Forty Years On His - Godel - Shoulders by Harvey M. Friedman
Forty Years On His - Godel - Shoulders by Harvey M. Friedman
Forty Years On His - Godel - Shoulders by Harvey M. Friedman
Harvey M. Friedman*
Ohio State University
Gödel Centenary
Delivered: April 29, 2006
Revised November 5, 2008
Revised August 10, 2009
Revised November 19, 2009
1. General Remarks.
5. Lengths of Proofs.
8. Wqo Theory.
9. Borel Selection.
We wish to thank Warren Goldfarb and Hilary Putnam for help with
1. GENERAL REMARKS
philosophical inquiry.
particularly in section 12, we believe that the potential impact of Gödel's work
this potential impact will have to wait for some new breakthroughs. We have
Gödel's legacy is most decisively overwhelming. However, there are some signs
that some of our most distinguished mathematicians recognize the need for some
We will not attempt to properly discuss the full impact of Gödel's work
and all of the ongoing important research programs that it suggests. This would
require a book length manuscript. Indeed, there are several books discussing the
Gödel legacy from many points of view, including, for example, (Wang 1987,
1996), (Dawson 2005), and the historically comprehensive five volume set (Gödel
1986-2003).
In sections 2-7 we briefly discuss some research projects that are suggested
The incompleteness phenomena lie at the heart of the Gödel legacy. Some
One particular issue that arises with regard to incompleteness has been a
driving force for a considerable portion of my work over the last forty years. This
has been the ongoing search for necessary uses of set theoretic methods in
normal mathematics.
of an unprovable sentence.
semiformal argument, that the sentence he constructs for his first incompleteness
4
consistency statement is not provable. It was later established that the two are in
by the general mathematical community, which naturally resists the notion that
the incompleteness theorem will have practical consequences for their own
research.
theoretic systems had to wait until the well known work on the axiom of choice
and the continuum hypothesis by Kurt Gödel and Paul Cohen. See (Gödel 1940),
(Cohen 1963-64).
piecewise continuous).
second look to see its overtly set theoretic character, is Kaplansky’s Conjecture
5
asserts that
of reals converging to 0 (under the sup norm) to any separable Banach algebra, is
continuous.
to H.G. Dales and J. Esterle), and later shown to be not refutable without the
continuum hypothesis; i.e., not refutable in the usual ZFC axioms (due to R.
Solovay). See (Dales 2001) for the refutation, and (Dales, Woodin 1987) for the
It is, of course, much easier for mathematicians to recognize the overtly set
theoretic character after they learn that there are set theoretic difficulties. By
It is clear that one is asking about the existence of an object that was well
known, even at the time, to necessarily have rather pathological properties. This
is the case even for discontinuous group homomorphisms from ℜ into ℜ (which
can be shown to exist without the continuum hypothesis). For instance, it is well
use fluff - as long as it doesn’t cause any trouble). Functions and sets are
normally Borel measurable within such so called Polish spaces. In fact, the sets
barrier into the relatively concrete world of Borel measurability - and even into
the countable and the finite world - with independence results of a mathematical
character.
Yet it must be said that our results to date are very limited in scope, and
1
Apparently, nonseparable arguments are being used in the proofs of certain
suggesting strongly that this is an area where logicians and number theorists
nonseparable arguments are. We have conjectured that they are not, and that
In particular, it is not yet clear how strongly and in what way the Gödel
Progress along these lines is steady but painfully slow. We are confident that a
much clearer assessment will be possible by the end of this century - and perhaps
In section 12, we take the opportunity to speculate far into the future.
completeness theorem for a standard version of the axioms and rules of first
1986-2003 44-59). These Introductory notes were written by Burton Dreben and
On page 52, the following passage from a letter from Gödel to Hao Wang,
consequence of Skolem 1923a. However, the fact is that, at the time, nobody
(including Skolem himself) drew this conclusion (neither from Skolem 1923a nor,
summarized as follows:
8
“provable” by a formal one ... and the explicit recognition that there is a question
to be answered.”2
“valid” or “valid in all set theoretic structures”. The appropriate fully formal
surprisingly the first clear statement in Tarski’s work of the formal semantics for
predicate calculus did not appear until (Tarski 1952) and (Tarski, Vaught 1957).
Let us return to the fundamental setup for the completeness theorem. The
of the notion of structure need be given. The proof requires only that we at least
admit the structures whose domain is an initial segment of the natural numbers
(finite or infinite). In fact, we need only admit structures whose relations and
functions are arithmetically defined; i.e., first order defined in the ring of
integers.
important to interpret logic with structures that may lie outside the realm of
2
Skolem 1923a above is (Skolem 1922) in our list of references.
9
demands that quantifiers range over absolutely everything. From this viewpoint,
everything.
The topic of logic in the universal domain has been taken up in the
We have not yet published on this topic, but unpublished reports on our
(Friedman 2002a 65-99). We plan to publish a monograph on this topic in the not
is proved there in detail for a specific variant of what is now known as the simple
theory of types (going back to Bertrand Russell), with natural numbers at the
lowest type. This is a rather strong system, nearly as strong as Zermelo set
theory.
10
this system.
(Gödel 1934) gives another treatment of the results in [Gödel 31], but also,
At the end of (Gödel 1931 195), Gödel writes that “The results will be
fact that due to A.M Turing’s work a precise and unquestionably adequate
definition of the general notion of formal system can now be given, a completely
general version of Theorems VI and XI is now possible. That is, it can be proved
rigorously that in every consistent formal system that contains a certain amount
that, moreover, the consistency of any such system cannot be proved in the
The sequel was never published at least partly because of the prompt
shown there that the hypothesis of 1-consistency can be replaced with the weaker
hypothesis of consistency.
