Beruflich Dokumente
Kultur Dokumente
. 16, No. 1 (Mar., 2010), pp. 37-84 Published by: Association for Symbolic Logic Stable URL: http://www.jstor.org/stable/25614509 . Accessed: 13/06/2011 17:54
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=asl. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Association for Symbolic Logic is collaborating with JSTOR to digitize, preserve and extend access to The Bulletin of Symbolic Logic.
http://www.jstor.org
The Bulletin
jr: V -+ V
PAUL CORAZZA
suggest a new approach for addressing the problem of establishing an ax for large cardinals. An axiom asserting the existence of a large cardinal can naturally be viewed as a strong Axiom of Infinity. However, it has not been clear on the basis of our knowledge of co itself, or of generally agreed upon intuitions about the true nature Abstract. We iomatic foundation universe, what the right strengthening of theAxiom of Infinity is?which large cardinals ought to be derivable? Itwas shown in the 1960s by Lawvere that the existence of an infinite set is equivalent to the existence of a certain kind of structure-preserving trans of themathematical formation from V to itself, not isomorphic to the identity.We use Lawvere's transformation, rather than w, as a starting point for a reasonably natural sequence of strengthenings and refinements, leading to a proposed in later work by Trnkova-Blass, strong Axiom of Infinity. A first refinement was discussed showing that if the preservation properties of Lawvere's tranformation are strengthened to the point of requiring it to be an exact functor, such a transformation is provably equivalent to the existence of a measurable cardinal. We propose to push the preservation properties as far as possible, short of inconsistency. The resulting V ? V is strong enough to account for virtually all large cardinals, but is at the same time a natural generalization of an assertion about transformations V ? V known transformation to be equivalent to theAxiom of Infinity.
is the right notion of "infinite" ?0. Introduction. What Prior to the work of Cantor, the only acceptable notion was not be collected
set. Both because of together to form a single, well-defined issues and seeming mathematical actual infinities philosophical paradoxes, were not considered legitimate objects of mathematical investigation. Can tor convinced the mathematical world that mathematics without an actual
infiniteis too impoverished. (See [15], [27], [28] for a discussion of these
points.)
Part of Cantor's legacy is the presence of theAxiom of Infinityin the standard set theory ZFC. The Axiom of Infinity asserts thatan infinite set
Received April 24, 2008. 2010Mathematics Subject Classification. Primary 03E55; Secondary 03E40. WA, Wholeness Axiom, large cardinal, exact Key words and phrases. Axiom of Infinity,
functor,
critical
The results of thispaper were presented at the 2008 Annual Meeting of the Association forSymbolic Logic, March 27-30, Irvine,California
? 2010, Association for Symbolic Logic 1079-8986/10/1601 -0002/$5.80
point,
Lawvere,
universal
element.
37
38
exists, and this axiom, in conjunction with the other ZFC to build Cantor's entire theory of transfinite cardinals. de facto
Ironically,even as ZFC
universal
foundation
Godel
being discovered: Hausdorff discoveredweakly inaccessible cardinals [18] Mahlo cardinals [29] in 1911. As in 1908 and Mahlo discovered (weakly)
would that is underivable It was later show, large cardinals from ZFC. are one type of mathematical entity new, clear even inHausdorfFs Cantor
notions of infinity, stronger giving rise to naturalmodels of set theory (see can truly asserted to exist in the but perhaps a strongertypeof infinity be
strong should the Axiom of Infinity be? the time of Godel's Second Incompleteness Theorem, which demon strated among other things that large cardinals are not derivable from ZFC, Since [20, Introduction]). gave us a "minimal" notion of actual infinite,
represented
universe. How
Some have been unconcerned with existence of large cardinals. have actively attempted to prove that some or all large cardinals are in fact inconsistent.2 And a third group attempted to devise heuristics, often vision of the structure of the universe, to legitimize the based on Cantor's presence of large cardinals in the universe. We consider two examples of such these heuristics were introduced by Godel and developed further by others
(see[21]):
One transition
of Cantor's
the universe is fairlyhomogenous ([15, Section 1.3], [20, Introduction]). would be too "accidental" for co to be the only From thispoint of view, it the largecardinal propertiesof (forexample) inaccessibility cardinal having
and measurability, surable cardinals and so we conclude that there are inaccessible is an example and mea in the universe. This of the generalization heuristic. ematical
universe is that the about the mathematical perspectives smooth and consequently, from finite to transfinite is reasonably
nals should be able to be captured by a singlefirst-order property [15,Sec claimed tion 1.3], [20, Introduction]. Therefore, if it can be legitimately
that V or ON
Reinhardt This isan intuitive the reflection principle',in [31], satisfy property. to formalize this typeof reflection. It can be argued on intuitive attempts
^ee [12] fora surveyof a varietyof contemporary views on these issues. 2 was Silver's early efforts prove to thatbore great fruits A striking example of such an effort the inconsistencyofmeasurable cardinals; his effortsresulted in many of the foundational
results on 0#.
some actual
also
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j :V
39
and perhaps much more, and so by grounds that ON is inaccessible, Mahlo, at least inaccessibles and Mahlos should exist in the the reflection principle, universe. See the work by Paul Bernays [1] where arguments of this kind are motivated For themost and developed much further. part, while these early heuristics may have been useful complicated large cardinals that are widely used cardinals. superstrong and supercompact in jus
wider
Steel, Martin, Kunen, and others, between some of the very strongest largest cardinals and sets of small cardinality, particularly sets of reals. One of the pioneering theorems in this domain, due toMartin [30], established determi sets of reals are Lebesgue extended measurable and have the property of Baire), as
found for these stronger types of infinities, there is in the universe. Though than ever of large cardinals this acceptance be due partly to long-time familiarity with the concepts, it development may ismost likely due to the deep connections that were discovered byWoodin,
heuristics
even the ever, [27], [28] for an excellent surveyof heuristics for justifying the fact thatcompletely satisfactory strongest axioms.) Yet, despite simple
have not been
which itfollows,forexample, thatall such nacy of all 11^ setsof reals (from
Woodin
In a different direction, Kunen [23] showed how to collapse a huge cardi nal to obtain in a forcing extension an (^-saturated ideal on co\. All these results were shown later to require much weaker large cardinal hypotheses, but the rich structural connections between longing to the earlier stages of the universe these strong axioms and sets be strengthened confidence in even
j: L(F^+i)
L(^+i),
that, assuming
an In embedding
of these results.) Reflecting similar sentiments, in his 2008 Godel Lecture,3 H. Woodin expressed his belief that the "hierarchy of large cardinal axioms has emerged as an intrinsic part of set theory." of large cardinals, more clear than ever is the Despite growing acceptance fact that there is no axiomatic foundation for their study. ZFC has proven itself to be a nearly universal foundation for the rest ofmathematics; it would be only natural to expect that an extension of ZFC should emerge that would provide a foundation for all ofmathematics, including large cardinals. Large cardinals have been studied long enough and with enough rigor to warrant a
of theorysimilar in spirittoperhaps the theory groups or topological spaces. the studyof large cardinals continues on without such a formal Although
one simply cannot ask without such a theory. theory, there are questions To make this point clearer, we consider here an example of such a question Presented
California.
40 and discuss answer it: Question. sequence? Which Test Laver ?> f\ k later in the paper
to address the Test Question, certain points need to be clarified. one needs to have a notion of Laver sequence that could apply to other First, types of large cardinals. Laver sequences for strong and extendible cardinals In order
have been definedand proven to exist [3], [4], [14]. But thequestion cannot
in general without knowing which large cardinals exist really be answered in the first place, and then, beyond mere existence, some form for candidate large cardinals that would permit a general definition of "Laver sequence." issues. We used as our foundational axiom the axiom (schema) discussed
general definition of a broad class of large cardinals that would be candidates the background Wholeness for admitting some form of Laver sequence; ensured the existence of the classes defined in this way. In Part 1we Axiom will
review some of the results and point out some of the added value that resulted from having a strong axiomatic framework inwhich to work. to account for large cardinals, In devising an appropriate extension of ZFC it is of course necessary to decide which large cardinals should be declared to "exist." From
the practical need of providing belief system, it could be viewed as meeting a useful robust framework for studying the things that are already being studied.4 Even ifwe agree to admit all of the better known large cardinals,
recent the trends we see in set theory research and Woodin's it is not perhaps too bold to seek an axiomatic foundation for all the remarks, have been studied over known to be inconsistent?that large cardinals?not to a an axiom system as a commitment the years. Rather than viewing such
but in some way be "natural" enough to belong the intended consequences, In this paper, we axioms for all of mathematics. the foundational among a case for its naturalness. will offer one such axiom and make
it is stillnecessary to decide which sorts of axiom candidates ought to be taken seriously;one would expect thata good axiom would provide not only
4An interesting parallel to thispoint of view is the fact that some set theoristsare not personally convinced of the consistency even of ZFC itself;some for example have argued that Z?-Replacement for n > 2 is unconvincing. (Solovay discusses this point of view use ZFC briefly inhis FOM post, August 17, 2007.) Nevertheless, these set theoristsfreely concern over thedegree of Replacement involved in theirarguments. A without meticulous foundation forall largecardinals could be viewed in a similar fashion.
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j \V ?+ V
41
to the task of building an extension of ZFC Our approach that accounts for large cardinals will be to return to simple heuristics, but we will do so by of Infinity. examining an equivalent and rather different form of the Axiom In our view, simpler heuristics are more convincing, if they truly can account for all large cardinals; so far, as we mentioned earlier, simpler heuristics have
fallen short,by quite a lot,of thisobjective. We will, in thispaper, examine can how a global version of the Axiom of Infinity be strengthened natural in
ways to lead to a candidate is divided
strong Axiom
outlines
form of
final version
hint about their many proofs, and for relevantdefinitionsthe readerwill in cases be redirectedtoPart 2. Part 2 beginswith definitions and background
theorems of the proofs of the theorems mentioned in Part 1. Most of these proofs can be found in the literature, and whenever certain details are omitted, together the essential paper. arguments from this diverse literature in one place
details theory; itwill then provide what we feel are themost interesting the references provided can be used to fill in gaps. We feel that collecting in a uniformcontextadds sufficient value to justifythe extra lengthof the Axiom of Infinity. ?1. In searchof theright
The Lawvere functor. We and
that may
be unfamiliar
to the reader,
particularly
in category
the result, we define possibly of the result for our program 1.1. Lawvere's Then the following Theorem
version
of the Axiom
(2) There
are equivalent:
thatfactors
as a composition
oF
of
(A) F H G (F is left adjoint toG), (B) F: V ^ V?, ? ?> V is the = A. (C) G: forgetful functor,definedbyG(A A)
all colimits and G preserves V-^ V all limits.
