Download Problems Connecting Logic and Number Theory and more Study notes Number Theory in PDF only on Docsity!
PROBLEMS CONNECTING LOGIC AND NUMBER
THEORY
BARRY MAZUR
In the recent conference (May 10, 11) at Harvard University^1 I was asked to take part in a question-and-answer session with Carol Wood and Bjorn Poonen regarding questions that relate Mathematical Logic to Number Theory.
- Our three “discussion problems.” Bjorn Poonen discussed the “recognition problem” for finitely gen- erated rings (and fields). That is, given two finitely generated commu- tative rings A and B, presented in terms of generators and relations, is there a decision procedure to determine whether or not these rings are isomorphic (this being, one would think, a basic issue for algebraic geometry!). Of course, if one drops the requirement of commutativity, one comes up against the unsolvablity of the corresponding problem for finitely generated groups (by taking A and B simply to be integral group rings). Carol Wood brought up cases where model theory, applied to number theoretic problems provided bounds that are impressively good! Model Theory—in some instances—yields significantly new proofs of theorems obtained by the number-theorists 2. In other instances, model theory achieves startling results for problems not yet considered by number theorists 3. Carol Wood discussed the recent article of Pila and Wilkie ([PW07]) that provides asymptotic upper bounds (as a function of the variable T ) for the number of Q-rational points of height ≤ T that
(^1) MAMLS@Harvard, a meeting on the intersections of logic and mathematics. I
want to thank Rehana Patel for organizing it and inviting us to participate.
(^2) For example, Hrushovski’s model-theoretic proof of the Manin-Mumford Con-
jecture was recently revisited and formulated as a (new) number theoretic proof in [PR06].
(^3) For example, explicit bounds for the number of transcendental points on the
intersection of subvarieties of semi-abelian varieties and a given finitely generated subgroup [HP00]. These bounds are double exponentials in the rank of the finitely generated group. 1
2 BARRY MAZUR
lie in a given semi-analytic set X ⊂ Rn^ but are outside any positive- dimensional semi-algebraic set contained in X.
I talked about examples of unsolvable problems given in Diophantine language (following Matiyasevich). Specifically, the negative solution of Hilbert’s Tenth Problem. Here are some notes on this^4.
- Comments on Hilbert’s Tenth Problem Following recent work on this problem it is useful to phrase the discussion for a general commutative ring A finitely presented over Z or over Q, and of infinite cardinality^5. The basic question is: Does there exist a finite algorithm to determine whether any finite system of polynomial equations in finitely many variables with coefficients in A has a solution in A or not? Here, in a nutshell, is the general status of this question we inherited from Hilbert and from “classical work” ( Julia Robinson/Davis/Putnam/ Matiyasevich). The culminating theorem is Matiyasevich’s:
Theorem 2.1. Every recursively enumerable^6 subset of Z is diophan- tine (relative to Z).
This fundamental result, of course, gives a negative answer to the question above, but does far more than just that. For example: (1) The result implies that relatively benign subsets of Z can be diophantinely described, as well. This is not as clear as one
(^4) I want to thank Bjorn and Carol for their comments about, and corrections to,
an early draft of these notes.
(^5) Of course, the historically interesting case for this problem is A = Z or A = Q. A close translation of Hilbert’s formulation of the problem is as follows:
Given a diophantine equation with any number of unknown quan- tities and with rational integral numerical coefficients: To devise a process according to which it can be determined in a finite number of operations whether the equation is solvable in rational integers. (^6) I’m told, by Bjorn, that logicians these days are suggesting that the terminology
“computably enumerable” replace “recursively enumerable”.
4 BARRY MAZUR
the 2-primary part of the Shafarevich-Tate Conjecture, Hilbert’s Tenth problem has a negative answer for the ring of integers in any number field. Since Kirsten Eisentr¨ager has, in her thesis, related Hilbert’s Tenth Problem over rings of integers in number fields to a much more general class of rings, one gets—thanks to her work:
Theorem 2.3. Conditional on the 2-primary part of the Shafarevich- Tate Conjecture, Hilbert’s Tenth problem has a negative answer for any commutative ring A that is of infinite cardinality, and is finitely generated over Z.
- The focus on cubics! One variable is OK; linear and quadratic are OK. There are unsolv- able fourth degree polynomials over Z (in a large number of variables). This focuses on the third degree, and there—to my knowledge—our knowledge essentially stops^9.
- Rational points on cubic plane curves A famous half-century-old example here is Selmer’s equation:
3 X^3 + 4Y 3 + 5Z^3 = 0
which has NO (nontrivial) solutions over Q even though all “local in- dicators” don’t rule out the possibility that a rational (nontrivial) so- lution exists^10 So that particular problem is “solved.” But, more gen- erally, we want to know:
Is there an algorithm to answer—for any third degree polynomial F (X, Y ) over Q—the question: is there a pair of rational numbers a, b such that F (a, b) = 0?
(^9) It is interesting how our lack of understanding of cubics seems to color lots of
mathematics, from the ancient concerns in the “one-variable case” having to do with “two mean proportionals,” and Archimedes’ Prop.4 of Book II of The Sphere and Cylinder and Eutocius’ commentaries on this, and–of course—the Italian 16c early algebraists.
(^10) This projective curve has points rational over every completion of Q.