Later, methods from recursion theory were used to prove yet more
general forms of first incompleteness, and where the proof avoids use of formal
self reference - although even in the recursion theory, there is, arguably, a trace
1952), and (Tarski, Mostowski, Robinson 1953), with the use of the formal system
Q.
axioms and rules of logic for this language, we have the nonlogical axioms
1. Sx ≠ 0.
2. Sx = Sy → x = y.
3. x ≠ 0 → (∃y)(x = Sy).
4. x + 0 = x.
5. x + Sy = S(x + y).
6. x • 0 = 0.
7. x • Sy = (x • y) + x.
8. x ≤ y ↔ (∃z)(z + x = y).
The last axiom is purely definitional, and is not needed for present
recursive.
For the proof, see (Robinson 1952), and (Tarski, Mostowski, Robinson
immediate Corollary.
calculus whose relational type and axioms are recursively enumerable. There is
obtain other forms of first incompleteness that are stronger in certain respects. In
fact, Hilbert’s tenth problem is still a great source of very difficult problems on
the border between logic and number theory, which we will discuss below.
commonly referred to as the MRDP theorem (in reverse historical order). See
13
(Davis 1973), (Matiyasevich 1993). The MRDP theorem was shown to be provable
Gaifman 1982).
calculus whose relational type and axioms are recursively enumerable. There is a
refutable in T.
(∃x1,...,xn)(s = t) that can be used in Theorem 3.3 for, say, T = PA or T = ZFC. Note
integers?”
Let us see what can be done on the purely recursion theoretic side with
integers suffices for recursive unsolvability. One form of the result (not the
14
strongest form) says that the problem of deciding whether or not a polynomial
Also (Jones 1982) proves that, e.g., the problem of deciding whether or not
1982), it is shown that degree 4 and 58 nonnegative integer unknowns suffice for
unsolvable.
<1.6 × 1045,9>.
nonnegative integers, the integers, and the rationals), going back to (Siegel 72).
For degree 3, the existence of an algorithm is wide open, even for three
variables (over the integers, the nonnegative integers, or the rationals). For
degree 3 in two integer variables, an algorithm is known, but it is wide open for
It is clear from this discussion that the gap between what is known and
what could be the case is enormous, just in this original context of deciding
number theory and recursion theory can be done in EFA. Although one obtains
upper bounds on pairs (number of variables, degree) in this way, this does not
address the question of the size of the coefficients needed in Theorem 3.3.
neither provable nor refutable in PA. (We can also use ZFC here instead of PA).
relationship between the “size” of a Gödel polynomial using PA and the “size”
theorem, after proving his first incompleteness theorem in detail. His sketch
depends on the fact that the proof of the first incompleteness theorem, which is
Gödel promised a part 2 of (Gödel 1931), but this never appeared. There is
In any case, the necessary details were carried out in (Hilbert, Bernays
1934,1939), and later in (Feferman 1960), and most recently, in (Boolos 1993).
We take the liberty of presenting our own particularly careful and clear
equality, with infinitely many constant, relation, and function symbols. For
i) variables xn, n ≥ 1;
v) connectives ¬,∧,∨,→,↔;
symbols.
substitution”.
statement”.
7. A distinguished formula PROV with at most the free variable x1, expressing
“provable in T”.
Here #(A) is the Gödel number of the formula A, as a closed term of RT.
18
THEOREM 4.1. (Self reference lemma). Let A be a formula of RT. There exists a
Note that
SSUB(#(B)) = #(B[x1/#(B)])
is provable in T. Hence
SSUB(s) = #(A[x1/SSUB(s)])
LEMMA 4.2. (“I am not provable” Lemma). There exists a closed term t such that
T proves t = #(¬PROV[x1/t]).
PROV[x1/#(¬PROV[x1/t])]
QED
PROV[x1/NEG(PR(t))].
PROV[x1/#(A)] → PROV[x1/PR(#(A))]
PROV[x1/t] → PROV[x1/PR(t)]
PR(t) = #(PROV[x1/t]).
By condition 8, T proves
NEG(#(PROV[x1/t])) = #(¬PROV[x1/t]) .
NEG(PR(t)) = t.
(∀x1)(¬(PROV ∧ PROV[x1/NEG(x1)])).
¬(PROV[x1/PR(t)] ∧ PROV[x1/NEG(PR(t))]).
Incompleteness are complicated and awkward. Even the abstract form of second
20
incompleteness given above using derivability conditions are rather subtle and
involved.
present new versions of Formal Second Incompleteness that are simple, and
Second Incompleteness.
Arithmetic.
We write L(prim) for the language based on 0,S and all primitive recursive
the language L(prim). I.e., the nonlogical axioms of PA(prim) consist of the
axioms for successor, primitive recursive defining equations, and the induction
that the induction scheme be applied only to quantifier free formulas of L(prim).
21
L(prim) such that every equation in L(prim) that is provable in PA(prim), is also
semiformal way from the above formal second incompleteness for PA(prim).
most n quantifiers.
The above development can be appropriately carried out for systems with
full induction. However, there is a more general treatment which covers finitely
We use the system EFA = exponential arithmetic for this more general
are bounded to terms. This is the same as the system IΣ0(exp) in (Hajek, Pudlak
1993 p. 37).
provable in EFA + A. Then EFA proves the consistency of every finite fragment
of T.