In particular, F preserves
42
PAULCORAZZA
In its original form, Lawvere's Theorem talked about cartesian closed rather than models of ZFC categories, Infinity. A cartesian closed category and, for any pair of objects A,B, has an exponential object AB, much
is one thathas all finite products (forsets, theseare just cartesian products)
as we
of his original.
of the
viewed
theoremcan be eliminated in theusual ways. In theproof of (1) (2), F and G are given explicitlyas classmaps. For (2) => (1), the resultcan be
as a schema of theorems, one for each F. In either direction, j
simplythecomposition G oF. We observe that in the theorem,the factors of j are endowedwith strong
properties. For instance F preserves all colimits; examples of
is
preservation
colimits in a category of sets are coproducts (which are disjoint unions) and direct limitsof directed systems.Dually, G preserves all limits; limits
include products (in the usual set-theoretic sense) and inverse limits. These are very strong preservation properties in the context of category theory, and
V, the arrows from F(a) to b are in 1-1 everyobject a of C and everyb in a to G(b), and thebijection is natural in correspondencewith those from a and b. A familiarexample of such an adjunction is foundwhen C is the
category of sets and V vector set V without
Then, if F(X)
is the category
of vector
spaces
is the
is the
Detailed definitionsare provided in Part 2. We will henceforthcall the functorj : V ?? V defined inpart (2) of the theoremtheLawverefunctor. in Axiom of Infinity Lawvere's The The proof of existenceof j fromthe F to thealready defined forgetful orem requiresone to definea left adjoint G. functor A definitionofF on objects is givenby
F{A) where succ is the successor = A x co-> A x co,
H G and G
Theorem.
can Axiom of Infinity) be used of definition recursion (derivablefromthe by H G. F to establish that Conversely, to establish theAxiom of Infinityfrom the existence of j, one can use the familiar trickof examining the criticalpoint of j. Arguing
forward. As we indicate
function on co. Defining F on arrows is straight version more detail in Part 2, a category-theoretic in
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j :V
43
category-theoretically, Lawvere's
one
shows
set. criticalpoint of the gives us an infinite mapping, and its image explicitly
result is striking because it implies that to understand theAxiom In to terms with tranformational dynamics of the universe as a whole. to believing: fact, it tells us something that we are not perhaps accustomed that the universe comes equipped with an auxiliary map j: V ?> V that has characteristics.5 strong structure-preserving to much stronger no Lawvere's Theorem opens the door to generalizing of Infinity does not. Recall tions of infinity, in a way that the usual Axiom come
that y (0) =
0 and
co, so 1 is the
that we are searching for the "right" version of the Axiom of Infinity.When limited to the usual version, the possible generalizations do not go very far. we ask, "What is the right version of Lawvere's Axiom of In When instead The task then becomes one of formulating finity?" we see new possibilities. the preservation properties of j as clearly as possible and then strengthening the strong preservation In Lawvere's Theorem, them as much as possible.
but to its factorsF and G. So our first propertiesbelong not to j itself, attemptat gettingthe "right"axiom is to ask thefollowing:
1.2. Generalization than its factors, with Step 1. What the properties ifwe endow j itself, rather happens of preserving limits and colimits?
we considered cardinals
Before attempting an answer, we should address the following philosophi cal issue: Why should stronger versions of Lawvere's Theorem be true?When
themotivation for concluding that certain large generalization, exist on the basis of generalizing properties of co was Cantor's in the uniformity of the universe?no cardinal should sight concerning single have many special properties not found elsewhere in the universe. Here, a that "more preservation is better" is principle thatmotivates us to conclude Cantor's principle of maximum possibility: As much as possible exists. (See As we will see, when Lawvere's functor is endowed with [15, pp. 20-23].) ever stronger preservation properties, the consequence is increasing combi natorial richness in the universe. This principle suggests to us that moving in the direction of strengthened preservation will lead to a more complete rendering of the universe. A theme that will develop as we move from Lawvere's result to refinements is a strengthening of what we call here critical point dynamics. Histori occurs in the context of the canonical elementary embedding = M, where M is the transitive collapse of the ultrapower
5 This point will be developed furtherin thepaper. What we have in mind here is that the adjointfactors of j have strong structure-preserving properties (as inLawvere's Theorem), and that this situation "points to" the possibility of a strongeraxiomatic formulation in which j itself has theseor similar strongpreservation properties.
44
PAULCORAZZA
/^-complete ultrafilter over an uncountable VK/U and U is a nonprincipal cardinal n. The first point to observe is that the large cardinal strength embedding?namely, k?which must be a measurable cardinal. The second
of such an embedding is located preciselyat the first ordinalmoved by the arises fromdynamics that "live" in thevicinity
there is a function / with domain k on sets in the vicinity of n.
M. InHamkins' terminology[16],k is a seed via j forU thatgenerates In studying and both theset-theoretic category criticalpointdynamicsfrom theoreticpoints of view, itwill be natural to define a criticalpoint of a which y< \j(y)\. (Themappings mapping j to be the least cardinal y for
j: V ?> V we
being defined
In particular,
for every y e M,
consider
here will
never have
the property
What
of which anyZFC universe is equipped, to be an intimation a typeof trans would havemore fully formationthat developed properties in thepresence of of largecardinals; thisfully the spectrum developed version could be seen as the evolutionarypinnacle of j. We have already suggested thatj embodies
dynamics of K, since this is what functions do. Since an of the transformational dynamics of V is the unfoldment of aspect that part of what j is "designed to do" is to generate all sets, we hypothesize that a seed of an elementary all sets, in the way (or by analogous means) ?> M M. We formulate this hypothesis generates all sets in embedding j: V below and then, as a secondary theme as we seek strengthenings of Lawvere's transformational obvious we track the extent to which Point Dynamics this hypothesis is verified in stronger
contexts.
purpose does it serve?" If we think of set theory with large cardinals as being full set theory, then one could consider this Lawvere functor, with
imprecise question " criticalpoint dynamics is, Why does V come equipped with a j: V ?> VI
Theorem,
1.3. Critical
Hypothesis.
in thevicinityof k.
functor and for each of the strengthenings we will consider. or, in the case of the Lawvere
We will take a look at the critical point dynamics both for Lawvere's
In each case, we functor, simply a countably
the seed property of themapping. at least roughly defined what we mean critical point dynamics, we Having in the case of the Lawvere functor. As we can now examine these dynamics this latter characteristic
infinite set) does so at itscriticalpoint, and that,on thebasis of dynamics We will call in thevicinityof thiscriticalpoint, "all" sets can be generated.
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j :V
45
with?in
we tellsus the "type" of infinity are dealing isa universalelementforG (butnot forj)\
B ?> B and any y G G(a), we can find seed property in this case arises from
a unique /: F(l) ?> a so thatj ? G(f)(0). Moreover, every set in V is expressibleas G(/)(0) for some F?-arrow /. That is,
V = {G(f)(0)\feAxr(V?)}.
We notice that the Critical Point Dynamics Hypothesis is only partially for verified theLawvere functorsince the seed propertydoes not belong to j We also note that the "type" of infinity itself but ratherto one of itsfactors. isgiven to us in thiscase as the imageof thecriticalpoint ratherthanby the
critical point 1. We itself. functor. We now attempt an answer to Generalization that, from a category-theoretic every set is Step perspective, a coproduct if a functor of singleton sets. Therefore, (i.e., disjoint union) a functor that preserves limits preserves singletons) preserves singletons (and The Trnkova-Blass note must
and preserves all coproducts, itmust preserve all sets; in other words, it be naturally isomorphic to the identity functor. Therefore, we do not attempt to endow j with such strong preservation properties. Trnkova, and
that preserve
definable inV fromD, that isnot isomorphicto the identity. (2) Suppose j: V ?> V is a definable {with parameters) exactfunctor not
isomorphic
cardinal."
there is a measurable cardinal in V and D is a nonprincipal on k. Then there is an exact functor j : V ?> V, n-complete ultrafilter
to the identityfunctor.
Then V
As before, the universe V plays the role of the category of sets. Again, the functor j is a class in V, definable with parameters. Henceforth, we will
usual
naturality requirement. We use this less natural version here because proofs and related results can be expressed more easily in the framework we easily from theirs.
46
paul
corazza
ifwe
functor j
we can define a be a nonprincipal /^-complete ultrafilter on ? K as follows: For any A", r and any A: Ar 7: jD: V
j(h)([g]) = [hog]
The For
also that
>
happens to be the least possible critical point among all critical points of For exact functors definedas in (*), itcan be shown thattheequivalence j as amember of nK/D, is functionid: k ?> class [id] G jM of the identity a weakly universal elementforj: that is,foranyB and any ye j(B) thereis
f:K-^B exact functors not isomorphic to the identity, then k must least measurable cardinal. We prove this fact in Part 2. in fact equal the
one shows that thecriticalpoint nofF must be at least the identity functor,
cardinal (which must exist). Moreover,
the converse,
whence j ^
id.
with k. We will show thatdespite weak universality [id] can be identified not thecase here thatevery set inV is expressibleas j(/)(?). of ft, it is
We to compare pause on a Trnkova-Blass our formulation functor is an exact functor of a stronger Axiom of Infinity
such that./(/)([id])
y. When
is a normal measure,
as usual
based
Axiom of Infinity. There Trnkova-Blass : V ?> V not isomorphic to the identity. j with the formulation Lawvere Axiom due
Axiom of Infinity:
to Lawvere,
provably
equivalent
to the (original)
of Infinity.
There
is a functor j:
V -> V where
V.
than Lawvere's
is the price we pay for this cleaner formulation tion properties. Moreover, not great: we are simply requiring j to be exact, and exactness is a natural geometric property of functors. we are seeking to strengthen the Lawvere formulation to the Nevertheless, fullest possible extent, and aspect of the Trnkova-Blass so we naturally seek further improvement. One that could be further refined, at formulation
since j itself (rather than its factors) is endowed with the key preserva
Certainly,
critical point of an exact functor not isomorphic to the identity is not always to be measurable (which is the large cardinal property associated guaranteed
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
?* V j :V
47
of Infinity with the Trnkova-Blass formulation is equivalent to the theory ZFC+"there is a measurable cardinal". One problem with this is its lack of robustness under set forcing: We might theory as a foundation a truly foundational expect theory to be indestructible under set forcing the usual Axiom
A natural question to ask as we seek to further refine our strong axioms of infinity, in the direction of a maximal strengthening of preservation proper
has thisproperty). But it is easy to destroya measurable (sinceZFC itself for (bycollapsing or via Prikry forcing, example).
embedding" Axiom.
the preservation
V that is definable
axioms plished byworking in theexpanded language {e, j} and formulating Embeddings, or BTEE.
axioms V ?> V having a asserting collectively that j is an elementary embedding critical point; these axioms are known [7] as the Basic Theory of Elementary regarding j are assumed (in particular, no or Replacement for j-formulas are assumed). This class theory) instances approach
in V. However, ifwe weaken definability sufficiently, inconsistency can be avoided and elementarity of j does not cause a problem. This can be accom
all instances of Separation since, under such an assumption, and Replacement for j-formulas must automatically hold true. 1.6. Axioms of BTEE [7]. </>(xi,X2,..., xm) is an e-formula: <=>
{CriticalPoint). "There is a least ordinalmoved by j". The theory isnot particularlystrong: transitive ZFC+BTEE A model can be found [7]assuming less than0*. This shows thatthe largecardinal strength of ZFC + "there is an elementaryembedding j: V ?> V" isnowhere near (2)
inconsistency j-formulas. in the absence of instances of Separation and Replacement for
We clarifya possibly confusingpoint here, having to do with the lack of definabilityof embeddings thatwe get fromZFC + BTEE. Suppose
48
paul
corazza
:V ?
is an elementary
embedding
because
must be for the Trnkova-Blass case j is not definable in the universe V as it Theorem makes use of to apply. In particular, the Trnkova-Blass Theorem an instance of Separation for j-formulas to define the ultrafilter D from j9
viewed as a functor, j is exact and not isomorphic to the identity. Therefore, one might be tempted to conclude Theorem, that, by the Trnkova-Blass k ismeasurable. in this does not follow, though, because This conclusion
To provide j with additional preserva strength as we search formaximal one further step we could take is to require j tion properties ofaj:V?>V, This additional to satisfy all instances of requirement forces j Separation]. to be very strong. Before investigating this "ultimate" theory, we first briefly survey the various strengths obtainable by adding certain natural instances of Separationj. We begin with several definitions; the reader is asked to con
C M. A cardinal k is extendible if, and is an inner model, of j, M Vj^ for every ordinal rj> k, there is an ordinal ? and an elementary embedding extendible
sult [19] and [20] forunfamiliar terms.A cardinal k is super strong if there is an elementaryembedding j: V ?? M such that n is the critical point i: Vrj?> V{ such thatcrit(z) = k and rj< map / is called an i(k) < ?; the
k is n-huge if embedding with critical point k. For each n G co, an inner model M and an elementary embedding j: V ?> M there exists
is called
if,
cardinal. The axiom I2M asserts there is an ?> M having critical point k so that the inner elementary embedding j: V includes as a subset Vx where X is the supremum of the critical model M sequence of j. The axiom Ii(k) asserts there is an elementary embedding an /3 embedding
and k < X is thecriticalpoint; in thiscase, j: Vx+\?> Vx+\where X isa limit and k isan I\ cardinal.The axiom In(?) asserts thereis j is an I\ embedding an elementaryembeddingL( V^\) -? L( Vx+\) having criticalpoint k,where and k isan lo cardinal. X> k is a limit;in thiscase j is an Iq embedding
Measurable Ultrafilter Axiom of Infinity. The
k G ](X)} is a set.
class {X c
k:
that asserts that the ultrafilter of adding the single instance of Separation] over k defined by j is a set. Perhaps surprisingly, for a theory to assert in
is the result
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : v ?> v
49
this way
that the critical point k of jmust be measurable stronger than measurable: 1.7. Measurable Ultrafilter Theorem
ZFC + BTEE + Measurable Ultrafilter Axiom: upperboundson the theory k ZFC + BTEE + Measurable Ultrafilter Axiom proves that (1) The theory
ismeasurable with Mitchell order > k.
