442: \cite{JrSh:875} will; actually \chaptercite{88r} deals in a limited way
4443: with this in a more restricted framework. \nl
4444: 4) On stability in $\lambda$ and existence of $(\lambda,\sigma)$-brimmed
4445: extensions see \scite{600-4a.1}. \nl
4446: \bn
4447: \relax From the rest of this section we shall use mainly the defintion of
4448: $K^{3,\text{bs}}_\lambda$ in Definition \scite{600-1.6}(3), also
4449: \scite{600-1.16} (restricting ourselves to a superlimit). We sometimes
4450: use implications among the axioms (in \scite{600-1.15} - \scite{600-1.16B}).
4451: The rest is, for now
4452: an exercise to familiarize the reader with $\lambda$-frames, in
4453: particular (\scite{600-1.5}-\scite{600-1.14}) to see what occurs to
4454: non-forking and basic types in cardinals $> \lambda$. This is
4455: easy (but see below).
4456: For this we first present the basic definitions.
4457: \medskip
4458:
4459: \demo{\stag{600-1.5} Convention} 1) We fix ${\frak s}$, a good
4460: $\lambda$-frame so $K = K^{\frak s},{\Cal S}^{\text{bs}} = {\Cal
4461: S}^{\text{bs}}_{\frak s}$.
4462: \nl
4463: 2) By $M \in K$ we mean $M \in K_{\ge \lambda}$ if not said
4464: otherwise.
4465: \enddemo
4466: \bigskip
4467:
4468: We lift the properties to ${\frak K}_{\ge \lambda}$ by reflecting to the
4469: situation in $K_\lambda$. But do not be too excited: the good properties
4470: do not lift automatically, we shall be working on that later (under
4471: additional assumptions). Of course, from the definition below later
4472: we shall use mainly $K^{3,\text{bs}}_{\frak s} = K^{3,\text{bs}}_\lambda$.
4473: \definition{\stag{600-1.6} Definition} 1)
4474:
4475: $$
4476: \align
4477: K^{3,\text{bs}} = K^{3,\text{bs}}_{\ge {\frak s}} :=
4478: \biggl\{ (M,N,a):&M \le_{\frak K} N,a \in N \backslash M
4479: \text{ and there is } M' \le_{\frak K} M \\
4480: &\text{satisfying } M' \in K_\lambda,\text{ such that for every }
4481: M'' \in K_\lambda\text{ we have:}\\
4482: &[M' \le_{\frak K} M'' \le_{\frak K} M \Rightarrow \text{ \ortp}
4483: (a,M'',N) \in {\Cal S}^{\text{bs}}(M'') \\
4484: &\text{does not fork over } M']; \text{ equivalently }
4485: [M' \le_{\frak K} M'' \le_{\frak K} M \\
4486: &\and M'' \le_{\frak K} N'' \le_{\frak K} N \and N'' \in K_\lambda
4487: \and a \in N'' \\
4488: &\Rightarrow \nonfork{}{}_{\lambda}(M',M'',a,N'')] \biggr\}.
4489: \endalign
4490: $$
4491: \mn
4492: 2) $K^{3,\text{bs}}_{= \mu} = K^{3,\text{bs}}_{{\frak s},\mu} :=
4493: \{(M,N,a) \in K^{3,\text{bs}}_{\ge {\frak s}}:M,N \in {\frak K}^{\frak
4494: s}_\mu\}$.
4495: \nl
4496: 3) $K^{3,\text{bs}}_{\frak s} := K^{3,\text{bs}}_{= \lambda,{\frak
4497: s}}$; and let $K^{3,\text{bs}}_\mu = K^{3,\text{bs}}_{= \mu}$, used
4498: mainly for $\mu = \lambda_{\frak s}$
4499: and $K^{3,\text{bs}}_{{\frak s},\ge \mu}$ is defined naturally.
4500: \enddefinition
4501: \bigskip
4502:
4503: \definition{\stag{600-1.7} Definition} We define
4504: $\nonfork{}{}_{< \infty}(M_0,M_1,a,M_3)$
4505: (rather than $\nonfork{}{}_{\lambda}$) as follows: it holds \ub{iff}
4506: $M_0 \le_{\frak K} M_1 \le_{\frak K}
4507: M_3$ are from $K$ (not necessarily $K_\lambda$), $a \in M_3 \backslash M_1$
4508: and there is $M'_0 \le_{\frak K} M_0$ which belongs to $K_\lambda$
4509: satisfying: if
4510: $M'_0 \le_{\frak K} M'_1 \le_{\frak K} M_1,M'_1 \in K_\lambda$, \newline
4511: $M'_1 \cup \{a\}
4512: \subseteq M'_3 \le_{\frak K} M_3$ and $M'_3 \in K_\lambda$ \underbar{then}
4513: $\nonfork{}{}_{\lambda}(M'_0,M'_1,a,M'_3)$. \nl
4514: We now check that $\nonfork{}{}_{< \infty}$ behaves correctly when
4515: restricted to $K_\lambda$.
4516: \enddefinition
4517: \bigskip
4518:
4519: \proclaim{\stag{600-1.8} Claim} 1) Assume $M \le_{\frak K} N$ are from
4520: $K_\lambda$ and $a \in N$. \underbar{Then} $(M,N,a) \in
4521: K^{3,\text{bs}}_{\frak s}$
4522: \ub{iff} ${\text{\rm \ortp\/}}(a,M,N) \in {\Cal S}^{\text{bs}}_{\frak
4523: s}(M)$.
4524: \newline
4525: 2) Assume $M_0,M_1,M_3 \in K_\lambda$ and $a \in M_3$. \underbar{Then}
4526: $\nonfork{}{}_{< \infty}(M_0,M_1,a,M_3)$ \underbar{iff} \nl
4527: $\nonfork{}{}_{\lambda}(M_0,M_1,a,M_3)$. \newline
4528: 3) Assume $M \le_{\frak K} N_1 \le_{\frak K} N_2$ and $a \in N_1$.
4529: \underbar{Then} \newline
4530: $(M,N_1,a) \in K^{3,\text{bs}}_{\ge {\frak s}} \Leftrightarrow
4531: (M,N_2,a) \in K^{3,\text{bs}}_{\ge {\frak s}}$. \newline
4532: 4) Assume
4533: $M_0 \le_{\frak K} M_1 \le_{\frak K} M_3 \le_{\frak K} M^*_3$ and
4534: $a \in M_3$ \ub{then}: $\nonfork{}{}_{< \infty}
4535: (M_0,M_1,a,M_3)$ \underbar{iff} \newline
4536: $\nonfork{}{}_{< \infty}(M_0,M_1,a,M^*_3)$.
4537: \endproclaim
4538: \bigskip
4539:
4540: \demo{Proof} 1) First assume \ortp$(a,M,N) \in
4541: {\Cal S}^{\text{bs}}_{\frak s}(M)$
4542: and check the definition of $(M,N,a) \in K^{3,\text{bs}}$.
4543: Clearly $M \le_{\frak K} N,
4544: a \in N$ and $a \in N \backslash M$; we have to find $M'$ as required
4545: in Definition \scite{600-1.6}(1);
4546: we let $M' = M$, so $M' \le_{\frak K} M,M' \in K_\lambda$ and
4547:
4548: $$
4549: \align
4550: M' \le_{\frak K} M''
4551: \le_{\frak K} M \and M'' \in K_\lambda &\Rightarrow M'' = M\\
4552: &\Rightarrow \text{ \ortp}_{{\frak K}_\lambda}(a,M'',N)
4553: = \text{ \ortp}_{{\frak K}_\lambda}(a,M,N) \in {\Cal
4554: S}^{\text{bs}}_{\frak s}(M) = {\Cal S}^{\text{bs}}_{\frak s}(M'')
4555: \endalign
4556: $$
4557: \mn
4558: so we are done.
4559:
4560: Second assume $(M,N,a) \in K^{3,\text{bs}}$
4561: so there is $M' \le_{\frak K} M$ as
4562: asserted in the definition \scite{600-1.6}(1) of
4563: $K^{3,\text{bs}}$ so $(\forall M'')[M' \le_{\frak K}
4564: M'' \le_{\frak K} M \and M'' \in K_\lambda
4565: \Rightarrow \text{ \ortp}(a,M'',N) \in {\Cal S}^{\text{bs}}_{\frak s}(M'')]$
4566: in particular this holds for $M'' = M$ and we get \ortp$(a,M,N) \in
4567: {\Cal S}^{\text{bs}}_{\frak s}(M)$ as required. \newline
4568: 2) First assume $\nonfork{}{}_{< \infty} (M_0,M_1,a,M_3)$. \newline
4569: So there is $M'_0$ as required in Definition \scite{600-1.7}; this means
4570:
4571: $$
4572: M'_0 \in K_\lambda,M'_0 \le_{\frak K} M_0 \text{ and}
4573: $$
4574:
4575: $$
4576: \align
4577: (\forall M'_1 \in K_\lambda)(\forall M'_3 \in K_\lambda)[M'_0
4578: \le_{\frak K} M'_1 \le M_1 &\and M'_1 \cup \{a\} \subseteq M'_3 \le_{\frak K}
4579: M_3 \\
4580: &\rightarrow \nonfork{}{}_{\lambda}(M'_0,M'_1,a,M'_3)].
4581: \endalign
4582: $$
4583: \mn
4584: In particular, we can choose $M'_1 = M_1,M'_3 = M_3$ so the antecedent holds
4585: hence $\nonfork{}{}_{\lambda}(M'_0,M'_1,a,M'_3)$ which means
4586: $\nonfork{}{}_{\lambda}(M'_0,M_1,a,M_3)$ and by clause $(E)(b)$ of
4587: Definition \scite{600-1.1}, $\nonfork{}{}_{\lambda}(M_0,M_1,a,M_3)$ holds,
4588: as required.
4589:
4590: Second assume $\nonfork{}{}_{\lambda}(M_0,M_1,a,M_3)$. So in Definition
4591: \scite{600-1.7} the demands $M_0 \le_{\frak K} M_1 \le_{\frak K} M_3,a \in
4592: M_3 \backslash M_1$ hold by clause $(E)(a)$ of Definition \scite{600-1.1}; and
4593: we choose $M'_0$ as $M_0$; clearly $M'_0 \in K_\lambda \and M'_0 \le_{\frak K}
4594: M_0$. Now suppose $M'_0 \le_{\frak K} M'_1 \le_{\frak K} M_1 \and M'_1 \in
4595: K_\lambda,M'_1 \cup \{a\} \le_{\frak K} M'_3 \le M_3$; by clause
4596: $(E)(b)$ of Definition
4597: \scite{600-1.1} we have $\nonfork{}{}_{\lambda}(M'_0,M'_1,a,M'_3)$; so $M'_0$
4598: is as required so really $\nonfork{}{}_{< \infty}(M_0,M_1,a,M_3)$. \newline
4599: 3) We prove something stronger: for any $M' \in {\frak K}_{\frak s}$
4600: which is $\le_{{\frak K}[{\frak s}]} M,M'$ witnesses
4601: $(M,N_1,a) \in K^{3,\text{bs}}$ iff $M'$ witnesses
4602: $(M,N_2,a) \in K^{3,\text{bs}}$ (witness means: as required in
4603: Definition \scite{600-1.6}). So we have to check the statement there for
4604: every $M'' \in K_\lambda$ such that $M' \le_{\frak s} M'' \le_{\frak
4605: K} M$. The equivalence holds
4606: because for every $M'' \le_{\frak K} M,M'' \in K_\lambda$ we
4607: have \ortp$(a,M'',N_1) = \text{\ortp}(a,M'',N_2)$, by
4608: \scite{600-0.12A}(2), more transparent as
4609: ${\frak K}_\lambda$ has the amalgamation property (by clause (C) of
4610: Definition \scite{600-1.1}) and so one is ``basic" iff the other is
4611: by clause $(E)(b)$ of Definition \scite{600-1.1}. \newline
4612: 4) The direction $\Leftarrow$ is because if $M'_0$ witness
4613: $\nonfork{}{}_{< \infty} (M_0,M_1,a,M^*_3)$ (see Definition
4614: \scite{600-1.7}), \ub{then} it witnesses $\nonfork{}{}_{< \infty}
4615: (M_0,M_1,a,M_3)$ as there are just fewer pairs $(M'_1,M'_3)$ to consider.
4616: For the direction $\Rightarrow$ the demands $M_0 \le_{\frak K} M_1
4617: \le_{\frak K} M_3,a \in M_3 \backslash M_1$, of course, hold and let $M'_0$
4618: be as required in the definition of $\nonfork{}{}_{< \infty}
4619: (M_0,M_1,a,M_3)$;
4620: let $M'_0 \le_{\frak K} M'_1 \le_{\frak K} M_1,M'_1 \cup \{a\} \subseteq
4621: M'_3 \le_{\frak K} M^*_3,M'_3 \in K_\lambda$.
4622: As $\lambda \ge \text{ LS}({\frak K})$ we can find $M''_3 \le_{\frak K} M_3$
4623: such that $M'_1 \cup \{a\} \subseteq M''_3 \in K_\lambda$ and then
4624: find $M'''_3 \le_{\frak s} M^*_3$
4625: such that $M'_3 \cup M''_3 \subseteq M'''_3 \in K_\lambda$.
4626: So by the choice of $M'_0$ and $M''_3$ clearly $\nonfork{}{}_{\lambda}
4627: (M'_0,M'_1,a,M''_3)$ and by clause $(E)(b)$ of Definition \scite{600-1.1}
4628: we have
4629:
4630: $$
4631: \nonfork{}{}_{\lambda}(M'_0,M'_1,a,M''_3) \Leftrightarrow
4632: \nonfork{}{}_{\lambda}(M'_0,M'_1,a,M'''_3) \Leftrightarrow
4633: \nonfork{}{}_{\lambda}(M'_0,M'_1,a,M'_3)
4634: $$
4635: \medskip
4636: \noindent
4637: (note that we know the left statement and need the right statement)
4638: so $M'_1$ is as required to complete the checking of
4639: $\nonfork{}{}_{< \infty}(M_0,M_1,a,M^*_3)$.
4640: \hfill$\square_{\scite{600-1.8}}$
4641: \enddemo
4642: \bn
4643: We extend the definition of ${\Cal S}^{\text{bs}}_{\frak s}(M)$
4644: from $M \in K_\lambda$ to arbitrary $M \in K$.
4645:
4646: \definition{\stag{600-1.9} Definition} 1) For $M \in K$ we let
4647:
4648: $$
4649: \align
4650: {\Cal S}^{\text{bs}}(M) = {\Cal S}^{\text{bs}}_{\ge {\frak s}}(M) =
4651: \biggl\{ p \in {\Cal S}(M):&\text{ for some }
4652: N \text{ and } a,\\
4653: &\,\,p = \text{ \ortp}(a,M,N) \text{ and } (M,N,a)
4654: \in K^{3,\text{bs}}_{\ge {\frak s}}\biggr\}
4655: \endalign
4656: $$
4657: \mn
4658: (for $M \in K_\lambda$ we get the old definition by \scite{600-1.8}(1); note that
4659: as we do not have amalgamation (in general) the meaning of types is
4660: more delicate. Not so in ${\frak K}_\lambda$ as in a good $\lambda$-frame we
4661: have amalgamation in ${\frak K}_\lambda$ but not
4662: necessarily in ${\frak K}_{\ge \lambda}$).
4663: \newline
4664: 2) We say that $p \in {\Cal S}^{\text{bs}}_{\ge {\frak s}}(M_1)$
4665: does not fork over $M_0
4666: \le_{\frak K} M_1$ \ub{if} for some $M_3,a$ we have \newline
4667: $p = \text{ \ortp}_{{\frak K}[{\frak s}]}(a,M_1,M_3)$
4668: and $\nonfork{}{}_{< \infty}(M_0,M_1,a,M_3)$.
4669: (Again, for $M \in K_\lambda$ this is equivalent to the old definition by
4670: \scite{600-1.8}). \nl
4671: 3) For $M \in K$ let ${\Cal E}^\lambda_M$ be the following two-place relation
4672: on ${\Cal S}(M):p_1 {\Cal E}^\lambda_M p_2$ iff $p_1,p_2 \in {\Cal
4673: S}^{\text{bs}}(M)$ and if $p_\ell = \text{ \ortp}(a_\ell,M,M^*),N
4674: \le_{\frak K} M,N \in K_\lambda$ then $p_1 \restriction N = p_2
4675: \restriction N$. Let ${\Cal E}^{\frak s}_M = {\Cal E}^{\lambda({\frak
4676: s})}_M \restriction {\Cal S}^{\text{bs}}(M)$.
4677: \nl
4678: 4) ${\frak K}$ is $(\lambda,\mu)$-local if every $M \in
4679: {\frak K}_\mu$ is $\lambda$-local which means that
4680: ${\Cal E}^\lambda_M$ is equality; let $({\frak s},\mu)$-local
4681: means $(\lambda_{\frak s},\mu)$-local.
4682: \mn
4683: Though we will prove
4684: below some nice things, having the extension property is
4685: more problematic.
4686: We may define ``the extension" in a
4687: formal way, for $M \in K_{> \lambda}$ but then it is not
4688: clear if it is realized in any $\le_{\frak K}$-extension of $M$. Similarly
4689: for the uniqueness property.
4690: That is, assume $M_0 \le_{\frak K} M \le_{\frak K} N_\ell$ and $a_\ell
4691: \in N_\ell \backslash M$, and $M_0 \in {\frak K}_{\frak s}$ and
4692: \ortp$(a_\ell,M,N_\ell)$ does not fork over $M_0$ for $\ell=1,2$ and
4693: \ortp$(a_1,M_0,N_1) = \text{\rm \ortp}(a_2,M_0,N_1)$. Now does it
4694: follow that $\ortp(a_1,M,N_1) = \text{\rm \ortp}(a_2,M,N_2)$?
4695: This requires the existence of some form of
4696: amalgamation in ${\frak K}$, which we are not justified in assuming.
4697: So we may prefer to define ${\Cal S}^{\text{bs}}(M)$ ``formally", the
4698: set of stationarization of $p \in {\Cal S}^{\text{bs}}
4699: (M_0),M_0 \in {\frak K}_{\frak s}$, see \cite{Sh:842}.
4700: We now note that in definition \scite{600-1.9}
4701: ``some" can be replaced by ``every".
4702: \enddefinition
4703: \bigskip
4704:
4705: \demo{\stag{600-1.10} Fact} 1) For $M \in K$
4706:
4707: $$
4708: \align
4709: {\Cal S}^{\text{bs}}_{\ge {\frak s}}(M) =
4710: \biggl\{ p \in {\Cal S}_{{\frak K}[{\frak s}]}(M):&\text{ for every } N,a \\
4711: &\text{ we have: if } M \le_{\frak K} N,a \in N \backslash M \text{ and} \\
4712: &\,p = \text{ \ortp}_{\frak K}(a,M,N)
4713: \text{ then } (M,N,a) \in K^{3,\text{bs}}_{\ge{\frak s}}\biggr\}.
4714: \endalign
4715: $$
4716: \mn
4717: 2) The type $p \in {\Cal S}_{{\frak K}[{\frak s}]}(M_1)$ does not
4718: fork over $M_0 \le_{\frak K} M_1$ iff for every $a,M_3$ satisfying
4719: $M_1 \le_{\frak K} M_3 \in K,a \in M_3 \backslash M_1$ and
4720: $p = \text{ \ortp}_{{\frak K}[{\frak s}]}(a,M_1,M_3)$ we have
4721: $\nonfork{}{}_{< \infty}(M_0,M_1,a,M_3)$. \nl
4722: 3) $(M,N,a) \in K^{3,\text{bs}}_{\ge{\frak s}}$ is preserved by
4723: isomorphisms.
4724: \nl
4725: 4) If $M \le_{\frak K} N_\ell,a_\ell \in N_\ell \backslash M$ for
4726: $\ell=1,2$ and $\ortp(a_1,M,N_1) {\Cal E}^{\frak s}_M
4727: \ortp(a_2,M,N_2)$ then $(M,N_1,a_1) \in
4728: K^{3,\text{bs}}_{\ge{\frak s}} \Leftrightarrow (M,N_2,a_2) \in
4729: K^{3,\text{bs}}_{\ge{\frak s}}$.
4730: \nl
4731: 5) ${\Cal E}^{\frak s}_M$ is an equivalence relation on ${\Cal
4732: S}^{\text{bs}}_{\ge {\frak s}}(M)$ and if $p,q \in
4733: {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M)$ do not fork over $N \in
4734: K_\lambda$ so $N
4735: \le_{\frak K} M$ then $p {\Cal E}^{\frak s}_M q \Leftrightarrow (p \restriction
4736: N = q \restriction N)$.
4737: \enddemo
4738: \bigskip
4739:
4740: \demo{Proof} 1) By \scite{600-1.8}(3) and the definition of type. \newline
4741: 2) By \scite{600-1.8}(4) and the definition of type. \nl
4742: 3) Easy.\nl
4743: 4) Enough to deal with the case $(M,N_1,a_1)
4744: E^{\text{at}}_M,(M,N_2,a_2)$ or (by (3)) even $a_1 = a_2,N_1
4745: \le_{\frak K} N_2$. This is easy.
4746: \nl
4747: 5) Easy, too. \hfill$\square_{\scite{600-1.10}}$
4748: \enddemo
4749: \bn
4750: We can also get that there are enough basic types, as follows:
4751: \proclaim{\stag{600-1.11} Claim} If $M \le_{\frak K} N$ and $M \ne N$,
4752: \underbar{then} for some $a \in N
4753: \backslash M$ we have ${\text{\rm \ortp\/}}_{\frak K}(a,M,N) \in
4754: {\Cal S}^{\text{bs}}(M)$.
4755: \endproclaim
4756: \bigskip
4757:
4758: \demo{Proof} Suppose not, so by clause $(D)(c)$ of Definition \scite{600-1.1}
4759: necessarily $\|N\| > \lambda$. If $\|M\| = \lambda < \|N\|$ choose
4760: $N'$ satisfying
4761: $M <_{\frak K} N' \le_{\frak K} N,N' \in K_\lambda$ and by clause $(D)(c)$ of
4762: Definition \scite{600-1.1} choose $a^* \in N' \backslash M$ such
4763: that \ortp$_{\frak s}(a^*,M,N') \in {\Cal S}^{\text{bs}}_{\frak s}(M)$.
4764: So we can assume $\|M\| > \lambda$; choose
4765: $a^* \in N \backslash M$. We choose by induction on
4766: $i < \omega,M_i,N_i,M_{i,c}$ (for $c \in N_i \backslash M_i)$
4767: such that:
4768: \mr
4769: \item "{$(a)$}" $M_i \le_{\frak K} M$ is $\le_{\frak K}$-increasing
4770: \sn
4771: \item "{$(b)$}" $M_i \in K_\lambda$
4772: \sn
4773: \item "{$(c)$}" $N_i \le_{\frak K} N$ is $\le_{\frak K}$-increasing
4774: \sn
4775: \item "{$(d)$}" $N_i \in K_\lambda$
4776: \sn
4777: \item "{$(e)$}" $a^* \in N_0$
4778: \sn
4779: \item "{$(f)$}" $M_i \le_{\frak K} N_i$
4780: \sn
4781: \item "{$(g)$}" if $c \in N_i \backslash M$, \ortp$_{\frak s}(c,M_i,N) \in
4782: {\Cal S}^{\text{bs}}_{\frak s}(M_i)$ and there is
4783: $M' \in K_\lambda$ such that $M_i \le_{\frak K} M' \le_{\frak K} M$ and
4784: \ortp$_{\frak s}(c,M',N)$ forks over $M_i$ then
4785: $M_{i,c}$ satisfies this, otherwise $M_{i,c} = M_i$
4786: \sn
4787: \item "{$(h)$}" $M_{i+1}$ includes the set
4788: $\dsize \bigcup_{c \in N_i \backslash M} M_{i,c} \cup(N_i \cap M)$.
4789: \ermn
4790: There is no problem to carry the definition. Let $M^* = \dsize \bigcup_{i <
4791: \omega} M_i$ and $N^* = \dsize \bigcup_{i < \omega} N_i$. It is easy to
4792: check that:
4793: \mr
4794: \widestnumber\item{(viii)}
4795: \item "{$(i)$}" $M_i \le_{\frak K} M^* \le_{\frak K} M$ for $i <
4796: \omega$
4797: \nl
4798: (by clause (a))
4799: \sn
4800: \item "{$(ii)$}" $M^* \in K_\lambda$ \newline
4801: (by clause (i) we have $M^* \in K$ and $\|M^*\| = \lambda$ by the
4802: choice of $M^*$ and clause (b))
4803: \sn
4804: \item "{$(iii)$}" $N_i \le_{\frak K} N^* \le_{\frak K} N$ \newline
4805: (by clause (c))
4806: \sn
4807: \item "{$(iv)$}" $N^* \in K_\lambda$ \newline
4808: (by clause (iii) we have $N^* \in K$ and $\|N^*\| = \lambda$ by the
4809: choice of $N^*$ and clause (d))
4810: \sn
4811: \item "{$(v)$}" $M_i \le_{\frak K} M^*
4812: \le_{\frak K} N^* \le_{\frak K} N$ \newline
4813: (by clauses (a) + (f) + (iii) we have $M_i \le_{\frak K} N^*$ hence by
4814: clause (a) and the choice of $M^*$ we have $M^* \le_{\frak K} N^*$,
4815: and $N^* \le_{\frak K} N$ by clause (iii))
4816: \sn
4817: \item "{$(vi)$}" $M^* = N^* \cap M$ \newline
4818: (by clauses (f) + (h) and the choices of $M^*,N^*$)
4819: \sn
4820: \item "{$(vii)$}" $M^* \ne N^*$ \nl
4821: (as $a^* \in N \backslash M$ and $a^* \in N_0 \le_{\frak K} N^*
4822: \le_{\frak K} N$ and $M^* = N^* \cap M$; \newline
4823: they hold by the choice of $a^*$, clause (e), choice of $N^*$, clause
4824: (iii) and clause (vi) respectively)
4825: \sn
4826: \item "{$(viii)$}" there is $b^* \in N^* \backslash M^*$ such that
4827: \ortp$(b^*,M^*,N^*) \in {\Cal S}^{\text{bs}}(M^*)$ \newline
4828: [why? by Definition \scite{600-1.1} clause (D)(c) (density)]
4829: \sn
4830: \item "{$(ix)$}" for some $i < \omega$
4831: we have $\nonfork{}{}_{}(M_i,M^*,b^*,N^*)$, so \nl
4832: \ortp$(b^*,M^*,N^*) \in {\Cal S}^{\text{bs}}_{\frak s}(M^*)$
4833: and \ortp$_{\frak s}(b^*,M_j,N^*) \in
4834: {\Cal S}^{\text{bs}}_{\frak s}(M_j)$ for $j \in [i,\omega)$ \nl
4835: [why? by Definition \scite{600-1.1} clause $(E)(c)$ (local character)
4836: applied to the sequence $\langle M_n:n < \omega \rangle \char 94
4837: \langle M^*,N^* \rangle$ and the element $b^*$, using of course (E)(a)
4838: of Definition \scite{600-1.1} and clause (viii)]
4839: \sn
4840: \item "{$(x)$}" $\nonfork{}{}_{}(M_i,M_{i,b^*},b^*,N^*)$ \newline
4841: [why? by clause (ix) and Definition \scite{600-1.1}$(E)(b)$ (monotonicity) as
4842: \newline
4843: $M_i \le_{\frak K} M_{i,b^*} \le_{\frak K} M_{i+1} \le_{\frak K} M^*$ by
4844: clause (g) in the construction]
4845: \sn
4846: \item "{$(xi)$}" if $M_i \le_{\frak K} M' \le_{\frak K} M$ and
4847: $M' \cup \{ b^*\} \subseteq N' \le_{\frak K} N$
4848: and $M' \in K_\lambda,N' \in K_\lambda$ then
4849: $\nonfork{}{}_{}(M_i,M',b^*,N')$ \newline
4850: [why? by clause (x) and clause (g) in the construction.]
4851: \ermn
4852: So we are done. \hfill$\square_{\scite{600-1.11}}$
4853: \enddemo
4854: \bigskip
4855:
4856: \proclaim{\stag{600-1.12} Claim} If $M \le_{\frak K} N,a \in N \backslash M$, and
4857: ${\text{\rm \ortp\/}}(a,M,N) \in {\Cal S}^{\text{bs}}_{\ge {\frak s}}(M)$
4858: \underbar{then} for some $M_0 \le_{\frak K} M$ we have
4859: \mr
4860: \item "{$(a)$}" $M_0 \in K_\lambda$
4861: \sn
4862: \item "{$(b)$}" ${\text{\rm \ortp\/}}(a,M_0,N) \in
4863: {\Cal S}^{\text{bs}}_{\frak s}(M_0)$
4864: \sn
4865: \item "{$(c)$}" if $M_0 \le_{\frak K} M' \le_{\frak K} M$,
4866: \ub{then} ${\text{\rm \ortp\/}}(a,M',N)
4867: \in {\Cal S}^{\text{bs}}_{\frak s}(M')$ does not fork over $M_0$.
4868: \endroster
4869: \endproclaim
4870: \bigskip
4871:
4872: \demo{Proof} Easy by now.
4873: \enddemo
4874: \bigskip
4875:
4876: \proclaim{\stag{600-1.13} Claim} 1) Assume $M_1 \le_{\frak K} M_2$ and $p \in
4877: {\Cal S}_{\frak K}(M_2)$. \ub{Then}
4878: $p \in {\Cal S}^{\text{bs}}_{\ge {\frak s}}(M_2)$ and $p$ does not fork
4879: over $M_1$ \underbar{iff} for some $N_1 \le_{\frak K} M_1,N_1 \in
4880: K_\lambda$ and $p$ does not fork over $N_1$ \ub{iff} for some $N_1
4881: \le_{\frak K} M_1,N_1 \in K_\lambda$ and we have
4882: $(\forall N)[N_1 \le_{\frak K} N \le_{\frak
4883: K} M_2 \and N \in K_\lambda \Rightarrow p \restriction N \in {\Cal
4884: S}^{\text{bs}}_{\frak s}(N)
4885: \and (p \restriction N$ does not fork over $N_1)]$; we
4886: call such $N_1$ a witness, so every $N'_1 \in K_\lambda,N_1 \le_{\frak
4887: K} N'_1 \le M_1$ is a witness, too. \newline
4888: 2) Assume $M^* \in K$ and $p \in {\Cal S}_{\frak K}(M^*)$. \newline
4889: \ub{Then}: $p \in {\Cal S}^{\text{bs}}_{\ge {\frak s}}(M^*)$ \ub{iff}
4890: for some $N^* \le_{\frak K} M^*$ we have
4891: $N^* \in K_\lambda,p \restriction N^* \in
4892: {\Cal S}^{\text{bs}}(N^*)$ and $(\forall N \in K_\lambda)
4893: (N^* \le_{\frak K} N \le_{\frak K} M^* \Rightarrow p
4894: \restriction N \in {\Cal S}^{\text{bs}}(N)$ and does not fork over $N^*)$
4895: (we say such $N^*$ is a witness, so any $N' \in K_\lambda,N^*
4896: \le_{\frak K} N' \le_{\frak K} M$ is a witness, too). \newline
4897: 3) (Monotonicity) \newline If $M_1 \le_{\frak K} M'_1 \le_{\frak K} M'_2
4898: \le_{\frak K} M_2$ and $p \in {\Cal S}^{\text{bs}}_{\ge {\frak
4899: s}}(M_2)$ does not fork over
4900: $M_1$, \underbar{then} \newline $p \restriction M'_2 \in
4901: {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M'_2)$ and it does not
4902: fork over $M'_1$. \newline
4903: 4) (Transitivity) \newline If $M_0 \le_{\frak K} M_1 \le_{\frak K} M_2$
4904: and $p \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_2)$
4905: does not fork over $M_1$ and $p
4906: \restriction M_1$ does not fork over $M_0$, \underbar{then} $p$ does
4907: not fork over $M_0$. \newline
4908: 5) (Local character)
4909: If $\langle M_i:i \le \delta + 1 \rangle$ is
4910: $\le_{\frak K}$-increasing continuous and $a \in M_{\delta + 1}$ and
4911: ${\text{\rm \ortp\/}}_{\frak K}(a,M_\delta,M_{\delta + 1}) \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_\delta)$ \underbar{then} for some $i < \delta$ we have
4912: ${\text{\rm \ortp\/}}_{\frak K}(a,M_\delta,M_{\delta + 1})$ does not fork
4913: over $M_i$. \newline
4914: 6) Assume that $\langle M_i:i \le \delta + 1 \rangle$ is
4915: $\le_{\frak K}$-increasing and
4916: $p \in {\Cal S}(M_\delta)$ and for every $i < \delta$ we have
4917: $p \restriction M_i \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_i)$
4918: does not fork over $M_0$. \ub{Then}
4919: $p \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_\delta)$ and
4920: $p$ does not fork over $M_0$.
4921: \endproclaim
4922: \bigskip
4923:
4924: \demo{Proof} 1), 2) Check the definitions.
4925: \newline
4926: 3) As $p \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_2)$ does not
4927: fork over $M_1$, there is $N_1
4928: \in K_\lambda$ which witnesses it.
4929:
4930: This same $N_1$ witnesses that $p \restriction M'_2$ does not fork over
4931: $M'_1$. \newline
4932: 4) Let $N_0 \le_{\frak K} M_0$ witness that $p \restriction M_1$ does not
4933: fork over $M_0$ (in particular $N_0 \in K_\lambda$);
4934: let $N_1 \le_{\frak K} M_1$ witness that $p$ does not fork over $M_1$
4935: (so in particular $N_1 \in K_\lambda$).
4936: Let us show that $N_0$ witnesses $p$ does not fork over $M_0$, so let
4937: $N \in K_\lambda$ be such that $N_0 \le_{\frak K} N \le_{\frak K} M_2$
4938: and we should just prove that $p \restriction N$ does not fork over
4939: $N_0$. We can find $N' \le_{\frak K} M_1,
4940: N' \in K_\lambda$ such that $N_0 \cup N_1 \subseteq N'$, we can also find
4941: $N'' \le_{\frak K} M_2$ satisfying $N'' \in K_\lambda$
4942: such that $N' \cup N \subseteq N''$.
4943: As $N_1$ witnesses
4944: that $p$ does not fork over $M_1$, clearly $p \restriction
4945: N'' \in {\Cal S}^{\text{bs}}_{\frak s}(N'')$
4946: does not fork over $N_1$, hence by
4947: monotonicity does not fork over $N'$. As $N_0$ witnesses
4948: $p \restriction M_1$ does not fork over
4949: $M_0$, clearly $p \restriction N'$ belongs to ${\Cal
4950: S}^{\text{bs}}(N')$ and does not fork over $N_0$, so by
4951: transitivity (in ${\frak K}_{\frak s}$) we know that
4952: $p \restriction N''$ does not fork over $N_0$; hence by
4953: monotonicity $p \restriction N$ does not fork over $N_0$. \newline
4954: 5) Let $p = \text{ \ortp}_{\frak K}(a,M_\delta,M_{\delta +1})$ and let $N^*
4955: \le_{\frak K} M_\delta$
4956: witness $p \in {\Cal S}^{\text{bs}}
4957: (M_\delta)$. Assume toward contradiction that
4958: the conclusion fails. Without loss of generality cf$(\delta) = \delta$.
4959: \enddemo
4960: \bn
4961: \ub{Case 0}: $\|M_\delta\| \le \lambda (=\lambda_{\frak s})$.
4962:
4963: Trivial.
4964: \bn
4965: \underbar{Case 1}: $\delta < \lambda^+,\|M_\delta\| > \lambda$.
4966:
4967: As $\|M_\delta\| > \lambda$, for some $i,\|M_i\| > \lambda$ so without loss
4968: of generality $i < \delta \Rightarrow \|M_i\| > \lambda$. We choose by
4969: induction on $i < \delta$, models $N_i,N'_i$ such that:
4970: \medskip
4971: \roster
4972: \item "{$(\alpha)$}" $N_i \in K_\lambda$
4973: \sn
4974: \item "{$(\beta)$}" $N_i \le_{\frak K} M_i$ (hence $N_i \le_{\frak K}
4975: M_j$ for $j \in [i,\delta))$
4976: \sn
4977: \item "{$(\gamma)$}" $N_i$ is $\le_{\frak K}$-increasing continuous
4978: \sn
4979: \item "{$(\delta)$}" $N'_i \in K_\lambda,N^* \le_{\frak K} N'_0$
4980: \sn
4981: \item "{$(\varepsilon)$}" $N_i \le_{\frak K} N'_i \le_{\frak K}
4982: M_\delta$,
4983: \sn
4984: \item "{$(\zeta)$}" $N'_i$ is $\le_{\frak K}$-increasing continuous
4985: \sn
4986: \item "{$(\eta)$}" $p \restriction N'_i$ forks over $N_i$ when $i \ne
4987: 0$ for simplicity
4988: \sn
4989: \item "{$(\theta)$}" $N_i \cup
4990: \bigcup_{j \le i}(N'_j \cap M_{i+1}) \subseteq N_{i+1}$.
4991: \ermn
4992: No problem to carry the induction, but we give details.
4993: \sn
4994: First, if $i = 0$ trivial. Second let $i$ be a limit ordinal.
4995:
4996: Let $N_i = \cup\{N_j:j < i\}$, now $N_i \le_{\frak K} M_i$ by clauses
4997: $(\beta) + (\gamma)$ and ${\frak K}$ being a.e.c. and
4998: $\|N_i\| = \lambda$ by clause $(\alpha)$, as
4999: $i \le \delta < \lambda^+$; so clauses $(\alpha), (\beta), (\gamma)$
5000: hold. Next, let $N'_i = \cup\{N'_j:j <i\}$ and similarly clauses
5001: $(\delta), (\varepsilon), (\zeta)$ hold. Lastly, we shall prove
5002: clause $(\eta)$ and assume toward contradiction that it fails; so
5003: $p \restriction N'_i$ does not fork over $N_i$ in
5004: particular $p \restriction N_i \in {\Cal S}^{\text{bs}}_{\frak
5005: s}(N_i)$ hence for some $j < i$ the type $p \restriction N'_i$ does not fork
5006: over $N_j \le_{\frak K} N_i$, (by (E)(c) of Definition \scite{600-1.1})
5007: hence by transitivity (for ${\frak K}_{\frak s}$), $p \restriction
5008: N'_i$ does not fork over $N_j$ hence by monotonicity
5009: $p \restriction N'_j$ does not fork over $N_j$ (see (E)(b) of
5010: Definition \scite{600-1.1}) contradicting the induction hypothesis.
5011:
5012: Lastly, clause $(\theta)$ is vacuous.
5013:
5014: Third assume $i=j+1$, so first choose $N_i$ satisfying clause
5015: $(\theta)$ (with $j,i$ here standing for $i,i+1$ there), and
5016: $(\alpha), (\beta), (\gamma)$; this is possible by the L.S. property.
5017: Now $N_i$ cannot witness ``$p$ does not fork over $M_i$" hence for
5018: some $N^*_i \in K_\lambda$ we have $N_i \le_{\frak K} N^*_i \le_{\frak
5019: K} M_\delta$ and $p \restriction N^*_i$ forks over $N_i$; again by
5020: L.S. choose $N'_i \in K_\lambda$ such that $N'_i \le_{\frak K} M_\delta$ and
5021: $N^* \cup N_i \cup N'_j \cup N^*_i \subseteq N'_i$, easily
5022: $(N_i,N'_i)$ are as required.
5023:
5024: Let $N_\delta = \dsize \bigcup_{i < \delta} N_i$, so
5025: by clause $(\beta), (\gamma)$ we have
5026: $N_\delta \le_{\frak K} M_\delta$ and by clause $(\alpha)$, as
5027: $\delta < \lambda^+$ we have $N_\delta \in K_\lambda$ and by
5028: clauses $(\delta) + (\theta)$ in the construction we have $i < \delta
5029: \Rightarrow N'_i = \cup\{N'_i \cap M_{j+1}:j \in [i,\delta)\}
5030: \subseteq N$ so by clause $(\delta),N^* \le_{\frak K} N'_0 \le_{\frak
5031: K} N_\delta$. Hence by the choice of $N^*,
5032: p \restriction N_\delta \in {\Cal S}^{\text{bs}}_{\frak s}(N_\delta)$
5033: and it does not fork
5034: over $N^*$. Now as $p \restriction N_\delta \in {\Cal S}^{\text{bs}}_{\frak s}
5035: (N_\delta)$ by local character, i.e.,
5036: clause $(E)(c)$ of Definition \scite{600-1.1}, for some $i < \delta,
5037: p \restriction
5038: N_\delta$ does not fork over $N_i$ (so
5039: $p \restriction N_i \in {\Cal S}^{\text{bs}}_{\frak s}(N_i)$). Now $N_i
5040: \le_{\frak K} N'_i \le_{\frak K}
5041: M_\delta$ and by clause $(\theta)$ of the construction $N'_i \subseteq
5042: N_\delta$ hence $N_i \le_{\frak K} N'_i \le_{\frak K} N_\delta$ hence by
5043: monotonicity of non-forking (i.e. clause (E)(b) of Definition \scite{600-1.1}),
5044: $p \restriction N'_i \in {\Cal S}^{\text{bs}}
5045: (N_i)$ does not fork over $N_i$. But this contradicts
5046: the choice of $N'_i$ (i.e., clause $(\eta)$ of the construction).
5047: \bn
5048: \underbar{Case 2}: $\delta = \text{ cf}(\delta) > \lambda$.
5049:
5050: Recall that $N^* \le_{\frak K} M_\delta,N^*$ is from $K_\lambda$ and $N^*
5051: \le_{\frak K} N \le_{\frak K} M_\delta \and N \in K_\lambda \Rightarrow$ \nl
5052: $p \restriction N \in {\Cal S}^{\text{bs}}_{\frak s}(N)$.
5053: Now as $\delta = \text{ cf}(\delta) > \lambda \ge \|N^*\|$ clearly
5054: for some $i < \delta$ we have $N^* \subseteq M_i$ hence $N^*
5055: \le_{\frak K} M_i$
5056: (hence $i \le j < \delta \Rightarrow
5057: p \restriction M_j \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_j)$), and $N^*$
5058: witnesses that $p \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_\delta)$
5059: does not fork over $M_i$ so we are clearly done.
5060: \nl
5061: 6) Let $N_0 \in
5062: K_\lambda,N_0 \le_{\frak K} M_0$ witness $p \restriction M_0 \in
5063: {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_0)$.
5064: By the proof of part (4) clearly $i < \delta \and N_0
5065: \le_{\frak K} N \in K_\lambda \and N \le_{\frak K} M_i \Rightarrow p \restriction N$
5066: does not fork over $N_0$. If cf$(\delta) > \lambda$ we are done, so assume
5067: cf$(\delta) \le \lambda$. Let $N_0 \le_{\frak K} N^* \in K_\lambda
5068: \and N^* \le_{\frak K} M_\delta$, and we shall prove that $p
5069: \restriction N^*$ does not fork over $N_0$, this clearly suffices. As
5070: in Case 1 in the proof of part (5) we can find
5071: $N_i \le_{\frak K} M_i$ for $i \in (0,\delta)$ such that $\langle
5072: N_i:i \le \delta \rangle$ is $\le_{\frak K}$-increasing with $i$,
5073: each $N_i$ belongs to ${\frak K}_\lambda$ and $N^* \cap M_i \subseteq
5074: N_{i+1}$, hence $N^* \subseteq N_\delta := \dbcu_{i < \delta} N_i$.
5075: Now $N_\delta \le_{\frak K} M_\delta$ and as said as $i < \delta
5076: \Rightarrow p \restriction N_i \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}
5077: (N_i)$ does not
5078: fork over $N_0$ hence $p \restriction N_\delta$ does not fork over
5079: $N_0$ and by monotonicity
5080: $p \restriction N$ does not fork over $N_0$, as required.
5081: \nl
5082: ${{}}$ \hfill$\square_{\scite{600-1.13}}$
5083: \bigskip
5084:
5085: \proclaim{\stag{600-1.13B} Lemma} If $\mu = {\text{\rm cf\/}}
5086: (\mu) > \lambda$ and $M \le_{\frak K} N$ are in $K_\mu$, \ub{then}
5087: we can find $\le_{\frak K}$-representations
5088: $\bar M,\bar N$ of $M,N$ respectively such that:
5089: \mr
5090: \item "{$(i)$}" $N_i \cap M = M_i$ for $i < \mu$
5091: \sn
5092: \item "{$(ii)$}" if $i < j < \mu \and a \in N_i$ \underbar{then}
5093: \ermn
5094: $$
5095: \align
5096: (a) \qquad
5097: {\text{\rm \ortp\/}}(a,M_i,N) \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}
5098: (M_i) &\Leftrightarrow {\text{\rm \ortp\/}}(a,M_j,
5099: N) \in {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_j) \\
5100: &\Leftrightarrow {\text{\rm \ortp\/}}(a,M,N) \text{ does not fork over } M_i \\
5101: &\Leftrightarrow { \text{\rm \ortp\/}}(a,M_j,N)
5102: \text{ is a non-forking extension of {\rm \ortp\/}} (a,M_i,N)
5103: \endalign
5104: $$
5105: $(b) \qquad M_i \le_{\frak K} N_i \le_{\frak K} N_j$ and $M_i
5106: \le_{\frak K} M_j \le_{\frak K} N_j$ \nl
5107:
5108: \hskip20pt (and obviously $M_i \le_{\frak K} N_j$ and $M_i \le_{\frak K} M,M_i
5109: \le_{\frak K} N,N_i \le_{\frak K} N$).
5110: \endproclaim
5111: \bigskip
5112:
5113: \remark{\stag{600-1.13C} Remark} In fact for any representations $\bar
5114: M,\bar N$ of $M,N$ respectively, for some club $E$ of $\mu$ the
5115: sequences $\bar M \restriction E,\bar N \restriction E$ are as above.
5116: \endremark
5117: \bigskip
5118:
5119: \demo{Proof} Let $\bar M$ be a $\le_{\frak K}$-representation of $M$.
5120: For $a \in N$ we define
5121: $S_a = \{\alpha < \mu:\text{\ortp}(a,M_\alpha,N) \in
5122: {\Cal S}^{\text{bs}}_{\ge{\frak s}}(M_\alpha)\}$.
5123: Clearly if $\delta \in S_a$ is a limit ordinal
5124: then for some $i(a,\delta) < \delta$ we
5125: have $i(a,\delta) \le i < \delta \Rightarrow i \in S_a \and
5126: (\text{\ortp}(a,M_i,N)$ does not fork over $M_{i_{(a,\delta)}})$ by
5127: \scite{600-1.13}(5).
5128: So if $S_a$ is stationary, then for some $i(a) < \mu$ the set $S'_a =
5129: \{\delta \in S_a:i(a,\delta)=i(a)\}$ is a stationary subset of
5130: $\lambda$ hence by monotonicity we have $i(a)
5131: \le i \le \mu \Rightarrow \text{ \ortp}(a,M_i,N)$ does not fork over
5132: $M_{i(a)}$. Let $E_a$ be a club of $\mu$ such that: if $S_a$ is not
5133: stationary (subset of $\mu$) then $E_a \cap S_a = \emptyset$ and if
5134: $S_a$ is not stationary then $S_a \cap E_a = \emptyset$.
5135: \nl
5136: Let $\bar N$ be a representation of $N$, and let
5137:
5138: $$
5139: \align
5140: E^* = \{\delta < \mu:&N_\delta \cap M = M_\delta \text{ and }
5141: M_\delta \le_{\frak K} M,N_\delta \le_{\frak K} N \\
5142: &\text{ and for every } a \in N_\delta \text{ we have } \delta \in E_a\}.
5143: \endalign
5144: $$
5145: \medskip
5146: \noindent
5147: Clearly it is a club of $\mu$ and $\bar M \restriction E^*,\bar N
5148: \restriction E^*$ are as required. \hfill$\square_{\scite{600-1.13B}}$
5149: \enddemo
5150: \bn
5151: \centerline{$* \qquad * \qquad *$}
5152: \bn
5153: We may treat the lifting of $K^{3,\text{bs}}_\lambda$ as a special
5154: case of the ``lifting" of ${\frak K}_\lambda$ to ${\frak K}_{\ge \lambda} =
5155: ({\frak K}_\lambda)^{\text{up}}$ in Claim \scite{600-0.31}; this may be
5156: considered a good exercise.
5157: \proclaim{\stag{600-1.13E} Claim} 1) $(K^{3,\text{bs}}_\lambda,
5158: \le_{\text{bs}})$ is a $\lambda$-a.e.c.
5159: \nl
5160: 2) $(K^{3,\text{bs}}_{\ge {\frak s}},
5161: \le_{\text{bs}})$ is $(K^{3,\text{bs}}_\lambda,\le_{\text{bs}})^{\text{up}}$.
5162: \endproclaim
5163: \bigskip
5164:
5165: \remark{Remark} What is the class in \scite{600-1.13E}(1)? Formally let
5166: $\tau^+ = \{R_{[\ell]}:R$ a predicate of $\tau_K,\ell=1,2\} \cup
5167: \{F_{[\ell]}:F$ a function symbolf rom $\tau_K\} \cup \{c\}$ where
5168: $R_{[\ell]}$ is an $n$-place predicate when $R \in \tau$ is an
5169: $n$-place predicate and similarly $F_{[\ell]}$ and $c$ is an
5170: individual constant. A triple $(M,N,a)$ is identified with the
5171: following $\tau^+$-model $N^+$ defined as follows:
5172: \mr
5173: \item "{$(a)$}" its universe is the universe of $N$
5174: \sn
5175: \item "{$(b)$}" $c^{N^+} = a$
5176: \sn
5177: \item "{$(c)$}" $R^{N^+}_{[2]} = R^N$
5178: \sn
5179: \item "{$(d)$}" $F^{N^+}_{[2]} = F^N$
5180: \sn
5181: \item "{$(e)$}" $R^{N^+}_{[1]} = R^M$
5182: \sn
5183: \item "{$(f)$}" $F^{N^+}_{[1]} = F^M$
5184: \ermn
5185: (if you do not like partial functions, extend them to functions with
5186: full domain by $F(a_0,\ldots) = a_0$ when not defined if
5187: $F$ has arity $>0$, if $F$ has arity zero it is an individual constant,
5188: $F^{N^+} = F^N$ so no problem).
5189: \endremark
5190: \bigskip
5191:
5192: \demo{Proof} Left to the reader (in particular this means that
5193: $K^{3,\text{bs}}_\lambda$ is closed under $\le_{\text{bs}}$-increasing
5194: chains of length $< \lambda^+$).
5195: \enddemo
5196: \bn
5197: Continuing
5198: \scite{600-0.31}, \scite{600-0.34} note that (and see more in \scite{600-1.16}):
5199: \proclaim{\stag{600-1.14} Lemma} Assume
5200: \mr
5201: \item "{$(a)$}" ${\frak K}$ is an abstract elementary class with
5202: ${\text{\rm LS\/}}({\frak K}) \le \mu$
5203: \sn
5204: \item "{$(b)$}" $K'_{\le \mu}$ is a class of $\tau_K$-model,
5205: $K'_{\le \mu} \subseteq K_{\le \mu}$ is non-empty and
5206: closed under $\le_{\frak K}$-increasing unions of length $< \mu^+$ and
5207: isomorphisms (e.g. the class of $\mu$-superlimit models of ${\frak
5208: K}_\mu$, if there is one)
5209: \sn
5210: \item "{$(c)$}" define $K' :=
5211: \{M \in K:M \text{ is a $\le_{\frak K}$-directed
5212: union of members of } K'_\mu\}$
5213: \sn
5214: \item "{$(d)$}" let ${\frak K}' = (K',\le_{\frak K} \restriction K')$
5215: so $\le_{{\frak K}'}$ is $\le_{\frak K} \restriction K'$, so
5216: ${\frak K}'_{\le \mu} := (K'_{\le \mu},\le_{\frak K} \restriction
5217: K'_{\le \mu})$; or $\le_{\frak K}$ is as in
5218: \scite{600-0.31}(1), see \scite{600-0.31}(4).
5219: \ermn
5220: \ub{Then}
5221: \mr
5222: \item "{$(A)$}" ${\frak K}'$ is an abstract elementary class,
5223: ${\text{\rm LS\/}}({\frak K}) \le { \text{\rm LS\/}}({\frak K}') \le \mu$
5224: \sn
5225: \item "{$(B)$}" If $\mu \le \lambda$ and
5226: $({\frak K},\nonfork{}{}_{},{\Cal S}^{\text{bs}})$
5227: is a good $\lambda$-frame and ${\frak K}'_\lambda$ has amalgamation
5228: and JEP and $M \in {\frak K}'_\lambda \Rightarrow
5229: {\Cal S}_{{\frak K}'}(M) = {\Cal S}_{\frak K}(M)$, \ub{then}
5230: $({\frak K}',\nonfork{}{}_{},{\Cal S}^{\text{bs}})$ (with $\nonfork{}{}_{},
5231: {\Cal S}^{\text{bs}}$ restricted to ${\frak K}'$) is a good $\lambda$-frame
5232: \sn
5233: \item "{$(C)$}" in clause $(B)$, instead ``$M \in {\frak K}'_\lambda
5234: \Rightarrow {\Cal S}_{{\frak K}'}(M)= {\Cal S}_{\frak K}(M)$, it
5235: suffices to require: if $M \in {\frak K}'_\lambda,M \le_{\frak K} N
5236: \in {\frak K}'_\lambda,p \in {\Cal S}^{\text{bs}}_{\frak s}(N),p$ does not
5237: fork over $M$ and $p \restriction M$ is realized in some $M',M
5238: \le_{{\frak K}'} M'$ \ub{then} $p$ is realized in some $N',N \le_{\frak
5239: K} N' \in {\frak K}'_\lambda$.
5240: \endroster
5241: \endproclaim
5242: \bigskip
5243:
5244: \remark{Remark} If in \scite{600-1.14}, $K'_\mu$ is not closed under
5245: $\le_{\frak K}$-increasing unions, we can close it but then the ``so
5246: ${\frak K}'_{\le \mu} = \ldots$" in clause (d) may fail.
5247: \endremark
5248: \bigskip
5249:
5250: \demo{Proof} \ub{Clause (A)}: As in \scite{600-0.31}.
5251: \mn
5252: \ub{Clauses (B),(C)}: Check. \hfill$\square_{\scite{600-1.14}}$
5253: \enddemo
5254: \bn
5255: \centerline{$* \qquad * \qquad *$}
5256: \bn
5257: Next we deal with some implications between the axioms in \scite{600-1.1}.
5258: \proclaim{\stag{600-1.15} Claim} 1) In Definition \scite{600-1.1} clause (E)(i) is
5259: redundant, i.e., follows from the rest, recalling
5260: \mr
5261: \item "{$(E)(i)$}" \ub{non-forking amalgamation}: \newline
5262: if for $\ell = 1,2,M_0 \le_{\frak K} M_\ell$ are in $K_\lambda,a_\ell \in
5263: M_\ell \backslash M_0,{\text{\rm \ortp\/}}
5264: (a_\ell,M_0,M_\ell) \in {\Cal S}^{\text{bs}}(M_0)$,
5265: \ub{then} we can find $f_1,f_2,M_3$ satisfying $M_0 \le_{\frak K} M_3 \in
5266: K_\lambda$ such that for $\ell =1,2$ we have $f_\ell$ is a
5267: $\le_{\frak K}$-embedding of $M_\ell$ into $M_3$ over
5268: $M_0$ and ${\text{\rm \ortp\/}}
5269: (f_\ell(a_\ell),f_{3-\ell}(M_{3-\ell}),M_3)$ does not fork over $M_0$.
5270: \ermn
5271: 2) In fact we use only Axioms (A),(C),(E)(b),(d),(f),(g) only.
5272: \bn
5273: \epsfbox{600pic.eps}
5274:
5275: \endproclaim
5276: \bigskip
5277:
5278: \demo{Proof} By Axiom (E)(g) (existence) applied with \ortp$(a_2,M_0,M_2),
5279: M_0,M_1$ here standing for $p,M,N$ there; there is $q_1$ such that:
5280: \mr
5281: \item "{$(a)$}" $q_1 \in {\Cal S}^{\text{bs}}(M_1)$
5282: \sn
5283: \item "{$(b)$}" $q_1$ does not fork over $M_0$
5284: \sn
5285: \item "{$(c)$}" $q_1 \restriction M_0 = \text{ \ortp}(a_2,M_0,M_2)$.
5286: \ermn
5287: By the definition of types and as ${\frak K}_\lambda$ has amalgamation
5288: (by Axiom (C)) there are $N_1,f_1$ such that
5289: \mr
5290: \item "{$(d)$}" $M_1 \le_{\frak K} N_1 \in K_\lambda$
5291: \sn
5292: \item "{$(e)$}" $f_1$ is a $\le_{\frak K}$-embedding of $M_2$ into $N_1$
5293: over $M_0$
5294: \sn
5295: \item "{$(f)$}" $f_1(a_2)$ realizes $q_1$ inside $N_1$.
5296: \ermn
5297: Now consider Axiom (E)(f) (symmetry) applied with $M_0,N_1,
5298: a_1,f_1(a_2)$ here standing
5299: for $M_0,M_3,a_1,a_2$ there; now as clause $(\alpha)$ of (E)(f) holds (use
5300: $M_1,N_1$ for $M_1,M'_3$) we get that clause $(\beta)$ of (E)(f) holds which
5301: means that there are $N_2,N^*_2$ (standing for $M'_3,M_2$ in clause $(\beta)$
5302: of (E)(f)) such that:
5303: \mr
5304: \item "{$(g)$}" $N_1 \le_{\frak K} N_2 \in K_\lambda$
5305: \sn
5306: \item "{$(h)$}" $M_0 \cup \{f_1(a_2)\} \subseteq N^*_2 \le_{\frak K} N_2$
5307: \sn
5308: \item "{$(i)$}" \ortp$(a_1,N^*_2,N_2) \in
5309: {\Cal S}^{\text{bs}}(N^*_2)$ does not fork over $M_0$.
5310: \ermn
5311: As ${\frak K}_\lambda$ has amalgamation (see Axiom (C)) and the
5312: definition of type and as
5313: \nl
5314: $\ortp(f_1(a_2),M_0,f_1(M_2)) =
5315: \text{ \ortp}(f_1(a_2),M_0,N_2)=
5316: \text{ \ortp}(f_1(a_2),M_0,N^*_2)$, we can find $N^*_3,f_2$ such that
5317: \mr
5318: \item "{$(j)$}" $N^*_2 \le_{\frak K} N^*_3 \in K_\lambda$
5319: \sn
5320: \item "{$(k)$}" $f_2$ is a $\le_{\frak K}$-embedding
5321: \footnote{we could have chosen $N^*_3 = N_2,f_2 = \text{ id}_{f_1(M_2)}$}
5322: of $f_1(M_2)$ into
5323: $N^*_3$ over $M_0 \cup \{f_1(a_2)\}$.
5324: \ermn
5325: As by clause (i) above \ortp$(a_1,N^*_2,N_2) \in {\Cal S}^{\text{bs}}(N^*_2)$,
5326: so by Axiom (E)(g) (extension existence) there are $N_3,f_3$ such that
5327: \mr
5328: \item "{$(l)$}" $N_2 \le_{\frak K} N_3 \in K_\lambda$
5329: \sn
5330: \item "{$(m)$}" $f_3$ is a $\le_{\frak K}$-embedding of $N^*_3$ into $N_3$
5331: over $N^*_2$
5332: \sn
5333: \item "{$(n)$}" \ortp$(a_1,f_3(N^*_3),N_3) \in {\Cal S}^{\text{bs}}
5334: (N^*_3)$ does not fork over $N^*_2$.
5335: \ermn
5336: By Axiom (E)(d) (transitivity) using clauses (i) + (n) above we have
5337: \mr
5338: \item "{$(o)$}" \ortp$(a_1,f_3(N^*_3),N_3) \in {\Cal S}^{\text{bs}}
5339: (N^*_3)$ does not fork over $M_0$.
5340: \ermn
5341: Letting $f = f_3 \circ f_2 \circ f_1$ as $f(M_2) \subseteq f_3(N^*_3)$
5342: by clauses $(e),(k),(m)$ we have
5343: \mr
5344: \item "{$(p)$}" $f$ is a $\le_{\frak K}$-embedding of $M_2$ into $N_3$
5345: over $M_0$.
5346: \ermn
5347: By (E)(b) (monotonicity) and clause (o) and clause (p)
5348: \mr
5349: \item "{$(q)$}" \ortp$(a_1,f(M_2),N_3) \in
5350: {\Cal S}^{\text{bs}}(f(M_2))$ does not
5351: fork over $M_0$.
5352: \ermn
5353: As \ortp$(f_1(a_2),M_1,N_3) = \text{ \ortp}(f_1(a_2),M_1,N_1) = q_1$ does not fork
5354: over $M_0$ by clauses (b) + (f), and $f_2(f_1(a_2)) = f_1(a_2)$ by
5355: clause (k) and $f_3(f_1(a_2)) = f_1(a_2)$ by clauses (m) + (h), we get
5356: \mr
5357: \item "{$(r)$}" \ortp$(f(a_2),M_1,N_3) \in {\Cal S}^{\text{bs}}(M_1)$ does not
5358: fork over $M_0$.