ϕ in T which leaves the meaning of all symbols unchanged, but where the
domain is allowed to consist of only some of the nonnegative integers from the
point of view of T.
that one is relying on a formalization of Con(T) within T via the indirect method
(and the relevant syntactic objects) ad hoc, but one is still being indirect and not
directly dealing with the objects at hand - which are syntactic and not numerical.
of any indirection. Thus in such an approach, one would add new sorts for the
relevant syntactic objects, and introduce the various relevant relations and
23
function symbols, together with the relevant axioms. Precisely this approach was
The first choice is to make sure that as one adds new sorts and new
relevant relations and function symbols and new axioms to T, associated with
syntax, one also somehow has already appropriately treated, directly, the new
syntactic objects and axioms beyond T that arise when one is performing this
addition to T.
The second choice is to be content with adding the new sorts and new
relevant relations and function symbols and new axioms to T, associated with the
syntax of T only - and not try to deal in this manner with the extended syntax
that arises from this very process. This is the choice made in (Quine, 1940, 1951,
Chapter 7).
We lean towards the opinion that the first choice is impossible to realize in
an appropriate way. Some level of indirection will remain. Perhaps the level of
indirection can be made rather weak and subtle. Thus we lean towards the
opinion that it is impossible to construct extensions of, say, PA that directly and
adequately formalize their entire syntax. We have not tried to prove such an
In any case, the second choice, upon reflection, turns out to be wholly
adequate for casting what may be called “direct second incompleteness”. This
formulation asserts that for any suitable theory T, if T’ is the (or any) extension of
We can recover the usual second incompleteness theorem for T from the
Thus under this view of second incompleteness, one does not view Con(T)
interpretations, all of which are ad hoc. This view would then eliminate ad hoc
foundational implications.
5. LENGTHS OF PROOFS.
asserts the following. Let RTT be Russell’s simple theory of types with the axiom
of infinity. Let RTTn be the fragment of RTT using only the first n types. Let f:N
such that
f(n) < m
where n is the least Gödel number of a proof of ϕ in RTTn+1 and m is the least
Gödel expressed the result in terms of lengths of proofs rather than Gödel
numbers or total number of symbols. Gödel did not publish any proofs of this
remarks by R. Parikh, it is likely that Gödel had inadvertently used lengths, and
In any case, the analogous result with Gödel numbers was proved in
1971) and (Parikh 1971). Also see (Parikh 1973) for results going in the opposite
positive integer n, the finite consistency statement Conn(T) expressing that “every
large. A more careful version of the argument gives the lower bound of n1/2 for
several directions. One direction is to give a treatment of a good lower bound for
launching point. A number of issues arise as to the best way to set this up, and
versions which are not asymptotic. I.e., they involve specific numbers of symbols
Although the very good upper bound of O(n) is given in (Pudlak 1985) for
significantly weaker than T. For specificity, consider how many symbols it takes
no subexponential upper bound here, we will have to refute this strong version
There are some other aspects of lengths of proofs that seem important.
What is not at all clear here is whether c,d can be made reasonably small.
There is clearly a lot of overhead involved on two counts. One is in the execution
The other overhead is that one must insert the proofs of various facts about the
The same remarks can be made with regard to NBG + GC + CH and NBG
+ GC, where GC is the global axiom of choice. Also, these remarks apply to ZFC
and ZF, and also to ZFC + CH and ZFC. Also they apply equally well to the
Cohen forcing method (Cohen 1963-1964), and proofs from ZF + ¬AxC, and from
ZFC + ¬CH.
axioms for real closed fields using quantifier elimination. This also provides a
decision procedure for recognizing the first order sentences in (ℜ,<,0,1,+,-,•). His
1) The language is 0,1,+,-,•. The axioms consist of the usual field axioms, together
with -1 is not the sum of squares, x or -x is a square, and every polynomial of odd
2) The language is 0,1,+,-,•,<. The axioms consist of the usual ordered field
axioms, together with every positive element has a square root, and every
3) The language is 0,1,+,-,•,<. The axioms consist of the usual ordered field
axioms, together with the axiom scheme asserting that if a first order property
holds of something, and there is an upper bound to what it holds of, then there is
1982a-c), (Basu, Pollack, Roy 2006). In terms of computational complexity, the set
28
nondeterministic exponential time hard. The gap has not been filled. Even the
first order theory of (ℜ,+) is nondeterministic exponential time hard. See Rabin
1977.
The work just cited concerns mainly the computational complexity of the
set of true sentences in the reals (sometimes with only addition). It does not
We conjecture that with the usual axioms and rules of logic, in all three cases,
there is a double exponential lower and upper bound on the number of symbols
many true sentences without < such that there is a double exponential reduction
system 3).
the quantifier structure of the sentence is restricted. For instance, the cases of
the matrix is particularly simple. Other cases of clear interest are ∀...∀∃...∃, and
sentences. The earliest ones were presented in (Goodstein 1944) and (Paris,
29
Harrington 1977), and are proved just beyond PA. We discovered many
Robertson, Seymour (Roberton, Seymour 1985, 2004), which are far stronger,
The basic idea is this. There are now a number of mathematically natural
remove the asymptotics. I.e., in many cases, we have verified that we can fix n to
be very small (numbers like 3 or 9 or 15), and consider the resulting Σ01 sentence
(∃m)(R(n,m)). The result is that any proof in T (or certain strong fragments of T)
of this Σ01 sentence must have an absurd number of symbols - e.g., an exponential
stack of 100 2’s. Yet if we go a little beyond T, we can prove the full Π02 sentence
proof just beyond T of the resulting Σ01 sentence R(n,m) with n fixed to be a small
3
See http://cs.nyu.edu/pipermail/fom/
30
on intuitionistic logic: (Gödel 1932a), (Gödel 1933a), (Gödel 1933b), (Gödel 1958).
cannot be viewed as a classical system with finitely many truth values. He shows
calculus. For more on intermediate logics, see (Hosoi, Ono 1973) and (Minari
1983).
logic. It can be axiomatized by taking the usual axioms and rules of intuitionistic
predicate logic, together with the axioms of PA as usual given. Of course, one
must be careful to present ordinary induction in the usual way, and not use the
propositional calculus.
predicate calculus.
