IN TH£ORY a-RECURSIC~ by RI CEAP.D j':,.. SHORE (1968) DEGREE OF DOC~OR OF PH2:LOSOPHY at the MASSACHUSETTS INSTITUTE OF ~ECHNO~OGY Septeri1ber, 1972 S i g rl a 'C u ~ e 0 ~f Ali, t 1"'1 \.) r '. . . . . . . . ,. . . . . .,. . ;'1 • • » '. .. ;Ol .. • .. .. • • .,.. " , • ..". ) • • • • • • • • • • ~ l~ <t. •• ."~ • ... • • , b • • 1> ... st -:ll. , 1,.::'7,2 Department of Mathematics, Au Certified. ... .....:- • .,, , / <'(. .,Q ... Thesis Supervisor Ch£tirm.2-r!;) T~!e p(3.1~t rn8n~: [:"1 c· n Gr acI Lt 2_ -:":·s r: ':r-"}~::!.. t t :-:: e ~S t.: ~J.(~1 ~,~ "', 1,,· r...,,} Richard, f'i.. Shore Subrr.itted to the !)epartrnE:rlt of thematiC's jn in partial fulfillment of the requirements for t, 1972 ~ne de~ree of Doctor of Philosophy. ABSTRACT The pri.ority method as a.pplieci to Y'(:::Cl,U"S:2.C{l I)n l1i~::sibJ.::ordinals is studied and developed by ~ereralizin~ to ~-recu~­ sion theor::7 several classtc theorer~ls of ordil 1 ;:1.ry y'eeUt'si.crl theory 'dhcse ppoofs r'eqlJire diff':;rent tY}JCS of' ~)riority 2..':'P;U-· ITlents. The siDplest form of the priority cethoi is the Fri I'·'IuCr})~lj.. l{ ty~)e of' fj.rlj_te i11 jur Jl ar l t gt.t!1iCl1t S 9 i~t~;C1~-ll'1:L i~. burg1~j,~3;0 been f;ener;:Llized to a--recur::;iDD tneory by :}. Sac~=s a.nd S. S:!.;';lpson. In cha~)t2rs I and II ~'l,-::': shoier hO~,I tbe aSSU?T,rJt:'OL of E~-ad~issibilit can be used to extend t~eir ~ethodG to ~ne r)J:)()rJel~~~~es~ Ir: tl"le f'o ll()f:;tir},~; Theorem: If a lS L 2 -admi s sible there i. S D ni nirrlal 0. -::1::: e e '" 0 n,r"l~' ) +- c ' "a - r . l.. :::. C1.""" {)l>""'~ ~ ~ '.: ~. r~ ~ )J f_ iJ 'S.... e LJ • '.-< v 1 • \..• .;... ......... ',~ L i.- .L '-.-. 1 , .. ". ,. the \~xlstenc(: of nj_nir:1al a-degree::> are <;11so establ::Ls[)ec;cjlxt without the usc: of ~J. priori ty 2.ri:~umerlt. In chaote2 2:1 '(·;t:: x a p,r i or i 't '::7 2. r' [; tlr:"le~J~C t/.~j~ ~~:1 3. fOl"C j_ns~ c c,' ~ 3 ~~ ]~U_ c~, i Cl~ t; c ;),t (J ·~.t (:; t. ~ e a----;;~;--.;; - c;; l~ d.1. ::;,r-...~..... \ r a '- ·l n J.. ,- ...., 1.11 i r· ..l- ,'....\./- \-, ('l A?' \.,; .C 4 Theo:r'er:L paraT)1 e exist Lj If CJ. lS En-admissible then Ln ~) u t: .~ e t s of a. Indeed, we unifor~lv i 1) J~ e sets .t\ a: ~-l far Inc }"12.. D "',_: e IJ I }~ I 21 (1 ttl j_ :3 ~: II '0 ~~~T (~ L r:; t: C t) '1' O\I e t rJ e a _~i r2~ S ;-:: 1 i~~~_j ~ t~': 'Z 1.11 a.r !.}£~~._ ~i~~~ e () r (~r~l O~ -1~ • z:. ::' ;~ t .r~ () ~r'l ~=. ~~~r e}~;1 E;~ .tl cl ::t 2.nd such th::.:. t In chanter 1'1 for all admissible of the VIe a: .~:.. 3- carE: s U ::~ {"l t: lie. t~ <0. EJ roo .f c~ t~ t,.: J. ~3 i.:,! -: f: C:' e 1~~ ) f~) C C 0,1'.' S e 1 nf i nit e in j u r ~v D r ~. 0 r it Y aI' p:- Lt T~ e n t <a. c. r (:: ~ 1,~ it") (: c ~J_ ;.:~ -'"7 (~l \~)~'._~> .:t ·CL f~~~~ ,'::\ .t"l for d.ll r:.c :L::; s i 1 (;' 0 I .~ n .,,;,:, S . 1 . 3 ArnonE;, those:"tho have officia.lly been ny te3.c.hsY's I 1\lould Ii ke to than~\: r:lave conveyed ~'lho tho'3e attituJes as well as formal naterial: insi~~ht and Prof. Burton Jreben., ny first ITH.=:nto-r allc]. continuinG: advisor in mathernatie;al logic, Prof. hartley Jr., who introduced me ~ogers, ~ost whc taugnt me set theory as it should be ta and COD\rictlon and, of course, Pro'f. Gerald Sacks, '/\iho taught me idodf:?l theory and set t~leory theory -' ",) u t mer'e importantJ.,! has ShOH'1. as :ie12. f;.S r'2CUTS ion me 1:0:"; beautiful and fruitful are the interconnections anong them. also like tO~}lank I \'Jould Prof. Sacks in his capG.city of thes.is advisor for encouragins me to work on and for rlcJj a-recursion theory the Questions dealt with in s u ~~ _~; (; s t: 1.1i ~:~:: Trian ~l thts thesis. Of those from w~om I have le2rned infor~ally, lowe 3peclal thanks to Stephen SiLlJscn for many lessons learned but e spes ially for my theory. often FOL' tl1ei.r half-ba~ed fi~s t i liS tru.e ';Jil1in:~nes;3 t iC!-j:in a-recur::;j. on to listen and react ta my ideas I would like to thank Lee Harrington anel <-John naclntyre. Important debts ~re also owed to Professors Joseph for 1:.1s " L,f clear an(~. s Ie presentations of difficult theorems of To the latter I r e C i..lr s J. 0 n t 11 e 0 r v in [ 11 ] indebted for his work on pleas~re much L [lJ from which I have a~ gai~ed and insight. For financial SUpDort I must thanJ::: the Depa::etn2nt of ~Iather:1atic.s at :/t.I.'l'. and the National Science ?ounc:a- tiona Finally I ~AJoulc.1 ::;'ilce to thank my v;ife., Haomi,. and my cat f'OI' the!!' conste.n de SUDport and COr1D8.n:i.onshJ.p. 1;1- , it is after my cat that all the important results in this t~esis are named. 5 TABLE OF CONTENTS 2 ACKNOT,'ILEDG;"1ENTS 3 5 IHTRODUCTIOi',y CHAPTER I: 7 ~INIMAL 13 a-DEGREES 1. Definitions and Bas~.c Lemmas 15 2. I·linimal Dezrees VTi thout Priorities 23 .. 0" . 1G '8J.SSlbl.e ra Ina 3 • L.'\' 2- Ad' 2_vQ 4. The Construction 32 5. The Priority Argument 39 6. 0 pen 1. The Q'J. est ion s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . It 1~ Forcin~ 48 Relation 2. The Cons true t ion 51 3. The Priority Argument 57 11 "t • P'.,...,eser~~;'10" ..l ' .• _ v ~ 1 ( '.) )... n _. ".rlmi··,...ir~l~f-n \.\.. . ,., i::> ..:; _ U -l.. ...... v.Y • • • • • • • • • • • • • • • hL -'.J... 65 1. 'llhe Cons true t ien . . . . . . . . . . . . . . . . . • . . . . . . . . . 68 2. The Priority Argument~ ~~ 73 TABLE: OF (cont:tnued) CHAPTER IV: rrI-IE DENSITY rrHEORETI 1. The General Plan 83 84 2. The Key Functions and Their Approximations 87 3. The Requirements and Construction 90 4. The Proof -- An Outline 92 BIBLIOGRAPHY a BIOGRAPHY 98 ,~ .... 0 7 Introduction In generalizing or aXionatizing a field of mathe~~tics one tries to capture in a more abstract form the basic concepts of the subject. Of course, one keeps in sind the essential properties and fundamental theorems that like to preserve. would o~e The first major test for the correctness or everl appropriateness of the then alsG seneralize the deeper atte~pt is whether one can theore~s and l~!Dortant methods of proof that are characteristic of the original field. Thus for recursion theory one atteoDts to reform- ulate in a new setting the basic notions of cocputation J recursive and recursively enumerable. In for~ulating s~ch an abstraction Olle is gUided, of course, by one's intuition about computation but one also has several basic facts and relatjonships th2t lnust be preserved. Thus one nakes sure that basic theoress such as the recursion theorem a~d enu- meration theorem are carried over. When it comes ti~e to apply the above test to an attenp- ted generalization of recursion theory we feel that the key technique to be handled is the priority method. pondin~ly, Corres- the crucial theorens are the ones that depend on the priority method in ordinary ~0cursion theory. Cur reason is that the priority method is the most distinctive and characteristic technique cf recursion theory, while t}le ~\!hich theorems for it needed .lS are :L.n veneral the deepest and most difficult in the field. In this thesis we concern ourselves with the generalization of recursion theory to recursi0D on For ordinals. the history of this subject adm~ssible an~ general esp~- background we refer the reader to [2J, [7J, [8J and cially to [13J and its bi.blio:.:-raphy. Af.i far 2_8 the Pl>l.CJ:'>j_t:l method is concerned the first success in this field comes in [10J, where G. Sacks and S. Simpson ~eneralize the Friedburg--l'luchnik solution to Pc,st' s problem to every admissible t~is any They do such C~. one problem. theore~ about rv:()~1E~, They develop r.e. hovrever, thS.H ~echniques t~at ,;'.l~3t ~ill L1 solve tand~e sets which in ordinary re:ursion theory calls for a simple finite injury priority argume~lt. Later, in [3J, M. Lerman developed an alternative arprcac!l to the finite inj~ry nethod. While Sacks and Simpson draw heavily on model theoretic notions is more pu~el~T recurs ion : '12oret ic . in [10J depend on sequences of r t arne t .1e approach Tl1us cruc ial cases L1 -submodels .L. . 2--proJ"ec t "lon +-' ",c..lces over papers significantJy viniicate Lcr~~n's • +1 lDS ",ear... a-recursion generalization of ordinary recursion theory, Poth of these ~heory as a ~articularly insofar as recursive enumerability is concerned. In chapters I and II of this thesis we use the assumpticn of Zn -admissibilitv to extend the methods of [10J to .> handle constructions of sets. E sets rather than just n In particular) assusin; l:dnil~lEJ_l to construct a in the connlete Cl,-,dezree \'Th1ch is C'i-r. e. set. construct ',:i thout pr:Lori ties ott) for from E 8 3 L':/ - a d r~: iss 1. b 1 e a C rl r2 0' to (belov: Th u s the i Tn rJ r 0 v e ~,1 e n t 0, • corresponds to Sacks' use of a a r:dniIEal degree recu:t'si ve in ordinary recursion theory [6J. rather than ex -de0~ree for ev-::ry illlprovem:::nt to produce a minir:lal adm~ssible. hO~'T c.;-cie[~ree r:-iinimal 3, .J to ex-recursive 'Je also in:licEJ te priori t;;r a:egument to construe t in E 1 Perhaps a ~o~e complex type of priority argu- ment is called for. Ie chapter II \'Ie .;i ve' another e;{aD'ple of this method but vlith tHO additional tFists. ment ;,vhich is uniform in Ct ?iT'st, VJe present an argu- ra.. ther than one ~'lhic:h spli ts into cases according to the internal nature of u. we mix a forcing construction into the priority a~gurnent. We then prove that for any (uniformly) a pair of L n 1: -adJ:lissi.ble n subsets of o. a. Second, there is whicl-. are incoLparable. Clearly the next step in the attack on priority methods In ry,-reCU.I'Slon cated types of theory TJa.S nrior~tv ar to g:eneral:tze the !"1ore ccmpltpts. The major other form of prior-i t/ argurnent metllod 0:' 2acks lS, of course, the inf:Lni te injury ..Ln. [16] Lerman and S<::c 1cs use an infinite injury al'guuent to construct sinimal pairs of a-r.e. degrees for many but not all Zl-admissible a. We attack the infinite injury method by beginning with an intermediate type of argument -- finite injuries but unbounded preservations. In particular we prove (chap- ter III) the Splitting Theorem for all El-admissi~le a. Our method seerrlS to be sufficiently [:eneral to handle other eYa~nples of this type of argument such as ernbedding partial ordering a-r.e. degree. into the a-r.e. degrees below any fixed Finally in chapter IV we sketch a proof of the Density Theoreu for all ~l-admissibles. In addi- tioD to the techniques of chanter I I I we develop the machinery needed to handle a full infinite injury priority argument. success fo~ Though we, of course, view this as an important a-recursion theory we feel that further exam- pIes of the infinite injury Method must be investigated before we will be in a position to claim a general method comp~rable to the ones prOVided in [10J and chapter III for the two types of finite injury arguments. We hope to pursue this course in later work. In closing, we would like to make a few remarks about other important tests for c-recursion theory. In addition to carrytn r : over:' basic tc1e3.s a. ?ood €:eneT'8.11z2.tion must .., .., ..L L offer 30!:,,(cthins nel'l as vTell. of investigat ion \tIi It shouJ.d offer ne,,'; areas thln t:le su.1Jj ec t i tsel f and snculd pro-- mote nevr interactions ':/i ttl other fields of mathematics. to have, success in both of these areas. the first ~'le As an example of cite the ne'!." probleLls presented by DOn-re[;u- larity and non-hyperregularity &nd refer to [13J for the beginninp;3 of in-lportant worle on these questions. feel that 01..1.1'" lemna IV. 2.1 l'IiII allo~'r al.30 further progres s in study of the non-regular and non-hyperregular degrees. ~:;e a-r.e. As far as the second criterion is concerned, \1e \'Jou1d cite on tile one hand the ir:'"(portant of model 2nd argunents in set~·thecretic mc~hocls ·:t-recursion theory. contri~ution~) to all '::'ork on p1"1 ori ty Indeed, on~~ could ~'lell claim that the subject is becoming part of set theory rather than recursion theory. the hope -:ha t the rnethocls of On the ot:'1er CJ. -recurs ion ~v1nd} '1','2 have t he'J=:"'Y 1.J1J..1 pro'le useful in set theory and perhaps also in sodel theory. As 8. ITdnOI" eXCl:1ple whose results ability theor~T 3ee~ of' this ..,,78 vTould cite our clls.pter II to be 2 part of set theory or defin- rather t:lan recursicn theory, although a priority arcu:'lent is cru,cial to the construction. f;. porc important exarlple is furnished by Si;npson f s report '1_'-13 ] of "fdO:'~ kin pros res ~) (; 11. U :31 the methods of a-recursion 12 admissible i~itial se s of a cant :1n",1.ed :1n terac t ion b 0t~Ieen L. ?inally, ~1e expect Ci.