5359: \ermn
5360: So by clauses (o) and (r) we have
5361: id$_{M_1},f,N_3$ are as required on $f_1,f_2,M_3$ in our desired
5362: conclusion. \hfill$\square_{\scite{600-1.15}}$
5363: \enddemo
5364: \bigskip
5365:
5366: \proclaim{\stag{600-1.16A} Claim} 1) In the local character
5367: Axiom (E)(c) of Definition
5368: \scite{600-1.1} if ${\Cal S}^{\text{bs}}_{\frak s} = {\Cal
5369: S}^{\text{na}}_{{\frak K}_{\frak s}}$
5370: where ${\Cal S}^{\text{na}}_{{\frak K}_s}(M) = \{\text{\rm
5371: \ortp}(a,M,N):M \le_{\frak s} N$ and $a \in N \backslash M\}$
5372: \ub{then} it
5373: suffices to restrict ourselves to the case that $\delta$ has
5374: cofinality $\aleph_0$ (i.e., the general case follows from this
5375: special case and the other axioms). \nl
5376: 2) In fact in part (1) we need only Axioms (E)(b),(h)
5377: and you may say (A),(D)(a),(E)(a). \nl
5378: 3) If ${\Cal S}^{\text{bs}} = {\Cal S}^{\text{na}}$ \ub{then} the
5379: continuity Axiom (E)(h) follows from the rest. \nl
5380: 4) In (3) actually we need only Axioms (E)(c), (local character)
5381: (d), (transitivity) and you may say (A),(D)(a),(E)(a).
5382: \endproclaim
5383: \bigskip
5384:
5385: \demo{Proof} 1), 2) Let $\langle M_i:i \le \delta +1 \rangle$ be
5386: $\le_{{\frak K}_\lambda}$-increasing, $a \in M_{\delta +1} \backslash
5387: M_\delta$ and without loss of generality
5388: $\aleph_0 < \delta = \text{ cf}(\delta)$, so for every
5389: $\alpha \in S := \{\alpha < \delta:\text{cf}(\alpha) = \aleph_0\}$,
5390: \ortp$(a,M_\alpha,M_{\delta +1}) \in {\Cal S}^{\text{bs}}(M_\alpha)$ by
5391: the assumption ``$S^{\text{bs}}_{\frak s} = {\Cal
5392: S}^{\text{na}}_{{\frak K}_{\frak s}}$ hence
5393: there is $\beta_\alpha < \alpha$ such that \ortp$(a,M_\alpha,M_{\delta
5394: +1})$ does not fork over $M_{\beta_\alpha}$, so for some
5395: $\beta < \delta$ the set $S_1 = \{\alpha \in S:\beta_\alpha = \beta$)
5396: is a stationary subset of $\delta$.
5397: By Axiom $(E)(b)$ (monotonicity) it follows that for any
5398: $\gamma_1 \le \gamma_2$ from $[\beta,\delta)$ the type
5399: $\ortp(a,M_{\gamma_2},M_{\delta +1}) \in {\Cal
5400: S}^{\text{bs}}(M_{\gamma_2})$ does not fork over $M_{\gamma_1}$. Now
5401: for any $\gamma \in [\beta,\delta)$ the type $\ortp(a,M_\delta,M_{\delta +1})$
5402: does not fork over $M_\gamma$ by applying $(E)(h)$ (continuity) to
5403: $\langle M_\alpha:\alpha \in [\gamma,\delta +1]$ so we have finished.
5404: \nl
5405: 3),4) So assume $\langle M_i:i \le \delta \rangle$ is
5406: $\le_{\frak K}$-increasing all in $K_\lambda$ and $\delta$ is a limit ordinal,
5407: $p \in {\Cal S}(M_\delta)$ and $p_i := p \restriction M_i \in
5408: {\Cal S}^{\text{bs}}(M_i)$ does not fork over $M_0$ for each $i <
5409: \delta$; we should prove
5410: that $p \in {\Cal S}^{\text{bs}}(M_\delta)$ and $p$ does not fork over
5411: $M_0$.
5412:
5413: First, for each $i < \delta,p_i \in {\Cal S}^{\text{bs}}(M_i)$ hence
5414: $p_i$ is not realized in $M_i$. As
5415: $M_\delta = \cup\{M_i:i < \delta\}$ clearly $p$ is not realized in
5416: $M_\delta$ so $p \in {\Cal S}^{\text{na}}(M_\delta) = {\Cal
5417: S}^{\text{bs}}(M_\delta)$.
5418:
5419: Second, by Ax(E)(c) the type $p$ does not fork over $M_j$
5420: for some $j < \delta$. As $p_j = p \restriction M_j$ does not fork
5421: over $M_0$ (by assumption) by the transitivity Axiom $(E)(d)$, we get that $p$
5422: does not fork over $M_0$, as required. \hfill$\square_{\scite{600-1.16A}}$
5423: \enddemo
5424: \bigskip
5425:
5426: \remark{Remark} So in some sense by \scite{600-1.16A} we can omit in
5427: \scite{600-1.1}, the local character Axiom $(E)(c)$ \ub{or} the continuity
5428: Axiom $(E)(h)$ but \ub{not} both. In fact (under reasonable
5429: assumptions) they are equivalent.
5430: \endremark
5431: \bigskip
5432:
5433: \proclaim{\stag{600-1.16B} Claim} In Definition \scite{600-1.1}, Clause
5434: (E)(d), i.e., transitivity of non-forking follows from
5435: (A),(C),(D)(a),(b),(E)(a),(b),(e),(g).
5436: \endproclaim
5437: \bigskip
5438:
5439: \demo{Proof} As ${\frak K}_\lambda$ is an $\lambda$-a.e.c. with amalgamation,
5440: types as well as restriction of types are not only well defined but
5441: are ``rasonable".
5442:
5443: So assume $M_0 \le_{\frak s} M'_0 \le_{\frak s} M''_0 \le_{\frak s}
5444: M_3,a \in M_3$ and $p'' := \text{ \ortp}_{\frak s}(a,M''_0,M_3)$ does not
5445: fork over $M'_0$ and $p' := \text{ \ortp}_{\frak s}(a,M'_0,M_3)$ does not
5446: fork over $M_0$. Let $p = p' \restriction M_0$. As $p'$ does not fork
5447: over $M_0$, by Axiom $(E)(a)$ we have
5448: $p' \in {\Cal S}^{\text{bs}}(M'_0)$ and $p =
5449: \text{ \ortp}(a,M_0,M_3) = p' \restriction M_0$ belongs to
5450: ${\Cal S}^{\text{bs}}(M_0)$. As $p''$ does not fork over $M'_0$ clearly $p''
5451: \in {\Cal S}^{\text{bs}}(M''_0)$ and recall $p'' \restriction M'_0 = p'$. By
5452: the existence axiom $(E)(g)$ the type $p$ has an extension $q'' \in {\Cal
5453: S}^{\text{bs}}(M''_0)$ which does not fork over $M_0$. By the
5454: monotonicity Axiom (E)(b) the type
5455: $q''$ does not fork over $M'_0$ and $q' = q''
5456: \restriction M'_0$ does not fork over $M_0$. As $p',q' \in {\Cal
5457: S}^{\text{bs}}(M'_0)$ do not fork over $M_0$ and $p' \restriction
5458: M_0 = p = q'' \restriction M_0 = q' \restriction M_0$, by the
5459: uniqueness Axiom Ax(E)(e), we have $p'=q'$.
5460: Similarly $p'' = q''$, but $q''$ does not fork over $M_0$ hence $p''$
5461: does not fork over $M_0$ as required. \hfill$\square_{\scite{600-1.16B}}$
5462: \enddemo
5463: \bigskip
5464:
5465: \proclaim{\stag{600-1.16E} Claim} 1) The symmetry axiom (E)(f) is
5466: equivalent to (E)(f)$'$ and to (E)(f)$'$ below if we assume
5467: (A),(B),(C),(D)(a0,(b),(E)(a),(b),(g) in Definition \scite{600-1.1}
5468: \mr
5469: \item "{$(E)(f)'$}" there are no
5470: $M_\ell(\ell \le 3)$ and $a_\ell(\ell \le 2)$ such that
5471: {\roster
5472: \itemitem{ $(a)$ } $M_0 \le_{\frak s} M_1 \le_{\frak s} M_2
5473: \le_{\frak s} M_3$
5474: \sn
5475: \itemitem{ $(b)$ } $\ortp(a_\ell,M_\ell,M_{\ell +1})$ does not fork over
5476: $M_0$ for $\ell = 0,1,2$
5477: \sn
5478: \itemitem{ $(c)$ } $\ortp_{\frak s}(a_0,M_0,M_1) =
5479: \ortp_{\frak s}(a_2,M_0,M_3)$
5480: \sn
5481: \itemitem{ $(d)$ } $\ortp_{\frak s}(\langle a_0,a_1\rangle,M_0,M_1) \ne
5482: \ortp_{\frak s}(\langle a_2,a_1\rangle,M_0,M_1)$.
5483: \endroster}
5484: \endroster
5485: \endproclaim
5486: \bigskip
5487:
5488: \demo{Proof} Easy.
5489: \enddemo
5490: \bn
5491: \centerline{$* \qquad * \qquad *$}
5492: \bn
5493: A most interesting case of \scite{600-1.14} is the following. In
5494: particular it tells us that the categoricity assumption is not so rare
5495: and it will have essential uses here.
5496: \proclaim{\stag{600-1.16} Claim} If
5497: ${\frak s} = ({\frak K},\nonfork{}{}_{\lambda},
5498: {\Cal S}^{\text{bs}})$ is a good $\lambda$-frame and $M \in K_\lambda$ is a
5499: superlimit model in ${\frak K}_\lambda$ and we define ${\frak s}' =
5500: {\frak s}^{[M]} = {\frak s}[M] = ({\frak K}[{\frak s}^{[M]}],
5501: \nonfork{}{}_{\lambda}[{\frak s}^{[M]}],{\Cal S}^{\text{bs}}
5502: [{\frak s}^{[M]}])$ by
5503:
5504: $$
5505: {\frak K}[{\frak s}^{[M]}] = {\frak K}^{[M]}, \text{ see Definition
5506: \scite{600-0.33} so } {\frak K}_{{\frak s}[M]} = {\frak K} \restriction
5507: \{N:N \cong M\}
5508: $$
5509:
5510: $$
5511: \nonfork{}{}_{\lambda}[{\frak s}^{[M]}] = \{(M_0,M_1,a,M_3) \in
5512: \nonfork{}{}_{\lambda}:M_0,M_1,M_3 \in K^{[M]}_\lambda\}
5513: $$
5514:
5515: $$
5516: \align
5517: {\Cal S}^{\text{bs}}[{\frak s}^{[M]}] =
5518: \bigl\{{\text{\rm \ortp\/}}_{{\frak K}[M]}(a,M_0,M_1):&M_0
5519: \le_{\frak K} M_1,M_0 \in K^{[M]}_\lambda,N \in K^{[M]}_\lambda \\
5520: &\text{and {\rm \ortp}}_{\frak K} (a,M_0,M_1) \in
5521: {\Cal S}^{\text{bs}}(M_0) \bigr\}.
5522: \endalign
5523: $$
5524: \mn
5525: \ub{Then}
5526: \mr
5527: \item "{$(a)$}" ${\frak s}'$ is a good $\lambda$-frame
5528: \sn
5529: \item "{$(b)$}" ${\frak K}[{\frak s}'] \subseteq {\frak K}_{\ge \lambda}
5530: [{\frak s}]$
5531: \sn
5532: \item "{$(c)$}" $\le_{{\frak K}[{\frak s}']} = \le_{\frak K} \restriction
5533: K[{\frak s}']$
5534: \sn
5535: \item "{$(d)$}" $K_\lambda[{\frak s}']$ is categorical.
5536: \endroster
5537: \endproclaim
5538: \bigskip
5539:
5540: \demo{Proof} Straight by \scite{600-0.31}, \scite{600-0.34}, \scite{600-1.14}.
5541: \hfill$\square_{\scite{600-1.16}}$
5542: \enddemo
5543: \bigskip
5544:
5545: %% end of file 600A.tex
5546: ------------------------------------------------------------
5547: % begin file 600B.tex
5548: \newpage
5549:
5550: \head {\S3 Examples} \endhead \resetall \sectno=3
5551: \spuriousreset
5552: \bigskip
5553:
5554: We show here that the context from \S2 occurs in earlier investigation: in
5555: \cite{Sh:88} = \chaptercite{88r}, \cite{Sh:576}, \cite{Sh:48} (and \cite{Sh:87a},
5556: \cite{Sh:87b}). Of course, also the class $K$ of models of
5557: a superstable (first order) theory $T$, with $\le_{\frak K} = \prec$ and
5558: ${\Cal S}^{\text{bs}}(M)$ being the set of
5559: regular types (also just ``the set non-algebraic types" works)
5560: with $\nonfork{}{}_{}(M_0,M_1,a,M_3)$ iff $M_0 \le_{\frak K} M_1
5561: \le_{\frak K} M_3$ are in $K_\lambda,a \in M_3$ and \ortp$(a,M_1,M_3) \in
5562: {\Cal S}^{\text{bs}}(M_1)$ does not fork over $M_0$, (in the sense of
5563: \cite[III]{Sh:c}, of course). The
5564: reader may concentrate on \scite{600-Ex.4} (or \scite{600-Ex.1}) below
5565: for easy life. \nl
5566: Note that \scite{600-Ex.1} (or \scite{600-Ex.1A}) will be used to continue
5567: \cite{Sh:88} = \chaptercite{88r} and also to give an alternative
5568: proof to the theorem of \cite{Sh:87a}, \cite{Sh:87b} + (deducing
5569: ``there is a model in $\aleph_n$" if there are not too many models in
5570: $\aleph_\ell$ for $\ell < n$) and note that
5571: \scite{600-Ex.1A} will be used to continue \cite{Sh:48}, i.e., on $\psi \in
5572: \Bbb L_{\omega_1,\omega}(\bold Q)$ and \scite{600-Ex.4} will be used
5573: to continue \cite{Sh:576}. Many of the axioms from \scite{600-1.1} are easy.
5574: \bigskip
5575:
5576: \subhead {(A) The superstable prototype} \endsubhead
5577:
5578: \proclaim{\stag{600-Ex.0} Claim} Assume $T$ is a first order
5579: complete theory and $\lambda$ be a cardinal $\ge |T| + \aleph_0$;
5580: let ${\frak K} = {\frak K}_{T,\lambda} =
5581: (K_{T,\lambda} \le_{{\frak K}_{T,\lambda}})$ be defined by:
5582: \mr
5583: \item "{$(a)$}" $K_{T,\lambda}$
5584: is the class of models of $T$ of cardinality $\ge \lambda$
5585: \sn
5586: \item "{$(b)$}" $\le_{{\frak K}_{T,\lambda}}$ is ``being elementary submodel".
5587: \ermn
5588: 0) ${\frak K}$ is an a.e.c. with {\rm LS}$({\frak K}) = \lambda$.
5589: \nl
5590: 1) If $T$ is superstable, stable in $\lambda$, \ub{then}
5591: ${\frak s} = {\frak s}_{T,\lambda}$ is a good $\lambda$-frame
5592: when ${\frak s} = ({\frak K}_{T,\lambda}{\Cal S}^{\text{bs}},\nonfork{}{}_{})$
5593: is defined by:
5594: \mr
5595: \item "{$(c)$}" $p \in {\Cal S}^{\text{bs}}(M)$ \ub{iff} $p =
5596: \ortp_{{\frak K}_{t,\lambda}}(a,M,N)$ for some $a,N$ such that
5597: $\sftp_{\Bbb L(\tau_T)}(a,M,N)$, see Definition \scite{600-EX.0.7} is a
5598: non-algebraic complete 1-type over $M$
5599: \sn
5600: \item "{$(d)$}" $\nonfork{}{}_{}(M_0,M_1,a,M_3)$ iff $M_0 \prec M_1
5601: \prec M_3$ are in $K_{T,\lambda}$ and $a \in M_3$ and
5602: $\sftp_{\Bbb L(\tau_T)}(a,M_1,M_3)$ is a type that does not fork over
5603: $M_0$ in the sense of \cite[III]{Sh:c}.
5604: \ermn
5605: 2) Let $\kappa = \text{\rm cf}(\kappa) \le \lambda$. The model
5606: $M$ is a $(\lambda,\kappa)$-brimmed model for ${\frak K}_{T,\lambda}$
5607: iff (i)+(ii) or (i)+(iii) where
5608: \mr
5609: \widestnumber\item{$(iii)$}
5610: \item "{$(i)$}" $T$ is stable in $\lambda$
5611: \sn
5612: \item "{$(ii)$}" $\kappa = { \text{\rm cf\/}}(\kappa) \ge \kappa(T)$
5613: and $M$ is a saturated model of $T$ of cardinality $\lambda$
5614: \sn
5615: \item "{$(iii)$}" $\kappa = { \text{\rm cf\/}}(\kappa) < \kappa(T)$
5616: and there is a $\prec$-increasing continuous sequence $\langle M_i:i
5617: \le \kappa \rangle$ (by $\prec$, equivalently
5618: by $\le_{\frak s}$) such that $M = M_\kappa$ and
5619: $(M_{i+1},c)_{c \in M_i}$ is saturated for $i < \kappa$.
5620: \ermn
5621: 2A) So there is a $(\lambda,\kappa)$-brimmed model for
5622: ${\frak K}_{T,\lambda}$ iff $T$ is stable in $\lambda$.
5623: \nl
5624: 3) $M$ is $(\lambda,\kappa)$-brimmed over $M_0$ in ${\frak K}_{T,\lambda}$
5625: iff $(M,c)_{c \in M_0}$ is $(\lambda,\kappa)$-brimmed.
5626: \nl
5627: 4) Assume $T$ is superstable first order complete theory stable in $\lambda$
5628: and we define ${\frak s}^{\text{reg}}_{T,\lambda}$ as above only
5629: ${\Cal S}^{\text{bs}}(M)$ is the set of regular types
5630: $p \in {\Cal S}_{{\frak K}_T}(M)$. \ub{Then}
5631: ${\frak s}^{\text{reg}}_{T,\lambda}$ is a good $\lambda$-frame.
5632: \nl
5633: 5) For $\kappa \le \lambda$ or $\kappa = \aleph_\varepsilon$ (abusing
5634: notation), ${\frak s}^\kappa_{T,\lambda}$ is defined similarly
5635: restricting ourselves to $\bold F^a_\kappa$-saturated models. (Let
5636: ${\frak s}^0_{t,\lambda} = {\frak s}_{T,\lambda}$.) If $T$ is
5637: superstable, stable in $\lambda$ \ub{then} ${\frak s}^\kappa_{t,\lambda}$ is
5638: a good $\lambda$ frame.
5639: \endproclaim
5640: \bigskip
5641:
5642: \remark{Remark} We can replace (c) of \scite{600-Ex.0} by:
5643: \mr
5644: \item "{$(c)'$}" $p \in {\Cal S}^{\text{bs}}(M)$ iff $p =
5645: \ortp_{{\frak K}_{T,\lambda}}(a,M,N)$ for some $a,N$ such that
5646: tp$_{\Bbb L(\tau_T)}(a,M,N)$ is a complete 1-type over $M$
5647: \ermn
5648: except that clause (D)(b) of Definition \scite{600-1.1} fail.
5649: In fact the proofs are easier in this case; of course, the two meaning
5650: of types essentially agree.
5651: \endremark
5652: \bigskip
5653:
5654: \demo{Proof} 0),1),2),2A),3) Obvious (see \cite{Sh:c}).
5655: \nl
5656: 4) As in (1), except density of regular types which holds
5657: by \cite{HuSh:342}). \nl
5658: 5) Also by \cite{Sh:c}. \hfill$\square_{\scite{600-Ex.0}}$
5659: \enddemo
5660: \bn
5661: Recall
5662: \definition{\stag{600-EX.0.7} Definition} 1) For a logic ${\Cal L}$ and
5663: vocabulary $\tau,{\Cal L}(\tau)$ is the set of ${\Cal L}$-formulas in
5664: this vocabulary.
5665: \nl
5666: 2) $\Bbb L = \Bbb L_{\omega,\omega}$ is first order logic.
5667: \nl
5668: 3) A theory in ${\Cal L}(\tau)$ is a set of sentences from
5669: ${\Cal L}(\tau)$ which we assume has a model if not said otherwise.
5670: Similarly in a language $L(\subseteq {\Cal L}(\tau))$
5671: \enddefinition
5672: \bn
5673: Very central in \chaptercite{88r} but
5674: peripheral here (except when in (parts of) \S3 we
5675: continue \chaptercite{88r} in our framework) is:
5676: \definition{\stag{600-0.4} Definition} Let $T_1$ be a theory in
5677: $\Bbb L(\tau_1), \tau \subseteq \tau_1$ vocabularies,
5678: $\Gamma$ a set of types in $\Bbb L(\tau_1)$; (i.e.
5679: for some $m$, a set of formulas $\varphi(x_0,\dotsc,x_{m-1}) \in
5680: \Bbb L(\tau_1)$). \newline
5681: 1) EC$(T_1,\Gamma) = \{M:M \text{ a } \tau_1 \text{-model of } T_1
5682: \text{ which omits every } p \in \Gamma\}$. \newline
5683: (So \wilog \, $\tau_1$ is reconstructible from $T_1,\Gamma$) and
5684:
5685: $$
5686: \text{PC}_\tau(T_1,\Gamma) = \text{ PC}(T_1,\Gamma,\tau)
5687: = \{M:M \text{ is a }
5688: \tau \text{-reduct of some } M_1 \in \text{ EC}(T_1,\Gamma)\}.
5689: $$
5690: \mn
5691: 2) We say that ${\frak K}$ is PC$^\mu_\lambda$ or PC$_{\lambda,\mu}$
5692: \ub{if} for some $T_1,T_2,\Gamma_1,\Gamma_2$ and $\tau_1$ and
5693: $\tau_2$ we have:
5694: ($T_\ell$ a first order theory in the vocabulary $\tau_\ell,
5695: \Gamma_\ell$ a set of types in $\Bbb L(\tau_\ell)$ and)
5696: $K = \text{ PC}(T_1,\Gamma_1,\tau_{\frak K})$ and
5697: $\{(M,N):M \le_{\frak K} N$ and $M,N \in K\} = \text{ PC}
5698: (T_2,\Gamma_2,\tau')$ where
5699: \newline
5700: $\tau' = \tau_{\frak K} \cup \{P\}$
5701: ($P$ a new one place predicate and $(M,N)$ means the $\tau'$-model
5702: $N^+$ expanding $N$ where $P^{N^+} = |M|$), $|T_\ell| \le \lambda,
5703: |\Gamma_\ell| \le \mu$ for $\ell = 1,2$.
5704: \nl
5705: 3) If $\mu = \lambda$, we may omit $\mu$.
5706: \enddefinition
5707: \bn
5708: \subhead {(B) The
5709: abstract elementary class which is PC$_{\aleph_0}$} \endsubhead
5710:
5711: \proclaim{\stag{600-Ex.1} Theorem} Assume
5712: $2^{\aleph_0} < 2^{\aleph_1}$ and consider the statements
5713: \medskip
5714: \roster
5715: \item "{$(\alpha)$}" ${\frak K}$ is an abstract elementary class with
5716: ${\text{\rm LS\/}}({\frak K}) = \aleph_0$
5717: (the last phrase follows by clause $(\beta))$ and $\tau = \tau({\frak
5718: K})$ is countable
5719: \sn
5720: \item "{$(\beta)$}" ${\frak K}$ is ${\text{\rm PC\/}}_{\aleph_0}$,
5721: equivalently for some $\psi_1,\psi_2 \in \Bbb L_{\omega_1,\omega}
5722: (\tau_1)$ where
5723: $\tau_1$ is a countable vocabulary extending $\tau$ we have
5724: $$
5725: \gather
5726: K = \{M_1 \restriction \tau:M_1 \text{ a model of } \psi_1\} \\
5727: \{(N,M):M \le_{\frak K} N\} = \{(N_1 \restriction \tau,M_1 \restriction
5728: \tau):(N_1,M_1) \text{ a model of } \psi_2\}
5729: \endgather
5730: $$
5731: \noindent
5732: \item "{$(\gamma)$}" $1 \le \dot I(\aleph_1,{\frak K}) < 2^{\aleph_1}$
5733: \sn
5734: \item "{$(\delta)$}" ${\frak K}$ is categorical in $\aleph_0$, has the
5735: amalgamation property in $\aleph_0$ and is stable in $\aleph_0$
5736: \sn
5737: \item "{$(\delta)^-$}" like $(\delta)$ but ``stable in $\aleph_0$" is
5738: weakened to: $M \in {\frak K}_{\aleph_0} \Rightarrow |{\Cal S}(M)|
5739: \le \aleph_1$
5740: \sn
5741: \item "{$(\varepsilon)$}" all models of ${\frak K}$ are
5742: $\Bbb L_{\infty,\omega}$-equivalent and $M \le_{\frak K} N \Rightarrow M
5743: \prec_{{\Bbb L}_{\infty,\omega}} N$.
5744: \ermn
5745: For $M \in {\frak K}_{\aleph_0}$ we define ${\frak K}'_M$ as
5746: follows: the class of members is \nl
5747: $\{N \in K:N \equiv_{{\Bbb L}_{\infty,\omega}} M\}$ and $N_1
5748: \le_{{\frak K}'_M} N_2$ iff $N_1 \le_{\frak K} N_2 \and N_1
5749: \prec_{{\Bbb L}_{\infty,\omega}} N_2$.
5750: \medskip
5751: \noindent
5752: 1) Assume $(\alpha) + (\beta) + (\gamma)$, \ub{then} for some
5753: $M \in {\frak K}_{\aleph_0}$ the class ${\frak K}'_M$ satisfies
5754: $(\alpha) + (\beta) + (\gamma) + (\delta)^- + (\varepsilon)$
5755: (any $M \in {\frak K}_{\aleph_0}$ such that
5756: $({\frak K}'_M)_{\aleph_1} \ne \emptyset$ will do and there are such
5757: $M \in K_{\aleph_0}$).
5758: Moreover, if ${\frak K}$ satisfies $(\delta)$ then also ${\frak
5759: K}'_M$ satisfies it; also trivially $K'_M \subseteq K$ and
5760: $\le_{{\frak K}'_M} \subseteq \le_{\frak K}$.
5761: \nl
5762: 1A) Also there is ${\frak K}'$ such that: ${\frak K}'$ satisfies
5763: $(\alpha) + (\beta) + (\gamma) + (\delta) + (\varepsilon)$, and for every
5764: $\mu$ we have $K'_\mu \subseteq K_\mu$.
5765: In fact, in the notation of \marginbf{!!}{\cprefix{88r}.\scite{88r-5.6}} for every $\alpha <
5766: \omega_1$ we can choose ${\frak K}' = {\frak K}_{\bold D_\alpha}$.
5767: \nl
5768: 2) Assume $(\alpha) + (\beta) + (\gamma) + (\delta)$. \ub{Then}
5769: $({\frak K},\nonfork{}{}_{},{\Cal S}^{\text{bs}})$
5770: is a good $\aleph_0$-frame for
5771: some $\nonfork{}{}_{}$ and ${\Cal S}^{\text{bs}}$. \nl
5772: 3) In fact, in part (2) we can choose ${\Cal S}^{\text{bs}}(M) =
5773: \{p \in {\Cal S}(M):p \text{ not algebraic}\}$ and
5774: $\nonfork{}{}_{}$ is defined by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.11}}
5775: (the definable extensions).
5776: \endproclaim
5777: \bigskip
5778:
5779: \remark{Remark} 1) In \marginbf{!!}{\cprefix{88r}.\scite{88r-5.23}} we use the
5780: assumption $\dot I(\aleph_2,K) < \mu_{\text{unif}}(\aleph_2)$.
5781: But this Theorem is not used here!
5782: \nl
5783: 2) Note that ${\frak K}'_M$ is related to $K^{[M]}$ from Definition
5784: \scite{600-0.33} but is different.
5785: \nl
5786: 3) In the proof we relate the types in the sense of
5787: ${\Cal S}_{\frak s}(M)$, and those in \sectioncite[\S5]{88r}. Now in
5788: \sectioncite[\S5]{88r} we have lift types, from ${\frak K}_{\aleph_0}$ to
5789: any ${\frak K}_\mu$, i.e., define $\bold D(N)$ for $N \in {\frak
5790: K}_\mu$. In $\mu > \aleph_0$, in general we do not know how to
5791: relate them to types ${\Cal S}_{{\frak K}_{\frak s}}(N)$. But when
5792: ${\frak s}^+$ is defined (in the ``successful" cases, see \S8 here and
5793: \sectioncite[\S1]{705}) we can get the parallel claim.
5794: \endremark
5795: \bn
5796: \ub{Discussion}: 1) What occurs if we do not pass in \scite{600-Ex.1} to the
5797: case ``$\bold D(N)$ countable for every $N \in K_{\aleph_0}$"? If we
5798: still assume ``${\frak K}$ categorical in $\aleph_0$" then as $|\bold
5799: D(N_0)| \le \aleph_1$, if we assume ``there is a superlimit model in
5800: ${\frak K}_{\aleph_1}$" we can find a good $\aleph_1$-frame ${\frak
5801: s}$. Not clear if we do not assume categoricity in $\aleph_1$.
5802:
5803: \demo{Proof} 1) Note that for any $M \in K_{\aleph_0}$, the class
5804: ${\frak K}'_M$ satisfies $(\alpha), (\beta), (\varepsilon)$
5805: and it is categorical in $\aleph_0$ and
5806: $(K'_M)_\mu \subseteq K_\mu$ hence $\dot I(\mu,K'_M) \le \dot I(\mu,K)$.
5807: By Theorem \marginbf{!!}{\cprefix{88r}.\scite{88r-3.6}}, (note: if you use the original version
5808: (i.e., \cite{Sh:88}) by its proof or use
5809: it and get a less specified class with the desired properties) for some
5810: $M \in K_{\aleph_0}$ we have $({\frak K}'_M)_{\aleph_1} \ne
5811: \emptyset$. By \marginbf{!!}{\cprefix{88r}.\scite{88r-3.5}} (or \cite[1.4]{Sh:576}(page 46)1.6(page 48))
5812: we get that ${\frak K}'_M$ has
5813: amalgamation in $\aleph_0$ and by \chaptercite{88r} \ub{almost} we get that in
5814: ${\frak K}'_M$ the set ${\Cal S}(M)$ is small; be careful - the types there
5815: are defined differently than here, but by the
5816: amalgamation (in $\aleph_0$)
5817: and the omitting types theorem in this case they are the same,
5818: see more in the proof of part (3) below. So by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.1}},\marginbf{!!}{\cprefix{88r}.\scite{88r-5.2}} we
5819: have $M \in ({\frak K}'_\mu)_{\aleph_0} \Rightarrow
5820: |{\Cal S}_{{\frak K}'_\mu}(M)| \le \aleph_1$.
5821: \nl
5822: Also the second and third sentences in (1) are easy.
5823: \nl
5824: 1A) Use \marginbf{!!}{\cprefix{88r}.\scite{88r-5.18}},\marginbf{!!}{\cprefix{88r}.\scite{88r-5.19}}.
5825: \nl
5826: In more detail, (but not much point in reading without some understanding
5827: of \sectioncite[\S5]{88r}, however we should not use \marginbf{!!}{\cprefix{88r}.\scite{88r-5.23}} as long as
5828: we do not strengthen our assumptions)
5829: by part (1) we can assume that clauses $(\delta)^- +(\varepsilon)$
5830: hold. (Looking at the old version \cite{Sh:88} of
5831: \chaptercite{88r} remember that there $\prec$ means $\le_{\frak K}$.)
5832: We can find $\bold D_* = \bold D^*_\alpha,\alpha < \omega_1$,
5833: which is a good countable diagram (see Definition \marginbf{!!}{\cprefix{88r}.\scite{88r-5.6.1}}
5834: and Fact \marginbf{!!}{\cprefix{88r}.\scite{88r-5.6}} or \marginbf{!!}{\cprefix{88r}.\scite{88r-5.16}}, \marginbf{!!}{\cprefix{88r}.\scite{88r-5.17}}.
5835: So in particular (give the
5836: non-maximality of models below) such
5837: that for some countable $M_0 <_{\frak K} M_1 <_{\frak K} M_2$ we have
5838: $M_m$ is $(\bold D^*(M_\ell),\aleph_0)$-homogeneous for $\ell < m \le 2$.
5839: In \marginbf{!!}{\cprefix{88r}.\scite{88r-5.18}} we define $(K_{\bold D_*},\le_{\bold D_*})$.
5840: By \marginbf{!!}{\cprefix{88r}.\scite{88r-5.19}} the pair $(K_{\bold D_*},\le_{\bold D_*})$
5841: is an abstract elementary class (the choice of $\bold D_*$ a part, e.g.
5842: transitivity = Axiom II which holds
5843: by the $M_\ell$'s above and \marginbf{!!}{\cprefix{88r}.\scite{88r-5.16}}) categorical in
5844: $\aleph_0$ and no maximal countable model (by $\le_{\bold D_*}$, see
5845: \marginbf{!!}{\cprefix{88r}.\scite{88r-5.6}}(2).
5846: Now $\aleph_0$-stability holds by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.19}}(2) and the
5847: equality of the three definitions of types in the proof of parts (2),(3)
5848: and $K_{\bold D_*} \subseteq K$ so we are done by part 3) below.
5849: \nl
5850: 2),3) The first part of the proof serves also part (1) of the
5851: theorem so we assume $(\delta)^-$ instead of $(\delta)$.
5852: We should be careful: the notion of type has three relevant
5853: meanings here.
5854: For $N \in K_{\aleph_0}$ the three definitions for
5855: ${\Cal S}^{< \omega}(N)$ and of $\sftp(\bar a,N,M)$ when $\bar a \in
5856: {}^{\omega >}M,N \le_{\frak K} M \in K_{\aleph_0}$ (of course we can
5857: use just 1-types) are:
5858: \mr
5859: \item "{$(\alpha)$}" the one we use here (recall \scite{600-0.12}) which uses
5860: elementary mappings; for the present proof we
5861: call them ${\Cal S}^{< \omega}_0(M),\sftp_0(\bar a,M,N)$
5862: \sn
5863: \item "{$(\beta)$}" ${\bold S}_1(N)$ which is (recall: materialzie is
5864: close to but different from realize)
5865: $$
5866: \align
5867: \bold D(N) = \bigl\{p:&\,p \text{ a complete }
5868: \Bbb L^0_{\aleph_1,\aleph_0}(N) \text{-type over } N \\
5869: &\text{ (so in each formula only finitely many parameters from } N
5870: \text{ appear)} \\
5871: &\text{ such that for some } M,\bar a \in {}^{\omega >}M,
5872: \bar a \text{ materializes } p \text{ in } (M,N) \bigr\}
5873: \endalign
5874: $$
5875: (``materializing a type" is defined in \marginbf{!!}{\cprefix{88r}.\scite{88r-4.2}}(2)) so
5876: $$
5877: {\bold S}_1(N) = \{\text{\sftp}_1(\bar a,N,M):\bar a \in {}^{\omega >}M
5878: \text{ and } N \le_{\frak K} M \in K_{\aleph_0}\}
5879: $$
5880: where
5881: $$
5882: \text{\sftp}_1(\bar a,N,M) = \{\varphi(\bar x) \in
5883: \Bbb L^0_{\aleph_1,\aleph_0}(N):
5884: M \Vdash^{\aleph_1}_{\frak K} \varphi(\bar a)\}
5885: $$
5886: (see \marginbf{!!}{\cprefix{88r}.\scite{88r-4.2}}(1)
5887: on the meaning of this forcing relation).
5888: \sn
5889: \item "{$(\gamma)$}" $\quad {\bold S}_2(N)$ which is
5890: $$
5891: \align
5892: \bold D^*(N) = \bigl\{p:&\,p \text{ a complete }
5893: \Bbb L^0_{\aleph_1,\aleph_0}(N;N)
5894: \text{-type over } N \\
5895: &\text{ (so in each formula all members of } N \text{ may appear)} \\
5896: &\text{ such that for some } M \in K_{\aleph_0} \text{ and} \\
5897: &\bar a \in {}^{\omega >}M \text{ satisfying } N \le_{\frak K} M
5898: \text{ the sequence} \\
5899: &\bar a \text{ materializes } p \text{ in } (M,N) \bigr\}
5900: \endalign
5901: $$
5902: \mn
5903: so
5904:
5905: $$
5906: {\bold S}_2(N) = \{\sftp_2(\bar a,N,M):\bar a \in
5907: {}^{\omega >}M \text{ and } N \le_{\frak K} M \in K_{\aleph_0}\}
5908: $$
5909:
5910: $$
5911: \text{\sftp}_2(\bar a,N,M) = \{\varphi(\bar x) \in \Bbb L^0_{\aleph_1,\aleph_0}
5912: (N,N):M \Vdash^{\aleph_1}_{\frak K} \varphi(\bar a)\}.
5913: $$
5914: \ermn
5915: As we have amalgamation in $K_{\aleph_0}$ it is enough to prove for
5916: $\ell,m < 3$ that
5917: \mr
5918: \item "{$(*)_{\ell,m}$}" if $k < \omega,
5919: N \le_{\frak K} M \in K_{\aleph_0}$ and $\bar a,\bar b \in {}^k M$, then \nl
5920: $\sftp_\ell(\bar a,N,M) = \sftp_\ell(\bar b,N,M)
5921: \Rightarrow \sftp_m(\bar a,N,M) = \sftp_m(\bar b,N,M)$.
5922: \ermn
5923: Now $(*)_{2,1}$ holds trivially (more formulas) and $(*)_{1,2}$ holds by
5924: \marginbf{!!}{\cprefix{88r}.\scite{88r-5.5}}. By amalgamation in ${\frak K}_{\aleph_0}$,
5925: if $\sftp_0(\bar a,N,M) = \sftp_0(\bar b,N,M)$, then for
5926: some $M',M \le_{\frak K} M' \in
5927: K_{\aleph_0}$ there is an automorphism $f$ of $M'$ over $N$ such that
5928: $f(\bar a) = \bar b$, so trivially $(*)_{0,1},(*)_{0,2}$ hold
5929: (we use the facts that $\sftp_\ell(\bar a,N,M)$ is preserved by
5930: isomorphism and by replacing $M$ by $M_1$ if $M \le_{\frak K} M_2 \in
5931: K_{\aleph_0}$ and $N \cup \bar a \subseteq M_1 \le_{\frak K}
5932: M_2$). Lastly we prove $(*)_{2,0}$. \nl
5933: So $N \le_{\frak K} M \in K_{\aleph_0}$, hence
5934: $\{\text{\ortp}_2(\bar c,N,M):\bar c \in {}^{\omega >} M\} \subseteq
5935: \bold D^*(N)$ is countable so by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.6}}(b),(c) for some
5936: countable $\alpha < \omega_1$
5937: we have $\{\sftp_2(\bar c,N,M):\bar c \in {}^{\omega >} M\} \subseteq
5938: \bold D^*_\alpha(N)$. Now
5939: there is $M' \in K_{\aleph_0}$ such that $M \le_{\frak K}
5940: M',M'$ is $(\bold D^*_\alpha,\aleph_0)$-homogeneous (by
5941: \marginbf{!!}{\cprefix{88r}.\scite{88r-5.6}}(e) see Definition \marginbf{!!}{\cprefix{88r}.\scite{88r-5.7}})
5942: hence $M'$ is $(\bold D^*_\alpha(N),\aleph_0)$-
5943: homogeneous (by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.6}}(f)), and $\sftp_2(\bar a,N,M') =
5944: \sftp_2(\bar b,N,M')$ by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.4.1}}(3)
5945: ($N$ here means $N_0$ there,
5946: that is increasing the model preserve the type). \nl
5947: Lastly by Definition \marginbf{!!}{\cprefix{88r}.\scite{88r-5.7}} there is an automorphism $f$ of $M'$
5948: over $N$ mapping $\bar a$ to $\bar b$, so we have proved $(*)_{2,0}$, so the
5949: three definitions of type are the same.
5950:
5951: Now we define for $M \in K_{\aleph_0}$:
5952: \mr
5953: \item "{$(a)$}" ${\Cal S}^{\text{bs}}(M) = \{p \in
5954: {\Cal S}_{\frak K}(M):p \text{ not
5955: algebraic}\}$
5956: \sn
5957: \item "{$(b)$}" for $M_0,M_1,M_3 \in K_{\aleph_0}$ and an element
5958: $a \in M_3$ we define:
5959: \nl
5960: $\nonfork{}{}_{}(M_0,M_1,a,M_3)$ \ub{iff} $M_0 \le_{\frak K} M_1
5961: \le_{\frak K} M_3$ and $a \in M_3 \backslash M_1$ and \nl
5962: $\sftp_1(a,M_1,M_3)(= \sftp(a,M_1,M_3)$ in \chaptercite{88r}'s notation)
5963: \nl
5964: is definable over some finite $\bar b \in {}^{\omega >} M_0$ (equivalently
5965: is preserved by every automorphism of $M_1$ over $\bar b$ (see
5966: \marginbf{!!}{\cprefix{88r}.\scite{88r-5.11}}) \nl
5967: (equivalently gtp$(a,M_1,M_3)$ is the stationarization of gtp$(a,M_0,M_3)$).
5968: \ermn
5969: Now we should check the axioms from Definition \scite{600-1.1}.
5970: \bn
5971: \ub{Clause (A)}: By clause $(\alpha)$ of the assumption.
5972: \mn
5973: \ub{Clauses (B),(C)}: By clause $(\delta)$ or $(\delta)^-$ of
5974: the assumption except ``the superlimit
5975: $M \in K_{\aleph_0}$ is not $\le_{\frak K}$-maximal" which holds by clause
5976: $(\gamma)+(\delta)$ or $(\gamma) + (\delta)^-$.
5977: \mn
5978: \ub{Clause (D)}: By the definition (note that about clause (d), bs-stability,
5979: that it holds by assumption $(\delta)$, and about clause (c), i.e.,
5980: the density is trivial by the way we have defined ${\Cal S}^{\text{bs}}$).
5981: \mn
5982: \ub{Subclause (E)(a)}: By the definition.
5983: \mn
5984: \ub{Subclause (E)(b)(monotonicity)}:
5985:
5986: Let
5987: $M_0 \le_{\frak K} M'_0 \le_{\frak K} M'_1 \le_{\frak K} M_1 \le_{\frak K}
5988: M_3 \le M'_3$ be all in ${\frak K}_{\aleph_0}$ and assume
5989: $\nonfork{}{}_{}(M_0,M_1,a,M_3)$. So
5990: $M'_0 \le_{\frak K} M'_1 \le_{\frak K} M_3 \le M'_3$ and
5991: $a \in M_3 \backslash M_1
5992: \subseteq M'_3 \backslash M'_1$. Now by the assumption and the
5993: definition of $\nonfork{}{}_{}$, for some $\bar b \in {}^{\omega >}
5994: (M_0)$, gtp$(a,M_1,M_3)$ is definable over $\bar b$. So the same
5995: holds for gtp$(a,M'_1,M_3)$ by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.13}}, in fact
5996: (with the same definition) and hence for gtp$(a,M'_1,M'_3) =
5997: \text{ gtp}(a,M'_1,M_3)$ by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.4.1}}(3), so as $\bar b \in
5998: {}^{\omega >}(M_0) \subseteq {}^{\omega >}(M'_0)$ we have gotten
5999: $\nonfork{}{}_{}(M'_0,M'_1,a,M'_3)$.
6000:
6001: For the additional clause in the monotoncity Axiom, assume
6002: in addition $M'_1 \cup \{a\} \subseteq
6003: M''_3 \le_{\frak K} M'_3$ again by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.4.1}}(3)
6004: clearly gtp$(a,M'_1,M''_3) = \text{
6005: gtp}(a,M'_1,M'_3)$, so (recalling the beginning of the proof) we are done.
6006: \mn
6007: \ub{Sublcause (E)(c)(local character)}:
6008:
6009: So let $\langle M_i:i \le \delta +1 \rangle$ be $\le_{\frak K}$-increasing
6010: continuous in $K_{\aleph_0}$ and $a \in M_{\delta +1}$ and
6011: \ortp$(a,M_\delta,M_{\delta +1}) \in {\Cal S}^{\text{bs}}
6012: (M_\delta)$, so $a \notin
6013: M_\delta$ and gtp$(a,M_\delta,M_{\delta +1})$ is definable over some $\bar b
6014: \in {}^{\omega >}(M_\delta)$ by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.4}}. \nl
6015: As $\bar b$ is finite, for some $\alpha < \delta$ we have
6016: $\bar b \subseteq M_\alpha$,
6017: hence as in the proof of (B)(b), we have
6018: (\ortp$(a,M_\beta,M_{\delta +1}) \in {\Cal S}^{\text{bs}}
6019: (M_\beta)$ trivially and)
6020: \ortp$(a,M_\delta,M_{\delta +1})$ does not fork over $M_\beta$.
6021: \mn
6022: \ub{Sublcause (E)(d)(transitivity)}:
6023:
6024: By \marginbf{!!}{\cprefix{88r}.\scite{88r-5.13}}(2) or even better \marginbf{!!}{\cprefix{88r}.\scite{88r-5.16}}.
6025: \mn
6026: \ub{Subclause (E)(e)(uniqueness)}:
6027:
6028: Holds by the Definition \marginbf{!!}{\cprefix{88r}.\scite{88r-5.11}}.
6029: \mn
6030: \ub{Subclause (E)(f)(symmetry)}:
6031:
6032: By \marginbf{!!}{\cprefix{88r}.\scite{88r-5.20}} + uniqueness we get (E)(f). Actually
6033: \marginbf{!!}{\cprefix{88r}.\scite{88r-5.20}} gives this more directly.
6034: \mn
6035: \ub{Subclause (E)(g)(extension existence)}:
6036:
6037: By \marginbf{!!}{\cprefix{88r}.\scite{88r-5.11}} (i.e., by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.4}} + all $M \in
6038: K_{\aleph_0}$ are $\aleph_0$-homogeneous).
6039:
6040: Alternatively, see \marginbf{!!}{\cprefix{88r}.\scite{88r-5.15}}.
6041: \mn
6042: \ub{Subclause (E)(h)(continuity)}:
6043:
6044: Suppose $\langle M_\alpha:\alpha \le \delta \rangle$ is $\le_{\frak K}$-
6045: increasingly continuous, $M_\alpha \in K_{\aleph_0},\delta < \omega_1,
6046: p \in {\Cal S}(M_\delta)$ and $\alpha < \delta \Rightarrow p
6047: \restriction M_\alpha$ does
6048: not fork over $M_0$. Now we shall use (E)(c)+(E)(d).
6049: As $p \restriction M_\alpha \in {\Cal S}^{\text{bs}}(M_\alpha)$ clearly
6050: $p \restriction M_\alpha$ is not realized in $M_\alpha$ hence $p$
6051: is not realized in $M_\alpha$;
6052: as $M_\delta = \dbcu_{\alpha < \delta} M_\alpha$ necessarily $p$ is not
6053: realized in $M_\delta$, hence $p$ is not algebraic.
6054:
6055: So $p \in {\Cal S}^{\text{bs}}(M_\delta)$. For some finite $\bar b \in
6056: {}^{\omega >}
6057: (M_\delta),p$ is definable over $\bar b$, let $\alpha < \delta$
6058: be such that $\bar b \in {}^{\omega >}(M_\alpha)$, so as in the proof
6059: of (E)(c), (or use it directly) the type $p$ does not fork over
6060: $M_\alpha$. As $p \restriction M_\alpha$ does not fork over $M_0$, by
6061: (E)(d) we get that $p$ does not fork over $M_0$ as required. Actually
6062: we can derive (E)(h) by \scite{600-1.16A}.
6063: \mn
6064: \ub{Subclause (E)(i)(non-forking amalgamation)}:
6065:
6066: One way is by \marginbf{!!}{\cprefix{88r}.\scite{88r-5.20}};
6067: (note that in \marginbf{!!}{\cprefix{88r}.\scite{88r-5.23}} we get more, but
6068: assuming, by our present notation $\dot I(\aleph_2,K) < \mu_{\text{wd}}
6069: (\aleph_2)$); \ub{but} another way is just to use \scite{600-1.15}. \nl
6070: ${{}}$ \hfill$\square_{\scite{600-Ex.1}}$
6071: \enddemo
6072: \bn
6073: \subhead {(C) The uncountable cardinality quantifier case,
6074: $\Bbb L_{\omega_1,\omega}(\bold Q)$} \endsubhead
6075:
6076: Now we turn to $\Bbb L$ sentences in $\Bbb L_{\omega_1,\omega}(\bold Q)$.
6077: \demo{\stag{600-Ex.1A} Conclusion} Assume $\psi \in
6078: \Bbb L_{\omega_1,\omega}(\bold Q)$ and $1 \le \dot I(\aleph_1,\psi)
6079: < 2^{\aleph_1}$ and $2^{\aleph_0} < 2^{\aleph_1}$. \nl
6080: \ub{Then} for
6081: some abstract elementary classes ${\frak K},{\frak K}^+$
6082: (note $\tau_\psi \subset \tau_{\frak K} = \tau_{{\frak K}^+}$) we have:
6083: \mr
6084: \item "{$(a)$}" ${\frak K}$ satisfies $(\alpha),(\beta),(\delta),
6085: (\varepsilon)$ from \scite{600-Ex.1} with $\tau_{\frak K} \supseteq
6086: \tau_\psi$ countable (for $(\gamma),(b)$ is a replacement)
6087: \sn
6088: \item "{$(b)$}" for every $\mu > \aleph_0,
6089: \dot I(\mu,{\frak K}(\aleph_1$-saturated)) $\le \dot I(\mu,\psi)$, where
6090: \footnote{much less than saturation suffice, like ``obeying" $<^{**}$}
6091: ``$\aleph_1$-saturated" is well defined as ${\frak K}_{\aleph_0}$ has
6092: amalgamation, see \scite{600-0.19}
6093: \sn
6094: \item "{$(c)$}" for some $\nonfork{}{}_{},
6095: {\Cal S}^{\text{bs}}$ (and $\lambda =
6096: \aleph_0$), the triple $({\frak K},\nonfork{}{}_{},S^{\text{bs}})$
6097: is as in \scite{600-Ex.1}(2) so is a good $\aleph_0$-frame
6098: \sn
6099: \item "{$(d)$}" every $\aleph_1$-saturated member of ${\frak K}$
6100: belongs to ${\frak K}^+$ and there is an $\aleph_1$-saturated
6101: member of ${\frak K}$ (and naturally it is uncountable)
6102: \sn
6103: \item "{$(e)$}" ${\frak K}^+$ is an a.e.c., has LS number $\aleph_1$
6104: and $\{M \restriction \tau_\psi:M \in {\frak K}^+\} \subseteq
6105: \{M:M \models \psi\}$ and every $\tau$-model $M$ of $\psi$ has a
6106: unique expansion in ${\frak K}^+$ hence $\mu \ge \aleph_1 \Rightarrow
6107: \dot I(\mu,\psi) = \dot I(\mu,{\frak K}^+)$ and ${\frak K}^+$ is the
6108: class of models of some complete $\psi \in \Bbb
6109: L_{\omega_1,\omega}(\bold Q)$.
6110: \endroster
6111: \enddemo
6112: \bigskip
6113:
6114: \demo{Proof} Essentially by \cite{Sh:48} and \scite{600-Ex.1}.
6115:
6116: I feel that upon reading the proof \cite{Sh:48} the proof should not
6117: be inherently difficult, much more so having
6118: read \scite{600-Ex.1}, but will give full details.
6119: \nl
6120: Recall Mod$(\psi)$ is the class of $\tau_\psi$-models of $\psi$. We
6121: can find a countable fragment ${\Cal L}$ of $\Bbb
6122: L_{\omega_1,\omega}(\bold Q)(\tau_\psi)$ to which $\psi$ belongs and a
6123: sentence $\psi_1 \in {\Cal L} \subseteq \Bbb L_{\omega_1,\omega}
6124: (\bold Q)(\tau_\psi)$ such that $\psi_1$ is ``nice" for
6125: \cite[Definition 3.1,3.2]{Sh:48}, \cite[Lemma 3.1]{Sh:48}
6126: \mr
6127: \item "{$\circledast_1$}" $(a) \quad \psi_1$ has uncountable models
6128: \sn
6129: \item "{${{}}$}" $(b) \quad \psi_1 \vdash \psi$, i.e., every model of
6130: $\psi_1$ is a model of $\psi$
6131: \sn
6132: \item "{${{}}$}" $(c) \quad \psi_1$ is $\Bbb
6133: L_{\omega_1,\omega}(\bold Q)$-complete
6134: \sn
6135: \item "{${{}}$}" $(d) \quad$ every model $M \models \psi_1$ realizes
6136: just countably many complete \nl
6137:
6138: \hskip25pt $\Bbb L_{\omega_1,\omega}(\bold
6139: Q)(\tau_\psi)$-types (of any finite arity, over the empty set),
6140: \nl
6141:
6142: \hskip25pt each isolated by a formula in ${\Cal L}$.
6143: \ermn
6144: The proof of $\circledast_1(d)$ is sketched in Theorem \ub{2.5} of
6145: \cite{Sh:48}. The reference to Keisler \cite{Ke71} is to the generalization of
6146: theorems 12 and 28 of Keisler's book from $\Bbb L_{\omega_1,\omega}$ to
6147: $\Bbb L_{\omega_1,\omega}(\bold Q)$, see \marginbf{!!}{\cprefix{88r}.\scite{88r-0.1}}.
6148: \nl
6149: Let
6150: \mr
6151: \item "{$\circledast_2$}" $(i) \quad {\frak K}_0 =
6152: (\text{Mod}(\psi),\prec_{\Cal L})$,
6153: \sn
6154: \item "{${{}}$}" $(ii) \quad {\frak K}_1 =
6155: (\text{Mod}(\psi_1),\prec_{\Cal L})$
6156: \sn
6157: \item "{$\circledast_3$}" ${\frak K}_\ell$ is an a.e.c. with
6158: L.S. number $\aleph_1$ for $\ell=0,1$.
6159: \ermn
6160: Toward defining ${\frak K}$, let $\tau_{\frak K} = \tau_\psi
6161: \cup\{R_{\varphi(\bar x)}:\varphi(\bar x) \in {\Cal
6162: L}\},R_{\varphi(\bar x)}$ a new $\ell g(\bar x)$-predicate and let
6163: $\psi_2 = \psi_1 \wedge \{(\forall \bar y)(R_{\varphi(\bar x)}(\bar y)
6164: = \varphi(\bar y):\varphi(\bar x) \in \Bbb L\}$. For every
6165: $M \in \text{ Mod}(\psi)$ we define $M^+$ by
6166: \mr
6167: \item "{$\circledast_4$}" $M^+$ is $M$ expanded to a
6168: $\tau_{\frak K}$-model by letting $R^{M^+}_{\varphi(\bar x)} = \{\bar
6169: a \in {}^{\ell g(\bar x)} M:M \models \varphi[\bar a]\}$
6170: \sn
6171: \item "{$\circledast_5$}" $(a) \quad {\frak K}^+_0 = (\{M^+:M \in \text{
6172: Mod}(\psi)\},\prec_{\Bbb L})$ is an a.e.c. with LS$({\frak K}^+_0) =
6173: \aleph_1$
6174: \sn
6175: \item "{${{}}$}" $(b) \quad {\frak K}^+_1
6176: = (\{M^+:M \in \text{ Mod}(\psi_1)\},\prec_{\Bbb L})$
6177: is an a.e.c. with LS$({\frak K}^+) = \aleph_1$.
6178: \ermn
6179: Clearly
6180: \mr
6181: \item "{$\circledast_6$}" if $M \models \psi_1$ then
6182: $M^+$ is an atomic model of the complete first-order theory
6183: $T_{\psi_1}$ where $T_{\psi_1}$ is the set of first order consequences
6184: in $\Bbb L(\tau_{\frak K})$ of $\psi_2$.
6185: \ermn
6186: So it is natural to define ${\frak K}$:
6187: \mr
6188: \item "{$\circledast_7$}" $(a) \quad N \in {\frak K}$ iff
6189: {\roster
6190: \itemitem{ $(i)$ } $N$ is a $\tau_{\frak K}$-model which is an atomic
6191: model of $T_{\psi_1}$
6192: \sn
6193: \itemitem{ $(ii)$ } if $\psi_1 \vdash (\forall \bar x)[\varphi_1(\bar
6194: x) = (\bold Q y) \varphi_2(y,\bar x)]$ and
6195: $\varphi_1,\varphi_2 \in {\Cal L}$ and
6196: $N \models \neg R_{\varphi_1(\bar x)}[\bar a]$ then $\{b \in N:N
6197: \models R_{\varphi_2(y,\bar x)}(b,\bar a)\}$ is countable
6198: \endroster}
6199: \item "{${{}}$}" $(b) \quad N_1 \le_{\frak K} N_2$ \ub{iff}
6200: $(N_1,N_2 \in K,N_1 \prec_{\Bbb L} N_2$ equivalently $N_1 \subseteq
6201: N_2$ and) for
6202: \nl
6203:
6204: \hskip25pt $\varphi_1(\bar x),\varphi_2(y,\bar x)$
6205: as in subclause (ii) of clause (a) above,
6206: if $\bar a \in {}^{\ell g(\bar x)}(N_1)$,
6207: \nl
6208:
6209: \hskip25pt $N_1 \models \neg R_{\varphi_1(\bar x)}[\bar a]$ and $b \in
6210: N_2 \backslash N_1$ then $N_2 \models \neg R_{\varphi_2(y,\bar x)}[b,\bar a]$.
6211: \ermn
6212: Observe
6213: \mr
6214: \item "{$\circledast_8$}" $N \in {\frak K}$ iff $N$ is an
6215: atomic $\tau_{\frak K}$-model of the first order
6216: $\Bbb L(\tau_{\frak K})$-consequences $\psi_2$ (i.e. of $\psi$ and
6217: every $\tau_{\frak K}$ sentence of the form $\forall \bar x
6218: [R_\varphi(\bar x) \equiv \varphi(\bar x)])$ and clause (ii) of
6219: $\circledast_7(a)$ holds
6220: \sn
6221: \item "{$\circledast_9$}" ${\frak K}$ is an a.e.c. with
6222: LS$({\frak K}) = \aleph_0$ and is PC$_{\aleph_0},
6223: {\frak K}$ is categorical in $\aleph_0$ (and $\le_{\frak K}$
6224: called $\le^*$ in \cite[Definition 3.3]{Sh:48}).
6225: \ermn
6226: Note that ${\frak K}_1,{\frak K}$ has the same number of models, but
6227: ${\frak K}$ has ``more models" than ${\frak K}^+_1$, in
6228: particular, it has countable members and ${\frak K}_0$ has at least as
6229: many models as ${\frak K}_1$. For $N \in {\frak K}$ to be in
6230: ${\frak K}^+_1 = \{M^+:M
6231: \in \text{ Mod}(\psi_1)\}$ what is missing is the other implications
6232: in $\circledast_7(a)(ii)$. \nl
6233: This is very close to \scite{600-Ex.1}, but ${\frak K}$ may have many
6234: models in $\aleph_1$ (as $\bold Q$ is not necessarily interpreted as
6235: expected). However,
6236: \mr
6237: \item "{$\circledast_{10}$}" constructing $M \in K_{\aleph_1}$ by the
6238: union as $\le_{\frak K}$-increasing continuous chain $\langle M_i:i <
6239: \omega_1 \rangle$, to make sure $M \in {\frak K}^+_1$ it is enough that for
6240: unboundedly many $\alpha < \omega_1,M_\alpha <^{**} M_{\alpha +1}$
6241: where
6242: \sn
6243: \item "{$\circledast_{11}$}" for $M,N \in {\frak K},M <^{**} N$ iff
6244: {\roster
6245: \itemitem{ $(i)$ } $M \le_{\frak K} N$
6246: \sn
6247: \itemitem{ $(ii)$ } in $\circledast_7(b)$ also the inverse direction
6248: holds.
6249: \endroster}
6250: \ermn
6251: It should be clear by now that we have proved clauses (a),(b),(d),(e)
6252: of \scite{600-Ex.1A} using ${\frak K}$.
6253: We have to prove clause (c); we cannot quote
6254: \scite{600-Ex.1} as clause $(\gamma)$ there is only almost true. The proof is
6255: similar to (but simpler than) that of \scite{600-Ex.1} quoting
6256: \cite{Sh:48} instead of \chaptercite{88r}; a marked difference is that in
6257: the present case the number of types over a countable model is
6258: countable (in ${\frak K}$) whereas in \chaptercite{88r} it seemingly could
6259: have been $\aleph_1$, generally \cite{Sh:48} situation is more similar
6260: to the first order logic case.
6261:
6262: Does ${\frak K}$ have amalgamation in $\aleph_0$?
6263: Now (\cite[Lemma 3.4]{Sh:48}, almost says this but it assumed
6264: $\diamondsuit_{\aleph_1}$ instead of $2^{\aleph_0} < 2^{\aleph_1}$;
6265: and \marginbf{!!}{\cprefix{88r}.\scite{88r-3.5}} almost says this, but the models are
6266: from ${\frak K}_{\aleph_1}$ rather than ${\frak K}^+_{\aleph_1}$
6267: but \marginbf{!!}{\cprefix{88r}.\scite{88r-3.8.4}} fully says this using the so called
6268: $K^F_{\aleph_1}$). So
6269: \mr
6270: \item "{$\circledast_{12}$}" ${\frak K}$ has the amalgamation
6271: property in $\aleph_0$.
6272: \ermn
6273: Recall that all models from ${\frak K}$ are atomic (in the
6274: first order sense) and we shall use below $\sftp_{\Bbb L}$.
6275:
6276: As ${\frak K}$ has $\aleph_0$-amalgamation (by $\circledast_{12}$), clearly
6277: \cite[\S4]{Sh:48} applies; now by \cite[Lemma 2.1]{Sh:48}(B)
6278: + Definition 3.5, being $(\aleph_0,1)$-stable as defined in
6279: \cite[Definition 3.5]{Sh:48}(A) holds. Hence all clauses of
6280: \cite[Lemma 4.2]{Sh:48} hold, in particular ($(D)(\beta)$ there and
6281: clause (A), i.e., \cite[Def.3.5]{Sh:48}(B)), so
6282: \mr
6283: \item "{$\circledast_{13}$}" $(i) \quad$ if
6284: $M \le_{\frak K} N$ and $\bar a \in N$ then
6285: $\sftp_{\Bbb L}(\bar a,M,N)$ is definable over a finite subset of $M$
6286: \sn
6287: \item "{${{}}$}" $(ii) \quad$ if $M \in {\frak K}_{\aleph_0}$ then
6288: $\sftp_{\Bbb L}(\bar a,M,N):\bar a \in {}^{\omega >} N$ and
6289: $M \le_{\frak K} N\}$ is countable.