⊥ as ⊥.
31
¬ as ¬.
∧ as ∧.
→ as →.
∨ as ¬¬∨.
∀ as ∀.
∃ as ¬¬∃.
successor axioms and the defining equations of PA are sent to theorems of HA,
(∀n)(F(n) = 0).
HA.
systems, most of them of the form T,T’, where T,T’ have the same nonlogical
on intuitionistic predicate calculus. For example, see (Kreisel 68a 344), (Kreisel
A much stronger result holds for PA over HA. Every Π02 sentence
provable in PA is provable in HA. The first proofs of this result were from the
proof theory of PA via Gentzen (see (Gentzen 1969), (Schütte 1977)), and from
1972).
However, for other pairs for which the negative interpretation shows that
they have the same provable Π01 sentences - say classical and intuitionistic
second order arithmetic - one does not have the required proof theory. In this
case, the Dialectica interpretation has been extended by Spector in (Spector 1962),
and the fact that these two systems have the same provable Π02 sentences then
follows.
Nevertheless, there are many appropriate pairs for which the negative
interpretation.
interpretation via what is now called the A translation. Also see (Dragalin 1980).
We illustrate the technique for PA over HA, formulated with primitive recursive
function symbols.
proves
(∃n)(F(n,m) = 0.
(∃n)(F(n,m) = 0).
intuitionistic logic. It can be axiomatized by taking the usual axioms and rules of
course, one must be careful to present ordinary induction in the usual way, and
finite type that is based on quantifier free axioms and rules, including a rule of
induction.
discusses this aspect in both papers, especially the second. The idea is that the
quantifiers in HA or PA, ranging over all natural numbers, are not finitary,
However, the objects of T are at least prima facie infinitary, and so there is the
difficult question of how to gauge this tradeoff. One idea is that the objects of T
refer the interested reader to the rather extensive Introductory notes to (Gödel
order system known as second order arithmetic, or Z2. This was carried out by
Clifford Spector in (Spector 1962). Here the idea is that one may construe such a
was not entirely satisfied that the quantifier free system Spector used was truly
constructive.
35
We believe that the Spector development has not been fully exploited. In
obtain sharper uniformities in certain areas of functional analysis that had been
Kohlenbach. See the five references to Kohlenbach (and joint authors) in the list
of references.
HYPOTHESIS.
hypothesis: Two abstracts, (Gödel 1938), (Gödel 1939a). One paper with sketches
of proofs, (Gödel 1939b). One research monograph with fully detailed proofs,
The normal abbreviations for the axiom of choice is AxC. The normal
AxC. However, note that in this form, CH can be naturally considered without
the presence of AxC. However, Solovay’s model satisfying ZFCD + “all sets are
36
Lebesgue measurable” also satisfies CH in the strong form that every set of reals
is countable or has a perfect subset (this strong form is incompatible with AxC).
Hausdorff. The GCH asserts that for all sets A, every subset of ℘(A) is either in
inconsistency in ZF.
constructible sets, is rather large, in the sense that when fully formalized, results
mechanisms are essential for the actual conduct of mathematics. In fact, current
proof assistants - where humans and computers interact to create verified proofs
¬CH, thus complementing Gödel’s results. See (Cohen 1963-1964). The proof
certain conditions (met here), if the consistency of every given finite subsystem of
one system is provable in another, then the first system is interpretable in the
of the interpretation? Again this is far from clear. And how does this question
primitive notation ∈,=, is decided in a weak fragment of ZF. See (Gogol 1979),
(Friedman 2003a). Also there is a 5 quantifier sentence in ∈,= that is not decided
in ZFC (it is equivalent to the existence of a subtle cardinal over ZFC). See
(Friedman 2003b). It is also known that AxC can be written with five quantifiers
The question is: how many quantifiers are needed to express CH over
ZFC, in ∈,=? We can also ask this and related questions where abbreviations are
allowed.
provable nor refutable from the standard axioms for mathematics (ZFC), the
ultimate status of CH has been settled and there is nothing left to ponder.
38
take a quite different view. This includes Kurt Gödel. They take the view that the
truth value. The problem is to determine just what this truth value is.
The idea here is that there is a definite system of objects that exists
independently of human minds, and that human minds can no more manipulate
the truth value of statements of set theory than they can manipulate the truth
This is the so called Platonist point of view that is argued so forcefully and
The late P.J. Cohen led a panel discussion at the Gödel Centenary called
question he asked was, roughly, “does the continuum hypothesis have a definite
The response from the audience appeared quite divided on the issue.
Of the panelists, the ones who have expressed very clear views on this
topic were most notably Cohen and Woodin. Cohen took a formalist viewpoint,
whereas Woodin takes a Platonist one. See their respective contributions to this
volume.
Platonist viewpoint - or, for that matter, what mixture of the two is true or
appropriate.