PROBLEMS CONNECTING LOGIC AND NUMBER THEORY 5
A proof of the Shafarevich-Tate Conjecture^11 would provide a proof that a certain algorithm works for the general third degree polyno- mial F (X, Y ) and—more generally—to find rational points on curves of genus one. The algorithm itself is currently known, and used quite extensively. If it (always) works, then it gives an answer to the question posed above, and indeed allows us to find the rational points. But we don’t know whether it will always terminate (in finite time) to provide us with an answer. The Shafarevich-Tate Conjecture would guarantee termination in finite time. This is a huge subject (the arithmetic theory of elliptic curves) and it would be good to understand it as well as we can possibly understand it. Note the curious irony in the formulation of Theorem 2.3:
If we have a proof of the (2-primary part of the) Shafarevich- Tate conjecture
—i.e., colloquially speaking: if the algorithm that en- ables us to deal with arithmetic of cubic plane curves can be proved to work—
then we have a proof of the non-existence of a general algorithm for the ring of integers over any number field.
- Integral points on plane curves of genus one Here one has a striking explicit result, thanks to Baker’s method. Let f (X 1 , X 2 ) ∈ Z[X 1 , X 2 ] be an absolutely irreducible polynomial such that the associated projective curve f = 0 has genus one. Let n := the (total) degree of f (X 1 , X 2 ) and let H := the maximum of the (ordinary) absolute values of the coefficients of f (X 1 , X 2 ). Then there are finitely many integral solutions (a 1 , a 2 ) of the equation f (X 1 , X 2 ) = 0 and they are bounded explicitly by the inequality
max{|a 1 |, |a 2 |} < exp exp exp { 2 H^10 n
10 }. For discussion about this, see section 4.4 of [B75].
- Polynomials of degree three in many variables, over Z
We are now left to ponder one of the big open problems in this area: Is there an algorithm to answer—with input third de- gree polynomials F (X 1 , X 2 ,... , Xn) over Z for arbitrary (^11) In fact, just, the p-primary part of the Shafarevich-Tate Conjecture, for any
single prime number p will do it.
PROBLEMS CONNECTING LOGIC AND NUMBER THEORY 7
- If V is of genus one, then
- D(V ) is 0 if V has no rational points,
- D(V ) is 1 if V has infinitely many rational points, and
- D(V ) is the order of the Mordell-Weil group of V over Q, if that group is finite. In all cases for V of genus one, then, (using [M77]) we get that D(V ) ≤ 16.
- If V is of genus > 1, by Faltings’ Theorem D(V ) is the (finite) number of rational points of V. Conditional on a conjecture of Lang, Caparoso, Harris and I have shown that D(V ) is bounded by a function that depends only on the genus of V.
In sum, we have that (conditional on a conjecture of Lang) for all algebraic varieties V of dimension one,
D(V ) is bounded from above by a function F (|VC|) that depends only on the homotopy type |VC| of the complex analytic space associated to V.
Is the above statement true or false for algebraic surfaces? Or, more generally, for algebraic varieties of arbitrary dimension?
Bibliography
[B75] A. Baker, Transcendental Number Theory, Cambridge Mathe- matical Library, Cambridge University Press (1975). [E03] A.K. Eisentr¨ager, Hilbert’s Tenth Problem and Arithmetic Ge- ometry, PhD Thesis, University of California, Berkeley, 2003. [Den80] J. Denef, Diophantine sets over algebraic integer rings. II, Trans. Amer. Math. Soc. 257 (1980), no. 1, 227-236. [DL78] J. Denef and L. Lipshitz, Diophantine sets over some rings of algebraic integers, J. London Math. Soc. (2) 18 (1978), no. 3, 385-391. [H96] E. Hrushovski, The Mordell-Lang conjecture for function elds. Journal of the American Mathematical Society 9 (1996), no. 3, 667-
[H98] E. Hrushovski, Proof of Manin’s theorem by reduction to posi- tive characteristic, Model theory and algebraic geometry, Lecture Notes in Math., 1696 , Springer, Berlin, (1998) 197-205. [H00] E. Hrushovski, Anand Pillay, Effective bounds for the num- ber of transcendental points on subvarieties of semi-abelian varieties. American Journal of Mathematics 122 (2000), no. 3, 439-450.
8 BARRY MAZUR
[ JSWW76] J. P. Jones, D. Sato, H. Wada and Do. Wiens, Dio- phantine Representation of the Set of Prime Numbers, The American Mathematical Monthly, 83 , No. 6 (Jun. - Jul., 1976), pp. 449-464. [M77] B. Mazur, Modular Curves and the Eisenstein ideal, Publ. IHES 47 (1977) 33-186. [MR09] B. Mazur and K. Rubin, Ranks of twists of elliptic curves and Hilbert’s Tenth Problem, (http://abel.math.harvard.edu/ mazur/) and also on Archiv (arXiv:0904.3709v2 [math.NT] 25 Apr 2009) [Phe88] Thanases Pheidas, Hilbert’s tenth problem for a class of rings of algebraic integers, Proc. Amer. Math. Soc. 104 (1988), no. 2, 611-620. [PR06] R. Pink, D. Roessler, On Hrushovski’s proof of the Manin- Mumford conjecture (preprint). [Po] Bjorn Poonen, Using elliptic curves of rank one towards the un- decidability of Hilbert’s Tenth Problem over rings of algebraic integers (... ) [PW07] J. Pila, A.J. Wilkie, The rational points of a definable set, MIMS Eprint .2007.198. [Shl89] Alexandra Shlapentokh, Extension of Hilbert’s tenth problem to some algebraic number fields, Comm. Pure Appl. Math. 42 (1989), no. 7, 939- 962. [Shl00b] Alexandra Shlapentokh, Hilbert’s tenth problem over num- ber fields, a survey, Hilbert’s tenth problem: relations with arithmetic and algebraic geometry (Ghent, 1999), Amer. Math. Soc., Providence, RI, 2000, pp. 107- 137.