6290: \ermn
6291: By \cite[Lemma 4.4]{Sh:48} it follows that
6292: \mr
6293: \item "{$\circledast_{14}$}" if $M \le_{\frak K} N$ are countable and
6294: $\bar a \in M$ then $\sftp_{\Bbb L}(\bar a,M,N)$ determine
6295: $\sftp_{\frak K}(\bar a,M,N)$.
6296: \ermn
6297: Now we define ${\frak s} = ({\frak K}_{\aleph_0},{\Cal
6298: S}^{\text{bs}},\nonfork{}{}_{})$ by
6299: \mr
6300: \item "{$\circledast_{15}$}" ${\Cal S}^{\text{bs}}(M)
6301: = \{\text{\ortp}_{\frak K}(\bar a,M,N):M \le_{\frak K} N$ are countable
6302: and $\bar a \in {}^{\omega >} N$ but $\bar a \notin {}^{\omega >}M\}$
6303: \sn
6304: \item "{$\circledast_{16}$}" \ortp$_{\frak K}(\bar a,M_1,M_3)$ does not
6305: fork over $M_0$ where $M_0 \le_{\frak K} M_1 \le_{\frak K} M_3 \in
6306: {\frak K}_{\aleph_0}$ iff $\sftp_{\Bbb L}(\bar a,M_1,M_3)$ is definable
6307: over some finite subset of $M_0$.
6308: \ermn
6309: Now we check ``${\frak s}$ is a good frame", i.e., all clauses of
6310: Definition \scite{600-1.1}.
6311: \enddemo
6312: \bn
6313: \ub{Clause (A)}: By $\circledast_9$ above.
6314: \bn
6315: \ub{Clause (B)}: As ${\frak K}$ is categorical in $\aleph_0$, has an
6316: uncountable model and LS$({\frak K}) = \aleph_0$ this should be clear.
6317: \bn
6318: \ub{Clause (C)}: ${\frak K}_{\aleph_0}$ has amalgamation by
6319: $\circledast_{12}$ and has the JEP by categoricity in $\aleph_0$ and
6320: ${\frak K}_{\aleph_0}$ has no maximal model by (categoricity and)
6321: having uncountable models (and LS$({\frak K}) = \aleph_0$).
6322: \bn
6323: \ub{Clause (D)}: Obvious; stability, i.e., (D)(d) holds by
6324: $\circledast_{13}(ii) + \circledast_{14}$.
6325: \bn
6326: \ub{Subclause (E)(a),(b)}: By the definition.
6327: \bn
6328: \ub{Subclause (E)(c)}: (Local character).
6329:
6330: If $\langle M_i:i \le \delta +1 \rangle$ is $\le_{\frak K}$-increasing
6331: continuous $M_i \in K_{\aleph_0},\bar a \in {}^{\omega >}
6332: (M_{\delta +1})$ and $\bar a \in {}^{\omega >}(M_\delta)$
6333: then for some finite $A \subseteq M_\delta,
6334: \sftp_{\Bbb L}(\bar a,M_\delta,M_{\delta +1})$
6335: is definable over $A$, so for some $i <
6336: \delta,A \subseteq M_\delta$ hence $j \in [i,\delta) \Rightarrow
6337: \sftp_{\Bbb L}(\bar a,M_i,M_{\delta +1})$ is definable over $A
6338: \Rightarrow \nonfork{}{}_{}(M_i,M_\delta,\bar a,M_{\delta +i})$.
6339: \bn
6340: \ub{Subclause (E)(d)}: (Transitivity).
6341:
6342: As if $M' \le_{\frak K} M'' \in {\frak K}_{\aleph_0}$, two definitions
6343: in $M'$ of complete types, which give the same result in $M'$ give the
6344: same result in $M''$.
6345: \bn
6346: \ub{Sublause (E)(e)}(uniqueness): By $\circledast_{14}$ and the
6347: justification of transitivity.
6348: \bn
6349: \ub{Subclause (E)(f)}(symmetry): By
6350: \cite[Theorem 5.4]{Sh:48}, we have the symmetry
6351: property see \cite[Definition 5.2]{Sh:48}. By \cite[5.5]{Sh:48} + the
6352: uniqueness proved above we can finish easily.
6353: \bn
6354: \ub{Subclause (E)(g)}: Extension existence.
6355:
6356: Easy, included in \cite[5.5]{Sh:48}.
6357: \bn
6358: \ub{Subclause (E)(h)}: Continuity.
6359:
6360: As ${\Cal S}^{\text{bs}}_{\frak s}(M)$ is the set of
6361: non-algebraic types this follows from ``finite character", that is by
6362: \scite{600-1.16A}(3)(4).
6363: \bn
6364: \ub{Subclause (E)(i)}: non-forking amalgamation
6365:
6366: By \scite{600-1.15}. \hfill$\square_{\scite{600-Ex.1A}}$
6367: \bigskip
6368:
6369: \remark{\stag{600-Ex.1B} Remark} So if $\psi \in \Bbb L_{\omega_1,\omega}
6370: (\bold Q)$ and
6371: $1 \le \dot I(\aleph_1,\psi) < 2^{\aleph_1}$, we essentially can apply
6372: Theorem \scite{600-0.A}, exactly see \scite{600-fc.4}.
6373: \endremark
6374: \bigskip
6375:
6376: \subhead {(D) Starting at $\lambda > \aleph_0$} \endsubhead
6377:
6378: The next theorem puts the results of \cite{Sh:576} in our context
6379: hence rely on it heavily.
6380: \proclaim{\stag{600-Ex.4} Theorem} Assume
6381: $2^\lambda < 2^{\lambda^+} < 2^{\lambda^{++}}$
6382: and
6383: \medskip
6384: \roster
6385: \item "{$(\alpha)$}" ${\frak K}$ is an abstract elementary class with
6386: ${\text{\rm LS\/}}({\frak K}) \le \lambda$
6387: \sn
6388: \item "{$(\beta)$}" ${\frak K}$ is categorical in $\lambda$ and in
6389: $\lambda^+$
6390: \sn
6391: \item "{$(\gamma)$}" ${\frak K}$ has a model in $\lambda^{++}$
6392: \sn
6393: \item "{$(\delta)$}" $\dot I(\lambda^{+2},K) < 2^{\lambda^{++}}$ and
6394: ${\text{\rm WDmId\/}}(\lambda^+)$ is not $\lambda^{++}$-saturated \ub{or}
6395: just some \footnote{alternatively we, in clause (d) assume just $\dot
6396: I(\lambda^{++},K) < \mu_{\text{unif}}(\lambda^{++},2^{\lambda^+})$, see
6397: \cite{Sh:838}, see on this cardinal in \marginbf{!!}{\cprefix{88r}.\scite{88r-0.wD}}(3).}
6398: consequences: density of minimal types (see
6399: \cite{Sh:603}) and $\otimes$ of \cite[6.4,pg.99]{Sh:576} proved by
6400: the conclusion of \cite[Th.6.7]{Sh:576}(pg.101).
6401: %%%tex reference =6.6t holds
6402: (so we can use all the results of \cite[\S8-\S10]{Sh:576})
6403: \ermn
6404: \ub{Then} 1) Letting $\mu = \lambda^+$ we can choose
6405: $\nonfork{}{}_{\mu},{\Cal S}^{\text{bs}}$ such that $({\frak K}_{\ge \mu},
6406: \nonfork{}{}_{\mu},{\Cal S}^{\text{bs}})$ is a $\mu$-good frame.
6407: \nl
6408: 2) Moreover, we can let
6409: \medskip
6410: \roster
6411: \item "{$(a)$}" ${\Cal S}^{\text{bs}}(M) :=
6412: \{ {\text{\rm \ortp\/}}_{\frak K}(a,M,N):
6413: \text{for some } M,N,a \text{ we have }(M,N,a) \in K^3_{\lambda^+}$ \nl
6414:
6415: $\qquad \qquad \qquad \qquad \qquad$ and for some
6416: $M' \le_{\frak K} M \text{ we have } M' \in K_\lambda$ \nl
6417:
6418: $\qquad \qquad \qquad \qquad \qquad$ and
6419: ${\text{\rm \ortp\/}}_{\frak K}(a,M',N) \in {\Cal S}_{\frak K}(M')$ is minimal$\}$
6420: \endroster
6421: \medskip
6422: \noindent
6423: (see Definition \cite[2.3]{Sh:576}(4)(pg.56) and
6424: \cite[2.5]{Sh:576}(1),(13),pg.57-58
6425: \mr
6426: \item "{$(b)$}" $\nonfork{}{}_{} = \nonfork{}{}_{\mu}$ be defined
6427: by: $\nonfork{}{}_{}(M_0,M_1,a,M_3)$ \underbar{iff}
6428: $M_0 \le_{\frak K} M_1 \le_{\frak K} M_3$ are from $K_\mu,a \in M_3
6429: \backslash M_1$ and for some $N \le_{\frak K} M_0$ of cardinality $\lambda$,
6430: the type ${\text{\rm \ortp\/}}_{\frak K}(a,N,M_3) \in
6431: {\Cal S}_{\frak K}(N)$ is minimal.
6432: \endroster
6433: \endproclaim
6434: \bigskip
6435:
6436: \demo{Proof} 1), 2) By clause $(\delta)$ of the assumption, we can
6437: use the ``positive" results of \cite{Sh:576} in particular
6438: \cite[\S8-\S10]{Sh:576} freely.
6439: Also by \cite{Sh:603}, the one point in \cite{Sh:576} in which we have used
6440: $\lambda \ne \aleph_0$ is eliminated from this assumption. Now (see
6441: Definition \scite{600-0.13}(2))
6442: \sn
6443: \mr
6444: \item "{$(*)$}" if $(M,N,a) \in K^3_{\lambda^+}$ and $M' \le_{\frak K} M,
6445: M' \in K_\lambda$ and $p = \text{ \ortp}_{\frak K}(a,M',N)$ is minimal
6446: (see Definition \scite{600-0.12}(0)) \ub{then}
6447: {\roster
6448: \itemitem{ $(a)$ } if $q \in {\Cal S}_{\frak K}(M)$ is not algebraic and
6449: $q \restriction M' = p$ then $q = \text{\rm \ortp}_{\frak K}(a,M,N)$
6450: \sn
6451: \itemitem{ $(b)$ } if $\langle M_\alpha:\alpha < \mu \rangle,\langle
6452: N_\alpha:\alpha < \mu \rangle$ are representations of $M,N$ respectively
6453: \ub{then} for a club of $\delta < \mu$ we have \ortp$_{\frak K}
6454: (a,M_\delta,N_\delta) \in {\Cal S}_{\frak K}
6455: (M_\delta)$ is minimal and reduced
6456: \endroster}
6457: [Why? For clause (b) let $\alpha^* = \text{ Min}\{\alpha:M' \le_{\frak K}
6458: M_\alpha\}$, so $\alpha^*$ is well defined and as $M$ is saturated
6459: (for ${\frak K}$), for a club of $\delta < \mu = \lambda^+$, the model
6460: $M_\delta$ is $(\lambda,\text{cf}(\delta))$-brimmed
6461: over $M'$ hence by \cite[7.5]{Sh:576}(2)(pg.106)
6462: %%%tex reference =7.4At(2)
6463: we are done.
6464:
6465: For clause (a) let $M^0 = M,M^1 = N$ and $a^1=a$ and $M^2,a^2 = a$
6466: be such that $(M^0,M^2,a^2) \in K^3_\mu = K^3_{\lambda^+}$
6467: and $q = \text{ \ortp}_{\frak K}(a^2,M^0,M)$. Now we repeat
6468: the proof of \cite[9.5]{Sh:576}(pg.120)
6469: %%%tex reference =9.4t
6470: but instead $f(a^2) \notin M^1$ we require
6471: $f(a^2) = a^1$; we are using \cite[10.5]{Sh:576}(1)(pg.125)
6472: %%%tex reference =10.4t(1)
6473: which says $<^*_{\lambda^+} = <_{\frak K} \restriction K_{\lambda^+}$.]
6474: \endroster
6475: \bn
6476: In particular we have used
6477: \mr
6478: \item "{$(**)$}" if $M_0 \le_{{\frak K}_\lambda} M_1,M_1$ is
6479: $(\lambda,\kappa)$-brimmed over $M_0,p \in {\Cal S}_{\frak K}(M_1)$ is not
6480: algebraic and $p \restriction M_0$ is minimal, \ub{then} $p$ is
6481: minimal and reduced.
6482: \endroster
6483: \bn
6484: \underbar{Clause $(A)$}:
6485:
6486: This is by assumption $(\alpha)$.
6487: \bigskip
6488: \noindent
6489: \underbar{Clause $(B)$}:
6490:
6491: As $K$ is categorical in $\mu = \lambda^+$, the existence of superlimit
6492: $M \in K_\mu$ follows; the superlimit is not maximal as LS$({\frak K})
6493: \le \lambda \and K_{\mu^+} = K_{\lambda^{++}} \ne \emptyset$ by
6494: assumption $(\gamma)$.
6495: \bigskip
6496: \noindent
6497: \underbar{Clause $(C)$}:
6498:
6499: $K_{\lambda^+}$ has the amalgamation property by \marginbf{!!}{\cprefix{88r}.\scite{88r-3.5}} or
6500: \cite[1.4]{Sh:576}(pg.46),1.6(pg.48) and
6501: %%%tex reference = 1.3t
6502: ${\frak K}_\lambda$ has the JEP in
6503: $\lambda^+$ by categoricity in $\lambda^+$.
6504: \bigskip
6505: \noindent
6506: \underbar{Clause $(D)$}: \newline
6507: \underbar{Subclause $(D)(a),(b)$}:
6508:
6509: By the definition of ${\Cal S}^{\text{bs}}(M)$ and of minimal types (in
6510: ${\Cal S}_{\frak K}(N),N \in K_\lambda$, \nl
6511: \cite[2.5]{Sh:576}(1)+(3)(pg.57),2.3(4)+(6)](pg.56)),
6512: %%%tex reference =2.3Bt
6513: %%%tex reference =2.3At
6514: this is clear.
6515: \bigskip
6516: \noindent
6517: \underbar{Subclause $(D)(c)$}:
6518:
6519: Suppose $M \le_{\frak K} N$ are from $K_\mu$ and $M \ne N$; let
6520: $\langle M_i:i < \lambda^+ \rangle, \langle N_i:i < \lambda^+ \rangle$ be a
6521: $\le_{\frak K}$-representation
6522: of $M,N$ respectively, choose $b \in N \backslash M$
6523: so $E = \{\delta < \lambda^+:N_\delta \cap M = M_\delta \text{ and } b \in
6524: N_\delta\}$ is a club of $\lambda^+$. Now for $\delta = \text{ Min}(E)$ we
6525: have $M_\delta \ne N_\delta,M_\delta \le_{\frak K} N_\delta$ and there
6526: is a minimal inevitable $p \in {\Cal S}_{\frak K}(M_\delta)$ by
6527: \cite[5.3,pg.94]{Sh:576}
6528: %%%tex reference = 5.2t
6529: and categoricity of $K$ in $\lambda$; so for some
6530: $a \in N_\delta \backslash M_\delta$ we have $p = \text{ \ortp}_{\frak K}
6531: (a,M_\delta,N_\delta)$. So \ortp$_{\frak K}(a,M,N)$ is non-algebraic as
6532: $a \in M \Rightarrow a \in M \cap
6533: N_\delta = M_\delta$, a contradiction, so \ortp$_{\frak K}(a,M,N) \in
6534: {\Cal S}^{\text{bs}}(M)$ as required.
6535: \bn
6536: \ub{Subclause $(D)(d)$}: If $M \in K_\mu$ let $\langle M_i:i < \lambda^+
6537: \rangle$ be a $\le_{\frak K}$-representation of $M$, so by $(*)(a)$ above
6538: $p \in {\Cal S}^{\text{bs}}(M)$ is
6539: determined by $p \restriction M_\alpha$ if $p \restriction M_\alpha$ is
6540: minimal and reduced.
6541: But for every such $p$ there is such $\alpha(p) < \lambda^+$ by the
6542: definition of ${\Cal S}^{\text{bs}}(M)$ and for each
6543: $\alpha < \lambda^+$ there are $\le \lambda$ possible such $p \restriction
6544: M_\alpha$ as ${\frak K}$ is stable in $\lambda$ by
6545: \cite[5.7]{Sh:576}(a)(pg.97), so the
6546: %%%tex reference =5.5t
6547: conclusion follows. Alternatively,
6548: $M \in K_\mu \Rightarrow |{\Cal S}^{\text{bs}}(M)| \le
6549: \mu$ as by \cite[10.5]{Sh:576}(pg.125), we have $\le^*_{\lambda^+} =
6550: %%%tex reference =10.4t
6551: \le_{\frak K} \restriction K_{\lambda^+}$, so we can apply
6552: \cite[9.7]{Sh:576}(pg.121); or use $(*)$ above.
6553: %%%tex reference = 9.6t
6554: \bn
6555: \underbar{Clause $(E)$}: \newline
6556: \underbar{Subclause $(E)(a)$}:
6557:
6558: Follows by the definition.
6559: \bn
6560: \underbar{Subclause $(E)(b)$}: (Monotonicity)
6561:
6562: Obvious properties of minimal types in ${\Cal S}(M)$ for $M \in K_\lambda$.
6563: \bn
6564: \underbar{Subclause $(E)(c)$}: (Local character)
6565:
6566: Let $\delta < \mu^+ =
6567: \lambda^{++}$ and $M_i \in K_\mu$ be $\le_{\frak K}$-increasing continuous
6568: for $i \le \delta$ and $p \in {\Cal S}^{\text{bs}}
6569: (M_\delta)$, so for some $N \le_{\frak K} M_\delta$ we have $N \in
6570: K_\lambda$ and $p \restriction N \in {\Cal S}_{\frak K}(N)$ is
6571: minimal. Without loss of generality
6572: $\delta = \text{ cf}(\delta)$ and if $\delta = \lambda^+$, there is
6573: $i < \delta$ such
6574: that $N \subseteq M_i$ and easily we are done.
6575: So assume $\delta = \text{ cf}
6576: (\delta) < \lambda^+$. \nl
6577: Let $\langle M^i_\zeta:\zeta < \lambda^+ \rangle$ be a $\le_{\frak K}$-representation of
6578: $M_i$ for $i \le \delta$, hence $E$ is a club of $\lambda^+$ where:
6579:
6580: $$
6581: \align
6582: E := \bigl\{ \zeta < \lambda^+:&\zeta \text{ a limit ordinal and for }
6583: j < i \le \delta \text{ we have} \\
6584: &M^i_\zeta \cap M_j = M^j_\zeta \text{ and for }
6585: \xi < \zeta,i \le \delta \text{ we have}: \\
6586: &M^i_\zeta \text{ is } (\lambda,\text{cf}(\zeta))
6587: \text{-brimmed over } M^i_\xi \text{ and } N \le_{\frak K}
6588: M^\delta_\zeta \bigr\}.
6589: \endalign
6590: $$
6591: \mn
6592: Let $\zeta_i$ be the $i$-th member of $E$ for $i \le \delta$, so
6593: $\langle \zeta_i:i \le \delta \rangle$ is increasing continuous,
6594: $\langle M^i_{\zeta_i}:i \le \delta \rangle$ is $\le_{\frak K}$-increasingly
6595: continuous in $K_\lambda$ and $M^{i+1}_{\zeta_{i+1}}$ is $(\lambda,
6596: \text{cf}(\zeta_{i+1}))$-brimmed over $M^{i+1}_{\zeta_i}$ hence also
6597: over $M^i_{\zeta_i}$. Also $p \restriction M^\delta_{\zeta_\delta}$
6598: is non-algebraic (as $p$ is) and extends
6599: $p \restriction N$ (as $N \le_{\frak K}
6600: M^\delta_{\zeta_\delta}$ as $\zeta_\delta
6601: \in E$) hence $p \restriction M^\delta_{\zeta_\delta}$ is minimal.
6602:
6603: Also $M^\delta_{\zeta_\delta}$ is
6604: $(\lambda,\text{cf}(\zeta_\delta))$-brimmed
6605: over $M^\delta_{\zeta_0}$ hence over $N$, hence by $(**)$ above we get
6606: that $p \restriction M^\delta_{\zeta_\delta}$ is not only minimal but also
6607: reduced. Hence by \cite[7.3]{Sh:576}(2)(pg.103) applied to $\langle
6608: %%%tex reference =7.2t
6609: M^i_{\zeta_i}:i \le \delta \rangle,p \restriction M^\delta_{\zeta_\delta}$
6610: we know that for some $i < \delta$ the type $p \restriction M^i_{\zeta_i} =
6611: (p \restriction M^\delta_{\zeta_\delta}) \restriction M^i_{\zeta_i}$ is
6612: minimal and reduced, so it witnesses that
6613: $p \restriction M_j \in {\Cal S}^{\text{bs}}(M_j)$ for every
6614: $j \in [i,\delta)$, as required.
6615: \bigskip
6616:
6617: \noindent
6618: \underbar{Subclause $(E)(d)$}: (Transitivity)
6619:
6620: Easy by the definition of minimal.
6621: \bigskip
6622:
6623: \noindent
6624: \underbar{Subclause $(E)(e)$}: (Uniqueness)
6625:
6626: By $(*)(a)$ above.
6627: \bn
6628: \underbar{Subclause $(E)(f)$}: (Symmetry)
6629:
6630: By the symmetry in the situation assume $M_0 \le_{\frak K} M_1 \le_{\frak K}
6631: M_3$ are from $K_\mu$, \newline
6632: $a_1 \in M_1 \backslash M_0,a_2 \in M_3 \backslash
6633: M_1$ and \ortp$_{\frak K}(a_1,M_0,M_3) \in {\Cal S}^{\text{bs}}(M_0)$
6634: and \ortp$_{\frak K}(a_2,M_1,M_3)
6635: \in {\Cal S}^{\text{bs}}(M_1)$ does not fork over $M_0$; hence for $\ell =
6636: 1,2$ we have \ortp$_{\frak K}(a_\ell,M_0,M_3) \in {\Cal S}^{\text{bs}}(M_0)$.
6637: By the existence of disjoint
6638: amalgamation (by \cite[9.11]{Sh:576}(pg.122),10.5(1)(pg.125))
6639: %%%tex reference =9.10t
6640: %%%tex reference =10.3t
6641: there are $M_2,M'_3,f$ such that
6642: $M_0 \le_{\frak K} M_2 \le_{\frak K} M'_3 \in K_\mu$,
6643: $M_3 \le_{\frak K} M'_3,f$ is an isomorphism from $M_3$ onto $M_2$
6644: over $M_0$, and $M_3 \cap M_2 = M_0$. By \ortp$_{\frak K}(a_2,M_0,M_3) \in {\Cal
6645: S}^{\text{bs}}(M_1)$ and as $f(a_2) \notin M_1$ being in $M_2
6646: \backslash M_0 = M_2 \backslash M_3$ and $a_2 \notin M_1$ by
6647: assumption and as $a_2,f(a_2)$ realize the same type from ${\Cal
6648: S}_{\frak K}(M_0)$ clearly by $(*)(a)$
6649: we have \ortp$_{\frak K}(a_2,M_1,M'_3) = \text{ \ortp}_{\frak K}(f(a_2),M_1,M'_3)$.
6650: \medskip
6651:
6652: Using amalgamation in ${\frak K}_\mu$ (and equality of types)
6653: there is $M''_3$ such that: \newline
6654: $M'_3 \le_{\frak K} M''_3 \in K_\mu$, and there is an
6655: $\le_{\frak K}$-embedding $g$ of $M'_3$ into $M''_3$ such that
6656: $g \restriction M_1 = \text{ id}_{M_1}$ and $g(f(a_2)) = a_2$.
6657: Note that as $a_1 \notin g(M_2),M_1 \le_{\frak K} g(M_2) \in K_\mu$
6658: and \ortp$_{\frak K}(a_1,M_1,M''_3)$ is minimal then necessarily
6659: \ortp$_{\frak K}(a_1,g(M_2),M''_3)$ is its non-forking extension.
6660: So $g(M_2),M''_3$ are models as required.
6661: \bn
6662: \underbar{Subclause $(E)(g)$}: (Extension existence)
6663:
6664: Claims \cite[9.11]{Sh:576}(pg.122),10.5(1)(pg.125) do even more.
6665: %%%tex reference =9.10t
6666: %%%tex reference =10.3t
6667: \bn
6668: \underbar{Subclause $(E)(h)$}: (Continuity)
6669:
6670: Easy.
6671: \bn
6672: \ub{Subclause $(E)(i)$}: (Non-forking amalgamation)
6673:
6674: Like $(E)(f)$ or use \scite{600-1.15}. \hfill$\square_{\scite{600-Ex.4}}$
6675: \enddemo
6676: \bigskip
6677: \noindent
6678: \margintag{600-Ex.5}\underbar{\stag{600-Ex.5} Question}: If ${\frak K}$ is categorical in $\lambda$
6679: and in $\mu$ and $\mu > \lambda \ge \text{LS}({\frak K})$, can we
6680: conclude categoricity in $\chi \in (\mu,\lambda)$?
6681: \bigskip
6682:
6683: \demo{\stag{600-Ex.6} Fact} In \scite{600-Ex.4}: \nl
6684: 1) If $p \in {\Cal S}^{\text{bs}}(M)$
6685: and $M \in K_\mu$, \ub{then} for some $N \le_{\frak K} M,N \in
6686: K_\lambda$ and $p \restriction N$ is minimal and reduced. \nl
6687: 2) If $M <_{\frak K} N,M \in K_\mu$ and $p \in {\Cal
6688: S}^{\text{bs}}(M)$, \ub{then} some $a \in N \backslash M$ realizes $p$,
6689: (i.e., ``a strong version of uni-dimensionality" holds).
6690: \enddemo
6691: \bigskip
6692:
6693: \demo{Proof} The proof is included in the proof of \scite{600-Ex.4}.
6694: \enddemo
6695: \bn
6696: \centerline{$* \qquad * \qquad *$}
6697: \bn
6698: \ub{(E) An Example}:
6699:
6700: A trivial example (of an approximation to good $\lambda$-frame) is:
6701: \definition{\stag{600-Ex.7.1} Definition/Claim} 1) Assume that ${\frak K}$
6702: is an a.e.c. and $\lambda \ge \text{\rm LS}({\frak K})$
6703: or ${\frak K}$ is a $\lambda$-a.e.c.
6704: We define ${\frak s} = {\frak s}_\lambda[{\frak K}]$ as the triple
6705: ${\frak s} = ({\frak K}_\lambda,{\Cal
6706: S}^{\text{na}},\nonfork{}{}_{\text{na}})$ where:
6707: \mr
6708: \item "{$(a)$}" ${\Cal S}^{\text{na}}(M) = \{\text{ \ortp}_{\frak
6709: K}(a,M,N),M \le_{\frak K} N$ and $a \in N \backslash M\}$
6710: \sn
6711: \item "{$(b)$}" $\nonfork{}{}_{\text{na}}(M_0,M_1,a,M_3)$ iff $M_0
6712: \le_{{\frak K}_\lambda} M_1 \le_{{\frak K}_\lambda} M_3$ and $a \in
6713: M_3 \backslash M_1$.
6714: \ermn
6715: 2) Then ${\frak s}$ satisfies Definition \scite{600-1.1} of good
6716: $\lambda$-frame except possibly: (B), existence of superlimits, (C)
6717: amalgamation and JEP, (D)(d) stability and (E)(e),(f),(g),(i) uniqueness,
6718: symmetry, extension existence and non-forking amalgamation.
6719: \enddefinition
6720: \newpage
6721:
6722: \head {\S4 Inside the framework} \endhead \resetall \sectno=4
6723: \spuriousreset
6724: \bigskip
6725:
6726: We investigate good $\lambda$-frames. We prove stability in $\lambda$
6727: (we have assumed in Definition \scite{600-1.1} only stability for basic
6728: types), hence the existence of a $(\lambda,\sigma)$-brimmed
6729: $\le_{\frak K}$-extension in $K_\lambda$ over $M_0 \in K_\lambda$ (see
6730: \scite{600-4a.1}), and we give a sufficient condition for ``$M_\delta$ is
6731: $(\lambda,\text{cf}(\delta))$-brimmed over $M_0$" (in \scite{600-4a.2}).
6732: We define again $K^{3,\text{bs}}_\lambda$ (like $K^3_\lambda$ from
6733: \scite{600-0.13}(2) but the type is
6734: basic) and the natural order $\le_{\text{bs}}$ on them as well as ``reduced"
6735: (Definition \scite{600-4a.3}),
6736: and indicate their basic properties (\scite{600-4a.4}). \nl
6737: We may like to construct sometimes pairs $N_i \le_{{\frak K}_\lambda}
6738: M_i$ such that
6739: $M_i,N_i$ are increasing continuous with $i$ and we would like to guarantee
6740: that $M_\gamma$ is $(\lambda,\text{cf}(\gamma))$-brimmed over
6741: $N_\gamma$, of course we need to carry more inductive assumptions.
6742: Toward this we may give a sufficient condition for building a
6743: $(\lambda,\text{cf}(\gamma))$-brimmed extension over $N_\gamma$
6744: where $\langle N_i:i \le \gamma \rangle$ is
6745: $\le_{{\frak K}_\lambda}$-increasing continuous,
6746: by a triangle of extensions of the $N_i$'s, with
6747: non-forking demands of course (see \scite{600-4a.5}). We
6748: also give conditions on a rectangle of models to get such pairs in
6749: both directions (\scite{600-4a.9}), for this we use nice extensions of
6750: chains (\scite{600-4a.7}, \scite{600-4a.8}).
6751:
6752: Then we can deduce that if ``$M_1$ is $(\lambda,\sigma)$-brimmed
6753: over $M_0$" then the isomorphism type of $M_1$ over $M_0$ does not
6754: depend on $\sigma$ (see \scite{600-4a.6}), so the brimmed $N$ over $M_0$
6755: is unique up to isomorphism (i.e. being $(\lambda,\sigma)$-brimmed
6756: over $M_0$ does not depend on $\sigma$). We finish giving conclusion
6757: about $K_{\lambda^+},K_{\lambda^{++}}$.
6758: \bigskip
6759:
6760: \demo{\stag{600-4a.0} Hypothesis} ${\frak s} = ({\frak K},
6761: \nonfork{}{}_{},{\Cal S}^{\text{bs}})$ is a good $\lambda$-frame.
6762: \enddemo
6763: \bigskip
6764:
6765: \proclaim{\stag{600-4a.1} Claim} 1) ${\frak K}$ is stable in $\lambda$,
6766: i.e., $M \in {\frak K}_\lambda \Rightarrow |{\Cal S}(M)| \le \lambda$.
6767: \nl
6768: 2) For every $M_0 \in K_\lambda$ and $\sigma \le \lambda$
6769: there is $M_1$ such that $M_0 \le_{\frak K} M_1 \in K_\lambda$ and $M_1$
6770: is $(\lambda,\sigma$)-brimmed over $M_0$ (see Definition \scite{600-0.21})
6771: and it is universal
6772: \footnote{in fact, this follows} over $M_0$.
6773: \endproclaim
6774: \bigskip
6775:
6776: \demo{Proof} 1) Let
6777: $M_0 \in K_\lambda$ and we choose by induction on
6778: $\alpha \in [1,\lambda],M_\alpha \in K_\lambda$ such that:
6779: \mr
6780: \item "{$(i)$}" $M_\alpha$ is $\le_{\frak K}$-increasing continuous
6781: \sn
6782: \item "{$(ii)$}" if $p \in {\Cal S}^{\text{bs}}(M_\alpha)$ then
6783: this type is realized in $M_{\alpha +1}$.
6784: \ermn
6785: No problem to carry this: for clause (i) use $Ax(A)$, for clause (ii)
6786: use Axiom $(D)(d)$ and amalgamation in ${\frak K}_\lambda$, i.e., Axiom (C).
6787: If every $q \in {\Cal S}(M_0)$ is realized in
6788: $M_\lambda$ we are done. So let $q$ be a counterexample, so let
6789: $M_0 \le_{\frak K} N \in K_\lambda$ be such that $q$ is
6790: realized in $N$. We now try to choose by induction on $\alpha < \lambda$
6791: a triple $(N_\alpha,f_\alpha,\bar{\bold a}_\alpha)$ such that:
6792: \mr
6793: \item "{$(A)$}" $N_\alpha \in K_\lambda$ is $\le_{\frak K}$-increasingly
6794: continuous
6795: \sn
6796: \item "{$(B)$}" $f_\alpha$ is a $\le_{\frak K}$-embedding of
6797: $M_\alpha$ into $N_\alpha$
6798: \sn
6799: \item "{$(C)$}" $f_\alpha$ is increasing continuous
6800: \sn
6801: \item "{$(D)$}" $f_0 = \text{ id}_{M_0}$ and $N_0 = N$
6802: \sn
6803: \item "{$(E)$}" $\bar{\bold a}_\alpha = \langle a_{\alpha,i}:i < \lambda
6804: \rangle$ lists the elements of $N_\alpha$
6805: \sn
6806: \item "{$(F)$}" if there are $\beta \le \alpha,i < \lambda$ such that
6807: \ortp$(a_{\beta,i},f_\alpha(M_\alpha),N_\alpha) \in {\Cal S}^{\text{bs}}
6808: (f_\alpha(M_\alpha))$ \ub{then} for some such pair $(\beta_\alpha,i_\alpha)$
6809: we have:
6810: {\roster
6811: \itemitem{ $(i)$ } the pair $(\beta_\alpha,i_\alpha)$ is minimal in
6812: an appropriate sense, that is:
6813: if $(\beta,i)$ is another such pair then $\beta + i >
6814: \beta_\alpha + i_\alpha$ or $\beta +i = \beta_\alpha + i_\alpha \and \beta
6815: > \beta_\alpha$ or $\beta+i = \beta_\alpha + i_\alpha \and \beta =
6816: \beta_\alpha \and i \ge i_\alpha$
6817: \sn
6818: \itemitem{ $(ii)$ } $a_{\beta_\alpha,i_\alpha}
6819: \in \text{ Rang}(f_{\alpha+1})$.
6820: \endroster}
6821: \ermn
6822: This is easy: for successor $\alpha$ we use the definition of type and
6823: let $N_\lambda := \cup\{N_\alpha:\alpha < \lambda\}$.
6824: Clearly $f_\lambda := \cup\{f_\alpha:\alpha < \lambda\}$ is a
6825: $\le_{\frak s}$-embedding of $M_\lambda$ into $N_\lambda$ over $M_0$.
6826:
6827: As in $N$, the type $q$ is realized and it is not realized in
6828: $M_\lambda$ necessarily $N \nsubseteq f_\lambda(M_\lambda)$ hence
6829: $N_\lambda \ne f_\lambda(M_\lambda)$ but easily $f_\lambda(M_\lambda)
6830: \le_{\frak K} N_\lambda$. So by
6831: Axiom $(D)(c)$ for some $c \in N_\lambda \backslash f_\lambda
6832: (M_\lambda)$ we have $p = \text{ \ortp}(c,f_\lambda(M_\lambda),N_\lambda)
6833: \in {\Cal S}^{\text{bs}}(f_\lambda(M_\lambda))$. As $\langle
6834: f_\gamma(M_\gamma):\gamma \le \lambda \rangle$ is $\le_{\frak
6835: K}$-increasing continuous, by
6836: Axiom (E)(c) for some $\gamma < \lambda$ we have
6837: \ortp$(c,f_\lambda(M_\lambda),N_\lambda)$ does not fork over
6838: $f_\gamma(M_\gamma)$, also as $c \in N_\lambda =
6839: \dbcu_{\beta < \lambda} N_\beta$ clearly $c \in N_\beta$ for some
6840: $\beta < \lambda$ and let $i < \lambda$ be such that
6841: $c = a_{\beta,i}$. Now if $\alpha \in
6842: [\text{max}\{\gamma,\beta\},\lambda)$ then $(\beta,i)$ is a legitimate
6843: candidate for $(\beta_\alpha,i_\alpha)$ that is
6844: \ortp$(a_{\beta,i},f_\alpha(M_\alpha),N_\alpha) \in {\Cal
6845: S}^{\text{bs}}(f_\alpha(M_\alpha))$ by monotonicity of non-forking,
6846: i.e., Axiom (E)(b). So $(\beta_\alpha,i_\alpha)$ is well defined for
6847: any such $\alpha$ and
6848: $\beta_\alpha + i_\alpha \le \beta+i$ by clause (F)(i). But $\alpha_1 <
6849: \alpha_2 \Rightarrow a_{\beta_{\alpha_1},i_{\alpha_1}} \ne
6850: a_{\beta_{\alpha_2},i_{\alpha_2}}$ (as one belongs to $f_{\alpha_1 +1}
6851: (M_{\alpha_1})$
6852: and the other not), contradiction by cardinality consideration. \nl
6853: 2) So ${\frak K}_\lambda$ is stable in $\lambda$ and has amalgamation, hence
6854: (see \scite{600-0.22}) the conclusion holds; alternatively use
6855: \scite{600-4a.2} below.
6856: \hfill$\square_{\scite{600-4a.1}}$
6857: \enddemo
6858: \bigskip
6859:
6860: \proclaim{\stag{600-4a.2} Claim} Assume
6861: \mr
6862: \item "{$(a)$}" $\delta < \lambda^+$ is a
6863: limit ordinal divisible by $\lambda$
6864: \sn
6865: \item "{$(b)$}" $\bar M = \langle M_\alpha:\alpha \le \delta \rangle$ is
6866: $\le_{\frak K}$-increasing continuous sequence in ${\frak K}_\lambda$
6867: \sn
6868: \item "{$(c)$}" if $i < \delta$ and
6869: $p \in {\Cal S}^{\text{bs}}(M_i)$, \ub{then}
6870: for $\lambda$ ordinals $j \in (i,\delta)$ there is $c' \in M_{j+1}$
6871: realizing the non-forking extension of $p$ in ${\Cal S}^{\text{bs}}(M_j)$.
6872: \ermn
6873: \ub{Then} $M_\delta$ is $(\lambda,{\text{\rm cf\/}}
6874: (\delta))$-brimmed over $M_0$ and universal over it.
6875: \endproclaim
6876: \bigskip
6877:
6878: \remark{\stag{600-4a.2A} Remark} 1) See end of proof of \scite{600-nf.17}. \nl
6879: 2) Of course, by renaming, $M_\delta$ is $(\lambda,\text{cf}
6880: (\delta))$-brimmed over $M_\alpha$ for any $\alpha < \delta$. \nl
6881: 3) Why in clause (c) of \scite{600-4a.2} we ask for ``$\lambda$ ordinals $j \in
6882: (i,\delta)$" rather than ``for unboundedly many $j \in (i,\delta)$"?
6883: For $\lambda$ regular there is no difference but for $\lambda$
6884: singular not so.
6885: Think of ${\frak K}$ the class of $(A,R),R$ an equivalence relation on
6886: $A$; (so it is not categorical) but for some $\lambda$-good frames
6887: ${\frak s},{\frak K}_{\frak s} = {\frak K}_\lambda$ and
6888: exemplifies a problem; some equivalence class of $M_\delta$ may be of
6889: cardinality $< \lambda$.
6890: \endremark
6891: \bigskip
6892:
6893: \demo{Proof} Like \scite{600-4a.1}, but we give details.
6894:
6895: Let $g:\delta \rightarrow \lambda$ be a one to
6896: one and choose by induction on $\alpha \le \delta$ a triple $(N_\alpha,
6897: f_\alpha,\bar{\bold a}_\alpha)$ such that
6898: \mr
6899: \item "{$(A)$}" $N_\alpha \in K_\lambda$ is $\le_{\frak K}$-increasing
6900: continuous
6901: \sn
6902: \item "{$(B)$}" $f_\alpha$ is a $\le_{\frak K}$-embedding of
6903: $M_\alpha$ into $N_\alpha$
6904: \sn
6905: \item "{$(C)$}" $f_\alpha$ is increasing continuous
6906: \sn
6907: \item "{$(D)$}" $f_0 = \text{ id}_{M_0},N_0 = M_0$
6908: \sn
6909: \item "{$(E)$}" $\bar{\bold a}_\alpha = \langle a_{\alpha,i}:i < \lambda
6910: \rangle$ list the elements of $N_\alpha$
6911: \sn
6912: \item "{$(F)$}" $N_{\alpha +1}$ is universal over $N_\alpha$
6913: \sn
6914: \item "{$(G)$}" if $\alpha < \delta$ and
6915: there is a pair $(\beta,i) = (\beta_\alpha,i_\alpha)$
6916: satisfying the condition $(*)^{\beta,i}_{f_\alpha,N_\alpha}$ stated below
6917: and it is minimal in the sense that \nl
6918: $(*)^{\beta',i'}_{f_\alpha,N_\alpha} \Rightarrow
6919: (**)^{\beta',i',\beta,i}_g$, see below,
6920: \ub{then} $a_{\beta,i} \in \text{ Rang}(f_{\alpha +1})$, \nl
6921: where
6922: {\roster
6923: \itemitem{ $(*)^{\beta,i}_{f_\alpha,N_\alpha}$ } $(a) \quad \beta \le
6924: \alpha$ and $i < \lambda$
6925: \sn
6926: \itemitem{ {} } $(b) \quad$ \ortp$(a_{\beta,i},f_\alpha(M_\alpha),N_\alpha)
6927: \in {\Cal S}^{\text{bs}}(f_\alpha(M_\alpha))$
6928: \sn
6929: \itemitem{ {} } $(c) \quad$ some $c \in M_{\alpha +1}$ realizes
6930: $f^{-1}_\alpha(\text{\ortp}(a_{\beta,i},f_\alpha(M_\alpha),N_\alpha)$, so
6931: by
6932: \nl
6933:
6934: \hskip25pt clause (b) it follows that
6935: $c \in M_{\alpha +1} \backslash M_\alpha$
6936: \sn
6937: \itemitem{ $(**)^{\beta',i',\beta,i}_g$ }
6938: $\quad \,\,\,\,[g(\beta) + i < g(\beta')+i'] \vee$ \nl
6939:
6940: $\qquad \quad [g(\beta) +i = g(\beta')+i'
6941: \and g(\beta) < g(\beta')] \vee [g(\beta)+i = g(\beta')+ i' \and$ \nl
6942:
6943: $\qquad \quad g(\beta) = g(\beta') \and i \le i']$.
6944: \endroster}
6945: \ermn
6946: There is no problem to choose $f_\alpha,N_\alpha$. Now in the end, by
6947: clauses (A),(F) clearly
6948: $N_\delta$ is $(\lambda,\text{cf}(\delta))$-brimmed over $N_0$, i.e.,
6949: over $M_0$, so it suffices to prove that $f_\delta$ is onto $N_\delta$.
6950: If not, then by Axiom (D)(c), the density,
6951: there is $d \in N_\delta \backslash f_\delta(M_\delta)$ such that
6952: $p := \text{\ortp}(d,f_\delta(M_\delta),N_\delta) \in {\Cal
6953: S}^{\text{bs}}(f_\delta(M_\delta))$ hence
6954: for some $\beta(*) < \delta$ we have $d \in N_{\beta(*)}$ so for some
6955: $i(*) < \lambda,d = a_{\beta(*),i(*)}$. Also by Axiom (E)(c), (the
6956: local character) for every $\beta < \delta$ large
6957: enough say $\ge \beta_d$ the type $p$ does not fork over $f_\delta(M_\beta)$,
6958: \wilog \, $\beta_d = \beta(*)$. Let $q = f^{-1}_\delta(\text{\rm
6959: \ortp}(d,f_\delta(M_\delta),N_\delta)$, so it $\in {\Cal
6960: S}^{\text{bs}}(M_\delta)$.
6961:
6962: Let $u = \{\alpha:\beta(*) \le \alpha < \delta$ and $q \restriction
6963: M_\alpha \in {\Cal S}^{\text{bs}}(M_\alpha)$ (note $\beta(*) \le
6964: \alpha$) is realized in $M_{\alpha +1}\}$. By clause (c) of the
6965: assumption clearly $|u| = \lambda$.
6966: Also by the definition of $v$ for every $\alpha \in u$ the
6967: condition $(*)^{\beta(*),i(*)}_{N_\alpha,f_\alpha}$ holds, hence in
6968: clause (F) the pair
6969: $(\beta_\alpha,i_\alpha)$ is well defined and is
6970: ``below" $(\beta(*),i(*))$ in the sense of clause (G).
6971: But there are only $\le |g(\beta(*)) \times i(*)| < \lambda$ such pairs hence
6972: for some $\alpha_1 < \alpha_2$ in $u$ we have $(\beta_{\alpha_1},
6973: i_{\alpha_1}) = (\beta_{\alpha_2},i_{\alpha_2})$, a contradiction:
6974: $a_{\beta_{\alpha_1},i_{\alpha_1}} \in \text{ Rang}(f_{\alpha_1+1}) \subseteq
6975: \text{ Rang}(f_{\alpha_2}) = f_{\alpha_2}(M_{\alpha_2})$
6976: hence \ortp$(a_{\beta_{\alpha_1},i_{\alpha_1}},
6977: f_{\alpha_2}(M_{\alpha_2}),N_{\alpha_2}) \notin {\Cal S}^{\text{bs}}
6978: (f_{\alpha_2}(M_{\alpha_2}))$, contradiction. So we are done.
6979: \hfill$\square_{\scite{600-4a.2}}$
6980: \enddemo
6981: \bn
6982: \centerline {$* \qquad * \qquad *$}
6983: \bn
6984: The following is helpful for constructions so that we can amalgamate
6985: disjointly preserving non-forking of a type; we first repeat the definition
6986: of $K^{3,\text{bs}}_\lambda,<_{\text{bs}}$.
6987: \definition{\stag{600-4a.3} Definition} 1) Recall
6988: $(M,N,a) \in K^{3,\text{bs}}_\lambda$
6989: if $M \le_{\frak K} N$ are models from $K_\lambda,a \in N \backslash M$
6990: and \ortp$(a,M,N) \in {\Cal S}^{\text{bs}}(M)$.
6991: Let $(M_1,N_1,a) \le_{\text{bs}} (M_2,N_2,a)$ or write
6992: $\le^{\frak s}_{\text{bs}}$, when: both triples are in
6993: $K^{3,\text{bs}}_\lambda,M_1
6994: \le_{\frak K} M_2,N_1 \le_{\frak K} N_2$ and \ortp$(a,M_2,N_2)$ does not fork
6995: over $M_1$. \nl
6996: 2) We say $(M,N,a)$ is bs-reduced \ub{when} if it belongs to
6997: $K^{3,\text{bs}}_\lambda$ and $(M,N,a) \le_{\text{bs}}
6998: (M',N',a) \in K^{3,\text{bs}}_\lambda \Rightarrow
6999: N \cap M' = M$. \nl
7000: 3) We say $p \in {\Cal S}^{\text{bs}}(N)$ is a (really the) stationarization
7001: of $q \in {\Cal S}^{\text{bs}}(M)$ if $M \le_{\frak K} N$ and $p$ is an
7002: extension of $q$ which does not fork over $M$.
7003: \enddefinition
7004: \bigskip
7005:
7006: \remark{Remark} 1) The definition of $K^{3,\text{bs}}_\lambda$ is
7007: compatible with the one in \scite{600-1.6} by \scite{600-1.8}(1).
7008: \nl
7009: 2) We could have strengthened the definition of bs-reduced
7010: (\scite{600-4a.3}), e.g., add: for no $b \in N' \backslash M'$, do we
7011: have \ortp$(b,M',N') \in {\Cal S}^{\text{bs}}(M')$ and there are
7012: $M'',N''$ such that $(M',N',a) \le_{\text{bs}} (M'',N'',a)$ and
7013: \ortp$(b,M'',N'')$ forks over $M'$.
7014: \endremark
7015: \bigskip
7016:
7017: \proclaim{\stag{600-4a.4} Claim} For parts (3),(4),(5) assume ${\frak s}$
7018: is categorical (in $\lambda$).
7019: \nl
7020: 1) If $\kappa \le \lambda,(M,N,a) \in
7021: K^{3,\text{bs}}_\lambda$, \ub{then}
7022: we can find $M',N'$ such that: $(M,N,a) \le_{\text{bs}}
7023: (M',N',a) \in K^{3,\text{bs}}_\lambda,M'$ is
7024: $(\lambda,\kappa)$-brimmed over
7025: $M,N'$ is $(\lambda,\kappa$)-brimmed over $N$
7026: and $(M',N',a)$ is {\rm bs}-reduced. \nl
7027: 1A) If $(M,N_\ell,a_\ell) \in K^{3,\text{bs}}_\lambda$ for
7028: $\ell = 1,2$, \ub{then}
7029: we can find $M^+,f_1,f_2$ such that: $M \le_{\frak K} M^+ \in
7030: K_\lambda$ and for $\ell \in \{1,2\},f_\ell$ is a $\le_{\frak
7031: K}$-embedding of $N_\ell$ into $M^+$ over $M$ and
7032: $(M,f_\ell(N_\ell),f_\ell(a_\ell)) \le_{\text{bs}}
7033: (f_{3-\ell}(N_{3-\ell}),M^+,f_\ell(a_\ell))$, equivalently
7034: ${\text{\rm \ortp\/}}(f_\ell(a_\ell),f_{3-\ell}(N_{3-\ell}),M^+)$ does
7035: not fork over $M$.
7036: \nl
7037: 2) If $(M_\alpha,N_\alpha,a) \in K^{3,\text{bs}}_\lambda$ is
7038: $\le_{\text{bs}}$-increasing for $\alpha < \delta$ and
7039: $\delta < \lambda^+$ is a limit ordinal \ub{then} their
7040: union $(\dbcu_{\alpha < \delta} M_\alpha,\dbcu_{\alpha < \delta}
7041: N_\alpha,a)$ is a $\le_{\text{bs}}$-lub. If each $(M_\alpha,N_\alpha,a)$
7042: is bs-reduced then so is their union. \nl
7043: 3) Let $\lambda$ divide $\delta,\delta < \lambda^+$. We can find $\langle
7044: N_j,a_i:j \le \delta,i < \delta \rangle$ such that: $N_j \in
7045: K_\lambda$ is $\le_{\frak K}$-increasing continuous,
7046: $(N_j,N_{j+1},a_j) \in K^{3,\text{bs}}_\lambda$ is bs-reduced and if $i <
7047: \delta,p \in {\Cal S}^{\text{bs}}(N_i)$ \ub{then} for
7048: $\lambda$ ordinals $j \in (i,i + \lambda)$ the type
7049: ${\text{\rm \ortp\/}}(a_j,N_j,N_{j+1})$ is a non-forking extension of
7050: $p$; so $N_\delta$ is $(\lambda,{\text{\rm cf\/}}(\delta))$-brimmed over
7051: each $N_i,i < \delta$. We can add ``$N_0$ is brimmed".
7052: \nl
7053: 4) For any $(M_0,M_1,a) \in K^{3,\text{bs}}_\lambda$ and
7054: $M_2 \in K_\lambda$ such that $M_0
7055: \le_{\frak K} M_2$ there are $N_0,N_1$ such that $(M_0,M_1,a)
7056: \le_{\text{bs}}
7057: (N_0,N_1,a),M_0 = M_1 \cap N_0$ and $M_2,N_0$ are isomorphic over
7058: $M_0$. (In fact, if $(M_0,M_2,b) \in K^{3,\text{bs}}_\lambda$ we can add that
7059: for some isomorphism $f$ from $M_2$ onto $N_0$ over $M_0$ we have
7060: $(M_0,N_0,f(a)) \le_{\text{bs}} (M_1,N_1,f(a))$.)
7061: \nl
7062: 5) If $M_0 \in K_\lambda$ is brimmed and $M_0 \le_{\frak s} M_\ell$
7063: for $\ell=1,2$ and there is a disjoint $\le_{\frak s}$-amalgamation of
7064: $M_1,M_2$ over $M_0$.
7065: \endproclaim
7066: \bigskip
7067:
7068: \demo{Proof} 1) We choose $M_i,N_i,b^\ell_i(\ell=1,2),
7069: \bar{\bold c}_i$ by induction on $i < \delta := \lambda$ such that
7070: \mr
7071: \item "{$(a)$}" $(M_i,N_i,a) \in K^{3,\text{bs}}_{\frak s}$ is
7072: $\le_{\text{bs}}$-increasing continuous
7073: \sn
7074: \item "{$(b)$}" $(M_0,N_0) = (M,N)$
7075: \sn
7076: \item "{$(c)_1$}" $b^1_i \in M_{i+1} \backslash M_i$ and \ortp$_{\frak
7077: s}(b^1_i,M_i,M_{i+1})\in {\Cal S}^{\text{bs}}(M_i)$,
7078: \sn
7079: \item "{$(c)_2$}" $b^2_i \in N_{i+1} \backslash N_i$ and
7080: \ortp$(b^2_i,N_i,N_{i+1}) \in {\Cal S}^{\text{bs}}(N_i)$
7081: \sn
7082: \item "{$(d)_1$}" if $i < \lambda$ and $p \in {\Cal
7083: S}^{\text{bs}}(M_i)$ \ub{then} the set $\{j:i \le j < \lambda$ and
7084: \ortp$_{\frak s}(b^1_j,M_j,M_{j+1})$ is a non-forking extension of $p\}$
7085: has order type $\lambda$
7086: \sn
7087: \item "{$(d)_2$}" if $i < \lambda$ and $p \in {\Cal
7088: S}^{\text{bs}}(N_i)$ then the set $\{j:i \le j < \lambda$ and
7089: \ortp$_{\frak s}(b^2_j,N_j,N_{j+1})$ is the non-forking extension of
7090: $p\}$ has order type $\lambda$
7091: \sn
7092: \item "{$(e)$}" $\bar{\bold c}_i = \langle c_{i,j}:j < \lambda
7093: \rangle$ list $N_i$
7094: \sn
7095: \item "{$(f)$}" if $\alpha < \lambda,i \le \alpha,j <
7096: \lambda,c_{i,j} \notin M_\alpha$ but for some $(M'',N'')$ we have
7097: $(M_{\alpha +1},N_{\alpha +1},a) \le_{\text{bs}} (M'',N'',a)$ and
7098: $c_{i,j} \in M''$ \ub{then} for some $i_1,j_1 \le \text{ max}\{i,j\}$
7099: we have $c_{i_1,j_1} \in M_{\alpha +1} \backslash M_\alpha$.
7100: \ermn
7101: Lastly, let $M' = \cup\{M_i:i < \lambda\},N' = \cup\{N_i:i <
7102: \lambda\}$, by \scite{600-4a.2} $M'$ is
7103: $(\lambda,\text{cf}(\lambda))$-brimmed over $M$ (using $(d)_1$), and
7104: $N'$ is $(\lambda,\text{cf}(\lambda))$-brimmed over $N$ (using
7105: $(d)_2$). \nl
7106: Lastly, bs-reduced is by clauses (e)+(f).
7107: \nl
7108: 1A),2) Easy.
7109: \nl
7110: 3) For proving part (3) use part (1) and the ``so" is
7111: by using \scite{600-4a.2}.
7112: \nl
7113: 4) For proving part (4), \wilog \, $M_2$ is
7114: $(\lambda,\text{\rm cf}(\lambda))$-brimmed over $M_0$, as we can replace
7115: $M_2$ by $M'_2$ if $M_2 \le_{\frak K} M'_2 \in K_\lambda$. By part
7116: (3) there is a sequence $\langle a_i:i < \delta \rangle$ and
7117: an $\le_{\frak K}$-increasing continuous $\langle N_i:i \le
7118: \delta \rangle$ with $N_0
7119: = M_0,N_\delta = M_2$ and $(N_i,N_{i+1},a_i) \in K^{3,\text{bs}}_\lambda$
7120: is reduced. Then use (1A) successively.
7121: \nl
7122: 5) By part (3). \hfill$\square_{\scite{600-4a.4}}$
7123: \enddemo
7124: \bigskip
7125:
7126: \proclaim{\stag{600-4a.5} Claim} Assume
7127: \mr
7128: \item "{$(a)$}" $\gamma < \lambda^+$ is a limit ordinal
7129: \sn
7130: \item "{$(b)$}" $\delta_i < \lambda^+$ is divisible by $\lambda$ for
7131: $i \le \gamma,\langle \delta_i:i \le \gamma \rangle$ is increasing continuous
7132: \sn
7133: \item "{$(c)$}" $\langle N_i:i < \gamma \rangle$ is
7134: $\le_{\frak K}$-increasing continuous in $K_\lambda$
7135: \sn
7136: \item "{$(d)$}" $\langle M_i:i < \gamma \rangle$ is
7137: $\le_{\frak K}$-increasing continuous in $K_\lambda$
7138: \sn
7139: \item "{$(e)$}" $N_i \le_{\frak K} M_i$ for $i < \gamma$
7140: \sn
7141: \item "{$(f)$}" $\langle M_{i,j}:j \le \delta_i \rangle$ is
7142: $\le_{\frak K}$-increasing continuous in $K_\lambda$ for each $i < \gamma$
7143: \sn
7144: \item "{$(g)$}" $M_{i,0} = N_i,M_{i,\delta_i} = M_i,
7145: a_j \in M_{i,j+1} \backslash M_{i,j}$ and ${\text{\rm \ortp\/}}
7146: (a_j,M_{i,j},M_{i,j+1}) \in {\Cal S}^{\text{bs}}(M_{i,j})$ when $i
7147: < \gamma,j < \delta_i$
7148: \sn
7149: \item "{$(h)$}" if $j \le \delta_{i(*)},i(*) < \gamma$ then
7150: $\langle M_{i,j}:i \in [i(*),\gamma) \rangle$ is
7151: $\le_{\frak K}$-increasing continuous
7152: \sn
7153: \item "{$(i)$}" ${\text{\rm \ortp\/}}
7154: (a_j,M_{\beta,j},M_{\beta,j+1})$ does not fork over
7155: $M_{i,j}$ when $i < \gamma,j < \delta_i,i \le \beta < \gamma$
7156: \sn
7157: \item "{$(j)$}" if $i < \gamma,j < \delta_i,p \in {\Cal S}^{\text{bs}}
7158: (M_{i,j})$ \ub{then} for $\lambda$ ordinals $j_1 \in [j,\delta_i)$
7159: we have \text{\rm tp}$(a_{j_1},M_{i,j_1},M_{i,j_1+1}) \in {\Cal
7160: S}^{\text{bs}}(M_{i,j_1})$ is a non-forking extension of $p$ or we can
7161: ask less
7162: \sn
7163: \item "{$(j)^-$}" if $i < \gamma,j < \delta_i$ and $p \in {\Cal
7164: S}^{\text{bs}}(M_{i,j})$ \ub{then} for $\lambda$ ordinals
7165: $j_1 \in [j,\delta_\gamma)$ for some $i_1 \in [i,\gamma)$ we have
7166: ${\text{\rm \ortp\/}}(a_{j_1},M_{i_1,j_1},M_{i_1,j_1+1})
7167: \in {\Cal S}^{\text{bs}}(M_{i_1,j_1})$
7168: is a non-forking extension of $p$.
7169: \ermn
7170: \ub{Then} $M_\gamma := \cup\{M_{i,j}:i < \gamma,j < \delta_i\} =
7171: \{M_i:i < \gamma\}$ is
7172: $(\lambda,{\text{\rm cf\/}}(\gamma))$-brimmed over $N_\gamma :=
7173: \cup\{N_i:i < \gamma\}$.
7174: \endproclaim
7175: \bigskip
7176:
7177: \demo{Proof} For $j < \delta_\gamma$ let $M_{\gamma,j} =
7178: \cup\{M_{i,j}:i < \gamma\}$, and let
7179: $M_{\gamma,\delta_\gamma} = M_\gamma$ be $\cup\{M_{\gamma,j}:j <
7180: \delta_\gamma\}$. Easily $\langle M_{\gamma,j}:j \le \delta_\gamma
7181: \rangle$ is $\le_{\frak K}$-increasing continuous, $M_{\gamma,j} \in
7182: K_\lambda$ and $i \le \gamma \wedge j < \delta_i \Rightarrow M_{i,j}
7183: \le_{\frak K} M_{\gamma,j}$. Also if $i < \gamma,j < \delta_i$ then
7184: \ortp$(a_j,M_{\gamma,j},M_{\gamma,j+1}) \in {\Cal S}^{\text{bs}}(M_{\gamma,j})$
7185: does not fork over $M_{i,j}$ by Axiom (E)(h), continuity. \nl
7186: Now if $j < \delta_\gamma$ and $p \in {\Cal
7187: S}^{\text{bs}}(M_{\gamma,j})$ then for some $i < \gamma,p$ does not
7188: fork over $M_{i,j}$ (by Ax(E)(c)) and \wilog \, $j < \delta_i$.
7189:
7190: Hence if clause (j) holds we have
7191: $u := \{\varepsilon:j < \varepsilon < \delta_i$ and
7192: \ortp$(a_\varepsilon,M_{i,\varepsilon},M_{i,\varepsilon+1})$ is a
7193: non-forking extension of $p \restriction M_{i,j}\}$ has $\lambda$
7194: members. But for $\varepsilon \in u$,
7195: \ortp$(a_\varepsilon,M_{\gamma,\varepsilon},M_{\gamma,\varepsilon +1})$
7196: does not fork over $M_{i,\varepsilon}$ (by clause (i) of the
7197: assumption) hence does not fork over $M_{i,j}$ and by monotonicity it
7198: does not fork over $M_{\gamma,i}$ and by uniqueness it extends $p$.
7199: If clause $(j)^-$ holds the proof is similar.