(2) If
Measurable
The critical point k of a j: V -> V will be forced to be a superstrong we supplement ZFC + BTEE instead cardinal if with thefollowingaxiom: Axiom of Infinity. The restrictionj f P(?) P(k)-Amenabilityj
is a set.
Given a model j: V -> V ofZFC + BTEE + P(k)-Amenability Axiom of Infinity, y= let For a g [y]<a>, defineEa fromj by Ea
by the usual arguments,
{XeP(*W):aej{X)}.
is a k-complete
\ P(k^)
inV?and we can define inV theusual (k, y)-extender derived fromj by E E = (Ea: a g [}>]<-).
usual arguments show that if j'e: v ?> Me is the resulting canonical
ultrafilter over
The
= y= embedding, the criticalpoint of j'e is ft,j'e(k) jM, and Vy C ME and [20] fordetails.) Once we know k is superstrong,itfollows that (See [6] {a < k :a is superstrong}belongs to thenormalmeasure derived fromj. 1.8.Restriction to P(k) theorem. The theory ZFC + BTEE + P(k)
Axiom proves that k is the nth superstrong.
Amenability^ conclude
The best known upper bound for this theoryis a 2-huge cardinal [7]. We
this quick survey of intermediate axioms with the following: Huge Amenability^
j \P(P(]n(k))))
Note
is a set.
Axiom
of
Infinity.
The
restriction number n.
such axiom
natural
1.9.Huge Amenability Theorem [7]. Bounds for ZFC + BTEE + Huge Amenability^ are givenby the following, for eachparticular n:
admits a normal measure that contains the set of n-huge cardinals below k. and an ele (2) Assuming an n + 2-huge cardinal k, there is a transitiveM
ZFC + BTEE + Huge Amenability nproves k is n-hugeand (1) The theory ? M mentary embedding M j:
Huge Amenability^.
ZFC + BTEE + \=
50
PAULCORAZZA
?> V with certain additional of an elementary embedding j: V for a foundational properties. Any of these could be considered candidates our aim has been to strengthen the preservation extension of ZFC. Still, existence possible extent. Our way of doing this is to introduce as additional
Each of the extensions of ZFC + BTEE just considered here asserts the
ioms to ZFC + BTEE all instancesof Separation for j-formulas (denoted BTEE + Separation] is known as theWholeness Separation]). The theory Axiom orWA. A model of the theory ZFC + WA provides us with a very
strong structure-preserving map of V to itself.We first establish that, rela is not tive to some of the very strongest large cardinals known, ZFC + WA inconsistent: 1.10. Consistency Theorem
i: Vx ?>
Vx,
meta-definition proofswill be given inPart 2. We beginwith the following ZFC + WA): A cardinal k is a WA-cardinal if there is a WA-embedding
?> V having critical point k. j: V 1.11. The Wholeness Axiom Theorem [4,7]. The following are character (which, as we discuss below, may be expressed formally in the language of
In this case, the model is (Vx, G, /). The following theorem the theory ZFC of the main results concerning -IWA;
summarizes outlines of
isticsof the theory ZFC + WA: ? (1) Assume ZFC + WA and that j: V
underset forcing. (2) ZFC + WA is indestructible model ofZFC + WA, ifthereisone at all, isV The only "natural" inner (3) (4) The criticalsequence (k,
for V. That subsequences critical sequence itself. is, <\)(x\,..., xm) and for any twofinite for any e-formula < "' < Pm of the < otm and fi\ < a\ < ai < of], (= </>[ai,...,am] <r+ (j)[pi,...,pm}. V
critical point k. Then k is super-n-huge for every n; moreover, proper class of cardinals that are super-n-huge for every n.
j(j(?)),...)
a cardinal
that have been studied that among large cardinal notions for every n is the strongest type that is super-w-huge
arises because of thepathologies thatare possible inpart (3) of the theorem, models models of the language {G, j}; the natural inner when considering
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : V ?*> V
51
the intuitive motivation for thisdefinitioncompelling, thenpart (3) of the theorem is quite striking: It suggests that a "good choice" fora model of ZFC + WA is V itself;that, in an unexpectedway, V is even the canonical
model,
of "natural model";
must
Part (4) of the theoremhighlightsthe sense of "ultimacy" thataccompa nies a large cardinal axiom such asWA. It tellsus for example (speaking somewhat loosely) that every large cardinal property thatholds true of k
also hold
being
inner model
of the theory.
cardinals
more on thispoint), thecritical sequence isunbounded in ON (seePart 2 for cardinal property true of k holds true forunboundedlymany every large
in the universe.
of every
in the critical
sequence
of j; moreover,
since
A slightly factof this kind is mentioned inPart (5) of the theorem: stronger The large cardinal propertyof k that is derived from the WA embedding j
also holds for each member of the critical
where
that kn is a WA as a ZFC + WA
= = j j, j (j j),... are definable inZFC + WA fromj application6 j, ingeneral (V,g, j[/1+1]) a model ofZFC + WA that is witnesses the fact and,
cardinal. Thus the metatheorem For can be schema as follows: each e-formula stated formally x2,..., </>(x\, xm)9
critical point k, then every kn in the critical sequence of j is aWA-cardinal, ? k and = The idea here is that the usual iterates by kn
of within property WA cardinals. This propertycan be formulated replicating the language {g, j} as follows:Given that WA holds with embedding j and
sequence.
We
call
These consequences together with the fact thatZFC + WA h \fng = co(crit(j[w+11) kn) state in a formalway that each term of the critical
sequence even though the consequences
for j is aWA-cardinal.
Separation^ toZFC+BTEE
alternative
j-c=
aeON
(J j(cna
In ZFC + WA, such definitions make sense sinceC n Va is a set. It is easy to see that
j-j= U id r aEON
52
paul corazza
x j \ isalso a set. In the literature,the theoryBTEE + Amenability is denotedWA0; note thatAmenability- is provable from Separation]. It is shown in [7] that ZFC + WAo has the same stronglargecardinal consequences as ZFC + WA; Wholeness Axiom Theorem stillholds. However, that is, part (1) of the under set forcing. This ZFC + WAo is not known to be indestructible us to believe that WA is thebetterchoice. ZFC + limitationleads As for thecriticalpoint dynamicsof an embeddingj thatcomes from WA, shows that thecriticalpoint has nearly the strongest possible large part (1) We will show inPart 2 of thepaper that itcannotbe thecase thateveryset is
for / e V. However, j has a much stronger property expressible as j than is found in the other examples we have considered. We state the result here and outline the proof in Part 2: 1.12. WA Critical Point Theorem k. cardinal properties. Such a,j:V?>V has a different sort of seed property.
-> V be a WA j: V [4]. Let there is a function /:??? Then VK such embedding with critical point ?> thatfor every set x, there is an elementary embedding i: satisfying
(2) /(/)(?)= x.
V = {/(/)(?)
Laver
cardinals sequences for supercompact [24], one can show [3], [4] that an "extendible" Laver sequence if there is an extendible cardinal k, there is ?> VK (definitions and details are given in Part 2). As the referee points f'.tv as given here, could be restated more theWA Critical Point Theorem, out,
WA Critical Point Theorem is that, in the spiritof The idea behind the
We will give a slightly with k + 1and ( with i(rt)+ 1. simplyby replacing rj will which this simplification enhanced version of this theorem inPart 2 for ZFC + WA to the testquestion. In this Applying theaxiomatic framework subsectionwe review some of thework done in [3], [4], [5], and [17] to address theTest Question ("Which large cardinals admit theirown brand WA. We offer ZFC 4of Laver sequence?") in the context of the theory
not be possible.
axiom for large cardinals this as an example to illustrate how a foundational contribute to the general program of large cardinal research. Here, may as part of the axiomatic framework not we indicate how the use of WA naturally arise, but also becomes an integral part of the solution
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : V ?* V
53
? as the criticalpoint of some elementaryembedding of the form Vp M set. where M is a transitive Our background theory ZFC + WA ensures the
it also guides our research into existence of many such classes; as ithappens, the question. Here is a more precise definition of the classes of embeddings that concern us:
that are candidates for admitting some type of Laver sequence. The intuition is that the large cardinals belonging to our class are those that arise
Let 9(x, y, z,w) be a first-order formula (in the language {e}) with all free variables displayed.We call 9 a suitable formula if the followingsentence is inZFC: provable
Vx, y, z, w [9(x, y,z,w) => A "x: Vz ?> w "w is a transitive set" A "z is an ordinal" is an elementary embedding with critical point y"]. For each cardinal k and each suitable 9{x, y,z,w), let
?eK={(i,M):
3pO(i,K,p,M)}.
such as strong, supercompact, extendible, superhuge and super can be characterized as classes of embeddings of this cardinals, almost-huge kind. Of course, having ZFC + WA as the background theory ensures these notions, and plenty of other such classes
Intuitively, consists of a class, definedby 9, of elementaryembeddings ?QK ? M; the need to pair up embeddingswith theircodomains in the Vp definitionof ?eK isa technicality that is explained in [4]. We show in [4] how many of the familiarglobally defined large cardinal
sequence, itneeds to have the additional property of regularity, which asserts ?> M that for any set x, there is some i: in the class ? forwhich x e M. Vp More precisely:
1.13.Definition (RegularClasses). A class ?6Kis regular if \ty>n^P>y 3i 3M [9(i, k, /S, M) A /(/c) yA Vy C M ]. >
above are in fact regular classes. We can large cardinal notions mentioned now define a notion of "Laver for any class can be shown sequence" ?eK\ it regular class.
that if such a class does admit such a Laver sequence, it must in fact be a
at k ifforeach setx and each X > k thereare ft> X, and /: Vp such that /(?) > X and /(g)(k) = x.
of Laver
e ?eK
for instance, if 9 is a formula that defines a class ?eK sequences; to supercompactness of k, then /: k ?> VK is Laver in the usual equivalent
We show in [4] that our definition coincides with the usual definitions
In [4]we observe thatan application of the WA-embedding j gives a hint about building Laver sequences generally: Let U be thenormal ultrafilter ?> over k derived from j: U = {X c k |k g j(Ar)}. If/: /c VK is a Laver sequence (in theusual sense), itfollows that ?> Ka isLaver at f a: a} g 17. {a < k \ \ a From thisobservation, one could attempt to build /: n ?? FK generallyby / a arranging it so / (a) is arbitraryif \ isLaver at a (and the intentionis condition holds for"most" cardinalsa), whereas f{a) is a witness that this
to the failure of Laverness usual Laver at a otherwise. This idea can be used to obtain the the and strong cardinals without for supercompact sequences one shows the use ofWA. However, in the presence of theWA-embedding, construction yields Laver sequences for virtually any globally defined large and extendible cardinal notion; for instance: superhuge, super-almost-huge, cardinals. In fact, we show [4] that whenever class ??K is regular and
of Laver sequence, we can ask, in the spirit classes ??K admit Laver sequences?"
is
must admit a Laver WA embedding, it sufficiently "compatible" with the Here we observe thehelpful role played by the embedding j that sequence.
axiomatic framework. For completeness, we of "compatibility."