-,·recur's i()n theory and Jensen's set,-theoretic :,vorlc [1J on t:18 fine structure of the constructible universe. Chapter I Minimal a-Degrees In this chapter we considRr the problem of constructing minimal a-degrees for admisslble ordinals Ct result wi.II consist of constructing, -:for every OUI' main L,,",-admi.ssible Co a , a set recursive in the complete a-r.e. set which is of minimal a-degree though some other cases will also be covered. vIe ass'L11!l€ here and throughout this thesis an acquaintence ~tlith the basic concepts of a-recursion theory such as the vario'l'.s notions of a-reducibility and regularity. All the necessary inf'ormation can be found, for example, in the introductory sections (1,2 and the beginning of 6) of [10]. For more general in:formation on a-recursion theory we refer the reader to [ 2 ], [ 8] and [ 9] • The first results on minimal degrees is that of Spector [14]. in On He proved the existence of a minlinal degree recursive .for ordinary recursion theory. Unfortunately, his proof is not suited to the needs of a-recursion theory [5]. Instead we rnlst turn to the later improvements introduced by Sacks [6] to construct a minimal degree recursive in Of • Though the most important change introduced by Sacks was the use of a priority argu.ment, his b,::tsic construction was also less re3trictive than that of Spector. It1i'laS this extra freedom that MacIntyre [4] exploited to construct miniJnal a-degrees for all counta.ble admissibles U. His construction, however, proceeded in w-many steps by using a counting of a and so produced a set of minimal degree whieh in general was not even constructible. On the other hand a minor combi- natorial twist (see section 1.7) allows one to carry out the argunlent for all section 2). recursive in ~3-a~missible ordinal3 (we do this in The set constructed is, as in Spector's argu.ment, 0". In section 4 we reintroduce the priority technique to make our construction recursive in a 0' is ~2-admissible Of • Now If we can essentially work recursively in without too much difficulty (seetion 3). Thus we can adopt the techniques of [10] to prove that the priority argument succeeds (section 5) and so degree as well as recursive in ou:~ 0' . set is indeed a minimal Definitions and the basic lenu'nas 1. A set and evers B is of minimal a-degree iff C a-recursive in B a-recursive in it. B is not a-recursive B is either ,a-recursive or has Our general plan is to construct B in stages and to see to it that the second condition is taken care of in one of two ways corresponding to the two possible values for the degree of Steps will, of course, have to be taken C. to assu:l"e ourselves that is not recursive but the assoc:'ated B requirements will be much less troublesome. This section ~ontains some det"'initions and the basic lenunas that 'will be used to handle the more crucial minimality requirements. 1.1 Sequences: An a-finite sequence disjoint a-finite sets called the length of :finite :sequences runction with and !!.!.!. = (lth 0-) • One should think of an (0"0 lJ {lth and CL- cr*,., in the obvious way: (J + i3 0*0 13 r € ,.o} , 01 U [lth 0'*1 and (J 0 by respectively to the ordinal lth 1 a we call extends a and In this section only (0" 'T ,. :) - 'f' or ,. c incompatible if p j (J + ~ J 13 € 1"1}> correspond to extending the representing function associated with say that a as being B·n initial segment of a representing 0 "., written and of 0 In particular 0 <0'0' vI> 0 as the zeros and 0 1 as the ones of the We define concatenation of two a·-finite sequences fUnction. a is a pair union is an in:itial segment of 1~hose 0 cr and a,-fini te seq.uenc2s exclusi.vely. 1 cr) assigning the values (1 . O"i -c '1"i .if cr ~~ In general we 'f' and :J i = 0,1 ,. !S. cr will be used to denote -1.,. -I.."') 1 2 Trees: 8 A tree is a partial a-recursive function T from a-finite sequences to a-finite sequences such that 1) If one of T (0') of l i { 0*0) T(O*O) J and In this case T(cr*l) will also denote by If 2) T(T) = 0,1 and T(o) i every ~ ~ for which we T(cr)l • J is a limit ordinal then lth cr (or union) of T(cr)O and '1' ( 0*0) T(cr) are proper incompatlble extensionH of T(cr*l) for is defined then so are all T(cr*l) J a , mo~e is the limit T(a) T(cr)i = precisely T(T) is defined iff U T(T)i 'fCO de~ined is for 1" C O . As examples of the many obvious but useful properties of trees we note the following: 3) If T(o) ~ T('f) 4) cr and ~ 1" T (if) is defined then so is and T (0) . If and (J '1" are incompatible and T (0) , T( '1") a""~ .L ..... defined then they are incompatible. We say that an a-finite sequence a C1 (necessarily unique) such that is said to be on T is on T if there is ,---- '1" = T(a) We call Bf3. 1.3 (~f3 <B n > y) A set B c a if there are unboundedly many initial T segments of its representing function on (vy) 'T. «B n f3 , B n t3> f3 , B n f3> is on a beginning of B i.e. ~r) • and denote it by Note that any set that is on some tree is regular. Reduction Procedures: In order to handle the minimality requirements we d.efine an Ct-r.e. approximation to [€]B: (Recall that Clearly if =y [e]B(x) R€ is e-th a-r.e. set) • If is a beginning of . Conversely if there is a beginning of indeed if , B B is regular and B , .- If =y [E]If(x) and then ( E] '\' (x) = y such that = y as well. =y [E]B(x) , a::> ff' then (e]O(x) Using this approximation to [€]B we define the key notion of splitting and prove the lemma that is associated with satisfying a minimality requirement bJi" making ,. and (j distinc·t and Yl. and [ E ] If (x) and If. split = Y2 Lemma 1.4: beginning for e . B t3 Then if °, iff 'T' ::> and there aloe p [E]O(X) x = Yl This is clearlJi" an a-1'.e. relation of Bca is on a tI'ee T and no pair of sequences on T split Suppose of € such that for some Y2 . for p recur sive: ( €] B B [E]B cr for som.e Bt) is a representing function it is a-recursive. Proof: extends [€]B(x) To compute [E]If(X) such that is defined. and so there is one extending ,. Be. [€] B • 0 on T which [E]'T{X) = [E]B(x) On the other hand, since extending 1.8 defj.ned must give this same answer. a-recursive so is ,. By our remarks B, ,. , such that above there 1s beginning of there is no splitting,any look for any Bt:) for which [E]'f(X) Since this search is We now turn to the second way of satisfying the minimality requirement - making sure that B ts a-recursive in 1.5 is a splitting tree for E if whenever are defined they split T(O) T T(o*l) Lenuna 1 .. 6: I f a hyperregular set a splitting tree for B B ~ a [€ J then E" and is regular (it:is on T) and B B to calculate initial segments of' T(O) • is on B T J ~rperregular ~ ca [€]B and so to show how B from [€ ]B step by step: been calculated to be a beginning of has and is a representing function [e]B it suffices [7] to prove that say € T (0,",0) • Since Proof: for [e]B. B. ira advance one more step all we have to do is decide which of ancl T( 0*0) T( 0*1) is a beginning of splitting tl"'ee for distinct Yl Y2 [e]T(a*l) (x) = Y2. such that [e]B Since theI'e is exactly one value for of .for which B for i = ° or 1 .,.. [E] '(x) and Since x and [€]T(a*O) (x) = Yl and is a representing flnlction [€] 1" (x) is defined. T(a*i) is a T we can a-recursively find an € and B" for any 1" a beginning Thus is then beginning of B. 0 This lemma can in fact be proven without the assumption that B is hyperregular [4] but the proof is more difficult and less intuitive. In any case our final set B will be hyperregular .. Now for som.e operations on trees that will indicate how 1t-le intend to expl.oi t lemr.las 1.4 and 1.6. 1 1.7 The Splitting rrrees: ,- on Sp(TJE,-r) (.0) is computed as follo1\Ts: has been computed as For the tree T T(p) , look for the least computation showing that some pair of a-finite sequences ni ~ T(p*i) set split for T(p) Sp(T,€,,-)(a*i) = they are undefined). set Sp(11 ,e,'T) (a) If 1". := Sp (T, €, 1") ~i 1 on T When this pair is found E. 00' with "'1 (Of course if there is no such pair Finally if Ith CJ is a Ii-mit ordinal we Sp(T,e,T)(cr) = U Sp(T,€,~)(p) • p=.cr This definition has one oddity that should be remarked on - the requirement that ~ '1i T(p*i) • Though not needed in ordinary minimal degree arguments it is inclu(led here to make sure that when the operation is iterated into the transfinite the resulting trees do not shrink down to a single path (see 1.9 and 1.10 for further details). This twist was introduced independently 'by MacIntyre [5] to handle the> case of 1.8 The Full Tree: €.. Thus, if for every B to be on some Clearly € no extension on Sp(T,€,,.) Sp(T,e,-r) ,. of € on T which is on some given Sp(T,€~,.) extending (recall that T) • TO 'f0 and ,.1 ,.1 T may have and so we would seem to be stuck. Sp(T, €,,.) and is a splitting tree for we would satisfy every mini.mality What we do in this case is fall back on on L .. This however majr not be possible - B no proper extension on a regular cardinal of· we could arrange for our final set requirement by lel1'Jna 1.6. some begimling a lenh~a 1.4. then there are no If has 7 110' '11 respectively which split ,. for are the immedj~ate extensiorls of T which spl.it T There may, hOi;,fevex, be PI' P2 on T for and so lemma 1.4 would not apply directly. e possibilities. ( €] p PI ( x) ~ [E] ~ ~O. If there is a split P2 ( x) then remarks in 1.3 On the other beginning of and that B [€]B B lies on E . In ei.ther case we will take The desired tree is is on Fu (T, is a subtree of (Tt p.. T; it lies above a the whj~ch B p = ,.0 or will hopefully :Fu (T, p) , the full subtree of p, which is defined by = 'jl have successfully handled the minimality "VIe as indIcated and form a new tree on T( PI) for then by our then lemma 1.4 \'1ill tell us that (modulo B requirement :for above T is not total and so does not concern us. hyperregularity of B) lie. ~O if there is no such split and we make ~and and is unde f in e d for all In this case we try to assure ourselves that is a beginning of ,.1 ~l extending ( E] P ( x) t'~'lO There are P T p) (0') = T (PI ~~(j) where if every sequence on if every sequence on T' Tf extends p) • 1.9 Intersecting Trees: ~ Since our constnlction will take place infinitely many steps we must provide a method to carry us through the limit stages (at successor stages we will essentially Fu). just use Sp or function RI (recursive intersection) whieh operates on a-finite sequences of trees iT \) 1,,~\ • v "-l.. . RI (T } (.0) = \) If (say RI (T } (u) To that end we deflne an a-recursive UT (¢) if they are compatible. \) <).. \) has been calculated and is on every ',) (,-) . - a) T \) \ 'I) then \ve set rr, ..I.. '\) 'I! \) (T \) *i) assuming they are compatible. ordinal U (~) ~ (J Since the is a limit sequ~nce is a-finite we can a-recursively eompute all of this union must be of ') Ith if they are in fact defined and if they are compatible v 'J RI {T } (,.) 'rco' {T,,1,,('X. T Of course if a. length less than o~ RI (T V-1 Thus by the admissiblity a is a partial a-recursive function. It may however be empty and even if not it may fail to be a tree since RI {T } (0'*0) " being defined. can be defined without The following jJnportant lemma obviates thes~ difficulties in our construction. Lemma 1.10: If is an a-fin:Lte sequence of non {T } 'J V(A \) < A , T Ll empty trees such that for every T y by v < '" an application of HI {T\}},,(A ' then is and indeed l::f every T Sp or Proof: Tv ls a non empty tree T 'V \) • T" (,. -*1) \) v on T" (¢) Thus = U T\)(¢). ('r *1) i = 0,1 \J If then are 'J By v < \)f is a nested (J:::: dE~fined al"e nested sequences. We have two possibilities p T a r e nested s i.e. if RI £T\)}(¢) and \) Ir\J' su:ffices to show that for any for som.e for limit is on is a subtree of RI {T} we claim. y Assume the lemma for.all limit ordinals less than sequence and on and every T Fu RI {TV}V(A It is then clear that the then is gotten from HI (T) and has a proper extension on 'V then it has one on RI (T ) • \) v cr ~p T,,('T"'V) is for every 'J , induction it < A T') (T \} *1) C - T'rr'l(\~ '\)-t-"l*i) • Tv +1 :; Sp (T. :J € -' p) or T~ +1 . ~ In either case it is clear that = :Fu(rr ~ , p) ')) '- trl' v·1 pc RI - (¢) and so is on 0 T above v case we recall from the definition of = Tv ('f) ~ p. In the fir~jt that since Sp '- fTl .. \)+1 (1" \)+1 ) T '1 ('f '1*i) ::> T l ''I JA-i) • In the second v+ \)T v\ v case it is immediate from the definition of FU that we have that T +l( 'f+1*i) 'V '\J and so = T'-' (,. '\) *i) . RI {T } (o*i) v We remark that = Sp Thus th~ T (1" -*i) 'J V are compatible U T ('f *i) . 0 v v J Fu and RI not only give recursive trees but that G5del numbers for these trees can be found uniformly from the arguments. 2 2. Minimal degrees without priorities In this section we give the basic min:Lmal degree argument without priorities as in [4,5] where MacIntyre proved that every countable ad.rnissible and every regular cardinal of minimal degrees. L have The method actually can be seen to give the result for every 'E -admissible "as well as for a few ordinals 3 which are not even L2 -admissible. examples. 2.1. The construction: procedures of' length our set in B Stage 0: Stage 0' ~ extension T + where i o 1 and = 0 or 1.8 and let Stage 1: (Ta}a<A. of every At each stage TO be the identi t~y map and let Since e5 of length To ~ U k( 0) ()<cr t3 0 be ¢. is total we can choose an of length at least U ken) which is T'~~5 Rk (5). If t3 cr = t3. let Sp(T 5 J k(o) , t3) Otherw'ise let is total let T 6+1 = Fu(T cr , f3i) is chosen as indicated by the discussion o:f 1 t3 cr = t3 i • In either case is a.-finite then by lennna 1.10 ~ A = U T (¢) cr <"- cr = T O+1 is total. If and Let and and a t')tal lies: B vIe eonstruct we \'1il1 have (j ~(J = 5+1: of k: y onto) a. given by of Let incompatible with it be <a redlction a 13 cr on which To Say vre have a list of the many stages. y an initial segment tree y We also include some relevant is a total subtree T~(¢) ;'- The only d.ifficulty arising i.n the construction is the -) II t::' .. ! possibility that at some limit stage the sequence might not be a-finite. A reasonably straightfor\'lard however" will sIlovl that if ~3-admissible then is the identity and k fT(j}cr<A is always a-f:Lnite. we can in a brute force \'lay require that v= cardinal of some model of of size less tha,n o:r a of' La. as a consider any = a. • Cc a Now f such that and more La is a in which it is also whose cardinality in the model is less cf(a) = U-' C Since V = L a To see that this last example is correct a is the last cardinal of subset of V = L such that the last cardinal of the cofinality of invert or a regular tl' well as all acl1nlssibles regular cardinal of some model of a Alternatively and that every subset is a regular cardinal of some model of than be is a Examples of ordinals with this second property include generally all As b (~) y cal~ulation-, in the model ts in fact a member y all countable admissibles a* (l'cr} cr<A = C <a is bounded below La fCC] ~{e can thus a C- L a • to obtain C itself as and so in fact a member of fCC] is a. bounded L= an a-finite set. 2.2 c~~ The minimality of B: Assuming the construction of 2,1 be carried out we Claim: B is of rr.ininla.l degree. Proof: B is not a-recursive nor evel''1l a-r.e. since 1-;e made sure at step 6 that ~. there ts an a-finite map assures us that on the a-finite set a, say by B I Rk (6) for each 6 < Y (of course t = key] k(5) • = B T6+1 ::: Sp ('1'5 J a) • Consider now any a-reduction procedure is clearly on every k( 5) ~ f3 0+1) and <k If on the other hand Fu (T 5- ' f3 0)' then the discussion of 1.8 assures us is that (modulo the hyperregulari t;yr of -B) if function it is recursive. to make and so if is a representing func'c ion [E]B B ~a [E]B. then lemma 1.6 tells us that T 5+1 Tv' v is a repre:3enti~g [E]B Though we have not taken any steps hyperregular vIe could ee4si1y do so as is done in [5]. B Alternativel,y we can again refer to a proof of le!l1~a 1.4 that does not use hyperregularity [4]. 