7200: By \scite{600-4a.2} the model $M_\gamma$ is $(\lambda,\text{cf}(\gamma))$-brimmed
7201: over $N_\gamma$. \hfill$\square_{\scite{600-4a.5}}$
7202: \enddemo
7203: \bigskip
7204:
7205: \proclaim{\stag{600-4a.6} Lemma} 1) If $M \in K_\lambda$ and the models
7206: $M_1,M_2 \in K_\lambda$ are $(\lambda,*)$-brimmed over $M$
7207: (see Definition \scite{600-0.21}(2)), \ub{then} $M_1,M_2$ are isomorphic
7208: over $M$. \nl
7209: 2) If $M_1,M_2 \in K_\lambda$ are $(\lambda,*)$-brimmed \ub{then}
7210: they are isomorphic.
7211: \endproclaim
7212: \bn
7213: We prove some claims before proving \scite{600-4a.6}; we will not much use
7214: the lemma, but it is of obvious interest and its proof is crucial in
7215: one point of \S6.
7216: \proclaim{\stag{600-4a.7} Claim} 1)
7217: \mr
7218: \item "{$(E)(i)^+$}" \ub{long non-forking amalgamation for
7219: $\alpha < \lambda^+$}: \nl
7220: if $\langle N_i:i \le \alpha
7221: \rangle$ is $\le_{\frak K}$-increasing continuous
7222: sequence of members of $K_\lambda,a_i \in N_{i+1} \backslash N_i$ for
7223: $i < \alpha,p_i = { \text{\rm \ortp\/}}
7224: (a_i,N_i,N_{i+1}) \in {\Cal S}^{\text{bs}}(N_i)$
7225: and $q \in {\Cal S}^{\text{bs}}(N_0)$, \ub{then} we can find a
7226: $\le_{\frak K}$-increasing continuous sequence $\langle N'_i:i \le \alpha
7227: \rangle$ of members of $K_\lambda$ such that: $i \le \alpha \Rightarrow
7228: N_i \le_{\frak K} N'_i$; some $b \in N'_0 \backslash N_0$ realizes
7229: $q,{\text{\rm \ortp\/}}(b,N_\alpha,N'_\alpha)$ does not fork over $N_0$
7230: and ${\text{\rm \ortp\/}}(a_i,N'_i,N'_{i+1})$ does not fork over $N_i$
7231: for $i < \alpha$.
7232: \ermn
7233: 2) Above assume in addition that there are $M,b^*$ such that
7234: $N_0 \le_{\frak K} M \in K_\lambda,b^* \in M$ and ${\text{\rm \ortp\/}}
7235: (b^*,N_0,M) = q$. \ub{Then} we can add: there is a
7236: $\le_{\frak K}$-embedding of $M$ into $N'_0$ over $N_0$ mapping $b^*$ to
7237: $b$.
7238: \endproclaim
7239: \bigskip
7240:
7241: \demo{Proof} Straight (remembering Axiom (E)(i) on non-forking
7242: amalgamation of Definition \scite{600-1.1}). In details
7243: \nl
7244: 1) Let $M_0,b^*$ be such that $N_0 \le_{{\frak K}[{\frak s}]} M_0$ and
7245: $q = \text{ \ortp}(b^*,N_0,M_0)$ and apply part (2).
7246: \nl
7247: 2) We choose $(M_i,f_i)$ by induction on $i \le \alpha$ such that
7248: \mr
7249: \item "{$\circledast$}" $(a) \quad M_i \in {\frak K}_{\frak s}$ is
7250: $\le_{\frak K}$-increasing continuous
7251: \sn
7252: \item "{${{}}$}" $(b) \quad f_i$ is a $\le_{\frak K}$-embedding of
7253: $N_i$ into $M_i$
7254: \sn
7255: \item "{${{}}$}" $(c) \quad f_i$ is increasing continuous with $i \le
7256: \alpha$
7257: \sn
7258: \item "{${{}}$}" $(d) \quad M_0 = M$ and $f_0 = \text{ id}_{N_0}$
7259: \sn
7260: \item "{${{}}$}" $(e) \quad$ \ortp$(b^*,f_i(N_i),M_i)$ does not fork
7261: over $N_0$
7262: \sn
7263: \item "{${{}}$}" $(f) \quad$ \ortp$(f_{i+1}(a_i),M_i,M_{i+1})$ does not
7264: fork over $f_i(N_i)$.
7265: \ermn
7266: For $i=0$ there is nothing to do. For $i$ limit take unions. Lastly,
7267: for $i=j+1$, we can find $(M'_i,f'_i)$ such that $f_j \subseteq f'_i$
7268: and $f'_i$ is an isomorphism from $N_i$ onto $M$. Hence $f_j(N_j)
7269: \le_{{\frak K}[{\frak s}]} N'_i$. Now use Ax(E)(i) for
7270: $f_j(N_j),M'_i,N_i,f'_i(a_j),b^*$.
7271:
7272: Having carried the induction, we rename to finish.
7273: \hfill$\square_{\scite{600-4a.7}}$
7274: \enddemo
7275: \bn
7276:
7277: In the claim below, we are given a
7278: $\le_{{\frak K}_\lambda}$-increasing
7279: continuous $\langle M_i:i \le \delta\rangle$
7280: and $u_0,u_1,u_2 \subseteq \delta$ such that: $u_0$ is where we are
7281: already given $a_i \in M_{i+1} \backslash M_i,u_1 \subseteq \delta$ is
7282: where we shall choose $a_i(\in M'_{i+1} \backslash M'_i)$ and $u_2
7283: \subseteq \delta$ is the place which we ``leave for future use"; main
7284: case is $u_1 = \delta;u_0 = u_2 = \emptyset$.
7285: \proclaim{\stag{600-4a.8} Claim} 1) Assume
7286: \mr
7287: \item "{$(a)$}" $\delta < \lambda^+$ is divisible by $\lambda$
7288: \sn
7289: \item "{$(b)$}" $u_0,u_1,u_2$ are disjoint subsets of $\delta$
7290: \sn
7291: \item "{$(c)$}" $\delta = \sup(u_1)$ and ${\text{\rm otp\/}}(u_1)$
7292: is divisible by $\lambda$
7293: \sn
7294: \item "{$(d)$}" $\langle M_i:i \le \delta \rangle$ is
7295: $\le_{\frak K}$-increasing continuous in ${\frak K}_\lambda$
7296: \sn
7297: \item "{$(e)$}" $\bold{\bar a} = \langle a_i:i \in u_0 \rangle,a_i \in
7298: M_{i+1} \backslash M_i,{\text{\rm \ortp\/}}(a_i,M_i,M_{i+1})
7299: \in {\Cal S}^{\text{bs}}(M_i)$.
7300: \ermn
7301: \ub{Then} we can find $\bar M' = \langle M'_i:i \le \delta \rangle$ and
7302: $\bold{\bar a}' = \langle a_i:i \in u_1 \rangle$ such that
7303: \mr
7304: \item "{$(\alpha)$}" $\bar M'$ is $\le_{\frak K}$-increasing continuous
7305: in $K_\lambda$
7306: \sn
7307: \item "{$(\beta)$}" $M_i \le_{\frak K} M'_i$
7308: \sn
7309: \item "{$(\gamma)$}" if $i \in u_0$ then ${\text{\rm \ortp\/}}
7310: (a_i,M'_i,M'_{i+1})$ is
7311: a non-forking extension of ${\text{\rm \ortp\/}}(a_i,M_i,M_{i+1})$
7312: \sn
7313: \item "{$(\delta)$}" if $i \in u_2$ then $M'_i = M'_{i+1}$
7314: \sn
7315: \item "{$(\varepsilon)$}" if $i \in u_1$ then ${\text{\rm \ortp\/}}
7316: (a_i,M'_i,M'_{i+1})\in {\Cal S}^{\text{bs}}(M'_i)$
7317: \sn
7318: \item "{$(\zeta)$}" if $i < \delta,p \in {\Cal S}^{\text{bs}}(M'_i)$
7319: then for $\lambda$ ordinals $j \in u_1 \cap (i,\delta)$ the type
7320: ${\text{\rm \ortp\/}}(a_j,M'_j,M'_{j+1})$ is a non-forking extension of $p$.
7321: \ermn
7322: 2) If we add in part (1) the assumption
7323: \mr
7324: \item "{$(g)$}" $M_0 \le_{\frak K} N \in K_\lambda$
7325: \ermn
7326: \ub{then} we can add to the conclusion
7327: \mr
7328: \item "{$(\eta)$}" there is an $\le_{\frak K}$-embedding $f$ of $N$ into
7329: $M'_0$ over $M_0$ and moreover $f$ is onto.
7330: \ermn
7331: 3) If we add in part (1) the assumption
7332: \mr
7333: \item "{$(h)^+$}" $M_0 \le_{\frak K} N \in K_\lambda$ and $b \in N
7334: \backslash M_0,{\text{\rm \ortp\/}}(b,M_0,N) \in {\Cal S}^{\text{bs}}(M_0)$
7335: \ermn
7336: \ub{then} we can add to the conclusion
7337: \mr
7338: \item "{$(\eta)^+$}" as in $(\eta)$ and ${\text{\rm \ortp\/}}
7339: (f(b),M_\delta,M'_\delta)$ does not fork over $M_0$.
7340: \ermn
7341: 4) We can strengthen clause $(\zeta)$ in part (1) to
7342: \mr
7343: \item "{$(\zeta)^+$}" if $i < \delta$ and $p \in {\Cal S}^{\text{bs}}
7344: (M'_i)$ then
7345: for $\lambda$ ordinals $j$ we have $j \in [i,\delta) \cap u_1$ and
7346: ${\text{\rm \ortp\/}}(a_j,M'_j,M'_{j+1})$ is a non-forking extension of
7347: $p$ and ${\text{\rm otp\/}}(u_1 \cap j \backslash i) < \lambda$.
7348: \ermn
7349: 5) If $i \in u_2 \Rightarrow M_i = M_{i+1}$ then we can add $i \in u_2
7350: \Rightarrow M'_i = M'_{i+1}$.
7351: \endproclaim
7352: \bigskip
7353:
7354: \demo{Proof} Straight. Note that we can find a sequence $\langle
7355: u_{1,i,\varepsilon}:i < \delta,\varepsilon < \lambda \rangle$ such
7356: that: this is a sequence of pairwise disjoint subsets of $u_1$ each of
7357: cardinality $\lambda$ satisfying $u_{1,i,\varepsilon}
7358: \subseteq \{j:i < j,j \in u_1 \text{ and }
7359: |u_1 \cap (i,j)| < \lambda\}$ (or we can demand that $i \le i_1 < i_2
7360: \le \delta \wedge |u_1 \cap (i_1,i_2)| = \lambda \Rightarrow
7361: |u_{1,i,\varepsilon} \cap (i_1,i_2)| = \lambda$).
7362: \hfill$\square_{\scite{600-4a.8}}$
7363: \enddemo
7364: \bn
7365: Toward building our rectangles of models with sides of difference
7366: lengths (and then we shall use \scite{600-4a.5}) we show
7367: (to understand the aim of the clauses in the conclusion of \scite{600-4a.9}
7368: see the proof of \scite{600-4a.6} below):
7369: \proclaim{\stag{600-4a.9} Claim} Assume
7370: \mr
7371: \item "{$(a)$}" $\delta_\ell < \lambda^+$ is divisible by $\lambda$
7372: for $\ell = 1,2$
7373: \sn
7374: \item "{$(b)$}" $\bar M^\ell = \langle M^\ell_\alpha:\alpha \le \delta_\ell
7375: \rangle$ is $\le_{\frak K}$-increasing continuous for $\ell = 1,2$
7376: \sn
7377: \item "{$(c)$}" $u^\ell_0,u^\ell_1,u^\ell_2$ are disjoint subsets of
7378: $\delta_\ell,{\text{\rm otp\/}}(u^\ell_1)$ is divisible by
7379: $\lambda$ and $\delta_\ell = \sup(u^\ell_1)$ for $\ell = 1,2$
7380: \sn
7381: \item "{$(d)$}" $\bold{\bar a}^\ell \equiv \langle a^\ell_\alpha:\alpha
7382: \in u^\ell_0 \rangle$ and ${\text{\rm \ortp\/}}(a^\ell_\alpha,M^\ell_\alpha,
7383: M^\ell_{\alpha +1}) \in {\Cal S}^{\text{bs}}(M^\ell_\alpha)$ for
7384: $\ell = 1,2,\alpha \in u^\ell_0$
7385: \sn
7386: \item "{$(e)$}" $M^1_0 = M^2_0$
7387: \sn
7388: \item "{$(f)$}" $\alpha \in u^\ell_1 \cup u^\ell_2
7389: \Rightarrow M^\ell_\alpha = M^\ell_{\alpha +1}$ for $\ell =1,2$.
7390: \ermn
7391: \ub{Then} we can find $\bar f^\ell = \langle f^\ell_\alpha:\alpha \le
7392: \delta_\ell \rangle,\bold{\bar b}^\ell = \langle b^\ell_\alpha:
7393: \alpha \in u^\ell_0 \cup u^\ell_1 \rangle$ for $\ell = 1,2$ and $\bar M =
7394: \langle M_{\alpha,\beta}:\alpha \le \delta_1,\beta \le \delta_2
7395: \rangle$ and functions $\zeta:u^1_1 \rightarrow \delta_2$ and
7396: $\varepsilon:u^2_1 \rightarrow \delta_1$ such that
7397: \mr
7398: \item "{$(\alpha)_1$}" for each
7399: $\alpha \le \delta_1,\langle M_{\alpha,\beta}:\beta \le \delta_2 \rangle$
7400: is $\le_{\frak K}$-increasing continuous
7401: \sn
7402: \item "{$(\alpha)_2$}" for each $\beta \le \delta_2,
7403: \langle M_{\alpha,\beta}:\alpha \le \delta_1 \rangle$
7404: is $\le_{\frak K}$-increasing continuous
7405: \sn
7406: \item "{$(\beta)_1$}" for $\alpha \in u^1_0,b^1_\alpha$ belongs to
7407: $M_{\alpha +1,0}$ and ${\text{\rm \ortp\/}}
7408: (b^1_\alpha,M_{\alpha,\delta_2},M_{\alpha +1,
7409: \delta_2}) \in {\Cal S}^{\text{bs}}(M_{\alpha,\delta_2})$ does not fork over
7410: $M_{\alpha,0}$
7411: \sn
7412: \item "{$(\beta)_2$}" for $\beta \in u^2_0,b^2_\beta$ belongs to
7413: $M_{0,\beta +1}$ and ${\text{\rm \ortp\/}}
7414: (b^2_\beta,M_{\delta_1,\beta},M_{\delta_1,\beta +1})
7415: \in {\Cal S}^{\text{bs}}(M_{\delta_1,\beta})$ does not fork over
7416: $M_{0,\beta}$
7417: \sn
7418: \item "{$(\gamma)_1$}" for $\alpha \in u^1_1,\zeta(\alpha) < \delta_2$
7419: and we have $b^1_\alpha \in
7420: M_{\alpha +1,\zeta(\alpha) + 1}$ and ${\text{\rm \ortp\/}}
7421: (b^1_\alpha,M_{\alpha,\delta_2},
7422: M_{\alpha +1,\delta_2})$ does not fork over $M_{\alpha,\zeta(\alpha)+1}$
7423: \sn
7424: \item "{$(\gamma)_2$}" for $\beta \in u^2_1,
7425: \varepsilon(\beta) < \delta_1$ and we have
7426: $b^2_\beta \in M_{\varepsilon(\beta)+1,\beta + 1}$ and
7427: ${\text{\rm \ortp\/}}
7428: (b^2_\beta,M_{\delta_1,\beta},M_{\delta_1,\beta +1})$ does not
7429: fork over $M_{\varepsilon(\beta)+1,\beta}$
7430: \sn
7431: \item "{$(\delta)_1$}" if $\alpha < \delta_1,\beta < \delta_2$ and $p \in
7432: {\Cal S}^{\text{bs}}(M_{\alpha,\beta})$ or just $p \in {\Cal
7433: S}^{\text{bs}}(M_{\alpha,\beta +1})$ \ub{then} for $\lambda$ ordinals
7434: \footnote{we can add ``and otp$(\alpha' \cap u^1_1 \backslash
7435: \alpha_2) < \lambda$"}
7436: $\alpha' \in [\alpha,\delta_1) \cap u^1_1$, the type
7437: ${\text{\rm \ortp\/}}
7438: (b^1_{\alpha'},M_{\alpha',\beta +1},M_{\alpha +1,\beta +1})$ is a
7439: (well defined) non-forking extension of $p$ and $\beta = \zeta(\alpha')$
7440: \sn
7441: \item "{$(\delta)_2$}" if $\alpha < \delta_1,\beta < \delta_2$ and $p \in
7442: {\Cal S}^{\text{bs}}(M_{\alpha,\beta})$ or just $p \in {\Cal
7443: S}^{\text{bs}}(M_{\alpha +1,\beta})$ \ub{then} for $\lambda$ ordinals
7444: \footnote{we can add ``and otp$(\beta' \cap u^2_1 \backslash \beta_2)
7445: < \lambda$"}
7446: $\beta' \in [\beta,\delta_2) \cap u^2_1$, the type
7447: ${\text{\rm \ortp\/}}(b^2_{\beta'},
7448: M_{\alpha +1,\beta'},M_{\alpha+1,\beta'+1})$ is a non-forking
7449: extension of $p$ and $\alpha = \varepsilon(\beta')$
7450: \sn
7451: \item "{$(\varepsilon)$}" $M_{0,0} = M^1_0 = M^2_0$
7452: \sn
7453: \item "{$(\zeta)_1$}" $f^1_\alpha$ is an isomorphism from $M^1_\alpha$ onto
7454: $M_{\alpha,0}$ such that $\alpha \in u^1_0 \Rightarrow f^1_\alpha
7455: (a^1_\alpha) = b^1_\alpha$ \nl
7456: $f^1_0 = { \text{\rm id\/}}_{M^1_0}$ and
7457: $f^1_\alpha$ is increasing continuous with $\alpha$
7458: \sn
7459: \item "{$(\zeta)_2$}" $f^2_\beta$ is an isomorphism from $M^2_\beta$ onto
7460: $M_{0,\beta}$ such that $\beta \in u^2_0 \Rightarrow f^2_\beta(a^2_\beta) =
7461: b^2_\beta$ \nl
7462: $f^2_0 = { \text{\rm id\/}}_{M^2_0}$ and
7463: $f^2_\alpha$ is increasing continuous with $\alpha$
7464: \sn
7465: \item "{$(\eta)_1$}" if $\alpha \in u^1_2$ then $M_{\alpha,\beta} =
7466: M_{\alpha +1,\beta}$ for every $\beta \le \delta_2$
7467: \sn
7468: \item "{$(\eta)_2$}" if $\beta \in u^2_2$ then $M_{\alpha,\beta} =
7469: M_{\alpha,\beta +1}$ for every $\alpha \le \delta_1$.
7470: \endroster
7471: \endproclaim
7472: \bigskip
7473:
7474: \demo{Proof} Straight, divide $u^\ell_1$ to $\delta_{3 - \ell}$
7475: subsets large enough), in fact, we can first choose the function
7476: $\zeta(-),\varepsilon(-)$. Now choose $\langle M_{\alpha,\beta}:\alpha
7477: \le \delta_1,\beta \le \beta^* \rangle,\langle f^1_\alpha:\alpha \le
7478: \delta_1 \rangle,\langle f^2_\beta:\beta \le \beta^* \rangle$ and
7479: $\langle b^1_\alpha:\zeta(\alpha) \in \beta^* \rangle,\langle
7480: b^2_\beta:\beta < \beta^* \rangle$ by induction on $\beta^*$ using
7481: \scite{600-4a.8}. \hfill$\square_{\scite{600-4a.9}}$
7482: \enddemo
7483: \bigskip
7484:
7485: \demo{Proof of \scite{600-4a.6}} By \scite{600-0.22}(3), i.e., uniqueness of
7486: the $(\lambda,\theta_\ell)$-brimmed model over $M$, it is
7487: enough to show for any regular
7488: $\theta_1,\theta_2 \le \lambda$ that there is a model $N \in K_\lambda$ which
7489: is $(\lambda,\theta_\ell)$-brimmed over $M$ for $\ell = 1,2$.
7490: Let $\delta_1 = \lambda
7491: \times \theta_1,\delta_2 = \lambda \times \theta_2$ (ordinal
7492: multiplication, of course), $M^1_\alpha = M^2_\beta =
7493: M$ for $\alpha \le \delta_1,\beta \le \delta_2,u^1_0 = u^2_0 = \emptyset,
7494: u^1_1 = \delta_1,u^2_1 = \delta_2,u^1_2 = u^2_2 = \emptyset$. So there
7495: are $\langle M_{\alpha,\beta}:\alpha \le \delta_1,\beta \le \delta_2 \rangle,
7496: \langle b^1_\alpha:\alpha < \delta_1 \rangle,\langle b^2_\beta:\beta <
7497: \delta_2 \rangle$ and $\langle f^1_\alpha:\alpha \le \delta_1
7498: \rangle,\langle f^2_\beta:\beta \le \delta_2 \rangle$
7499: as in Claim \scite{600-4a.9}. Without loss of generality
7500: $f^1_\alpha = f^2_\alpha = \text{ id}_M$. Now
7501: \mr
7502: \item "{$(*)_1$}" $\langle M_{\alpha,\delta_2}:\alpha \le \delta_1
7503: \rangle$ is $\le_{\frak K}$-increasing continuous in $K_\lambda$ (by
7504: clause $(\alpha)_1$, of \scite{600-4a.9}). Also
7505: \sn
7506: \item "{$(*)_2$}" if $\alpha < \delta_1$ and
7507: $p \in {\Cal S}(M_{\alpha,\delta_2})$
7508: \ub{then} for $\lambda$ ordinals
7509: $\alpha' \in (\alpha,\delta_1) \cap u^1_1$ the type
7510: \ortp$(b^1_{\alpha',\delta_2},
7511: M_{\alpha',\delta_2},M_{\alpha' +1,\delta_2})$ is a
7512: non-forking extension of $p$.
7513: \ermn
7514: (Easy, by Axiom (E)(c) for some $\beta < \delta_2,p$ does not fork over
7515: $M_{\alpha,\beta +1}$ and use clause $(\delta)_1$ of \scite{600-4a.9}).
7516:
7517: So by \scite{600-4a.5}, $M_{\delta_1,\delta_2}$ is
7518: $(\lambda,\text{cf}(\delta_1))$-brimmed over $M_{0,\delta_2}$ which is $M$.
7519:
7520: Similarly $M_{\delta_1,\delta_2}$ is
7521: $(\lambda,\text{cf}(\delta_2))$-brimmed over $M_{\delta_1,0}$ which
7522: is $M$; so together we are done. \nl
7523: ${{}}$ \hfill$\square_{\scite{600-4a.6}}$
7524: \enddemo
7525: \bigskip
7526:
7527: \proclaim{\stag{600-4a.10A} Claim} 1) If $M \in K_{\lambda^+}$ and $p \in {\Cal
7528: S}^{\text{bs}}(M_0),M_0 \le_{\frak K} M$ (so $M_0 \in K_\lambda$),
7529: \ub{then} we can find
7530: $b,\langle N^0_\alpha:\alpha \le \lambda^+\rangle$ and $\langle
7531: N^1_\alpha:\alpha \le \lambda^+ \rangle$ such that
7532: \mr
7533: \item "{$(a)$}" $\langle N^0_\alpha:\alpha < \lambda^+ \rangle$ is a
7534: $\le_{\frak K}$-representation of $N^0_{\lambda^+} = M$
7535: \sn
7536: \item "{$(b)$}" $\langle N^1_\alpha:\alpha < \lambda^+ \rangle$ is a
7537: $\le_{\frak K}$-representation of $N^1_{\lambda^+} \in K_{\lambda^+}$
7538: \sn
7539: \item "{$(c)$}" $N^1_{\alpha +1}$ is $(\lambda,\lambda)$-brimmed over
7540: $N^1_\alpha$ (hence $N^1_{\lambda^+}$ is saturated over $\lambda$ in
7541: ${\frak K}$)
7542: \sn
7543: \item "{$(d)$}" $M_0 \le N^0_0$ and $N^0_\alpha \le_{\frak K} N^1_\alpha$
7544: \sn
7545: \item "{$(e)$}" ${\text{\rm \ortp\/}}_{\frak
7546: s}(b,N^0_\alpha,N^1_\alpha)$ is a non-forking extension of $p$ for
7547: every $\alpha < \lambda^+$.
7548: \ermn
7549: 2) We can add
7550: \mr
7551: \item "{$(f)$}" for $\alpha < \beta < \lambda^+,N^1_\beta$ is
7552: $(\lambda,*)$-brimmed over $N^0_\beta \cup N^1_\alpha$.
7553: \endroster
7554: \endproclaim
7555: \bigskip
7556:
7557: \demo{Proof} 1) Easy by long non-forking amalgamation \scite{600-4a.7}
7558: (see \scite{600-4a.10}).
7559: \nl
7560: 2) Use \scite{600-4a.5}. \hfill$\square_{\scite{600-4a.10A}}$
7561: \enddemo
7562: \bigskip
7563:
7564: \demo{\stag{600-4a.12} Conclusion} 1) $K_{\lambda^{++}} \ne \emptyset$. \nl
7565: 2) $K_{\lambda^+} \ne \emptyset$. \nl
7566: 3) {\rm No} $M \in K_{\lambda^+}$ is $\le_{\frak K}$-maximal.
7567: \enddemo
7568: \bigskip
7569:
7570: \demo{Proof} 1) By (2) + (3). \nl
7571: 2) By $(B)$ of \scite{600-1.1}. \nl
7572: 3) By \scite{600-4a.10A}. \hfill$\square_{\scite{600-4a.12}}$
7573: \enddemo
7574: \bn
7575: \margintag{600-4a.14F}\ub{\stag{600-4a.14F} Exercise}: 1) Let $M \in K_{\frak s}$ be superlimit
7576: and ${\frak t} = {\frak s}_{[M]}$, so $K_{\frak t}$ is categorical.
7577: If $(M,N,a) \in K^{\text{bs}}_{\frak t}$ is reduced for ${\frak t}$,
7578: \ub{then} it is reduced for ${\frak s}$.
7579: \nl
7580: 2) In \scite{600-4a.4}(3),(4),(5), we can omit the assumption ``${\frak s}$
7581: is categorical" if:
7582: \mr
7583: \item "{$(a)$}" we add in aprt (3), each $N_i$ is superlimit
7584: (equivalently brimmed)
7585: \sn
7586: \item "{$(b)$}" in parts (4),(5) add the assumption ``$M_0$ is
7587: superlimit".
7588: \ermn
7589: 2) Some extra assumption in \scite{600-4a.4}(5) is needed.
7590: \newpage
7591:
7592: \head {\S5 Non-structure or some unique amalgamation} \endhead \resetall \sectno=5
7593: \spuriousreset
7594: \bn
7595: We shall get from essentially $\dot I(\lambda^{++},K) < 2^{\lambda^{++}}$ or
7596: just $\dot I(\lambda^{++},K(\lambda^+$-saturated)) $< 2^{\lambda^{++}}$,
7597: many cases of uniqueness of amalgamation
7598: assuming WDmId$(\lambda^+)$ is not $\lambda^{++}$-saturated.
7599: The proof is similar to \cite{Sh:482}, \cite[\S3]{Sh:576}.
7600:
7601: We define $K^{3,\text{bt}}_\lambda$, it is a brimmed relative of
7602: $K^{3,\text{bs}}_\lambda$
7603: hence the choice of bt; it guarantees much brimness
7604: (see Definition \scite{600-nu.1}) hence it guarantees some uniqueness,
7605: that is, if $(M,N,a) \in K^{3,\text{bt}}_\lambda,M$ is unique
7606: (recalling the uniqueness of the brimmed model) and more crucially, we consider
7607: $K^{3,\text{uq}}_\lambda$, (the family of members of
7608: $K^{3,\text{bs}}_\lambda$ for which we have
7609: uniqueness in relevant extensions). Having enough such triples is the
7610: main conclusion of this section (in \scite{600-nu.6} under ``not too many
7611: non isomorphic models" assumptions). In \scite{600-nu.2} we give some
7612: properties of $K^{3,\text{bt}}_\lambda,K^{3,\text{uq}}_\lambda$.
7613:
7614: To construct models in $\lambda^{++}$ we use approximations of cardianlity in
7615: $\lambda^+$ with ``obligation" on the further construction, which are
7616: presented as pairs $(\bar M,\bar{\bold a}) \in K^{\text{sq}}_\lambda$ ordered
7617: by $\le_{\text{ct}}$, see Definition \scite{600-nu.3}, Claims \scite{600-nu.3A},
7618: \scite{600-nu.4}. We need more: the triples
7619: $(\bar M,\bar{\bold a},\bold f) \in K^{\text{mqr}}_S,K^{\text{nqr}}_S$
7620: in Definition \scite{600-nu.7}, Claim \scite{600-nu.8}.
7621: All this enables us to quote results of \cite[\S3]{Sh:576}, but apart
7622: from believing the reader do not need to know \cite{Sh:576}.
7623: \bigskip
7624:
7625: \demo{\stag{600-nu.0} Hypothesis}
7626: \mr
7627: \item "{$(a)$}" ${\frak s} =
7628: ({\frak K},\nonfork{}{}_{},
7629: {\Cal S}^{\text{bs}})$ is a good $\lambda$-frame.
7630: \endroster
7631: \enddemo
7632: \bigskip
7633:
7634: \definition{\stag{600-nu.1} Definition} 1) Let $K^{3,\text{bt}}_\lambda =
7635: K^{3,\text{bt}}_{\frak s}$
7636: be the set of triples $(M,N,a)$ such that for some $\sigma =
7637: \text{ cf}(\sigma) \le \lambda,
7638: M \le_{\frak K} N$ are both $(\lambda,\sigma)$-brimmed members of
7639: $K_\lambda,a \in N \backslash M$ and \ortp$(a,M,N) \in
7640: {\Cal S}^{\text{bs}}(M)$. \nl
7641: 2) For $(M_\ell,N_\ell,a_\ell) \in K^{3,\text{bt}}_\lambda$ for $\ell =1,2$
7642: let $(M_1,N_1,a_1) <_{\text{bt}} (M_2,N_2,a_2)$ mean $a_1 = a_2$,
7643: \ortp$(a_1,M_2,N_2)$ does not fork over $M_1$ and for
7644: some $\sigma_2 = \text{ cf}(\sigma_2) \le \lambda$, the model $M_2$ is
7645: $(\lambda,\sigma_2)$-brimmed over $M_1$ and the model $N_2$ is
7646: $(\lambda,\sigma_2)$-brimmed over $N_1$. Finally $(M_1,N_1,a_2)
7647: \le_{\text{bt}}(M_2,N_2,a_2)$ means
7648: $(M_1,N_1,a_1) <_{\text{bt}}
7649: (M_2,N_2,a_2)$ or $(M_1,N_1,a_1) = (M_2,N_2,a_2)$.
7650: \enddefinition
7651: \bigskip
7652:
7653: \definition{\stag{600-nu.1A} Definition} 1) Let ``$(M_0,M_2,a) \in
7654: K^{3,\text{uq}}_\lambda$" mean: $(M_0,M_2,a) \in
7655: K^{3,\text{bs}}_\lambda$ and:
7656: for every $M_1$ satisfying $M_0 \le_{\frak K} M_1 \in K_\lambda$,
7657: the amalgamation $M$ of $M_1,M_2$ over $M_0$, with \ortp$(a,M_1,M)$ not
7658: forking over $M_0$, is unique, that is:
7659: \mr
7660: \item "{$(*)$}" \ub{if} for $\ell =1,2$ we have $M_0 \le_{\frak K} M_1
7661: \le_{\frak K} M^\ell \in K_\lambda$ and $f_\ell$ is a
7662: $\le_{\frak K}$-embedding of $M_2$ into $M^\ell$ over $M_0$
7663: (so $f_1 \restriction M_0 = f_2 \restriction
7664: M_0 = \text{ id}_{M_0}$) such that \ortp$(f_\ell(a),M_1,M^\ell)$ does not
7665: fork over $M_0$, \underbar{then}
7666: {\roster
7667: \itemitem{ $(a)$ } [uniqueness]: \nl
7668: for some $M',g_1,g_2$ we have:
7669: $M_1 \le_{\frak K} M' \in K_\lambda$ and \newline
7670: $g_\ell$ is a $\le_{\frak K}$-embedding of $M^\ell$ into $M'$
7671: over $M_1$ for $\ell =1,2$ such that $g_1 \circ f_1 \restriction M_2 =
7672: g_2 \circ f_2 \restriction M_2$
7673: \sn
7674: \itemitem{ $(b)$ } [being reduced] $f_\ell(M_2) \cap M_1 = M_0$ \nl
7675: [this is ``for free" in the proofs; and is not really necessary so the
7676: decision if to include it is not important].
7677: \endroster}
7678: \ermn
7679: 2) $K^{3,\text{uq}}_\lambda$ is dense (or ${\frak s}$ has density for
7680: $K^{3,\text{uq}}_\lambda$) when $K^{3,\text{uq}}_\lambda$ is dense
7681: in $(K^{3,\text{bs}}_\lambda,\le_{\text{bs}})$, i.e., for every
7682: $(M_1,M_2,a) \in K^{3,\text{bs}}_\lambda$ there is $(M_1,N_2,a) \in
7683: K^{3,\text{uq}}_\lambda$ such that $(M_1,M_2,a) \le_{\text{bs}}
7684: (N_1,N_2,a) \in K^{3,\text{uq}}_\lambda$.
7685: \nl
7686: 3) $K^{3,\text{uq}}_\lambda$ has existence or ${\frak s}$ has
7687: existence for $K^{3,\text{uq}}_\lambda$ when for every $M_0 \in
7688: K_\lambda$ and $p \in {\Cal S}^{\text{bs}}(M_0)$ for some $M_1,a$ we
7689: have $(M_0,M_1,a) \in K^{3,\text{uq}}_\lambda$ and $p = \text{\rm
7690: \ortp}(a,M_0,M_1)$.
7691: \nl
7692: 4) $K^{3,\text{uq}}_{\frak s} = K^{3,\text{uq}}_\lambda$.
7693: \enddefinition
7694: \bigskip
7695:
7696: \proclaim{\stag{600-nu.2} Claim} 1) The relation $\le_{\text{bt}}$ is
7697: a partial order on $K^{3,\text{bt}}_\lambda$ that is transitive
7698: and reflexive (but not necessarily the parallel of Ax V of
7699: {\rm a.e.c.} see Definition \scite{600-0.2}). \nl
7700: 2) If $(M_\alpha,N_\alpha,a) \in K^{3,\text{bt}}_\lambda$
7701: is $\le_{\text{bt}}$-increasing
7702: continuous for $\alpha < \delta$ where $\delta$ is a limit ordinal
7703: $< \lambda^+$ \ub{then} $(M,N,a) = (\dbcu_{\alpha < \delta} M_\alpha,
7704: \dbcu_{\alpha < \delta} N_\alpha,a)$ belongs to
7705: $K^{3,\text{bt}}_\lambda$ and
7706: $\alpha < \delta \Rightarrow (M_\alpha,N_\alpha,a) \le_{\text{bt}}
7707: (M,N,a)$ and
7708: $(M,N,a)$ is a $\le_{\text{bt}}$-upper bound of
7709: $\langle (M_\alpha,N_\alpha,a):\alpha < \delta \rangle$. \nl
7710: 3) In $(*)$ of \scite{600-nu.1A}, clause (b) follows from (a).
7711: \endproclaim
7712: \bigskip
7713:
7714: \demo{Proof} Easy, e.g. (3) by the uniqueness (i.e., clause (a))
7715: and \scite{600-4a.4}(4). \hfill$\square_{\scite{600-nu.2}}$
7716: \enddemo
7717: \bn
7718: We now define $K^{\text{sq}}_{\lambda^+}$, a family of $\le_{\frak
7719: K}$-increasing continuous sequences (the reason for sq) in $K_\lambda$
7720: of length $\lambda^+$, will be used to approximate stages in
7721: constructing models in $K_{\lambda^{++}}$.
7722: \definition{\stag{600-nu.3} Definition} 1) Let $K^{\text{sq}}_{\lambda^+}
7723: = K^{\text{sq}}_{\frak s}$
7724: be the set of pairs $(\bar M,\bold{\bar a})$ such that (sq stands for
7725: sequence):
7726: \mr
7727: \item "{$(a)$}" $\bar M = \langle M_\alpha:\alpha < \lambda^+ \rangle$
7728: is a $\le_{\frak K}$-increasing continuous sequence of models
7729: from $K_\lambda$
7730: \sn
7731: \item "{$(b)$}" $\bold{\bar a} = \langle a_\alpha:\alpha \in S \rangle$,
7732: where $S \subseteq \lambda^+$ is stationary in $\lambda^+$ and
7733: $a_\alpha \in M_{\alpha +1} \backslash
7734: M_\alpha$
7735: \sn
7736: \item "{$(c)$}" for some club $E$ of $\lambda^+$ for every $\alpha
7737: \in S \cap E$ we have
7738: \ortp$(a_\alpha,M_\alpha,M_{\alpha +1}) \in {\Cal S}^{\text{bs}}(M_\alpha)$
7739: \sn
7740: \item "{$(d)$}" if $p \in {\Cal S}^{\text{bs}}
7741: (M_\alpha)$ \ub{then} for stationarily
7742: many $\delta \in S$ we have: \ortp$(a_\delta,M_\delta,M_{\delta +1}) \in
7743: {\Cal S}^{\text{bs}}(M_\delta)$ does not fork over $M_\alpha$ and extends $p$.
7744: \ermn
7745: In such cases we let $M = \dbcu_{\alpha < \lambda^+} M_\alpha$. \nl
7746: 2) When for $\ell = 1,2$ we are given $(\bar M^\ell,\bold{\bar a}^\ell) \in
7747: K^{\text{sq}}_{\lambda^+}$ we say $(\bar M^1,\bold{\bar a}^1)
7748: \le_{\text{ct}} (\bar M^2,
7749: \bold{\bar a}^2)$ \ub{if} for some club $E$ of $\lambda^+$, letting
7750: $\bold{\bar a}^\ell = \langle a^\ell_\delta:\delta \in S^\ell \rangle$ for
7751: $\ell =1,2$, of course, we have
7752: \mr
7753: \item "{$(a)$}" $S^1 \cap E \subseteq S^2 \cap E$
7754: \sn
7755: \item "{$(b)$}" if $\delta \in S^1 \cap E$ then
7756: {\roster
7757: \itemitem{ $(\alpha)$ } $M^1_\delta \le_{\frak K} M^2_\delta$,
7758: \sn
7759: \itemitem{ $(\beta)$ } $M^1_{\delta +1} \le_{\frak K} M^2_{\delta +1}$
7760: \sn
7761: \itemitem{ $(\gamma)$ } $a^2_\delta = a^1_\delta$
7762: \sn
7763: \itemitem{ $(\delta)$ } \ortp$(a^1_\delta,M^2_\delta,M^2_{\delta +1})$ does
7764: not fork over $M^1_\delta$, so in particular $a^1_\delta \notin M^2_\delta$.
7765: \endroster}
7766: \endroster
7767: \enddefinition
7768: \bigskip
7769:
7770: \demo{\stag{600-nu.3A} Observation} 1) If $(\bar M,\bold{\bar a}) \in
7771: K^{\text{sq}}_{\lambda^+}$ then $M := \dbcu_{\alpha < \lambda^+} M_\alpha \in
7772: K_{\lambda^+}$ is saturated. \nl
7773: 2) $K^{\text{sq}}_{\lambda^+}$ is partially ordered by
7774: $\le_{\text{ct}}$.
7775: \hfill$\square_{\scite{600-nu.3A}}$
7776: \enddemo
7777: \bigskip
7778:
7779: \proclaim{\stag{600-nu.4} Claim} Assume
7780: $\langle (\bar M^\zeta,\bold{\bar a}^\zeta):\zeta < \zeta^* \rangle$
7781: is $\le_{\text{ct}}$-increasing in $K^{\text{sq}}_{\lambda^+}$,
7782: and $\zeta^*$ is a limit ordinal $< \lambda^{++}$, \ub{then} the
7783: sequence has a $\le_{\text{ct}}$-{\rm lub} $(\bar M,\bold{\bar a})$.
7784: \endproclaim
7785: \bigskip
7786:
7787: \demo{Proof} Let $\bold{\bar a}^\zeta = \langle a^\zeta_\delta:\delta \in
7788: S_\zeta \rangle$ for $\zeta < \zeta^*$ and \wilog \, $\zeta^* = \text{ cf}
7789: (\zeta^*)$ and for $\zeta < \xi < \zeta^*$ let $E_{\zeta,\xi}$ be a club of
7790: $\lambda^+$ witnessing $(\bar M^\zeta,\bold{\bar a}^\zeta) \le_{\text{ct}}
7791: (\bar M^\xi,\bold{\bar a}^\xi)$.
7792: \enddemo
7793: \bn
7794: \ub{Case 1}: $\zeta^* < \lambda^+$.
7795:
7796: Let $E = \cap\{E_{\zeta,\xi}:\zeta < \xi < \zeta^*\}$ and for $\delta \in E$
7797: let $M_\delta = \cup\{M^\zeta_\delta:\zeta < \zeta^*\}$ and
7798: $M_{\delta +1} = \cup\{M^\zeta_{\delta +1}:\zeta < \zeta^*\}$ and for any
7799: other $\alpha,M_\alpha = M_{\text{Min}(E \backslash \alpha)}$. Let $S =
7800: \dbcu_{\zeta < \zeta^*} S_\zeta \cap E$ and for $\delta \in S$ let
7801: $a_\delta = a^\zeta_\delta$ for every $\zeta$ for which
7802: $\delta \in S_\zeta$. Clearly
7803: $M_\alpha \in K_\lambda$ is $\le_{\frak K}$-increasing continuous
7804: and $\zeta < \zeta^* \wedge
7805: \delta \in E \Rightarrow M^\zeta_\delta \le_{\frak K} M_\delta \and
7806: M^\zeta_{\delta +1} \le_{\frak K} M_{\delta +1}$.
7807:
7808: Now if $\delta \in E \cap S_\zeta$ then $\xi \in [\zeta,\zeta^*)$ implies
7809: \ortp$(a_\delta,M^\xi_\delta,M_{\delta+1}) = \text{ \ortp}(a^\zeta_\delta,
7810: M^\xi_\delta,M^\xi_{\delta+1})$ does not fork over $M^\zeta_\delta$
7811: (and $\langle M^\xi_\delta:\xi \in [\zeta,\delta)\rangle,\langle
7812: M^\xi_{\delta+1}:\xi \in [\zeta,\delta)\rangle$ are $\le_{\frak
7813: K}$-increasing continuous); hence
7814: by Axiom (E)(h) we know that \ortp$(a_\delta,M_\delta,M_{\delta +1})$ does not
7815: fork over $M^\zeta_\delta$ and in particular $\in {\Cal
7816: S}^{\text{bs}}(M_\delta)$. Also if $N \le_{\frak K} M := \dbcu_{\alpha <
7817: \lambda^+} M_\alpha,N \in K_\lambda$ and $p \in {\Cal S}^{\text{bs}}(N)$ then
7818: for some $\delta(*) \in E,N \le_{\frak K}
7819: M_{\delta(*)}$, let $p_1 \in {\Cal S}
7820: ^{\text{bs}}(M_{\delta(*)})$ be a non-forking
7821: extension of $p$, so for some
7822: $\zeta < \zeta^*,p$ does not fork over $M^\zeta_{\delta(*)}$ hence for
7823: stationarily many $\delta \in S_\zeta,q^0_\delta = \text{ \ortp}(a_\delta,
7824: M^\zeta_\delta,M^\zeta_{\delta +1})$ is a non-forking extension of $p_1
7825: \restriction M^\zeta_{\delta(*)}$, hence this holds for stationarily
7826: many $\delta \in S \cap E$ and for each such $\delta,
7827: q_\delta = \text{ \ortp}(a_\delta,
7828: M_\delta,M_{\delta +1})$ is a non-forking extension of $p_1 \restriction
7829: M^\zeta_{\delta(*)}$, hence of $p_1$ hence of $p$. Looking at the
7830: definitions, clearly $(\bar M,\bar{\bold a}) \in
7831: K^{\text{sq}}_{\lambda^+}$ and $\zeta < \zeta^* \Rightarrow (\bar
7832: M^\zeta,\bar{\bold a}^\zeta) \le_{\text{ct}} (\bar M,\bar{\bold a})$.
7833:
7834: Lastly, it is easy to check the $\le_{\text{ct}}$-e.u.b.
7835: \bn
7836: \ub{Case 2}: $\zeta^* = \lambda^+$.
7837:
7838: Similarly using diagonal union, i.e., $E = \{\delta <
7839: \lambda^+:\delta$ is a limit ordinal such that $\zeta < \xi < \delta
7840: \Rightarrow \delta \in E_{\zeta,\varepsilon}\}$ and we choose $M_\alpha =
7841: \cup\{M^\zeta_\alpha:\zeta < \alpha\}$ when $\alpha \in E$ and
7842: $M_\alpha = M_{\text{min}(E \backslash (\alpha +1))}$ otherwise.
7843: \hfill$\square_{\scite{600-nu.4}}$
7844: \bigskip
7845:
7846: \demo{\stag{600-nu.13.1} Observation} Assume $K^{3,\text{uq}}_\lambda$ is
7847: dense in $K^{3,\text{bs}}_\lambda$, i.e., in
7848: $(K^{3,\text{bs}}_\lambda,\le_{\text{bs}})$ and even in
7849: $(K^{3,\text{bt}}_\lambda,<_{\text{bt}})$. \ub{Then}
7850: \mr
7851: \item "{$(a)$}" if $M \in K_\lambda$ is superlimit and $p \in {\Cal
7852: S}^{\text{bs}}(M)$ then there are $N,a$ such that $(M,N,a) \in
7853: K^{3,\text{uq}}_\lambda$ and $p = \text{\rm \ortp}(a,M,N)$
7854: \sn
7855: \item "{$(b)$}" if in addition $K_{\frak s}$ is categorical (in
7856: $\lambda$) \ub{then} ${\frak s}$ has existence for
7857: $K^{3,\text{uq}}_\lambda$ (recall that this means
7858: that for every $M \in K_{\frak s}$ and
7859: $p \in {\Cal S}^{\text{bs}}(M)$ for some pair $(N,a)$ we have
7860: $(M,N,a) \in K^{3,\text{uq}}_\lambda$ and $p = \text{\rm \ortp}(a,M,N))$.
7861: \endroster
7862: \enddemo
7863: \bigskip
7864:
7865: \demo{Proof} Should be clear.
7866: \enddemo
7867: \bn
7868: Now the assumption of \scite{600-nu.13.1} are justified by the following
7869: theorem (and the categoricity in (b) is justified by Claim \scite{600-0.34}).
7870: \proclaim{\stag{600-nu.6} First Main Claim} Assume that
7871: \mr
7872: \item "{$(a)$}" as in \scite{600-nu.0}
7873: \sn
7874: \item "{$(b)$}" ${\text{\rm WDmId\/}}
7875: (\lambda^+)$ is not $\lambda^{++}$-saturated and
7876: $2^\lambda < 2^{\lambda^+} < 2^{\lambda^{++}}$ (or the parallel versions
7877: for the definitional weak diamond).
7878: \ermn
7879: If $\dot I(\lambda^{++},K) <
7880: 2^{\lambda^{++}}$, \ub{then} for every $(M,N,a) \in
7881: K^{3,\text{bs}}_\lambda$ there is $(M^*,N^*,a) \in
7882: K^{3,\text{bt}}_\lambda$ such that
7883: $(M,N,a) <_{\text{bt}} (M^*,N^*,a)$ and
7884: $(M^*,N^*,a) \in K^{3,\text{uq}}_\lambda$.
7885: \endproclaim
7886: \bigskip
7887:
7888: \demo{\stag{600-nu.6.2} Explanation} The reader who agrees to believe in
7889: \scite{600-nu.6} can ignore the rest of this section (though it can still
7890: serve as a good exercise).
7891:
7892: Let $\langle S_\alpha:\alpha < \lambda^{++}\rangle$ be a sequence of
7893: subsets of $\lambda^+$ such that $\alpha < \beta \Rightarrow |S_\alpha
7894: \backslash S_\beta| \le \lambda$ and $S_{\alpha +1} \backslash
7895: S_\alpha \ne \emptyset$ mod WDmId$(\lambda^+)$, exists by assumption.
7896:
7897: Why having $(M,N,a)$ failing the conclusion of \scite{600-nu.6} helps us to
7898: construct many models in $K_{\lambda^{++}}$? The point is that we can
7899: choose $(\bar M^\alpha,\bar{\bold a}^\alpha) \in
7900: K^{\text{sq}}_{\lambda^+}$ with Dom$(\bar{\bold a}^\alpha) = S_\alpha$
7901: for $\alpha <\lambda^{++},<_{\text{ct}}$-increasing continuous
7902: (see \scite{600-nu.4}).
7903:
7904: Now for $\alpha=\beta +1$, having $(\bar M^\beta,\bar{\bold
7905: a}^\beta)$, \wilog \, $M^\beta_{i+1}$ is brimmed over $M^\beta_i$
7906: and we shall choose $M^\alpha_i$
7907: by induction on $i < \lambda^+$ (for simplicity we pretend $M^\alpha_i
7908: \cap \cup\{M^\beta_j:j < \lambda^+\} = M^\beta_i$) and $M^\beta_i
7909: \le_{\frak K} M^\alpha_i \in K_\lambda$ and
7910: \ortp$(a_i,M^\alpha_i,M^\alpha_{i+1})$ does not fork over $M^\beta_i$ and
7911: $M^\alpha_{i+1}$ is brimmed over $M^\alpha_i$).
7912:
7913: Given $(\bar M^\beta,\bar{\bold a}^\beta),\bar M^\beta = \langle
7914: M^\beta_i:i < \lambda^+\rangle,\bar{\bold a}^\beta_i$ toward building
7915: $(\bar M^\alpha,\bar{\bold a}^\alpha),\alpha_{\beta +1}$.
7916:
7917: We start with choosing $(M^\alpha_0,b)$ such that no member of
7918: $K^{3,\text{bs}}_\lambda$ which is $\le_{\text{bs}}$-above
7919: $(M^\beta_0,M^\alpha_0,b) \in K^{3,\text{bs}}_\lambda$ belongs to
7920: $K^{3,\text{uq}}_\lambda$ and will choose $M^\beta_i$ by induction on
7921: $i$ such that $(M^\beta_i,M^\alpha_i,b) \in K^{3,\text{bs}}_\lambda$
7922: is $\le_{\text{bs}}$-increasing continuous and even
7923: $<_{\text{bt}}$-increasing hence in particular that
7924: \ortp$(b,M^\beta_i,M^\alpha_i)$ does not fork over $M^\alpha_0$. Now in
7925: each stage $i=j+1$, as $M^\beta_i$ is universal over $M^\beta_j$, and
7926: the choice of $M^\alpha_0,b$ we have some freedom. So it makes sense
7927: that we will have many possible outcomes, i.e., models $M =
7928: \cup\{M^\alpha_i:\alpha < \lambda^{++},i < \lambda^+\}$ which are
7929: in $K_{\lambda^{++}}$. The combination of what we have above and
7930: \cite[\S3]{Sh:576} gives
7931: that $2^\lambda < 2^{\lambda^+} < 2^{\lambda^{++}}$ is enough to
7932: materialize this intuition. If in addition $2^\lambda = \lambda^+$
7933: and moreover $\diamondsuit_{\lambda^+}$ it is considerably easier. In
7934: the end we still have to define $\bar{\bold a}^\alpha \restriction
7935: (S_\alpha \backslash S_\beta)$ as required in Definition \scite{600-nu.3}.
7936:
7937: Alternatively when \cite{Sh:839} materializes, it will do it in more
7938: transparent way. Another alternative is to force a model in
7939: $\lambda^{++}$. Now below we replace $K^{3,\text{sq}}_{\lambda^+}$ by
7940: $K^{\text{mqr}}_{\lambda^+},K^{\text{nqr}}_S$ but
7941: actually $K^{3,\text{sq}}_{\lambda^+}$ is enough,
7942: but not in the way \cite{Sh:576} is done. So we need a somewhat more
7943: complicated relative as elaborated below which anyhow seems to me more natural.
7944: \enddemo
7945: \bigskip
7946:
7947: \proclaim{\stag{600-nu.6A} Second Main Claim} Assume
7948: $2^\lambda < 2^{\lambda^+} < 2^{\lambda^{++}}$
7949: (or the parallel versions for the definitional weak
7950: diamond). If $\dot I(\lambda^{++},K) < \mu_{\text{wd}}(\lambda^{++},
7951: 2^{\lambda^+})$, \ub{then} for
7952: every $(M,N,a) \in K^{3,\text{bt}}_\lambda$ there is $(M^*,N^*,a) \in
7953: K^{3,\text{bt}}_\lambda$ such that $(M,N,a) <_{\text{bt}} (M^*,N^*,a)$ and
7954: $(M^*,N^*,a) \in K^{3,\text{uq}}_\lambda$.
7955: \endproclaim
7956: \bn
7957: We shall not prove here \scite{600-nu.6A} and shall not use it; toward
7958: proving \scite{600-nu.6} (by quoting) let
7959: \definition{\stag{600-nu.7} Definition} Let $S \subseteq \lambda^+$ be a
7960: stationary subset of $\lambda^+$. \nl
7961: 1) Let $K^{\text{mqr}}_S$ or $K^{\text{mqr}}_{\lambda^+}[S]$
7962: be the set of triples
7963: $(\bar M,\bold {\bar a},\bold f)$ such that:
7964: \mr
7965: \item "{$(a)$}" $\bar M = \langle M_\alpha:\alpha < \lambda^+ \rangle$ is
7966: $\le_{\frak K}$-increasing continuous, $M_\alpha \in K_\lambda$ \nl
7967: (we denote $\dsize \bigcup_{\alpha < \lambda^+} M_\alpha$ by $M$) and
7968: demand $M \in K_{\lambda^+}$
7969: \sn
7970: \item "{$(b)$}" $\bold {\bar a} = \langle a_\alpha:\alpha < \lambda \rangle$
7971: with $a_\alpha \in M_{\alpha + 1}$
7972: \sn
7973: \item "{$(c)$}" $\bold f$ is a function from $\lambda^+$ to $\lambda^+$
7974: such that for some club $E$ of $\lambda^+$ for every
7975: $\delta \in E \cap S$ and ordinal
7976: $i < \bold f(\delta)$ we have \ortp$(a_{\delta +i},M_{\delta +i},
7977: M_{\delta +i+ 1}) \in {\Cal S}^{\text{bs}}(M_{\delta +i})$
7978: \sn
7979: \item "{$(d)$}" for every $\alpha < \lambda^+$ and $p \in
7980: {\Cal S}^{\text{bs}}(M_\alpha)$, stationarily many
7981: $\delta \in S$ satisfies: for some $\varepsilon < \bold f(\delta)$ we have
7982: \ortp$(a_{\delta + \varepsilon},M_{\delta + \varepsilon},
7983: M_{\delta + \varepsilon +1})$ is a
7984: non-forking extension of $p$.
7985: \ermn
7986: 1A) $K^{\text{nqr}}_{\lambda^+}[S] =
7987: K^{\text{nqr}}_S$ is the set of triples $(\bar M,
7988: \bar{\bold a},\bold f) \in K^{\text{mqr}}_S$ such that:
7989: \mr
7990: \item "{$(e)$}" for a club of $\delta < \lambda^+$, if $\delta \in S$ then
7991: $\bold f(\delta)$ is divisible by $\lambda$ and \footnote{if we have an
7992: a priori bound $\bold f^*:\lambda^+ \rightarrow \lambda^+$ which is
7993: a $<_{{\Cal D}_{\lambda^+}}$-upper bound of the ``first"
7994: $\lambda^{++}$ functions in ${}^{\lambda^+}(\lambda^+)/D$, we can use
7995: bookkeeping for $u_i$'s as in the proof of \scite{600-4a.8}}
7996: for every $i < \bold f(\delta)$ if $q \in {\Cal S}^{\text{bs}}
7997: (M_{\delta + i})$ then for $\lambda$ ordinals $\varepsilon \in [i,
7998: \bold f(\delta))$ the type
7999: \ortp$(a_{\delta + \varepsilon},M_{\delta + \varepsilon},
8000: M_{\delta + \varepsilon +1}) \in {\Cal S}^{\text{bs}}
8001: (M_{\delta + \varepsilon})$ is a stationarization of $q$ (=
8002: non-forking extension of $q$, see Definition \scite{600-4a.3}).
8003: \ermn
8004: 2) Assume $(\bar M^\ell,\bold {\bar a}^\ell,
8005: \bold f^\ell) \in K^{\text{mqr}}_S$ for
8006: $\ell =1,2$; we say
8007: $(\bar M^1,\bold{\bar a}^1,\bold f^1) \le^0_S (\bar M^2,\bold{\bar a}^2,
8008: \bold f^2)$ \underbar{iff} for some club $E$ of $\lambda^+$, for every
8009: $\delta \in E \cap S$ we have:
8010: \medskip
8011: \roster
8012: \item "{$(a)$}" $M^1_{\delta +i} \le_{\frak K} M^2_{\delta + i}$ for
8013: \footnote{could have used (systematically) $i < \bold f^1(\delta)$}
8014: $i \le \bold f^1(\delta)$
8015: \sn
8016: \item "{$(b)$}" $\bold f^1(\delta) \le \bold f^2(\delta)$
8017: \sn
8018: \item "{$(c)$}" for $i < \bold f^1(\delta)$ we have
8019: $a^1_{\delta+i} = a^2_{\delta+i}$ and \newline
8020: \ortp$(a^1_{\delta+i},M^2_{\delta +i},M^2_{\delta +i+1})$
8021: does not fork over $M^1_{\delta +i}$.
8022: \ermn
8023: 3) We define the relation $<^1_S$ on $K^{\text{mqr}}_S$ as in part (2) adding
8024: \mr
8025: \item "{$(d)$}" if $\delta \in E$ and $i < \bold f^1(\delta)$ then
8026: $M^2_{\delta +i+1}$ is $(\lambda,*)$-brimmed over $M^1_{\delta +i+1} \cup
8027: M^2_{\delta +i}$.
8028: \endroster
8029: \enddefinition
8030: \bigskip
8031:
8032: \proclaim{\stag{600-nu.8} Claim} 0) If $(\bar M,\bar{\bold a},\bold f) \in
8033: K^{\text{mqr}}_S$ \ub{then}
8034: $\dbcu_{\alpha < \lambda^+} M_\alpha \in K_{\lambda^+}$ is
8035: saturated. \nl
8036: 1) The relation $\le^0_S$ is a quasi-order \footnote{quasi order $\le$
8037: is a transitive relation, so we waive $x \le y \le x \Rightarrow x=y$}
8038: on $K^{\text{mqr}}_\lambda$; also $<^1_S$ is. \nl
8039: 2) $K^{\text{mqr}}_S \supseteq K^{\text{nqr}}_S
8040: \ne \emptyset$ for any stationary $S
8041: \subseteq \lambda^+$. \nl
8042: 3) For every $(\bar M,\bar{\bold a},\bold f) \in
8043: K^{\text{mqr}}_\lambda[S]$ for
8044: some $(\bar M',\bar{\bold a},\bold f') \in K^{\text{nqr}}_\lambda[S]$ we have
8045: $(\bar M,\bar{\bold a},\bold f) <^1_S (\bar M',\bar{\bold a},\bold f')$. \nl
8046: 4) For every $(\bar M^1,\bold{\bar a}^1,\bold f^1) \in K^{\text{mqr}}_S$ and
8047: $q \in {\Cal S}^{\text{bs}}(M^1_\alpha),\alpha < \lambda^+$,
8048: \ub{there is} $(M^2,\bold{\bar a}^2,\bold f^2) \in K^{\text{mqr}}_S$
8049: such that $(\bar M^1,\bold{\bar a}^1,\bold f^1) <^1_S
8050: (\bar M^2,\bold{\bar a}^2,\bold f^2) \in K^{\text{nqr}}_S$
8051: and $b \in M^2_\alpha$ realizing $q$ such that for
8052: every $\beta \in [\alpha,\lambda^+)$ we have
8053: {\rm \ortp}$(b,M^1_\beta,M^2_\beta) \in {\Cal S}^{\text{bs}}
8054: (M^1_\beta)$ does not fork over $M^1_\alpha$. \nl
8055: 5) If $\langle(\bar M^\zeta,\bold{\bar a}^\zeta,\bold f^\zeta):\zeta <
8056: \xi(*) \rangle$ is $\le^0_S$-increasing continuous in $K^{\text{mqr}}_S$ and
8057: $\xi(*) < \lambda^{++}$ a limit ordering, \ub{then} the sequence has
8058: a $\le^0_S$-{\rm lub}.
8059: \endproclaim
8060: \bigskip
8061:
8062: \demo{Proof} 0, 1) Easy.
8063: \nl
8064: 2) The inclusion $K^{\text{mqr}}_S \supseteq K^{\text{nqr}}_S$ is
8065: obvious, so let us prove $K^{\text{nqr}}_S \ne \emptyset$.
8066: We choose by induction on
8067: $\alpha < \lambda^+,a_\alpha,M_\alpha,p_\alpha$ such that
8068: \mr
8069: \item "{$(a)$}" $M_\alpha \in K_\lambda$ is a super limit model,
8070: \sn
8071: \item "{$(b)$}" $M_\alpha$ is $\le_{\frak K}$-increasingly continuous,
8072: \sn
8073: \item "{$(c)$}" if $\alpha = \beta +1$, then $a_\beta \in M_\alpha
8074: \backslash M_\beta$ realizes $p_\beta \in {\Cal S}^{\text{bs}}(M_\beta)$,
8075: \sn
8076: \item "{$(d)$}" if
8077: $p \in {\Cal S}^{\text{bs}}(M_\alpha)$, then for some $i <
8078: \lambda$, for every $j \in [i,\lambda)$ for at least one ordinal
8079: $\varepsilon \in [j,j + i),p_{\alpha + \varepsilon} \restriction
8080: M_\alpha = p$ and $p_{\alpha + \varepsilon}$ does not fork over $M_\alpha$.