My ideas are not very well developed, but I will offer at least something
formalism/Platonism in anything like its present terms. What may come out is a
fundamental mental picture for the axioms of ZFC, even with some large
cardinals, along with a theorem to the effect that there is no fundamental mental
8. WQO THEORY.
quasi order) is a qo (A,≤) such that for all infinite sequences x1,x2,... from A, ∃ i < j
The highlights of wqo theory are that certain qo’s are wqo’s, and certain
(Kruskal 1960), treats finite trees as finite posets, and studies the qo
THEOREM 8.1. (Kruskal 1960). The above qo of finite trees as posets is a wqo.
We observed that the connection between wqo’s and well orderings can
Weyl, and others railed against impredicative mathematics. See (Weyl 1910),
FS.
J.B. Kruskal actually considered finite trees whose vertices are labeled
label(h(v)).
THEOREM 8.3. (Kruskal 1960). The qo of finite trees as posets, with vertices
labeled KT.
THEOREM 8.4. Labeled KT does not hold in the hyperarithmetic sets. In fact,
in preparation).
41
vertices of Ti.
sequence of finite trees. ∃ i < j such that Ti is inf preserving embeddable into Tj.
a given complexity class. There exists i < j such that Ti is inf preserving
THEOREM 8.7. Corollary 8.5 cannot be proved in FS. This holds even for linear
THEOREM 8.8. Corollary 8.6 cannot be proved in FS, even for linear time,
logarithmic space.
strong uniformities.
bounded sequence of finite trees. There exists i < j ≤ n such that Ti is inf
preserving embeddable into Tj, where n depends only on the given linear bound,
With this kind of strong uniformity, we can obviously strip Theorem 8.9 of
infinite sequences of trees. Using the linear bounds n+k, k fixed, we obtain:
42
THEOREM 8.10. (finite KT). Let n >> k. For all finite trees T1,...,Tn with each |Ti|
≤ i+k, there exists i < j such that Ti is inf preserving embeddable into Tj.
Since Theorem 8.10 → Theorem 8.9 → Corollary 8.5 (using bounds n+k,
involves a finite label set and a gap embedding condition. Only here the strength
We said that the gap condition was natural (i.e., EKT was natural). Many
Soon later, EKT became a tool in the proof of the well known graph minor
THEOREM 8.11. Let G1,G2,... be finite graphs. There exists i < j such that Gi is
knew implied full EKT, just from GMT. They complied, and we wrote the triple
The upshot is that GMT is not provable in Π11-CA0. Just where GMT is
provable is unclear, and recent discussions with Robertson have not stabilized.
can be proved.
43
theorem. A subcubic graph is a graph where every vertex has valence ≤ 3. (Loops
THEOREM 8.12. Let G1,G2,... be subcubic graphs. There exists i < j such that Gi is
Robertson, Seymour 1987)). Therefore the subcubic graph theorem (even in the
from {1,2,3}, where each |Ti| ≤ i. There exists i < j such that Ti is inf and label
each |Ti| ≤ i+13. There exists i < j such that Gi is homeomorphically embeddable
into Gj.
THEOREM 8.13. Every proof of *) in FS uses at least 2[1000] symbols. Every proof
9. BOREL SELECTION.
graph is contained in S.
44
THEOREM 9.1. Let S ⊆ ℜ2 be a symmetric Borel set. Then S or ℜ2\S has a Borel
selection.
determinacy, due to D.A. Martin. See (Martin 1975), (Martin 1985), and (Kechris
1994 137-148).
THEOREM 9.2. (Friedman 1981). Theorem 9.1 is provable in ZFC, but not
Debs and Saint Raymond of Paris VII. They take the general form: if there is a
nice selection for S on compact subsets of E, then there is a nice selection for S on
E. See the five papers of Debs and Saint Raymond in the references.
selection for S on every compact subset of E, then there is a Borel selection for S
on E.
45
THEOREM 9.5. (Friedman 2005). Theorem 9.3 is provable in ZFC but not without
the axiom scheme of replacement. Theorem 9.4 is neither provable nor refutable
in ZFC.
through every countable ordinal is sufficient to prove Theorems 9.1 and 9.3.
by some g:N → N.
THEOREM 9.7. ZFC + Theorem 9.4 implies DOM (Friedman 2005). ZFC + DOM
The principal reference for this section is the forthcoming book Friedman
2010.
We begin with two examples of statements in BRT of special importance for the
theory.
THIN SET THEOREM. Let k ≥ 1 and f:Nk → N. There exists an infinite set A ⊆ N
x ∈ Nk, f(x) > max(x). There exists an infinite set A ⊆ N such that f[Ak] = N\A.
46
THIN SET THEOREM. For all f ∈ MF there exists A ∈ INF such that fA ≠ N.
fA = N\A.
The thin set theorem lives in IBRT in A,fA. There are only 22^2 = 16
The complementation theorem lives in EBRT in A,fA. There are only 22^2 =
statements. This is entirely unmanageable. It would take several major new ideas
independent of ZFC. It can be proved in SMAH+ but not in SMAH, even with
The particular example is far nicer than any “typical” statement in EBRT
suitably natural.
47
too hard.
We also look for a bonus: a striking feature of the classification that is itself
ZFC.
In order to carry this off, we need to use the function class ELG of
These are functions f:Nk → N such that there exist constants c,d > 1 such
that
TEMPLATE. For all f,g ∈ ELG there exist A,B,C ∈ INF such that
X ∪. fY ⊆ V ∪. gW
P ∪. fR ⊆ S ∪. gT.