0 2.3 Some cou.nterexamples: Our first example shows that the man~l construction. of 2.1 does not handle as might at first believe. 1'Iore specifically we indicate h.ow to construct exampl.es of adrnissible ordinals is a regular cardinal of have the property let = u {rt}" ~ U (002 1 .A- Ai" U i<w x = L6 Working in = U A n w2 (A. L~ • i) ~. tf3} , = Ai +1 ~"2 ) = a but a ~ does not be the cofinality and w1 Define sets as follo'T;ls lU3 A n S[ (u A.l. n , ltJ 2) +1 U A~] ..L of w is an initial segment of 2 is the transitive collapse of A has cofinality w since 1J)2 = . U Ai n then w2 • J..(t.fj is a regular cardinal of 11)2 1- u L a sueh that L, we let which is o:C w2 a is clearly an elementary submodel of (WI < 6 < w2) Since U A and cardinality 1.1)2 cf(a) be a Sk.olem-function for S A = w O 1 A Land (*) first admiss:ible after ordina,l s as one is not a member of L (). . L 6 , however, the sequence and so of course not in on the other hand is adlnissible and of cofinall.ty by the elementariness of It is our desired example. A • Our second example touches on the que8tion of hY::9crregu1.arity and exhibit s various ordinal s 1,'lhich do not h9.ve non-·h:yperregular minimal degrees. Claim: La. (Assume cf(a) = and is the last cardinal o! If V == L). then any nQn-hyperregu1ar set tV i.e. every a-r.e. set is a-recursive in Proof: f Since B is B • non-hyperregula~there is a function < !J)a B whose domaln is an initial segment of a and whose - we can clearly take = hand, cf(a) As a range is unbounded in dam f Let = w. wI. Finally, since a. L, this subset is constructlble in dam f ~ C be any o~ On the otner implies that some countable subset of t') is sent unboundedly into v= is the last cardinal such that f is complete B an-d w a) dom f 1 La a-r. e. set., whj&ch we can B,ssume to be regular without any loss of generality [9]. For some limited formula We define a functi.on j: w ~ qJ(x,y), y ill Vy < fen) [~x e La C is Using j as a B. C (::.:) La 1== 3:x w(x,y) j(n) as follows: since in € ~(x,y) regula~ Given a set ~x € (=> while paramete~ is the least V = L Lf(m) :ts a y then ~r € 101 Mn C ~(x,y)] . j it is easy to see that and an . exists C m and the task of deciding if :f(n) > U M In such that assures us that just compute finitely many values of such that m until an f n c ill n < j (n) and if not Mn c such that = ¢ ~ CJ tp (x, y) ¢ J is reached No,., we just check Cl'.-recursi vely x = . ~.' j..p ,.L ......~ .i.l. there yes 2 This argument is essentially just a simplification of one in [13] where Simpson proves that any non-hype:cregular a-r. e. degree is complete for va.2~lous at s . Though he uses a special representati.on theorem for a-r.e. sets whieh is not tr1J.e for arbitrary subsets of a, it is not really needed. We note" for example, that the argu.m.ent given here also works for and that a's such that ~w is the last cardinal of La and sJ.ch cf(a) = w • Finally we would like to point out that no a-r.e. degree can be minimal. by Cor III.3.4 2 'E2-Admissible Ordinals 3.. An ord2nal !:2 over L(l, is L2 -admissible if there is no function which maps a proper in:'tial segment. of a onto a an unbounded subset of Though we will be using a,. 2: 2 functions and so L -aCL'1lisibili ty ratt}er heavily, V.re would like 2 to disguise this fact and make things look as much like the case as possible. We will then be able to ca.rry out most l of our argu.l11ents formally as in the case of ~l -functions ano. L Thus we follow Jensen [1] ion making a' reduction admissibility. from 2: to 2 structure i: by adding on an extra predicate l and working in the expanded La specific we J.et be a complete A 1:]. -set, to t':1e A To b(~ <La.'A>. e. g. A = {<n, \» I r\ E or for the more model-theoretically oriented \ve could let L satisfaction for 1 note the following facts: La, in a more obvious way.. 3.1: 1: 2 admissible code Prop. <=> admissi.ble For such Be 2) over is 1) over 2 over <La,A> , over La 3) <=> is regular and hyperregular. a we also have a is B La iff . a-A Now is recursively enumerable, i. e. is recursively enumerable in Similarly B <La,A> B A 1: A iff B is 1 iff is a-A recursive, i. e. is a-recursive in A B ~l fj2 . On the other ha.nd bounded, iff :recursive notion of is A <La' A>, iff' i: c(, Ii\)} a-~ B B c a is a-A finite, i"e. a-A is a-flnite. Thus for example the cardinal coincides with that of a-cardinal. 2 Except for the second equivalence of (1), which can be found i.n (7,8], the proofs of all these facts are quantifier manipulations using the so are omitted. to 1: out in <La' A> 2 afuaissibility of a and A remark on the roles these facts will play (2) tells us that adjo:Lning is, however, useful. reduce ~~ c:.. by straightforw3.rd A really does 2: , while (1) assures us that vIe can still ca'l"'ry l most of the familiar operations of recursion Finally (3), though essientially a l restatement of the L: -admissibj.lity of a, embodies the form 2 in which this property will be most heavily used. theory on L: -admissibles. The basic recursion theoretic facts such as G5del numJering, the enmneration theorem and the recursion theorem are of course true in any ad..missible structure, but <La,A> is much closer to an admissible ord.inal than an arbitrary admissible set. We give examples of the analogy and the reduction of tltlO that is introduced. First considt:;r the basic lemrn.a 1 of the Sacks-Simpson priority argument [lO] in this generalized to I: setting. Lel1'.ma 3 . 2: [Iv I ') < p} Let ~ for some p be a regular a-A cardinal and let <~ be a unifornLly a-A-r. e. sequenc.e of a-A-finite sets each of a-A-cardinality less than Then U I v<p ') Proof: is a-A finite and of a-A ca.rdinality Since 'He N. < • ~ 'will later need the proof of this lemma Let as well as the statement we repeat the proof from (10]. g be a 1-1 a-A recursive enumeration of U \;<p x {') 1 .. I 'V ~ If the domain o~ g admissibility). g is is less than ~ So say c dom g a-A recursive and :finite and of a-A where g[ ~J n I J is \j N we are of course done (by <La,A> is admissible.} cardinality On the other hand, U J 'J ~. ~ this lemma into the language of {Iv I v < p} Let for some o:f a-cardinality < U I Then 3.3 v ~ < g[~] = ~~ to ~2 ~l we translate La be a sequence of a-finite seta 2 over La . is a-finite and of a-ce.rdinality less than ~. ~ which is uniformly The a-A projection: to be the least ordinal function a-A be a regular a-cardinal and let ~ p is • 0 To ilDlstrate the reduction of 3.2~: g[~] Since P~"A such that there is a partial a , Pa,A 1 (La,A) • For l:2-adJnissible enjoys most of the p:roperties generally associated with and will act as its analog. 1 a* 1 Pa,A < a f: p onto> a '5.: Following Jensen 'we def:tne p set bounded below pI a,A reduction procedures of Of course in of which is it is the last 2: Thus for example, if a-A (and so a) cardinal, any a-A r.e. is a-A finite, and we can index the <La' A> Pa1 , A by a 1 Pu,A list recursively is just the usual ~ ~2 projectl1,m a • All such facts as 3~27 course be proven directly in • a-A-fecursi.ve union and so ,..,re ha-.:.re contradicted the regularity of Lemma Clearly g[N] • U J v<P 'J ' x {\}}, and so is of a-A cardinality < is an v and consider and those about 1 Pa,A could of Lo:, though the manipulations become more involved. The advantage of worktng in <La,A> is that there is no need to reprove them - they all become For example we can essentially m:Lm:tc the Sacks- obvious. Simpson solution of Posts problem in a-recursion theory [10] to get an analogous :L"'1comparability resul t for One word of' warning is in order". however t ~2 sets (eh. II). Though recurs:i.on theoretic and combinatorial arguments tend to carryover to <La' A> without any changes, model theoretlc ones do not. main obstacles arise because a ~l (La,A) -,Skolem hull of The 8, transitive set is not necessarily transitive, while one taken relative to is always transitive. La In particula~the notion of an a-stable ordinal (i. e. one 1"lhi.ch is a of 1: 1 su'omodel La ) does not prove useful in this setting and will be replaced. Remark 3.4: Indeed everJ~hing throughout 1:01" one goes from. any Ll Nothing in this section i.s specific to remains correct when n > 2. to 1: 2 2 ~s replaced by :E 2 n The only real break occurs when • 4-.1: assume We 4. The Construction is L a admissible and work in 2 the sense that our construction will be recursive and our a.G·A priority arguments will be done accordingly in the other hand,$ we are trying to build a set <La,A>. ~J'hich Or: 1s of minimal and so '\'1e are basically interested in reduction degree in procedures in La on our notation: He in rather than <La/A> and all reduction procedures will be assumed to be ones from La so we compromise [E] and r~e. sets and all trees are a-recursive functions. At the end of each stage of our construction we will cr tT~}i<f have an a-A-finite sequence of' trees for some f , - (J such that each of Sp or Fu T~+l is obtained from and each T~ have an a-finite sequence ~a' ~ ~o desired for all by an application RI (T~1i<A • is such that v'lhich is on every f3 cr crt < a • T~ B will of course be our = set of minimal degree. from the straightfoT'\'lard one of <' cr~ The construction will differ ~2 in that the trees will no longer be required to be total functions since such decisions are not a-A recursive. As a result strictly increasing function. can no longer be a fa The Lllpact of the priority argument in the next section will be that though f O' is not monotone it eventually stays above each point of' interest, and so for each i there is eventually a of' the requirement of priority i. T.1 1tlhich takes care Unfort1mately, we cannot in all cases recursively specj_fy 'the priori.ties in adv8.nce, but cr must content ourselves, as in section 5 of [10], vlith an ap- proximation procedure which eventually orders the This approximation will be given by an a-A recursive correctly.. ~ction e priori~ies k(a,e) at stage which tells us what requirement has priority a. The precise definition of on the nature of k(a,e) and will be given in §5. a depends For the p'.lrposes of' understanding the construction it will suffice to note that k(a,€) will have the following properties: 1) 5 <a There is a k(a,e), as a function of initial segment of o. called the doma:in of k such that is eventually constant on ever.f €, (0 should be thought of as the fir.. a.l order ty-pe of" the priorities). He let k(e) = lim k(a,e) (J-?a for < € 2) k(e) For everJ Y<a there is an e < 6 such that =5 • 3) then 6 • ,. If' ,. is the least e (~O" ~ such that e)(k(o,€)=k(e») is zero or a successor. The first tlvO properties are to make the prj_ori ty argu.r!lent work, while the last is .just technically convenient for the construction . At first readj.ng it is probably best to think of identity function as it will in fact be '\"lhen k as the 1 Pa,A = ex, and to ignore the complications introduced by changes in the va.lue 4.2 We begin the construction at stage o = identity and TO 13,) = .0 \ a W'e continue o 9-E by setting follows: Stage cr y + 1. =: when one sets x Keeping in mind the convention that equal to the and there Is no such least < € f such that V < to the least· i of ~y <t y 'i1 +1 then x <t y with some property t, we let = k(cr,E) I- k(V,E) be the "(1 cr equal such that there is no proper extension and set f on Case 1: f (J =: +1 . ,.., TY one 0 f 'T") R ( k (J, has a 11) =: ~i Ch , ~ (J We let TY1. for < i AO. ·1 y ~y' 18 necessar1 say WI1 ~Oy =: fl"t' v =: and set n+1 fa \) + 1 • ~lere any sequence such as extension on T~ the definition of !3 ~y < 11. By lemma 1.10 f O' :Ls again a successor, of the form In (T~,t3) then Now if TY \)+1 Case 2: for some • .' T'! 1 f cr ' By 1.2.1 it has two incompatible n proper extensions on incompatible with By the minimalif:r of TY. proper extension on say le~st ~ 13 y has one on f f3 y ' rr , v t3 y on T~+l. Since this 'Vrould contradict TY -- Sp(T Y , k(Y,\J) , 8) \>+ 1 ') these are the only trees we ever put we must have as in at successor places. successor of which is on both and has a proper We now let T~ t3(J be the immediate dictated by 1.8 and set .0' Y Ti = Ti Note thg,t all the informe"tion needed to carry out these steps is a-A-recursive. Essentially we ask only if certain partial a-recursive functions are defined a.t specified points, and except for these questions which are trivially recursive in and computing the a-A recursive function A we proceed k a-recursively. Let stage X: TO" Since constant as < 11. = J.L8({V I f y < e} is unbounded in A) can change at stage i definition of i i1 assures us that for "l cr ~ only if cr A. We now set lL~ Similarly f, = .\ i < fei '1 k(a,i) f3\ = u f3 0' (j<). . f3 a ( (J (J.L9 < -n) <\) exists for eveTy (3:\) < e) (lim k(cr,v) I k(:\,'J)). T~l = lim T~J. are nested and so we can set i < fA for 0"-+).. can happen when the changes, in f).. < T).. other hand if f If f,,. we set fA is a limit ordinal we set m). lo.p J.).. = Sp ( 'I').. of course:J but f).. = 'v + 1 (which are u..l1bounded in k(cr,\J) ,, 11, is eventually (j-)A there are two possibilities for or ",hen v rr -i ~ 1\ By induction the , < i, the J k ( A.., \) ) ') " t3,:\) TA. :fA. . A- On the = RI {r.[1~1. J. ~ J. <.l.)~ -f' Note that since our construction is a-A recursive everything ~ d b y some sage t ·lS u- A genera-lie. In particular This is the f'" lnle •t iT~}i<f is a-finite A c~J.cial .p' • t -:z 1 -. i\ " an d so a-J.lnl" e ('0 ... rop • ./".;; aDd so T~ is a tree. A use of the L2 -admissibility of a in the construction . The construction is no\v complete. we will make the choice of k explicit and show that the priority argument succeeds :i.e. for every bounded. Before vI8 In the next section E < dom k to' I ! do this ho-:,.. . ever "hOC vrllI show .p . . . a < €} that this is suffices to establish the minimality of Lemma 4 . 3: For every for which Proof: =U y {a < dom k there is a last stage € Moreover if Since I fa < + l} E I fa < € {a B. + l} and then look at ~ f y of U {a I fa < € + I} < \) < y the possibility that limit stage I fa < {a B 4.2 . 1 and 4.2.2 ken) = e and also the least By ~ T)(k(a,~) such ordinal." f If = and ~ € we can find an f Y > y and so the T ~ ,., 'Tl + 1. 'T, \ve claim that that cannot be least. T ~ Since is a limit ordinal greater than assures us that f y-< E - Consider a such that '1'1 is the least and so by lemma 4 .~) there is a first +1 = .. a fa O' In €. such that T such that and so then Finally we have is not a-r.e 'f' cr y - v + 1 0 • B k(~,~) = k(~» implies, by our chaie e of in fy = We in fact show that R€. > > y is not a-recursive. any a f part of the construction tells us that Lemma 4.4: (va would contradict the definition must be unbound.ed in again a contradiction and so Proof: shows us that is a limit ordinal and + l} An i.nspectj_on €. a contradiction. y € fy - constru~tion y + 1 and so f y < € On the other hand if t hi s cas e To prove the f v • I of the successor stage of the Y+l €' is bounded vIe can let lemma it clearly suffices to show that f < € is a successor. ty I ,. f Y ::: cr is a successor: then fa 11 +- I} = ,,+ 1 is unbounded On the other hand 4.2.3 :f ,. < +1 't"l a,nd cr is the first st.age > in vrhich 'T f 0' in case 1 of the construction at stage t3(J that was incompatible with t3cr' cannot equal = ." + 1 cr and so we made sure and so R e ·,,;e mnst have been B, which extends • 0 Re Note however that since our construction is a-A-recursj*ve so is Thus by Proposition 3.1 B. which is hyperregular" and so B 1s c:l-recursive in is hyperregular. B In particUlar we can employ Lemma 1.6 once we choose our trees. The trees we need are obviously the limits of O'-7"U. as ~he ~ A The limits exist because of the priority argument and the fact > i for all 0' > 'T th2D T(ji = Ti'T for all We set T. = lim and note tha.t B is ()n every T.1 that if fa ~ t3 a T~ is on = Ti Lemma 4.5: for every If Proof: Let we can take a > and so TO' 1 = T n+ J TO ~ TO' ,., = T ~ = T"n t3 0) = Sp ( T'n, ' E , (again b:~l 4.3 By len~a be as in lemma 4.2. 'T and 1 n+~ two possibilities for Sp (T ~ , k ( 0, ,.,) Now for the final le:rmla .. to be the last stage at which 'T since is recursive in it. B and n • ~ is a representing function then it [€]B is either recursive or > ~ cr > cr f cr == "'". +1 4.3). The re are it can be t3 a) or In(fJ~~, ~ 0) = In(T 'I In the first case (which can occur either if' cr :I ,.., t3 0.) .. is a limit ordinal or if it is a successor and we were on case 1 at stage a), B is recursive in ,. ]B LE: must be a successor and and so 1.8 tells us that by 1 eIIhTIla 1. 6 • 1'le In the second case must have been in case 2 at stage [€]B is recursive" 0 (J Thus modulo the proof of the priori.ty argument in 95 Vie have proven our main result. Theorem 4.6: If a is ~2 admissible there is a minimal a-degree recursive in the complete a-r.e. set. 0 5. The priority argQment In this section we prove Theorem 5 . 1: For each < dam E k the set IE = try i fa <E} is bou.nded. The proof splits into three cases as in [10]. we will use a different For each case to assign priorities and then prove k that we stop returning to each initial segment of priorities. Before we split into cases, however~ there is one basic fact about the construction which we need in every case. Let Lemma 5.2: 0' E ['rj~'J), then fa = "1 k(o.,y) f' "'0 > E for at most two Proof: E, and suppose that for all < 'J < a and that a's It is clear that for all f '. = k\Tl,Y) y < J [ n~ \») i.n can be a limit ordinal fa € A I f y < >"'+1) is unbounded on a. This ~'lould imply however that ty I f <)..} is also unbounded in ( J . Thus our y assumption that fa L € for cr E [,.,,,v) assures us that only if {y a (,;,'.1) fa :/ € for let cr be the least E if cr € are two possibilities for SP(T~, k{cr,y) , fjo.) is a limit ordinal.. E en,\)) rnO' .I.E Since such that .• f 0 In(T~ , f' cr) > does not change in this interva.l. return to € after a f and before for € ~ E cr = and € = Y+ 1 There E • [~,\)) , T~ Thus in the first case a v ",·ould require some ~6+1 (0 < e < v) but not on If to have a proper extension on T~ = T~ cr ~5 Since t-"A uR +J. ~ A~v this is -Y+ 1 = In(T .\. . - y' 6 G .. 'J' (-I impossible and we do not return to T~ tively assume to cr < e < stage or fa > e In this case see that we can not return to Proof and that we first return T~ = In(T~+l'~o). and so 6 o. Alterna- \i. for so we must be in case 2 of the construction and 6 'before SP(T~,k(cr,y) ,t3 oJ = at a later stage € e 5.1: Now just as above we bef-ore stage 1" 'J. 0 <a Let f P~,A, onto a Let k(cr,€) = f(e) case 1. a-A recursive map from e Pa,A a~ be a partial if there is a computation with GBdel number less than, cr which shows that f(e} is defined and a G5del number for the empty functi.on otherwise. If we take domain k pl a, A ' k has all the properties required of it in seetion ll·.2~ The = only one which is perhaps not inunediately obvious is the It :follows, however, directly from the fact that any first. P~,A is a-A finite (3.3). a-A r.e. subset of this case (~) has the additional property k at most once for each Note that in k(a,e) changes 1 e < Pa,A • The theorem is now an immediate consequence of' an in- duction on after € i€ that €. Ie €+ be the first infinite a-A cardinal (it is, of course, regular) and inductively assume < €+ which takes each of Let above it. At a successor stage we consider the map a such that By 5.2 and fa = (*) E to the least element this map is at worst three- to-one on its domain and there are at most three fa = E domain. which are above all elements of Thus IE cr's with and so not in its Finally, at limit levels, leIlh"1la 3.2 carries the i.nduction s.long for us. 0 l~ 1 " 1 Case 2: ' Pa,A = a, and there is no last a-A eardlnal" This is the simplest case. (j , The domain of €. k k( 0', €} We let is now for all E and it trivially has a (*) the properties required in 4.2 as well as there is no last cardinal vie 'can argue 1.. := ~ exa~tly Since as in case 0 P~,A = a Case 3: N. and there is a last a-A cardinal As usual this is the most inter~sting case. Indeed it is the only one in which we deviate from [10]. Since L1 <La' A> is admissi.ble it is partial function all h 1 formulas of sequence (5, } " 60 . 6)., in 6, . .- to define a which is a Skolem function for <La/A> [1]. 2: eas~r U h[~] 1 Pa,A - a , vJ'e use °,+1 - U 1:1 hE to define a 6, + 1] and this sequence is unbounded C<'X. a (an easy consequence o:f 3" 3) " We now think of the = U Since priorities as being arranged in blocks [0, 6,+1) and within each block as arranged in an N-list by using canon i ca11 y Let Ch osen f ' "• ~ H maps h(x) We approxirnate (6,1 6~+1 = U hcr [ 0 ~ + 1] be the least map f. vr+l " if there is a computation < cr hO (x) :;-:: h(x) showing that onto> is defined and hO(X) = 0 by sequences (oC}: ~g and f: N ~ 0~ = U 5~ C<'X. ~ 0' 6,+1 • otherwise . = hO" [~] Finally let , , fO" We would now naturally defille our approximation to the priority listing by "There that this would not satisfy 4.2.3. we perturb +n , "1 < ~) except To accomodate this requirement k ( a + 1 , ~ .. \) + "'1) = fa+ 1 ( ,.., ) slightly and set k = ~ • v € ') and k(\, N· \)+.,.,) = fA(,..,) if this,value is also \)' f~+l( Tl) + 1 and k that (j--» otherwise. We must now check the other requirements. The computation of values of v A This alteration clearly assures us satisfies 4.2.3. 1) lim fa(n} is entirely determined by the k except for the above perturbation not affect the requirements other than 4.2.:3. wh~ch does t5~},,<y Thus if reaches a constant value so does k(o,e) for E < N · Y • (1 Once 5( = Cc for , ~ y all we r.eed to correctly compute are the computations of of course values of the 5~+1 compute Finally if 6(1+1 C , such that (j > , ~ • 6(, + 1 , 6, ~ + \)) = + Thus at stage (J are all correct for e~+l < E ~'+l 1 , k «(J , in a ~ •\)+" = ~Y+l +1 we ~ 5(+1 ., + \",) - T - 6 - -~~. "J+'!! " 1'+-1 then by definition for each and so for some = e there is a v < N f e(\) = e correctly '\V'e will have Indeed we can say E. C<l more~r fOT every e • 0 To finish this :final case vle induct on J and so by is also correct. 0 Now once we have computed k( (j h t&C 1 is unbounded As They are € correctly and of course never change again. alone we see that 2) X function ~l-Skolem ~Y+l.· definition occur by for h(x) we prove that ~. \) + 'T1. Setting is bounded below an~ 6V + 2 ~ b) J~.v+~ < ~+ • v + 11. By induction calculation for < .~+ u_T 6\)+2 there is an ~·\)+~+l Moreover as x ~ 6\)+1 By lemma 5.2 for some above h(x). L l since' h is a f O' ~.\)+~ h(x) ~ I N. v+ • n there are exactly n a's = ~ • \) +~+1 'rhis can be < 6\)+2 = h[ 0\)+1 + 1]. limit ordinal less than ~~ for the cardlnality of J domain of J ~. \;+il <Lo,A> as 6\)+2 = h[ ()\J+l + 1] " the funetion .., . \,rrn (ax)(vy g is y < ~ recursive. 0 J l:l J~.~+il At stages clearly implies fact empty. a ~. I N• + £ ~.y v n c 6 - y+l o\) +..J.. • The is certainly Thus a bound on Skolern fU..l'lction = g[y] y which g g[ yJ c ~\)+2 • and so ~ y)(g(y) < x) € a-A and thing proceeds trivially: < ~ For is defined from parameters < on some ordinal < 0V+l v . J~ is given as the value of the g( y] Thus An examination of the proof of lem.ma 3.2 Moreover the sentence true in . As for the rest ".. e must • is of course some g • \)..L we can again argue as in case 1 ~. \)+f1 shows that when applied to enumerates 6 +., Skolem function they must all occur below the limit ordinal look more closely. sho~ Jib . S oun d e d ln sentence with parameters ~l to 5~2 such that <2 n ,.3uch that written as a and so using our I\.\\"\. \) c - a v+l above we can employ lemma k that Assume we have succeeded for h is bounded below of the induction every6~+2 for every and so J~.y '" v <y , is in 6~ 1. Open Questions The basic problem is of course to construct a minimal degree for arbitrary 2:1 -admissibles. In partlcular '\Ale view ~~w (of L) as a good test case for new approaches. It is not hard to convince oneself that something different is actually needed: consider the functions B(x) if I Lx F 3: [n]B < n given by cardinals o othervrise • In any procedure like that of section splitting tree until one reaches stage 4 Tn will be a ~ n at which point we must return to it and make it the full subtree of' Thus there will never be a stage at l'lhich all of \'1111 have reached their limits. T. 1. Tn - 1 • i < to Thus there will never be a chance to properly attack requirements a.fter IJ). On the other ha.nd cotu1terexam.ples seem even less likely. 2. In connection ~ji th the second example of 2.3 we ask: what admissible ordinals a-degrees? a For are there non-hyperregulax' minimal Indeed doing much of anything with non-hyperregu.lar sets is something of a challenge. See (13] where some important first steps have been taken. 3. When can we embed arbitrary finite lattices (or even distributive ones) as initial segments of the a-degrees of unsolvability? Chc.pter II G. Sacks and S. Simoson ha~e asked [10, §7 Question 5J if it is obvious that there ar€, for each Cf., L L ) (over n subsets of a ble. If one understands '2 there 2,1"'e L~ n -adY:1issible which are a in :Ls 0'1 \,...' ~ and I II, n reductton nroceclures ':':1icn 1Jut out ex when one feeds in B unqualified !1 -' ., then the answer is an In this sense yes~·. generalization of ('1 B PI::, in '. fl',:.-recursi ve in'( n (replace is a dirEct II by in the definition) and so amenable to the methods of and A 1.5. Indeed one simply choeses a complete a.nd minic s the cons truc tion of [10 J 0. -- A- r . e. set s Band in the other. 1::111 i I:-'7l e:::1 E produc e 1,::·0 1.3 n-l t~'!o neither of which is By the renJ.rks on ti.'Ei.nslation (1.3) this i ate <1 y g i vet h e des ire d res u 1 t for t his d 8 f' j, n 1- tion of There is, however, the more obvious and natural notion of are The .,. /\ -n in '< L and n 3. n Sit] e r> B to be consider formulas of IT d n :i nth i s C cu3 e definition is easily seen to bility than the for~er (L a ,C) i s s t 111 ;; yes ~ no longer quite so obvious. for ~ is n in C iff there which define f but the Dr 0 0 3. f is First of all, this later ve a much n > 2. stron~er reduci- Consider, for eX22ple, ani'" Hi th cor-:nlc:te 11: A and B. Cle2:elv }J £ D, to the second de -f'::Lni tion. ~n ever, any thins A in should also note that as n~ n in" (A) n and 2.C T.,!:! ttl the fir s t is ~n c:L.::f'ini tio:':, hO'..J - and so 4 ~n(0), B 20rd inc t B ~n(A). (One the second notion of is not transitive vrhile the first clea:rl J is). r The second definition, moreover, is really model- theoretic in nature and one can ga5n no direct hold on it via purely recursion-theoretic constructions. son for this is that individual fnets about via the SeCODe} definition of is 13 in general, on unboundedly much of C. very nature the first definition uses about C for any sinsle decision. a construction inside much informatio~~e L a ~le depend, Of course by its on~y a-finitely much Since at any stage of a-finitely would seem to be at a loss as far as is concer'ned. It ~'lould could hope to handle would be reduction procedures which were L /L . n a. The solution to this difficulty, by Prof. C" we car only handle the strong notion of seem that all derived 3 in tIn The rea- ~ili~ch was suggested Sacks, is to mix a forcing construction into the priority argument to produce sets which are sufficiently generic so that the titlO notions of reduc=_bility coincide. More specifically, one hones to make the truth of sentences about C depend on only L n a-finitely much cf via a forcing re12.. t ion H.hic h is it self Thts ":0'.110. C 1.'~ ,.."{ then give us a method of handling the rotion of stron~ directly in a construction inside T oW a. . carry out this suggestion below to prove the theorem. We also answer question 3 of [10J in this chanter carrying out our construction of sets uniformly. integers k Thus for each and 1 a. , Godel numbers for incomparable. which are, rn !J n -- inc ornparab 1e . n fo~ by L n we show that there are every subsets of a 2: -admissible n which are ~- n (N. Leman and S. Simpson have informed us that they have also found a unlf'ol"In solut:tcn for the case n,= 1.) 4 1. The Forclnrr ~elation He assume that the reader has a gen2ral familiD.rity with various standard forcin~ descrlbe the partlcular systen but ~ive procedures. ~'Ji th no proofs for the standard '.[hie::l therefore ~e 'liTe are concern.2d The basic thec~e~s. techniques and lemmas for the subject in various recursiontheoretic settings can be found for examnle in 1.1 Defini t ions: ?he gener2.1 f:"'arllev.7ork will be working is that of Cohen style ~Iore models of "VJeal: f:ct theories. be fore ins over L 's G sions (see I.l). is that the for ex ex \'ii thin c=~ass intend specifically ~:Jhich 'de ~de viill Our forcing E ·-adnisslble. n inelu-- j.nterp:cet::.1tion, of course, seq~ence lS a-finite 1• forcing over -finite sequences ordered ~he rL '--,q an initial segment of the represent inn; :''llnct::Lon for our Q"e:neric precLtcate. far 'as lansu2se is concerned ours ~'lill r"S contatn, i,n addi- tion to the usual logical symbols, bounded quantifiers, £, =, constant symbols predicate symbol G a for each and a unary for the generic predicate. ~e will, however, in our construction only consider formulas in strict pren2x nor0al forri, i.e. bounded quanti.fiers C'l¢ we ~'lill follo~'T ones in which all all unbounded ones. T:'lUS by med.n the forT1ula r;otten by crlanzing ;':.11 quan- fore on the 60 relation unrestrjct nu~t:er For ¢, forcins eqhals truth. formulas ifr every constant symbol in cr\~ ~ and L,~~ -'- v 6 ~ a J 1 The inducti7estep is as usual: C3a ~ 10.)(0\1- ()(§:.)) belongs to representin~ ~unction. at\- :jjxQ(x) iff ifJ~ and 0) (T >¥ -3 X'1' (x) ) . (1fT J 6 nrec isel;1 ~ 0 ...'"'0 is internreted as having G ~hen as an initial segment of its a der~d-led b~/ induct10n GIl- Q ls T/ith thls c.efinition lt is easy to prove all the standard facts about forcin;; in tlJe usual inducti ve ;'lay. ¢ Cn > 1) En is a \l- ¢ relation while for Thus in every case, for formulas in is (J (J h- iP the is rn One should also note that for formulas =a n Z n-,IV it IT n-- l ' These basic facts will enable us to use ~. n rn-admissibility of a to carry out the forcing part of our construction. 1.2 r n -genericitv: r, iff .for every (La' G >F 9 Band C not:i.on of ditions. that if we call a subset <J 'J sentence n <=> (3 a ~ - r. n _r)"A-ne""'ic .)..':",_ . . .1 n!J. n ..... ,-1, f G) ( (j 'J- rp ) • 'Fa 'I ' - in!! via 'i' OUI' of En-g._.eneric a j language By making our L: n sets \,r111 be able to handle the strong.,., .,J a-finite facts, i.e. As an examnle of the is of G • then un-generlc po~er forcins con- this gives us we note is fJ n G strong sense iff it is also true in the weak sense. in the We must l~rove and 1[; x G of and <=> Say there are (x) (x € 8 <=> <La <Lo.,G) rL ~)(x)). i B <=> is b. 11 in x, x G En e 0(x)). formulas n ,G') 1= j (30 ~ G)(ot~ sentences is [ Then b:'T the we have that for all x En B such that B .J,. 'f; the if Dart: ¢ 2: ¢ (x) Ln -~enerici+;'v ' ,- B <=> (30 C 0)(0 I~ ¢(x») Since forcin~ for this is precisely the statenent that in the v!eak sense. Finally we remark that by the usual lemmas about forcing (in particular the \Iforcinr; equals truth;; lemma) being I _ n l a,~ rn-generic is equivalent to there being, for each formula ,¢. ¢, a . a c G such that al~ or ¢ Indeed these are the requirements that we will try to satisfy in our construction. 1.3 Before proceeding we should comment on the restriction to strict prene~ formulas. This corresncnds to yet another variation in tl1e meaning: of (or ;! 11 n in'1 ) . Though we will for the sake of sjrnplicity prinarily concern ourselves ~!Ti th strict En formulas, we will indicate in section 4 how to add on additional requirements to the construction and handle E n formulas in which there are no restrictions on the positions of the bounded quantifiers. !W "., --",' 2. i I The Construction We first glve an jntuitive oicture of what we are trying to do 2nd then rri ve the actual details of the con·In struc~tion. sections, 13 bot~1 ally syT'1T1etr-ic roles and ~;,Je C cmd generall~l 'Hill attentions to the 3 necessary we shall dlstin~uish bet~een 2.1 n assume and each L sure that no '.'1 hen the n-l En a is sentence. ~he towards our first ~oal goals. tNO co~ple~ent To beg in, '/! e 'Ar :i.ll assi~n The 0 of C rod e r priorities Once this has been done we nroceed n, by, at each stage current guess at an initial segment of r n- 1 nota·- second will be to make sentence defines the to these reauire2ents. C forcing conditions contained sentences in some way and so n and rn-admissible. G i s in t e r pre t e cl a s ' ) . E ~~e~ Throughout the rest of this chapt0r first will be to decide by in 3 The construction will have The idea: confine our side of the construction. tions by subscripting. we fix Tllill play virt;"l" B extendin~ our to force the formula of highest priority not yet decided or its nega t.ion ~ tension By al~ia.y s sentences extenslon. ~t3 the asual J8!:lmas about exi::; t s . A ~n the~e ~oreover is a 6n for(:in,~ -' such an ex-" as forcing for L: n·-l U 1 . n-l function givinz such an ~or in the usual fashion. have a potential witness X X define C. into £: for~ula 6 ye will s E: our current a0proximation to we put rn eac~ B forces -"\i - thus making sure that ~'Je Of course in both parts necessary negative information about then (X-l"G) does not ·.·r:i.ll protect the B only against later encroachments by reouirements of lower priority. The problem with making this construction independent of a arises, of course, with the specification of ority ordering. 1.Te pri- In chanter I as in [lOJ various orderings depending on internal characteristics of in advance. t~e a are specified will here uniformly generate (v:La approx- imations, of course) a priority llstin~ that, for each turns out to be essentially the same as the one ~iven a, in I. 5 .1. The idea ls to :niriic the proof for case 3 in I. 5 (there is a last a parameterless a-cardinal and P~ \'1e first use to generate blocks Then within eRch block we act as if we were indeed in the situation of case 0 a). r. n of requirements. block is = 3, that is, if the we consider the reauirements fer formulas in =: h[o] ordered in a d type list. The proofs will then follow 1.5 and [10J. 2.2 The priority listing: Slnce one can easily get a parameterless a is r. -admissible n E -Skolem function n [1 ~ . h in T ~~. , z:; .,..1 1 ...J... (')). \. -:. • U l:<A If thpre is a last 1':11~~ blocks 11- a-cardinal arld P = a then these a eventually coineide wi th those of I. 5.3. If there is no last cardinal then the bounded in . lng . .,... ,- ] ..L.~). a last (L ') ) ' 8~ sit ua t 1 (, n tion a \":ill be un- and l'lill incluje e·ach Finallvv if • on < a , CI. ~nto which projects be as 1 n I . 5 • 1 \"1 iII a-c:ardinal 'thus mlmic-· then there will be a 0 f' c:) u r s e in the con s t rue -- (1 ) . but this is Tdhat ca.n only approxinate the 1'le "'.jill happen "tn the lirr,it li • UsinG the notation of 1.5.1 (3) ~:je let 1-1 map froT:1 u onto least ordinal for . \" l • by stage en e• for some and y = enc; s < E' hnf~(Ef). ~'Jhere W D ,\ hnrnCe::) :Ls the S theI'~ ls such a map constructed (requirement ~) at stage R y a-card :tnal i ty if' n' I to have S) B = hnriCs) ~ ~',E' and there is no pair < and ~oreover for fixed c; £' end l·-itb either < E: E < '? ment 8n '? RQ We now define higher priority than (o;+l)n x be the lec1st n<Cl ' t es t'ne approxlma ~ be the ant,~i .. r.: n. at stage proxirr.at:ton to o~ associated B wi~~ has hi7he~ ~:;.uch enr; that a reauire- priority t~an I"'" 'v • 2.3 rrhe reqt}lrer:lents: 'iTe be n by making a.n Ci -re:;urs :1ve will be two types of requirements If 'E:: ~ En _ 1 sentence,then at sta~A cr n , our current initial segment of that or is a ¢. (for both <P E: • then ..P \... S If ¢ £ fled at stage C) • rerplires n B, forces is a reouires that say that any Band does not define require~ent R He of the first type is satis- s n It ls a. stann + 1 dard fact about forcing that at any stage we can satisfy any unsatisfied requirenent of this type by choosing an • a proper extension of Moreover, by the remarks at the end of 1.1, this operation is ~. (and so relatively harm- n less). Satisfying the second type of requirenent is More dif- ficul t but follar,;!s the U3\J2.1 DC. ttern of the Priedl:>:.~:r}.;_· !·1uchnik theorem. ~ We first split into ~any a disjoint recursive subsets will come fro~ Z stap;e quantifiE:!' ','Ihich COLl.rSe j~f ..-:nB \ Ir:.J j.3 in n "'" '1" E: we will have a for every from requir~~ent n + 1, l.r does not define ¢E:: At each stage E: specific potential witness He can The idea. is that < a. E:: potential witnesses to the fact that C unbounded, nqirwise ~ < a. of the second type at E: n (·'1' vi Cx_£.' ~, E: ~ ~ ,-1. ':1 '-- "~ii tnos"" . ,~ I < 4 ,.) f'oY> ~ .J- into the first c. 1s also since we have nut a bound on the witness for the init 1 existential ~ Of course if then H 2.4 t: ntlfler. is n0t of either of these forrs E: and '/Ie never cone ern ourseJ..ve s 5.s n.ull The construction: sta~e At each a-finite sequences approxi~,ate for C. Band one sten L: n ~1ich ~D :1. t . Ne will have an are intended to of the I'epresent segm,.~nt s rnhese sequ.::nc2s 'cTill be changed froY:l s':ep to tLeir length. increas be initial n ~J1 th the ~)ince the next will he ~n of DrOCf-?SS ~;oin~ from the sets we enumerate ~ill • We begin qt stage o by setting, for both and f"\ V X each < a. E: E: equal to the least elements of n + At stage 1 we let fled requrie0ent of hiChsst pl'tori ty 1rJi th be the unsatis- B. 1:'!e n0 1;.J satisf~T some changes in the causes. _. .A- nI S :·1.ore prec isel:" D ;"p as dictated ~33.t- is associated by 2 .3 2nd IEake to reflect the injuries this is a if require~ent of the extension of exT1)c E: . set nO~:J " i:\B first type we let priority t"lan can \'Jhi~h He fro be specific let us assume that lsfy. for Fer o~ for all of hirsher E: ;-' lower nric:rity "nan . eoua.l to the leFl.st ele 1nent of '. '7 1.J S above ste~'; and :tno.l1y all € < a. If 1\,-.; ~ to ("1 V • . for is cd."' tne second Formally ':Ie t~TDf.~ then we a.dd set Everything else remains unchanged. Had T? J S been"'" . 0. requireaent we would of course have proceeded dually. At lir:11t sta.g:es we sinply set, for both an} 3. ~he Priority Argunent In this section we will show that, for act to satisfy require~ents with h1~he~ e~ch we B~ priority than only boundedly often. T',TS C2D then e8.S j.ly conclud.e S , and constructed above ~2ve the detv t:la t the set s R .... sired properties. 3.1 1]:111e first point to consider ::'8 our apprcxiL-:ti.tion for the 6~. U l:;~y 0~ Let = a. be the least y it ordi~al suc~ t~at ca~i be elthe:c a l.i;:lit or successor ordInal. In either case the implies tliat:> for eaC~l such tha.t for all such a stage has been stase 3. v < ~eQched orn~al"d lie in any block < every ~ y and 'J all and < v n' \) 11 > (..J ) o .,.. _ - ~ fOI" (J .,.. Once all s < 'V. ~j1inally for every x ~)riorities Hhcse He note that :"'or n v , E < 0V+1 there is a stage such th2.t < 0 \)..l.. +1 V ":> ':> requj.rer"ent~3 ctre fixed. s < 8v T) , we can, of course, find a nv < 8 v +1 ' by which later stage Prom such v < y, there is a st3.£:e < s and n > - nE 4 r:I:hus vIe c a.n al ~Ja'y s as s uyne any J.ni t ial s egr::ent 0 f 'Che pr>i-- ority listing to be fixed from some now prove the main T~~m8 ~-':'=.:.:.:... __ )~. :-~~_~.... sta~e onw~rd. We can lerr~rna. -.~o~. .. . 0a._c~ '- _ 1! v < Y, ~ _ < g ..) _ , ~.te~p .... _ ~ _ l·~ ~.' ~.J.--. hClj·.r.. ,d. ~ on the stages at we act to ~hich s~tisf~ requirc~ents of higler prior1ty t an Proof: We proceed by induction to prove an even stron~er result: at 1,{hich He than 1) JV,E J the set of stages after nV,E act to sati::.fy requirements of hi;ller priority 2\ J < E+ • , is bounded belo,;; <5 p l'hf veE) v+l 'llhe proof is siI:dlar to 1.5.1. lemrYla 1.5.2 / Corresponding to have the fact thc"l.t once VIe V,E h2,ve stopped ~'Je acting to satisfy all requiremerts of higher priority than R ' S we act at most once more to satisfy R a successor (2) is iIT~ediQte by acting to satisfy an at vThich hull. • +- I-> h :i a S~~olem s a also continues to hold For lS llO induction. L n ~ ~ ioreover our ,~ fact and so the stage occurs lies inside the appropriate lv As ::> Thus for S fu.nction, 'v'le L: Skolem have that (1) by the definition of 0\>+1. a limit ordinal lemma 1.3.2' (for E n L -admisr. sibles) carries along the induction for (2) since our construction is ~n. To see that (1) remains true we argue more closely from the proof of 1.3.2' as we did from that of 1.3.2 in the last paragraph of l.5e Of cour2e;1 at stages of the induction corresponding to trivially as in I . 5 : that c J\>+l , J -- for limit v s u V J V,E: c °V+l - +l < °\)+2 and s::: 0 Vle proceed ror every ::: J\>+l,O ::: O. The a.rgument is equally trivial .. to ou!" 21 via so~e simple lemmas. E < B V such that Proof: If y y o y = U h[e \) +lJ - a.. D = hf \) (E). is a limit ordinal the result is obvious, so assume tells us that = S = v + 1. By the definition of 'Thus t,he don h A 0v + 1 is not ",' .. i • parameters Lemma Vie 3 .l~: Proof: By lemma ~1ave that L -generic ~ n Band' Care Consider any L: and E 3.3 there a.re \) < y It Of course, Thus immeGiately a-finite. and so over ;ts tran- is, however, definable over sitive collapse y, sentence ~ n-J. < ev ¢ Qp By 1 errl.llla • such that hfv(E) = S. 3.2 we can go to a stage after which we never act A.t to satisfy a requirenent of higher priority than such a stage we if it is not yet satis- must satisfy Moreover, the procedure for Doving the fied. so protecting this requirenent assures us that remain satisfied. can now conclude the s ~S and will is in fact decided by Thus each an initial segment of each set ~e x , cons~ructed. D~oof. Proof -- - L LeL, he any 1>13 variable and let 'J and s For the sa:{e of definiteness 1!Ihen define is G the s t a C. e F; i ve n for" from tbe n given by lemma 3 . 3 . be 3.S ~:Je sho;/J inte~preted \) cO~-lstruct5""on formdJ.a \,:1 th one free \' i.> l- and [: / L ljJ K" t· , ...J/'- r;,~ (\"i\. v n o.-~ is Zn -1 , p~7 '/) -- as n' n if Rn ~ be ':! e see n Say • , n· > vThere be C?_ 1'[:-:.. t n e s s fer the in i t i 2.1 and so represent nr , there anG rk(a) such that an initial seg~ent of is not satisfied at stage not yet in 3. 2 • does no"C then by assunptlon an cl 1 e t a an p p is constant for that Quant ifi:2r. 1s a stace ~) Let b Y 1 e mE; 3. Let its value be ~ t~lat D. We now see that n + 1 (i.e. vAS is C) 't're can and indeed must tilen satisfy it. To do this we put into C -- a contradiction. Finally ve note that our construction does not denend on a nor does it mention any infinite parameters. we have in fact establi procised at the beginning of Thus the theorem with the uniformity t~e chapter. 4. As r ...:.., Pres ervin:-~ 0 i1 t sed in 1. 3 n n O~,J i nd i i:; E. Gat e 1'1o:-r but not recessarily in s~rict pre~ex for~. 4.1 ~?tructure If the E~(B) then every v T .l~: n ad cL:'ng used in are a. ~(~Tnaj .. ns stri~t formula is equivalent to cne in as an additional (' E T .LI .fCJT'h ~ of: the r t n20 r e L ;12red to can be s prenex 0U • sirQ5-1a.r to those [lO,§6] and III.4 to make sure that t\'JO destr~y We achieve this recult by uire:-;lents to our cOLstruction vario~s sets We treat each st ict hyperregul~r. mUl.-3. vlith a does not p~e~icate for'- free varJ..ELbles as a function and try to preserve computations on cortain initial segments on which it is tot Of course, this is handled via the forcing relation and so the construction remains priori ty art·;urnent conputat ( 2, C for each '\ralne ~ 1 ts evei.Tt:;ually keep ~:Je ran;e~!8uD.ject. (I obvjou31~l the first n in. t h 1 S con t the rune tlon itself This By the preser~,tln~ t ua 11 y a for c 5_ n ~s c c' n d i t i T~nus ~n' bEcome;:" suffIces to make Lin t~ X t ) and ¢o, preser- that it suffices to consider, for a function tJ vations on initial segments up to the cofinality (in of ev (where = (3 hf \) (€:) we add on requ.irements rn a (strict prenex) Let cr ll (S) be the '0 ). ~ p y~,s Also let ,'rhi-ch all does not force a= n n hif (s) v R pQ of our neVI kind if rro for such that Q' f..) C" >.J La-cofinality of 11 • If . £: ¢ a ) this (-< " AV is B formula witn two free variables. stage value. < G\) ) for some L b~T as COillDuted 0 ':: y < cfT)e en) be the least fer v to have a specific Q }-. ..1 Tl-r1!(s) .. \> . we say that we can satisfy an '\TTl ,}\).> this stage an extension of £ < cfT)(e n ) and V (J il VIe vrhich forces can find a.t (v n , Oq",\) ~.J to have a specific value. ) S ) These new requirements fit natu- rally into the priority ordering. The construction pro- ceeds as before by always trying to satisfy the requirement movi~g of highest priority possible and then appropriately corresponding to requirements of lower priority. the 4.3 r~ehe proof: have only to indicate how the new re- quirements are handled in the proof of the priority lemnla. Everything else in §3 remains the same. foJlows as indicated in with the required Lemma l~. 1{ : once the r n -admissibility priority lemma is establishea the <L a ,B) ~oreover, 4.1. of We therefore conclude le~~a. For each v < y, E < e v there is a bou~d on the star;2S at -;orh1ch VIe act to satisf;! requirements of higher priority than Rhf (c). \) First note that we can adjust Proof: Cf1\;,E: (8 sary to make sure that mind the proof of' 3.2, e that for each U{~J"\I,t:-c-,!s' < K£ - immediat~2. of 3. 2 { T U v ,c.,... f If have "'r(:~ ~ £ T R ....Tr'" the set & R, E: < £ f' n~. (c" v '- For 2. bound £ & R hf (£:) v . I~eePinG is of the nev.]" typ~} I < <5 v +1..l. T1 aT Any on is of an old type} for s' < £: and . TIle conslde:t' a member of T y < cf 8 one corresponding element, T, to force T 1'n f v ( E t) (v Co\) rh, i~ l\ n. is a En K£ - n ~ En -admissibilitv of . a v £: < 6 n mao from ~ < for which £ with v < pn a W x T cf 8"v ~he bound is a member of the 1\ £: - T) • a-finite. onto the Thu.s by the ' En yr £: it is its range is bounded. - the existence of a bound is itself a and so Thus at which we acted corresponding to them. eler:1ents of < 6 £ ' cree v ) subset of We nO·tl ha.v8 a total 1-1 T1 there can be at most of H unbounded in there are W v of 0- S to have a value. ,) ,2 We rirst handle the set Since is a successor this fact is E 1s necessarily permanent by the definition of for each in is a limi t ordinal then by the aI'guill('rlts s n v ev < E ) = cf(e ). if neces- £ see that it suffj.c es to Drove T'!1e 6 v +l . bounded below () < y, v nv Moreover, fact in parameters )' "1'1 2,kolem huJ 1 c of _. n In pa.rticu1ar, not in Now we deal with the elements of vIe clain that the assoclated 8 \) . If cf(e V ) < 8 ''0 Yv,s' T 8 is a regular v are unbounded in ,S into D 9. by sendin0; s Y a-+ €' is the least such member of ~ n and a 3 < 8 a subset of s x is in f 2. C t a -. f ~ "'- j - contrad~cted 'J lS the ~he . The correspondence is to. ::1 'tTi th and the As j. n:i. teo P" f\S of this bound is a 0v T are in a natural 1-1 correSpODc.ence E: n tells us that and Since this nan ~o Thus 0 \) 8 be the. bound just established. v L L s Gvo elements of subset is KE:. - the segr;lent of i1he:re n regularity of Let segnent of ~inal E -admissible we have is a-cardinal and the maD frOf:1 trJis final ~l On the ','Ie see then tha t P \)v· involved nust cover a vre can cIeri-ne 7,J are bounded below then thi.s is iIi1Inedlate" other hand, suppose J.s _. n rn r n -admissibilitv,. of 1.s bounded. a As before the existence fact depending on parameters below and so is indeed attained below 6 V +'. .i. Chapter III The Splitting eoren In ordinary recursion theory one can distinGuish various types of priority arguments. between fin~_ te The najor split is andinfini tein,i ury constructions but finer distinction3 can and should be Thus for example one d~awn. should mark the difference tJet-'reer, the arguments for the Friedburl,,;->,Iuchnik solution of ?ost' s problem [10 J sp l 1·t~· vln~ theorem r~ LO, l~l ~j. and Sac:(;3' In the former there 1.s an a ETiori recurslve bound on the preservat10ns initiated for a given requirement and so on tlle injuries it inflicts on lesser requircnents. There are, however, no such bounds a va i 1 a b 19 1 nth e proof 0 f t h ,2 1 a t t e r t:1 e o:e e til. As long as one renalns ','Ii thiE the confines o~· o:;':-'GJ.nary recursion theory these distinctions are relatively unimpor- tanto All that one ever needs in the proofs is that the injuries to each initial seFment of requirenents are bounded. Since the union over an initial segoent of w of finite sets is finite, the proofs go through without regard to the more subtle questions of the existence of recursive bounds on the inj ury set 3. The 81 tuG. ~~ion changes draLa t ically \;Jnen one enters the realm of recursion tllGOry generalized to all admissible ordinals r~l1e 0,. an ini t :.0.1 se gr.;ent of a problen: is that the union of set s each neI~al Q O'-ler' f ;/ihich iso, -1'1ni t e tIle cell. etlon itself must be given as an. C'k-flnite u.nion of one to be sure that the proor of the UniOY1 is ?riedburg-Muchn~k a-finj_te sets for a-·fini te. theorem, towever, one is not so lucky as to have everything presented on a silver platter. of the In the Indeed} a cruciD.l point theore~ S2.c>~s-;:)1;".pson proof a-recursion theory [10J is their lemma in 2.3 (see I.3) which enables them to handle unions which are only Unfortuna.tely, their lef:1l.-rna only a?plies a-r. e. . . bound on the size to collections with a un-,- f arm §:. Rrlorl ~ of the .i1erILers. It is tl181'efore admirab1y sui ted to pric1r-- i ty argurle:nts of the Friedbure;-i\'~uchnik ty'pe bu t does not seem to suffice for ones like the splitting theorem lack appropriate bounds. ~~ich Indeed to date the only priority arguments that have been carried out for all admissibles have been of the Friedburg-T'ruchnl.k type. 1~:re In thls ehapter exhibi t a construction that enables one to handle nrior-- ity arguments of the second type. a In particular we Drove strong form of the splitting theoren for all admissible ordinals. TheoreF1.: be a non-recursive a-r.e. set,'S A,B B. <~ C i\ be a regular Let and a-r.e. set. 8 and such that D Then there are regular such that D a-r.e. set and is nat AVB =: C, AI\, B :::: ¢, a-recursive in A or Note: We have been informed tI. Lerman that C. Teo Chong has independently proven that there are incoLipa.rable a-r. e. degrees belo~'l every stronger result is a s 1 c,:-r. e. degree. C. .L ..... Though an even corollary of our theorem (Cor. 3.3) the same type of priority argument is needed even for the weaker result. ~e have net seen his proof but have been giv2n to understand that his methods are rather dif- ferent from ours. 1. The Construction The intuitive picture: 1.1 Ve ber1n by noting that since a-r.e. degree contains a regular set we can assume every ~·.;ithout loss of generality that c and d be cne--one C and D respectively. approximations to DO = n is regular. .i./ I:Ie let Cl,-·recursiv2 func;tions which enumerate As usual, and (' v denote the associated \'Te IU CO = by c (1) and 1<0 LJ r1(lO) \ u • 1<0 The general plan of the construction calls for pu.tting c(o) into exactly one of vlill assure us that A , 3, -c~ < We will also v - 0v . to keep out of as soc iated A. or n 0. AU13 = B v~riGus ...... (? 0- This and that elements that we wish for the sake of requ.ir'2r.lcnts the co:yli t i on that These at stage = C, A/"tE h~ve ~ and 't:j_ th and fA ~equ1rements ~ill D not be recurs J. ve 1n be ordered ty a priority sys tern that \;il1 be develoDed along '.11 th the construe tioD ° At any stage we will have an approximation to the final a priority listing lirhich we \'7illu~)e c(o) B. is put into \<Th1chever ~.',rill A or pre servc~ the to determine whether Essentially we will choose r~?qu 1renent s of tlH: highest possible priority. In line with Sacks' original approach to this theorem [6J \'Ie try to insure that "":/,"": ":" 4" 1. / .. j,.:to. A \ S ) by the rounciabcut in~tiaJ on ~1he sez~ents as lon~ ey seeE to agree Nith 2S t [') idea is that if for'" SO;-'l"? E: = [s l'~ j D, then TiTe J. 'lIould eventually be preserving the first available comrutation each D basi s 0 would then be able to a-recursively -- a contradiction. hO~'iever) ,;::e f such c on3 idel"2ttions, that there is SO:llC of C01~j.t·DJu~atl·on,~~ _ ~ _ _ '-' _ ~ x. priori ~alue bound on the ~!'._cJA\/X'). c. • On the can only- ar p:ue for Hhich uc preserve x' s cannot as s =1.3n any unj.form ~'le to this bound and so find ourselves heir to the problerns discussed in the introduction to this chapter. Our strategy for handlins these difficulties is to first arrange the requirerLents then to consider requirement. [S " lJR r~ ~~ 1\ [s J1->. f or a 11 D in in blocks ? ? and as a single Since the preservations for such requirements only involve keeping elements out of conflict v:ithin a given block. whole block t P as a si~gle with our ability to recover associated with P A, there can be no Moreover, considering the requirement will not interfere D should the preservations be unbounded. Finally we will have the determina:tion of' the size aDd number of these blocks interwoveD with the construction in suc~ a way that the blocks progress through the list of reduction procedures cofinally with the progression of preservations and injuries through a. 1.2 actual construction: ~he Since and f entire explicitly only the that s par~. : It is of course understood lIar steps are to be taken on behalf of ., bec;:tnnin c the constructlon ",[e let - f:o. cr.-recursive pro(.iecticn. W~ will be ~n the least x for with argument in De an ~.Jill For eDC h of 1,'!hic h e~ch ~e y fin0 y - A reouiren--jent associated with a y - A active reduction y A active reduction nrocedure - 1 1'0 [f- eJd ex) for "'lJhich such computation and create a _u = y - D (x), we take the least A the elements required to be out of of * there 18 no If there is a nrocedure. a~ of r;0" Y blocks h~ve will ~'!hich x ('J. Before 1;i1tr the const:ruction. se~nont initial l_--..L) B. of our terrn1nolor:",1 T'::~--rest si:'~'ulta~eousl~T be defined ... by the chosen A comDutatiQn. If at any ment into cedure is a x A E < y ":le destr<?"y' a* A is x the requIrement. Y - A require~ent associated with 1.e. we put an element of a sta~e E active at y - reauctton nro-- A a sta~e A require- unless there (as yet unde~3troyed) Hith arpument such that r-. ,..,-1 LIE: J ..,.p.~(j I \. X ) I,,; has been enumerated in D since the ( ) r....t -,0 J .X _.- reauire~ent was created. The idea is that as long as we seem to have a computation showin~ that r-.;', L-l. -1 E: In" r n -t jj \V'e need Day no '71 ( ..L further attention to €. Now for the definition of the blocks where Y~ stage T = U{T<Y' V C at stage (~8<y) (a P 5-requirenent is created at 6-re~uirement an element of a, is enumerated in T}. The idea is that (for some A) ea'~h block eventually reach a constant value which will reflect (via injur~es a bound on all f) c-requiremc;nts for shOvT D that caused by our having to C < A. :Jnce v;e have such a ()ound alluded to in 1.1. :'ioreover if U Py = a* A :Ls the least ordi-- then every P handled and we will succeed in getting n Y<A Finallv we take sider the sets y ed r-u :1~ Io(I o ) ,4.... D,. lU t: - be the least ordinal a y-requirement. A(B~ of J.._' t' l' n• ,0",'._ . If c (, rr) .i -~ y < ~ 8 ~ 3 or . fI- 2 so ;;~ore precisel~Y) ~'Te j~"\. ( ;3 'I J J • 3uch th2.t 8 A, B. con- requirements which would j .. n" t 0 '.1 A 'Nill be so y and put it into C(0) as to preserve as l':luch as Dossible. '0" pst ~.. 0 ~... can N2 is bounded vic. the arguluent 2.bout recovering nal such that b '-~~ preser~e we put L~ P~ ,...v .r ( U.r B ) u.;1 contains c(o) into d; require;:'"1ent priority ave!:' a y 2 8. structiorJL. 0 - B r'equirener.t iff our description of tho con- ~he 2. Priority Arsument Our prima.ry concern is to show that enough bloc I\:8 eventually settle down so that we can 8rgue that [f-1sJA ~ D s < G*. for every Lemma 2.1: If by some stabilized at values N* v.. ~ < .L () 0 an d sta~e then ~ _1 pO 1 Y U .J cr S V have respectively for Yya ~.nd Co C < "'( eventually stabilize \'t~~ and a respectively). U ?" < a* irnDlies that o<y o since no sequence unbounded in a can oroject T';ote that U Y8 < a 8<y down to one bounded in a*. To establish the a clearly suffices to shov that For and and o<y (of course at values less than Proof: We proceed inductively. Yv le~ma it eventually stabilizes. a limit ordinal this is immediate from our assump- y We tians and the definition of = consider the~efore v + 1 and show that there is a bound on the stages at which v-requirements are created and at the case y which elements of such requirements are enumerated in Let stage for y 8 < \) y = U Y6 o<y onvlard. and look at the co~struction Everythlns connected vi th has stopped actln~ up by starre f; y. we never have to worry about nreserving any for 8 < v after sta~e y. mhus any v - A C. from --requlrements In nart:tcu1ar- 8-requirement reauirement 711 I ' existi~g at stage strayed. or created thereafter is never de- v cO~Dutatiorl The associated with such a require1'\0 r f - 1 E:"1 0 ment is, therefore, correct, i.e. - _1 h r.p-l e JA( = L.l. x ). (x) liTe must sho'l>J that there are not too many of then. Con ~) ide 1" the set is v-inact~ve subset of D at t'J'hich e3.ch and so any of W. y. This is a no v -- A e:: in liV requirement Hhj can be created. ment created after stage r-Tcreover by the W a-finite subset is an After stage T. ch is associa ted "'I'li th any 2T1Y \) A - requtre- gives the correct value of The point is that the computation from changed while the only change in "that a nev·; element of D is never A that can occur is rrJ. • ~~1lS is enuTr:era.ted. , \'lould put the reduction procedure involved into definition contradicting our choice of Finally it is the only way ~ould cl~ar ho"rE:ver, T/T by T. from the construction that a-infinitely nany v - A requirenents be created would be to e~~ntually have ODe associated with a v-active requirsffient for each to occur we could calculate cide 1 v--inactive are contained Poreover T E Of course the stases becomes Since such that -v this hull is bounded,- sa',l.·'y bv.. T D. p v-inc?ct i ve after becomes forever. H in a-·flnlte. € 30 in E: ~l-hull P'.) < a* • of * it remains y in the J. on~e above remarks stage at sone stace after < v of ~f K D = 0) be D x < a. Were this recursively as fellows: at stage T and proceed until a to de- v - A requirer'",()'+in ha \tIith arfSUment been fer' creG~ted assoc;iated vrith an x xsi{. evel:~Y SLot (2ince sU.l2h cJ stage is recurs i ve, th'2rt:2' is one stage by 1:-rhie :'1 it has all stages at ""I]rd.ch the collect1on of ::-,eaulrer'ents are created. \,\ - A which elerents of these ct-fin1te requi 1-'erJents 1s an re0uire~ents Thus the contribution to Sf.~t from are enumerated in also stabillzGa. va 1 u e s bel 0 vI 'l hus i and ct P8 = a* y (J attain and hence y can let Ne T U . y - B res p e c t i v ely. a ;: In vie ,<: of this lenm,::l ordinal such that \..' A requirements y - argument now shows that the contributions from requireme~ts I' Eecinnins c,t :3uch 2 stase the sane eventuall;:'" 2tabilizes. con s t cLn t :i~oreoifer is reGular there is a bound on the stages at C and SlDce A v -, A 0 be the least ~ and be assured that is 8<A does in fact exist for each a limit ordinal and that ~ < A. tion ~e s are now in a nosition to nrove that our construc- Of)V:tOllS such ·'tha.t . r;(J 'v AU B ir:lnediate "J/hile f ..-' B --(J. < 'I'o check if' ri A f\ 13 less before p\s noted SUCCeE'ct80. 1"\ 1 ...-- Cl.. = r' v r\ y: 1'" 1\ 1\ " ask and. == (f; if C is only slip;htly ju,st find a stage {lCf 41. clearly renresents a reduction procedure .8 == l''"\ K == ¢ . ~or from are re A :\s (J This C A and 7' 6 Bare a-r.e. by construction we only have to Drove the follo:'rinr; : Lemma 2.2: Proof: ts not Let that E: < v - A active [ f --1 £ JA = aftex') D, v - A 2. s' v - A cannot be E: = x ~ such which is requireT;lent associated '.vi th at any stage after 1\ H Yv' As inactive at any stage is regular there is a stage (J by which [f-lEJ~ (x) = [f-1s]A(x). course also assume that \ D(x) n J.l, be the least ordinal < v [I[oreaver since a ~ Yv l" By lemma 2.1 there is a least P'J " not the arc:ument of a e cur s i "'I e vor the sake of definiteness Assume not. say Ct -!") r... l·o-1c~.A\,X'" e- _ quirement with r;'~e D0(x) crea:t ,e, a t ar~U8ent x active reduction nrocedure our choice of x. 0 = D(x). S t ,ap'e T • ',Ie Since a, a " v A _, associated with the E. may of ::'e- v - A This of C0urse contradicts 7? 3. Some Corollaries about a-de~rees In order to derive sone interestlnp; about splittin~ a-degrees from the theorem consequenc~s ~e need a simnle lemma. If LemT1~~~: a .-r • e. set s the n Proof: de g it suffic~s SInce and is otvj.ous. is Av ( B) • ir~ To prove the other direction A and ...... observation that ( A) V de g B AC are (and a-recursive in from Be) l\, U B. c;.isjoint this is ilnnediate from the a~e B are disjoint regular a-recursive is that Vie can get • B 3 ) = de g B Pi (I to show that A () B, i. € A and (1\ U That deg (1-\) V der; (B) A :::: U (/\ a-r.e. and regular. "r ,C :3) n (J tl and the fact tha.t 13 0 We can now nrove the usual corollaries of the ~plitting theore~. C<?rolla!'y 3.2: such that degrees Ci ia b. d a is not and b Let c a!:1d a-recursive. such tha.-!'; C = d (x.-r.e~ degrees there are a-r.e. be ~hen a \( b, d J +a a 0 and b and - ...-. - ~...... b, 0 < "'.,) C1.. 0 f t i f the n = D s et ~) a ,3. l1C.i V and the i saor b theore~ recursive is incosparablc a and b Du t b . ~. complete Let a the theore~. ar:~ and b Qe~rees b = 1. a e ..L.. .!:: b. 0 a V a-r.e. c de~ree of the D theore~ be a ~ be reEular set of aegree re~ular be the degrees of the sets given by If both a ~nd b a r e comDarable with recursjve in it and so ,::1 and d. Let the and be the by the 1 e 2m 2. c a-r.e. set and let d. c one is a-r.e. degree then there is an Proof: a v t' = c and ruaranteed assures us 1ncomnarable with then ',)cth B Let are comp&r2.ble " (2. b c C'i be r2gular set of degree and j~ < t; ct. in the theoren. 0 :"'-'::0 .1 ....._ a 0 < C, e:t Let Proof: let < (1 C't a V b --a, < d ~a c --.- a contradic t ion. 0 but d 4• 4.1 be Stren fl. enin~ ~'.-e ~/'!i11 shoH 1101',r In this section strengthenec. to fro!'! or f~ r'i8.1~e Si~ce hyperrep~ular ()lJre usu~lly not i~ a--calculable a-calculable ~rom of OJ.r theoren and invoked to specifi_cally, [s]A(x) ::' our ]:1a1n theorern car. a re set He vV'i 11 dot Ii ish v ad dIn 0' :~ore that a set is suffice to 8ake the vations a-calculability for ~:e ~:Jill 0 o(?rrerular. n a s r e w r e qui r e r1 e n t s the ~~ke sets preSer"18 hyperre~ular cOr:1putatto:n~"3 Dr e s e r- [lOJ. for roughly on the lon~est i~itial se~Qen~ on ~hich r it is total. L€ on this sefment a~d so its I'. l. ,. ran~e essentially a-recursjve wIll be bounded. Thus (i d. will be hyperregular by definition. 4.2 ~e The construction: notin~ besin by that we can place an a pr:i.ori hound on the 1erlgt:-l of an InitIal which we must concern ourselves. that there is no need to nreserve ger than s .~1oreover' cofinality in [e JA on y. add:i.ttc,ns to At the least 1s L (1 we can at;out Trivial manipulations show f\ on any [E]n if there is a last y se~~P1ent ~et by seg~ent 10n- .'X-cardinal vThose with preserving Keeninq this in mind we make the following OtlP construction: we find) (J for each and such that there is no x < s(y) ..:'\. ctS oci t E" each v' - A requireIf there is such an x and 1s converrr,ent requirement with argument x crea tea 1:Jf~ associated with s. v' - Ii ~he qUirement consists of the nega.t1ve facts about bound the creation of by Of course a v' - A A. ments into the priority A a~d not confl~ct (Since" listin~ v - A by and c(o) .' H into The priority arguDent; that the givin~ v - that e: < and any Pv ' of [r-1s]A.(x) of s(y) troyed. v' E y < 'J. ·-require~-nents or B Our main stabilize. < a*. Let v X .... are made into < <:,('''\1') ...... 1 We now determine as before . ~oal is again to show Once that has been done A as follows: be the least ordinal such Once we are beyond stage for requirements do A we can argue for the hyperregularlty of ~onsider s for Finallv we insert these require- this presents no Droblems.) whether to put 4.3 r -~equi renent the saDe priority as the annropriate B) requlrer'1ent s . y requirement is destroyed if any cf its elements is put into (for to y'-requirements and the enuneration of element s contai red In C used \'!e adJust the defini tton of P~ in this conputation. re- il' ~I v any- cOTTloutations covering an initial segment require~ents which are never des- is defined at all we even... :':orcover tually get such a corroutation (A 1s still rs-gl:lar). Thus we can compute on which it is defined by just going through the construction 2ince t is is an ~-recursive proce:Jure :ts bounded on this ser:rr:ent and hyperre8'u ing 19m~a ~e 2.1 ~herefore this i~ LeTTlrna J.l.li: contex~. stabilized at values Proof: =v y + 1 conclude our proof by e2tablish- If bv some staF(e then D '- and,' o" 0.1 Y t efore, ,f\.s ~'la ""no", pO' ... requi:re;;lent s are bounded. sta~es creat j ,. Y respectivel~,'1 0 - and eventually stabjlize. need only considel... \'7e h3. 7 e only to T'Ie at which case •..1 .• , ="...,.) (;of ShO\-'1 v- that there v'-requirerents are takes ("',L' CO!'HjOnent ~T have and and we know that the contributions for is a bound on the 1s of the otier ca~e If ther'e is no last t)/.-cardinc.l or trh; .I last bound.. a-ca~dinal is sinGular we can subset of 01') D ;. \) X V, {X * r)"ll as sumpt ton. and so enunerated in houn~ej ~lhl1S ~.- fini the set is v j.:3 r e ~; u J.::t r 'J c;; fir s teo n - x iT 'v and so t E:~.J a-finite. f) £ . Si~ilarlv te tine. ~:~et -.7ith e. each o~ which is strictly , If tl'le last c~-~r. (the associated reouireMent, the argument of the requiresen~) s idep the directly for this \), -requIrements created corresnoncl to 2.n rrhe les s than ar~ue \. all these ') is < y ":hich a,-r.2. recuire~ents aSSOCl- for -'lhi ch ar~uments t he ';Ie cre~~t e could arsue as above, since \' · -rc'Qu ire:Y'":ent s, ',Ie y < a*. Or the other hand if there t'lere no such bOLtnd, i're could stve ma:9 y -+ t\rith the flrst P v x by sendinr:r; \) f -requirement over p,. \) < vI of to th2 ~:.Jj_th Since this would exhibit ~. y (~.efj_ne?in argument as an y in T ..1./ a deed conclude as in the nrevious cases. Of course there is n()~I; created after x a-finite union Thus • y it i'!ould 'He ean in·- 0 no difficul t;r in the corollaries of section 3 for \~Jell. ?:3soclated a-finite sets bounded below contradict the regularity of as E (x-recur·- crt v:in9: 8.1.1 ct-calc.ula.biJit:'l cier:y·ees recursion theory surely the most; powerful (and difficult) is the i.nfinite injury r:lethod developed by Sacks [6J. PIS a prime example of this method we cite Sack's theoree that the r e, degrees C~! ;::Jr' E. "-"" 0'-, Y'J ...... c:; [1 C\ Il.-,.............J.... 'I· J""'. • dee d:. bee n see n a s a t 0 11 ~ h s ton;2 f method in C'(. -, be s t res -l J t t the t er r e cur s i 0 the 0 r y [ 1 0, 0 (l cis. t e has [) e e 11 ~) r i SC0 a-r.e. de[rees are dense for \; e 2 i<: etc 1:1 apr 00 f adrnissitle u. 0 f detailed ~heorer t :~ e ~t n fin i t e . Lt J, [ 16, ~~. a* = w. uy' y ')l L J • The In this chap- [1 eo rem Eopefully tilL::; hezinninEs ~-recur3ion has, in- 11 ? s t [} e 0 r e ~J [1 5 J t 11 a t the d t~ n~) i t Y t some insight into the gener argum~"?nt3 0 Y' ~his ~'lill for eve".C' y also give L1etllcd of infinite inj \J.ry theory. iscussion will aDD2ar A full elsew~ere ~roof [13J. 111th more ." 1. Gene.t>al ;-:) an Let two re2ular a-r.e. sets s < 0. C be ~iven. ~\Ie sueh that < --"Ci. A ( 0) =:. {~A Seconaly.} II ((' ~) .yo....... 1 \./ 'de C. To ~nsure that A has 1 the to nake su.re thclt c· (;.. ne:~~ative rei1u.irer.-:ent~3 rC3cF'blin~"S \;r5.11 have cuara~tee those of chapter III to c -La { t. that Pine.Ily we r· ,.' into certain areas of ~ lines of chanter III. The idea is that, should out to be of C < --a a-recursive in into A < -a ns long as it B a~ in. sir:lple a~pears that turn B, we VJould tben have put all ~·Ia~~. This would imply that a contradiction and so As in chapter III, the main overall problem ',..., l:::> +-vO arrange tl:e requirements in a sufficiently short sequence. Because the complexity of the construction we are forced to L 2 ( E) - c: f ( ;~ ):> as 0 ur bas i cor d e r in g 0 r ~) rio r i tie s . this is a natural choice it is difficult to implement it 1s gi v,en tJy a L3 ~' h 011 SL sin~e fune t jon \'Jhi Ie our cons t:ruc t ion nus t dure for the priority T+.l..\., converges only in the sense that the correct value least one returned to unboundedly often. • lS +-1-- vue This Dhenomena is typical of the irtfinite injury ffi?thod in ordinary recurs10n theory and it waB only to b2 exrected that it be manifested in the priority listing i tsel f :Ln a-~recu.rs iOIl theory. Of course, we also have the usual prohlec of infinite injury arguments. We must not allow positive requirements to be thwarted by an unbounded succession of temporary negative requirements. This probleD 1s solved roughly as in [12J by not allowins elements associated with positive requirements into negative requirements. Instead we put in a collect ien of other elernent S \'Jhic:n are also being kept out by related negative this is vague.) require~ents. (Admittedly The machj_nations necessary to carr~:l this out THill cause a split -in the proof (tl10UZh not in the construction) accordin[ to dtlether or not the w equals or not. The idea is that we will have to be able to close under the operations involved in form1ng negative in w require~ents. many steps of a all will be vJel.L. If, on This closure is done naturally E 2 tl-;.e process. ~hus other hand, if j~ ~ c::::. - E2 - cr(G) > w cf (C~) =: we will Drove that only finitely many steps are needed (.,), essentially be / T 'J-! relative to D, a ~.) .. ::r~ recursions. by restrictjr.g We handle ourselves to recursive cof'inality - (B) several parts of the construction. cial well. its 2ble to iter2te s not be as ~crking le~=a 2.1 which allows us to many steps for Eve~ this urastic ~ark belo~ .,, p ~~, p kX that {v II =:. u of r ('{ 1J.~ ,~ , '.f) for every Cf. a preDel" i.Iii tJ ,;:;1 lS .. x < e lJ but = a. x<:8 b) = Y the 1e2.3 t 8 ;::uch ths. t thE:reI.8 a Dar·'t ial function, k(x,Y), a-recursive in B such that k is .i~ a proper initial segment of U k x<0 x < y but = S. x -f[·r g,LYJlJ every,S < y, f[ g[ 0] ] 2.4: for each is bounded in t.'3 for aporoximating approxir.--i3. ti n§~ that gives us a-recursive in = that [efJB f(Jex) = [efJ~ B~ at stage IJ B = ep -"- > h,} Ide jegin by a-recursive enumeration As a. there are f, [e and for p. and f, L!" as usual via an 3(J p is unbounded in e f-l, and f, ,,-1 r- l and [ehJ and e h c - h. hare such We can ftherefore approximate them q.ui te easily Ly set tin(~ p(J for f- 1 and (x) h. for ever:r x and acting similarly (Note that "He al-r:;ays tal:::e tl'.e least (J [eJ~ (x) available computation giving a value for ide approxlnate determine a value 4) lemma 2.2 ( a) • As and He set c~ cr tS divergent. We add a later stage rnation about (x') -\.1 (J T Ie is least g via the k to of a-recG.rsive in y such that the proviso, however, that if at "He dlscover tnat some incorrect infor- reverts to the least abou.t B for y ~'Jh:l.ch "vas used in the cOrlnutat ion for The approximation to f and rect on each initial se7ment of" a As ror are eventually cor- h B and respectively. f-l, the approximation is·eventually correct O~ for each can say 1.s that any to be incorrect. y tneorrect inforuat5.on such that <5 < rr. propo~:;ed "Finally for gO(x) = y For x don f -1- ,t. ~ value is eventually seen ,. \ x < u, g,Xj is the for unboundedly Many o. 3. ~Che Eequ We will try to give an intuitive picture of the ~etails requirements and leave the formal this we supress t}le facl1itat~ ..... and Nrite simply f .L, -1 approxi~ation and "g T,',1. e all other 3.1 <,~. 0 : x) . +ln~o ~ A. requirerr~ents anu D 1.1I.;_ v ,-1" _nlS AG(x) find the least Ath(y) froD 1 n I I I I . 2 .) priority square as a S y < is enumerated " caKes :or p~sceuence ~n over e2.C;1 x < \' l!Jf2 (~on- 2.Y'ra.~T. such that W2 canrot cODDute B via some p-active reduction procedure i;J e the n i:n i t i G:. t e po sit i v (.; r e qUi r e:-:'c en t S "7 i t 11 r \j that incorrect information about Arh(z) I C h ( y) - the least fran be;innjn~ r 1} ~. (v) \ '- into the o th Of course, if at any later stage we discover of some nrocedures -(i, that try to copy x x To ;::) < 'l\i.e pos5 tive requtremr::nts: sider Be~ore l1 • in earnest we insist that when any B to [18J. A C ,7 was used to compute we then give up trying to copy lh (z ) v B <: 1 nt B Vla SODS 0 c~Js (x ) ) er • such t t 'J e c 2. n not co::::.:' l~ t e n-active roduction prcc~dure < fg(x). For esch us :in the c r~a:(e these 1nto about 111:::e to 'poulll allo~.; s tee:. d a~.: nr~;;ativc eJerlent s to be soc i ate priority ,;e x i,~c; t requ1r?:;'.snt::3 but th2 t O~). t of b~? A. ~, J r[lhe a,x,z ma:Ln l."'e s t J.>j_,~ that no element with an associated positive ments lnto tive ~) .:.\ 0,X)Z 'o""x,z is later put into R 0,X,Z Crh(z) B used to compute destroy for ~he construction: ority x priority ~hich SI1G[} e1e-- w p La , x , z • 'l',":-- ~J aj~~~z If any ele'nent or any information about is found to be incorrect, we > z. We SiMply act as dictated by the requirements described above. element for than puttinr; any of high priority whicll are currently an operation to get su~h 3.3 i~::tther rcq~ir2Gent Indeed, ue basically close keeping then out. of . ti.or: is we put in all uther elements of nega- ~1 req~irements under a su cc e2· S lon n e ~ ,:-1. t i v e r f~ qui reI:; e n t 2, with each gets put into ':JOU 1 e). At each stage we put in any we have a positive requirement wit~ pri- unless it belongs to a negative requirement with < x. 4. ~tline The Proof -- An rrhe priorIty Argument 3nc3 Droof t}l.-lt OUy' constructi.O"il. s u c e e e d s a. r e qui t e l 0 n g and we have j e =' e n ~j. so 1:1 e.-I h ,q ton de t a t1 s rrhus 11'1 this section ~':e merely gi.ve neglected. the sequence of ler1rnas \Firi '\fhlcrl the proof proceed2 ':Ii th only a general idea as to hO\\f one Lemma 4.1: begins to copy C a.-recursively 5_n stoDS tryin~ One caD tell into a prove~3 the lel11ma:3. a-recurs:'_";lelv :tf one eve:" glven part of' =·1oreover, 1\. Bone ce;.n tell whether one ever to copy C into this Dart of ,1\ J..j. • late~:-' Finallv " .' if one never does give this up the assoc:i.ated computation relative to is correct. B Lemna 4.2: 'rhis lemma is irnmedinte. If a negative requirement R da,X,Z permanent the associated computation relative to is corIOlect. The proof depends on the well-foundedness (derived from the priority ordering) of the procedure used to forn from Na,x,£:' n" Now for the heart of the procf one proceeds by a 51mul tar:.eOllS induction on Lemrri2. 4. 1) x < y We can tell to establish the follo'>tJins: a-recursively in B if a given negative requirement of priority < X Xl lS per- manent. 2) a-fini. tely many :oe!·I"~a.n.ent nega.- There are only x' < x. tive requriements of priority 3) x' < x A permanent positive requirement gets put into negative requirement of A unless it belongs to a permanent ~riority T.here are only 4) C1. x" < x'. --·fini ~ely many eler1snts at VJhich we permanently try to copy x' < x. of nriority C in A with priority a-fini tely many elements I"Ioreover, there are onl y for which we actually create permanent positive requirements with priority x' < x. At successor (~ staces, the fact that we can theA \.1 ) (from ~ \ \j) and (LI.) plus 0-recursively check if a nega- tive requirement is destroyed by a quirement. t en positive re- tempo~ary Of course, in every part we can tell iMmediately if any incorrect information about B) was used. b The proofs of (2) and (4) are similar to 111.2.1. failed we could sho"J that C < -a would give first that If (2) B., w11ile a failure of ~ind c then B. < -a (J.~) In either case v,,:e contradict our t)asic as sumpticr: that B < a. ('I V • e 1 e r:l en t p l" "(1 a ..:.... ."..to. ~r i J l"l _~. (3 II . J+"'0 1 1-'---. 0"· 0~ . J .f'" ._ Y '../.. ; 0:-'" :. _.. t ~> ~ ..L ... '_ t h a p 0 sit i v ere qui r e ;~J: n t tive requirement thereafter. to which it belon~s r> _ 0 r'.l--t......... (-1 l t-V .J, 4 n, _..... '_ .i.Y"1 +v 'r '1.....-\.. ~ tV Y' ,.... J. 1 V i () r: uti n to aLe.::: a- Thus all the ne~ative are eventually destroyed. ones At linit levels we note that the whole procedure was a --r e cur s i ve 1 n ments of LJ, ~_J. X C C were also ne C r) t t L1a tee r t eJ. in i t i a 13 e [>- 25. 11 t~e It turns out th2t ar:lount of funct1on. the definition of proper initial seccent of set is a-fini t2 and one r',-I. C(:.Hl c Since no',- ,::;0 th:"'ou to this sta.se uniformly to get the requ:ireJ bound:::;,. ide are nOH in pOf:jition to 2. tLat the set shO~J ~ .c\ has all the desired .properties. Lemma 4. 1+: I, of course, well founded and t~le sure the tree generated ty any a. under are If only problem is to make a·-fini t~; set In general the oper&tions l: 2 E2 - efCa) and so =w Sar; icaJ.ly (ine must, however, sco.:ccl'l the idea is to use below 1', ~~ C. is regu:tar and then i.t out lS touLdcd th~t one nust close > a 11 is 0J tha~ \I e 11 . finitely D2ny steps suffice to close off, since in thjs case all priorities are intezers and each sten in priority. 4.5: ----"-.-'- Lenr:1a i~surcs a strict decrease I, Lenma ,.. <+.0: In both of t~le3e ler;;FIas the id.ea is like that of fl'he main ne,;,-] ohstElcle for 4.5 is to 81101'[ that 111.2.2. for certai.n corre(~t CO::1Dutatj_ons tually gets an assoclatec. ne permanent. cedure for of again faced with a ~ust analyze the pro- negative requirements. We are cperation problem and again closu~e the proof splits ac.eordin.g to Hhether or not. one even- p" ive requirement Hhich is In addition to 4.3 one constructin~ from ::'"\c - ef(Ct) = w Since it is easy to see that the relevant correct computations of ,~ H from B eventually produce permanent positive requirements, there are no extra difficulties in the proof of 4.6. As we insured that B < the required properties for -a. A 3, we now have all in B < ex A < ,.... c(. v • Bibliography 1. R. B. Jensen, The r; n uni:formizatlon lemma, unpublished mimeographed notes. 2. G. Kreisel and G. E. Sacks, Metarecursive sets, Jour. Symb. Log. 30 (1965), 318-338. 3. M. Lerman, On suborderings ·of the a-recursively en~~erable a-degrees, Ann. Math. Log., to appear. 4. J. MacIntyre, Contributions to metarecursion theory, Ph~D. Thesis, M.l.T. 1968. 5. , Minimal a-recursion theo:retic degrees, ,Jour" Symb. Log., to appear. 6. G. E. Sacks, Degrees of Unsolvability, Ann. Math. Studies No. 55 (Princeton, 1963). 7. , Metarecursion theory, in: Sets, Models and Recursion Theory, ed. J. N. Crossley (North-Holland, 1967). 8" , Higher Recursion Theory. (Springer-Verlag, to -~--- appear) .. ------, Trans. 124 (1966), 1-23. and S .. G. Simpson, The a-finite 10" ..l \..rm... 11. ~~~S Posts problem, admissible ordinals and regularity, injurJ~ method, !,:ath" Log., tc appear. J. R .. Shoenfield, A theorem on minimal degrees, Jour o S:rmb. Log., 31 (1966), 539-544~ 12~ -------_.-..,,--"'-;, Degrees 0:' Unsolva,bility (North-Holland 1971) ~ ~! ;" 13. S. G. Simpson, A~m1issible ordinals and recuTsion theory, Thesis, M.I.T., 1971. 14. C. Spector, On degrees of recursive unso1vability, Arm. Math. 64, (1956), 581-592. Additions for Chapter IV 15. G. Driscoll, Meta-r.e. sets and their metadegrees, Jour. Symb. Log. 33 ,(1968), 389-411. 16. M. Lerman and G.. E. Sacks, Some minima1 pairs of a-recursively enmnerable degrees, Annals of Ma"::h. Log., to appear. 17. G. E. Sacks, The recursively enlrrnerable degrees are dense, Annals of Math." 80 (1964), 300--314. 18. R. A. Shore, The a-recu!sive1y enumerable a-degrees are dense, to appear. ""'.....•'. ',i'>·. j Richa~('cl A. Shore on August 18, 1946. .,. '1, .lJc til'l SChO()'" '1 fl •• ..L.... ~'ras He was i"(')0 U LL.";;', lr',c.c., >':) the fol:l.o :!1ng ~)eptember. 1965 fY'OI:1 l le8 ve OE ..... ".~ J..ll T'ln (.,. ~ .... 8, ¢BK. ., ,.-.. .J... ') ~Ilhich r< i...c)) ,J On' • e~ of Technolo;y. f~~,JD' and entered Harvard there cons tit uted his ~) tl.lci;.;· he recei 3. 3eDte~ber ~athe2atics teachirL~~ he be~an fo~r years of at the Massachusetts 7nstit~te as.sist2.ntsllip 2.nd a staff ?ar't~ti.on ::~el,~tjon~~j, ~ wa~ scholc~r-·· supported by rJ~r8.tneeshlp 21 ct'2c1 to f:'111 r:v::rnr)ershi.p 1.::: ,,-. J~..tn1cT> cum 18.U.d .•-:: 1n anci Square Eracket ?a:etitioD t. 1961~.- B.J.Ed. In June" 1966. rec ~2 i ved an National Science }i1 ou21rlati:)n '- Unive~3ity He spent the academic yea.r In his third and fourth years he Cardlnals3.nd the~~oston Dur beJ.el a fulJ.···time ship. hi s The following graduate work in J uatecl ~"la~3sacrlUf;etts Har,rard at the Hebre·~.f Univers i t~! in I..T ex,usa 1 er:>.} Israel. setts, from =Ln Eoston, bOl'"'r1 '-" and Ii'.Te?k CC;~T)actn2sS _elat5()ns p z~X and is 2. r:1cnber 0J.-' the a