8081: \ermn
8082: For $\alpha = 0$ choose $M_0 \in K_\lambda$. For $\alpha$ limit,
8083: $M_\alpha = \dbcu_{\beta < \alpha} M_\beta$ is as required.
8084: Then use Axiom(E)(g) to take care of clause (d) (with careful
8085: bookkeeping).
8086: Lastly, let $\bold f:\lambda^+ \rightarrow \lambda^+$ be constantly
8087: $\lambda,\bar M = \langle M_\alpha:\alpha < \lambda \rangle,\bar{\bold
8088: a} = \langle a_\alpha:\alpha < \lambda \rangle$; now for any
8089: stationary $S \subseteq \lambda^+$, the triple $(\bar M,\bar{\bold a}
8090: \restriction S,\bold f \restriction S)$ belong to $K^{\text{nqr}}_S$.
8091: \nl
8092: 3) Let $E$ be a club witnessing $(\bar M^1,\bold{\bar a}^1,\bold f^1) \in
8093: K^{\text{mqr}}_S$ such that
8094: $\delta \in E \Rightarrow \delta + \bold f^1(\delta) <
8095: \text{ Min}(E \backslash (\delta +1))$. Choose $\bold f^2:\lambda^+
8096: \rightarrow \lambda^+$ such that $\alpha < \lambda^+$ implies
8097: $\bold f^1(\alpha) < \bold f^2(\alpha) < \lambda^+$ and $\bold
8098: f^2(\alpha)$ is divisible by $\lambda$. We
8099: choose by induction on $\alpha < \lambda^+,f_\alpha,M^2_\alpha,p_\alpha,
8100: a^2_\alpha$ such that:
8101: \mr
8102: \widestnumber\item{$(a),(b),(c)$}
8103: \item "{$(a),(b),(c)$}" as in the proof of part (2)
8104: \sn
8105: \item "{$(d)$}" $f_\alpha$ is a $\le_{\frak K}$-embedding of
8106: $M^1_\alpha$ into $M^2_\alpha$
8107: \sn
8108: \item "{$(e)$}" $f_\alpha$ is increasing continuous
8109: \sn
8110: \item "{$(f)$}" if $\delta \in E \cap S$ and $i < \bold f^1(\delta)$ hence
8111: \ortp$(a^1_{\delta +i},M^1_{\delta +i},M^1_{\delta +i+1}) \in
8112: {\Cal S}^{\text{bs}}(M^1_{\delta +i})$, \ub{then} $f_{\delta +i+1}
8113: (a^1_{\delta +i}) =
8114: a^2_{\delta +i}$ and $p_{\varepsilon +i} = \text{ \ortp}(a^2_{\delta +i},
8115: M^2_{\delta +i},M^2_{\delta + i+1}) \in {\Cal S}^{\text{bs}}
8116: (M^2_{\delta +i})$
8117: is a stationarization of \ortp$\bigl(f_{\delta + i+1}(a^1_{\delta +i}),
8118: f_{\delta +i}
8119: (M^1_{\delta+i}),f_{\delta +i+1}(M^1_{\delta +i+1}) \bigr) =$
8120: \nl
8121: $\text{ \ortp}(a^2_{\delta +i},
8122: f_{\delta +i}(M^1_{\delta +i}),M^2_{\delta +i+1})$
8123: \sn
8124: \item "{$(g)$}" if $\delta \in E$ and $i < \bold f^2(\delta),
8125: q \in {\Cal S}^{\text{bs}}(M^2_{\delta +i})$ then for some
8126: $\lambda$ ordinals $\varepsilon \in (i,\bold f^2(\delta))$ the type
8127: $p_{\delta + \varepsilon}$ is a stationarization of $q$
8128: \sn
8129: \item "{$(h)$}" if $\delta \in E,i < \bold f^2(\delta)$ then
8130: $M_{\delta +i+1}$ is $(\lambda,*)$-brimmed over $M_{\delta +i} \cup
8131: f_{\delta +i+1}(M^1_{\delta +i+1})$.
8132: \ermn
8133: The proof is as in part (2) only the bookkeeping is different.
8134: At the end without loss of generality
8135: $\dbcu_{\alpha < \lambda^*} f_\alpha$ is the identity and we are done. \nl
8136: 4) Similar proof but in some cases we have to use
8137: Axiom (E)(i), the non-forking amalgamation
8138: of Definition \scite{600-1.1}, in the appropriate cases. \nl
8139: 5) Without loss of generality cf$(\xi(*)) = \xi(*)$. First
8140: assume that $\xi(*) \le \lambda$. For $\varepsilon < \zeta < \xi(*)$ let
8141: $E_{\varepsilon,\zeta}$ be a club of $\lambda^+$ witnessing
8142: $\bar M^\varepsilon <^0_S \bar M^\zeta$. Let \newline
8143: $E^* = \dsize \bigcap_{\varepsilon < \zeta < \xi(*)} E_{\varepsilon,\zeta}
8144: \cap \{ \delta < \lambda^+:\text{for every } \alpha < \delta \text{ we
8145: have }
8146: \underset{\varepsilon < \xi(*)} {}\to \sup \bold f^\varepsilon(\alpha) <
8147: \delta\}$, it is a club of $\lambda^+$.
8148: Let $\bold f^{\xi(*)}:\lambda^+ \rightarrow \lambda^+ \text{ be }
8149: \bold f^{\xi(*)}(i) = \underset{\varepsilon < \xi(*)} {}\to \sup
8150: \bold f^\varepsilon(i)$ now define $M^{\xi(*)}_i$ as follows:
8151: \mn
8152: \ub{Case 1}: If $\delta \in E^*$ and $\varepsilon < \xi(*)$ and $i \le
8153: \bold f^\varepsilon(\delta)$ and $i \ge \dbcu_{\zeta < \varepsilon}
8154: \bold f^\zeta(\delta)$ then
8155: \mr
8156: \item "{$(\alpha)$}" $M^{\xi(*)}_{\delta +i} = \bigcup
8157: \bigl\{ M^\zeta_{\delta+i}:\zeta \in [\varepsilon,\xi(*)) \bigr\}$
8158: \sn
8159: \item "{$(\beta)$}" $i < \bold f^\varepsilon(\delta)
8160: \Rightarrow a^{\xi(*)}_{\delta+i} = a^\varepsilon_{\delta+i}$.
8161: \ermn
8162: (Note: we may define $M^{\xi(*)}_{\delta +i}$ twice if $i = \bold
8163: f^\varepsilon(\delta)$, but the two values are the same).
8164: \sn
8165: \ub{Case 2}: If $\delta \in E^*,i = \bold f^{\xi(*)}(\delta)$ is a limit
8166: ordinal let
8167:
8168: $$
8169: M^{\xi(*)}_{\delta +i} = \dsize \bigcup_{j<i} M^{\xi(*)}_{\delta+i}.
8170: $$
8171: \mn
8172: \ub{Case 3}: If $M^{\xi(*)}_i$ has not been defined yet, let it be
8173: $M^{\xi(*)}_{\text{Min}(E^* \backslash i)}$.
8174: \mn
8175: \ub{Case 4}: If $a^{\xi(*)}_i$ has not been defined yet, let
8176: $a^{\xi(*)}_i \in M^{\xi(*)}_{i+1}$ be arbitrary.
8177: \medskip
8178:
8179: Note that Case 3,4 deal with the ``unimportant" cases. \nl
8180: Let $\varepsilon < \xi(*)$, why
8181: $(\bar M^\varepsilon,\bold{\bar a}^\varepsilon,
8182: \bold f^\varepsilon) \le^0_S (\bar M^{\xi(*)},\bold{\bar a}^{\xi(*)},
8183: \bold f^{\xi(*)}) \in K^{\text{mqr}}_S$? Enough to check
8184: that the club $E^*$ witnesses it. \nl
8185: Why \ortp$(a_{\delta +i},M^{\xi(*)}_{\delta +i},M^{\xi(*)}_{\delta +i+1}) \in
8186: {\Cal S}^{\text{bs}}(M^{\xi(*)}_{\delta +i})$ and when $\delta \in E^*,i <
8187: \bold f^{\xi(*)}(i)$, and does not fork over $M^\varepsilon_{\delta +i}$ when
8188: $i < \bold f^\varepsilon(\delta)$ ? by Axiom (E)(h) of Definition
8189: \scite{600-1.1}. \nl
8190: Why clause (e) of Definition \scite{600-nu.7}(1A)? By Axiom (E)(c), local
8191: character of non-forking.
8192:
8193: The case $\xi(*) = \lambda^+$ is similar using diagonal intersections.
8194: \hfill$\square_{\scite{600-nu.8}}$
8195: \enddemo
8196: \bigskip
8197:
8198: \remark{Remark} If we use weaker versions of ``good
8199: $\lambda$-frames", we should
8200: systematically concentrate on successor $i < \bold f(\delta)$.
8201: \endremark
8202: \bigskip
8203:
8204: \demo{Proof of \scite{600-nu.6}} The use of $\lambda^{++} \notin \text{
8205: WDmId}(\lambda^{++})$ is as in the proof of
8206: \cite[3.19]{Sh:576}(pg.79)=3.12t. But
8207: now we need to preserve saturation in limit stages $\delta <
8208: \lambda^{++}$ of cofinality $< \lambda^+$, we use $<^1_S$, otherwise
8209: we act as in \cite[\S3]{Sh:576}. \hfill$\square_{\scite{600-nu.6}}$
8210: \enddemo
8211: \bn
8212: Let us elaborate
8213: \definition{\stag{600-nu.11} Definition} We define $\bold C =
8214: ({\frak K}^+,\bold S eq,\le^*)$ as follows:
8215: \mr
8216: \item "{$(a)$}" $\tau^+ = \tau \cup \{P,<\},{\frak K}^+$ is the set of
8217: $(M,P^M,<^M)$ where $M \in {\frak K}_{< \lambda},P^M \subseteq M,<^M$
8218: a linear ordering of $P^M$ (but $=^M$ may be as in
8219: \cite[3.1]{Sh:576}(2) and $M_1 \le_{{\frak K}^+} M_2$
8220: iff $(M_1 \restriction \tau) \le_{\frak K} (M_2
8221: \restriction \tau)$ and $M_1 \subseteq M_2$
8222: \sn
8223: \item "{$(b)$}" ${\bold Seq}_\alpha = \{\bar M:\bar M = \langle M_i:i \le
8224: \alpha \rangle$ is an increasing continuous sequence of members of
8225: ${\frak K}^+$ and $\langle M_i \restriction \tau:i \le \alpha \rangle$ is
8226: $\le_{\frak K}$-increasing, and for \newline
8227: $i < j < \alpha:P^{M_i}$ is a proper initial segment of
8228: $(P^{M_j},<^{M_j})$ and there is a first element in the difference$\}$ \nl
8229: we denote
8230: the $<^{M_{i+1}}$-first element of $P^{M_{i+1}} \backslash P^{M_i}$,
8231: by $a_i[\bar M]$ and we demand \ortp$(a_i(\bar M),M_i \tau
8232: \restriction,M_{i+1} \restriction \tau) \in {\Cal S}^{\text{bs}}(M_i
8233: \restriction \tau)$ and if $\alpha = \lambda,M = \cup\{M_i
8234: \restriction \tau:i < \lambda^+\}$ is saturated
8235: \sn
8236: \item "{$(c)$}" $\bar M <^*_t \bar N$ \underbar{iff} \newline
8237: $\bar M = \langle M_i:i < \alpha^* \rangle,\bar N = \langle N_i:
8238: i < \alpha^{**} \rangle$ are from ${\bold Seq},t$ is a set of pairwise
8239: disjoint closed intervals of $\alpha^*$ and for any $[\alpha,\beta] \in t$
8240: we have $(\beta < \alpha^*$ and):
8241: \sn
8242: $\gamma \in [\alpha,\beta) \Rightarrow M_\gamma \le_{\frak K} N_\gamma \and
8243: a_\gamma[\bar M] \notin N_\gamma$, moreover \newline
8244: $a_\gamma[\bar M] = a_\gamma[\bar N]$ and \ortp$(a_j[\bar M],N_\gamma
8245: \restriction \tau,N_{\gamma+1},\tau)$ does not fork over $M_\gamma
8246: \restriction \tau$.
8247: \endroster
8248: \enddefinition
8249: \bigskip
8250:
8251: \proclaim{\stag{600-nu.12} Claim} 1) $\bold C$ is a
8252: $\lambda^+$-construction framework (see \cite[3.3]{Sh:576}(pg.68). \nl
8253: 2) $\bold C$ is weakly nice (see Definition \cite[3.14]{Sh:576}(2)(pg.76). \nl
8254: 4) $\bold C$ has the weakening $\lambda^+$-coding property.
8255: \endproclaim
8256: \bn
8257: \ub{Discussion}: Is it better to use (see \cite[3.14]{Sh:576}(1)(pg.75))
8258: stronger axiomatization in \cite[\S3]{Sh:576} to cover this? \nl
8259: But at present this will be the only case.
8260: \bigskip
8261:
8262: \demo{Proof} Straight. \hfill$\square_{\scite{600-nu.12}}$
8263: \enddemo
8264: \bn
8265: Now \scite{600-nu.6A} follows by \cite[3.19]{Sh:576}(pg.79).
8266: \newpage
8267:
8268: \head {\S6 Non-forking amalgamation in ${\frak K}_\lambda$} \endhead \resetall \sectno=6
8269: \spuriousreset
8270: \bigskip
8271:
8272: We deal in this section only with ${\frak K}_\lambda$. \newline
8273: We would like to, at least, approximate ``non-forking amalgamation of
8274: models" using as a starting point the conclusion of \scite{600-nu.6}, i.e.,
8275: $K^{3,\text{uq}}_\lambda$ is dense.
8276: We use what looks like a stronger hypothesis: the existence for
8277: $K^{3,\text{uq}}_\lambda$ (also called ``weakly successful");
8278: but in our application we can assume
8279: categoricity in $\lambda$; the point being that as we have a
8280: superlimit $M \in K_\lambda$ this assumption is reasonable when we restrict
8281: ourselves to ${\frak K}^{[M]}$, recalling that we believe in first
8282: analyzing the saturated enough models; see \scite{600-nu.13.1}.
8283: By \scite{600-4a.6}, the ``$(\lambda,\text{cf}
8284: (\delta))$-brimmed over" is the same for all limit ordinals $\delta <
8285: \lambda^+$, (but not for $\delta = 1$); nevertheless for possible
8286: generalizations we do not use this.
8287: \bigskip
8288:
8289: It may help the reader to note, that if there is a 4-place relation
8290: NF$_\lambda(M_0,M_1,M_2,M_3)$ on $K_\lambda$, satisfying the expected
8291: properties of ``$M_1,M_2$ are amalgamated in a non-forking = free way over
8292: $M_0$ inside $M_3$", i.e., is a ${\frak K}_\lambda$-non-forking
8293: relation from Definition \scite{600-nf.0X}
8294: below then Definition \scite{600-nf.2} below (of NF$_\lambda$)
8295: gives it (provably!). So we have ``a definition" of
8296: NF$_\lambda$ satisfying that: if desirable non-forking relation
8297: exists, our definition gives it (assuming the hypothesis \scite{600-nf.0}). So
8298: during this section we are trying to get better and better
8299: approximations to the desirable properties; have the feeling of going
8300: up on a spiral, as usual.
8301:
8302: For the readers who know on non-forking in stable first order theory
8303: we note that in such context NF$_\lambda(M_0,M_1,M_2,M_3)$ says that
8304: \ortp$(M_2,M_1,M_3)$, the type of $M_2$ over $M_1$ inside $M_3$, does not
8305: fork over $M_0$. It is natural to say that there are $\langle
8306: N_{1,\alpha},N_{2,\alpha}:\alpha \le \alpha^*\rangle,N_{\ell,\alpha}$
8307: is increasing continuous. $N_{1,0} = M_0,N_{2,0} = M_2,M_1 \subseteq
8308: M_{1,\alpha},M_3 \subseteq M'_3,N_{2,\alpha} \subseteq
8309: M'_3,N_{\ell,\alpha +2}$ is prime over $N_{\ell,\alpha} + a_\alpha$
8310: for $\ell=1,2$ and \ortp$(a_\alpha,N_{2,\alpha})$ does not fork over
8311: $N_{1,\alpha}$ but this is not available. The
8312: $K^{3,\text{uq}}_\lambda$ is a substitute.
8313: \bigskip
8314:
8315: \definition{\stag{600-nf.0X} Definition} 1) Assume that ${\frak K} =
8316: {\frak K}_\lambda$ is a $\lambda$-a.e.c.
8317: We say NF is a non-forking relation on
8318: ${}^4({\frak K}_\lambda)$ or just a ${\frak K}_\lambda$-non-forking
8319: relation \ub{when}:
8320: \mr
8321: \item "{$\boxtimes_{\text{NF}}(a)$}" NF is a 4-place relation on
8322: $K_\lambda$ and NF is preserved under isomorphisms
8323: \sn
8324: \item "{$(b)$}" NF$(M_0,M_1,M_2,M_3)$ implies $M_0 \le_{\frak K}
8325: M_\ell \le_{\frak K} M_3$ for $\ell = 1,2$
8326: \sn
8327: \item "{$(c)_1$}" (monotonicity): if NF$(M_0,M_1,M_2,M_3)$ and $M_0
8328: \le_{\frak K} M'_\ell \le_{\frak K} M_\ell$
8329: for $\ell = 1,2$ then NF$(M_0,M'_1,M'_2,M_3)$
8330: \sn
8331: \item "{$(c)_2$}" (monotonicity): if NF$(M_0,M_1,M_2,M_3)$ and
8332: $M_3 \le_{\frak K} M'_3 \in K_\lambda,M_1 \cup M_2 \subseteq M''_3
8333: \le_{\frak K} M'_3$ \ub{then} NF$(M_0,M_1,M_2,M''_3)$
8334: \sn
8335: \item "{$(d)$}" (symmetry) NF$(M_0,M_1,M_2,M_3)$ iff
8336: NF$(M_0,M_2,M_1,M_3)$
8337: \sn
8338: \item "{$(e)$}" (transitivity) if NF$(M_i,N_i,M_{i+1},N_{i+1})$ for
8339: $i < \alpha,\langle M_i:i \le \alpha \rangle$ is $\le_{\frak
8340: K}$-increasing continuous and $\langle N_i:i \le \alpha \rangle$ is
8341: $\le_{\frak K}$-increasing continuous \ub{then}
8342: \nl
8343: NF$(M_0,N_0,M_\alpha,N_\alpha)$
8344: \sn
8345: \item "{$(f)$}" (existence) if $M_0 \le_{\frak K} M_\ell$ for $\ell
8346: =1,2$ (all in $K_\lambda)$ \ub{then} for some $M_3 \in K_\lambda,f_1,
8347: f_2$ we have
8348: $M_0 \le_{\frak K} M_3,f_\ell$ is a $\le_{\frak K}$-embedding of
8349: $M_\ell$ into $M_3$ over $M_0$ for $\ell = 1,2$ and
8350: NF$(M_0,f_1(M_1),f_2(M_2),M_3)$
8351: \sn
8352: \item "{$(g)$}" (uniqueness) if
8353: NF$(M^\ell_0,M^\ell_1,M^\ell_2,M^\ell_3)$ and for $\ell =
8354: 1,2$ and $f_i$ is an isomorphism from $M^1_i$ onto $M^2_i$ for $i =
8355: 0,1,2$ and $f_0 \subseteq f_1,f_0 \subseteq f_2$ \ub{then} $f_1 \cup
8356: f_2$ can be extended to an embedding $f_3$ of $M^1_3$ into some
8357: $M^2_4,M^2_3 \le_{{\frak K}_\lambda} M^2_4$.
8358: \ermn
8359: 2) We say that NF is a weak non-forking relation on ${}^4(K_\lambda)$
8360: or a weak ${\frak K}_\lambda$-non-forking relation
8361: if clauses (a)-(f) of $\boxtimes_{\text{NF}}$ above holds but not
8362: necessarily clause (g).
8363: \nl
8364: 3) Assume ${\frak s}$ is a good $\lambda$-frame and NF is a
8365: non-forking relation on ${\frak K}$ or just a formal one.
8366: We say that NF respects ${\frak s}$ or NF is an ${\frak
8367: s}$-non-forking relation \ub{when}:
8368: \mr
8369: \item "{$(h)$}" if NF$(M_0,M_1,M_2,M_3)$ and $a \in M_2 \backslash
8370: M_0$, \ortp$_{\frak s}(a,M_0,M_2) \in {\Cal S}^{\text{bs}}(M_0)$ then
8371: \ortp$_{\frak s}(a,M_1,M_3)$
8372: does not fork over $M_0$ in the sense of ${\frak s}$.
8373: \endroster
8374: \enddefinition
8375: \bigskip
8376:
8377: \demo{\stag{600-nf.0X.2} Observation} Assume ${\frak K}_\lambda$ is a
8378: $\lambda$-a.e.c. and NF is a non-forking relation on ${}^4({\frak
8379: K}_\lambda)$.
8380: \nl
8381: 1) Assume ${\frak K}$ is stable in $\lambda$. If in clause (g) of
8382: \scite{600-nf.0X}(1) above we assume in addition that $M^\ell_3$ is
8383: $(\lambda,\sigma)$-brimmed over $M^\ell_1 \cup M^\ell_2$, \ub{then} in
8384: the conclusion of (g) we can add $M^2_3 = M^2_4$, i.e., $f_1 \cup f_2$
8385: can be extended to an isomorphism from $M^1_3$ onto $M^2_3$. This
8386: version of (g) is equivalent to it (assuming stability in $\lambda$;
8387: note that ``${\frak K}_\lambda$ has amalgamation" follows by clause (h)
8388: of Definition \scite{600-nf.0X}).
8389: \nl
8390: 2) If $M_0 \le_{\frak K} M_1 \le_{\frak K} M_3$ are from $K_\lambda$
8391: then NF$(M_0,M_0,M_1,M_3)$.
8392: \nl
8393: 3) In Definition \scite{600-nf.0X}(1), clause (d), symmetry, it is enough
8394: to demand ``if".
8395: \enddemo
8396: \bigskip
8397:
8398: \demo{Proof} 1) Chase arrows.
8399: \nl
8400: 2) By clause (f) of $\boxtimes_{\text{NF}}$ of \scite{600-nf.0X}(1) and
8401: clause (c)$_2$, i.e., first apply existence
8402: with $(M_0,M_0,M_2)$ here standing for $(M_0,M_1,M_2)$
8403: there, then chase arrows and use the monotonicity in (c)$_2$.
8404: \nl
8405: 3) Easy. \hfill$\square_{\scite{600-nf.0X.2}}$
8406: \enddemo
8407: \bn
8408: The main point of the
8409: following claim shows that there is at most one non-forking
8410: relation respecting ${\frak s}$; so it justifies the definition of
8411: NF$_{\frak s}$ later. The assumption ``NF respects ${\frak s}$" is
8412: not so strong by \scite{600-nf.0.z.1}.
8413: \proclaim{\stag{600-nf.0Y} Claim} 1) If ${\frak s}$ is a
8414: good $\lambda$-frame and
8415: {\rm NF} is a non-forking relation on ${}^4({\frak K}_{\frak s})$
8416: respecting ${\frak s}$ and $(M_0,N_0,a) \in K^{3,\text{uq}}_\lambda$ and
8417: $(M_0,N_0,a) \le_{\text{bs}} (M_1,N_1,a)$ then
8418: ${\text{\rm NF\/}}(M_0,N_0,M_1,N_1)$. \nl
8419: 2) If ${\frak s}$ is a good $\lambda$-frame, weakly successful (which means
8420: $K^{3,\text{uq}}_{\frak s}$ has
8421: existence in $K^{3,\text{uq}}_{\frak s}$, i.e.,
8422: ${\frak s}$ satisfies hypothesis \scite{600-nf.0} below) and
8423: {\rm NF} is a non-forking relation on ${}^4({\frak K}_{\frak s})$
8424: respecting ${\frak s}$ \ub{then} the relation
8425: {\rm NF}$_\lambda = \text{\rm NF}_{\frak s}$, i.e.,
8426: $\nonforkin{N_1}{N_2}_{N_0}^{N_3}$ defined in Definition \scite{600-nf.2} below is
8427: equivalent to ${\text{\rm NF\/}}(N_0,N_1,N_2,N_3)$.
8428: [But see \scite{600-nf.20.7A}]
8429: \nl
8430: 3) If ${\frak s}$ is a weakly successful
8431: good $\lambda$-frame and for $\ell=1,2$, the
8432: relation {\rm NF}$_\ell$ is a non-forking relation on ${}^4({\frak
8433: K}_{\frak s})$ respecting ${\frak s}$,
8434: \ub{then} ${\text{\rm NF\/}}_1 = { \text{\rm NF\/}}_2$.
8435: \endproclaim
8436: \bigskip
8437:
8438: \demo{Proof} Straightforward but we elaborate.
8439: \nl
8440: 1) We can find $(M'_1,N'_1)$ such that NF$(M_0,N_0,M'_1,N'_1)$ and
8441: $M_1,M'_1$ are isomorphic over $M_0$, say $f_1$ is such an
8442: isomorphism from $M_1$ onto $M'_1$ over $M_0$; why such
8443: $(M'_1,N'_1,f_1)$ exists? by
8444: clause (f) of $\boxtimes_{\text{NF}}$ of Definition \scite{600-nf.0X}.
8445:
8446: As NF respects ${\frak s}$, see Definition \scite{600-nf.0X}(2), recalling
8447: \ortp$(a,M_0,N_0) \in {\Cal S}^{\text{bs}}(M_0)$ we know
8448: that \ortp$_{\frak s}(a,M'_1,N'_1)$ does not fork over $M_0$,
8449: so by the definition of
8450: $\le_{\text{bs}}$ we have $(M_0,N_0,a) \le_{\text{bs}}
8451: (M'_1,N'_1,a)$.
8452:
8453: As $(M_0,N_0,a) \in K^{3,\text{uq}}_\lambda$, by the definition of
8454: $K^{3,\text{uq}}_\lambda$ (and chasing arrows) we conclude that there
8455: are $N_2,f_2$ such that:
8456: \mr
8457: \item "{$(*)$}" $N_1 \le_{{\frak K}[{\frak s}]} N_2 \in K_\lambda$ and $f_2$
8458: is a $\le_{\frak K}$-embedding of $N'_1$ into $N_2$ extending
8459: $f^{-1}_1$ and id$_{N_0}$.
8460: \ermn
8461: As NF$(M_0,N_0,M'_1,N'_1)$ and NF is preserved under isomorphisms (see
8462: clause (a) in \scite{600-nf.0X}(1)) it follows that
8463: NF$(M_0,N_0,M_1,f_2(N'_1))$. By the monotonicity of NF (see clause
8464: $(c)_2$ of Definition \scite{600-nf.0X}) it follows that
8465: NF$(M_0,N_0,M_1,N_2)$. Again
8466: by the same monotonicity we have NF$(M_0,N_0,M_1,N_1)$, as required.
8467: \nl
8468: 2) First we prove that NF$_{\lambda,\bar \delta}(N_0,N_1,N_2,N_3)$,
8469: which is defined in Definition \scite{600-nf.1}
8470: below implies NF$(N_0,N_1,N_2,N_3)$. By
8471: definition \scite{600-nf.1}, clause (f) there are $\langle
8472: (N_{1,i},N_{2,i}:i \le \lambda \times \delta_1 \rangle),\langle c_i:i
8473: < \lambda \times \delta_1 \rangle$ as there. Now we prove by
8474: induction on $j\le \lambda \times \delta_1$ that $i \le j \Rightarrow
8475: \text{ NF}(N_{1,i},N_{2,i},N_{1,j},N_{2,j})$. For $j=0$ or more
8476: generally when $i=j$ this is trivial by \scite{600-nf.0X.2}(2).
8477: For $j$ a limit ordinal use the induction hypothesis and
8478: transitivity of NF (see clause (e) of \scite{600-nf.0X}(1)).
8479:
8480: Lastly, for $j$ successor by the demands in Definition \scite{600-nf.1} we
8481: know that $N_{1,j-1} \le_{\frak K} N_{1,j} \le_{\frak K}
8482: N_{2,j},N_{1,j-1} \le_{\frak K} N_{2,j-1} \le_{\frak K} N_{2,j}$ are
8483: all in $K_\lambda$, \ortp$(c_{j-1},N_{2,j-1},N_{2,j})$ does not fork over
8484: $N_{1,j-1}$ and $(N_{1,j-1},N_{1,j},c_{j-1}) \in
8485: K^{3,\text{uq}}_\lambda$. By part (1) of this claim we deduce that
8486: NF$(N_{1,j-1},N_{1,j},N_{2,j-1},N_{2,j})$ hence by symmetry (i.e.,
8487: clause (d) of Definition \scite{600-nf.0X}(1)) we deduce
8488: NF$(N_{1,j-1},N_{2,j-1},N_{1,j},N_{2,j})$.
8489:
8490: So we have gotten $i < j \Rightarrow$
8491: NF$(N_{1,i},N_{2,i},N_{1,j},N_{2,j})$. \nl
8492: [Why? If $i=j-1$ by the previous sentence and for $i < j-1$ note that
8493: by the induction hypothesis NF$(N_{1,i},N_{2,i},N_{1,j-1},N_{1,j-1})$
8494: so by transitivity (clause (e) of \scite{600-nf.0X}(1) of Definition
8495: \scite{600-nf.0X}) we get NF$(N_{1,i},N_{2,i},N_{1,j},N_{2,j})$].
8496:
8497: We have carried the induction so in particular for $i=0,j=\alpha$ we get
8498: NF$(N_{1,0},N_{2,0},N_{1,\alpha},N_{2,\alpha})$ which means
8499: NF$(N_0,N_1,N_2,N_3)$ as promised. So we have proved NF$_{\lambda,\bar
8500: \delta}(N_0,N_1,N_2,N_3) \Rightarrow$ NF$(N_0,N_1,N_2,N_3)$.
8501:
8502: Second, if NF$_\lambda(N_0,N_1,N_2,N_3)$ as defined in Definition
8503: \scite{600-nf.2} then there are
8504: $M_0,M_1,M_2,M_3 \in K_\lambda$ such that
8505: NF$_{\lambda,\langle \lambda,\lambda\rangle}
8506: (M_0,M_1,M_2,M_3),N_\ell \le_{\frak K} M_\ell$ for $\ell <
8507: 4$ and $N_0 = M_0$. By what we have proved above we can conclude
8508: NF$(M_0,M_1,M_2,M_3)$. As $N_0 = M_0 \le_{\frak K} N_\ell \le_{\frak
8509: K} M_\ell$ for $\ell=1,2$ by clause $(c)_1$ of Definition
8510: \scite{600-nf.0X}(1) we get NF$(M_0,N_1,N_2,M_3)$ and by clause $(c)_2$ of
8511: Definition \scite{600-nf.0X}(1) we get NF$(N_0,N_1,N_2,N_3)$. So we have
8512: proved the implication NF$_\lambda(N_0,N_1,N_2,N_3) \Rightarrow$
8513: NF$(N_0,N_1,N_2,N_3)$. Using this, the equivalence
8514: follows by the existence, uniqueness and monotonicity.
8515: \nl
8516: 3) By the rest of this section, i.e., the main conclusion
8517: \scite{600-nf.20.7}, the relation NF$_\lambda$ defined in \scite{600-nf.2} is
8518: a non-forking relation on ${}^4(K_{\frak s})$ respecting ${\frak s}$.
8519: Hence by part (2) of the present claim we have NF$_1 = \text{ NF}_\lambda =
8520: \text{ NF}_2$. \hfill$\square_{\scite{600-nf.0Y}}$
8521: \enddemo
8522: \bn
8523: \margintag{600-nf.0Z}\ub{\stag{600-nf.0Z} Example}: 1) Do we need ${\frak s}$ in \scite{600-nf.0Y}(3)?
8524: Yes.
8525:
8526: Let ${\frak K}$ be the class of graphs and $M \le_{\frak K} N$ iff $M
8527: \subseteq N$; so ${\frak K}$ is an a.e.c. with LS$({\frak K}) =
8528: \aleph_0$. For cardinal $\lambda$ and
8529: $\ell=1,2$ we define NF$^\ell =
8530: \{(M_0,M_1,M_2,M_3):M_0\le_{\frak K} M_1 \le_{\frak K} M_3$ and $M_0
8531: \le_{\frak K} M_2 \le_{\frak K} M_3$ and $M_1 \cap M_2 =M_0$ and if $a
8532: \in M_1 \backslash M_0,b \in M_2 \backslash M_0$ then $\{a,b\}$ is an
8533: edge of $M_3$ iff $\ell=2\}$ and NF$^\ell_\lambda := \{(M_0,M_1,M_2,M_3)
8534: \in \text{ NF}:M_0,M_1,M_2,M_3 \in K_\lambda\}$.
8535: Then NF$^\ell_\lambda$ is a non-forking
8536: relation on ${}^4({\frak K}_\lambda)$ but NF$^1_\lambda \ne$ NF$^2_\lambda$.
8537: \nl
8538: 2) So the assumption on ${\frak K}_\lambda$ that for some good
8539: $\lambda$-frame ${\frak s}$ we have ${\frak K}_{\frak s} =
8540: {\frak K}_\lambda$ is quite a strong demand on ${\frak K}_\lambda$.
8541:
8542: However, the assumption ``respect" essentially is not necessary as it can be
8543: deduced when ${\frak s}$ is good enough.
8544: \bn
8545: Below on ``good$^+$" see \sectioncite[\S1]{705} in particular Definition
8546: \marginbf{!!}{\cprefix{705}.\scite{705-stg.1}}.
8547: \proclaim{\stag{600-nf.0.z.1} Claim} Assume that ${\frak s}$ is a good$^+
8548: \lambda$-frame and {\rm NF} is a non-forking relation on
8549: ${}^4({\frak K}_{\frak s})$. \ub{Then} {\rm NF} respects ${\frak s}$.
8550: \endproclaim
8551: \bigskip
8552:
8553: \remark{Remark} The construction in the proof is similar to the ones
8554: in \scite{600-4a.7}, \scite{600-nf.4}.
8555: \endremark
8556: \bigskip
8557:
8558: \demo{Proof} Assume NF$(M_0,M_1,M_2,M_3)$ and $a \in M_2 \backslash
8559: M_0$, \ortp$_{\frak s}(a,M_0,M_2) \in {\Cal S}^{\text{bs}}_{\frak s}(M_0)$. We
8560: define $(N_{0,i},N_{1,i},f_i)$ for $i < \lambda^+_{\frak s}$ as
8561: follows:
8562: \mr
8563: \item "{$\otimes_1$}" $(a) \quad N_{0,i}$ is $\le_{\frak
8564: s}$-increasing continuous and $N_{0,0} = M_0$
8565: \sn
8566: \item "{${{}}$}" $(b) \quad N_{1,i}$ is $\le_{\frak s}$-increasing
8567: continuous and $N_{1,0} = M_1$
8568: \sn
8569: \item "{${{}}$}" $(c) \quad$
8570: NF$(N_{0,i},N_{1,i},N_{0,i+1},N_{1,i+1})$
8571: \sn
8572: \item "{${{}}$}" $(d) \quad f_i$ is a $\le_{\frak K}$-embedding of
8573: $M_2$ into $N_{0,i+1}$ over $M_0 = N_{0,0}$ such that
8574: \nl
8575:
8576: \hskip25pt \ortp$_{\frak s}(f_i(a),N_{0,i},N_{0,i+1})$
8577: does not fork over $M_0 = N_{0,0}$.
8578: \ermn
8579: We shall choose $f_i$ together with $N_{0,i+1},N_{1,i+1}$.
8580: \nl
8581: Why can we define? For $i=0$ there is nothing to do. For $i$ limit
8582: take unions. For $i=j+1$ choose $f_j,N_{0,i}$ satisfying clause (d)
8583: and $N_{0,j} \le_{\frak s} N_{0,i}$, this is possible for
8584: ${\frak s}$ as we have the existence of
8585: non-forking extensions of \ortp$_{\frak s}(a,M_0,M_2)$ (and
8586: amalgamation).
8587:
8588: Lastly, we take care of the rest
8589: (mainly clause (c) of $\otimes_1$ by clause (f) of Definition \scite{600-nf.0X}(1),
8590: existence). Now
8591: \mr
8592: \item "{$\circledast_2$}" for $i<j<\lambda^+$ we have
8593: NF$(N_{0,i},N_{1,i},N_{0,j},N_{1,j})$
8594: \nl
8595: [why? by transitivity for NF, i.e., clause (e) of Definition
8596: \scite{600-nf.0X}(1), transitivity]
8597: \sn
8598: \item "{$\circledast_3$}" for some $i$, \ortp$_{\frak s}(f_i(a),N_{1,i},
8599: N_{1,i+1})$ does not fork over $M_0$
8600: \nl
8601: [why? by the definition of good$^+$].
8602: \ermn
8603: So for this $i,M_0 \le_{\frak s} f_i(M_2) \le_{\frak s} N_{0,i+1}$ by clause
8604: (d) of $\otimes_1$, hence by clause $(c)_1$ of Definition
8605: \scite{600-nf.0X}, monotonicity
8606: we have NF$(M_0,M_1,f_i(M_2),N_{1,i+1})$. Now again by the choice of
8607: $i$, i.e., by $\circledast_3$ we have
8608: \ortp$_{\frak s}(f_i(a),M_1,N_{1,i+1})$ does
8609: not fork over $M_0$. By clause (g) of Definition \scite{600-nf.0X}(1), i.e.,
8610: uniqueness of NF (and preservation by isomorphisms)
8611: we get \ortp$_{\frak s}(a,M_1,M_3)$ does not fork over $M_0$ as required.
8612: \hfill$\square_{\scite{600-nf.0.z.1}}$
8613: \enddemo
8614: \bn
8615: We turn to our main task in this section proving that such NF exist;
8616: till \scite{600-nf.20.7} we assume:
8617: \demo{\stag{600-nf.0} Hypothesis} 1) ${\frak s} = ({\frak K},\nonfork{}{}_{},
8618: {\Cal S}^{\text{bs}})$ is a good $\lambda$-frame. \nl
8619: 2) ${\frak s}$ is weakly successful which just means that it has
8620: existence for $K^{3,\text{uq}}_\lambda$: for every $M \in
8621: K_\lambda$ and $p \in {\Cal S}^{\text{bs}}(M)$
8622: there are $N,a$ such that $(M,N,a) \in K^{3,\text{uq}}_\lambda$
8623: (see Definition \scite{600-nu.1A}) and $p = \text{ \ortp}(a,M,N)$. (This
8624: follows by $K^{3,\text{uq}}_{\frak s}$ is dense in
8625: $K^{3,\text{bs}}_{\frak s}$; when ${\frak s}$ is categorical, see
8626: \scite{600-nu.13.1}.)
8627:
8628: In this section we deal with models from $K_\lambda$ only.
8629: \enddemo
8630: \bigskip
8631:
8632: \proclaim{\stag{600-nf.0A} Claim} If $M \in K_\lambda$ and $N$ is
8633: $(\lambda,\kappa)$-brimmed over $M$, \ub{then} we can find
8634: $\bar M = \langle M_i:i \le \delta \rangle,\le_{\frak K}$-increasing
8635: continuous, $(M_i,M_{i+1},c_i) \in
8636: K^{3,\text{uq}}_\lambda,M_0 = M,M_\delta = N$ and $\delta$
8637: any pregiven limit ordinal $< \lambda^+$ of cofinality $\kappa$
8638: divisible by $\lambda$.
8639: \endproclaim
8640: \bigskip
8641:
8642: \demo{Proof} Let $\delta$ be given, e.g.,
8643: $\delta = \lambda \times \kappa$. By
8644: \scite{600-nf.0}(2) we can find a $\le_{\frak K}$-increasing sequence $\langle
8645: M_i:i \le \delta \rangle$ of members of $K_\lambda$ and
8646: $\langle a_i:i < \delta \rangle$ such that
8647: $M_0 = M$ and $i < \delta \Rightarrow (M_i,M_{i+1},a_i) \in
8648: K^{3,\text{uq}}_\lambda$ and for every $i < \delta,p \in {\Cal
8649: S}^{\text{bs}}(M_i)$ for $\lambda$ ordinals $j \in (i,i+ \lambda)$ we have
8650: \ortp$(a_j,M_j,M_{j+1})$ is a non-forking extension of $p$.
8651: So the demands in \scite{600-4a.2} hold hence $M_\delta$
8652: is $(\lambda,\kappa)$-brimmed over $M_0=M$. Now we are done by the
8653: uniqueness of $N$ being $(\lambda,\kappa)$-brimmed over $M_0$, see
8654: \scite{600-0.22}(3). \hfill$\square_{\scite{600-nf.0A}}$
8655: \enddemo
8656: \bigskip
8657:
8658: \proclaim{\stag{600-nf.0B} Claim} If
8659: $M^\ell_0 \le_{\frak K} M^\ell_1 \le_{\frak K} M^\ell_3$ and
8660: $M^\ell_0 \le_{\frak K} M^\ell_2 \le_{\frak K} M^\ell_3,c_\ell \in M^\ell_1$
8661: and $(M^\ell_0,M^\ell_1,c_\ell) \in K^{3,\text{uq}}_\lambda$
8662: and ${\text{\rm \ortp\/}}(c_\ell,M^\ell_2,M^\ell_3) \in
8663: {\Cal S}^{\text{bs}}(M^\ell_2)$ does not fork over $M^\ell_0$ and
8664: $M^\ell_3$ is $(\lambda,\sigma)$-brimmed over $M^\ell_1 \cup
8665: M^\ell_2$ all this for
8666: $\ell=1,2$ and $f_i$ is an isomorphism from $M^1_i$ onto $M^2_i$ for
8667: $i=0,1,2$ such that $f_0 \subseteq f_1,f_0 \subseteq f_2$ and $f_1(c_1) =
8668: c_2$, \ub{then} $f_1 \cup
8669: f_2$ can be extended to an isomorphism from $M^1_3$ onto $M^2_3$.
8670: \endproclaim
8671: \bigskip
8672:
8673: \demo{Proof} Chase arrows (and recall definition of
8674: $K^{3,\text{uq}}_\lambda$), that is by \scite{600-nf.0X}(1) and Definition
8675: \scite{600-nf.0X.2}(1) and \scite{600-0.22}(3). \hfill$\square_{\scite{600-nf.0B}}$
8676: \enddemo
8677: \bigskip
8678:
8679: \definition{\stag{600-nf.1} Definition} Assume $\bar \delta = \langle \delta_1,
8680: \delta_2,\delta_3 \rangle,\delta_1,\delta_2,\delta_3$
8681: are ordinals $< \lambda^+$, maybe 1. We say that
8682: NF$_{\lambda,\bar \delta}(N_0,N_1,N_2,N_3)$ or, in other words $N_1,N_2$ are
8683: \ub{brimmedly smoothly amalgamated} in $N_3$ over $N_0$ for
8684: $\bar \delta$ when:
8685: \mr
8686: \item "{$(a)$}" $N_\ell \in K_\lambda$ for $\ell \in \{ 0,1,2,3\}$
8687: \sn
8688: \item "{$(b)$}" $N_0
8689: \le_{\frak K} N_\ell \le_{\frak K} N_3$ for $\ell = 1,2$
8690: \sn
8691: \item "{$(c)$}" $N_1 \cap N_2 = N_0$ (i.e. in disjoint amalgamation,
8692: actually follows by clause (f))
8693: \sn
8694: \item "{$(d)$}" $N_1$ is ($\lambda$,cf$(\delta_1)$)-brimmed over
8695: $N_0$; recall that if
8696: cf$(\delta_1)=1$ this just means $N_0 \le_{\frak K} N_1$
8697: \sn
8698: \item "{$(e)$}" $N_2$ is ($\lambda$,cf$(\delta_2)$)-brimmed over
8699: $N_0$; so that if cf$(\delta_2) =1$ this just
8700: means $N_0 \le_{\frak K} N_2$ and
8701: \sn
8702: \item "{$(f)$}" there are $N_{1,i},N_{2,i}$ for $i \le \lambda \times
8703: \delta_1$ and $c_i$ for $i < \lambda \times \delta_1$
8704: (called witnesses and $\langle N_{1,i},N_{2,i},c_j:i \le \lambda \times
8705: \delta_1,j < \lambda \times \delta_1 \rangle$ is called
8706: a witness sequence as well as $\langle
8707: N_{1,i}:i \le \lambda \times \delta_1 \rangle,\langle N_{2,i}:i \le
8708: \lambda \times \delta_1 \rangle$) such that:
8709: {\roster
8710: \itemitem{ $(\alpha)$ } $N_{1,0} = N_0,N_{1,\lambda \times \delta_1} = N_1$
8711: \sn
8712: \itemitem{ $(\beta)$ } $N_{2,0} = N_2$
8713: \sn
8714: \itemitem{ $(\gamma)$ } $\langle N_{\ell,i}:i \le \lambda \times
8715: \delta_1 \rangle$ is a $\le_{\frak K}$-increasing continuous sequence of
8716: models for $\ell = 1,2$
8717: \sn
8718: \itemitem{ $(\delta)$ } $(N_{1,i},N_{1,i+1},c_i) \in
8719: K^{3,\text{uq}}_\lambda$
8720: \sn
8721: \itemitem{ $(\varepsilon)$ } \ortp$(c_i,N_{2,i},N_{2,i+1})
8722: \in {\Cal S}^{\text{bs}}
8723: (N_{2,i})$ does not fork over $N_{1,i}$ and $N_{2,i} \cap N_1 =
8724: N_{1,i}$, for $i < \lambda \times \delta_1$
8725: (follows by Definition \scite{600-nu.1A})
8726: \sn
8727: \itemitem{ $(\zeta)$ } $N_3$ is $(\lambda$,cf$(\delta_3)$)-brimmed
8728: over $N_{2,\lambda \times \delta_1}$; so for cf$(\delta_3)
8729: = 1$ this means just $N_{2,\lambda \times \delta_1} \le_{\frak K} N_3$
8730: \endroster}
8731: \endroster
8732: \enddefinition
8733: \bigskip
8734:
8735: \definition{\stag{600-nf.2} Definition} 1) We say
8736: $\nonforkin{N_1}{N_2}_{N_0}^{N_3}$ (or
8737: $N_1,N_2$ are \ub{smoothly amalgamated} over $N_0$ inside $N_3$ or
8738: NF$_\lambda(N_0,N_1,N_2,N_3)$ or NF$_{\frak s}(N_0,N_1,N_2,N_3)$)
8739: \ub{when} we can find $M_\ell \in
8740: K_\lambda$ (for $\ell < 4$) such that:
8741: \medskip
8742: \roster
8743: \item "{$(a)$}" NF$_{\lambda,\langle \lambda,\lambda,\lambda \rangle}
8744: (M_0,M_1,M_2,M_3)$
8745: \sn
8746: \item "{$(b)$}" $N_\ell \le_{\frak K} M_\ell$ for $\ell < 4$
8747: \sn
8748: \item "{$(c)$}" $N_0 = M_0$
8749: \sn
8750: \item "{$(d)$}" $M_1,M_2$ are $(\lambda,\text{cf}(\lambda))$-brimmed
8751: over $N_0$ (follows by (a) see clauses (d), (e) of \scite{600-nf.1}).
8752: \ermn
8753: 2) We call $(M,N,a)$ \ub{strongly bs-reduced} if
8754: $(M,N,a) \in K^{3,\text{bs}}_\lambda$
8755: and $(M,N,a) \le_{\text{bs}}
8756: (M',N',a) \in K^{3,\text{bs}}_\lambda \Rightarrow
8757: \text{ NF}_\lambda(M,N,M',N')$; not used.
8758: \enddefinition
8759: \bn
8760: Clearly we expect ``strongly bs-reduced" to be
8761: equivalent to ``$\in K^{3,\text{uq}}_\lambda$", e.g. as this occurs in the
8762: first order case.
8763: We start by proving existence for NF$_{\lambda,\bar \delta}$ from
8764: Definition \scite{600-nf.1}.
8765: \proclaim{\stag{600-nf.3} Claim} 1) Assume
8766: $\bar \delta = \langle \delta_1,\delta_2,
8767: \delta_3 \rangle,\delta_\ell$ an ordinal $< \lambda^+$ and
8768: $N_\ell \in K_\lambda$ for $\ell < 3$ and $N_1$ is
8769: $(\lambda,{\text{\rm cf\/}}(\delta_1)$)-brimmed over $N_0$ and $N_2$ is
8770: $(\lambda,{\text{\rm cf\/}}(\delta_2)$)-brimmed
8771: over $N_0$ and $N_0 \le_{\frak K} N_1$ and $N_0 \le_{\frak K} N_2$
8772: and for simplicity
8773: $N_1 \cap N_2 = N_0$. \underbar{Then} we can find $N_3$ such that
8774: ${\text{\rm NF\/}}_{\lambda,\bar \delta}(N_0,N_1,N_2,N_3)$. \newline
8775: 2) Moreover, we can choose any $\langle N_{1,i}:i \le \lambda \times
8776: \delta_1 \rangle,\langle c_i:i < \lambda \times \delta_1\rangle$
8777: as in \scite{600-nf.1} subclauses (f)$(\alpha),
8778: (\gamma),(\delta)$ as part of the witness.
8779: \nl
8780: 3) If {\rm NF}$_\lambda(N_0,N_1,N_2,N_3)$ then $N_1 \cap N_2 = N_0$.
8781: \endproclaim
8782: \bigskip
8783:
8784: \demo{Proof} 1) We can find $\langle N_{1,i}:i \le \lambda \times
8785: \delta_1 \rangle$ and $\langle c_i:i < \lambda \times \delta_1 \rangle$
8786: as required in part (2) by Claim \scite{600-nf.0A},
8787: the $(\lambda,\text{cf}(\lambda \times \delta_1))$-brimness holds by
8788: \scite{600-4a.2} and apply part (2). \nl
8789: 2) We choose the $N_{2,i}$ (by induction on $i$) by \scite{600-4a.7}
8790: preserving $N_{2,i} \cap N_{1,\lambda \times \delta_2} = N_{1,i}$; in
8791: the successor case use Definition \scite{600-nu.1A} + Claim
8792: \scite{600-nu.2}(3). We then choose $N_3$ using \scite{600-4a.1}(2).
8793: \nl
8794: 3) By the definitions of NF$_\lambda$, NF$_{\lambda,\bar \delta}$.
8795: \hfill$\square_{\scite{600-nf.3}}$
8796: \enddemo
8797: \bn
8798: The following claim tells us that if we have
8799: ``$(\lambda,\text{cf}(\delta_3))$-brimmed" in the end, then we can have
8800: it in all successor stages.
8801: \proclaim{\stag{600-nf.4} Claim} In Definition
8802: \scite{600-nf.1}, if $\delta_3$ is a
8803: limit ordinal and $\kappa = { \text{\rm cf\/}}(\kappa) \ge \aleph_0$,
8804: \ub{then} without loss of generality
8805: (even without changing $\langle N_{1,i}:i \le \lambda \times
8806: \delta_1\rangle,\langle c_i:i < \lambda \times \delta_1\rangle)$
8807: \mr
8808: \item "{$(g)$}" $N_{2,i+1}$ is $(\lambda,\kappa)$-brimmed over
8809: $N_{1,i+1} \cup N_{2,i}$ (which means that it is \newline
8810: $(\lambda,\kappa)$-brimmed over some $N$,
8811: where $N_{1,i+1} \cup N_{2,i} \subseteq N \le_{\frak K} N_{2,i+1}$).
8812: \endroster
8813: \endproclaim
8814: \bigskip
8815:
8816: \demo{Proof} So assume NF$_{\lambda,\bar \delta}(N_0,N_1,N_2,N_3)$ holds
8817: as being witnessed by $\langle N_{\ell,i}:i \le \lambda \times \delta_1
8818: \rangle,\langle c_i:i < \lambda \times \delta_1 \rangle$
8819: for $\ell = 1,2$. Now we choose by induction on $i \le \lambda \times
8820: \delta_1$ a model $M_{2,i} \in K_\lambda$ and $f_i$ such that:
8821: \mr
8822: \widestnumber\item{ (iii) }
8823: \item "{$(i)$}" $f_i$ is a $\le_{\frak K}$-embedding of
8824: $N_{2,i}$ into $M_{2,i}$
8825: \sn
8826: \item "{$(ii)$}" $M_{2,0} = f_i(N_2)$
8827: \sn
8828: \item "{$(iii)$}" $M_{2,i}$ is $\le_{\frak K}$-increasing continuous and
8829: also $f_i$ is increasing continuous
8830: \sn
8831: \item "{$(iv)$}" $M_{2,j} \cap f_i(N_{1,i}) =
8832: f_i(N_{1,j})$ for $j \le i$
8833: \sn
8834: \item "{$(v)$}" $M_{2,i+1}$ is $(\lambda,\kappa)$-brimmed over
8835: $M_{2,i} \cup f_i(N_{2,i+1})$
8836: \sn
8837: \item "{$(vi)$}" \ortp$(f_{i+1}(c_i),M_{2,i},M_{2,i+1}) \in
8838: {\Cal S}^{\text{bs}}(M_{2,i})$ does not fork over $f_i(N_{1,i})$.
8839: \ermn
8840: There is no problem to carry the induction. Using in the successor case
8841: $i=j+1$ the existence Axiom (E)(g) of
8842: Definition \scite{600-1.1} there is a model $M'_{2,i} \in K_{\frak s}$
8843: such that $M_{2,j} \le_{\frak K} M'_{2,i}$ and $f_i
8844: \supseteq f_j$ as required in clauses (i), (iv), (vi) and then use Claim
8845: \scite{600-4a.1} to find a model $M_{2,i} \in K_\lambda$ which is
8846: $(\lambda,\kappa)$-brimmed over $M_{2,j} \cup f_i(N_{2,i})$.
8847:
8848: Having carried the induction, \wilog \, $f_i = \text{ id}_{N_{2,i}}$.
8849: Let $M_3$ be such that
8850: $M_{2,\lambda \times \delta_1} \le_{\frak K} M_3 \in K_\lambda$ and
8851: $M_3$ is $(\lambda$,cf$(\delta_3)$)-brimmed over $M_{2,\lambda \times
8852: \delta_1}$, it exists by \scite{600-4a.1}(2) but $N_{2,\lambda \times
8853: \delta_1} \le_{\frak K} M_{2,\lambda \times \delta_1}$, hence it
8854: follows that $M_3$ is $(\lambda,\kappa)$-brimmed over $N_{1,\lambda
8855: \times \delta_1}$.
8856: So both $M_3$ and $N_3$ are $(\lambda$,cf$(\delta_3)$)-brimmed over
8857: $N_{2,\lambda \times \delta_1}$, hence they are isomorphic over
8858: $N_{2,\lambda \times \delta_1}$ (by \scite{600-0.22}(1)) so
8859: let $f$ be an isomorphism from $M_3$
8860: onto $N_3$ which is the identity over $N_{2,\lambda \times \delta_1}$.
8861: \newline
8862: Clearly $\langle N_{1,i}:i \le \lambda \times \delta_1 \rangle,
8863: \langle f(M_{2,i}):i \le \lambda \times \delta_1 \rangle$ are also
8864: witnesses for \newline
8865: NF$_{\lambda,\bar \delta}(N_0,N_1,N_2,N_3)$ satisfying the extra demand
8866: $(g)$ from \scite{600-nf.4}. \hfill$\square_{\scite{600-nf.4}}$
8867: \enddemo
8868: \bn
8869: The point of the following claim is that having uniqueness in every
8870: atomic step we have uniqueness in the end (using the same ``ladder"
8871: $N_{1,i}$ for now).
8872: \proclaim{\stag{600-nf.5} Claim} (Weak Uniqueness).
8873:
8874: Assume that for $x \in \{ a,b\}$, we have
8875: ${\text{\rm NF\/}}_{\lambda,\bar \delta^x}
8876: (N^x_0,N^x_1,N^x_2,N^x_3)$ holds as witnessed by
8877: $\langle N^x_{1,i}:i \le \lambda \times \delta^x_1 \rangle,
8878: \langle c^x_i:i < \lambda \times \delta^x_1 \rangle,
8879: \langle N^x_{2,i}:i \le \lambda \times \delta^x_1 \rangle$
8880: and $\delta_1 := \delta^a_1 = \delta^b_1,{\text{\rm cf\/}}
8881: (\delta^a_2) = { \text{\rm cf\/}}(\delta^b_2)$ and
8882: ${\text{\rm cf\/}}(\delta^a_3) = { \text{\rm cf\/}}
8883: (\delta^b_3) \ge \aleph_0$. \nl
8884: (Note that ${\text{\rm cf\/}}(\lambda \times \delta^a_1) \ge \aleph_0$
8885: by the definition of {\rm NF}).
8886: \medskip
8887:
8888: Suppose further that $f_\ell$ is an isomorphism from $N^a_\ell$ onto
8889: $N^b_\ell$ for $\ell = 0,1,2$, moreover: $f_0 \subseteq f_1,f_0 \subseteq
8890: f_2$ and $f_1(N^a_{1,i}) = N^b_{1,i},f_1(c^a_i) = c^b_i$.
8891:
8892: \underbar{Then} we can find an isomorphism $f$ from $N^a_3$ onto $N^b_3$
8893: extending $f_1 \cup f_2$.
8894: \endproclaim
8895: \bigskip
8896:
8897: \demo{Proof} Without loss of generality for each $i < \lambda \times
8898: \delta_1$, the model
8899: $N^x_{2,i+1}$ is $(\lambda,\lambda)$-brimmed over
8900: $N^x_{1,i+1} \cup N^x_{2,i}$ (by \scite{600-nf.4},
8901: note there the statement ``without changing the $N_{1,i}$'s"). Now we
8902: choose by induction on $i \le \lambda \times \delta_1$ an isomorphism
8903: $g_i$ from
8904: $N^a_{2,i}$ onto $N^b_{2,i}$ such that: $g_i$ is increasing with $i$ and
8905: $g_i$ extends $(f_1 \restriction N^a_{1,i}) \cup f_2$. \newline
8906: For $i = 0$ choose $g_0 = f_2$ and for $i$ limit let $g_i$ be
8907: $\dsize \bigcup_{j < i} g_j$ and for $i = j+1$ it exists by \scite{600-nf.0B},
8908: whose assumptions hold by $(N^x_{1,i},N^x_{1,i+1},c^x_i)
8909: \in K^{3,\text{uq}}_\lambda$ (see \scite{600-nf.1}, clause (f)$(\delta)$) and
8910: the extra brimness clause from \scite{600-nf.4}. Now by \scite{600-0.22}(3)
8911: we can extend
8912: $g_{\lambda \times \delta_1}$ to an isomorphism from
8913: $N^a_3$ onto $N^b_3$ as $N^x_3$ is $(\lambda,\text{cf}(\delta_3))$-brimmed
8914: over $N^x_{2,\lambda \times \delta_1}$ (for $x \in \{ a,b\}$).
8915: \hfill$\square_{\scite{600-nf.5}}$
8916: \enddemo
8917: \bn
8918: Note that even
8919: knowing \scite{600-nf.5} the choice of $\langle N_{1,i}:i \le \lambda \times
8920: \delta_1 \rangle,\langle c_i:i < \lambda \times \delta_1 \rangle$
8921: still possibly matters. Now we prove an ``inverted"
8922: uniqueness, using our ability to construct a ``rectangle" of models
8923: which is a witness for NF$_{\lambda,\bar \delta}$ in two ways.
8924: \proclaim{\stag{600-nf.6} Claim} Suppose that
8925: \mr
8926: \item "{$(a)$}" for $x \in \{ a,b\}$ we have
8927: ${\text{\rm NF\/}}_{\lambda,\bar \delta^x}(N^x_0,N^x_1,N^x_2,N^x_3)$
8928: \sn
8929: \item "{$(b)$}" $\bar \delta^x = \langle \delta^x_1,\delta^x_2,\delta^x_3
8930: \rangle,\delta^a_1 = \delta^b_2$,
8931: $\delta^a_2 = \delta^b_1,{\text{\rm cf\/}}(\delta^a_3) =
8932: { \text{\rm cf\/}}(\delta^b_3)$, all limit ordinals
8933: \sn
8934: \item "{$(c)$}" $f_0$ is an isomorphism from $N^a_0$ onto $N^b_0$
8935: \sn
8936: \item "{$(d)$}" $f_1$ is an isomorphism from $N^a_1$ onto $N^b_2$
8937: \sn
8938: \item "{$(e)$}" $f_2$ is an isomorphism from $N^a_2$ onto $N^b_1$
8939: \sn
8940: \item "{$(f)$}" $f_0 \subseteq f_1$ and $f_0 \subseteq f_2$.
8941: \endroster
8942: \medskip
8943: \noindent
8944: \underbar{Then} there is an isomorphism from $N^a_3$ onto $N^b_3$ extending
8945: $f_1 \cup f_2$.
8946: \endproclaim
8947: \bn
8948: Before proving we shall construct a third ``rectangle" of models such
8949: that we shall be able to construct appropriate isomorphisms each of
8950: $N^a_3,N^b_3$
8951: \proclaim{\stag{600-nf.6A} Subclaim} Assume
8952: \mr
8953: \item "{$(a)$}" $\delta^a_1,\delta^a_2,\delta^a_3 < \lambda^+$ are limit
8954: ordinals
8955: \sn
8956: \item "{$(b)_1$}" $\bar M^1 = \langle M^1_\alpha:\alpha \le \lambda \times
8957: \delta^a_1 \rangle$ is $\le_{\frak K}$-increasing continuous in $K_\lambda$
8958: \nl
8959: and $(M^1_\alpha,M^1_{\alpha +1},c_\alpha) \in K^{3,\text{bs}}_\lambda$
8960: \sn
8961: \item "{$(b)_2$}" $\bar M^2 = \langle M^2_\alpha:\alpha \le \lambda \times
8962: \delta^a_2 \rangle$ is $\le_{\frak K}$-increasing continuous in
8963: $K_\lambda$ and $(M^2_\alpha,M^2_{\alpha +1},d_\alpha)
8964: \in K^{3,\text{bs}}_\lambda$
8965: \sn
8966: \item "{$(c)$}" $M^1_0 = M^2_0$
8967: we call it $M$ and $M^1_\alpha \cap M^2_\beta = M$ for $\alpha \le
8968: \lambda \times \delta^a_1,\beta \le \lambda \times \delta^a_2$.