Note that there are 6561 such statements. We have shown that all of these
PROPOSITION A. For all f,g ∈ ELG there exist A,B,C ∈ INF such that
A ∪. fA ⊆ C ∪. gB
A ∪. fB ⊆ C ∪. gC.
ACA’.
Inspection shows that all of the non exotic cases come out with the same
truth value in the two classifications, and that is of course provable in RCA0.
Furthermore, the exotic case comes out true in the second classification.
THEOREM 10.1. The following is provable in SMAH+ but not in SMAH, even
with the axiom of constructibility. An instance of the Template holds if and only
We caution the reader that this is intensively ongoing research, which has
max(x) < max(y). We say that E ⊆ Qk is order invariant if and only if membership
49
We write SDOI(Qk) for the family of all strictly dominating order invariant
R ⊆ Qk × Qk.
For A ⊆ Qk, write cube(A,0) for the least set Vk such that A ⊆ Vk ∧ 0 ∈ V.
Con(SUB), in WKL0.
Here SUB+ = ZFC + "for all k there exists a k-subtle cardinal". SUB = ZFC
+ {there exists a k-subtle cardinal}k. For the definition of k-subtle cardinal, due to
J. Baumgartner, see Friedman 2001. WKL0 is one of the principal five systems of
SDOI(Qk), there exist finite A1,A2,... ⊆ Qk such that for all i ≥ 1, Ai ∪ us(Ai) ⊆ Ai+1
= cube(Ai+1,0).
50
all R ∈ SDOI(Qk), there exist finite A1,...,Ak ⊆ Qk such that for all 1 ≤ i ≤ k-2, Ai ∪
SDOI(Qk), there exist finite A1,...,Ak ⊆ Qk such that for all 1 ≤ i ≤ k-2, Ai ∪ us(Ai) ⊆
Note that the second of these is explicitly Π02 and the third of these is
explicitly Π01.
THEOREM 11.2. WKL0 proves that all four Upper Shift Fixed Point
Sequential Upper Shift Fixed Point Propositions and the Estimated Upper Shift
come a long way. The same is true of the related phenomenon of recursive
unsolvability, also part of the Gödel legacy. The phenomena is so deep, and rich
in possibilities, that we expect the future to eclipse the past and present.
the Gödel legacy (as is common today) on the basis of his letter of March 20,
1956, to John von Neumann (see (Gödel 1986-2003 Vol. V, letter 21, 373-377).
51
conjecture that we will understand this in the future, and that we will find,
solution to Hilbert’s tenth problem, replete with new deep ideas. This would
result in a sharp increase in the level of interest for the working mathematician
that this will be answered in the negative, and that the solution will involve some
iii. Consistency of the AxC. Consistency of the most basic, and once
controversial, early candidate for a new axiom of set theory. (Gödel 1940).
52
iv. Consistency of the CH. Consistency of the most basic set theoretic
higher type primitive recursion, without quantifiers. (Gödel 1958), (Gödel 1972).
Starting soon after (Cohen 1963-1964), starting dramatically with R.M. Solovay
(e.g., his work on Lebesgue measurability (Solovay 1970), and his independence
many others. See the rather comprehensive (Jech 2006). Also see the many set
used to prove statements in and around Borel mathematics. See (Friedman 1971),
and some Borel selection theorems of Debs and Saint Raymond (see section 9
above).
Includes some Borel selection theorems of Debs and Saint Raymond (see section
1944), (Paris, Harrington 1977), and, most recently, with (Friedman 2002b), and
(Buchholz, Feferman, Pohlers, Seig 1981). Includes Kruskal’s theorem, the graph
minor theorem of Robertson, Seymour (Robertson, Seymour, 1985, 2004), and the
xv. Large cardinals necessarily used to prove explicitly Π01 sentences. See
the creative (and high quality) study of U.S. tax law dwarfs the effort devoted to
Through my efforts over 40 years, I can see, touch, and feel a certain
mathematically.
But I don’t quite have the right way to express it. I likely need some richer
context than the completely primitive combinatorial settings that I currently use.
This difficulty will definitely be overcome in the future, and that will make a
huge difference in the quality, force, and relevance of the results to mathematical
practice.
everywhere.
decided using well studied extensions of ZFC, but not all of which
This program has been carried out, to some very limited extent, by BRT –
This may seem like a ridiculously ambitious conjecture, which goes totally
quite new. Let’s say 100 years old – although even that is a stretch.
Assuming the human race thrives, what is this compared to, say, 1000
time. A more reasonable number is 1M years. And what does our present
nothing in astronomical time. This Sun has several billion good years left
Mathematics in 1B years time? Who can know what that will be like. But I
am convinced that the Gödel legacy will remain very much alive – at least as
REFERENCES
Atiyah, M. 2008a. Mind, Matter and Mathematics, The Royal Society of Edinburgh,
2008/presidential_address.pdf
http://www.dailystar.com.lb/article.asp?edition_id=1&categ_id=2&article_id=9
7190
11:257-284, 2003.
Basu, S., R. Pollack, R., M. Roy, M. 2006. Algorithms in Real Algebraic Geometry. Springer
Verlag. http://www.math.gatech.edu/~saugata/index.html#books_authored
Press.
Buchholz, W., S. Feferman, S., Pohlers, W., Seig, W. 1981. Iterated Inductive Definitions
Cohen, P.J. 1963-64. “The independence of the continuum hypothesis”. Proc. Nat. Acad.
Cohen, P.J. 1969. “Decision procedures for real and p-adic fields”. Comm. Pure Appl.
Dales, H.G. 2001. Banach algebras and automatic continuity. London Mathematical Society
57
Dales, H.G., and W.H. Woodin, W.H. 1987. An introduction to independence results for
analysts. London Mathematical Society Lecture Note Series, vol. 115, Cambridge,
Dawson, J. 2005. Logical Dilemmas: The Life and Work of Kurt Gödel. Wellesley,
Debs, G. and Saint Raymond, J. 1996. “Compact covering and game determinacy”.
Debs, G., and Saint Raymond, J. 1999. “Cofinal Σ11 and Π11 subsets of NN”. Fund. Math.
159, 161-193.
Debs, G., and Saint Raymond, J. 2001. “Compact covering mappings and cofinal
Debs, G., and Saint Raymond, J. 2004. “Compact covering mappings between Borel
sets and the size of constructible reals”. Transactions of the American Mathematical
Debs, G., and Saint Raymond, J. 2007. Borel Liftings of Borel Sets: Some Decidable and
the MRDP theorem”. In: Logic and Algorithmic, An International Symposium Held in
Honour of Ernst Specker, Monogr. Ensign. Math. Univ. Geneve, 1982, 317-329.
Dragalin, A. 1980. “New forms of realizability and Markov’s rule”. Russian: Doklady,
Feferman, S. 1998. In the Light of Logic. Logic and Computation in Philosophy. Oxford:
http://math.stanford.edu/~feferman/papers.html
Friedman, H. 1971. “Higher set theory and mathematical practice”. Ann. Math. Logic, 2,
325-357.
Friedman, H. 1973. “The consistency of classical set theory relative to a set theory with
G.H. Müller and D.S. Scott (eds.), Higher Set Theory, Lecture Notes in Mathematics,
Friedman, H. 1981. “On the Necessary Use of Abstract Set Theory”. Advances in Math.,
59
Friedman, H. 1998. “Finite Functions and the Necessary Use of Large Cardinals”,
state.edu/%7Efriedman/
Friedman, H. 2001. "Subtle Cardinals and Linear Orderings". Annals of Pure and
http://www.math.ohio-state.edu/%7Efriedman/
Mathematics: Essays in honor of Solomon Feferman, ed. Sieg, Sommer, Talcott, Lecture
Notes in Logic, volume 15, 62-93. Providence, Rhode Island: Association for
Symbolic Logic.
240.
Friedman, H. 2005. “Selection for Borel relations”. In: Logic Colloquium ’01, ed. J.
Friedman H. 2006a. “271: Clarification of Smith article”, FOM Archives, March 22, 2006,
http://www.cs.nyu.edu/pipermail/fom/2006-March/010244.html
http://www.cs.nyu.edu/pipermail/fom/2006-March/010260.html
60
http://www.cs.nyu.edu/pipermail/fom/2006-March/010279.html
Friedman, H. 2006d. “n(3) < Graham's number < n(4) < TREE[3]”. FOM Archives,
March/010290.html
Friedman, H. 2006e. “274: Subcubic Graph Numbers”. FOM Archives”. April 1, 2006,
http://www.cs.nyu.edu/pipermail/fom/2006-April/010305.html
2006, http://www.cs.nyu.edu/pipermail/fom/2006-April/010362.html
Friedman, H. 2006g. “Integer Thresholds in FFF”. FOM Archives, June 22, 2006,
http://www.cs.nyu.edu/pipermail/fom/2006-June/010627.html
http://www.math.ohio-state.edu/%7Efriedman/manuscripts.html
Friedman H. 2007b. “New Borel Independence Results”. May 7, 2007, 20 pages, proof
sketch, #57,
http://www.math.ohio-state.edu/%7Efriedman/manuscripts.html
Friedman, H. 2009. "Upper Shift Fixed Points and Large Cardinals", "Upper Shift Fixed
http://www.cs.nyu.edu/pipermail/fom/2009-October/014119.html
http://www.cs.nyu.edu/pipermail/fom/2009-October/014123.html
Friedman, H. 2010. Boolean Relation Theory and the Incompleteness Phenomena, to appear,
http://www.math.ohio-state.edu/%7Efriedman/manuscripts.html
Friedman, H., Robertson, N., and P. Seymour P. 1987. “The Metamathematics of the
Gentzen, G. 1969. The collected papers of Gerhard Gentzen. Edited and translated into
Gödel, K. 1929. “On the completeness of the calculus of logic”. In (Gödel 1986-2003
Vol. I, 61-101).
Gödel, K. 1932a. “On the intuitionistic propositional calculus”. In: (Gödel 1986-2003
Vol. I 223-225).
Gödel, K. 1932b. “On completeness and consistency”. In: (Gödel 1986-2003 Vol. I 235-
237).
Gödel, K. 1933a. “On intuitionistic arithmetic and number theory”. In: (Gödel 1986-
Gödel, K. 1936. “On the lengths of proofs”. In: (Gödel 1986-2003 Vol. I 397-399).
Gödel, K. 1938. “The consistency of the axiom of choice and of the generalized
Gödel K. 1939a. The consistency of the generalized continuum hypothesis, 1939, in:
Gödel, K. 1939b. “Consistency proof for the generalized continuum hypothesis”. In:
62
Gödel, K. 1940. The consistency of the axiom of choice and the generalized continuum
hypothesis with the axioms of set theory. Annals of mathematics studies, vol. 3
Gödel, K. 1944. “Russell’s mathematical logic”. In: (Gödel 1986-2003 Vol. II 119-141).
Gödel, K. 1958. “On a hitherto unutilized extension of the finitary standpoint”. In:
Gödel, K. 1972. “On an extension of finitary mathematics which has not yet been
Gödel, K. 1986-2003. Collected works. Vol. I-V. Oxford: Oxford University Press (S.
Gogol, D. 1979. “Sentences with three quantifiers are decidable in set theory”.
Goodstein, R.L. 1944. “On the restricted ordinal theorem”. Journal of Symbolic Logic 9,
1944, 33-41.
Grunewald, F.J., and Segal, D. 1981. “How to solve a quadratic equation in integers”.
Hilbert, D., and Bernays, P. 1934,1939. Grundlagen der Mathematik, Vol. 1 (1934), Vol II
Hosoi, T., and Ono, H. 1973. “Intermediate propositional logics (a survey)”. Journal of
Jech, T. 2006. Set Theory, Third Edition, Springer Monographs in Mathematics. Springer.
Gödel’s second incompleteness theorem”. Journal of Symbolic Logic, Vol. 38, 1973,
359-367.
Jones, J.P. 1982. “Universal Diophantine equation”. Journal of Symbolic Logic, 47(3):549-
571.
Kechris, A.S. 1994. Classical Descriptive Set Theory. Graduate Texts in Mathematics,
Springer-Verlag, 1994.
analysis”. Trans. Amer. Amth. Soc. vol. 357, no. 1, 89-128, 2005.
analysis”. In: B. Cooper, B. Loewe, A. Sorbi (eds.), CIE 2005 New Computational
http://www.mathematik.tu-darmstadt.de/~kohlenbach/
darmstadt.de/~kohlenbach/
Kohlenbach, U., and Oliva, P. 2003. “Proof mining: a systematic way of analyzing
64
proofs in mathematics”. Proc. Steklov Inst. Math., vol. 242, 136-164, 2003.
Kreisel, G. 1968a. “A survey of proof theory”. The Journal of Symbolic Logic, 33, 321-388.
Kreisel, G. 1968b. “Functions, ordinals, species”. In: Staal and van Rootselaar, eds.
143-158.
Kruskal, J.B. 1960. “Well-quasi-ordering, the tree theorem, and Vazsonyi’s conjecture”.
Choice”. Preprint.
Martin, D.A. 1985. “A purely inductive proof of Borel determinacy”. In: Recursion
Martin, D.A., and J.R. Steel, J.R. 1989. “A proof of projective determinacy”. J. Amer.
Masser, D.W. 1998. “How to solve a quadratic equation in rationals”. Bull. London
Siena). 1983.
correction”. In: T. Jech, ed., Axiomatic set theory, Proceeding of symposia in pure
185-188.
Parikh, R. 1971. “Existence and feasibility in arithmetic”. Journal of Symbolic Logic 36,
494-508, 1971.
Parkih, R. 1973. “Some results on the lengths of proofs”. Transactions of the American
1133-1142.
Quine, W.V.O. 1940, 1951. Mathematical Logic. New York: W. W. Norton & Co. Revised
Algebraic Sets. The Decision Problem for the Existential Theory of the Reals. J.
Theory of the Reals, Part II: The General Decision Problem. Preliminaries for
(3): 329-352.
Robertson, N., and Seymour, P.D. 1985. “Graph minors---a survey”. In: Surveys in
combinatorics 1985 (Glasgow, 1985), 153--171, London Math. Soc. Lecture Note Ser.,
Robertson, N., and Seymour, P.D. 2004. “Graph minors. XX. Wagner's conjecture”. J.
Robinson, R.M. 1952. “An essentially undecidable axiom system”. In: Inernational
Rosser, J.B. 1936. “Extensions of some theorems of Gödel and Church”. Journal of
Rayo, A., and T. Wiliamson, T. 2003. “A Completeness Theorem for Unrestricted First-
Order Languages”. In J.C. Beall, ed., Liars and Heaps (Oxford: Clarendon Press),
pp. 331-356.
Scott, D.S. 1961. “Measurable cardinals and constructible sets”. Bull. Acad. Polon. Sci. 9
(1961), 521-524.
Siegel, C.L. 1972. “Zur Theorie der quadratischen Formen. Nachricten der Akademie
3:21-46 (German).
Skolem, T. 1922. “Some remarks on axiomatized set theory”. In: From Frege to Gödel: A
Smith, R.L. 1985. “The consistency strengths of some finite forms of the Higman and
Mathematics, ed. L.A. Harrington et al. New York: Elsevier Science Publishers,
119-159.
Solovay, R.M. 1970. “A model of set theory in which every set of reals is Lebesgue
Society).
Stanley, L.J. 1985. “Borel Diagonalization and Abstract Set Theory: Recent Results of
Mathematics, ed. L.A. Harrington et al. New York: Elsevier Science Publishers. 11-
68
86.
Tarski, A., and McKinsey, J.C.C. 1951. A decision method for elementary algebra and
Tarski, A. 1952. “Some notions and methods on the borderline of algebra and
Tarski, A., Mostowski, A., and Robinson, R.M. 1953. Undecidable Theories. Amsterdam:
North-Holland.
Tarski, A., and Vaught, R.L. 1957. “Arithmetical extensions of relational systems”.
Weyl, H. 1918. Das Kontinuum. Kritische Untersuchungen über die Grundlagen der
Wiedijk F. 2006. (ed.), The Seventeen Provers of the World. Foreword by Dana S. Scott,
Williamson, T. 2006. “Absolute Identity and Absolute Generality”. In: A. Rayo and G.
Uzquiano, eds., Absolute Generality, (Oxford: Oxford University Press), pp. 369-
69
389.
*This research was partially supported by NSF Grant DMS 0245349 and by Grant