(Compatibility). ?dK is said to be compatiblewith the if for each X, k < X < j(k), there exist f},i such that WA-embedding j /: Vp -> M, (/, X < ft < Af) g ?BK,and i is compatible with j \ ? N such that ?> N = in the sense that there is k: Af Vp: Vp Kj(^) k o i= j rVp and k \ nM = id^nM Vx 1.15.Definition
Vp-^
vm
\ k / Af
classes
is a summary
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : V ?> V
55
Laver Theorem [3,4]. (ZFC + WA). with the is WA-embedding, compatible (1) If?QK a regularclass ofemebeddings 1.16.Generalized model of ZFC (2) There isa transitive
embeddings then there is an ?eK-Laver sequence. that does not admit an ?%-Laver sequence. in reviewing these results is to illustrate the role of an ax
Our purpose In this example, the theory ZFC+WA iomatic framework for large cardinals. the existence of themany classes of large cardinals that could ad guarantees mit Laver
It then also plays a direct and natural role in the sequences. construction of a general notion of Laver sequence as well as providing a criterion to decide which classes admit Laver sequences. A third contri results of this kind are obtained under weaker hypotheses technical thanWA.
we For instance, obtained in [4] similarresultsby replacing jwith a class of The point is that theaxiomatic framework ZFC + WA not only provides in
this case an "ontology" of large cardinals in the background but also plays a a solution to the question and even serves to guide direct role in formulating further research beyond the direct use ofWA.7 The naturalness are used ofWA as a foundational axiom. Even we if think it is rea embeddings of the form i: V ?> N, modulo several conditions.
to state this acceptance that simply asserts the existence of certain large cardinals lacks the cogency we might expect to find in a founda In addition, as we have pointed out, an optimal extension of tional axiom. ZFC would have other desirable properties, such as indestructibility under about how set forcing. It would also be more an axiom that is clearly motivated this characteristic, of the universe has many in the spirit of foundations to introduce our knowledge of the structure of by obtained from Cantor's early vision
theuniverse on thebasis of ZFC. Certainly the early simpleheuristicshad (and thoseof others) of theuniverseas guidelines forenhancing therichness kind, thesolutionprovidedbyZFC+WA Taking intoaccount issuesof this
advantages. As we have seen, it is strong enough via large cardinals. to derive vir using basic principles
7Evenwithout a formal theoryof largecardinals, the strongestlargecardinal notions have already been used in a similarway. For instance, as we mentioned earlier, the firstproof of determinacy of setswas obtained byMartin assuming an h embedding, and the first Woodin using an lo embedding. Many of proof of projective determinacywas obtained by the insightsin theseproofswere refinedto obtain theoptimal large cardinal hypotheses for these results thatare known today. The differenceis that the strong large cardinal axioms used to obtain theearly resultshad tobe invokedas ad hoc and somewhat amazing hypotheses
rather than as part of a formal axiomatic system.
56
PAULCORAZZA
has long been recognized [22], [28] that an axiom of the form "there is a
nontrivial ?> V" represents a kind of ultimate elementary embedding j: V cardinal axiom in another sense: The strongest large cardinals are de large
by set forcing. It also bears the character of the "ultimate" large cardinal axiom because of the self-replicating feature that if there is aWA cardinal there must be unboundedly such cardinals above k. Moreover, it many
finedusing elementaryembeddingsof the formj: V ?> M; and one of the characteristicsof the strongerlarge cardinals is that the image model M of more of V. For instance, the correspondingembedding(s) tends to include C M and a huge embedding re a superstrongembedding requires V^ bedding j: V ?> V would be thenatural culmination of all large cardinal
axioms.
C M. in a sense already familiar to large cardinal Therefore, some strong but not inconsistent version of an elementary em
We Axiom
consider
"naturalness"
WA of
Our analysis began with this natural and incon properties. preservation trovertible starting point and proceeded by asking, "How can this axiom in refinements of the axiom always proceeded be optimized?" Moreover, richness of the universe, adhering the direction of increasing combinatorial fairly closely to Cantor's principle of maximum possibility. Therefore, since the form of the axiom WA, asserting the existence of a certain j: V ?> V, meets the requirement of "naturalness"
natural one, we suggest thatWA a new foundational axiom. of ZFC that accounts
of
one can reasonably for large cardinals, argue that there is still room for improvement, and also that there are other acceptable alternatives; we consider these points in the next section. Limitations of the theory ZFC + WA and an alternative and then discuss axiom schema.
We make
areas.
ZFC + WA. concerns about the theory The following is a listof legitimate in to WA, introduced Woodin, which offersimprovements some of these by
1.17. Limitations of ZFC + WA. some remarks about each point, an alternative
are super-w-huge for every n are deriv (1) From ZFC + WA, cardinals that able, but none of the cardinals given by the axioms I3-I0 are derivable, a significant though, as was discussed earlier, some of these have played
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j:
57
of reals.
concerning
properties
of sets
(2) ZFC + WA is formulatedin theexpanded language {e, j}. This dimin (3) In a ZFC + WA universe (KG, j), the critical sequence of j is cofinal in the ordinals. This again detracts from its intuitive appeal since it
violates our intuition that the universe is "inaccessible"?an intuition ishes its intuitive appeal.
Points (l)-(3)
For (1), one could invokeCantor's reflection principle tomotivate the existenceof an I3 cardinal in theuniverse,assumingZFC + WA holds true;
even this approach, "Is there a nontrivial
however,
I2?lo
Since it isnot possible to formalizethequestion in the language ofZFC, even with theuse of proper classes (withoutquantification), it is inevitablethat
some extended framework must worked so forces us to regard j as a class, which in a class theory. Doing must obey the instances of Separation and Replacement; then this additional axiomatic To baggage plays a crucial role in the proof of inconsistency. examine the actual strength of elementarity therefore requires a different context to expand
be used
In Kunen's
all. The original intuition about Replacement was that, because the universe is vast and, in Cantor's words p. 44]), "beyond mathematical (cf. [15, in forming a countable determination," (or longer) sequence of sets, the resulting continues should not be cofinal in the universe. This intuition sequence to be realized in a ZFC+WA universe relative to e-formulas, since for such formulas continues to hold true inZFC+WA. Indeed,
?> V without the study elementary embeddings V impact of class theory. To address point (3), we suggest an alternative perspective, arguing that
for studying the question. The only way we know to do this is the language in the way that we have. Therefore, though the
Replacement And,
never require theuse of j. Because j is a verydifferent which typeof entity, encodes (via an extendibleLaver sequence) all sets in theuniverse, it may not be too heretical to consider thatV could appear different fromj's point of view, and that thispoint of view is not accessible to inhabitantsof V.
situations in set theory are commonly encountered. For instance,
Analogous
58
consider L's view of the F-cardinal means that from "inside" the world
This
is not available
What grants / its special status is the existence of 0#? In a similar way, in of L. the comprehension that is beyond something a ZFC + WA its the co-sequence universe, what grants k,,)(k), j2(/s),... that is not special status in V (the status of being a cofinal co-sequence that are in V) is the presence of j?transformational G-definable dynamics to this universe. outside
sequence is an altogetherdifferent typeof entityrelative to V\\ though it nor a (definable) class relative a a subcollection of is V^, it is neither set
unknown phenomena we might find ina ZFC + WA universe remainforever mathematics of sets and to the inhabitantsof V. For thebusiness of the
G-classes,
universe.
the comprehension
of V.
Therefore,
any non-Replacement-like
no violation
of Replacement
can
ever occur
in a ZFC
+ WA
To conclude
WA.8. Woodin which could be used inplace of schema due to Weak Reinhardt Axiom (WRA). There is a j: VM ?
with critical point k such that VK <V.
VM
Woodin calls the critical point k in this axiom a weak Reinhardt cardinal. The notation WR(?) means thatk isa weak Reinhardt cardinal. The axiom motivated in the following WRA is technically ways: Kunen's inconsistency shows that (in ZFC) there is no nontrivial elementary embedding proof
j: derivable to be inconsistency appears As a j: VM Vx+\, i.e., an Ii embedding. since the is unsatisfactory foundational however, an Ii embedding axiom, strongest large cardinals cannot be proven from this axiom to exist outright Vx+2- Kunen from a nontrivial observed ? ?> that no
in V, but only in themodel Vx+\ (or in Vx, which is a model of ZFC). reflect the model Vx does not sufficiently is the Intuitively, difficulty that the of truth V. This shortcomingis remediedby requiringVK -< V. This extra
greatly enhances the strength of the axiom; for completeness,
requirement
one can prove that there From the theory ZFC + WRA, 1.18. Proposition. in the universe, and also arbitrarily large are arbitrarily large I\ cardinals cardinals that are super-n-huge for every n. Woodin has not (as far as we know) proposed his axiom as 8Wehasten to point out that the "right" or "best" strong large cardinal axiom or as a means of fillingthe need for an as optimal extension ofZFC to account for largecardinals. His axiom was simplypresented which largecardinals can be derived axioms from an example of one of the strongest possible
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j: V
59
Proof.
cardinal.
? = = <f>(x) "x is an ordinal" A 3y> x 3d > y3e: Vs+\ K^+i (crit(e) y). First we claim thatVK \=Va 0(a): Fix a < /s. Since k is an Ii cardinal, was arbitrary,the V |= 0(a). Since VK < V, VK \=0(a). Since a < claim follows. But now, from the claim and the fact thatVK < V again, it follows thatV |=Va 0(a), and we are done. For the second part, let /: Vx ?> Vx be the restriction an Ii embedding of = k. One can argue as in to Vx,where X is a limitand crit(i') part (1) of the Wholeness Axiom Theorem (given inpart 2 of thispaper), insideV^9 to
or [20, Exercise
the following
s-huge for everyn" holds in Vx. Since VK -< Vx (see [4,Proposition 3.12] statement holds inV. H
24.5]), the statement holds in VK, and because VK -< V, the
large cardinals
of "seed"
thiswithout any kind of violation of Replacement. Also, it has similar criticalpoint dynamics to thoseof does not play therole WA, thoughk itself
in this case (see below). Nevertheless, cardinals throughout the universe (by Proposition there are a and /: a ?> Va such that since there are extendible 1.18), one can prove that
the very strongest largest cardinals, even can be accounted for. And it accomplishes
V = {/(/)(a)
One
to WRA is consistent;unlike that it isdifficult assess whether WA, it is not tobe consistentrelativeto any of theusual very stronglargecardinal known
axioms.
disadvantage
is lo- Thus, even this seemingly all-encompassing able to account for all the known large cardinals. quite fromWRA
therefore V were, let k be the least weakly Reinhardt cardinal; then VK?and itself?satisfies the sentence "there is no weakly Reinhardt cardinal", which
more One could argue that the language inwhichWRA is formulated is natural than {e, j}, but note that WRA is not formalizable inZFC: If it
is impossible. WRA can be formulatedin the language {?, c}, where c is a new constant symbol intendedtodenote thecriticalpoint k of an embedding ? collection of axioms could be used to assert j: Vx+\ Vx+\. An infinite
formally that VK -< V.
we found forWA
cardinals. is Indeed, the existence of two such cardinals than the existence of just one, as one can see from classical strictly stronger If k < S are weak Reinhardt cardinals, notice that since Vs -< V, reasoning: we must have VK -< V$ and also that, in Vs, there is an Ii embedding having
60
PAULCORAZZA
(This
a similar vein, froma model of ZFC together with arbitrarilylargeweak Reinhardt cardinals one can obtain a model of ZFC inwhich theweak Reinhardt cardinals are bounded in the universe: Startingwith a model (M9E) of arbitrarilylargeweak Reinhardt cardinals?formalized in some (in the language {G,ci,C2}) inwhich k and S are both weak Reinhardt cardinals, k < S9 and forall ywith k < y < S, y is not weak Reinhardt. ZFC+WRA and also "theweak Reinhardt cardinals N9 Then, in Vs satisfies
are bounded in the universe." cardinal does not imply the existence of others expansion of the language of set theory?one can obtain a model (N, E,k,S)
k to a countable we conjecture that collapsing by set forcing?moreover 1.18 Of course, as Proposition ordinal would be enough to destroy WRA. no such forcing can alter the fact that Ii cardinals and cardinals that shows, are super-w-huge for every n must pervade the universe.
ZFC + WRA can be destroyed above it.This fact leads us to conjecture that
The conclusion thatwe draw from these bits of reasoning is that the
The reasoning just given also shows thata weak Reinhardt cardinal typi globally defined largecardinal proper cally fails tohavemost of the strong, and superhugeness.Suppose forexample thatk isboth likeextendibility ties, ? extendibleand weak Reinhardtwith Ii embeddingj: Vx+\ V^+i. Lete) be Now an inaccessibleabove Xwith VK < Vs (thisfollowsfromextendibility).
is strictly + "/c is extendible" Thus, WR(?) do not know if the same can be said for super (We since that k is strongly compact It does follow fromWR(/c) compactness. as mentioned limit of supercompacts.) must be a measurable it Therefore, that there is an /: n ?> VK such that above, it does not follow fromWR(/c) ofWRA.
V = {i(f)M
existence
axiom. be entirely confident about it as a candidate for a new foundational relative strengths and with a table comparing this section We conclude ofWA and WRA. weaknesses
motivation WRA in our view is that its Perhaps thebiggest drawback to we know) is purely technical: It arises from taking the strongest (as far as formof an elementaryembedding froma rank to a rank not known to be and ensuringa global effect requiringVK -< V. It is not inconsistent, by clear from "first principles"why an axiom such asWRA should hold true WRA is not known to be consistent relative to in theuniverse. And since to other very strong (and well-known) large cardinal axioms, it is difficult
of such an /
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : v -* v
61
Property Consistency
of Axiom
Schema
Not
known
to be
Naturalness
Motivated by
strengthening preservation properties of
Primarily techni
j: V ->
is a
culmina
tion of
criticalpoint k is strongly
though
global large
cardinal erties, prop as such
extendibility and
_superhugeness.
un
62
PAULCORAZZA
"Inaccessibility" of the
universe
cofinal
Replacement
e-formulas,
{ ?, j}
{ ? >c)
WA cardinals WRA "Ultimacy" of theaxiom implies schema have the self- there is a proper
replicating WA
erty:
If k is a
un-
prop-
class
nals and of
cardinals
of Ii cardi
cardinal,
super w-huge for every n. But the exis tence of one WR cardinal does not it.
others above
implyexistenceof
proofs of the main results. In this section we give a concise review of category-theoretic Preliminaries. that are used in the paper. We assume the reader knows the def concepts ?2. Selected
We also terminal objects, initial objects, and exponentials. coequalizers, assume the reader knows the definitions of monic, epic and iso arrows, and of functor and natural transformation. See [26] as necessary. all arrows a ?> b in C.
inition of a category
and
the assortment
of standard
constructions
that
For any categoryC and objects a, b inC, the setHome (a, b) is the set of
A diagram D in a category C is a collection of C objects {df \ i El} with some C arrows d\ ?> dj between some (or all) pairs of the together
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
?+ V j :V
63
objects inD. A cone for a diagram D consists of a C-object c together with a C-arrow /,-: c ? di for each object d\ inD such that the diagram di->dj
c The notation commutes, whenever k is an arrow in the diagram D. {/, : ?> c ?> a cone for D. The dual concept is a co-cone {gi: c}, di} signifies ?> c for each di in Z>. consisting of an object c and arrows g/: di for a diagram D is a Z)-cone {/, : c ?> rf,-}with the property that A /i/w# ?> : for any other D-cone </, there is exactly one arrow f:c'?>c }, {/* c'
/
commutes Dually, for every rf,-in Z>. a colimit for a diagram
property that for any other co-cone g: c ?> c; such that the diagram
? Z> is a co-cone c} with the {#,-: ^ ? there is exactly one arrow di c7}, {g[:
gi/
c-^c1
di
\Zi
commutes
The constructions of products of finitely many objects, equalizers, and terminal objects are examples of finite limits. The constructions of coprod ucts of finitely many objects, coequalizers, and initial objects are examples products and has exponentiation.
A diagram is many objects and finitely many finite if ithas only finitely arrows.A limit (or colimit) is ifitsdiagram isfinite. finite
IfD isa diagram inC and F: C ?> V is a functor, F(D) is the D-diagram whose objects aiQF(d),d is Gi), and whose arrows areF(fi) whenever an arrow in F is said topreserve (finite)limitsifforevery (finite) D. diagram D and limitcone {/}: c^ di) for inC, {F(fi): F(c) -+ F(di)} is a limit D cone for V. F(D) in The dual notion ofpreserving(finite)colimitsisdefined
similarly.
A functor F: C -> V is left (right) exact if preserves all finite limits F F is exact if it isboth leftexact and rightexact. (colimits).
64 Suppose F: C ?> V
is a functor and
F(c) is a universalarrow ifforany x G C and any g: d is a unique /: c ?> x inC such thatg = F(/) o w. c </ ?^F(c)
F(x)
inV, there
V and G: D ?> C are functors.Then F is leftadjoint Suppose F: C toG, and we write F H G, if there is, for each c G C, d G D, a bijection In this ^: Homp(F(c),rf) ?> Homc(c, G(d)) that isnatural in c and
case
arrow living in #cir(c)(l/r(c)); that is,^c is the imageunder # of the identity is a universal arrow One shows that rjc: c ?> G(F(c)) Homx>(F(c), F(c)). for each c and that the collection {rjc \c G C} are the components of a
77is called the unit of the adjunction. is completely determined by its unit. That is, given functors adjunction F: C -+V and G: V ?> C and a natural transformation rj: \q ?> G oF such : c ?> G(F(c)) is a universal arrow, then FHG. that, for each c G C, ^ F H G, F preserves all colimits of C and G preserves all limits Whenever natural An tranformation 77: \q of V. category of sets, denoted Set, has as objects all sets and as arrows all functions between sets. For any category C, the category of endos from C, ? denoted C^, has as objects all C-arrows c ?> c. Given /: c ?> c,g: d an arrow a: / ?> g is a C-arrow ea\ c d thatmakes the following G commute: diagram The
c-^
(F,G99)
is said
to be an adjunction.
For
each
c G C,
let //cdenote
?> G o F.
d-^d ?> Set is a functorand c G C. An object u G F(c) is a Suppose F: C universalelement F ifforeach J G C and each y e F(d) there is for weakly F : c ? d inC so thatF(fd)(u) = j>;more verbosely, is said to be an f by weakly represented c withweakly universalelementu. Moreover, if ^ is
u is a universal element for F; again, one unique for each choice of d, then case that F is represented by c with universal element u. also says in this universal element
way.We will useful to know whether everyset inSet can be reached in this -> Set is cofinal iffor everyx G Set there is F: C declare that a functor c G C such thatx G F(c). One easily verifiesthat ifu is a weakly universal
vast expanse
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : V ?> V
65
is rather unnatural (as the referee point of view, this definition of "cofinal" has pointed out) because it is not preserved by natural transformations. to a universal This notion and its (set-theoretic) connection element has turned out to be conceptually helpful, so we have used it advisedly; Theorem 2.17.)
F: element fora functor C ?> Set, thenevery set is expressibleas F(f){u) some arrow/ inC if F for and only if iscofinal. (From thecategory-theoretic
see
category M as follows: The objects are the elements ofM. f a ?> M b is an arrow in the category if and only if function".
the usual set-theoretic product Since, internal toM, is cartesian closed. ab operations can be carried out, M exponentiation of this kind as a category of sets. Lawvere's Theorem. As we mentioned in Part
fact a theorem about cartesian closed categories. we restrict our attention to categories obtained ZFC ? Infinity, which we treat as a category
to the proof of Lawvere's Theorem, we need to introduce a category-theoretic construction for the set of natural numbers. Recall
category are isomorphic and that, in a category of sets, any singleton is a terminal object. Terminal objects are typically denoted with the numeral 1.
X,
with the followinguniversal property:For any diagram 1 U -UU inC, thereis a unique arrowX ^ U such that thefollowingdiagram commutes:
X---^X
u-'?^
66
PAULCORAZZA
ists, and we can define the usual n U as well as the map z: {?},
We outline briefly holds why the assumption that theAxiom of Infinity in V implies there is an NNO in V. The Axiom of Infinityimplies co ex
successor function co : 0 i-? 0. 1 ?> succ: co ?> co : n Now, the assertion
i?>
CO
h
u--?^ u
(1)
t: U ?> U by primitive
one can obtain an NNO (namely, (co,z, succ)) from the Axiom of Infinity, inZFC ? Infinity. working To prove the converse inZFC - Infinity?that the existenceof an NNO implies theAxiom of Infinity?startwith an NNO (X,z,s). We show thatX must be infinite. Define by recursion the class sequence F =
(X0,Xi,*2,..., *?,...) by
x0 = z(0) = s?(xo),
xn+i = s(xn) = sn+l(xo).
By Separation, the range of F is a set (being a subclass of the setX). One now uses inductionand theuniversal propertyofNNOs to show that, for uses the formula thatdefinesF to define the inductionP(x):
each n, xo>*i,*2> ,xn are distinct. (For the induction, formally, one
P(n) asserts
the
axiom
of
infinity
and
transformations
67 j\V?*V
\n +
1 has n +
hence
that xn,x\9x29...,xn
Now, lettingt*(x) denote the z'thiterateof t at x, and lettingr?(x) = x, it follows thatfor 0 < i < n, tl(xo) ? x, = s1(xq)9 but ly"+1(^o)^ u ?
tn+1(x0). ing (1) commute. To show that xn, x\9..., xn9 x?+i are distinct, assume in stead that x?+i = Xi for some /, < /< n\ in other words, sn+l (xq) = 0 (xq) for some z, 0 < /< n. Tracing through the diagram, we have
By theuniversal propertyofNNOs,
there is a unique h :X
U mak
X We have completes the induction and establishes that must be infinite. shown inZFC ? Infinity that theexistenceof anNNO impliesthe Axiom of
Infinity. These observations 2.1. NNO Lemma. are summarized The theory ZFC in the following ? lemma: that thefollowing
Infinity proves
(2) There exists a natural numbers object. P. Freyd noticed how to enhance the diagram for an NNO to capture definition by recursion with a parameter. His result holds in any cartesian the proof of Lawvere's theorem. We context of models of set theory. 2.2. Freyd's Theorem
is a function definedby x${a) = (a, 0) and Ia x succ: Axco?>Axcois = = definedby (Ia x succ)(a,?) (a9 succ(n)) (a9n + 1). Then, given any
? there is a unique h: A x co B such that
[13]. Suppose
is a set9 x e A, xn: A
A x co
68
A
PAULCORAZZA
x co-> \AXsucc A x co
We do not give theproof here. The main idea restson the connection to we letB = A, the concept of definitionby recursion. In the diagram, if ? A x co = ?> A = 1^, then definexo: A (a, 0), and define/: A by xo(a) theunique h: A x co?> A that is givenby the theorem is theunique h given Definition By Recursion Theorem for thedata by the = x, = h(x,n + 1) g(h(x,n)), h(x, 0)
where We x g A and g: ^4 ?> A are given. proceed now to the proof of Lawvere's we restate the theorem: Theorem. Suppose V is a model Theorem. oj ZFC For the reader's the
?-?Ji
are equivalent:
Infinity. Then
Axiom of Infinity. (1) V satisfiesthe G There is a functor j: V ?> V that factors as a composition o F of (2)
functors
In particular, F preserves
adjoint toG), (A) F HG (F is left -> V?9 (B) F : V -> K w rteforgetful functor,definedbyG(A ^ (C) G: V?
all colimits and G preserves
satisfying:
all limits.
A)
= A.
V-^
V0
= ea: a ?> fe. = b, and = adjoint Suppose F is a left G(a) a9G(g) G(f) = G o F be theLawvere functor. We exhibit to G, defined in V. Let j which it follows that the arrows z, s such that (y(1), z, s) is an NNO, from NNO Lemma). Moreover, we show thatthe holds (by the Axiom of Infinity criticalpoint forj is 1.
?> a9 ?> V denote the forgetful functor. If /: a Infinity. Let G: b ?> b are objects in V?, and a :f ?> g is an arrow in V?, note that
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : V ?* V
69
First we observe that y(0) = 0: Being a leftadjoint, F preserves all colimits,and so, inparticular,preserves the initialobject 0 of V. It is easy to verifythat the initialobject of V? is the empty map 0 ?> 0. It follows, = = = 0. thaty(0) therefore, G(0 -+ 0) G(F(Q)) we apply j to 1and show thatj (1) is theunderlyingsetof anNNO Next, and, inparticular, that |y(l)| > 1. F(l) is an object s: X ?> X inV? and = X. Let rj: 1y ?> G oF be theunit of the adjunction ofF and G(F(1)) (?, and let z = rj\: 1?? (/(.F(l)) be itscomponent at the object 1.We use
the fact that z is a universal arrow to show that
1 il^I
is an NNO: Suppose we are given F-maps 1 {/ (7. Note that s and g are objects in V?. By universalityof z, there is a unique a : s g such
that j> = G(a) o z = ea o z.
1-^GF(l)
G(a)
1 ea ea
u-?>u
In particular, (X, z, s) is an NNO, as required.As j(0) = 0 and y (1) = X, we which is infinite, have also established that 1 is thecriticalpoint forj.
We now prove the other direction 1?> of the theorem. Assuming is an NNO, where x: co: 0 i?> 0 and succ is the usual successor
our the Axiom of Infinity, proofof the NNO Lemma givesus that (co, succ) x, We definea left functor V?-> V as G: adjointF: V ?> V? fortheforgetful follows:For any setA, F(A) isdefined tobe theendo Axco 1-4XSUCC) Axco. Given a F-arrow /: A ?> B, let (f) be theF?-arrow / x lw: l^x succ ?? F
1 x succ making # the following diagram A x co->
* Uxsucc
commute:
A x co
*
x co-^
l^xsucc
x co
70
PAULCORAZZA
For each A e V and each w#:B ? B G V?, we obtain a bijection ?a,uB:nomvo(F(A),uB)9*Homv(A,G(uB)) natural inA and
f:A?> G(ub) = B, we
. Given
in the
w^, as
following:
A x co->iA XSUCC
A x co
e<f>
map h guaranteed by Freyd's Theorem. It is routine to verify that this correspondence is a bijection that isnatural inA and H As we observed in the proof, the critical point of theLawvere functor = co. The seed behavior of j is given by the j: V ?> V is 1, and j(l)
let j = G o F be the Lawvere functor. Then 0 e G(co co) is a universal element for G. In = B there is a B and every y G G{ub) unique particular, for every endo B ? ?> ub so that G a: succ y. Moreover, (a)(0) 2.3. Lawvere Seed Proposition. Working in ZFC,
following:
V = {G(a)(0)
Proof. ? 1 map Let x:
1 ?> co: 0 ?->0. Identify the element >> G G(wjb) with the isomor with succ via the canonical 2?: 0 i-> >>, and identify F(l) a:
G{F{\))
of
succ ?>
ub
succ
1?^GF(l)
G(a) G{?B)
H
of Blass [2], showing that there is a measurable cardinal ifand only ifthere is an exact functorfromV to V not (naturally) isomorphic to the identity to (the reader is also referred Trnkova's work [32] fora somewhatdifferent
In some cases we have re-formulated his results to emphasize argument). that we pursue in this paper; we include the framework of generalization
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j:
V ?> V
71
omission
or proofs for these points of departure when neces is our omission of the naturality of such a re-formulation example in the original statement of the Trnkova-Blass Theorem. This condition additional sary. An discussion was discussed in the comments 1, and the proof of its harmlessness Theorem 2.12. following in Part is given
was also made inPart 1) is that,since every A helpfulobservation (which set is (isomorphic to) a coproduct of l's, ifa functorpreserves 1 (as it must if it preserves all finite limits) and all set-indexedcoproducts, it is functor. As is thecase with elementary naturally isomorphic to the identity
embeddings F
a functor is right preserves all finite products and all equalizers. Dually, exact if it preserves all finite co-limits; equivalently, if it preserves all finite co-products and all co-equalizers.
we statewithout proof We beginwith severaldefinitionsand results that can be found in [2]).As was done for V Lawvere's Theorem,we treat (proofs as the category of sets. Recall (a model of ZFC or even ZFC Infinity) thata functor is leftexact if itpreserves all finite limits;equivalently, if it
V ?> M,
a measurable
cardinal
is thecriticalpoint ofF.
that, ifF:
such that k
is an exact functor
X the identity, then if is any set for which F(X) ^ X, it follows that is X infinite (sinceF preservesall finitecoproducts of 1). This observation can be established in the theory ZFC ? Infinity, and allows us to conclude that the Axiom of Infinity provable from is ZFC ? Infinity "there is an exact functorfromV to V +
not isomorphic to the identity." We will often state in hypotheses of theorems that "F is an exact func tor"; to be precise, we think of F as a class defined possibly with param eters. A theorem that assumes the existence of such a functor should be viewed as a schema of theorems, one for each class-defining formula F.
isomorphic
to
Suppose F: V ?> V is leftexact and F(0) = 0. For any setA and any a e F(A), define FA,a(X) = {F(f)(a)\f:A-+X}.
A, a.
Formally, froma formula thatdefinesF, we have a finitistic procedure for on Suppose D isa filter a nonempty setA.lff and g are partial functions on A, we write/ ~ g ifand only ifthe set of a for which / and g are both
obtaining a formula defining FAta, with extra parameters
Z)-Prod(jr)
= =
XA/D A / {[f] I is a partial function -+ D-Prod(r): X with domain in D}. ^ [g] [hog].
D-Vrod(h): D-Fvod(X)
D-Pvod Z>-Prod proof: 2.4. Subfunctor Suppose X is a set. Lemma. is called is a left exact
a reduced power mod D. It is straightforward to show that state the next several lemmas without functor. We F: V ?> V is left exact andF($) = 0.
Suppose
(1) For every A9a9FAta(X) QF(X). x (2) For every G F(X), x G FyfJC(JSr).
: V ?> V
and only (4) Suppose I isa set. Then F preservesall I-indexed coproductsif
all I-indexed
is left exact.
^(U*)-II*w i i
for each A9 a ifand only if
FA^jlB^^llFAABi). i i
over all A9 a of F/4,a(Ar). Given
is theunion
A is a set
D = {X QA\a ? ranFfo)},
Z> is the filter derived from F9A9a. to preserve F if happens the subset relation
of definition ameasurable ultrafilter Notice thatthisisessentiallythefamiliar all F(ix)" becomes (moreprecisely, inclusion maps), theexpression"a G ran more familiar"a G F(X)". the = 0. exactwith V 2.5. Equivalence Lemma. Suppose F \ -* V is left F(0) set Let A be a nonempty and a G F{A). Let D be the F9A9a. filterderivedfrom Then
Fa,u 2.6. Reduced Power is naturally Filter isomorphic to Z>-Prod. D is a filter. Lemma. Suppose from an elementary
where
finite coproducts ifand only ifD isan ultrafilter. (1) D-Prod preserves
(2) Suppose
X is an infinite cardinal
D-Prod
preserves
X-indexed coproducts
the
axiom
of
infinity
and
transformations
j : V -> V
73
2.7. Coproduct
Coequalizer
Lemma.
Suppose
F:
V ?>
F w fe/f exac/1
D (1) IfD-Pvo&{k) ^ k, then isnon-principal. ? k andD is and n-complete, then (2) IfX nonprincipal D-Prod(ft) ^ k. a thatforany total/: X?> k, if = f{y), then [/] = [ca]. Also notice that for everypartially defined g : X such that k, there is a totalf:X?>k = as Therefore |Z)-Prod(?)| = ac, required. [g] [/]. For (2), assume Z) is nonprincipal and suppose (fa: a < k) are such a thatforeach a, [fa] is an elementofD-Prod(k) and if ^ /?, ^ [/p]. assume eachfa is total.Define g: k ?> /c letting Moreover,WLOG, by g(/?) be the least elementnot belonging to {/<*(/?) |a < /?}. Then for everya > and every /? a, g(/?) 7^ /<*(/?) Since Z) is a k-complete nonprincipal it fromeach elementof ([fa] ultrafilter, follows that [g] isdifferent > inparticular,Z)-Prod(/c)^ k. H Therefore, |/)-Prod(?)|
2.9. Trnkova-Blass Theorem there is a measur (First Half). Suppose able cardinal k in V and D is a nonprincipal n-complete ultrafilter on k. Then Z)-Prod is an exact functor V ?> V that is not isomorphic to the identity. In particular, k is the critical point of 2)-Prod. Outline of Proof. As we remarked is an ultrafilter, Z>-Prod and k is uncountable, earlier, Z>-Prod is left exact. Since D finite coproducts. Since D is ^-complete preserves Z>-Prod preserves countably indexed coproducts; Proof. For a < k, let ca: is principal with generator y < X. For each (1), assume D ?> k be the constant function with value a. Now observe X
hence, by the Coproduct Lemma, Z>-Prod preserves coequaliz Coequalizer ers. Therefore, Z>-Prod is right exact. Morever, Power Filter by the Reduced Lemma and ^-completeness of D, Z)-Prod preserves all A-indexed coprod = X for all such X. ucts, for X < k, and so Z)-Prod(A) By the Nonprincipal Filter Lemma, is not isomorphic to the k, and so Z>-Prod ^ Z>-Prod(/s) k is the critical point of Z)-Prod. H identity; indeed, ?> V is 2.10. Trnkova-Blass Theorem (Second Half). Suppose F : V an exact class functor not Then V isomorphic to the identity functor. (= "there exists a measurable cardinal". Indeed, ifk is the least cardinalfor which
ex measurable (whichmust F(k) ^ k, then,inV, k isat leastas bigas theleast is theleastelement {a: a is thecritical Moreover, ifk ist). of point of an exact
to the identity}, then k is the least mea
By our earlier remarks, k must be infinite. By the Coproduct k must be uncountable. since F preserves coequalizers, Lemma,
^ k.
(4)
Coequalizer By part
since F
Z)-Prod(tt)^ FAm(k) ? ?.
so must Z)-Prod, and by the Reduced Since F preserves coequalizers, Power Filter Lemma, D is a>\-complete. By theNonprincipal Filter Lemma, D must also be nonprincipal. It follows that there must exist a measurable argument takes place inside V since both F and FAm are definable
cardinal in V (see for example [11,Corollary 2.12]). (Note that the entire Let y denote the leastmeasurable cardinal in K. It follows thatD is y-complete (see for example [11,Theorem 2.11]). By theReduced Power
again, it follows that Z>-Prod preserves /^-indexed coproducts over V.)
Filter Lemma
= y. It sufficesto show that k < y. Let U identity}.We show that k on be a nonprincipal y-complete ultrafilter y. By theproof of theFirstHalf
of the Trnkova-Blass Theorem, U-Prod is an exact functor V ?> V with
forall p < y. Since D-Prod does not preserve ]JK 1, itfollows thaty < k. us Finally, let suppose thatk happens tobe the leastelementof {a: a is the
V ?> V that is not isomorphic to the
(/-Prod(y) ^ y. By leastnessof k < y,as required. H We remarkhere that thedefinitionof n as the least elementof the setS = V ?> Fthat isnot isomorphic {a: a is thecriticalpoint of an exact functor to the identity} is formallyexpressible inZFC. Consider the following set, defined inZFC:
T = {a : 3A, D (D
is an a>\-complete
nonprincipal
ultrafilter
We
denote the To begin, we let show k is the least element of T. so that Z)-Prod(tto) least element of T with witnesses A,D, /^o- Since ^ ? V be an exact let F: V V ?> V is exact, k < kq. Conversely, D-Prod: of k in S. Since F(k) functor witnessing membership then, as we ar ^ D-Prod
^ k and a corresponding
^ k. Therefore, kq < k.
We have shown k is the least elementof T. We also mention here that theproof ofTrnkova-Blass Theorem (Second F Half) makes essential use of the fact that is a class over V. In order for D thederived filter to be a set in V, F must be given by a formula. Later, will be we will discuss other functorsthat are not definable over V and it toD will not provide us with a measurable cardinal clear that theanalogue in V. (A rough example along these lines thatcan be given at thispoint is where L is the an elementaryembedding j: L ? L with criticalpoint as, constructibleuniverse and j is given by the existence of 0#. Though one can define (in V) the subcollectionD = {X G Pl(k) | k G j{X)} of L,
and D-Prod(k)
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : v ?> v
75
The prove that the critical point of any exact functor V ?> V ismeasurable. obvious strategy to accomplish this is to observe, as in the proof, that, given an exact F: V ?> V, not isomorphic to the identity, there is, over some A,
is not a set inL preciselybecause it is defined in termsof the external embeddingj.) In theproof ofTrnkova-Blass Theorem (SecondHalf), onemight hope to
an coi-completenonprincipalultrafilter for D which Z)-Prod(Ac) k (where ^ k is the criticalpoint of F). By leastness of we have thatF preserves
A-indexed /^-complete), but this is not generally the case, as A. Blass author. We formulate the issue as an open question: 2.11. Open Question. Suppose F: V ?> V
coproducts
X <
k.
The
natural
hope
out to the
to the identity. Without assuming other conditions on F, is itnecessarily true that thecriticalpoint ofF is a largecardinal?
in Part 1, our version of the Trnkova-Blass Theorem from the original theorems which referred to "natural iso easily instead of merely "isomorphism" in the theorem statement. We morphism" prove this now. follows As we mentioned
functorV ?
Assume Suppose
2.12. Original
TrnkovA-Blass cardinal.
Theorem
there is a measurable
is
an
exact
V not isomorphic
to the identity.
is a ^-complete ultrafilter over a measurable nonprincipal V ?> V is an exact functor with Z)-Prod(^) which ^ the existence of an exact functor not isomorphic to the identity,
As our proof of theTrnkova-Blass Theorem indicates, it is often the case that thecriticalpoint of an exact functor V ? V that isnot isomorphic j:
to the identity is a measurable cardinal k. We show, however, that such
D-Prod.
element at First, if such a j does have a weakly universal Proposition: must be naturally isomorphic to one of the reduced all, j product functors
a functor is unable to generate all sets in V with seed k; in otherwords, = V 7^ {j'(/)(?) |dom / k}. The reason is summarized in the following Also, whenever D is co\complete, the class \JX D-Prod(X) is
PAULCORAZZA
ultrapower VK/D whose transitive collapse M does In particular, when D is normal over k, ifwe identify the in the usual way, k, itself is the seed, with its transitive collapse V ?> V is an exact func
power Z>-Prod, then admits a (1) If j isnaturally isomorphictoa reduced j j (2) If j admits a weakly universalelement,then isnaturally isomorphicto
a reduced power. is a normal measure (3) If D weakly universal element.
of for D-Prod {via theusual identification D-Prod(n) with its transitive collapse), but for some set Y, Y ^ Z>-Prod(/)(?) for anyf.
of (1). We show that for any reduced power D-Prod, the element
on k, k g
Z)-Prod(ft)
is weakly
universal
= [id] g D-Prod(A) (whereA \JD) isweakly universal forD-Prod: = assume Suppose [/] g D-Prod(Ar) XA/D. Without loss of generality, ?> X is total; but now / is a candidate towitness weak universality: /: A We must verifythat Z)-Prod(/)([id]) = [/]; but this follows immediately a =
from the definition of Z>-Prod on functions.
Proof
Proof of (2). Suppose F = j: V -? V is exact, not isomorphic to the and has a weakly universal element a g F(A). We show that identity, to F = FA,a- By theEquivalence Lemma, this suffices prove (2). Suppose = x. x g F(X). By weak universality, thereisf:A?> X such that F(f) (a) g F^(X). By definition,then,x g FA,a{X). Conversely, supposeF(/)(?) g F(JT). F(X), F(f)(a) Clearly, sinceF(f); F{A) -U Proof of (3). We have seen that [id] isweakly universal for 2)-Prod in
general. If D
D-?roA{X)
M for D-Prod. As usual (see [20,Proposition 5.7] forexample), if denotes = the transitive Z)-Prod(X), thenD g M. So, via collapse of VK/D \JX D our identification, belongs to no D-Prod(X), whence D ^ D-Prod(/)(k) forany/. h
The Wholeness strengthen Axiom. As we discussed in Part 1, as we attempt to it the Trnkova-Blass functor j to an elementary embedding,
so, identifying ultrapowers with their transitive collapse Therefore k g Z>-Prod(/s) follows that k> g Z>-Prod(/s).
is a normal measure,
we have
that each
of the ultrapowers
is helpful to work in the context of the language {g,j} ZFC + BTEE. We recall theaxioms of BTEE: (l)^
is an axiom, where Vxi,x2,...,xm x2,..., </>(x\, ...,xm) xm) <=> is an g-formula, 0(j(xi),
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j : v -+ v
11
(2) {CriticalPoint). "There is a least ordinalmoved by j". we showed thata transitive model for ZFC + BTEE can be obtained In [7], froman co-Erdoscardinal (which isweaker than 0#). By adding to BTEE
the axiom schema
The important must be definable Kunen's result point forus is that thej in
in K.
one's assumptions it about the surrounding universe. To avoid pathologies, A canonical is helpful to identify a class of natural models. example of a natural model arises from Kunen's classical result that the existence of 0# is ? L. equivalent to the existence of a nontrivial elementary embedding j: L In particular, ifwe are given a j: L ?> L, in order to carry out a
(3)^ {SeparationSchema for yformulas). Each instanceof theusual Sepa is ration schema involving<j> an axiom (where0 is a j-formula). Different kinds ofmodels of ZFC + BTEE are possible, depending on
Separationj
we obtain
theWholeness
Axiom
orWA.
definable inV. Indeed, proof (see [19]) that0#exists,j must be sufficiently in [7]we give an example of a model Af forwhich there is an elementary embedding i: LM ?> LM and 0#does not exist.The point of theexample is
that /does not satisfy the instances of Separation for the standard argument to go through. The Kunen embedding can be represented and Replacement necessary (though it in our present context
cannot be formalized inZFC) as a model (L, g, j) ofZFC + BTEE where j is definable in V. In [7]we call such a model sharp-like; our intention here is that the sharp-like models (models (Af, i) of ZFC + BTEE for E, which / isdefinable inV) should be includedamong thenaturalmodels for ZFC + BTEE. We would also like to declare that anymodel of ZFC + BTEE of the
form {V,e,j) our motivation inconsistent. Nevertheless, for viewing this type of ?> L. To see as natural also comes from our canonical example j: L we attempt tomimic a standard proof that such an why, suppose embedding implies the existence of 0#; that is, suppose we attempt to prove in a standard hence that
is natural.
Certainly
such a model
is not sharp-like
because
model
way
then exists. 0#
assume that
Ifwe restrict ourselves to just sharp-like models (so that the only natural
then, at the outset, we must result, it is unnecessarily indeed from Kunen's
assume this as part of the hypothesis. A more faithfulrepresentationof Kunen's resultwould therefore leave open the possibility thatV = L and then let theproof itself demonstrate that this isnot possible. (The fact that V ^ L falls out of theproof is easy to see: Given j: L ?> L with critical D point k, form in V theL-ultrafilter = {X e Pl{k) |k g j{X)}. Then L
to
78
PAULCORAZZA
embeddingjp: L ?> L plays a centralrole in therestof theproof). Therefore These considerations lead to the followingdefinitionof natural K^L.) model and (we believe) a faithfulrepresentationof Kunen's result in our
present context:
in V?of the ultrapower L*/D, is equal to the transitive collapse?formed for the usual reason, cannot contain D as an element (the ultrapower which,
2.14. Definition (Naturalmodels). A model (M,E, j) of ZFC + BTEE is a naturalmodel ofZFC + BTEE ifeitherj: M ? M isdefinable inV or M = V and E is the standard membership relationg. model ofZFC + BTEE of 2.15. Kunen's 0#Theorem. If thereisa natural the 0#exists. form (L, g,y), ZFC + WA seems dangerously close to being an in Although the theory
theory via Kunen's inconsistency result, as we remarked does not apply to WA because WA for j-formulas. Confidence in Part does 1, not
consistent
lihood of consistencyof ZFC + WA is furtherincreasedby the fact that,as we mentioned before, any I3 embedding i: Vx ? Vx gives rise to a model (Kg,/) ofZFC + WA.
a preliminary to the main of the theory ZFC consequences As 2.16. Wholeness WA-embedding, Lemma theorem about WA, + WA, we list some known + WA. ZFC where j: V ?> V > a. is the
in the like
v-
<
v\w
-<-'
<
Vn?y
(3) (Metatheorem)
vk < We now restate and outline Axiom Theorem viM -< vPM -< -< V Axiom Theorem. of
Wholeness
are characteristics
ZFC + WA: the theory -+ V is the WA-embedding, with (1) Assume ZFC + WA and that j: V underset forcing. (2) ZFC + WA is indestructible model ofZFC + WA, ifthereisone at all, isV The only "natural" inner (3) (4) The criticalsequence (k, ]{k),
for V, relative to e-formulas. itself. That critical point k. Then k is super-n-huge for every n\moreover, proper class of cardinals that are super-n-huge for every n. there is a
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
j: v ?+ v
79
and for any twofinite subsequences cx\< 0,2 < " < fim of the critical sequence of], V |= (j)[ah ..., am]
<
/?m].
We
a stationary subset S\ of \(k) each of whose elements is the targetof an we with criticalpoint k; then apply a suitable elementary ?-huge embedding embedding repeatedlyto Si to show that similar stationarysets existbelow Let D = {X c \(k) : g j )(X)}. D is a normal ultrafilter over j(ac). Let Si = < )(k) :a is a target some ft-huge of embeddinghaving critical {a Then Si g D since is a target an w-hugeembeddinghaving of point k}.
critical point k, as we just showed. Hence,
each km.
to prove super-fl-hugeness for every n, it suffices to show that for Now, all m, n g co, k is a-huge with km targets. We will first show that there is
1, letkn= f (k).
Si
is stationary.
* j ](Sm)*
By elementarity, Sm is a stationary subset of nm each of whose elements is a target of an w-huge embedding with critical point k. cardinals, apply Finally, to see that there is a proper class of super-^-huge theWholeness Lemma. H
statement h(rz) ispreservedby any notion of forcingthatbelongs to Vx. 3k To see this,first note that ifthenotion of forcing is an elementof VK (in P theground model V), j can be liftedto/: V[G]x ?> V[G]x in the standard P way, by definingj'(xg) = j'(x)g- Next, if the notion of forcing is an = having criticalpoint kn (namely,letk j^n+l\ as in thediscussion inPart 1 of the Wholeness Axiom Theorem, part (5)). But now, as before,k can be liftedin the standardway sinceP g VKn,and so in the extension,3kh(rv)
continues to hold. element of Vx\ VK, then for some n, P g VKn, where kn = jn(ri), and one ? can obtain another Vx definable from j and elementary embedding k: Vx
Outline of Proof of (2). The proof of (2) is like the more easily proven
/: Vx ?> Vx is an I3 embedding with critical point k, the
To prove theparallel resultfor ZFC + WA, theargumentjust givenhas to be "internalized." The hard part is, inproving indestructibility ZFC+WA of notions of forcing to VK, to show that by belonging Separation] stillholds in
the extension. To do this, boolean values of j-formula need to be defined; the
80
PAUL CORAZZA
to extending
the usual
forcing methodology
to the
H is
re
Wholeness Lemma, the critical sequence lation g. By part (1) of the = is cofinal in theordinals of M, and hence in theor / (k, j2{^),...) models always includeall theordinals). dinals of V as well (recall that inner model M, j isdefinable inV. But now thenby naturalnessof the lfV^M,
the formula defining j can be used
in particular,
(forV) to show that the range of/ must be a set inV, and thiscontradicts that thefact that iscofinal in theordinals. It followstherefore M = V. H /
Proof of (4). The key technical lemma for the proof is the following, which isproven in [4]:
Lemma.
1< m < s}). A j-class functioni,definedfromj-j,can be specified having the followingproperties: (1) i: V ?> V is an elementaryembedding; (2) crit(i) > Kn[;
(3) The < for 1 < m < s, i(?wJ = ??1+(/w_1)r. < nms and Kmi < up high enough with the Lemma ?/i2 < says that, given sequences Kmx < '* can Pusri these cardinals < K"s' one
Suppose
n\ <
n-i <
<
nm:
values agree; indiscernibility appropriate choice of i so thattheirtransformed followseasily fromthisobservation.H Theorem given Proof of (5). The proofwas given in thediscussion of this 1of thepaper. H inPart We turnto thecriticalpoint dynamics givenby aWA-embedding. Part (1) of Wholeness Axiom Theorem indicates the strength thecriticalpoint of the
of theWA-embedding. We also observe here a significant strengthening
in
the seed behavior of thistypeof mapping. We begin by showing that it isnot for a function in V, the case thatall sets in V are expressibleas / a different much strongertypeof but but thendemonstrate that j exhibits
seed behavior. 2.17. Restricted Seed Theorem For WA. Suppose]: V ? VistheWA
for any embeddinghaving criticalpoint k. Then thereis a set x such that, for j. ^ x. Moreover, k isnot a weakly universalelement functionf, j(/)(?)
Proof. because, The "moreover" clause as a functor, j is cofinal from the rest of the proposition = section). Let D (see the Preliminaries follows
?> V for which {X c k Ik g ](X)}. We firstobserve that forevery/: A = Let k g dom there is a g: k -> V such that j(/)(k)
THE AXIOM
OF
INFINITY
AND
TRANSFORMATIONS
? j :F
81
K by
= belongs toD. There Clearly, {a |f(a) g(a)} includesS and therefore = fore, j(g)(k). such To complete the proof, it sufficesto show that there is z g k -+ V} z for any g: ac > V. Let i? - {](g)M that j(g)M ^ g: \ k and r = {j(/*)(?) | //: ?> J^}. Notice thatwhenever g: k ?> V and g then {a | rank(/(a)) < k} g i). It therefore follows that
RnviMcr.
we of By inaccessibility j(/c), have It follows that some z g H does not belong toR, as required. Fj^j Although ac is not a weakly universal element for j, j exhibits a stronger
sort of seed behavior at which arises because
i^i<i^i<i^)I
natural
of the presence
of Laver
sequence,
A function /: ac ?> VK (Extendible Laver Sequences). is a weakly extendible Laver sequence at ac if for any set x, there are ordinals ? i: such that rj,C and an elementary embedding 2.18. Definition
(c) /(/)(*)=x.
are //, and i: Vn ?> such that (a) and (c) hold as well as { < (b/) rank(x) < X< rj /(?) < ?.
We will consider sequences, and more only weakly extendible as defined above, were Laver studied
is an extendible Laver
sequence
k, there
Laver context
directly generalize Laver's X forces there to be witnessing extendible embeddings the critical point are cofinal in the ordinals. See Definition 1.14. than for su
because thenotion of extendibility follows immediately from WA: one ob tains a sufficiently collection of extendible embeddings (with critical large meet the requirementsin thedefinitionof "extendible cardinal" point k) to
by considering all restrictions of j and its iterates to ranks above VK.
82
PAULCORAZZA
= with critical point k}. particular,V {/(/)?: i isan extendibleembedding Proof. Let </>(g,x) the formula be
3a [a is a cardinal ? A g: a
?? Va A V//VCV/: Vn
^i(g)(a)^x)].
ness x. Define
^ ^
f0 1x
D = {X C /c|? G j(JT)}.
D isa setby Separation forj-formulas. Define setsSi and Si by = a {a < k |/ r isweakly extendibleLaver at a}, S2 = {a<k\4>{f \a,f(a)}.
Clearly, Si U S2 G D.
to and for this, it suffices showS2 ^ D. Toward a contradiction,suppose S2 G D. Then, <f>(f, j(/)(?)) holds inK. so Let i = \\ Vn: Pick 77 that rank(x) < rj< Let x = < embeddingwith criticalpoint k. Clearly, inV, rank(x) < rj i(k) < C and = x, holds inV. Therefore contradictingthefact that0(/,j(/)(?)) /(/)(?) as required. H S2 & D,
REFERENCES
in der axiomatischen Zur Frage der Unendlichkeitsschemata Mengenlehre, [1] P. Bernays, and M.O. E. I. J. Poznanski, on thefoundations Rabin, (Y Bar-Hillel, of mathematics Essays 1961, pp. 3-49. A. Robinson, Press, Jerusalem, editors), Magnes Journal and measurable Exact functors cardinals, of Mathematics, Pacific [2] A. Biass,
To complete
the proof,
V^, where
j(^).
Because
of Separation},
/ is a set, and
is an elementary
for
extendible
and
super-almost-huge
cardinals,
The Jour
OF
AND
TRANSFORMATIONS Annals
j:
83
vol. 105 (2000), pp. 157-260. = HOD with the wholenessaxiom, Archive Mathematical Consistency ofV [5]-, for Logic, vol. 39 (2000), pp. 219-226. vol. 179 (2003), pp. 43-60. The spectrumof elementaryembeddings j [7]-, Applied Logic, vol. 139 (2006), pp. 327-399.
The wholeness axiom. Consciousness-based [8]-, and learning in the academic (P. Corazza, ing disciplines [6]-, The gap between h and the wholeness axiom, Fundamenta Mathematicae,
[4]-,
laver sequences,
of Pure
Logic,
editor),
[10]P. Dehornoy,
f r/~ dehornoy/surveys. at http: //www. math.unicaen. html.). and A. H. Mekler, set-theoretic methods, modules, [11] P. C Eklof Almost-free Holland Mathematical B. V., Amsterdam, Science Publishers Library, vol. 46, Elsevier [12] S. Feferman, this Bulletin, axioms'], H. Friedman, vol. 6 (2000), P. Maddy, and J. Steel, Does mathematics no. 4, pp. 401-433.
of wholeness,
[13]P. Freyd, Aspects of topoi,Bulletin of theAustralianMathematical Society, vol. 7 (1972), pp. 1-76. Mathematical Logic, vol. 28 (1989), pp. 35-42. of Hajnal, Archives of 1988.
[15] M. Hallett, Cantorian set theory and the limitation of size, Clarendon [14] M. Gitik and S. Shelah, On certain indestructibility of strong cardinals and a question Press, Oxford,
[16] J.D. Hamkins, Canonical seeds and Prikry trees,The Journal of Symbolic Logic, vol. 62 (1997), no. 2, pp. 373-396. vol. 40 (2001), pp. 1-8.
[18] F. Hausdorff, Grundzuge einer Theorie [17]-, The wholeness axioms and V ? HOD, Archive for Mathematical Mathematische Logic, An
nalen, vol. 65 (1908), pp. 435-505. Berlin, 2003. [19]T. Jech,Set theory, Springer-Verlag, Berlin, 1994. [20]A. Kanamori, The higher infinite, Springer-Verlag, The Journal ofSym combinatorics, [22]K. Kunen, Elementary embeddingsand infinitary bolic Logic, vol. 36 (1971), pp. 407-413. Saturated ideals,The Journal ofSymbolic Logic, (1978), pp. 65-76. [23]-,
forcing, [21]-, Godel and set theory, this Bulletin, vol. 13 (2007), no. 2, pp. 153-188.
der geordneten
Mengen,
[25]F.W. Lawvere, Adjointness in foundations,Dialectica, vol. 23 (1969), pp. 281-296. New York, mathematician, Springer-Verlag, [26] S.Mac Lane, Categoriesfor theworking 1971. [27]P. Maddy, Believing theaxioms, I, The Journal of Symbolic Logic, vol. 53 (1988), [28]-, Believing theaxioms, II, The Journal ofSymbolic Logic, vol. 53 (1988), no. 3,
of k
under
^-directed
closed
(1978),
385-388.
liber lineare transfinite [29]P. Mahlo, Mengen, Berichte tiber die Verhandlungen der Koniglich Sachsischen Gesellschaft derWissenschaften zu Leipzig,Mathematische-Physische Klasse, vol. 63 (1911), pp. 187-225. Math [30]D. A. Martin, Infinitegames, Proceedings of the InternationalCongress of ematicians,Helsinki, (1978) (Olli Lehto, editor), vol. 1,Academia Scientiarum Fennica,
pp.
736-764.
PAULCORAZZA
and
Society,
II, Commentationes
DEPARTMENT OF MATHEMATICS AND COMPUTER SCIENCE MAHARISHI UNIVERSITY OF MANAGEMENT USA FAIRFIELD, IA, 52557,
E-mail: paul_corazza@yahoo.com