8969: \ermn
8970: \ub{Then} we can find $M_{i,j}$ (for $i \le \lambda \times
8971: \delta^a_1 \text{ and } j \le \lambda \times \delta^a_2)$
8972: and $M_3$ such that:
8973: \mr
8974: \item "{$(A)$}" $M_{i,j} \in K_\lambda$ and $M_{0,0} = M$ and
8975: $M_{i,0} = M^1_i,M_{0,j} = M^2_j$
8976: \sn
8977: \item "{$(B)$}" $i_1 \le i_2 \and j_1 \le j_2 \Rightarrow M_{i_1,j_1}
8978: \le_{\frak K} M_{i_2,j_2}$
8979: \sn
8980: \item "{$(C)$}" if $i \le \lambda \times \delta^a_1$ is a limit
8981: ordinal and $j \le \lambda \times \delta^a_2$ \ub{then} $M_{i,j} =
8982: \dsize \bigcup_{\zeta < i} M_{\zeta,j}$
8983: \sn
8984: \item "{$(D)$}" if $i \le \lambda \times \delta^a_1$ and $j \le \lambda
8985: \times \delta^a_2$ is a limit ordinal \ub{then} $M_{i,j} =
8986: \dsize \bigcup_{\xi < j} M_{i,\xi}$
8987: \sn
8988: \item "{$(E)$}" $M_{\lambda \times \delta^a_1,j+1}$ is
8989: $(\lambda,{\text{\rm cf\/}}
8990: (\delta^a_1))$-brimmed over $M^a_{\lambda \times \delta^a_1,j}$
8991: for $j < \lambda \times \delta^a_2$
8992: \sn
8993: \item "{$(F)$}" $M_{i+1,\lambda \times \delta^a_2}$ is
8994: $(\lambda,{\text{\rm cf\/}}(\delta^a_2))$-brimmed over
8995: $M_{i,\lambda \times \delta^a_2}$ for $i < \lambda \times \delta^a_1$
8996: \sn
8997: \item "{$(G)$}" $M_{\lambda \times \delta^a_1,\lambda \times \delta^a_2}
8998: \le_{\frak K} M_3 \in K_\lambda$ moreover \newline
8999: $M_3$ is $(\lambda,{\text{\rm cf\/}}(\delta^a_3))$-brimmed over
9000: $M_{\lambda \times \delta^a_1,\lambda \times \delta^a_2}$
9001: \sn
9002: \item "{$(H)$}" for $i < \lambda \times \delta^a_1,j \le \lambda
9003: \times \delta^a_2$ we have
9004: ${\text{\rm \ortp\/}}(c_i,M_{i,j},M_{i+1,j})$ does not fork over $M_{i,0}$
9005: \sn
9006: \item "{$(I)$}" for $j < \lambda \times \delta^a_2,i \le \lambda
9007: \times \delta^a_1$ we have
9008: ${\text{\rm \ortp\/}}(d_j,M_{i,j},M_{i,j+1})$ does not fork over $M_{0,j}$.
9009: \ermn
9010: We can add
9011: \mr
9012: \item "{$(J)$}" for $i < \lambda \times \delta^a_1,j < \lambda \times
9013: \delta^b_2$ the model
9014: $M_{i+1,j+1}$ is $(\lambda,*)$-brimmed over $M_{i,j+1} \cup
9015: M_{i+1,j}$.
9016: \endroster
9017: \endproclaim
9018: \bigskip
9019:
9020: \remark{Remark} 1) We can replace in \scite{600-nf.6A} the ordinals $\lambda
9021: \times \delta^a_\ell \, (\ell =1,2,3)$ by any ordinal $\alpha^a_\ell <
9022: \lambda^+$ (for $\ell = 1,2,3$) we use the
9023: present notation just to conform with its use in
9024: the proof of \scite{600-nf.6}.
9025: \nl
9026: 2) Why do we need $u^\ell_1$ in the proof below? This is used to get
9027: the brimmness demands in \scite{600-nf.6A}.
9028: \endremark
9029: \bigskip
9030:
9031: \demo{Proof} We first change our towers, repeating models to give
9032: space for bookkeeping.
9033: That is we define ${}^*M^1_\alpha$ for $\alpha \le \lambda \times \lambda
9034: \times \delta^a_1$ as follows:
9035: \sn
9036:
9037: if $\lambda \times \beta < \alpha \le \lambda \times
9038: \beta + \lambda$ and $\beta < \lambda \times \delta^a_1$ then
9039: ${}^* M^1_\alpha = M^1_{\beta + 1}$
9040: \sn
9041:
9042: if $\alpha = \lambda \times \beta$, then ${}^*M^1_\alpha = M^1_\beta$.
9043: \mn
9044: Let $u^1_0 = \{\lambda \beta:\beta < \delta^a_1\},u^1_1 = \lambda
9045: \times \lambda \times
9046: \delta^a_1 \backslash u^1_0,u^1_2 = \emptyset$ and for $\alpha = \lambda
9047: \beta \in u^1_0$ let $a^1_\alpha = c_\beta$.
9048:
9049: Similarly let us define ${}^*M^2_\alpha$ (for $\alpha \le \lambda
9050: \times \lambda \times \delta^a_2$),$u^2_0,u^2_1,u^2_2$ and
9051: $\langle a^2_\alpha:\alpha \in u^2_0
9052: \rangle$.
9053:
9054: Now apply \scite{600-4a.9} (check) and get ${}^*M_{i,j},
9055: (i \le \lambda \times \lambda \times
9056: \delta^a_1,j \le \lambda \times
9057: \lambda \times \delta^a_2)$. Lastly, for $i \le
9058: \delta^a_1,j \le \delta^a_2$ let $M_{i,j} = {}^*M_{\lambda \times
9059: i,\lambda \times j}$.
9060: By \scite{600-4a.2} clearly ${}^* M_{\lambda \times i + \lambda,
9061: \lambda \times j+ \lambda}$ is
9062: $(\lambda,\text{cf}(\lambda))$-brimmed over
9063: ${}^*M_{\lambda \times i+1,\lambda \times
9064: j+1}$ hence $M_{i+1,j+1}$ is $(\lambda,\text{cf}(\lambda))$-brimmed over
9065: $M_{i+1,j} \cup M_{i,j+1}$.
9066: And, by \scite{600-4a.1}(1) choose $M_3 \in K_\lambda$ which is
9067: $(\lambda,\text{cf}(\delta^a_3))$-brimmed over
9068: $M_{\lambda \times \delta^a_1,\lambda \times \delta^a_2}$.
9069: \hfill$\square_{\scite{600-nf.6A}}$
9070: \enddemo
9071: \bigskip
9072:
9073: \demo{Proof of \scite{600-nf.6}} We shall let $M_{i,j},M_3$ be as in
9074: \scite{600-nf.6A} for $\bar \delta^a$ and $\bar M^1,\bar M^2$ determined below.
9075: For $x \in \{ a,b\}$ as
9076: NF$_{\lambda,\bar \delta^x}(N^x_0,N^x_1,N^x_2,N^x_3)$, we know that
9077: there are witnesses $\langle N^x_{1,i}:i \le \lambda \times
9078: \delta^x_1 \rangle,\langle c^x_i:i < \lambda \times \delta^x_1\rangle,
9079: \langle N^x_{2,i}:i \le \lambda \times \delta^x_1
9080: \rangle$ for this. So $\langle N^x_{1,i}:i \le \lambda \times
9081: \delta^x_1 \rangle$ is $\le_{\frak K}$-increasing continuous
9082: and $(N^x_{1,i},N^x_{1,i+1},c^x_i) \in K^{3,\text{uq}}_\lambda$ for $i
9083: < \lambda \times \delta^x_1$. Hence by the freedom we
9084: have in choosing $\bar M^1$ and $\langle c_i:i < \lambda \times
9085: \delta_1\rangle$ without loss of generality there is an
9086: isomorphism $g_1$ from $N^a_{1,\lambda \times \delta^a_1}$ onto
9087: $M_{\lambda \times \delta^a_1}$ mapping $N^a_{1,i}$ onto $M^1_i
9088: = M_{i,0}$ and $c^a_i$ to $c_i$; remember that
9089: $N^a_{1,\lambda \times \delta^a_1} = N^a_1$.
9090: Let $g_0 = g_1 \restriction N^a_0 = g_1 \restriction N^a_{1,0}$ so
9091: $g_0 \circ f^{-1}_0$ is an isomorphism from $N^b_0$ onto $M_{0,0}$.
9092: \medskip
9093:
9094: Similarly as $\delta^b_1 = \delta^a_2$,
9095: and using the freedom we have in choosing $\bar M^2$ and $\langle
9096: d_i:i < \lambda \times \delta^b_1\rangle$
9097: without loss of generality there is an isomorphism $g_2$ from
9098: $N^b_{1,\lambda \times \delta^a_2}$ onto $M^2_j = M_{0,\lambda \times
9099: \delta^a_2}$ mapping $N^b_{1,j}$ onto $M_{0,j}$
9100: (for $j \le \lambda \times \delta^a_2)$ and mapping $c^b_i$ to $d_i$
9101: and $g_2$ extends $g_0 \circ f^{-1}_0$. \newline
9102: Now would like to use the weak uniqueness \scite{600-nf.5} and for this note:
9103: \mr
9104: \item "{$(\alpha)$}" NF$_{\lambda,\bar \delta^a}(N^a_0,N^a_1,N^a_2,N^a_3)$
9105: is witnessed by the sequences
9106: $\langle N^a_{1,i}:i \le \lambda \times \delta^a_1 \rangle$, and
9107: $\langle N^a_{2,i}:i \le \lambda \times \delta^a_1 \rangle$ \newline
9108: [why? an assumption]
9109: \sn
9110: \item "{$(\beta)$}" NF$_{\lambda,\bar \delta^a}(M_{0,0},M_{\lambda \times
9111: \delta^a_1,0},M_{0,\lambda \times \delta^a_2},M_3)$
9112: is witnessed by the sequences \newline
9113: $\langle M_{i,0}:i \le \lambda \times \delta^a_1 \rangle,\langle
9114: M_{i,\lambda \times \delta^a_2}:i \le \lambda \times \delta^a_1
9115: \rangle$ \newline
9116: [why? check]
9117: \sn
9118: \item "{$(\gamma)$}" $g_0$ is an isomorphism from $N^a_0$ onto $M_{0,0}$
9119: \newline
9120: [why? see its choice]
9121: \sn
9122: \item "{$(\delta)$}" $g_1$ is an isomorphism from $N^a_1$ onto
9123: $M_{\lambda \times \delta^a_1,0}$ mapping $N^a_{1,i}$ onto
9124: $M_{i,0}$ for $i < \lambda \times \delta^a_1$ and $c^a_i$ to $c_i$
9125: for $i < \lambda \times \delta^a_1$ and extending $g_0$ \newline
9126: [why? see the choice of $g_1$ and of $g_0$]
9127: \sn
9128: \item "{$(\varepsilon)$}" $g_2 \circ f_2$ is an isomorphism from
9129: $N^a_2$ onto $M_{0,\lambda \times \delta^a_2}$ extending $g_0$ \newline
9130: [why? $f_2$ is an isomorphism from $N^a_2$ onto $N^b_1$ and $g_2$ is an
9131: isomorphism from $N^b_1$ onto $M_{0,\lambda \times \delta^a_1}$
9132: extending $g_0 \circ f^{-1}_0$ and $f_0 \subseteq f_2$].
9133: \endroster
9134: \medskip
9135:
9136: So there is by \scite{600-nf.5} an isomorphism $g^a_3$ from $N^a_3$ onto
9137: $M_3$ extending both $g_1$ and $g_2 \circ f_2$.
9138: \medskip
9139:
9140: We next would like to apply \scite{600-nf.5} to the $N^b_i$'s; so note:
9141: \mr
9142: \item "{$(\alpha)'$}" NF$_{\lambda,\bar \delta^b}(N^b_0,N^b_1,N^b_2,N^b_3)$
9143: is witnessed by the sequences
9144: $\langle N^b_{1,i}:i \le \lambda \times \delta^a_2
9145: \rangle$, \newline
9146: $\langle N^b_{2,i}:i \le \lambda \times \delta^a_2 \rangle$
9147: \sn
9148: \item "{$(\beta)'$}" NF$_{\lambda,\bar \delta^b}(M_{0,0},M_{0,\lambda
9149: \times \delta^a_2},M_{\lambda \times \delta^a_1,0},M_3)$ is witnessed by the
9150: sequences \newline
9151: $\langle M_{0,j}:j \le \lambda \times \delta^a_2 \rangle,
9152: \langle
9153: M_{\lambda \times \delta^a_1,j}:j \le \lambda \times \delta^a_2 \rangle$
9154: \sn
9155: \item "{$(\gamma)'$}" $g_0 \circ (f_0)^{-1}$ is an
9156: isomorphism from $N^b_0$ onto $M_{0,0}$ \newline
9157: [why? Check.]
9158: \sn
9159: \item "{$(\delta)'$}" $g_2$ is an isomorphism from $N^b_1$ onto
9160: $M_{0,\lambda \times \delta^a_2}$ mapping $N^b_{1,j}$ onto $M_{0,j}$
9161: and $c^a_j$ to $d_j$ for
9162: $j \le \lambda \times \delta^a_2$ and extending $g_0 \circ (f_2)^{-1}$
9163: \newline
9164: [why? see the choice of $g_2$: it maps $N^b_{1,j}$ onto $M_{0,j}$]
9165: \sn
9166: \item "{$(\varepsilon)'$}" $g_1 \circ (f_1)^{-1}$ is an isomorphism from
9167: $N^b_2$ onto $M_{\lambda \times \delta^a_0}$ extending $g_0$ \newline
9168: [why? remember $f_1$ is an isomorphism from $N^a_1$ onto $N^b_2$ extending
9169: $f_0$ and the choice of $g_1$: it maps $N^a_1$
9170: onto $M_{\lambda \times \delta^a_1,0}$].
9171: \endroster
9172: \medskip
9173:
9174: \noindent
9175: So there is an isomorphism $g^b_3$ form $N^b_3$ onto $M_3$ extending
9176: $g_2$ and $g_1 \circ (f_1)^{-1}$. \newline
9177: Lastly $(g^b_3)^{-1} \circ g^a_3$ is an isomorphism from $N^a_3$ onto
9178: $N^b_3$ (chase arrows). Also
9179:
9180: $$
9181: \align
9182: ((g^b_3)^{-1} \circ g^a_3) \restriction N^a_1 &= (g^b_3)^{-1}(g^a_3
9183: \restriction N^a_1) \\
9184: &= (g^b_3)^{-1} g_1 = ((g^b_3)^{-1} \restriction M_{\lambda \times
9185: \delta^a_1,0}) \circ g_1 \\
9186: &= (g^b_3 \restriction N^b_2)^{-1} \circ g_1 = ((g_1 \circ (f_1)^{-1})^{-1})
9187: \circ g_1 \\
9188: &= (f_1 \circ (g_1)^{-1}) \circ g_1 = f_1.
9189: \endalign
9190: $$
9191: \medskip
9192:
9193: \noindent
9194: Similarly $((g^b_3)^{-1} \circ g^a_3) \restriction N^a_2 = f_2$.
9195: \newline
9196: So we have finished. \hfill$\square_{\scite{600-nf.6}}$
9197: \enddemo
9198: \bn
9199: But if we invert twice we get straight; so
9200: \proclaim{\stag{600-nf.7} Claim} [Uniqueness]. Assume for
9201: $x \in \{ a,b\}$ we have \newline
9202: ${\text{\rm NF\/}}_{\lambda,\bar \delta^x}(N^x_0,N^x_1,N^x_2,N^x_3)$ and
9203: ${\text{\rm cf\/}}(\delta^a_1) =
9204: { \text{\rm cf\/}}(\delta^b_1),{\text{\rm cf\/}}(\delta^a_2)
9205: = { \text{\rm cf\/}}(\delta^b_2),{\text{\rm cf\/}}(\delta^a_3)
9206: = { \text{\rm cf\/}}(\delta^b_3)$, all
9207: $\delta^x_\ell$ limit ordinals $< \lambda^+$.
9208:
9209: If $f_\ell$ is an isomorphism from $N^a_\ell$ onto $N^b_\ell$ for $\ell < 3$
9210: and $f_0 \subseteq f_1,f_0 \subseteq f_2$ \underbar{then} there is an
9211: isomorphism $f$ from $N^a_3$ onto $N^b_3$ extending $f_1,f_2$.
9212: \endproclaim
9213: \bigskip
9214:
9215: \demo{Proof} Let $\bar \delta^c = \langle \delta^c_1,\delta^c_2,
9216: \delta^c_3 \rangle = \langle \delta^a_2,\delta^a_1,\delta^a_3 \rangle$;
9217: by \scite{600-nf.3}(1) there are
9218: $N^c_\ell$ (for $\ell \le 3$) such that NF$_{\lambda,\bar \delta^c}(N^c_0,
9219: N^c_1,N^c_2,N^c_3)$ and $N^c_0 \cong N^a_0$.
9220: There is for $x \in \{ a,b\}$ an isomorphism $g^x_0$
9221: from $N^x_0$ onto $N^c_0$ and without loss of generality
9222: $g^a_0 = g^b_0 \circ f_0$. Similarly
9223: for $x \in \{a,b\}$
9224: there is an isomorphism $g^x_1$ from
9225: $N^x_1$ onto $N^c_2$ extending $g^x_0$ (as $N^x_1$ is
9226: $(\lambda,\text{cf}(\delta^x_1))$-brimmed over $N^x_0$ and also
9227: $N^c_2$ is $(\lambda,\text{cf}(\delta^c_2))$-brimmed over $N^c_0$ and
9228: $\text{cf}(\delta^c_2) = \text{cf}(\delta^a_1) = \text{cf}(\delta^x_1))$
9229: and without loss of generality $g^b_1 = g^a_1 \circ f_1$. Similarly for
9230: $x \in \{a,b\}$
9231: there is an isomorphism $g^x_2$ from $N^x_2$ onto $N^c_1$ extending $g^x_0$
9232: (as $N^x_2$ is
9233: $(\lambda,\text{cf}(\delta^x_2))$-brimmed over $N^x_0$ and also $N^c_1$
9234: is $(\lambda,\text{cf}(\delta^c_1))$-brimmed over $N^c_0$ and
9235: $\text{cf}(\delta^c_1) = \text{cf}(\delta^a_2) = \text{cf}(\delta^x_2))$
9236: and without loss of generality $g^a_2 = g^b_2 \circ f_2$.
9237: \newline
9238: So by \scite{600-nf.6} for $x \in \{a,b\}$ there is an isomorphism
9239: $g^x_3$ from $N^x_3$ onto $N^c_3$
9240: extending $g^x_1$ and $g^x_2$.
9241: Now $(g^b_3)^{-1} \circ g^a_3$ is an isomorphism
9242: from $N^a_3$ onto $N^b_3$ extending $f_1,f_2$ as required.
9243: \hfill$\square_{\scite{600-nf.7}}$
9244: \enddemo
9245: \bn
9246: So we have proved the uniqueness for NF$_{\lambda,\bar \delta}$ when all
9247: $\delta_\ell$ are limit ordinals; this means that the arbitrary choice of
9248: $\langle N_{1,i}:i \le \lambda \times \delta_1 \rangle$ and
9249: $\langle c_i:i < \lambda \times \delta_1 \rangle$ is immaterial; it
9250: figures in the definition and, e.g. existence proof
9251: but does not influence the net
9252: result. The
9253: power of this result is illustrated in the following conclusion.
9254: \demo{\stag{600-nf.8} Conclusion} [Symmetry]. \newline
9255:
9256: If
9257: NF$_{\lambda,\langle \delta_1,\delta_2,\delta_3 \rangle}(N_0,N_1,N_2,N_3)$
9258: where $\delta_1,\delta_2,\delta_3$ are limit ordinals $< \lambda^+$
9259: \underbar{then} NF$_{\lambda,\langle \delta_2,\delta_1,
9260: \delta_3 \rangle}(N_0,N_2,N_1,N_3)$.
9261: \enddemo
9262: \bigskip
9263:
9264: \demo{Proof} By \scite{600-nf.6A} we can find $N'_\ell (\ell \le 3)$
9265: such that: $N'_0 = N_0,N'_1$ is
9266: $(\lambda,\text{cf}(\delta_1))$-brimmed over $N'_0,N'_2$ is
9267: $(\lambda,\text{cf}(\delta_2))$-brimmed over $N'_0$ and $N'_3$ is
9268: $(\lambda,\text{cf}(\delta_3))$-brimmed over $N'_1 \cup N'_2$ and
9269: NF$_{\lambda,\langle \delta_1,\delta_2,\delta_3
9270: \rangle}(N'_0,N'_1,N'_2,N'_3)$ and NF$_{\lambda,\langle
9271: \delta_2,\delta_1,\delta_3 \rangle}(N'_0,N'_2,N'_1,N'_3)$. Let
9272: $f_1,f_2$ be an isomorphism from $N_1,N_2$ onto $N'_1,N'_2$ over
9273: $N_0$, respectively. By \scite{600-nf.7} (or \scite{600-nf.6})
9274: there is an isomorphism $f'_3$
9275: form $N_3$ onto $N'_3$ extending $f_1 \cup f_2$. As isomorphisms
9276: preserve NF we are done. \hfill$\square_{\scite{600-nf.8}}$
9277: \enddemo
9278: \bigskip
9279:
9280: Now we turn to smooth amalgamation (not necessarily brimmed, see
9281: Definition \scite{600-nf.2}). If we use Lemma \scite{600-4a.6}, of course, we
9282: do not really need \scite{600-nf.10}.
9283: \proclaim{\stag{600-nf.10} Claim} 1) If
9284: ${\text{\rm NF\/}}_{\lambda,\bar \delta}(N_0,N_1,N_2,N_3)$ and
9285: $\delta_1,\delta_2,\delta_3$ are limit ordinals,
9286: \underbar{then} ${\text{\rm NF\/}}_\lambda
9287: (N_0,N_1,N_2,N_3)$ (see Definition \scite{600-nf.2}). \newline
9288: 2) In Definition \scite{600-nf.2}(1) we can add:
9289: \mr
9290: \item "{$(d)^+$}" $M_\ell$ is $(\lambda,{\text{\rm cf\/}}(\lambda))$-brimmed
9291: over $N_0$ and moreover over $N_\ell$,
9292: \sn
9293: \item "{$(e)$}" $M_3$ is $(\lambda,{\text{\rm cf\/}}(\lambda))$-brimmed
9294: over $M_1 \cup M_2$ (actually this is given by clause $(f)(\zeta)$ of
9295: Definition \scite{600-nf.1}).
9296: \ermn
9297: 3) If $N_0 \le_{\frak K} N_\ell$ for $\ell=1,2$ and $N_1 \cap N_2 =
9298: N_0$, \ub{then} we can find
9299: $N_3$ such that {\rm NF}$_\lambda(N_0,N_1,N_2,N_3)$.
9300: \endproclaim
9301: \bigskip
9302:
9303: \demo{Proof} 1) Note that even if every $\delta_\ell$ is limit and we
9304: waive the ``moreover" in clause $(d)^+$,
9305: the problem is in the case that e.g.
9306: $(\text{cf}(\delta^a),\text{cf}(\delta^b),\text{cf}(\delta^c)) \ne
9307: (\text{cf}(\lambda),\text{cf}(\lambda),\text{cf}(\lambda))$.
9308: For $\ell=1,2$ we can find $\bar M^\ell = \langle M^\ell_i:i \le
9309: \lambda \times (\delta_\ell + \lambda) \rangle$ and $\langle
9310: c^\ell_i:i < \lambda \times (\delta_i + \lambda)\rangle$ such that
9311: $M^\ell_0 = N_0,\bar M^1$ is $\le_{\frak K}$-increasing continuous
9312: $(M^\ell_i,M^\ell_{i+1},c_i) \in K^{3,\text{uq}}_{\frak s}$ and if $p
9313: \in {\Cal S}^{\text{bs}}(M^\ell_i)$ and $i < \lambda \times (\delta_\ell +
9314: \lambda)$ then for $\lambda$ ordinals $j < \lambda$,
9315: \ortp$(c_i,M^\ell_{i+j},M^\ell_{i+j+1})$ is a non-forking extension of
9316: $p$. So $M^\ell_{\lambda \times \delta_\ell}$ is
9317: $(\lambda,\text{cf}(\delta_\ell))$-brimmed over $M^\ell_0 = N_0$ and
9318: $M^\ell_{\lambda \times (\delta_\ell + \lambda)}$ is
9319: $(\lambda,\text{cf}(\lambda))$-brimmed over $M^\ell_{\lambda \times
9320: \delta_\ell}$; so \wilog \, $M^\ell_{\lambda \times \delta_\ell} =
9321: N_\ell$ for $\ell=1,2$. \nl
9322: By \scite{600-nf.6A} we can find $M_{i,j}$ for
9323: $i \le \lambda \times (\delta_1 + \lambda),j \le \lambda \times
9324: (\delta_2 + \lambda)$ for $\bar \delta' := \langle \delta_1 + \lambda,
9325: \delta_2 + \lambda,\delta_3 \rangle$ such that they are as in
9326: \scite{600-nf.6A} for $\bar M^1,\bar M^2$ so
9327: $M_{0,0} = N_0$; then choose $M'_3 \in K_\lambda$ which
9328: is $(\lambda,\text{cf}(\delta_3))$-brimmed over $M_{\lambda \times
9329: \delta_1,\lambda \times \delta_2}$. So
9330: NF$_{\lambda,\bar \delta}(M_{0,0},M_{\lambda \times
9331: \delta_1,0},M_{0,\lambda \times \delta_2},M'_3)$, hence by \scite{600-nf.7}
9332: without loss of generality $M_{0,0} = N_0,M_{\lambda \times \delta_1,0}
9333: = N_1,M_{0,\lambda \times \delta_2} = N_2$, and $N_3 = M'_3$.
9334: Lastly, let $M_3$
9335: be $(\lambda,\text{cf}(\lambda))$-brimmed over $M'_3$.
9336: Now clearly also \newline
9337: NF$_{\lambda,\langle \delta_1 + \lambda,\delta_2 + \lambda,\delta_3 +
9338: \lambda \rangle}(M_{0,0},M_{\lambda \times (\delta_1 + \lambda),0},
9339: M_{0,\lambda \times (\delta_2 + \lambda)},M_3)$ and \newline
9340: $N_0 = M_{0,0},N_1 = M_{\lambda \times \delta_2,0} \le_{\frak K}
9341: M_{\lambda \times (\delta_2 + \lambda),0},
9342: N_2 = M_{0,\lambda \times \delta_2} \le_{\frak K}
9343: M_{0,\lambda \times (\delta_2 + \lambda)}$ \newline
9344: and $M_{\lambda \times (\delta_1 + \lambda),0}$ is
9345: $(\lambda,\text{cf}(\lambda))$-brimmed over
9346: $M_{\lambda \times \delta_1,0}$ and
9347: $M_{0,\lambda \times (\delta_2 + \lambda)}$ is \newline
9348: $(\lambda,\text{cf}(\lambda))$-brimmed over
9349: $M_{0,\lambda \times \delta_2}$
9350: and $N_3 = M'_3 \le_{\frak K} M_3$. So
9351: we get all the requirements for
9352: NF$_\lambda(N_0,N_1,N_2,N_3)$ (as witnessed
9353: by $\langle M_{0,0},M_{\lambda \times (\delta_1 + \lambda),0},
9354: M_{0,\lambda \times (\delta_2 + \lambda)},M_3 \rangle$).
9355: 2) Similar proof. \nl
9356: 3) By \scite{600-nf.3} and the proof above. \hfill$\square_{\scite{600-nf.10}}$
9357: \enddemo
9358: \bn
9359: Now we turn to NF$_\lambda$; existence is easy.
9360: \proclaim{\stag{600-nf.10.3} Claim} {\rm NF}$_\lambda$ has existence,
9361: i.e., clause (f) of \scite{600-nf.0X}(1).
9362: \endproclaim
9363: \bigskip
9364:
9365: \demo{Proof} By \scite{600-nf.10}(3). \hfill$\square_{\scite{600-nf.10.3}}$
9366: \enddemo
9367: \bn
9368: Next we deal with real uniqueness
9369: \proclaim{\stag{600-nf.11} Claim} [Uniqueness of smooth amalgamation]: \nl
9370: 1) If
9371: ${\text{\rm NF\/}}_\lambda(N^x_0,N^x_1,
9372: N^x_2,N^x_3)$ for $x \in \{ a,b\},f_\ell$ an
9373: isomorphism from $N^a_\ell$ onto $N^b_\ell$ for $\ell < 3$ and
9374: $f_0
9375: \subseteq f_1,f_0 \subseteq f_2$ \underbar{then} $f_1 \cup f_2$ can be
9376: extended to a $\le_{\frak K}$-embedding of $N^a_3$ into some
9377: $\le_{\frak K}$-extension of $N^b_3$. \nl
9378: 2) So if above $N^x_3$ is $(\lambda,\kappa)$-brimmed
9379: over $N^x_1 \cup N^x_2$ for $x = a,b$, we can
9380: extend $f_1 \cup f_2$ to an isomorphism from $N^a_3$ onto $N^b_3$.
9381: \endproclaim
9382: \bigskip
9383:
9384: \demo{Proof} 1) For $x \in \{ a,b\}$ let the sequence $\langle M^x_\ell:
9385: \ell < 4 \rangle$ be a witness to \newline
9386: NF$_\lambda(N^x_0,N^x_1,N^x_2,N^x_3)$ as in \scite{600-nf.2}, \scite{600-nf.10}(2),
9387: so in particular \newline
9388: NF$_{\lambda,\langle \lambda,\lambda,\lambda \rangle}
9389: (M^x_0,M^x_1,M^x_2,M^x_3)$. By chasing arrows (disjointness) and
9390: uniqueness, i.e.
9391: \scite{600-nf.7} without loss of generality $M^a_\ell = M^b_\ell$ for $\ell < 4$
9392: and $f_0 = \text{ id}_{N^a_0}$. As $M^a_1$ is
9393: $(\lambda,\text{cf}(\lambda))$-brimmed over $N^a_1$ and also
9394: over $N^b_1$ (by clause $(d)^+$ of \scite{600-nf.10}(2))
9395: and $f_1$ is an isomorphism from $N^a_1$ onto $N^b_1$, clearly
9396: by \scite{600-0.22} there is an automorphism $g_1$ of $M^a_1$ such that $f_1 \subseteq g_1$,
9397: hence also $\text{id}_{N^a_0} = f_0 \subseteq f_1 \subseteq g_1$.
9398: Similarly there is an
9399: automorphism $g_2$ of $M^a_2$ extending $f_2$ hence $f_0$. So $g_\ell \in
9400: \text{ AUT}(M^a_\ell)$ for $\ell = 1,2$ and
9401: $g_1 \restriction M^a_0 = f_0 =
9402: g_2 \restriction M^a_0$. By the uniqueness of NF$_{\lambda,\langle
9403: \lambda,\lambda,\lambda \rangle}$ (i.e. Claim \scite{600-nf.7}) there is an
9404: automorphism $g_3$ of $M^a_3$
9405: extending $g_1 \cup g_2$. This proves the desired conclusion.
9406: \nl
9407: 2) Should be clear. \hfill$\square_{\scite{600-nf.11}}$
9408: \enddemo
9409: \bn
9410: We now show that in the cases the two notions of non-forking
9411: amalgamations are meaningful then they coincide, one implication
9412: already is a case of \scite{600-nf.10}.
9413: \proclaim{\stag{600-nf.12} Claim} Assume
9414: \mr
9415: \item "{$(a)$}" $\bar \delta = \langle \delta_1,\delta_2,\delta_3 \rangle,
9416: \delta_\ell < \lambda^+$ is a limit ordinal for $\ell = 1,2,3$; \newline
9417: $N_0 \le_{\frak K} N_\ell \le_{\frak K} N_3$ are in $K_\lambda$ for
9418: $\ell = 1,2$
9419: \sn
9420: \item "{$(b)$}" $N_\ell$ is $(\lambda,{\text{\rm cf\/}}
9421: (\delta_\ell))$-brimmed over $N_0$ for $\ell =1,2$
9422: \sn
9423: \item "{$(c)$}" $N_3$ is ${\text{\rm cf\/}}(\delta_3)$-brimmed
9424: over $N_1 \cup N_2$.
9425: \ermn
9426: \ub{Then} ${\text{\rm NF\/}}_\lambda(N_0,N_1,N_2,N_3)$ iff
9427: {\rm NF}$_{\lambda,\bar \delta}(N_0,N_1,N_2,N_3)$.
9428: \endproclaim
9429: \bigskip
9430:
9431: \demo{Proof} The ``if" direction holds by \scite{600-nf.10}(1).
9432: As for the ``only if" direction, basically it follows from the existence for
9433: NF$_{\lambda,\bar \delta}$ and uniqueness for NF$_\lambda$; in details
9434: by the proof of \scite{600-nf.10}(1) (and Definition \scite{600-nf.1},
9435: \scite{600-nf.2}) we can find $M_\ell(\ell \le 3)$ such that $M_0 = N_0$ and
9436: NF$_{\lambda,\bar \delta}(M_0,M_1,M_2,M_3)$ and clauses (b), (c), (d) of
9437: Definition \scite{600-nf.2} and $(d)^+$ of \scite{600-nf.10}(2) hold so by
9438: \scite{600-nf.10} also
9439: NF$_\lambda(M_0,M_1,M_2,M_3)$. Easily there are for $\ell < 3$, isomorphisms
9440: $f_\ell$ from $M_\ell$ onto $N_\ell$ such that $f_0 = f_\ell
9441: \restriction M_\ell$ where $f_0 = \text{ id}_{N_0}$.
9442: By the uniqueness of smooth amalgamations
9443: (i.e., \scite{600-nf.11}(2)) we can find an isomorphism $f_3$ from $M_3$
9444: onto $N_3$ extending $f_1 \cup f_2$. So as
9445: NF$_{\lambda,\bar \delta}(M_0,M_1,M_2,M_3)$ holds also NF$_{\lambda,
9446: \bar \delta},(f_0(M_0),f_3(M_1),f_3(M_2),f_3(M_3))$; that is
9447: NF$_{\lambda,\bar \delta}(N_0,N_1,N_2,N_3)$ is as required.
9448: \hfill$\square_{\scite{600-nf.12}}$
9449: \enddemo
9450: \bigskip
9451:
9452: \proclaim{\stag{600-nf.13} Claim} [Monotonicity]: If
9453: ${\text{\rm NF\/}}_\lambda(N_0,N_1,N_2,N_3)$ and
9454: $N_0 \le_{\frak K} N'_1 \le_{\frak K} N_1$ and $N_0 \le_{\frak K}
9455: N'_2 \le_{\frak K} N_2$ and $N'_1 \cup N'_2
9456: \subseteq N'_3 \le_{\frak K} N''_3,N_3 \le_{\frak K} N''_3$
9457: \underbar{then} ${\text{\rm NF\/}}_\lambda(N_0,N'_1,N'_2,N'_3)$.
9458: \endproclaim
9459: \bigskip
9460:
9461: \demo{Proof} Read Definition \scite{600-nf.2}(1). \hfill$\square_{\scite{600-nf.13}}$
9462: \enddemo
9463: \bigskip
9464:
9465: \proclaim{\stag{600-nf.14} Claim} [Symmetry]:
9466: ${\text{\rm NF\/}}_\lambda(N_0,N_1,N_2,N_3)$
9467: holds \underbar{if and only if} \newline
9468: {\rm NF}$_\lambda(N_0,N_2,N_1,N_3)$ holds.
9469: \endproclaim
9470: \bigskip
9471:
9472: \demo{Proof} By Claim \scite{600-nf.8} (and Definition \scite{600-nf.2}).
9473: \hfill$\square_{\scite{600-nf.14}}$
9474: \enddemo
9475: \bn
9476: We observe
9477: \demo{\stag{600-nf.14.1} Conclusion} If NF$_\lambda(N_0,N_1,N_2,N_3),N_3$
9478: is $(\lambda,\sigma)$-brimmed over $N_1 \cup N_2$ and $\lambda \ge
9479: \sigma,\kappa \ge \aleph_0$, \ub{then} there is $N^+_2$ such that
9480: \mr
9481: \item "{$(a)$}" NF$_\lambda(N_0,N_1,N^+_2,N_3)$
9482: \sn
9483: \item "{$(b)$}" $N_2 \le_{\frak K} N^+_2$
9484: \sn
9485: \item "{$(c)$}" $N^+_2$ is $(\lambda,\kappa)$-brimmed over $N_0$ and
9486: even over $N_2$
9487: \sn
9488: \item "{$(d)$}" $N_3$ is $(\lambda,\sigma)$-brimmed over $N_1 \cup N^+_2$.
9489: \endroster
9490: \enddemo
9491: \bigskip
9492:
9493: \demo{Proof} Let $N^+_2$ be $(\lambda,\kappa)$-brimmed over $N_2$ be
9494: such that $N^+_2 \cap N_3 = N_2$. So by existence \scite{600-nf.10.3}
9495: there is $N^+_3$ such that NF$_\lambda(N_0,N_1,N^+_2,N^+_3)$ and
9496: $N^+_3$ is $(\lambda,\sigma)$-brimmed over $N_1 \cup N^+_2$. By
9497: monotonicity \scite{600-nf.13} we have NF$_\lambda(N_0,N_1,N_2,N^+_3)$.
9498: So by uniqueness (i.e., \scite{600-nf.11}(2)) \wilog \, $N_3 = N^+_3$, so
9499: we are done. \hfill$\square_{\scite{600-nf.14.1}}$
9500: \enddemo
9501: \bn
9502: The following claim is a step toward proving transitivity for
9503: NF$_\lambda$; so we first deal with NF$_{\lambda,\bar \delta}$. Note
9504: below: if we ignore $N^c_i$ we have problem showing NF$_{\lambda,\bar
9505: \delta}(N^a_0,N^a_\alpha,N^b_0,N^b_\alpha)$. Note that it is not clear
9506: at this stage whether, e.g. $N^b_\omega$ is even universal over
9507: $N^a_\omega$, but $N^c_\omega$ is; note that the $N^c_i$ are $\le_{\frak
9508: K}$-increasing with $i$ but not necessarily continuous.
9509: However once we finish proving that NF$_\lambda$ is a non-forking
9510: relation on ${\frak K}_{\frak s}$ respecting ${\frak s}$ this claim
9511: will lose its relevance.
9512: \proclaim{\stag{600-nf.15} Claim} Assume $\alpha < \lambda^+$ is an ordinal and
9513: for $x \in \{a,b,c\}$ the sequence $\bar N^x = \langle N^x_i:i \le \alpha
9514: \rangle$ is a $\le_{\frak K}$-increasing sequence of members of
9515: $K_\lambda$, and for $x = a,b$ the sequence $\bar N^x$ is
9516: $\le_{\frak K}$-increasing
9517: continuous, $N^b_i \cap N^a_\alpha = N^a_i,N^c_i \cap N^a_\alpha = N^a_i,
9518: N^a_i \le_{\frak K} N^b_i \le_{\frak K} N^c_i$ and $N^b_0$ is
9519: $(\lambda,\delta_2)$-brimmed over $N^a_0$ and
9520: ${\text{\rm NF\/}}_{\lambda,\bar \delta^i}(N^a_i,N^a_{i+1},N^c_i,N^b_{i+1})$
9521: (so necessarily $i < \alpha \Rightarrow N^c_i \le_{\frak K}
9522: N^b_{i+1}$) where \newline
9523: $\bar \delta^i = \langle \delta^i_1,\delta^i_2,\delta^i_3 \rangle$
9524: with $\delta^i_1,\delta^i_2,\delta^i_3$ are ordinals $< \lambda^+$
9525: and $\delta_3 < \lambda^+$ is limit, $N^c_\alpha$ is
9526: $(\lambda,{\text{\rm cf\/}}(\delta_3))$-brimmed over $N^b_\alpha,\delta_1 =
9527: \dsize \sum_{\beta < \alpha} \delta^\beta_1$ and $\delta_3 =
9528: \delta^\alpha_3$ and $\delta_2 = \delta^0_2,
9529: \bar \delta = \langle \delta_1,\delta_2,\delta_3 \rangle$. \nl
9530: \underbar{Then} ${\text{\rm NF\/}}_{\lambda,\bar \delta}
9531: (N^a_0,N^a_\alpha,N^b_0,N^c_\alpha)$.
9532: \endproclaim
9533: \bigskip
9534:
9535: \demo{Proof} For $i < \alpha$ let
9536: $\langle N^i_{1,\varepsilon},N^i_{2,\varepsilon},d^i_\zeta:\varepsilon
9537: \le \lambda \times \delta^i_1,\zeta < \lambda \times \delta^i_1
9538: \rangle$ be a witness to
9539: NF$_{\lambda,\bar \delta^i}(N^a_i,N^a_{i+1},N^c_i,N^b_{i+1})$. Now we
9540: define a sequence $\langle N_{1,\varepsilon},N_{2,\varepsilon},
9541: d^i_\zeta:\varepsilon \le \lambda \times \delta_1$ and $\zeta <
9542: \lambda \times \delta_1 \rangle$ where
9543: \mr
9544: \item "{$(a)$}" $N_{1,0} = N^a_0,N_{2,0} = N^b_0$ and
9545: \sn
9546: \item "{$(b)$}" if $\lambda \times (\dsize \sum_{j<i}
9547: \delta^j_1) < \zeta \le \lambda \times (\dsize \sum_{j \le i}
9548: \delta^j_1)$ then we let $N_{1,\zeta} =
9549: N^i_{1,\varepsilon_\zeta},N_{2,\zeta} =
9550: N^i_{2,\varepsilon_\zeta}$ where
9551: $\varepsilon_\zeta = \zeta - \lambda \times ( \dsize \sum_{j < i}
9552: \delta^j_1)$ and
9553: \sn
9554: \item "{$(c)$}" if $0 < \zeta = \lambda \times \dsize \sum_{j
9555: < \alpha} \delta^j_1$ we let $N_{1,\zeta} = N^a_i,N_{2,\zeta} = N^b_i
9556: = \alpha$ (if $i$ is non-limit we should note that this is compatible
9557: with clause (b), note that by this if $i = \alpha$ then
9558: $N_{1,\zeta} = N^a_\alpha,N_{2,\zeta} =
9559: \cup\{N^i_{2,\lambda \times \delta_1}:i < \alpha\}$
9560: \sn
9561: \item "{$(d)$}" if $\lambda \times (\dsize \sum_{j<i} \delta^j_1) \le
9562: \zeta < \lambda \times (\dsize \sum_{j \le i} \delta^j_1)$ then we let
9563: $d_\zeta = d^i_{\varepsilon_\zeta}$ where $\varepsilon_\zeta = \zeta -
9564: \lambda \times (\dsize \sum_{j<i} \delta^j_j) =
9565: \cup\{N^*_{2,\zeta}:\zeta < \lambda \times (\dsize \sum_{j<\alpha}
9566: \delta^j_1)$.
9567: \ermn
9568: Clearly $\langle
9569: N_{1,\zeta}:\zeta \le \lambda \times \delta_1 \rangle$ is $\le_{\frak
9570: K}$-increasing continuous, and also $\langle N_{2,\zeta}:\zeta \le \lambda
9571: \times \delta_1 \rangle$ is. Obviously $(N_{1,\zeta},N_{1,\zeta
9572: +1},d_\zeta) \in K^{3,\text{uq}}_\lambda$ as this just means
9573: $(N^i_{1,\varepsilon_\zeta},N^i_{1,\varepsilon_\zeta +1},d^i_\zeta)
9574: \in K^{3,\text{uq}}_\lambda$ when $\lambda \times \dsize \sum_{j<i}
9575: \delta^j_1:j \le \zeta < \lambda \times \dsize \sum_{j \le i}
9576: \delta^j_1$ and $\varepsilon_\zeta$ as above.
9577:
9578: Why \ortp$(d_\zeta,N_{2,\zeta},N_{2,\zeta +1})$ does not fork over
9579: $N_{1,\zeta}$ for $\zeta,i$ such that $\lambda \times (\dsize
9580: \sum_{j<i} \delta^j_1) \zeta < \lambda \times (\dsize \sum_{j \le i}
9581: \delta^j_j)$? If $\lambda \times \dsize \sum_{j<i}
9582: \delta^j_1 < \zeta$ this holds as it means
9583: \ortp$(d^i_{\varepsilon_\zeta},N^i_{2,\varepsilon_\zeta},N^i_{2,\varepsilon_\zeta
9584: +1})$ does not fork over $N^i_{1,\zeta}$. If $\lambda \times \dsize
9585: \sum_{j<i} \delta^j_1 = \zeta$ this is not the case but $N^i_{1,0} =
9586: N_{1,\zeta} \le_{\frak K} N_{2,\zeta} \le_{\frak K} N^c_i = N^i_{2,0}$
9587: and we know that \ortp$(d_\zeta,N^i_{2,0},N^i_{2,1})$ does not fork over
9588: $N^i_{1,0} = N_{1,\zeta}$ hence by monotonicity of non-forking
9589: \ortp$(d_\zeta,N_{2,\zeta},N_{2,\zeta +1})$ does not fork over
9590: $N_{1,\zeta}$ is as required.
9591: \nl
9592: Note that we have not demanded or used
9593: ``$\bar N^c$ continuous"; the $N^c_i$ is really needed for $i$ limit
9594: as we do not know that $N^b_i$ is brimmed over $N^a_i$.
9595: \hfill$\square_{\scite{600-nf.15}}$
9596: \enddemo
9597: \bigskip
9598:
9599: \proclaim{\stag{600-nf.16} Claim} [transitivity] 1) Assume
9600: that $\alpha < \lambda^+$ and for $x \in
9601: \{ a,b\}$ we have $\langle N^x_i:i \le \alpha \rangle$ is a
9602: $\le_{\frak K}$-increasing continuous sequence of members
9603: of $K_\lambda$. \newline
9604: If {\rm NF}$_\lambda(N^a_i,N^a_{i+1},N^b_i,N^b_{i+1})$ for each $i < \alpha$
9605: \underbar{then} {\rm NF}$_\lambda(N^a_0,N^a_\alpha,N^b_0,N^b_\alpha)$. \nl
9606: 2) Assume that $\alpha_1 < \lambda^+,\alpha_2 < \lambda^+$ and
9607: $M_{i,j} \in K_\lambda$ (for $i \le \alpha_1,j \le \alpha_2$)
9608: satisfy clauses (B), (C), (D),
9609: from \scite{600-nf.6A}, and for each $i < \alpha_1,j < \alpha_2$ we have:
9610:
9611: $$
9612: \nonforkin{M_{i,j+1}}{M_{i+1,j}}_{M_{i,j}}^{M_{i+1,j+1}}.
9613: $$
9614: \medskip
9615:
9616: $$
9617: \text{\underbar{Then} }
9618: \nonforkin{M_{i,0}}{M_{0,j}}_{M_{0,0}}^{M_{\alpha_1,\alpha_2}} \text{ for }
9619: i \le \alpha_1, j \le \alpha_2.
9620: $$
9621: \endproclaim
9622: \bigskip
9623:
9624: \demo{Proof} 1) We first prove special cases and use them to prove more
9625: general cases. \newline
9626: \smallskip
9627:
9628: \noindent
9629: \underbar{Case A}: $N^a_{i+1}$ is $(\lambda,\kappa_i)$-brimmed
9630: over $N^a_i$ and $N^b_{i+1}$ is $(\lambda,\sigma_i)$-brimmed
9631: over $N^a_{i+1} \cup N^b_i$ for $i < \alpha$ ($\sigma_i$ infinite, of
9632: course).
9633:
9634: In essence the problem is that we do not know ``$N^b_i$ is brimmed
9635: over $N^a_i$" ($i$ limit) so we shall use \scite{600-nf.15}; for this we
9636: introduce appropriate $N^c_i$.
9637:
9638: Let $\delta^i_1 = \kappa_i,\delta^i_2 = \kappa_i,\delta^i_3 =
9639: \sigma_i$ where we stipulate $\sigma_\alpha = \lambda$.
9640: For $i \le \alpha$ we can choose
9641: $N^c_i \in K_\lambda$ such that
9642: \mr
9643: \item "{$(a)$}" $N^b_i \le_{\frak K} N^c_i \le_{\frak K} N^b_{i+1},
9644: N^c_i$ is $(\lambda,\kappa_i)$-brimmed over $N^b_i$, and \newline
9645: ${\text{\rm NF\/}}_{\lambda,\langle \delta^i_1,\delta^i_2,\delta^i_3
9646: \rangle}(N^a_i,N^a_{i+1},N^c_i,N^b_{i+1})$
9647: \sn
9648: \item "{$(b)$}" $N^c_\alpha \in K_\lambda$ is
9649: $(\lambda,\delta^\alpha_3)$-brimmed over $N^b_\alpha$
9650: \sn
9651: \item "{$(c)$}" $\langle N^c_i:i < \alpha \rangle$ is
9652: $\le_{\frak K}$-increasing (in fact follows)
9653: \ermn
9654: (Possible by \scite{600-nf.14.1}). Now we can use \scite{600-nf.15}.
9655: \bn
9656: \underbar{Case B}:
9657: For each $i < \alpha$ we have: $N^a_{i+1}$ is $(\lambda,\kappa_i)$-brimmed
9658: over $N^a_i$.
9659:
9660: In essence our problem is that we do not know anything about brimmness of
9661: the $N^b_i$, so we shall ``correct it".
9662:
9663: Let $\bar \delta^i = (\kappa_i,\lambda,\lambda)$. \newline
9664: We can find a $\le_{\frak K}$-increasing sequence
9665: $\langle M^x_i:i \le \alpha \rangle$ of models in $K_\lambda$
9666: for $x \in \{ a,b,c\}$, continuous for $x=a,b$ such that
9667: $i < \alpha \Rightarrow M^a_i \le_{\frak K} M^b_i \le_{\frak K}
9668: M^c_i \le_{\frak K} M^b_{i+1}$ and $M^b_\alpha \le_{\frak K} M^c_\alpha$ and
9669: $M^c_i$ is $(\lambda,\kappa_i)$-brimmed over $M^b_i$ (hence over
9670: $M^a_i$) and
9671: NF$_{\lambda,\bar \delta^i}(M^a_i,M^a_{i+1},M^c_i,M^b_{i+1})$ by
9672: choosing $M^a_i,M^b_i,M^c_i$ by induction on
9673: $i,M^a_0 = N^a_0$ and $M^b_0$ is universal over $M^a_0$ recalling that
9674: the NF$_{\lambda,\bar \delta^i}$ implies some brimness condition,
9675: e.g. $M^b_{i+1}$ is $(\lambda,\text{cf}(\delta^i_3))$-brimmed over
9676: $M^a_{i+1} \cup M^b_i$.
9677: By Case A we know that
9678: NF$_\lambda(M^a_0,M^a_\alpha,M^b_0,M^c_\alpha)$ holds.
9679: \medskip
9680:
9681: We can now choose an isomorphism $f^a_0$ from $N^a_0$ onto $M^a_0$, as
9682: the identity (exists as $M^a_0 = N^a_0$) and then a $\le_{\frak K}$-embedding $f^b_0$
9683: of $N^b_0$ into $M^b_0$ extending $f^a_0$. Next we choose by induction on
9684: $i \le \alpha,f^a_i$ an isomorphism from $N^a_i$ onto $M^a_i$ such that:
9685: $j < i \Rightarrow f^a_j \subseteq f^a_i$, possible by ``uniqueness of the
9686: $(\lambda,\kappa_i)$-brimmed model over $M^a_i$" so here we are
9687: using the assumption of this case.
9688: \medskip
9689:
9690: Now we choose by induction on $i \le \alpha$, a $\le_{\frak K}$-embedding
9691: $f^b_i$ of
9692: $N^b_i$ into $M^b_i$ extending $f^a_i$ and $f^b_j$ for $j < i$. For
9693: $i = 0$ we have done it, for $i$ limit use $\dsize \bigcup_{j < i} f^b_j$,
9694: lastly for $i$ a successor ordinal let $i = j+1$, now we have
9695: \mr
9696: \item "{$(*)_2$}" NF$_\lambda(M^a_j,M^a_{j+1},f^b_j(N^b_j),M^b_{j+1})$
9697: \newline
9698: [why? because NF$_{\lambda,\bar \delta^j}(M^a_j,M^a_{j+1},M^c_j,M^b_{j+1})$
9699: by the choice of the \newline
9700: $M^x_\zeta$'s hence by \scite{600-nf.12} we have
9701: NF$_\lambda(M^a_j,M^a_{j+1},M^c_j,M^b_{j+1})$ and as \newline
9702: $M^a_j = f^a_j(N^a_j) \le_{\frak K} f^b_j(N^b_j) \le M^b_j
9703: \le_{\frak K} M^c_j$ by \scite{600-nf.13} we get $(*)_2$.]
9704: \ermn
9705: By $(*)_2$ and the uniqueness of smooth amalgamation \scite{600-nf.11} and as
9706: $M^b_{j+1}$ is $(\lambda,\text{cf}(\delta^3_j))$-brimmed over
9707: $M^a_{j+1} \cup M^b_j$ hence over $M^a_{j+1} \cup f^b_j(N^b_j)$ clearly
9708: there is $f^b_i$ as required. \newline
9709: So without loss of generality $f^a_\alpha$ is the identity, so we have
9710: $N^a_0 = M^a_0,N^a_\alpha = M^a_\alpha,N^b_0 \le_{\frak K} M^b_0,N^b_\alpha
9711: \le_{\frak K} M^b_\alpha$; also as said above NF$_\lambda(M^a_0,M^a_\alpha,
9712: M^b_0,M^b_\alpha)$ holds (using Case A) so by
9713: monotonicity, i.e., \scite{600-nf.13} we get
9714: NF$_\lambda(N^a_0,N^a_\alpha,N^b_0,N^b_\alpha)$ as required.
9715: \mn
9716: \underbar{Case C}: General case.
9717:
9718: We can find $M^\ell_i$ for $\ell < 3,i \le \alpha$ such that (note
9719: that $M^1_0 = M^0_0$):
9720: \medskip
9721: \roster
9722: \item "{$(a)$}" $M^\ell_i \in K_\lambda$
9723: \sn
9724: \item "{$(b)$}" for each $\ell < 3,M^\ell_i$ is $\le_{\frak
9725: K}$-increasing in $i$ (but for $\ell=1,2$ they are not required to be
9726: continuous)
9727: \sn
9728: \item "{$(c)$}" $M^0_i = N^a_i$
9729: \sn
9730: \item "{$(d)$}" $M^{\ell +1}_{i+1}$ is $(\lambda,\lambda)$-brimmed over
9731: $M^\ell_{i+1} \cup M^{\ell + 1}_i$ for $\ell < 2,i < \alpha$
9732: \sn
9733: \item "{$(e)$}" NF$_\lambda(M^\ell_i,M^\ell_{i+1},M^{\ell + 1}_i,
9734: M^{\ell + 1}_{i+1})$ for $\ell < 2,i < \alpha$
9735: \sn
9736: \item "{$(f)$}" $M^1_0 = M^0_0$ and
9737: $M^2_0$ is $(\lambda,\text{cf}(\lambda))$-brimmed over $M^1_0$
9738: \sn
9739: \item "{$(g)$}" for $\ell < 2$ and $i < \alpha$ limit we have
9740: $$
9741: M^{\ell +1}_i \text{ is } (\lambda,\lambda) \text{-brimmed over }
9742: \dsize \bigcup_{j < i} M^{\ell +1}_j \cup M^\ell_i
9743: $$
9744: \item "{$(h)$}" for $i < \alpha$ limit we have
9745: $$
9746: {\text{\rm NF\/}}_\lambda
9747: (\dsize \bigcup_{j < i} M^1_j,M^1_i,\dsize \bigcup_{j < i} M^2_j,
9748: M^2_i).
9749: $$
9750: \noindent
9751: [How? As in the proof of \scite{600-nf.6A} or just do by hand.]
9752: \ermn
9753: Now note:
9754: \mr
9755: \item "{$(*)_3$}" $M^{\ell +1}_i$ is $(\lambda,\text{cf}(\lambda \times
9756: (1 + i)))$-brimmed over $M^\ell_i$ if $\ell = 1 \vee i \ne 0$
9757: \newline
9758: [why? If $i=0$ by clause $(f)$, if $i$ a successor ordinal by clause $(d)$
9759: and if $i$ is a limit ordinal then by clause (g)]
9760: \sn
9761: \item "{$(*)_4$}" for $i < \alpha,{\text{\rm NF\/}}_\lambda
9762: (M^0_i,M^0_{i+1},M^2_i,M^2
9763: _{i+1})$. \newline
9764: [Why? If $i=0$ by clause (e) for $\ell=1,i=0$ we get
9765: NF$_\lambda(M^1_0,M^1_1,M^2_0,M^2_1)$ so by clause (f) (i.e., $M^1_0 =
9766: M^0_0$) and monotonicity (i.e., Claim \scite{600-nf.13}) we
9767: have NF$_\lambda(M^0_0,M^1_0,M^2_0,M^2_1)$ as
9768: required. If $i>0$ we use
9769: Case B for $\alpha = 2$ with $M^0_i,M^0_{i+1},M^1_i,M^1_{i+1},
9770: M^2_i,M^2_{i+1}$ here standing for $N^a_0,N^b_0,N^a_1,N^b_1,
9771: N^a_2,N^b_2$ there (and symmetry).]
9772: \ermn
9773: Let us define $N^\ell_i$ for $\ell < 3,i \le \alpha$ by: $N^\ell_i$ is
9774: $M^\ell_i$ if $i$ is non-limit and $N^\ell_i = \cup\{N^\ell_j:j < i\}$
9775: if $i$ is limit.
9776: \mr
9777: \widestnumber\item{$(*)_5(iii)$}
9778: \item "{$(*)_5(i)$}" $\langle N^\ell_i:i \le \alpha \rangle$ is
9779: $\le_{\frak K}$-increasing continuous, $N^0_i = N^a_i$ and $N^\ell_i
9780: \le_{\frak K} M^\ell_i$
9781: \sn
9782: \item "{$(ii)$}" for $i < \alpha$, NF$_\lambda(N^0_i,N^0_{i+1},
9783: N^2_i,N^2_{i+1})$ \nl
9784: [why? by $(*)_4 +$ monotonicity of NF$_\lambda$]
9785: \sn
9786: \item "{$(iii)$}" for $i < \alpha,N^2_{i+1}$ is
9787: $(\lambda,\text{cf}(\lambda))$-brimmed over $N^0_{i+1} \cup N^2_i$ and
9788: even over $N^1_{i+1} \cup N^2_i$
9789: \nl
9790: [why? by clause (d)]
9791: \sn
9792: \item "{$(*)_6$}" NF$_{\lambda,\langle \lambda,\lambda,1
9793: \rangle}(N^1_0,N^1_\alpha,N^2_0,N^2_\alpha)$. \nl
9794: [Why? As we have proved case A (or, if you prefer, by \scite{600-nf.15};
9795: easily the assumption there holds).]
9796: \ermn
9797: Choose $f^a_i = \text{ id}_{N^a_i}$ for $i \le \alpha$ and let $f^b_0$
9798: be a $\le_{\frak K}$-embedding of $N^b_0$ into $N^2_0$.
9799:
9800: Now we continue as in Case B defining by induction on $i$
9801: a $\le_{\frak K}$-embedding $f^b_i$ of $N^b_i$ into $N^2_i$,
9802: the successor case is possible by $(*)_5(ii) + (*)_5(iii)$.
9803: In the end by $(*)_6$ and monotonicity of NF$_\lambda$ (i.e., Claim
9804: \scite{600-nf.13}) we are done. \newline
9805: 2) Apply for each $i < \alpha_2$ part (1) to the sequences $\langle M_{\beta,i}:\beta
9806: \le \alpha_1 \rangle,\langle M_{\beta,i+1}:\beta \le \alpha_1 \rangle$ so we
9807: get $\nonforkin{M_{\alpha_1,i}}{M_{0,i+1}}_{M_{0,i}}^{M_{\alpha_1,i+1}}$
9808: hence by symmetry (i.e., \scite{600-nf.11})
9809: we have $\nonforkin{M_{0,i+1}}{M_{\alpha_1,i}}
9810: _{M_{0,i}}^{M_{\alpha_1,i+1}}$. Applying part (1) to the sequences
9811: $\langle M_{0,j}:j \le \alpha_2 \rangle,\langle M_{\alpha_1,j}:j \le \alpha_2
9812: \rangle$ we get $\nonforkin{M_{0,\alpha_2}}{M_{\alpha_1,0}}_{M_{0,0}}
9813: ^{M_{\alpha_1,\alpha_2}}$ hence by symmetry (i.e. \scite{600-nf.11}) we have
9814: $\nonforkin{M_{\alpha_1,0}}{M_{0,\alpha_2}}_{M_{0,0}}
9815: ^{M_{\alpha_1,\alpha_2}}$; so we get
9816: the desired conclusion. \hfill$\square_{\scite{600-nf.16}}$
9817: \enddemo
9818: \bigskip
9819:
9820: \proclaim{\stag{600-nf.17} Claim} Assume $\alpha < \lambda^+,
9821: \langle N^\ell_i:i \le \alpha \rangle$
9822: is $\le_{\frak K}$-increasing continuous sequence of models
9823: for $\ell = 0,1$ where $N^\ell_i \in K_\lambda$ and $N^1_{i+1}$ is
9824: $(\lambda,\kappa_i)$-brimmed over $N^0_{i+1} \cup N^1_i$ and
9825: ${\text{\rm NF\/}}_\lambda(N^0_i,N^1_i,N^0_{i+1},N^1_{i+1})$.
9826:
9827: \ub{Then} $N^1_\alpha$ is $(\lambda,{\text{\rm cf\/}}(\dsize
9828: \sum_{i < \alpha} \kappa_i))$-brimmed over $N^0_\alpha \cup N^1_0$.
9829: \endproclaim
9830: \bigskip
9831:
9832: \remark{\stag{600-nf.17A} Remark} 1) If our framework is uni-dimensional
9833: (see \sectioncite[\S2]{705}; as for example
9834: when it comes from \cite{Sh:576}) we can simplify the proof. \nl
9835: 2) Assuming only ``$N^1_{i+1}$ is
9836: universal over $N^0_{i+1} \cup N^1_i$" suffices when $\alpha$ is a
9837: limit ordinal, i.e., we get $N^1_\alpha$ is $(\lambda,\text{\rm
9838: cf}(\alpha))$-brimmed over $N^0_\alpha$. Why? We choose $N^2_j$ for
9839: $j \le i$ such that $N^2_j = N^1_j$ if $j=0$ or $j$ a limit ordinal
9840: and $N^2_j$ is a model $\le_{\frak s} N^1_j$ and
9841: $(\lambda,\kappa_1)$-brimmed over $N^0_j \cup N^1_i$ when $j=i+1$.
9842: Now $\langle N^2_j:j \le \alpha\rangle$ satisfies all the requirements
9843: in $\langle N^1_j:j \le \alpha\rangle$ in \scite{600-nf.17}.
9844: \nl
9845: 3) We could have proved this earlier and used it, e.g. in \scite{600-nf.16}.
9846: \endremark
9847: \bigskip
9848:
9849: \demo{Proof} The case $\alpha$ not a limit ordinal is trivial so
9850: assume $\alpha$ is a
9851: limit ordinal. We choose by induction on $i \le \alpha$, an ordinal
9852: $\varepsilon(i)$ and a sequence
9853: $\langle M_{i,\varepsilon}:\varepsilon \le \varepsilon(i) \rangle$ and
9854: $\langle c_\varepsilon:\varepsilon < \varepsilon(i) \text{ non-limit}
9855: \rangle$ such that:
9856: \mr
9857: \item "{$(a)$}" $\langle M_{i,\varepsilon}:\varepsilon \le \varepsilon(i)
9858: \rangle$ is (strictly) $<_{\frak K}$-increasing continuous in $K_\lambda$
9859: \sn
9860: \item "{$(b)$}" $N^0_i \le_{\frak K} M_{i,\varepsilon} \le_{\frak K}
9861: N^1_i$
9862: \sn
9863: \item "{$(c)$}" $N^0_i = M_{i,0}$ and $N^1_i = M_{i,\varepsilon(i)}$
9864: \sn
9865: \item "{$(d)$}" $\varepsilon(i)$ is (strictly) increasing continuous in
9866: $i$ and $\varepsilon(i)$ is divisible by $\lambda$
9867: \sn
9868: \item "{$(e)$}" $j < i \and \varepsilon \le \varepsilon(j) \Rightarrow
9869: M_{i,\varepsilon} \cap N^1_j = M_{j,\varepsilon}$
9870: \sn
9871: \item "{$(f)$}" for $j<i$ and $\varepsilon \le \varepsilon(j+1)$, the
9872: sequence $\langle M_{\beta,\varepsilon}:\beta \in
9873: (j,i] \rangle$ is $\le_{\frak K}$-increasing continuous
9874: \sn
9875: \item "{$(g)$}" for $j<i,\varepsilon < \varepsilon(j)$ non-limit; the type
9876: \ortp$(c_\varepsilon,M_{i,\varepsilon},M_{i,\varepsilon +1}) \in
9877: {\Cal S}^{\text{bs}}(M_{i,\varepsilon})$ does not fork over
9878: $M_{j,\varepsilon}$ (actually, here allowing all $\varepsilon$ is O.K., too)
9879: \sn
9880: \item "{$(h)$}" $M_{i+1,\varepsilon +1}$ is $(\lambda,
9881: \text{cf}(\lambda))$-brimmed over $M_{i+1,\varepsilon} \cup
9882: M_{i,\varepsilon +1}$
9883: \sn
9884: \item "{$(i)$}" if $\varepsilon < \varepsilon(i)$ and $p \in
9885: {\Cal S}^{\text{bs}}(M_{i,\varepsilon})$ then for $\lambda$ successor
9886: ordinals $\xi \in [\varepsilon,\varepsilon(i))$ the type
9887: \ortp$(c_\xi,M_{i,\xi},M_{i,\xi +1})$ is a non-forking extension of $p$.
9888: \ermn
9889: If we succeed, then $\langle M_{\alpha,\varepsilon}:\varepsilon \le
9890: \varepsilon(\alpha) \rangle$ is a (strictly)
9891: $<_{\frak K}$-increasing continuous sequence of models from $K_\lambda,
9892: M_{\alpha,0} = N^0_\alpha$,
9893: and $M_{\alpha,\varepsilon(\alpha)} = N^1_\alpha$.
9894: We can apply \scite{600-4a.2}
9895: and we conclude that $N^1_\alpha = M_{\alpha,\varepsilon(\alpha)}$ is
9896: $(\lambda,\text{cf}(\alpha))$-brimmed over $M_{\alpha,\varepsilon(j)}$
9897: hence over $N^0_\alpha \cup N^1_0$ (both $\le_{\frak K} M_{\alpha,1}$).
9898:
9899: Carrying the induction is easy. For $i=0$,
9900: there is not much to do. For $i$ successor we use ``$N^j_{i+1}$ is
9901: brimmed over $N^0_{i+1} \cup N^1_i$" the existence of
9902: non-forking amalgamations and \scite{600-4a.1}, bookkeeping and the extension
9903: property $(E)(g)$. For $i$ limit we have no problem.
9904: \hfill$\square_{\scite{600-nf.17}}$
9905: \enddemo
9906: \bigskip
9907:
9908: \demo{\stag{600-nf.18} Conclusion} 1) If NF$_\lambda(N_0,N_1,N_2,N_3)$ and
9909: $\langle M_{0,\varepsilon}:\varepsilon \le \varepsilon(*) \rangle$ is an
9910: $\le_{\frak K}$-increasing continuous sequence of models from $K_\lambda$,
9911: $N_0 \le_{\frak K} M_{0,\varepsilon} \le_{\frak K} N_2$ \underbar{then}
9912: we can find $\langle M_{1,\varepsilon}:
9913: \varepsilon \le \varepsilon(*) \rangle$ and $N'_3$ such that:
9914: \medskip
9915: \roster
9916: \item "{$(a)$}" $N_3 \le_{\frak K} N'_3 \in K_\lambda$
9917: \sn
9918: \item "{$(b)$}" $\langle M_{1,\varepsilon}:\varepsilon \le \varepsilon(*)
9919: \rangle$ is $\le_{\frak K}$-increasing continuous
9920: \sn
9921: \item "{$(c)$}" $M_{1,\varepsilon} \cap N_2 = M_{0,\varepsilon}$
9922: \sn
9923: \item "{$(d)$}" $N_1 \le_{\frak K} M_{1,\varepsilon} \le_{\frak K} N'_3$
9924: \sn
9925: \item "{$(e)$}" if $M_{0,0} = N_0$ then $M_{1,0} = N_1$
9926: \sn
9927: \item "{$(f)$}"
9928: NF$_\lambda(M_{0,\varepsilon},M_{1,\varepsilon},N_2,N'_3)$, for every
9929: $\varepsilon \le \varepsilon(*)$.
9930: \ermn
9931: 2) If $N_3$ is universal over $N_1 \cup N_2$, then without loss of generality
9932: $N'_3 = N_3$.
9933: \nl
9934: 3) In part (1) we can add
9935: \mr
9936: \item "{$(g)$}" $M_{1,\varepsilon +1}$ is brimmed over
9937: $M_{0,\varepsilon +1} \cup M_{1,\varepsilon}$.
9938: \endroster
9939: \enddemo
9940: \bigskip
9941:
9942: \demo{Proof} 1) Define $M'_{0,i}$ for $i \le \varepsilon^* := 1 +
9943: \varepsilon(*) +1$ by $M'_{0,0} = N_0,M'_{0,1 + \varepsilon} =
9944: M_{0,\varepsilon}$ for $\varepsilon \le \varepsilon(*)$ and $M'_{0,1 +
9945: \varepsilon(*)+1} = N_2$. By existence (\scite{600-nf.10.3})
9946: we can find an $\le_{\frak K}$-increasing continuous sequence $\langle
9947: M'_{1,\varepsilon}:\varepsilon \le \varepsilon^* \rangle$ with
9948: $M'_{1,0} = N_1$ and $\le_{\frak K}$-embedding $f$ of $N_2$ into
9949: $M'_{1,\varepsilon^*}$ such that $\varepsilon < \varepsilon^*
9950: \Rightarrow \text{\rm NF}_\lambda(f(M'_{0,\varepsilon}),M'_{1,0},
9951: f(M'_{0,\varepsilon +1}),M'_{1,\varepsilon +1})$. By transitivity we have
9952: NF$_\lambda(f(M'_{0,0}),M'_{1,0},
9953: f(M'_{0,\varepsilon^*}),M'_{1,\varepsilon^*})$.
9954: By disjointness (i.e., $f(M'_{0,\varepsilon^*}) \cap M'_{1,0} =
9955: M'_{0,0}$, see \scite{600-nf.3}(3))
9956: \wilog \, $f$ is the identity. By uniqueness for NF there are
9957: $N'_3,N_3 \le_{\frak K} N'_3 \in K_\lambda$ and $\le_{\frak
9958: K}$-embedding of $M'_{1,\varepsilon^*}$ onto $N'_3$ over $N_1 \cup N_2
9959: = M'_{0,\varepsilon^*} \cup M'_{1,0}$ so we are done.
9960: \nl
9961: 2) Follows by (1).
9962: \nl
9963: 3) Similar to (1). \hfill$\square_{\scite{600-nf.18}}$
9964: \enddemo
9965: \bigskip
9966:
9967: \proclaim{\stag{600-nf.19} Claim} {\rm NF}$_\lambda$ respects ${\frak s}$;
9968: that is assume
9969: ${\text{\rm NF\/}}_\lambda(M_0,M_1,M_2,M_3)$ and
9970: $a \in M_1 \backslash M_0$ satisfies {\rm \ortp}$(a,M_0,M_3)
9971: \in {\Cal S}^{\text{bs}}(M_0)$,
9972: \ub{then} ${\text{\rm \ortp\/}}(a,M_2,M_3) \in {\Cal S}^{\text{bs}}(M_2)$
9973: does not fork over $M_0$.
9974: \endproclaim
9975: \bigskip
9976:
9977: \demo{Proof} Without loss of generality $M_1$ is $(\lambda,*)$-brimmed
9978: over $M_0$. [Why? By the existence we can find $M^+_1$ which is a
9979: $(\lambda,*)$-brimmed extension of $M_1$. By the existence for
9980: NF$_\lambda$ without loss of generality we can find $M^+_3$ such that
9981: NF$_\lambda(M_1,M^+_1,M_3,M^+_3)$, hence by transitivity for
9982: NF$_\lambda$ we have NF$_\lambda(M_0,M^+_1,M_2,M^+_3)$.]
9983: By the hypothesis of the section there are $M'_1,a'$ such that
9984: $M_0 \cup \{a'\} \subseteq M'_1$ and $\ortp(a',M_0,M'_1) =
9985: \text{\rm \ortp}(a,M_0,M_1)$ and $(M_0,M'_1,a) \in
9986: K^{3,\text{uq}}_\lambda$; as $M^+_1$ is
9987: $(\lambda,*)$-brimmed over $M_0$ without loss of generality
9988: $M' \le_{\frak K} M^+_1$
9989: and $a' = a$ and $M_1$ is $(\lambda,*)$-brimmed over
9990: $M'_1$. We can apply \scite{600-nf.0A} to $M'_1,M^+_1$ getting $\langle
9991: M^*_i,a_i:i \le \delta < \lambda^+ \rangle$ as there. Let $M'_i$ be:
9992: $M_0$ if $i=0,M^*_j$ if $1+j = i$ so $M'_1 = M^*_0 = M'_1$ and let $a_i$
9993: be $a$ if $i=0,a_j$ if $1+j=i$. So we can find $M'_3$ and $f$ such
9994: that $M_2 \le_{\frak K} M'_3,f$ is a $\le_{\frak K}$-embedding of
9995: $M^+_1$ into $M'_3$ extending id$_{M_0}$ such that NF$_{\lambda,\langle
9996: \delta,\lambda,\lambda \rangle}
9997: (M_0,f(M^+_1),M_2,M'_3)$ and $M'_3$, this is witnessed by
9998: $\langle f(M'_i):i \le \delta \rangle,\langle M''_i:i \le \delta \rangle,
9999: \langle f(a_i):i < \delta \rangle$ and $M''_0 = M_2$; this is
10000: possible by \scite{600-nf.3}(2). Hence NF$_\lambda(M_0,f(M^+_1),M_2,N) =
10001: \text{\rm NF}_\lambda(f(M'_0),f(M'_\delta),M''_0,N)$
10002: hence by the uniqueness for NF$_\lambda$ \wilog \,
10003: $f = \text{ id}_{M^+_1}$ and $M_3 \le_{\frak K} N$.
10004: By the choice of $f,N$ we have \ortp$(a,M_2,M_3) = \text{\ortp}(a_0,M_2,N) =
10005: \text{ \ortp}(a_0,M''_0,M'_1) \in {\Cal S}^{\text{bs}}(M''_0)
10006: = {\Cal S}^{\text{bs}}(M_2)$ does not fork over $M'_0 = M_0$ as required.
10007: \hfill$\square_{\scite{600-nf.19}}$
10008: \enddemo
10009: \bigskip
10010:
10011: \demo{\stag{600-nf.20} Conclusion} If $M_0 \le_{\frak K} M_\ell
10012: \le_{\frak K} M_3$ for $\ell = 1,2$ and $(M_0,M_1,a) \in
10013: K^{3,\text{uq}}_\lambda$ and ${\text{\rm \ortp\/}}(a,M_2,M_3) \in
10014: {\Cal S}^{\text{bs}}(M_2)$ does not fork over $M_0$ \ub{then}
10015: ${\text{\rm NF\/}}(M_0,M_1,M_2,M_3)$.
10016: \enddemo
10017: \bigskip
10018:
10019: \demo{Proof} By the definition of $K^{3,\text{uq}}_\lambda$ and
10020: existence for NF$_\lambda$ and \scite{600-nf.19} (or use \scite{600-nf.0Y} +
10021: \scite{600-nf.20.7}. \hfill$\square_{\scite{600-nf.20}}$
10022: \enddemo
10023: \bn
10024: We can sum up our work by
10025: \demo{\stag{600-nf.20.7} Main Conclusion} NF$_\lambda$ is a non-forking
10026: relation on ${}^4({\frak K}_\lambda)$ which respects ${\frak s}$.
10027: \enddemo
10028: \bigskip
10029:
10030: \demo{Proof} We have to check clauses (a)-(g)+(h) from \scite{600-nf.0X}.
10031: Clauses (a),(b) hold by the Definition \scite{600-nf.2} of NF$_\lambda$.
10032: Clauses $(c)_1,(c)_2$, i.e., monotonicity hold by \scite{600-nf.13}.
10033: Clause (d), i.e., symmetry holds by \scite{600-nf.14}.
10034: Clause (e), i.e., transitivity holds by \scite{600-nf.16}.
10035: Clause (f), i.e., existence hold by \scite{600-nf.10.3}.
10036: Clause (g), i.e., uniqueness holds by \scite{600-nf.11}.
10037:
10038: Lastly, clause (h), i.e., NF$_\lambda$ respecting ${\frak s}$ by
10039: \scite{600-nf.19}. \hfill$\square_{\scite{600-nf.20.7}}$
10040: \enddemo
10041: \bn
10042: The following definition is not needed for now but is
10043: natural (of course, we can omit ``there is
10044: superlimit" from the assumption and the conclusion). For the rest of
10045: the section we stop assuming Hypothesis \scite{600-nf.0}.
10046: \definition{\stag{600-nf.20.7A} Definition} 1) A good $\lambda$-frame
10047: ${\frak s}$ is type-full when for $M \in {\frak K}_{\frak s},{\Cal
10048: S}^{\text{bs}}(M) = {\Cal S}^{\text{na}}_{{\frak K}_\lambda}(M)$.
10049: \nl
10050: 2) Assume ${\frak K}_\lambda$ is a $\lambda$-a.e.c. and NF is a
10051: 4-place relation on $K_\lambda$. We define ${\frak t} = {\frak
10052: t}_{{\frak K}_\lambda,\text{NF}} = (K_{\frak t},\nonfork{}{}_{\frak
10053: t},{\Cal S}^{\text{bs}}_{\frak t})$ as follows:
10054: \mr
10055: \item "{$(a)$}" ${\frak K}_{\frak t}$ is the $\lambda$-a.e.c. ${\frak
10056: K}_\lambda$
10057: \sn
10058: \item "{$(b)$}" ${\Cal S}^{\text{bs}}_{\frak t}(M)$ is ${\Cal
10059: S}^{\text{na}}_{{\frak K}_\lambda}(M)$ for $M \in {\frak K}_\lambda$
10060: \sn
10061: \item "{$(c)$}" $\nonfork{}{}_{\frak t}$ is defined by:
10062: $(M_0,M_1,a,M_3) \in \nonfork{}{}_{\frak t}$ when we can find
10063: $M_2,M'_3$ such that $M_0 \le_{{\frak K}_\lambda} M_2 \le_{{\frak
10064: K}_\lambda} M'_3,M_3 \le_{{\frak K}_\lambda} M'_3,a \in M_2 \backslash M_0$ and
10065: NF$(M_0,M_1,M_2,M'_3)$.
10066: \endroster
10067: \enddefinition
10068: \bigskip
10069:
10070: \proclaim{\stag{600-nf.20.8} Claim} 1) Assume that
10071: \mr
10072: \item "{$(a)$}" ${\frak K}_\lambda$ is a $\lambda$-a.e.c. with
10073: amalgamation and a superlimit model
10074: \sn
10075: \item "{$(b)$}" ${\frak K}_\lambda$ is stable
10076: \sn
10077: \item "{$(c)$}" {\rm NF} is a ${\frak K}_\lambda$-non-forking
10078: relation, see Definition \scite{600-nf.0X}(1).
10079: \ermn
10080: \ub{Then} ${\frak t} = {\frak t}_{{\frak K}_\lambda,\text{NF}}$ is a
10081: type-full good $\lambda$-frame.
10082: \nl
10083: 2) Assume that ${\frak s}$ is a good $\lambda$-frame which has
10084: existence for $K^{3,\text{uq}}_\lambda$ (see \scite{600-nf.0}(2)) and {\rm
10085: NF} = {\rm NF}$_\lambda$. \ub{Then} ${\frak t}$ is very close to
10086: ${\frak s}$, i.e.:
10087: \mr
10088: \item "{$(a)$}" ${\frak K}_{\frak s} = {\frak K}_{\frak t}$
10089: \sn
10090: \item "{$(b)$}" if $p \in {\Cal S}^{\text{bs}}_{\frak s}(M_1)$ and
10091: $M_0 \le_{{\frak K}_\lambda} M_1$ \ub{then} $p \in {\Cal
10092: S}^{\text{bs}}_{\frak t}(M_1)$ and $p$ forks over $M_0$ for ${\frak
10093: s}$ iff $p$ forks over $M_0$ for ${\frak t}$.
10094: \endroster
10095: \endproclaim
10096: \bigskip
10097:
10098: \demo{Proof} For the time being, left to the reader (but before it is
10099: really used it is proved in \marginbf{!!}{\cprefix{705}.\scite{705-9.11A}}).
10100: \enddemo
10101: \bigskip
10102:
10103: \remark{Remark} Note that this actually says that from now on we
10104: could have used type-full ${\frak s}$, but it is not necessary for a
10105: long time.
10106: \endremark
10107: \bigskip
10108:
10109: \definition{\stag{600-nf.20.9} Definition} 1) Let ${\frak s}$ be a good
10110: $\lambda$-frame. We say that NF is a weak
10111: ${\frak s}$-non-forking relation where
10112: \mr
10113: \item "{$(a)$}" NF is weak ${\frak K}_{\frak s}$-non-forking
10114: relation, see Definition \scite{600-nf.0X}(2), i.e., uniqueness is omitted
10115: \sn
10116: \item "{$(b)$}" NF respects ${\frak s}$, see Definition
10117: \scite{600-nf.0X}(3)
10118: \sn
10119: \item "{$(c)$}" NF satisfies \scite{600-nf.18}, (NF-lifting of an
10120: $\le_{\frak K}$-increasing sequence).
10121: \ermn
10122: 2) We say ${\frak s}$ is pseudo-successful if some NF is a weak
10123: ${\frak s}$-non-forking relation witnesses it.
10124: \enddefinition
10125: \bigskip
10126:
10127: \demo{\stag{600-nf.20.10} Observation} 1) If ${\frak s}$ is a good
10128: $\lambda$-frame which is weakly successful (i.e., has
10129: existence for $K^{3,\text{uq}}_\lambda$, i.e., \scite{600-nf.0}) \ub{then}
10130: {\rm NF}$_\lambda = \text{\rm NF}_{\frak s}$ is a weak ${\frak
10131: s}$-non-forking relation.
10132: \nl
10133: 2) If ${\frak s}$ is a good $\lambda$-frame and {\rm NF} is a weak
10134: ${\frak s}$-non-forking relation then \scite{600-nf.20} holds.
10135: \nl
10136: 3) If ${\frak s}$ is a good $\lambda$-frame and NF is an
10137: ${\frak s}$-forking relation \ub{then} NF is a weak
10138: ${\frak s}$-non-forking relation.
10139: \enddemo
10140: \bigskip
10141:
10142: \demo{Proof} Straight.
10143: \nl
10144: 1) Follows by \scite{600-nf.20.7}, NF$_\lambda$ satisfies clauses (a)+(b)
10145: and by \scite{600-nf.18} it satisfies also clause (c) of Definition
10146: \scite{600-nf.0X}(1).
10147: \nl
10148: 2) Also easy.
10149: \nl
10150: 3) We have just to check the proof of \scite{600-nf.18} still works.
10151: \enddemo
10152: \bigskip
10153:
10154: \remark{\stag{600-nf.21.3} Remark} 1) In \chaptercite{705} ,\S1 -\S11 we can
10155: use ``${\frak s}$ is pseudo successful as witnessed by NF + lifting of
10156: decompositions" instead of ``${\frak s}$ is weakly successful". We
10157: shall return to this elsewhere, see \cite{Sh:838}, \cite{Sh:842}.
10158: \endremark
10159: \newpage
10160:
10161: \head {\S7 Nice extensions in $K_{\lambda^+}$} \endhead \resetall \sectno=7
10162: \spuriousreset
10163: \bigskip
10164:
10165: \demo{\stag{600-ne.0} Hypothesis} Assume the hypothesis \scite{600-nf.0}. \nl
10166: \enddemo
10167: \bn
10168: So by \S6 we have reasonable control on \underbar{smooth} amalgamation in
10169: $K_\lambda$. We use this to define ``nice" extensions in
10170: $K_{\lambda^+}$ and prove some basic properties.
10171: This will be treated again in \S8.
10172: \bigskip
10173:
10174: \definition{\stag{600-ne.1} Definition} 1) $K^{\text{nice}}_{\lambda^+}$
10175: is the class of saturated $M \in K_{\lambda^+}$. \nl
10176: 2) Let $M_0 \le^*_{\lambda^+} M_1$ mean:
10177: \mr
10178: \item "{{}}" $M_0 \le_{\frak K} M_1$ and they are from
10179: $K_{\lambda^+}$ and we can find $\bar M^\ell = \langle M^\ell_i:i < \lambda^+
10180: \rangle$, a $\le_{\frak K}$-representation of $M_\ell$ for $\ell = 0,1$
10181: such that: \newline
10182: NF$_\lambda(M^0_i,M^0_{i+1},M^1_i,M^1_{i+1}$) for $i < \lambda^+$.
10183: \endroster
10184: \medskip
10185: \noindent
10186: 3) Let $M_0 <^+_{\lambda^+,\kappa} M_1$ mean
10187: \footnote{Note that $M_0 <^+_{\lambda^+,\kappa} M_1$ implies $M_1 \in
10188: K^{\text{nice}}_{\lambda^+}$ but in general $M_0 \in
10189: K^{\text{nice}}_{\lambda^+}$ does not follow.}
10190: that
10191: $(M_0,M_1 \in K_{\lambda^+}$ and) $M_0 \le^*_{\lambda^+} M_1$ by
10192: some witnesses $M^\ell_i$ (for $i < \lambda^+,\ell < 2$) such that
10193: NF$_{\lambda,\langle 1,1,\kappa \rangle}(M^0_i,M^0_{i+1},
10194: M^1_i,M^1_{i+1})$ for $i < \lambda^+$;
10195: of course $M_0 \le_{\frak K} M_1$ in this case.
10196: Let $M_0 \le^+_{\lambda^+,\kappa} M_1$ mean $(M_0 = M_1 \in
10197: K_{\lambda^+}) \vee (M_0 <^+_{\lambda^+,\kappa} M_1)$.
10198: If $\kappa = \lambda$, we may omit it. \nl
10199: 4) Let $K^{3,\text{bs}}_{\lambda^+} = \{(M,N,a):M \le^*_{\lambda^+} N$
10200: are from $K_{\lambda^+}$ and $a \in N \backslash M$ and for some $M_0
10201: \le_{\frak K} M,M_0 \in K_\lambda$ we have $[M_0 \le_{\frak K} M_1
10202: \le_{\frak K} M \and M_1 \in K_\lambda$ \ub{implies} $\text{ \ortp}(a,M_1,N) \in
10203: {\Cal S}^{\text{bs}}(M_1)$ and does not fork over $M_0]\}$.
10204: We call $M_0$ or \ortp$(a,M_0,N)$ a witness for $(M,N,a) \in
10205: K^{3,\text{bs}}_{\lambda^+}$. (In fact this definition
10206: on $K^{3,\text{bs}}_{\lambda^+}$ is compatible with the
10207: definition in \S2 for triples such that $M \le^*_{\lambda^+} N$ but we
10208: do not know now whether even
10209: $(K^{\text{nice}}_{\lambda^+},\le^*_{\lambda^+})$ is a $\lambda^+$-a.e.c..)
10210: \enddefinition
10211: \bigskip
10212:
10213: \proclaim{\stag{600-ne.2} Claim} 0) $K^{\text{nice}}_{\lambda^+}$ has one
10214: and only one model up to isomorphisms and
10215: $M \in K^{\text{nice}}_{\lambda^+}$ implies
10216: $M \le^*_{\lambda^+} M$ and $M \le^+_{\lambda^+} M$;
10217: moreover, $M \in K_{\lambda^+} \Rightarrow
10218: M \le^*_{\lambda^+} M$. Also $\le^*_{\lambda^+}$ is a partial order
10219: and if $M_\ell \in K_{\lambda^+}$ for $\ell=0,1,2$ and
10220: $M_0 \le_{\frak K} M_1 \le_{\frak K} M_2$ and $M_0 \le^*_{\lambda^+}
10221: M_2$ then $M_0 \le^*_{\lambda^+} M_1$.
10222: \nl
10223: 1) If $M_0 \le^*_{\lambda^+} M_1$ and
10224: $\bar M^\ell = \langle M^\ell_i:
10225: i < \lambda^+ \rangle$ is a representation of $M_\ell$ for $\ell=0,1$ \ub{then}
10226: \mr
10227: \item "{$(*)$}" for some club $E$ of $\lambda^+$,
10228: {\roster
10229: \itemitem{ $(a)$ } for every $\alpha < \beta$ from $E$ we have
10230: {\rm NF}$_\lambda(M^0_\alpha,M^0_\beta,M^1_\alpha,M^1_\beta)$
10231: \sn
10232: \itemitem{ $(b)$ } if $\ell < 2$ and $M_\ell \in
10233: K^{\text{nice}}_{\lambda^+}$ then for $\alpha < \beta$ from $E$ the
10234: model $M^\ell_\beta$ is $(\lambda,*)$-brimmed over $M^\ell_\alpha$.
10235: \endroster}
10236: \ermn
10237: 2) Similarly for $<^+_{\lambda^+,\kappa}$: if $M_0
10238: <^+_{\lambda^+,\kappa} M_1,\bar M^\ell = \langle \bar M^\ell_i:i <
10239: \lambda^+ \rangle$ a representation of $M_\ell$ for $\ell = 0,1$
10240: \ub{then} for some club $E$ of $\lambda^+$ for every $\alpha < \beta$
10241: from $E$ we have {\rm NF}$_{\lambda,\langle 1,1,\kappa\rangle}
10242: (M^0_\alpha,M^0_\beta,M^1_\alpha,M^1_\beta)$, moreover
10243: NF$_{\lambda,\langle 1,\text{cf}(\lambda \times (1 +
10244: \beta)),\kappa\rangle}(M^0_\alpha,M^0_\beta,M^1_\alpha,M^1_\beta)$ and if
10245: $(M_\alpha,\bar M^0_\beta,M^1_\alpha,M^1_\beta),M_0 \in
10246: K^{\text{nice}}_{\lambda^+}$ then we can add NF$_{\lambda,\langle
10247: \lambda,\text{cf}(\lambda \times(1 + \beta)),\kappa)}(M^0_\alpha,M^0_\beta,M'_\alpha,M'_\beta)$.
10248: \nl
10249: 3) The $\kappa$ in Definition \scite{600-ne.1}(3) does not matter.
10250: \nl
10251: 4) If $M_0 <^+_{\lambda^+,\kappa} M_1$, \ub{then} $M_1 \in
10252: K^{\text{nice}}_{\lambda^+}$. \nl
10253: 5) If $M \in K_{\lambda^+}$ is saturated, equivalently $M \in
10254: K^{\text{nice}}_{\lambda^+}$ \ub{then} $M$ has a $\le_{\frak K}$-representation
10255: $\bar M = \langle M_\alpha:\alpha < \lambda^+ \rangle$ such that
10256: $M_{i+1}$ is $(\lambda,\lambda)$-brimmed over $M_i$ for $i <
10257: \lambda^+$ and also the inverse is true.
10258: \nl
10259: 6) If $M \le^*_{\lambda^+} N$ and $N_0 \le_{\frak K} N,N_0 \in
10260: K_\lambda$ \ub{then} we can find $M_1 \le_{\frak K} N_1$ from
10261: $K_\lambda$ such that $M_1 \le_{\frak K} M,N_0 \le_{\frak K} N_1
10262: \le_{\frak K} N$ and: for every $M_2 \in K_\lambda$ satisfying
10263: $M_1 \le_{\frak K} M_2 \le_{\frak K} M$
10264: there is $N_2 \le_{\frak K} N$ such that
10265: {\rm NF}$_{\frak s}(M_1,M_2,N_1,N_2)$.
10266: \endproclaim
10267: \bigskip
10268:
10269: \demo{Proof} 0) Obvious by now (for the second sentence use part (1)
10270: and NF$_{\frak s}$ being a non-forking relation on ${\frak K}_{\frak
10271: s}$) in particular transitivity and monotonicity.
10272: \nl
10273: 1) Straight by \scite{600-nf.16} as any two representations
10274: agree on a club. \newline
10275: 2) Up to ``moreover" quite straight. For the ``moreover" use
10276: \scite{600-nf.17} to show that $M^1_\beta$ is
10277: $(\lambda,\text{cf}(\beta))$-brimmed over $M^0_\beta$. Lastly, for
10278: the ``we can add" just use part (5), choosing thin enough club $E$ of
10279: $\lambda^+$ then use $\{\alpha \in E:\text{otp}(\alpha \cap E)$ is
10280: divisible by $\lambda\}$.
10281: \nl
10282: 3) By \scite{600-nf.17}. \nl
10283: 4) By \scite{600-nf.17}. \nl
10284: 5) Trivial.
10285: \nl
10286: 6) Easy. \hfill$\square_{\scite{600-ne.2}}$
10287: \enddemo
10288: \bigskip
10289:
10290: \proclaim{\stag{600-ne.3} Claim} 0) For every $M_0 \in K_{\lambda^+}$ for some
10291: $M_1 \in K^{\text{nice}}_{\lambda^+}$ we have $M_0 \le_{\frak K} M_1$. \nl
10292: 1) For every $M_0 \in K_{\lambda^+}$ and
10293: $\kappa = { \text{\rm cf\/}}(\kappa) \le \lambda$
10294: for some $M_1 \in K_{\lambda^+}$ we have $M_0 <^+_{\lambda^+,\kappa} M_1$ so
10295: $M_1 \in K^{\text{nice}}_{\lambda^+}$. \newline
10296: 1A) Moreover, if $N_0 \le_{\frak K} M_0 \in K_{\lambda^+},N_0 \in K_\lambda,
10297: p \in {\Cal S}^{\text{bs}}(N_0)$ \ub{then} in (1) we can add that for
10298: some $a,(M_0,M_1,a) \in K^{3,\text{bs}}_\lambda$ as witnessed by $p$. \nl
10299: 2) $\le^*_{\lambda^+}$ and $<^+_{\lambda^+,\kappa}$ are transitive. \newline
10300: 3) If $M_0 \le_{\frak K} M_1 \le_{\frak K} M_2$ are in $K_{\lambda^+}$
10301: and $M_0 \le^*_{\lambda^+} M_2$, \ub{then}
10302: $M_0 \le^*_{\lambda^+} M_1$. \newline
10303: 4) If $M_1 <^+_{\lambda^+,\kappa} M_2$, \ub{then}
10304: $M_1 <^*_{\lambda^+} M_2$. \nl
10305: 5) If $M_0 <^*_{\lambda^+} M_1
10306: <^+_{\lambda,\kappa} M_2$ \ub{then} $M_0 <^+_{\lambda,\kappa} M_2$.
10307: \endproclaim
10308: \bigskip
10309:
10310: \demo{Proof} 0) Easy and follows by the proof of part (1) below. \nl
10311: 1), 1A) Let $\langle M^0_i:i < \lambda^+ \rangle$
10312: be a $\le_{\frak K}$-representation of $M_0$ with $M^0_i$ brimmed and
10313: brimmed over $M^0_j$ for $j <i$ and for part (1A) we have
10314: $M^0_0 = N_0$, and for part (1) let $p$ be any member of
10315: ${\Cal S}^{\text{bs}}(M^0_0)$.
10316: We choose by induction on $i$ a model $M^1_i \in K_\lambda$ and
10317: $a \in M^1_0$ such that $M^1_i$ is $(\lambda,\text{cf}(\lambda \times
10318: (1+i)))$-brimmed over
10319: $M^0_i,\langle M^1_i:i < \lambda^+ \rangle$ is
10320: $<_{\frak K}$-increasing continuous, $M^1_i \cap M_0 = M^0_i$ and
10321: \ortp$(a,M^0_0,M^1_0) = p$ and $M^1_{i+1}$ is
10322: $(\lambda,\kappa)$-brimmed over $M^0_{i+1} \cup M^1_i$ and
10323: NF$_{\lambda,\langle 1,\text{cf}(\lambda \times (1+i)),\kappa
10324: \rangle}(M^0_i,M^0_{i+1},
10325: M^1_i,M^1_{i+1})$ for $i < \lambda^+$.
10326: Note that for limit $i$, by \scite{600-nf.17}, $M^1_i$ is
10327: $(\lambda,\text{cf}(i))$-brimmed over $M^0_i \cup M^1_j$ for any $j<i$.
10328:
10329: Note that for $i < \lambda^+$, the type \ortp$(a,M^0_i,M^1_i)$ does not
10330: fork over $M^0_0 = N_0$ and extends $p$ by \scite{600-nf.19} (saying
10331: NF$_\lambda$ respects ${\frak s}$)
10332: \scite{600-nf.14} (symmetry) and \scite{600-nf.12}. So clearly we are done. \nl
10333: 2) Concerning $<^+_{\lambda^+,\kappa}$ use \scite{600-ne.2}
10334: and \scite{600-nf.16} (i.e. transitivity for smooth amalgamations).
10335: The proof for $<^*_{\lambda^+}$ is the same. \newline
10336: 3) By monotonicity for smooth amalgamations in ${\frak K}_\lambda$;
10337: i.e., \scite{600-nf.13}. \newline
10338: 4), 5) Check. \hfill$\square_{\scite{600-ne.3}}$
10339: \enddemo
10340: \bigskip
10341:
10342: \proclaim{\stag{600-ne.2A} Claim} 1) If $(M_0,M_1,a) \in
10343: K^{3,\text{bs}}_{\lambda^+}$
10344: and $M_1 \le^*_{\lambda^+} M_2 \in K_{\lambda^+}$ \ub{then} \nl
10345: $(M_0,M_2,a) \in K^{3,\text{bs}}_{\lambda^+}$. \nl
10346: 2) If $M_0 <^*_{\lambda^+} M_1$, \ub{then} for some $a,(M_0,M_1,a) \in
10347: K^{3,\text{bs}}_{\lambda^+}$.
10348: \endproclaim
10349: \bigskip
10350:
10351: \demo{Proof} 1) By the transitivity of $\le^*_{\lambda^+}$ which
10352: holds by \scite{600-ne.3}(2).
10353: \nl
10354: 2) As in the proof of \scite{600-1.11}, in fact it follows from it. \nl
10355: 3) Easy (and is included in \scite{600-ne.3}(1A)).
10356: \enddemo
10357: \bigskip
10358:
10359: \remark{Remark} Note that the parallel to \scite{600-ne.3}(1A) is
10360: problematic in \S2 as, .e.g. locality may fail, i.e., $(M,N_i,a_i) \in
10361: K^{3,\text{bs}}_{\lambda^+}$ and $M' \le_{\frak K} M \wedge M' \in
10362: K_\lambda \Rightarrow \text{ \ortp}_{\frak s}(a_1,M',N_1) = \text{
10363: \ortp}_{\frak s}(a_2,M',N_2)$ but \ortp$_{K^{\frak
10364: s}_{\lambda^+}}(a_1,M,N_1) \ne
10365: \ortp_{K^{\frak s}_{\lambda^+}}
10366: (\bar a_2,M,N_2)$. \hfill$\square_{\scite{600-ne.2A}}$
10367: \endremark
10368: \bigskip
10369:
10370: \proclaim{\stag{600-ne.4} Claim} 1) [Amalgamation of $\le^*_{\lambda^+}$
10371: and toward extending types]
10372: If $M_0 \le^*_{\lambda^+} M_\ell$ for
10373: $\ell = 1,2,\kappa = { \text{\rm cf\/}}(\kappa) \le
10374: \lambda$ and $a \in M_2 \backslash
10375: M_0$ is such that $(M_0,M_2,a) \in K^{3,\text{bs}}_{\lambda^+}$
10376: is witnessed by $p$,
10377: \underbar{then} for some $M_3$ and $f$ we have: $M_1 <^+_{\lambda^+,
10378: \kappa}M_3$ and $f$ is an $\le_{\frak K}$-embedding of $M_2$ into
10379: $M_3$ over $M_0$ with $f(a) \notin M_1$, moreover, $f(M_2)
10380: \le^*_{\lambda^+}M_3$ and $(M_1,M_3,f(a)) \in K^{3,\text{bs}}_{\lambda^+}$ is
10381: witnessed by $p$. \newline
10382: 2) [uniqueness] Assume $M_0 <^+_{\lambda^+,\kappa}M_\ell$ for $\ell = 1,2$
10383: \underbar{then} there is an isomorphism $f$ from $M_1$ onto $M_2$ over
10384: $M_0$. \nl
10385: 3) [locality] Moreover
10386: \footnote{the meaning of this will be that types over
10387: $M \in K^{\text{nice}}_{\lambda^+}$
10388: can be reduced to basic types over a model in
10389: $K_\lambda$, i.e., locality}, in (2) if $a_\ell
10390: \in M_\ell \backslash M_0$ for $\ell = 1,2$
10391: and $[N \le_{\frak K} M_0 \and N \in K_\lambda \Rightarrow
10392: { \text{\rm \ortp\/}}(a_1,N,M_1) = { \text{\rm \ortp\/}}
10393: (a_2,N,M_2)]$, \ub{then} we can demand $f(a_1) = a_2$
10394: (so in particular ${\text{\rm \ortp\/}}(a_1,M_0,M_1) =
10395: { \text{\rm \ortp\/}}(a_2,M_0,M_2)$ where the types are as defined in
10396: ${\frak K}_{\lambda^+}$ and even in $(K_{\lambda^+},\le^*_{\lambda^+}$). \nl
10397: 4) Moreover in (2), assume further that for $\ell=1,2$, the following hold:
10398: $N_0 \le_{\frak K} N_\ell \le_{\frak K} M_\ell,
10399: N_0 \in K_\lambda,N_0 \le_{\frak K} M_0,
10400: N_\ell \in K_\lambda$ and $(\forall N \in K_\lambda)
10401: [N_0 \le_{\frak K} N \le_{\frak K} M_0 \rightarrow (\exists N' \in K_\lambda)
10402: (N \cup N_\ell \subseteq N' \le_{\frak K} M_\ell \wedge
10403: { \text{\rm NF\/}}_\lambda(N_0,N_\ell,N,N')]$. If $f_0$
10404: is an isomorphism from $N_1$ onto $N_2$ over $N_0$
10405: \ub{then} we can add $f \supseteq f_0$.
10406: \endproclaim
10407: \bigskip
10408:
10409: \demo{Proof} We first prove part (2). \newline
10410: 2) By \scite{600-ne.2}(1) + (2) there are representations $\bar M^\ell = \langle
10411: M^\ell_i:i < \lambda^+ \rangle$ of $M_\ell$ for $\ell < 3$
10412: such that for $\ell =1,2$ we have: $M^\ell_i \cap M_0 = M^\ell_0$
10413: and NF$_{\lambda,\langle 1,1,\kappa \rangle}
10414: (M^0_i,M^0_{i+1},M^\ell_i,M^\ell_{i+1})$ and \wilog \, $M^\ell_0$ is
10415: $(\lambda,\kappa)$-brimmed over $M^0_0$ for $\ell = 1,2$.
10416: \medskip
10417: \noindent
10418: Now we choose by induction on $i < \lambda^+$ an isomorphism $f_i$ from
10419: $M^1_i$ onto $M^2_i$, increasing with $i$ and being the identity over
10420: $M^0_i$. For $i=0$ use ``$M^\ell_0$ is $(\lambda,\kappa)$-brimmed over
10421: $M^0_0$ for $\ell = 1,2$" which we assume above.
10422: For $i$ limit take unions, for $i$ successor ordinal use
10423: uniqueness (Claim \scite{600-nf.7}).
10424: \enddemo
10425: \bigskip
10426:
10427: \demo{Proof of part (1)} By \scite{600-ne.3}(1) there
10428: are for $\ell = 1,2$ models
10429: $N^*_\ell \in K_{\lambda^+}$ such that $M_\ell <^+_{\lambda^+,\kappa}
10430: N^*_\ell$. Now let $\bar M^\ell = \langle M^\ell_i:i < \lambda^+ \rangle$
10431: be a representation of $M_\ell$ for $\ell = 0,1,2$ and let
10432: $\bar N^\ell = \langle N^\ell_i:i < \lambda^+ \rangle$ be a representation of
10433: $N^*_\ell$ for $\ell = 1,2$. By \scite{600-ne.3}(4) and \scite{600-ne.2}(2)
10434: without loss of generality $N^\ell_0$ is $(\lambda,\kappa)$-brimmed
10435: over $M^\ell_0$ and
10436: NF$_\lambda(M^0_i,M^0_{i+1},M^\ell_i,M^\ell_{i+1})$ and
10437: NF$_{\lambda,\langle 1,1,\kappa \rangle}
10438: (M^\ell_i,M^\ell_{i+1},N^\ell_i,N^\ell_{i+1})$ respectively for $\ell
10439: = 1,2$. Let $M^*_0$ be such that $p \in {\Cal S}^{\text{bs}}(M^*_0),M^*_0 \in
10440: K_\lambda,M^*_0 \le_{\frak K} M_0$; \wilog \, $M^*_0 \le_{\frak K}
10441: M^0_0$ and $a \in N^2_0$.
10442: Now $N^\ell_0$ is $(\lambda,\kappa)$-brimmed over
10443: $M^\ell_0$ hence over
10444: $M^0_0$ (for $\ell = 1,2$) so there is an isomorphism $f_0$ from
10445: $N^2_0$ onto $N^1_0$ extending id$_{M^0_0}$. There is
10446: $a' \in N^1_0$ such that \ortp$(a',M^1_0,N^1_0)$ is a non-forking
10447: extension of $p$ and \wilog \, $f_0(a)
10448: = a'$ hence \ortp$(f_0(a),M^1_0,N^1_0) \in
10449: {\Cal S}^{\text{bs}}(M^1_0)$ does not fork over $M^0_0$.
10450: \medskip
10451:
10452: We continue as in the proof of part (2). In the end
10453: $f = \dsize \bigcup_{i < \lambda^+} f_i$ is an isomorphism of $N^*_2$ onto
10454: $N^*_1$ over $M_0$ and as $f_0(a)$ is well defined and in $N^1_0 \backslash
10455: M^1_0$ clearly \ortp$(f(a),M^1_i,N^1_i)$ does not fork over
10456: $M^1_0$ and extends $p$ hence the pair
10457: $(N^*_1,f \restriction M_2)$ is as required.
10458: \enddemo
10459: \bigskip
10460:
10461: \demo{Proof of part (3), (4)} Like part (2). \hfill$\square_{\scite{600-ne.4}}$
10462: \enddemo
10463: \bigskip
10464:
10465: \proclaim{\stag{600-ne.5} Claim} 1) If $\delta$ is a limit ordinal
10466: $< \lambda^{+2}$ and $\langle M_i:i < \delta \rangle$ is a
10467: $\le^*_{\lambda^+}$-increasing continuous (in $K_{\lambda^+}$) and
10468: $M_\delta = \dbcu_{i < \delta} M_i$ (so $M_\delta \in K_{\lambda^+}$),
10469: \ub{then} $M_i \le^*_{\lambda^+} M_\delta$ for each $i < \delta$.
10470: \nl
10471: 2) If
10472: $\delta$ is a limit ordinal $< \lambda^{+2}$ and $\langle M_i:i < \delta
10473: \rangle$ is a $\le^*_{\lambda^+}$-increasing sequence, each $M_i$ is in
10474: $K^{\text{nice}}_{\lambda^+}$, \ub{then} $\dbcu_{i < \delta} M_i$ is
10475: in $K^{\text{nice}}_{\lambda^+}$. \nl
10476: 3) If $\delta$ is a limit ordinal $< \lambda^{+2}$ and $\langle M_i:i <
10477: \delta \rangle$ is a $<^+_{\lambda^+}$-increasing continuous
10478: (or just $<^*_{\lambda^+}$-increasing continuous, and $M_{2i+1}
10479: <^+_{\lambda^+} M_{2i+2}$ for $i < \delta$), \ub{then}
10480: $i < \delta \Rightarrow M_i <^+_{\lambda^+} \dbcu_{j < \delta} M_j$.
10481: \endproclaim
10482: \bigskip
10483:
10484: \demo{Proof} 1) We prove it by induction on $\delta$.
10485: Now if $C$ is a club of $\delta$, (as $\le^*_{\lambda^+}$ is transitive)
10486: then we can replace $\langle M_j:j < \delta \rangle$ by $\langle M_j:j \in
10487: C \rangle$ so without loss of generality $\delta = \text{ cf}(\delta)$, so
10488: $\delta \le \lambda^+$; similarly it is enough to prove $M_0
10489: \le^*_{\lambda^+} M_\delta :=
10490: \dsize \bigcup_{j < \delta} M_j$. For each $i \le \delta$ let
10491: $\langle M^i_\zeta:\zeta < \lambda^+ \rangle$ be a
10492: $<^*_{\frak K}$-representation of $M_i$.
10493: \enddemo
10494: \bigskip
10495: \noindent
10496: \underbar{Case A}: $\delta < \lambda^+$. \newline
10497: Without loss of generality (see \scite{600-ne.2}(1))
10498: for every $i < j < \delta$ and $\zeta < \lambda^+$ we have: \newline
10499: $M^j_\zeta \cap M_i = M^i_\zeta$ and NF$_\lambda(M^i_\zeta,M^i_{\zeta +1},
10500: M^j_\zeta,M^j_{\zeta +1})$. Let
10501: $M^\delta_\zeta = \dsize \bigcup_{i < \delta} M^i_\zeta$, so \newline
10502: $\langle M^\delta_\zeta:\zeta < \lambda^+ \rangle$
10503: is $\le_{\frak K}$-increasing continuous sequence of members of $K_\lambda$
10504: with limit $M_\delta$, and for $i < \delta,M^\delta_\zeta \cap M_i =
10505: M^i_\zeta$. By symmetry (see \scite{600-nf.14}) we have NF$_\lambda(M^i_\zeta,
10506: M^{i+1}_\zeta,M^i_{\zeta +1},M^{i+1}_{\zeta + 1})$ so as $\langle M^i_\zeta:
10507: i \le \delta \rangle$,$\langle M^i_{\zeta +1}:i \le \delta \rangle$ are
10508: $\le_{\frak K}$-increasing continuous, by \scite{600-nf.16} the transitivity
10509: of NF$_{\frak s}$, we know NF$_\lambda(M^0_\zeta,M^\delta_\zeta,
10510: M^0_{\zeta +1},M^\delta_{\zeta +1})$
10511: hence by symmetry (\scite{600-nf.14}) we have
10512: NF$_\lambda(M^0_\zeta,M^0_{\zeta +1},M^\delta_\zeta,M^\delta_{\zeta +1})$.
10513: \newline
10514: So $\langle M^0_\zeta:\zeta < \lambda^+ \rangle,\langle M^\delta_\zeta:
10515: \zeta < \lambda^+ \rangle$ are witnesses to $M_0 \le^*_{\lambda^+} M_\delta$.
10516: \bn
10517: \underbar{Case B}: $\delta = \lambda^+$.
10518:
10519: By \scite{600-ne.2}(1) (using normality of the club filter, restricting
10520: to a club of $\lambda^+$ and renaming), without loss
10521: of generality for $i < j \le 1 + \zeta < 1 + \xi < \lambda^+$ we have
10522: $M^j_\zeta \cap M_i = M^i_\zeta$, and NF$_\lambda(M^i_\zeta,M^i_\xi,M^j
10523: _\zeta,M^j_\xi)$. Let us define $M^{\lambda^+}_\zeta = \dsize \bigcup
10524: _{j < 1 + \zeta} M^j_\zeta$.
10525: So $\langle M^{\lambda^+}_\zeta:\zeta < \lambda^+ \rangle$ is
10526: a $<_{\frak K}$-representation of $M_{\lambda^+} = M_\delta$ and
10527: continue as before. \nl
10528: 2) Again without loss of generality $\delta = \text{ cf}(\delta)$ call it
10529: $\kappa$. Let $\langle M^i_\zeta:\zeta < \lambda^+ \rangle$ be a
10530: $<_{\frak K}$-representation of $M_i$ for $i < \delta$.
10531: \bn
10532: \ub{Case A}: $\delta = \kappa < \lambda^+$.
10533:
10534: Easy by now, yet we give details.
10535: So \wilog \, (see \scite{600-ne.2}(1)) for every $i < j < \delta$ and $\zeta <
10536: \xi < \lambda^+$ we have: $M^j_\zeta \cap M_i = M^i_\zeta$,
10537: NF$_\lambda(M^i_\zeta,M^i_\xi,M^j_\zeta,M^j_\xi)$ and $M^i_{\zeta +1}$
10538: is $(\lambda,\lambda)$-brimmed over $M^i_\zeta$. Let $M^\delta_\zeta =
10539: \dbcu_{\beta < \delta} M^\beta_\zeta$. Let $\xi < \lambda^+$.
10540: Now if $p \in {\Cal S}^{\text{bs}}(M^\delta_\xi)$
10541: then by the local character Axiom (E)(c) + the uniqueness Axiom
10542: (E)(e), for some $i < \delta,p$ does not fork over
10543: $M^i_\xi$.
10544: As $M_i$ is $\lambda^+$-saturated above $\lambda$, the type $p \restriction
10545: M^i_\xi$ is realized in
10546: $M_i$. So let $b \in M_i$ realize $p \restriction M^i_\xi$ and
10547: by Axiom $(E)(h)$, continuity, it
10548: suffices to prove that for every $j \in (i,\delta),b$ realizes
10549: $p \restriction M^j_\xi$ in $M_j$ which holds by \scite{600-nf.19}
10550: (note that $b \in M_i \le_{\frak K} M_j$ as $j \in [i,\delta)$).
10551: So $p$ is realized in $M_\delta = \dbcu_{i < \delta} M_i$. As this holds
10552: for every $\xi < \lambda^+$ and $p \in {\Cal S}^{\text{bs}}
10553: (M^\delta_\xi)$, the model $M_\delta$ is saturated.
10554: \bn
10555: \ub{Case B}: cf$(\delta) = \lambda^+$.
10556:
10557: Straight, in fact true for ${\frak K}$ a.e.c. with the
10558: $\lambda$-amalgamation property. \nl
10559: 3) Similar. \hfill$\square_{\scite{600-ne.5}}$
10560: \bigskip
10561:
10562: \proclaim{\stag{600-ne.6} Claim} 1) If
10563: $M_0 \in K_{\lambda^+}$ \ub{then} there is $M_1$
10564: such that $M_0 <^+_{\lambda^+} M_1 \in K^{\text{nice}}_{\lambda^+}$,
10565: and any such $M_1$ is universal over $M_0$ in $(K_{\lambda^+},
10566: \le^*_{\lambda^+})$.
10567: \nl
10568: 2) Assume $\boxtimes_{\bar N_1,\bar N_2,M_1,M_2}$ below holds. \ub{Then}
10569: $M_1 <^+_{\lambda^+} M_2$ \ub{iff} for every $\alpha < \lambda^+$ for
10570: stationarily many $\beta < \lambda^+$ there is $N$ such that
10571: $N^1_\beta \cup N^2_\alpha \subseteq N \le_{\frak K} N^2_\beta$ and
10572: $N^2_\beta$ is $(\lambda,*)$-brimmed over $N$ where
10573: \mr
10574: \item "{$\boxtimes_{\bar N_1,\bar N_2,M_1,M_2}$}" $M_1
10575: \le^*_{\lambda^+} M_2$ is being witnessed by $\bar N_1,\bar N_2$ that
10576: is $\bar N_\ell = \langle N^\ell_\alpha:\alpha < \lambda^+ \rangle$ is
10577: a $\le_{\frak K}$-representation of $M_\ell$ for $\ell =1,2$ and
10578: $\alpha < \lambda^+ \Rightarrow { \text{\rm NF\/}}_\lambda
10579: (N^1_\alpha,N^1_{\alpha +1},N^2_\alpha,N^2_{\alpha +1})$
10580: (hence $\alpha \le \beta < \lambda^+
10581: \Rightarrow { \text{\rm NF\/}}_\lambda
10582: (N^1_\alpha,N^1_\beta,N^2_\alpha,N^2_\beta))$.
10583: \endroster
10584: \endproclaim
10585: \bigskip
10586:
10587: \demo{Proof} 1) The existence by \scite{600-ne.3}(1). Why ``any such
10588: $M_1,\ldots$?" if $M_0 \le^*_{\lambda^+}
10589: M_2$ then for some $M^+_2 \in K^{\text{nice}}_{\lambda^+}$ we have $M_2
10590: <^+_\lambda M^+_2 \in K^{\text{nice}}_{\lambda^+}$ so $M_0 \le^*
10591: _{\lambda^+} M_1 <^+_{\lambda^+} M^+_2$ hence by \scite{600-ne.3}(5) we have
10592: $M_0 <^+_\lambda M^+_2$; so by \scite{600-ne.4}(2) the models
10593: $M^+_2,M_1$ are isomorphic over $M_0$, so $M_2$ can be
10594: $\le^*_{\lambda^+}$-embedded into $M_1$ over $M_0$, so we are
10595: done. \nl
10596: 2) Not hard. \hfill$\square_{\scite{600-ne.6}}$
10597: \enddemo
10598: \newpage
10599:
10600: \head {\S8 Is $\le^*_{\lambda^+}$ equal to $\le_{\frak K}$ on
10601: $K^{\text{nice}}_{\lambda^+}$?} \endhead \resetall
10602: \spuriousreset
10603: \bigskip
10604:
10605: \demo{\stag{600-rg.0} Hypothesis} The hypothesis \scite{600-nf.0}.
10606:
10607: An important issue is whether
10608: $(K^{\text{nice}}_{\lambda^+},\le^*_{\lambda^+})$ satisfies Ax IV of
10609: a.e.c. So a model $M \in K_{\lambda^{++}}$ may be the union of a
10610: $\le^*_{\lambda^+}$-increasing chain of length $\lambda^{++}$, but we
10611: still do not know if there is a continuous such sequence.
10612:
10613: E.g. let $\langle M_\alpha:\alpha < \lambda^{++}\rangle$ be
10614: $\le^*_{\lambda^+}$-increasing with union $M \in K_{\lambda^{++}}$ let
10615: $M'_n = M_n,M'_{\omega + \alpha +1} = M_{\omega + \alpha}$ and
10616: $M'_\delta = \cup \{M_\beta:\beta < \delta\}$ for $\delta$ limit. So
10617: $\langle M'_\alpha:\alpha < \lambda^{++}\rangle$ is $\le_{\frak
10618: K}$-increasing continuous, $\langle M'_{\alpha +1}:\alpha <
10619: \lambda^{++}\rangle$ is $\le^*_{\lambda^+}$-increasing, but we do not
10620: know whether $M'_\delta \le^*_{\lambda^+} M'_{\delta +1}$ for limit
10621: $\delta < \lambda^{++}$.
10622: \enddemo
10623: \bigskip
10624:
10625: \definition{\stag{600-sg.1} Definition} Let $M \in {\frak K}_{\lambda^{++}}$
10626: be the union of an $\le_{\frak K}$-increasing chain from
10627: $(K^{\text{nice}}_{\lambda^+},\le^*_{\lambda^+})$ or just
10628: $(K_{\lambda^+},\le^*_{\lambda^+}),
10629: \bar M = \langle M_i:i < \lambda^{++} \rangle$ such that
10630: $\langle M_i:i < \lambda^{++}$ non-limit$\rangle$ is
10631: $\le^*_{\lambda^+}$-increasing. \nl
10632: 1) Let $S(\bar M) = \{\delta:M_\delta \nleq^*_{\lambda^+}
10633: M_{\delta+1}$ (see \scite{600-sg.2}(3) below)$\}$, so
10634: $S(\bar M) \subseteq \lambda^{++}$.
10635: \nl
10636: 2) For such $M$ let $S(M)$ be $S(\bar M)/{\Cal D}_{\lambda^{++}}$
10637: where $\bar M$ is a $\le_{\frak K}$-representation of $M$ and
10638: ${\Cal D}_{\lambda^{++}}$ is the club filter on $\lambda^{++}$; it
10639: is well defined by \scite{600-sg.2} below. \nl
10640: 3) We say $\langle M_i:i < \delta \rangle$ is non-limit
10641: $<^*_{\lambda^+}$-increasing \ub{if} for non-limit $i <j < \delta$ we have
10642: $M_i \le^*_{\lambda^+} M_j$.
10643: \enddefinition
10644: \bigskip
10645:
10646: \proclaim{\stag{600-sg.2} Claim} 1) If $\bar M^\ell = \langle M^\ell_i:i <
10647: \lambda^{++}\rangle$ for
10648: $\ell \in \{1,2\}$ is $\le_{\frak K}$-increasing continuous and
10649: $i < j < \lambda^{++} \Rightarrow M_0 \le^*_{\lambda^+} M_{i+1}
10650: \le^*_{\lambda^+} M_{j+1}$ and
10651: $M = \dbcu_{i < \lambda^{++}} M^1_i = \dbcu_{i < \lambda^{++}} M^2_i$
10652: has cardinality $\lambda^{++}$ \ub{then} $S(\bar M^1) = S(\bar M^2)$
10653: {\rm mod} ${\Cal D}_{\lambda^{++}}$. \nl
10654: 2) If $M,\bar M$ are as in \scite{600-sg.1} hence $M = \dbcu_{i < \lambda^{++}}
10655: M_i$ \ub{then}
10656: $S(\bar M)/{\Cal D}_{\lambda^{++}}$ depends just on $M/\cong$. \nl
10657: 3) If $\bar M$ is as in \scite{600-sg.1} and $i < j < \lambda^{++}$,
10658: \ub{then} $M_i \le^*_{\lambda^+} M_{i+1} \Leftrightarrow M_i
10659: \le^*_{\lambda^+} M_j$. \nl
10660: 4) If $M \in {\frak K}_{\lambda^{++}}$ is the union of a
10661: $\le^*_{\lambda^+}$-increasing chain from
10662: $(K^{\text{nice}}_{\lambda^+},\le^*_{\lambda^+})$, not necessarily
10663: continuous, \ub{then} there is $\bar M$ as in Defintion \scite{600-sg.1},
10664: that is $\bar M = \langle M_i:i < \lambda^{++}
10665: \rangle$, a $\le_{\frak K}$-representation of $M$ with $M_i
10666: \le^*_{\lambda^+} M_j$ for non-limit $i<j$.
10667: \endproclaim
10668: \bigskip
10669:
10670: \demo{Proof} 1) We can find a club $E$ of $\lambda^{++}$ consisting
10671: of limit ordinals such that $i \in E \Rightarrow M^1_i = M^2_i$. Now
10672: if $\delta_1 < \delta_2$ are from $E$ then $\delta_1 \in S(\bar M^1)
10673: \Leftrightarrow M^1_{\delta_1} \le^*_{\lambda^+} M^1_{\delta_1+1}
10674: \Leftrightarrow M^1_{\delta_1} \le^*_{\lambda^+} M^1_{\delta_2}
10675: \Leftrightarrow M^2_{\delta_1} \le^*_{\lambda^+} M^2_{\delta_2}
10676: \Leftrightarrow M^2_{\delta_1} \le^*_{\lambda^+} M^2_{\delta_1+1}
10677: \Leftrightarrow \delta_1 \in S(M^2)$.
10678: \nl
10679: [Why? By the definition of $S(\bar M^1)$, by part (3), by
10680: ``$\delta_1,\delta_2 \in E$", by part (3), by the definition of
10681: $S(\bar M^2)$, respectively.] So we are done.
10682: \nl
10683: 2) Follows by part (1). \nl
10684: 3) The implication $\Leftarrow$ is by \scite{600-ne.3}(3); for the implication
10685: $\Rightarrow$, note that assuming $M_i <^*_{\lambda^+} M_{i+1}$, as
10686: $\le^*_{\lambda^+}$ is a partial order, noting that by the assumption
10687: on $\bar M$ we have $M_{i+1} \le^*_{\lambda^+} M_{j+1}$,
10688: and by \scite{600-ne.3}(3) we are done. \nl
10689: 4) Trivial. \hfill$\square_{\scite{600-sg.2}}$
10690: \enddemo
10691: \bigskip
10692:
10693: \proclaim{\stag{600-sg.3} Claim} If $(*)$
10694: below holds \ub{then} for every stationary
10695: $S \subseteq S^{\lambda^{++}}_{\lambda^+} (= \{\delta < \lambda^{++}:
10696: {\text{\rm cf\/}}(\delta) =
10697: \lambda^+\})$ for some $\lambda^+$-saturated $M \in K_{\lambda^{++}}$
10698: we have $S(M)$ is well defined and equal
10699: to $S/{\Cal D}_{\lambda^{++}}$, where
10700: \mr
10701: \item "{$(*)$}" we can find $\langle M_i:i \le \lambda^+ +1 \rangle$
10702: which is $<_{\frak K}$-increasing continuous of members of
10703: $K^{\text{nice}}_{\lambda^+}$ such that
10704: $i < j \le \lambda^+ +1 \and (i,j) \ne
10705: (\lambda^+,\lambda^+ + 1) \Rightarrow M_i <^+_{\lambda^+} M_j$ but
10706: $\neg(M_{\lambda^+} \le^*_{\lambda^+} M_{\lambda^+ +1})$.
10707: \endroster
10708: \endproclaim
10709: \bigskip
10710:
10711: \demo{Proof} Fix
10712: $S \subseteq S^{\lambda^{++}}_{\lambda^+}$ and $\langle M_i:i \le
10713: \lambda^+ +1 \rangle$ as in $(*)$. \nl
10714: Without loss of generality $|M_{\lambda^+ +1} \backslash M_{\lambda^+}|
10715: = \lambda^+$.
10716:
10717: We choose by induction on $\alpha < \lambda^{+2}$ a model $M^S_\alpha$
10718: such that:
10719: \mr
10720: \item "{$(a)$}" $M^S_\alpha \in K^{\text{nice}}_{\lambda^+}$
10721: has universe an ordinal $< \lambda^{++}$
10722: \sn
10723: \item "{$(b)$}" for $\beta < \alpha$ we have $M^S_\beta \le_{\frak K}
10724: M^S_\alpha$
10725: \sn
10726: \item "{$(c)$}" if $\alpha = \beta + 1$, $\beta \notin S$ then $M^S_\beta
10727: <^+_{\lambda^+} M^S_\alpha$
10728: \sn
10729: \item "{$(d)$}" if $\alpha = \beta + 1,\beta \in S$ then $(M^S_\beta,
10730: M^S_\alpha) \cong (M_{\lambda^+},M_{\lambda^+ +1})$
10731: \sn
10732: \item "{$(e)$}" if $\beta < \alpha,\beta \notin S$ then $M^S_\beta
10733: \le^+_{\lambda^+} M^S_\alpha$
10734: \sn
10735: \item "{$(f)$}" if $\alpha$ is a limit ordinal, then $M_\alpha =
10736: \cup\{M_\beta:\beta < \alpha\}$.
10737: \ermn
10738: We use freely the transitivity and continuity of $\le^*_\lambda$
10739: and of $<^+_\lambda$. \nl
10740: \sn
10741: For $\alpha = 0$ no problem.
10742: \mn
10743: For $\alpha$ limit no problem; choose an increasing continuous
10744: sequence $\langle \gamma_i:i < \text{\rm cf}(\alpha)\rangle$ of
10745: ordinals with limit $\alpha$ each of cofinality $< \lambda,\gamma_i
10746: \notin S$, and use \scite{600-ne.5}(3) for clause (e).
10747: \mn
10748: For $\alpha = \beta +1,\beta \notin S$ no problem.
10749: \mn
10750: For $\alpha = \beta + 1,\beta \in S$ so cf$(\beta) = \lambda^+$, let
10751: $\langle \gamma_i:i < \lambda^+ \rangle$ be increasing continuous with limit
10752: $\beta$ and cf$(\gamma_i) \le \lambda$, hence $\gamma_i \notin S$ and
10753: each $\gamma_{i+1}$ a successor ordinal. By
10754: clause (e) above and \scite{600-ne.3}(5) we have
10755: $M^S_{\gamma_i} <^+_{\lambda^+}
10756: M^S_{\gamma_{i+1}}$, hence $\langle M_{\gamma_i}:i < \lambda^+ \rangle$
10757: is $<^+_{\lambda^+}$-increasing continuous. Now there is an
10758: isomorphism $f_\beta$ from $M_{\lambda^+}$ onto $M^S_\beta$
10759: mapping $M_i$ onto $M^S_{\gamma_i}$ for $i < \lambda$ (why? choose
10760: $f_\beta \restriction M_i$ by induction on $i$, for $i=0$ by
10761: \scite{600-ne.2}(0), for $i$ successor $M^S_{\gamma_i} <^+_\lambda
10762: M^S_{\gamma_{i+1}}$ by \scite{600-ne.3}(3) as $M^S_{\gamma_i}
10763: <^*_{\lambda^+} M^S_{\gamma_{i+1}} <^+_{\lambda^+} M^S_{\gamma_{i+1}}$
10764: so we can use \scite{600-ne.4}(2)). So we can choose a
10765: one-to-one function $f_\alpha$ from $M_{\lambda^+ +1}$ onto some
10766: ordinal $< \lambda^{++}$ extending
10767: $f_\beta$ and let $M_\alpha = f_\alpha(M_{\lambda^+ +1})$.
10768:
10769: Finally having carried the induction, let
10770: $M_S = \dbcu_{\alpha < \lambda^{+2}} M^S_\alpha$,
10771: it is easy to check that $M_S \in K_{\lambda^{++}}$ is $\lambda^+$-saturated
10772: and $\bar M = \langle M^S_\alpha:\alpha < \lambda^{++} \rangle$
10773: witnesses that
10774: $S(M_S)/{\Cal D}_{\lambda^{++}}$ is well defined and
10775: $S(M_S)/{\Cal D}_{\lambda^{++}} = S(\langle M^S_\alpha:
10776: \alpha < \lambda^{++} \rangle)/{\Cal D}_{\lambda^{++}} =
10777: S/{\Cal D}_{\lambda^{++}}$ as required. \hfill$\square_{\scite{600-sg.3}}$
10778: \enddemo
10779: \bn
10780: Below we prove that some versions of non-smoothness are equivalent.
10781: \proclaim{\stag{600-sg.10} Claim} 1) We have $(**)_{M^*_1,M^*_2} \Rightarrow
10782: (***)$ (see below). \nl
10783: 2) If $(*)$ then $(**)_{M^*_1,M^*_2}$ for some $M^*_1,M^*_2$ and
10784: trivially $(***) \Rightarrow (*)$.
10785: \nl
10786: 3) In part (1) we get $\langle M_i:i \le \lambda^+ +1 \rangle$ as in
10787: $(***)$, see below,
10788: such that $M_{\lambda^+} = M^*_1,M_{\lambda^+ +1} = M^*_2$ if we waive
10789: $i < \lambda^+ \Rightarrow M_i <^+_\lambda M_{\lambda+1}$ or assume
10790: $M^*_1 <_{\frak K} M^* <^+_\lambda M^*_2$ for some $M^*$.
10791: \nl
10792: 4) If $M^*_1 \le^*_{\lambda^+} M^*_2$ and $M^*_1
10793: \in K^{\text{nice}}_{\lambda^+}$
10794: and $N_1 \le_{\frak K} N_2 \in K_\lambda,N_\ell \le M^*_\ell$ for
10795: $\ell =1,2$ and $p \in
10796: {\Cal S}^{\text{bs}}(N_2)$ does not fork over $N_1$ \ub{then} some $c \in
10797: M^*_1$ realizes $p$ \nl
10798: \ub{where}
10799: \smallskip
10800:
10801: $(*) \quad$ there are limit $\delta < \lambda^{++},N$ and $\bar M =
10802: \langle M_i:i \le \delta \rangle$ a $\le^*_{\lambda^+}$-increasing \nl
10803:
10804: \hskip33pt continuous sequence with
10805: $M_i,N \in K^{\text{nice}}_{\lambda^+}$ such that: $M_i \le^*_{\lambda^+} N
10806: \Leftrightarrow i < \delta$
10807: \sn
10808:
10809: $(**)_{M^*_1,M^*_2} \quad (i) \quad
10810: M^*_1 \in K^{\text{nice}}_{\lambda^+},
10811: M^*_2 \in K^{\text{nice}}_{\lambda^+}$
10812: \sn
10813:
10814: \hskip58pt $(ii) \quad M^*_1 \le_{\frak K} M^*_2$
10815: \sn
10816:
10817: \hskip58pt $(iii) \quad M^*_1 \nleq^*_{\lambda^+} M^*_2$
10818: \sn
10819:
10820: \hskip58pt $(iv) \quad$ if
10821: $N_1 \le_{\frak K} N_2$ are from $K_\lambda$, \nl
10822:
10823: \hskip80pt $N_\ell \le_{\frak K} M^*_\ell$ for $\ell =1,2$ and
10824: $p \in {\Cal S}^{\text{bs}}(N_2)$ does not \nl
10825:
10826: \hskip80pt fork over $N_1$,
10827: \ub{then} some $a \in M^*_1$ realizes $p$ in $M^*_2$
10828: \sn
10829:
10830: $(***) \quad$ there is $\bar M = \langle M_i:i \le \lambda^+ +1
10831: \rangle,\le_{\frak K}$-increasing continuous, every \nl
10832:
10833: \hskip40pt $M_i \in K^{\text{nice}}_{\lambda^+}$
10834: and $M_{\lambda^+} \nleq^*_{\lambda^+} M_{\lambda^+ +1}$ but \nl
10835:
10836: \hskip40pt $i < j \le \lambda^+ +1 \and i \ne
10837: \lambda^+ \Rightarrow M_i <^+_{\lambda^+} M_j$.
10838: \endproclaim
10839: \bigskip
10840:
10841: \demo{Proof} 1),3) Let $\langle a^\ell_i:i < \lambda^+ \rangle$ list the
10842: elements of $M^*_\ell$ for $\ell = 1,2$.
10843: Let $\langle N^*_{2,i}:i < \lambda^+ \rangle$ be a
10844: $\le_{\frak K}$-representation of $M^*_2$.
10845: \nl
10846: Let $\langle (p_\zeta,N^*_\zeta,\gamma_\zeta):\zeta < \lambda^+ \rangle$
10847: list the triples $(p,N,\gamma)$ such that $\gamma < \lambda^+,p \in
10848: {\Cal S}^{\text{bs}}(N),N \in \{N^*_{2,i}:i < \lambda^+\}$ with each
10849: such triple appearing $\lambda^+$ times. By induction on $\alpha
10850: < \lambda^+$ we choose $\langle N^\alpha_i:i \le \alpha \rangle,
10851: N_\alpha$ such that:
10852: \mr
10853: \item "{$(a)$}" $N^\alpha_i \in K_\lambda$ and $N^\alpha_i \le_{\frak K}
10854: M^*_1$ \nl
10855: \sn
10856: \item "{$(b)$}" $N_\alpha \le_{\frak K} M^*_2$
10857: \sn
10858: \item "{$(c)$}" $\langle N^\alpha_i:i \le \alpha \rangle$ is
10859: $\le_{\frak K}$-increasing continuous
10860: \sn
10861: \item "{$(d)$}" $N^\alpha_\alpha \le_{\frak K} N_\alpha,N_\alpha \cap
10862: M^*_1 = N^\alpha_\alpha$
10863: \sn
10864: \item "{$(e)$}" if $i \le \alpha$ then $\langle N^\beta_i:\beta \in [i,\alpha]
10865: \rangle$ is $\le_{\frak K}$-increasing continuous
10866: \sn
10867: \item "{$(f)$}" $\langle N_\beta:\beta \le \alpha \rangle$ is
10868: $\le_{\frak K}$-increasing continuous
10869: \sn
10870: \item "{$(g)$}" if
10871: $\alpha = \beta +1,i \le \beta$ then NF$_\lambda(N^\beta_i,
10872: N_\beta,N^\alpha_i,N_\alpha)$
10873: \sn
10874: \item "{$(h)$}" if $\alpha = 2\beta +1$ then $a^2_\beta \in N_{\alpha +1}$
10875: \sn
10876: \item "{$(i)$}" if $\alpha = 2\beta +2$ and $i < \alpha$ then
10877: $N^\alpha_{i+1}$ is brimmed over $N^\alpha_i \cup N^{2 \beta
10878: +1}_{i+1}$ and $N^\alpha_0$ is brimmed over $N^{2\beta}_0$.
10879: \endroster
10880: \bn
10881: \ub{Why is this enough}?
10882:
10883: We let $M_{\lambda^+} = M^*_1,M_{\lambda^+ +1} = M^*_2$ and let
10884: $M'_{\lambda^+ +1} \in K^{\text{nice}}_{\lambda^+}$ be such that
10885: $M_{\lambda^+ +1} <^+_{\lambda^+} M'_{\lambda^+ +1}$ and for
10886: $i < \lambda^+$ we let
10887: $M_i = \cup \{N^\alpha_i:\alpha \in [i,\lambda^+)\}$; now
10888: \mr
10889: \item "{$(\alpha)$}" $M^*_1 = \dbcu_{\alpha < \lambda^+} N^\alpha_\alpha
10890: = \dbcu_{i < \lambda^+} M_i$
10891: and $M^*_2 = \dbcu_{\alpha < \lambda^+} N_\alpha$ \nl
10892: [why? the second by clause (h) (and (b) of course),
10893: the first as $N_\alpha \cap M^*_1 = N^\alpha_\alpha$].
10894: \ermn
10895: Now:
10896: \mr
10897: \item "{$(\beta)$}" $\langle M_i:i \le \lambda^+ +1 \rangle$ is
10898: $\le_{\frak K}$-increasing continuous \nl
10899: [trivial by clauses (c) + (e) if $i < \lambda^+$ and (d) if $i
10900: = \lambda^+$]
10901: \sn
10902: \item "{$(\gamma)$}" for $i < \lambda^+,M_i$ is saturated, i.e.,
10903: $\in K^{\text{nice}}_{\lambda^+}$. \nl
10904: [Why? Clearly $\langle N^\alpha_i:\alpha \in (i,\lambda^+)\rangle$ is
10905: a $\le_{\frak K}$-representation of $M_i$ by clause (e) and the choice
10906: of $M_i$. If $i=0$ this follows by clauses (i) + (e).
10907: If $i=j+1$ this follows by clauses (e) + (i). If $i$ is a
10908: limit ordinal use \scite{600-ne.5}(2) and clause (g)]
10909: \sn
10910: \item "{$(\delta)$}" for $i < \lambda^+,i < j \le \lambda^+ +1$ we
10911: have $M_i \le^*_{\lambda^+} M_j$. \nl
10912: [Why? Let $N^\alpha_{\lambda^+} :=
10913: N^\alpha_\alpha,N^\alpha_{\lambda^+ +1} = N_\alpha$ for $\alpha <
10914: \lambda^+$ and let $\gamma$ be $i$ if $j = \lambda^+,\lambda^+ +1$ and
10915: be $j$ if $j < \lambda^+$; so in any case $\gamma < \lambda^+$. Now
10916: as $\langle N^\alpha_i:\alpha \in [\gamma,\lambda^+)\rangle$ is a
10917: $\le_{\frak K}$-representation of $M_i$ and $\langle N^\alpha_j:\alpha
10918: \in [\gamma,\lambda^+) \rangle$ is a $\le_{\frak K}$-representation of
10919: $M_j$ and if $\gamma \le \beta < \lambda^+$ then by clause (g) we have
10920: NF$_\lambda(N^\beta_i,N_\beta,N^{\beta +1}_i,N_{\beta +1})$ hence by
10921: symmetry NF$_\lambda(N^\beta_i,N^{\beta +1}_i,N_\beta,N_{\beta +1})$
10922: hence by monotonicity \nl
10923: NF$_\lambda(N^\beta_i,N^{\beta
10924: +1}_i,N^\beta_j,N^{\beta +1}_j)$; this suffices]
10925: \sn
10926: \item "{$(\varepsilon)$}" if $i < j \le \lambda^+$ then
10927: $M_i <^+_{\lambda^+} M_j$ \nl
10928: [why? by \scite{600-ne.5}(1) it suffices to prove this in the cases
10929: $j=i+1$. Now claim \scite{600-ne.6}(2), clause (i) guaranteed this.]
10930: \ermn
10931: Clearly $\langle M_i:i \le \lambda^+ +1 \rangle$ is as required for
10932: part (3) and $\langle M_i:i \le \lambda^+ \rangle \char 94 \langle
10933: M'_{\lambda^+ +1}\rangle$ is as required in part (1).
10934: \nl
10935: So we are done.
10936: \bn
10937: So let us carry the construction.
10938: \mn
10939: For $\alpha = 0$ trivially.
10940: \mn
10941: For $\alpha$ limit: straightforward.
10942: \mn
10943: For $\alpha = 2 \beta + 1$ we let $N^\alpha_i = N^{2 \beta}_i$ for $i \le
10944: 2 \beta$ and $N_\alpha \in K_\lambda$ is chosen such that $N_{2 \beta} \cup
10945: \{a^2_\beta\} \subseteq N_\alpha \le_{\frak K} M^*_2$ and $N_\alpha
10946: \restriction M^*_1 \le_{\frak K} M^*_1$, easy by the properties of abstract
10947: elementary class and we let $N^\alpha_{2 \beta +1} = N_\alpha \restriction
10948: M^*_1$.
10949: \mn
10950: For $\alpha = 2 \beta + 2$ we choose by induction on
10951: $\varepsilon < \lambda^2$, a triple $(N^\oplus_{\alpha,\varepsilon},
10952: N^\otimes_{\alpha,\varepsilon},a_{\alpha,\varepsilon})$ such that:
10953: \mr
10954: \item "{$(A)$}" $N^\otimes_{\alpha,\varepsilon} \le_{\frak K} M^*_2$
10955: belongs to $K_\lambda$ and is $\le_{\frak K}$-increasing continuous with
10956: $\varepsilon$
10957: \sn
10958: \item "{$(B)$}" $N^\otimes_{\alpha,0} = N_{2 \beta +1}$ and
10959: $N^\otimes_{\alpha,\varepsilon} \restriction M^*_1 \le^*_{\frak K} M^*_1$
10960: \sn
10961: \item "{$(C)$}" $N^\oplus_{\alpha,\varepsilon} \le_{\frak K} M^*_1$ belongs
10962: to $K_\lambda$ and is $\le_{\frak K}$-increasing continuous with
10963: $\varepsilon$
10964: \sn
10965: \item "{$(D)$}" $N^\oplus_{\alpha,0} = N^{2 \beta +1}_{2 \beta +1}$
10966: \sn
10967: \item "{$(E)$}" $(N^\oplus_{\alpha,\varepsilon},
10968: N^\oplus_{\alpha,\varepsilon + 1},
10969: a_{\alpha,\varepsilon}) \in K^{3,\text{uq}}_\lambda$
10970: \sn
10971: \item "{$(F)$}" \ortp$(a_{\alpha,\varepsilon},N^\otimes_{\alpha,\varepsilon},
10972: M^*_2)$ does not fork over $N^\oplus_{\alpha,\varepsilon}$
10973: \sn
10974: \item "{$(G)$}" $N^\oplus_{\alpha,\varepsilon} \le_{\frak K}
10975: N^\otimes_{\alpha,\varepsilon}$
10976: \sn
10977: \item "{$(H)$}" for every $p \in {\Cal S}^{\text{bs}}
10978: (N^\oplus_{\alpha,\varepsilon})$ for some odd $\zeta \in [\varepsilon,
10979: \varepsilon + \lambda)$ the type \ortp$(a_{\alpha,\zeta},
10980: N^\otimes_{\alpha,\zeta},N^\otimes_{\alpha,\zeta +1})$ is a
10981: non-forking extension of $p$.
10982: \ermn
10983: No problem to carry this. [Why? For $\varepsilon =0$ and
10984: $\varepsilon$ limit there are no problems. In stage $\varepsilon +1$
10985: by bookkeeping gives you a type $p_\varepsilon \in {\Cal
10986: S}^{\text{bs}}(N^\oplus_{\alpha,\varepsilon})$ and let $q_\varepsilon
10987: \in {\Cal S}^{\text{bs}}(N^\otimes_{\alpha,\varepsilon})$ be a
10988: non-forking extension of $p_\varepsilon$. By assumption (iv) of
10989: $(**)_{M^*_1,M^*_2}$ there is an element $a_{\alpha,\varepsilon} \in
10990: M^*_1$ realizing $q_\varepsilon$. Now $M^*_1$ is saturated hence
10991: there is a model $N^\oplus_{\alpha,\varepsilon +1} \in K_\lambda$ such
10992: that $N^\oplus_{\alpha,\varepsilon +1} \le_{\frak K} M^*_1$ and
10993: $(N^\oplus_{\alpha,\varepsilon},N^\oplus_{\alpha,\varepsilon
10994: +1},a_{\alpha,\varepsilon}) \in K^{3,\text{uq}}_\lambda$.
10995:
10996: Lastly, choose $N^\otimes_{\alpha,\varepsilon +1}$ satisfying
10997: clauses (A),(B),(G) so we have carried the induction on
10998: $\varepsilon$.] Note that NF$_\lambda(N^\oplus_{\alpha,\varepsilon},
10999: N^\otimes_{\alpha,\varepsilon},N^\oplus_{\alpha,\varepsilon
11000: +1},N^\otimes_{\alpha,\varepsilon +1})$ for each $\varepsilon <
11001: \lambda^2$ by clauses (E),(F) of \scite{600-nf.20},
11002: hence NF$(N^{2 \beta +1}_{2\beta +1},N_{2 \beta +1},
11003: \cup\{N^\oplus_{\alpha,\varepsilon}:\varepsilon < \lambda^2\},
11004: \cup\{N^\otimes_{\alpha,\varepsilon}:\varepsilon <
11005: \lambda^2\})$ by \scite{600-nf.16} as
11006: $(N^\oplus_{\alpha,0},N^\otimes_{\alpha,0}) = (N^{2 \beta +1}_{2
11007: \beta+1},N_{2 \beta+1})$ and the sequence $\langle
11008: N^\oplus_{\alpha,\varepsilon}:\varepsilon < \lambda^+\rangle,\langle
11009: N^\otimes_{\alpha,\varepsilon}:\varepsilon < \lambda^+\rangle$ are
11010: increasing continuous.
11011:
11012: Now let $N_\alpha = \bigcup \{N^\otimes_{\alpha,\varepsilon}:
11013: \varepsilon < \lambda^2\},N^\alpha_\alpha = N_\alpha \cap M^*_1$
11014: recalling clauses (A)+(B).
11015:
11016: Now $\cup\{N^\oplus_{\alpha,\varepsilon}:\varepsilon < \lambda^2\}
11017: \le_{\frak K} M^*_1$ is $(\lambda,*)$-brimmed over $N^{2 \beta +1}
11018: _{2 \beta +1}$ by \scite{600-4a.2} (and clause (H) above). Hence
11019: there is no problem to choose
11020: $N^\alpha_i \le_{\frak K} N^\alpha_\alpha$
11021: for $i \le 2 \beta +1$ as required, that is $N^{2 \beta+1}_i
11022: \le_{\frak K} N^\alpha_i,\langle N^\alpha_i:i \le 2 \beta +1 \rangle$
11023: is $\le_{\frak K}$-increasing continuous, NF$_\lambda(N^{2 \beta
11024: +1}_i,N^{2 \beta +1}_{i+1},N^\alpha_i,N^\alpha_{i+1})$ and $N^\alpha_{i+1}$ is
11025: $(\lambda,*)$-brimmed over $N^{2 \beta +1}_{i+1} \cup N^\alpha_i$ and
11026: $N^\alpha_0$ is $(\lambda,*)$-brimmed over $N^{2 \beta +1}_0$.
11027: So we have finished the induction step on $\alpha = 2 \beta +2$.
11028:
11029: Having carried the induction we are done.
11030: \nl
11031: 2) So assume $(*)$ and let $M_{\delta +1} := N$ from $(*)$.
11032: It is enough to prove that $(**)_{M_\delta,M_{\delta +1}}$ holds. Clearly
11033: clauses (i), (ii), (iii) hold, so we should prove (iv).
11034: Without loss of generality $\delta =
11035: \text{ cf}(\delta)$ so $\delta = \lambda^+$ or $\delta \le \lambda$.
11036: For $i \le \delta + 1$ let $\langle M_{i,\alpha}:\alpha <
11037: \lambda^+\rangle$ be a $\le_{\frak K}$-representation
11038: of $M_i$ and for $i < \delta,j \in
11039: (i,\delta +1]$ let $E_{i,j}$ be a club of $\lambda^+$ witnessing
11040: $M_i \le^*_{\lambda^+} M_j$ for $\bar M^i,\bar M^j$.
11041: First assume $\delta \le \lambda$. Let $E = \cap\{E_{i,j}:i < \delta,j \in
11042: (i,\delta +1]\}$, it is a club of $\lambda^+$. So assume
11043: $N_2 \le_{\frak K} M_{\delta +1},N_1 \le_{\frak K} N_2,N_1
11044: \le_{\frak K} M_\delta$ and $N_1,N_2 \in K_\lambda$ and
11045: $p \in {\Cal S}^{\text{bs}}(N_2)$ does not fork over
11046: $N_1$. We can choose $\zeta \in E$ such that $N_2 \subseteq
11047: M_{\delta +1,\zeta}$, let $p_1 \in {\Cal S}^{\text{bs}}(M_{\delta +1,\zeta})$
11048: be a non-forking extension of $p$, so $p_1$ does not fork over $N_1$
11049: hence (by monotonicity)
11050: over $M_{\delta,\zeta}$ so $p_2 := p_1 \restriction M_{\delta,\zeta} \in
11051: {\Cal S}^{\text{bs}}(M_{\delta,\zeta})$. By Axiom $(E)(c)$
11052: for some $\alpha < \delta,p_2$ does not fork over $M_{\alpha,\zeta}$ hence
11053: $p_2 \restriction M_{\alpha,\zeta} \in {\Cal S}^{\text{bs}}
11054: (M_{\alpha,\zeta})$. As $M_\alpha \in K^{\text{nice}}_{\lambda^+}$,
11055: i.e., $M_\alpha$ is $\lambda^+$-saturated (above $\lambda$), clearly
11056: for some $\xi \in (\zeta,\lambda^+) \cap E$
11057: some $c \in M_{\alpha,\xi}$ realizes $p_2 \restriction M_{\alpha,\zeta}$
11058: but NF$_\lambda(M_{\alpha,\zeta},M_{\delta +1,\zeta},M_{\alpha,\xi},
11059: M_{\delta +1,\xi})$ hence by \scite{600-nf.19} we know that
11060: \ortp$(c,M_{\delta +1,\zeta},
11061: M_{\delta +1,\xi})$ belongs to ${\Cal S}^{\text{bs}}(M_{\delta +1,\zeta})$ and
11062: does not fork over $M_{\alpha,\zeta}$ hence $c$ realizes
11063: $p_2$ and even $p_1$ hence $p$ and we are done. \nl
11064:
11065: Second, assume $\delta = \lambda^+$, then for some $\delta^* < \delta$
11066: we have $N_1 \le_{\frak K} M_{\delta^*}$, and use the proof above for
11067: $\langle M_i:i \le \delta^* \rangle,M_{\delta +1}$ (or use
11068: $M_{\delta^*} \le^*_{\lambda^+} M_{\delta +1}$). \nl
11069: 4) Straight, in fact included the proof of \scite{600-ne.5}(2).
11070: \hfill$\square_{\scite{600-sg.10}}$
11071: \enddemo
11072: \bn
11073: The definition below has affinity to ``blowing ${\frak K}_\lambda$ to
11074: ${\frak K}^{\text{up}}_\lambda$" in \S1.
11075: \definition{\stag{600-rg.6} Definition} 0) $K^{3,\text{cs}}_{\lambda^+} =
11076: \{(M,N,a) \in K^{3,\text{bs}}_{\lambda^+}:M,N$ are from
11077: $K^{\text{nice}}_{\lambda^+}\}$; we say $N' \in K_\lambda$ (or $p'$)
11078: witness $(M,N,a) \in K^{3,\text{cs}}_{\lambda^+}$ if it witnesses
11079: $(M,N,a) \in K^{3,\text{bs}}_\lambda$.
11080: \nl
11081: 1) ${\Cal S}^{\text{cs}}_{\lambda^+} :=
11082: \{\text{\ortp}(a,M,N):M \le^*_{\lambda^+} N$ are in
11083: $K^{\text{nice}}_{\lambda^+},a \in N$ and $(M,N,a) \in
11084: K^{3,\text{cs}}_{\lambda^+}\}$, the type being for ${\frak
11085: K}^{\text{nice}}_{\lambda^+} = (K^{\text{nice}}_{\lambda^+},
11086: \le^*_{\lambda^+}$), see below
11087: \footnote{actually to define \ortp$_{{\frak K}_\lambda}(a,M,N)$ where $M
11088: \le_{{\frak K}_\lambda} N,\bar a \in N$ we need less that ``${\frak
11089: K}_\lambda$ is a $\lambda$-a.e.c.", and we know on
11090: $(K^{\text{nice}}_{\lambda^+},\le^*_{\lambda^+})$ more than enough}
11091: so the notation is justified by \scite{600-rg.7}(2). \nl
11092: 2) We define ${\frak K}^\otimes = (K^\otimes,\le^\otimes)$ as follows
11093: \mr
11094: \item "{$(a)$}" $K^\otimes = {\frak K} \restriction
11095: \{M \in K:M$ is the union of a directed family of $\le_{\frak K}$-submodels
11096: each from $K^{\text{nice}}_{\lambda^+}\}$
11097: \sn
11098: \item "{$(b)$}" Let
11099: $M_1 \le^\otimes M_2$ if $M_1,M_2 \in K^\otimes,M_1 \le_{\frak K} M_2$ and:
11100: {\roster
11101: \itemitem{ $(*)_{M_1,M_2}$ } if $N_\ell \in K_\lambda,
11102: N_\ell \le_{\frak K} M_\ell$,
11103: for $\ell = 1,2,p \in {\Cal S}^{\text{bs}}(N_2)$ does not fork over $N_1$ and
11104: $N_1 \le_{\frak K} N_2$ \ub{then} some $a \in M_1$ realizes $p$ in $M_2$
11105: \endroster}
11106: \item "{$(c)$}" let $\le^\otimes_{\lambda^+} =
11107: \le^\otimes \restriction K^\otimes_{\lambda^+}$.
11108: \ermn
11109: 3) $\nonfork{}{}_{\lambda^+} = \{(M_0,M_1,a,M_3):M_0 \le^*_{\lambda^+} M_1
11110: \le^*_{\lambda^+} M_3$ are in $K^{\text{nice}}_{\lambda^+}$ and $(M_1,M_3,a)
11111: \in K^{3,\text{cs}}_{\lambda^+}$ as witnessed by some $N \le_{\frak K} M_0$
11112: from $K_\lambda\}$.
11113: \nl
11114: 4) ${\frak K}^{\text{nice}}_{\lambda^+} = (K^{\text{nice}}_{\lambda^+},
11115: \le^*_{\lambda^+})$, that is
11116: $(K^{\text{nice}}_{\lambda^+},\le^*_{\lambda^+} \restriction
11117: K^{\text{nice}}_{\lambda^+})$.
11118: \nl
11119: 5) We say that $M'$ or $p'$ witness $p =
11120: \text{ \ortp}_{{\frak K}^{\text{nice}}_{\lambda^+}}(a,M,N)$ when $M'
11121: \le_{\frak K} M,M' \in K_\lambda$ and $[M' \le_{{\frak K}_\lambda} M''
11122: \le_{\frak K} M \Rightarrow \text{\rm \ortp}_{\frak s}(a,M'',N)$ does not
11123: fork over $M'$ and $p' = \text{\rm \ortp}_{\frak s}(a,M',N)$.
11124: \enddefinition
11125: \bigskip
11126:
11127: \demo{\stag{600-rg.7} Conclusion} Assume \footnote{this is like
11128: $(**)_{M_1,M_2}$ from \scite{600-sg.10}, particularly see clause (iv)
11129: there} (recalling \scite{600-sg.3}):
11130: \mr
11131: \item "{$\boxtimes$}" not for every
11132: $S \subseteq S^{\lambda^{++}}_{\lambda^+}$ is there
11133: $\lambda^+$-saturated $M \in K_{\lambda^{++}}$
11134: such that $S(M) = S/{\Cal D}_{\lambda^{++}}$.
11135: \ermn
11136: 0) On $K^{\text{nice}}_{\lambda^+}$, the relations $\le^*_{\lambda^+},
11137: \le^\otimes$ agree. \nl
11138: 1) ${\frak K}^{\text{nice}}_{\lambda^+} =
11139: (K^{\text{nice}}_{\lambda^+},\le^*_{\lambda^+})$ is a
11140: $\lambda^+$-abstract elementary class and is categorical in $\lambda^+$ and
11141: has no maximal member and has amalgamation.
11142: \nl
11143: 2) $K^\otimes$ is the class of $\lambda^+$-saturated
11144: models in ${\frak K}$ so $K^\otimes_{\lambda^+} = K^{\text{nice}}
11145: _{\lambda^+}$. \nl
11146: 3) ${\frak K}^\otimes$ is an a.e.c. with LS$(K^\otimes) = \lambda^+$
11147: and is the lifting of ${\frak K}^{\text{nice}}_{\lambda^+}$.
11148: \nl
11149: 4) On $K^{\text{nice}}_{\lambda^+},({\Cal S}^{\text{cs}}_{\lambda^+},
11150: \nonfork{}{}_{\lambda^+})$ are equal to $({\Cal S}^{\text{bs}}
11151: \restriction K^{\text{nice}}_{\lambda^+},
11152: \nonfork{}{}_{< \infty} \restriction K^{\text{nice}}_{\lambda^+})$
11153: where they are defined in \scite{600-1.6}, \scite{600-1.7}. \nl
11154: 5) $({\frak K}^{\text{nice}}_{\lambda^+},{\Cal S}^{\text{cs}}_{\lambda^+},
11155: \nonfork{}{}_{\lambda^+})$ is a good $\lambda^+$-frame. \nl
11156: 6) For $M_1 \le^*_{\lambda^+} M_2$ from $K^\otimes_{\lambda^+}$
11157: and $a \in M_2 \backslash M_1$, the
11158: type \ortp$_{K^\otimes}(a,M_1,M_2)$ is determined by
11159: \ortp$_{{\frak K}_\lambda}(a,N_1,M_2)$ for all
11160: $N_1 \le_{\frak K} M_1,N_1 \in K_\lambda$.
11161: \enddemo
11162: \bigskip
11163:
11164: \demo{Proof} 0) By \scite{600-sg.3} and our assumption $\boxtimes$, we have
11165: $M_1,M_2 \in K^{\text{nice}}_{\lambda^+} \and M_1 \le^\otimes M_2
11166: \Rightarrow M_1 \le^*_{\lambda^+} M_2$ (otherwise $(**)_{M_1,M_2}$ of \scite{600-sg.10} holds
11167: hence $(***)$ of \scite{600-sg.10} holds and by \scite{600-sg.3} we get $\neg \boxtimes$,
11168: contradiction). The other direction is easier just see \scite{600-sg.10}(4).
11169: \nl
11170: 1) We check the axioms for being a $\lambda^+$-a.e.c.:
11171: \sn
11172: \ub{Ax 0}: (Preservation under isomorphisms) Obviously.
11173: \sn
11174: \ub{Ax I}: Trivially.
11175: \sn
11176: \ub{Ax II}: By \scite{600-ne.3}(2).
11177: \sn
11178: \ub{Ax III}: By \scite{600-ne.5}(2) the union belongs to
11179: $K^{\text{nice}}_{\lambda^+}$ and it $\le^*_{\lambda^+}$-extends each
11180: member of the union by \scite{600-ne.5}(1).
11181: \sn
11182: \ub{Ax IV}: Otherwise $(*)$ of \scite{600-sg.10} holds, hence by \scite{600-sg.10}
11183: also $(***)$ of \scite{600-sg.10} holds. So
11184: by \scite{600-sg.3} our assumption $\boxtimes$ fail, contradiction; this
11185: is the only place we use $\boxtimes$ in the proof of (1).
11186: \sn
11187: \ub{Ax V}: By \scite{600-ne.3}(3) and Ax V for ${\frak K}$.
11188:
11189: Also ${\frak K}^{\text{nice}}_{\lambda^+}$ is categorical by the
11190: uniqueness of the saturated model in $\lambda^+$ for ${\frak K}$
11191: has no maximal model by
11192: \scite{600-ne.3}(1). ${\frak K}^{\text{nice}}_{\lambda^+}$ has
11193: amalgamation by \scite{600-ne.4}(1).
11194: \nl
11195: 2) Every member of $K^\otimes$ is $\lambda^+$-saturated in ${\frak K}$ by
11196: \scite{600-ne.5}(2) (prove by induction on the cardinality of the directed
11197: family in Definition \scite{600-rg.6}(2), i.e. by the LS-argument it is
11198: enough to deal with the index family of $\le \lambda^+$ models each
11199: of cardinality $\lambda^+$, which holds by part (0) + (1)).
11200: If $M \in K$ is $\lambda^+$-saturated, then as in \S1, it is easy
11201: to find a directed system of saturated $\le_{\frak K}$-submodels from
11202: $K_{\lambda^+}$ using \scite{600-ne.5}(2) (using the stability of ${\frak K}$
11203: in $\lambda$).
11204: \nl
11205: 3),4) Easy by now (or see \S1). \nl
11206: 5) We have to check all the clauses in Definition \scite{600-1.1}. We shall
11207: use parts (0)-(3) freely.
11208: \bn
11209: \ub{Axiom (A)}:
11210:
11211: By part (3) (of \scite{600-rg.7}).
11212: \bn
11213: \ub{Axiom (B)}:
11214:
11215: There is a superlimit model in $K^\otimes_{\lambda^+} =
11216: K^{\text{nice}}_{\lambda^+}$ by part (1) and uniqueness of the
11217: saturated model.
11218: \bn
11219: \ub{Axiom (C)}:
11220:
11221: By part (1), i.e., \scite{600-ne.4}(1) we have amalgamation;
11222: JEP holds as $K^{\text{nice}}
11223: _{\lambda^+}$ is categorical in $\lambda^+$.
11224: ``No maximal member in ${\frak K}^\otimes_{\lambda^+}$" holds by
11225: \scite{600-ne.3}(1).
11226: \bn
11227: \ub{Axiom (D)(a),(b)}:
11228:
11229: By the definition \scite{600-rg.6}(1).
11230: \bn
11231: \ub{Axiom (D)(c)}:
11232:
11233: By \scite{600-1.11} (and Definition \scite{600-rg.6}(1)). Clearly
11234: $K^{3,\text{cs}}_{\lambda^+} = K^{3,\text{bs}} \restriction
11235: K^{\text{nice}}_{\lambda^+}$.
11236: \bn
11237: \ub{Axiom (D)(d)}:
11238:
11239: For $M \in {\frak K}^\otimes_{\lambda^+}$ let $\bar M = \langle M_i:
11240: i < \lambda^+ \rangle \, \le_{\frak K}$-represent $M$, so if
11241: $M \le^\otimes N \in K^\otimes_{\lambda^+}$, (hence $M \le^*_{\lambda^+}
11242: N \in K^\otimes_{\lambda^+} = K^{\text{nice}}_{\lambda^+}$)
11243: and $a \in N$, \ortp$_{{\frak K}^{\text{nice}}_{\lambda^+}}(a,M,N) \in
11244: {\Cal S}^{\text{cs}}_{\lambda^+}(M)$, we let
11245: $\alpha(a,N,\bar M) = \text{ Min}\{\alpha:\text{\ortp}(a,M_\alpha,N) \in
11246: {\Cal S}^{\text{bs}}(M_\alpha)$ and for every $\beta \in (\alpha,\lambda^+)$,
11247: \ortp$(a,M_\beta,N) \in {\Cal S}^{\text{bs}}(M_\beta)$
11248: is a non-forking extension of \ortp$(a,M_\alpha,N)\}$.
11249: \nl
11250: Now
11251: \mr
11252: \item "{$(a)$}" $\alpha(a,N,\bar M)$ is well defined for $a,N$ as above \nl
11253: [Why? By Defintion \scite{600-1.9} + \scite{600-rg.6}(1)]
11254: \sn
11255: \item "{$(b)$}" if $a_\ell,N_\ell$ are above for $\ell = 1,2$ and
11256: $\alpha(a_1,N_1,\bar M) = \alpha(a_2,N_2,\bar M)$ call it $\alpha$
11257: and \ortp$_{\frak s}(a_1,M_\alpha,N) = \text{ \ortp}_{\frak s}
11258: (a_2,M_\alpha,N_2)$ then
11259: {\roster
11260: \itemitem{ $(*)$ } for $\beta < \lambda^+$ we have
11261: \ortp$_{\frak s}(a_1,M_\beta,N_1) = \text{ \ortp}_{\frak s}(a_1,M_\beta,N_2) \in
11262: {\Cal S}^{\text{bs}}(M_\beta)$ \nl
11263: [Why? By $\nonfork{}{}_{}$-calculus when $\beta \ge \alpha$ by
11264: monotonicity if $\beta \le \alpha$]
11265: \endroster}
11266: \item "{$(c)$}" if $a_\ell,N_\ell$ are as above for $\ell=1,2$ and
11267: $(*)$ above holds then
11268: {\roster
11269: \itemitem{ $(**)$ } \ortp$_{{\frak K}^\otimes_{\lambda^+}}
11270: (a_1,M,N_1) = \text{\rm \ortp}_{{\frak K}^\otimes_{\lambda^+}}(a_2,M,N_2)$ \nl
11271: [Why? Use \scite{600-ne.4}(3) or by part (6) below].
11272: \endroster}
11273: \ermn
11274: As $\alpha < \lambda \Rightarrow
11275: |{\Cal S}^{\text{bs}}_{\frak s}(M_\alpha)| \le \lambda$ (by the
11276: stability Axiom (D)(d) for ${\frak s}$), clearly
11277: $|{\Cal S}^{\text{cs}}_{\lambda^+}(M)| \le \dsize \sum_{\alpha < \lambda^+}
11278: |{\Cal S}^{\text{bs}}(M_\alpha)| \le \lambda^+ = \|M\|$ as required.
11279:
11280: The reader may ask why do we not just quote the parallel result from
11281: \S2: The answer is that the equality of types there is ``a formal, not
11282: the true one". The crux of the matter is that we prove locality (in
11283: clause (c) above).
11284: \bn
11285: \ub{Axiom (E)(a)}:
11286:
11287: By \scite{600-1.6} - \scite{600-1.9}.
11288: \bn
11289: \ub{Axiom (E)(b); monotonicity}:
11290:
11291: Follows by Axiom (E)(b) for ${\frak s}$ and the definition.
11292: \bn
11293: \ub{Axiom (E)(c)}; local character:
11294:
11295: By \scite{600-1.13}(5) or direct by translating it to the ${\frak s}$-case.
11296: \bn
11297: \ub{Axiom (E)(d); (transitivity)}:
11298:
11299: By \scite{600-1.13}(4).
11300: \bn
11301: \ub{Axiom (E)(e); uniqueness}:
11302:
11303: By \scite{600-ne.4}(3) or by part (6) below.
11304: \bn
11305: \ub{Axiom (E)(f); symmetry}:
11306:
11307: So assume $M_0 \le^*_{\lambda^+} M_1 \le^*_{\lambda^+} M_2$ are from
11308: $K^\otimes_{\lambda^+}$ and for $\ell=1,2$ we have $a_\ell \in M_\ell$,
11309: \ortp$_{{\frak K}^{\text{nice}}_{\lambda^+}}(a_\ell,
11310: M_0,M_\ell) \in {\Cal S}^{\text{cs}}_{\lambda^+}(M_0)$ as witnessed by
11311: $p_\ell \in {\Cal S}^{\text{bs}}_{\frak s}(N^*_\ell),N^*_\ell \in
11312: K_\lambda,N^*_\ell \le_{\frak K} M_0$ and
11313: \ortp$_{{\frak K}^\otimes_{\lambda^+}}(a_2,
11314: M_1,M_2)$ does not fork (in
11315: the sense of $\nonfork{}{}_{\lambda^+}$) over $M_0$ (note that
11316: $M_0,M_1,M_2$ here stand for $M_0,M_1,M'_3$ in (E)(f)(i) from
11317: Definition \scite{600-1.1}). As we know the monotonicity \wilog \,
11318: $M_1 <^+_{\lambda^+} M_2$.
11319: We can finish by \scite{600-ne.4}(4) (and Axiom (E)(e)
11320: for ${\frak K}_\lambda$).
11321:
11322: In more details, we can find $N_0,N_1,N_2$ such that: $N_\ell
11323: \le_{\frak K} M_\ell$ and $N_\ell \in K_\lambda$ for $\ell = 0,1,2$
11324: and $N^*_1 \cup N^*_2 \subseteq N_0 \le_{\frak K} N_1 \le_{\frak K}
11325: N_2$ and $a_1 \in N_1,a_2 \in N_2$
11326: and $N_2$ is $(\lambda,*)$-brimmed over $N_1$ hence over $N_0$,
11327: and $(\forall N \in K_\lambda)[N_0 \le_{\frak K} N
11328: \le_{\frak K} M_0 \rightarrow
11329: (\exists M \in K_\lambda)(M \le_{\frak K} M_2 \and
11330: \text{ NF}_\lambda(N_0,N,N_2,M))]$. \nl
11331: By Axiom (E)(f) for $({\frak K},{\Cal S}^{\text{bs}},
11332: \nonfork{}{}_{\lambda})$ we can
11333: find $N'$ such that $N_0 \le_{\frak K} N' \le_{\frak K} N_2$ and $a_2
11334: \in N'$ and \ortp$_{\frak s}(a_1,N',N_2)$ does not fork over
11335: $N_0$. Now we can find $f'_0,M'_1$ such that $M_0 \le^+_{\lambda^+}
11336: M'_1,f'_0$ is a $\le_{\frak K}$-embedding of $N'$ into $M'_1$ and
11337: $(\forall N \in K_\lambda)[N_0 \le_{\frak K} N \le_{\frak K} M_0
11338: \rightarrow (\exists M \in K_\lambda)(M \le_{\frak K} M'_1 \and
11339: \text{\rm NF}_\lambda(N_0,N,f'_0(N'),M))]$. Next we can find $f''_0,M'_2$ such
11340: that $M'_1 <^+_{\lambda^+} M'_2,f''_0 \supseteq f'_0,f''_0$ is a
11341: $\le_{\frak K}$-embedding of $N_2$ into $M'_2$ and $(\forall N \in
11342: K_\lambda)[N_0 \le_{\frak K} N \le_{\frak K} M_0 \rightarrow (\exists
11343: M \in K_\lambda)(M \le_{\frak K} M'_2 \and \text{\rm NF}_\lambda
11344: (N_0,N,f''_0(N_2),M)]$. \nl
11345: Lastly, by \scite{600-ne.4}(4) there is an isomorphism $f$ from $M_2$
11346: onto $M'_2$ over $M_0$ extending $f''_0$. Now $f^{-1}(M'_1)$ is a
11347: model as required.
11348: \bn
11349: \ub{Axiom (E)(g); extension existence}:
11350:
11351: Assume $M_0 \le^*_{\lambda^+} M_1$ are from
11352: $K^{\text{nice}}_{\lambda^+},
11353: p \in {\Cal S}^{\text{cs}}_{\lambda^+}(M_0)$, hence
11354: there is $N_0 \le_{\frak K} M_0,N_0 \in K_\lambda$ such that $(\forall N \in
11355: K_\lambda)(N_0 \le_{\frak K} N <_{\frak K} M_0 \rightarrow p \restriction N$
11356: does not fork over $N_0)$. By \scite{600-ne.3}(1A) there are $M_2 \in K^\otimes
11357: _{\lambda^+}$ and $a \in M_2$ such that $M_1 \le^*_{\lambda^+} M_2$ and
11358: \ortp$_{{\frak K}^{\text{nice}}_{\lambda^+}}(a,M_1,M_2)
11359: \in {\Cal S}^{\text{cs}}_{\lambda^+}(M_1)$
11360: is witnessed by $p \restriction N_0$ and by part (6) we have
11361: \ortp$_{{\frak K}^{\text{nice}}_{\lambda^+}}(a,M_0,M_2)=p$.
11362: Checking the definition of does not
11363: fork, i.e., $\nonfork{}{}_{\lambda^+}$ we are done.
11364: \bn
11365: \ub{Axiom (E)(h), (continuity)}:
11366:
11367: By \scite{600-1.13}(6).
11368: \bn
11369: \ub{Axiom (E)(i)}:
11370:
11371: It follows from the rest by \scite{600-1.15}.
11372: \nl
11373: 6) So assume $M \le^*_{\lambda^+} M_\ell,a_\ell \in M_\ell \backslash
11374: M$ for $\ell=1,2$ and $N \le_{\frak K} M \wedge N \in K_\lambda
11375: \Rightarrow \text{\rm \ortp}_{\frak K}(a_1,N,M_1) = \text{\rm \ortp}_{\frak
11376: K}(a_2,N,M_2)$. By \scite{600-ne.3}(1) there are $M^+_1,M^+_2 \in
11377: K^{\text{nice}}_{\lambda^+}$ such that $M_\ell <^+_{\lambda^+}
11378: M^+_\ell$ for $\ell=1,2$. By \scite{600-ne.4}(2),(3) there is an
11379: isomorphism $f$ from $M^+_1$ onto $M^+_2$ over $M$ which maps $a_1$ to
11380: $a_2$. This clearly suffices. \hfill$\square_{\scite{600-rg.7}}$
11381: \enddemo
11382: \newpage
11383:
11384: \head {\S9 Final conclusions} \endhead \resetall \sectno=9
11385: \spuriousreset
11386: \bigskip
11387:
11388: We now show that we have actually solved our specific test questions
11389: about categoricity and few models. First we deal with good $\lambda$-frames.
11390: \proclaim{\stag{600-fc.1} Main Lemma} Assume
11391: \mr
11392: \item "{$(a)$}" $2^\lambda < 2^{\lambda^+} < 2^{\lambda^{++}} < \ldots <
11393: 2^{\lambda^{+n}}$, and $n \ge 2$ and ${\text{\rm WDmId\/}}
11394: (\lambda^{+ \ell})$ is not
11395: $\lambda^{+ \ell +1}$-saturated (normal ideal on $\lambda^{+ \ell}$)
11396: for $\ell = 1,\dotsc,n-1$
11397: \sn
11398: \item "{$(b)$}" ${\frak s} = ({\frak K},{\Cal S}^{\text{bs}},
11399: \nonfork{}{}_{})$ is a good $\lambda$-frame
11400: \sn
11401: \item "{$(c)$}" $\dot I(\lambda^{+ \ell},{\frak K}(\lambda^+$-saturated))
11402: $< 2^{\lambda^{+ \ell}}$ for $\ell = 2,\dotsc,n$.
11403: \ermn
11404: \ub{Then}
11405: \mr
11406: \item "{$(\alpha)$}" $K$ has a member of cardinality $\lambda^{+n+1}$
11407: \sn
11408: \item "{$(\beta)$}" for $\ell < n$ there is a good
11409: $\lambda^{+ \ell}$-frame ${\frak s}_\ell =
11410: ({\frak K}^\ell,{\Cal S}^{\text{bs}}_{{\frak s}_\ell},
11411: \nonfork{}{}_{{\frak s}_\ell})$ such that
11412: $K^\ell_{\lambda^{+ \ell}} \subseteq K_{\lambda^{+ \ell}}$ and
11413: $\le_{{\frak K}^\ell} \subseteq \le_{\frak K}$
11414: \sn
11415: \item "{$(\gamma)$}" ${\frak s}_0 = {\frak s}$ and if $\ell < m < n$
11416: then $K^\ell_{\lambda^{+m}} \supseteq K^m_{\lambda^{+ m}} \and
11417: \le_{{\frak K}^\ell} \restriction K^m \supseteq \le_{{\frak K}^m}$.
11418: \endroster
11419: \endproclaim
11420: \bigskip
11421:
11422: \demo{Proof} We prove this by induction on $n$.
11423:
11424: For $n=m+1 \ge 2$, by the induction hypothesis for $\ell = 0,\dotsc,m-1$, there
11425: is a frame ${\frak s}_\ell = ({\frak K}^\ell,\nonfork{}{}_{{\frak s}_\ell},
11426: {\Cal S}^{\text{bs}}_{{\frak s}_\ell})$ which is $\lambda^{+ \ell}$-good and
11427: $K_{{\frak s}_\ell} \subseteq K^{\frak s}_{\lambda^{+ \ell}}$ and
11428: $\le_{{\frak K}^\ell} \subseteq \le_{\frak K} \restriction {\frak K}^\ell$.
11429: By \scite{600-nu.6} and clause (c) of the assumption we know that ${\frak
11430: s}$ has density for $K^{3,\text{uq}}_{\frak s}$.
11431: Now \wilog \, $K^{m-1}$ is categorical in $\lambda^{+(m-1)}$
11432: (by \scite{600-1.16} really necessary only for $\ell = 0$) and by
11433: Observation \scite{600-nu.13.1} we get the assumption \scite{600-nf.0} of \S6
11434: hence the results of \S6, \S7, \S8 apply.
11435: Now apply \scite{600-rg.7} to $({\frak K}^{m-1},
11436: {\Cal S}^{\text{bs}}_{{\frak s}_{m-1}},\nonfork{}{}_{{\frak
11437: s}_{m-1}})$ and get a $\lambda^{+ m}$-frame
11438: ${\frak s}_m$ as required in clause $(\beta)$. By \scite{600-4a.12} we have
11439: $K^m_{\lambda^{+m+1}} \ne \emptyset$ which is clause $(\alpha)$ in the
11440: conclusion. Clause $(\beta)$ has already been proved and clause
11441: $(\gamma)$ should be clear. \hfill$\square_{\scite{600-fc.1}}$
11442: \enddemo
11443: \bn
11444: Second (this fulfills the aim of \cite{Sh:576}).
11445: \proclaim{\stag{600-fc.2} Theorem} 1) Assume $2^{\lambda^{+ \ell}} <
11446: 2^{\lambda^{+(\ell+1)}}$ for $\ell = 0,\dotsc,n-1$ and the normal
11447: ideal ${\text{\rm WDmId\/}}(\lambda^{+ \ell})$ is not
11448: $\lambda^{+\ell+1}$-saturated for $\ell = 1,\dotsc,n-1$.
11449:
11450: If ${\frak K}$ is an abstract elementary class with
11451: ${\text{\rm LS\/}}({\frak K}) \le \lambda$ which is
11452: categorical in $\lambda,\lambda^+$ and $1 \le \dot I
11453: (\lambda^{+2},K)$ and $\dot I(\lambda^{+m},
11454: {\frak K}) < 2^{\lambda^{+m}}$
11455: for $m \in [2,n)$ (or just $\dot I(\lambda^{+m},
11456: {\frak K}(\lambda^+$-saturated)) $< 2^{\lambda^{+m}}$),
11457: \ub{then} ${\frak K}_{\lambda^{+n}} \ne \emptyset$ (and there are
11458: ${\frak s}_\ell(\ell < n)$ as in $(\gamma)$ of \scite{600-fc.1}).
11459: \nl
11460: 2) We can omit the assumption ``not $\lambda^{+ \ell +1}$-saturated",
11461: if we strengthen $\dot I(\lambda^{+m},{\frak K}) <
11462: 2^{\lambda^{+m}}$ to $\dot I(\lambda^{+m},{\frak K}) <
11463: \mu_{\text{unif}}(\lambda^{+m},2^{\lambda^{+(m-1)}})$, see
11464: \marginbf{!!}{\cprefix{88r}.\scite{88r-0.wD}}(3).
11465: \endproclaim
11466: \bigskip
11467:
11468: \demo{Proof} 1) By \scite{600-Ex.4} and \scite{600-fc.1}.
11469: \nl
11470: 2) See \cite{Sh:838}. \hfill$\square_{\scite{600-fc.2}}$
11471: \enddemo
11472: \bn
11473: Next we fulfill an aim of \chaptercite{88r}.
11474: \proclaim{\stag{600-fc.3} Theorem} 1) Assume $2^{\aleph_\ell} <
11475: 2^{\aleph_{(\ell +1)}}$ for $\ell = 0,\dotsc,n-1$ and $n \ge 2$
11476: and {\rm WDmId}$(\lambda^{+ \ell})$ is not $\lambda^{+ \ell +1}$-saturated
11477: for $\ell = 1,\dotsc,n-1$.
11478:
11479: If ${\frak K}$ is an abstract
11480: elementary class which is ${\text{\rm PC\/}}_{\aleph_0}$
11481: and $1 \le \dot I(\aleph_1,{\frak K}) < 2^{\aleph_1}$ and $\dot I
11482: (\aleph_\ell,
11483: {\frak K}) < 2^{\aleph_\ell}$ for $\ell =2,\dotsc,n$, \ub{then} ${\frak K}$
11484: has a model of cardinality $\aleph_{n+1}$ (and there are ${\frak
11485: s}_\ell(\ell < n)$ as in \scite{600-fc.2}.
11486: \nl
11487: 2) We can omit the assumption ``not $\lambda^{+\ell+1}$-saturated" if
11488: we strengthen ``$\dot I(\aleph_\ell,{\frak K}) < 2^{\aleph_\ell}$"
11489: to ``$\dot I(\aleph_\ell,K) <
11490: \mu_{\text{unif}}(\aleph_\ell,2^{\aleph_{\ell -1}})$.
11491: \endproclaim
11492: \bigskip
11493:
11494: \remark{Remark} Compared with Theorem \scite{600-fc.2} our gains are no
11495: assumption on $\dot I(\lambda,K)$ and weaker assumption on $\dot
11496: I(\lambda^+,K)$, i.e., $< 2^{\aleph_1}$ (and $\ge 1$) rather than
11497: $=1$. The price is $\lambda = \aleph^+_0$ and being PC$_{\aleph_0}$.
11498: \endremark
11499: \bigskip
11500:
11501: \demo{Proof} 1) By \scite{600-Ex.1} and \scite{600-fc.1}.
11502: \nl
11503: 2) See \cite{Sh:838}. \hfill$\square_{\scite{600-fc.3}}$
11504: \enddemo
11505: \bn
11506: Lastly, we fulfill an aim of \cite{Sh:48}.
11507: \proclaim{\stag{600-fc.4} Theorem} 1) Assume $2^{\aleph_\ell} <
11508: 2^{\aleph_{\ell +1}}$ for $\ell \le n -1$
11509: and ${\text{\rm WDmId\/}}
11510: (\lambda^{+ \ell})$ is not $\lambda^{+ \ell +1}$-saturated
11511: for $\ell = 1,\dotsc,n-1,\psi \in \Bbb L_{\omega_1,\omega}(\bold Q),
11512: \dot I(\aleph_1,\psi) \ge 1$ and $\dot I(\aleph_\ell,\psi) <
11513: 2^{\aleph_\ell}$ for
11514: $\ell = 1,\dotsc,n$. \ub{Then} $\psi$ has a model in $\aleph_{n+1}$
11515: and there are ${\frak s}_1,\dotsc,{\frak s}_{n-1}$ as in \scite{600-fc.3}
11516: for $K = \text{\rm Mod}_\psi$ and appropriate $\le_{\frak K}$.
11517: \nl
11518: 2) We can omit the assumption ``not $\lambda^{+ \ell +1}$-saturated"
11519: if we strengthen ``$\dot I(\aleph_\ell,{\frak K}) < 2^{\aleph_\ell}$
11520: for $\ell=2,\dotsc,n$" to ``$\dot I(\aleph_\ell,{\frak K}) <
11521: \mu_{\text{unif}}(\aleph_\ell,2^{\aleph_{\ell-1}})$".
11522: \endproclaim
11523: \bigskip
11524:
11525: \demo{Proof} 1) By \scite{600-Ex.1A} mainly clauses (c)-(d) and
11526: \scite{600-fc.1}. Note that this time in \scite{600-fc.1} we use the \nl
11527: $\dot I(\lambda^{+ \ell},{\frak K}
11528: (\lambda^+$-saturated)) $< 2^{\lambda^{+ \ell}}$.
11529: \nl
11530: 2) See \cite{Sh:838}. \hfill$\square_{\scite{600-fc.4}}$
11531: \enddemo
11532:
11533: \nocite{ignore-this-bibtex-warning}
11534: %% you may want to move the following lines up a bit
11535: \newpage
11536:
11537: REFERENCES.
11538: \bibliographystyle{lit-plain}
11539: \bibliography{lista,listb,listx,listf,liste}
11540:
11541: \enddocument %%
11542:
11543:
11544:
11545:
11546:
11547: