Moscow Mathematical Journal Volume 4, Number 1, January-March 2004, Pages 245-305
Moscow Mathematical Journal Volume 4, Number 1, January-March 2004, Pages 245-305
Moscow Mathematical Journal Volume 4, Number 1, January-March 2004, Pages 245-305
Abstract. Diophantine Analysis is a very active domain of mathematical research where one nds more conjectures than results. We collect here a number of open questions concerning Diophantine equations (including Pillais Conjectures), Diophantine approximation (featuring the abc Conjecture) and transcendental number theory (with, for instance, Schanuels Conjecture). Some questions related to Mahlers measure and Weil absolute logarithmic height are then considered (e. g., Lehmers Problem). We also discuss Mazurs question regarding the density of rational points on a variety, especially in the particular case of algebraic groups, in connexion with transcendence problems in several variables. We say only a few words on metric problems, equidistribution questions, Diophantine approximation on manifolds and Diophantine analysis on function elds.
2000 Math. Subj. Class. Primary 11Jxx; Secondary 11Dxx, 11Gxx, 14Gxx. Key words and phrases. Diophantine problems, transcendence, linear and algebraic independence, Schanuel conjecture, four exponential conjecture, multizeta values, abc conjecture, Waring problem, Diophantine approximation, continued fractions, ThueSiegelRothSchmidt, irrationality measures, Marko spectrum, height, Lehmer problem, Mazur density conjecture, abelian varieties, special functions, function elds.
1. Diophantine Equations 1.1. Points on Curves. Among the 23 problems posed by Hilbert [Hi], [Gu] the tenth one has the shortest statement. Given a Diophantine equation with any number of unknown quantities and with integral numerical coecients: To devise a process according to which it can be determined by a nite number of operations whether the equation is solvable in rational integers. An equation of the form f (x) = 0, where f Q[X1 , . . . , Xn ] is a given polynomial, while the unknowns x = (x1 , . . . , xn ) take rational integer values, is a Diophantine equation. To solve this equation amounts to determining the integer points on the corresponding hypersurface of the ane space. Hilberts Tenth Problem is to
Received October 16, 2002.
c 2004 Independent University of Moscow
245
246
M. WALDSCHMIDT
nd an algorithm which tells us whether or not such a Diophantine equation has a solution. There are other types of Diophantine equations. First of all one may consider rational solutions instead of integer ones. In this case, one considers rational points on a hypersurface. Next, one may consider integer or rational points over a number eld. There is a situation which is intermediate between integer and rational points, where the unknowns take S-integral point values. This means that S is a xed, nite set of prime numbers (rational primes, or prime ideals in the number eld), and that the denominators of the solutions are restricted to those belonging to S. Examples are the ThueMahler equation F (x, y) = pz1 pzk 1 k where F is a homogeneous polynomial with integer coecients and p1 , . . . , pk are xed primes (the unknowns are x, y, z1 , . . . , zk and take rational integer values with zi 0) and the generalized RamanujanNagell equation x2 + D = pn , where D is a xed integer, p a xed prime, and the unknowns are x, n which take rational integer values with n 0 (see for instance [ST], [Ti3], [Sh1], and [BS] for these and other similar questions). Also, it is interesting to deal with simultaneous Diophantine equations, i. e., to study rational or integer points on algebraic varieties. The nal answer to Hilberts original Tenth Problem was given in 1970 by Matiyasevich, following the works of Davis, Putnam and Robinson. This was the culminating stage of a rich and beautiful theory (see [DMR], [Ma] and [Mat]). The solution is negative, there is no hope of producing a complete theory of the subject. But one may still hope that there is a positive answer if one restricts Hilberts initial question to equations in a limited number of variables, say n = 2, which amounts to considering integer points on a plane curve. In this case, deep results were achieved during the 20th century and many results are now known, but many more remain to be discovered. The most basic results are those of Siegel (1929) and Faltings (1983). Siegels Theorem deals with integer points and produces an algorithm to decide whether the set of solutions forms a nite or an innite set. Faltingss result, solving Mordells Conjecture, does the same for rational solutions, i. e., rational points on curves. To these two outstanding achievements of the 20th century, one may add Wiless contribution, which not only settles the Last Fermat Theorem, but also provides a quantity of similar results for other curves [K]. Some natural questions arise. (a) To answer Hilberts tenth Problem for this special case of plane curves, which means to nd an algorithm to decide whether a given Diophantine equation f (x, y) = 0 has a solution in Z (and the same problem in Q). (b) To nd an upper bound for the number of either rational or integral points on a curve. (c) To nd an algorithm for solving explicitly a given Diophantine equation in two unknowns.
247
Further questions may be asked. For instance in question b) one might ask for the exact number of solutions; it may be more relevant to consider more generally the number of points on any number eld, or the number of points of bounded degree and to investigate the related generating series. . . The number of open problems is endless! Our goal here is not to describe in detail the state of the art regarding these questions (see for instance [La8]). It suces to say that a complete answer to question (a) is not yet available. There is no algorithm (not even a conjectural one) to decide whether a curve has a rational point or not, that a number of results are known about question (b), the latest work on this topic being due to G. Rmond [Re] who produces an eective upper bound for the e number of rational points on a curve of genus 2, and that question (c) is unanswered even for integer points, and even for the special case of curves of genus 2. We do not request a practical algorithm, but only (to start with) a theoretical one. So our rst open problem will be an eective renement to Siegels Theorem. Problem 1.1. Let f Z[X, Y ] be a polynomial such that the equation f (x, y) = 0 has only nitely many solutions (x, y) Z Z. Give an upper bound for max{|x|, |y|} when (x, y) is such a solution, in terms of the degree of f and of the maximal absolute value of the coecients of f . That such a bound exists is part of the hypothesis, but the problem is to state it explicitly (and, if possible, in a closed form). Further similar questions might also be asked regarding equations involving more variables (rational points on varieties), for instance Schmidts norm form equations. We refer the reader to [La3] and [La8] for such questions, including the LangVojta Conjectures. Even the simplest case of quadratic forms suggests open problems. The determination of all positive integers which are represented by a given binary form is far from being solved. It is also expected that innitely many real quadratic elds have class number one, but it is not even known that there are innitely many number elds (without restriction on the degree) with class number one. Recall that the rst complete solution of Gausss class number 1 and 2 Problems (for imaginary quadratic elds) has been obtained by transcendence methods (A. Baker and H. M. Stark), so it may be considered to be a Diophantine problem. Nowadays more ecient methods (Goldfeld, GrossZagier, . . . see [La8, Chap. V, 5]) are available. A related open problem is the determination of Eulers numeri idonei [Ri2]. Fix a positive integer n. If p is an odd prime for which there exist integers x 0 and y 0 with p = x2 + ny 2 , then (i) gcd(x, ny) = 1, (ii) the equation p = X 2 + nY 2 in integers X 0 and Y 0 has the only one solution, X = x and Y = y. Now let p be an odd integer such that there exist integers x 0 and y 0 with p = x2 + ny 2 and such that the conditions (i) and (ii) above are satised. If these
248
M. WALDSCHMIDT
properties imply that p is prime, then the number n belongs to the set of so-called numeri idonei. Euler found 65 such integers n 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 13, 15, 16, 18, 21, 22, 24, 25, 28, 30, 33, 37, 40, 42, 45, 48, 57, 58, 60, 70, 72, 78, 85, 88, 93, 102, 105, 112, 120, 130, 133, 165, 168, 177, 190, 210, 232, 240, 253, 273, 280, 312, 330, 345, 357, 385, 408, 462, 520, 760, 840, 1320, 1365, 1848. It is known that there is at most one more number in the list, but one expects there be no other. Here is just one example ([Sie1, problem 58, p. 112], [Guy, D18]) of an open problem dealing with simultaneous Diophantine quadratic equations. Is there a perfect integer cuboid ? The existence of a box with integer edges x1 , x2 , x3 , integer face diagonals y1 , y2 , y3 and integer body diagonal z, amounts to solving the system of four simultaneous Diophantine equations in seven unknowns 2 2 2 x1 + x2 = y3 , 2 x + x2 = y 2 , 2 3 1
2 x2 + x2 = y2 , 3 1 2 x1 + x2 + x2 = z 2 2 3
in Z. We dont know whether there is a solution, but it is known that there is no perfect integer cuboid with the smallest edge 231 . 1.2. Exponential Diophantine Equations. In a Diophantine equation, the unknowns occur as the variables of polynomials, while in an exponential Diophantine equation (see [ST]), some exponents also are variables. One may consider the abovementioned RamanujanNagell equation x2 +D = pn as an exponential Diophantine equation. A famous problem which was open until 2002 is Catalans one which dates back to 1844 [Cat], the same year where Liouville constructed the rst examples of transcendental numbers (see also [Sie1, problem 77, p. 116], [Sie2, n 60, p. 42], [ST, Chap. 12], [N1, Chap. 11], [Ri1], [Guy, D9], [Ri2, Chap. 7]). The Note extraite dune lettre adresse ` lEditeur par Monsieur E. Catalan, Rptiteur ` e a e e a lcole polytechnique de Paris, published in Crelle Journal [Cat], reads: e Je vous prie, Monsieur, de bien vouloir noncer, dans votre recueil, e le thor`me suivant, que je crois vrai, bien que je naie pas encore e e russi ` le dmontrer compl`tement: dautres seront peut-tre plus e a e e e heureux. Deux nombres entiers conscutifs, autres que 8 et 9, ne peuvent e tre des puissances exactes; autrement dit: lquation xm y n = e e 1, dans laquelle les inconnues sont enti`res et positives, nadmet e quune seule solution. This means that the only example of consecutive numbers which are perfect powers xp with p 2 should be 8 and 9. Further information on the history of this question is available in Ribenboims book [Ri1]. Tijdemans result [Ti2] in 1976
249
shows that there are only nitely many solutions. More precisely, for any solution x, y, p, q, the number max{p, q} can be bounded by an eectively computable absolute constant. Once max{p, q} is bounded, only nitely many exponential Diophantine equations remain to be considered, and there are algorithms to complete the solution (based on Bakers method). Such a bound has been computed, but it is somewhat large: M. Mignotte proved that any solution x, y, p, q to Catalans equation should satisfy max{p, q} < 8 1016 . Catalans claim was nally substantiated by P. Mihilescu [Mi] (see also [Bi] a and [Me]). Theorem 1.2 (Catalans Conjecture). The equation xp y q = 1, where the unknowns x, y, p and q take integer values all 2, has only one solution, namely (x, y, p, q) = (3, 2, 2, 3). The nal solution by Mihilescu involves deep results from the theory of cya clotomic elds. Initially sharp measures of linear independence of logarithms of algebraic numbers were required, namely a specic estimate for two logarithms due to M. Laurent, M. Mignotte and Yu. V. Nesterenko, but then a solution using neither results from transcendental number theory nor the help of a computer was derived. Catalan asked for integral solutions, like in Siegels Theorem, while Faltingss Theorem deals with rational points. D. Prasad suggested that the set of tuples (x, y, p, q) in Q2 N2 satisfying the conditions xp y q = 1, and the curve X p Y q = 1 has genus 1 should be nite evidence for this is provided by the abc Conjecture (see Section 2.1). The fact that the right hand side in Catalans equation is 1 is crucial. Nothing is known if one replaces it by another positive integer. The next conjecture was proposed by S. S. Pillai [Pi] at a conference of the Indian Mathematical Society held in Aligarh (see also [Sie1, problem 78, p. 117], [ST], [Ti3], [Sh1]). Conjecture 1.3 (Pillai). Let k be a positive integer. The equation xp y q = k, where the unknowns x, y, p and q take integer values, all 2, has only nitely many solutions (x, y, p, q). This means that in the increasing sequence of perfect powers xp , with x 2 and p 2: 4, 8, 9, 16, 25, 27, 32, 36, 49, 64, 81, 100, 121, 125, 128, 144, 169, . . . , the dierence between two consecutive terms tends to innity. It is not even known whether for, say, k = 2, Pillais equation has only nitely many solutions. A related open question is whether the number 6 occurs as a dierence between two perfect
250
M. WALDSCHMIDT
powers: Is there a solution to the Diophantine equation xp y q = 6? (see [Sie2, problem 238a, p. 116]). A conjecture which implies Pillais has been suggested by T. N. Shorey in [Sh2]. This is the very problem which motivated C. L. Siegel in [Si1]. Let f Z[X] be a polynomial of degree n with at least two distinct roots and f (0) = 0. Let L be the number of non-zero coecients of f . Write f (X) = b1 X n1 + + bL1 X nL1 + bL with n = n1 > n2 > > nL1 > 0 and bi = 0 (1 i L). Set H = H(f ) = max1iL |bi |. Conjecture 1.4 (Shorey). There exists a positive number C which depends only on L and H with the following property. Let m, x and y be rational integers with m 2 and |y| > 1 satisfying y m = f (x). Then either m C, or else there is a proper sub-sum in y m b1 xn1 bL1 xnL1 bL which vanishes. An example with a vanishing proper sub-sum is y m = xn1 + x 2 where H = 2, L = 3 and a solution (m, x, y) = (n1 , 2, 2). Consider now the positive integers which are perfect powers y q , with q 2, and such that all digits in some basis x 2 are 1s. Examples are 121 in basis 3, 400 in basis 7 and 343 in basis 18. To nd all solutions amounts to solving the exponential Diophantine equation xn 1 = yq , x1 where the unknown x, y, n, q take positive, rational, integer values with x 2, y 1, n 3 and q 2. Only 3 solutions are known (x, y, n, q) = (3, 11, 5, 2), (7, 20, 4, 2), (18, 7, 3, 3), corresponding to the above-mentioned three examples. One does not know whether these are the only solutions (see [ST], [Guy, D10], [Ti3], [Sh1], [BM] and [Sh2]), but it is expected that there are no others. The next question is to determine all the perfect powers with identical digits in some basis, which amounts to solving the equation xn 1 = yq , x1 where the unknown x, y, n, q, z take positive, rational, integer values with x 2, y 1, n 3, 1 z < x and q 2. Another type of exponential Diophantine equation has been studied in a joint paper by H. P. Schlickewei and C. Viola [SV] where they state the following conjecture. z
251
Conjecture 1.5. Let k 2 be an integer and 1 , . . . , n be non-zero elements in a eld K of zero characteristic, such that, no quotient i /j with j = i is a root of unity. Consider the function X1 X k 1 1 F (X1 , . . . , Xk ) = det . . . . . . . . . . . . . . . . X X 1 k k k Then the equation F (0, x2 , . . . , xk ) = 0 has only nitely many solutions (x2 , . . . , xk ) Zk1 such that, in the corresponding determinant, all (k 1) k and all k (k 1) submatrices have rank k 1. Other exponential Diophantine equations are worth of study. See for instance [N1, Chap. III], [ST], [Ti3] and [Sh1]. Among the numerous applications of Bakers transcendence method are several questions related to the greatest prime factors of certain numbers. In this connexion we mention Grimms Conjecture ([Gri], [N1, Chap. III, 3], [Guy, B32]). Conjecture 1.6 (Grimm). Given k consecutive, composite integers, n + 1, . . . , n + k, there exist k distinct primes p1 , . . . , pk such that n + j is divisible by pj , 1 j k. This conjecture may be rephrased as follows. Given an increasing sequence of positive integers n1 < < nk for which the product n1 nk has fewer than k distinct prime factors, there is a prime p in the range n1 p nk . The equivalence of this with the original formulation follows from the marriage theorem. According to P. Erds and J. L. Selfridge, a consequence of Conjecture 1.6 is o that between two consecutive squares there is always a prime number. A weaker form of Conjecture 1.6, which is also an open problem, is Conjecture 1.7. If there is no prime in the interval [n+1, n+k], then the product (n + 1) (n + k) has at least k distinct prime divisors. M. Langevin (personal communication) pointed out that Grimms Conjecture cannot be extended to arithmetical progressions without a proviso. For instance the numbers 12, 25, 38, 51, 64, 77, 90 belong to an arithmetic progression of ratio 13, but the number of distinct prime factors of 12256490 is only 3. However in [Lan1] he proposed a stronger conjecture than Conjecture 1.6. Conjecture 1.8 (Langevin). Given an increasing sequence n1 < n2 < < nk of positive integers such that n1 , n2 , . . . , nk are multiplicatively dependent, there exists a prime number in the interval [n1 , nk ]. Even if they may not be classied as Diophantine questions, the following open problems (see [La10]) are related to this topic: the twin prime conjecture, the Goldbach Problem (is every even integer 4 the sum of two primes?), Bouniakovskys conjecture, Schinzels hypothesis (H) (see also [Sie1, 29]) and the BatemanHorn Conjecture. The Diophantine equation xp + y q = z r
252
M. WALDSCHMIDT
has also a long history in relation with Fermats last Theorem ([K], [Ri2, 9.2.D]). If we look at the solutions in positive integers (x, y, z, p, q, r) for which 1 1 1 + + <1 p q r and such that x, y, z are relatively prime, then only 10 solutions1 are known, 1 + 23 = 32 , 35 + 114 = 1222 , 25 + 72 = 34 , 73 + 132 = 29 , 27 + 173 = 712 , 14143 + 22134592 = 657 , 338 + 15490342 = 156133 . 177 + 762713 = 210639282 , 438 + 962223 = 300429072 ,
1 1 1 + + <1 p q r
implies
1 1 1 41 + + , p q r 42
the abc Conjecture (see Section 2.1) predicts that the set of such solutions is nite (the FermatCatalan Conjecture formulated by Darmon and Granville see [Mau]). For all known solutions, one of p, q, r is 2; this led R. Tijdeman and D. Zagier to conjecture2 that there is no solution with the further restriction that each of p, q and r is 3. A Diophantine tuple is a tuple (a1 , . . . , an ) of distinct positive integers such that ai aj + 1 is a square for 1 i < j n (see [Guy] and [G]). Fermat gave the example (1, 3, 8, 120), and Euler showed that any Diophantine pair (a1 , a2 ) can be extended to a Diophantine quadruple (a1 , a2 , a3 , a4 ). It is not known whether there exists a Diophantine quintuple (a1 , a2 , a3 , a4 , a5 ), but A. Dujella [Du] proved that each 26 Diophantine quintuple has max{a1 , a2 , a3 , a4 , a5 } 1010 . He also proved that there is no Diophantine sextuple. 1.3. Marko Spectrum. The original Marko3 equation (1879) is x2 + y 2 + z 2 = 3xyz (see [Ca, Chap. II], [CF, Chap. 2], [Guy, D12] and [Ri2, 10.5.B]). This is an algorithm which produces all solutions in positive integers. Given any solution (x, y, z) = (m, m1 , m2 ), we x two of the three coordinates; then we obtain a quadratic equation in the third coordinate, for which we already know a solution. By the usual process of cutting with a rational line we deduce another solution. From one solution (m, m1 , m2 ), this produces three other solutions (m , m1 , m2 ), where m = 3m1 m2 m, m1 = 3mm2 m1 , m2 = 3mm1 m2 . (m, m1 , m2 ), (m, m1 , m2 ),
These three solutions are called neighbors of the original one. Apart from the two so-called singular solutions (1, 1, 1) and (2, 1, 1), the three components of
1Up to obvious symmetries; in particular 1 + 23 = 32 counts only for one solution. 2This conjecture is also known as Beals Conjecture see [Mau]. 3His name is spelled Markov in probability theory.
253
(m, m1 , m2 ) are pairwise distinct, and the three neighbors of (m, m1 , m2 ) are pairwise distinct. Assuming m > m1 > m2 , one deduces m2 > m1 > m > m . Hence there is a neighbor of (m, m1 , m2 ) with maximum component less than m, and two neighbors, namely (m1 , m, m2 ) and (m2 , m, m1 ), with maximum component greater than m. It follows that one produces all solutions, starting from (1, 1, 1), by taking successively the neighbors of the known solutions. Here is the Marko tree, in the notation of H. Cohn [Coh], where (m1 , m, m2 ) is written on the right and (m2 , m, m1 ) on the left. (1, 1, 1) (2, 1, 1) (5, 2, 1) (29, 5, 2) (433, 29, 5) . . . . . . . . . (169, 29, 2) . . . . . . (194, 13, 5) . . . . . . (13, 5, 1) (34, 13, 1) . . .
The main open problem in this topic ([Ca, p. 33], [CF, p. 11] and [Guy, D12]) is to prove that each largest component occurs only once in a triple of this tree. Conjecture 1.9. Fix a positive integer m for which the equation m2 + m2 + m2 = 3mm1 m2 1 2 has a solution in positive integers (m1 , m2 ) with 0 < m1 m2 m. Then such a pair (m1 , m2 ) is unique. This conjecture has been proven for m 10105 . The sequence 1, 2, 5, 13, 29, 34, 89, 169, 194, 233, 433, 610, 985, 1325, 1597, . . . of integers m satisfying the hypotheses of Conjecture 1.9 is closely related to the question of the best rational approximation to quadratic, irrational, real numbers: for each m in this sequence, there is an explicit quadratic form fm (x, y) such that fm (x, 1) = 0 has a root m for which lim sup (q qm ) =
qZ, q
m . 9m2 4
(1.10)
254
M. WALDSCHMIDT
The third row gives the continued fraction expansion for m , where 2211, for instance, stands for [2, 2, 1, 1, 2, 2, 1, 1, 2, 2, 1, 1, . . . ]. Conjecture 1.9 amounts to claiming that there is no ambiguity in the notation fm : given m, two quadratic numbers m satisfying (1.10) should be roots of equivalent quadratic forms. Hence the Marko spectrum is closely related to rational approximation to a single real number. A generalization to simultaneous approximation is considered in Section 2.2 below. 2. Diophantine Approximation In this section we restrict ourselves to problems in Diophantine approximation which do not require introducing a notion of height for algebraic numbers, those will be discussed in Section 4. 2.1. The abc Conjecture. For a positive integer n, we denote by R(n) =
p|n
the radical or square free part of n. The abc Conjecture resulted from a discussion between D. W. Masser and J. sterl ([, p. 169]; see also [Mas], as well as [La7], [La8, Chap. II, 1], [La9, e Ch. IV, 7], [Guy, B19], [Bro], [Ri2, 9.4.E], [V], [Maz4] and [Ni]). Conjecture 2.1 (abc Conjecture). For each > 0 there exists a positive number () which has the following property: if a, b and c are three positive rational integers which are relatively prime and satisfy a + b = c, then c < ()R(abc)1+ . Conjecture 2.1 implies a previous conjecture by L. Szpiro on the conductor of elliptic curves. Given any > 0, there exists a constant C() > 0 such that, for every elliptic curve with minimal discriminant and conductor N , || < C()N 6+ . When a, b and c are three positive relatively prime integers satisfying a + b = c, dene log c (a, b, c) = log R(abc) and log abc (a, b, c) = . log R(abc)
255
Here are the six largest known values for (abc) (in [Bro] p. 102105 as well as in [Ni], one can nd all the 140 known values of (a, b, c) which are 1.4). a+b=c 1 2 2+3
10
(a, b, c)
5
author(s)
109 = 23
112 + 32 56 73 = 221 23
3 19 1307 + 7 292 318 = 28 322 54 1.623490. . . J. Browkin, J. Brzezinski J. Browkin, J. Brzezinski; 4 283 + 511 132 = 28 38 173 1.580756. . . A. Nitaj 7 4 5 1+23 =5 7 1.567887. . . B. de Weger 6 73 + 310 = 211 29 1.547075. . . B. de Weger
Here are the six largest known values for (abc), according to [Ni], where one can nd the complete list of 46 known triples (a, b, c) with 0 < a < b < c, a + b = c and gcd(a, b) = 1 satisfying (a, b, c) > 4. a+b=c 1 2 3 4 5 6 2 3
35 2 18 4 5 2
37 47 = 3 7
11 2
743
2 2
13 103 + 7
27 3
=3
11 15 10 19
5 11 5 7
2
7 17 19 + 3
3 29
107 = 5
8 2
37 2311
15
23 2269 + 17 29 31 = 2
17 79 211 + 2
23 29 = 5
As observed by M. Langevin [Lan2], a consequence of the abc Conjecture is the solution of the following open problem [E1]. Conjecture 2.2 (ErdsWoods). There exists a positive integer k such that, for o m and n positive integers, the conditions R(m + i) = R(n + i) imply m = n. Conjecture 2.2 is motivated by the following question raised by J. Robinson: Is rst order arithmetic denable using only the successor function S : x x + 1 and the coprimarity x y (x, y) = 1? It would suce to decide whether the function x 5x can be dened in the language (S, ); see [Woo], [Guy, B29 and B35], [BLSW]. From the abc Conjecture (or even the weaker version with some < 1 rather than all > 0), it follows that, apart from possibly nitely many exceptions (i = 0, . . . , k 1)
256
M. WALDSCHMIDT
(m, n), k = 3 is an admissible value. Indeed, assume m > n. Using the abc Conjecture with a = m(m + 2), b = 1, c = (m + 1)2 , we obtain m2 ()R m(m + 1)(m + 2)
1+
Now if R(m + i) = R(n + i) for i = 0, 1, 2 then R(m + i) divides m n, hence the number R m(m + 1)(m + 2) = lcm R(m), R(m + 1), R(m + 2) divides m n and therefore m2 ()m1+ . This shows that m is bounded. One suspects that there is no exception at all with k = 3. This would mean that if m and n have the same prime divisors, m + 1, n + 1 have the same prime divisors and m + 2, n + 2 have the same prime divisors, then m = n. That k = 2 is not an admissible value is easily seen: 75 and 1215 have the same prime divisors, and this is true also for 76 and 1216, R(75) = 15 = R(1215), R(76) = 2 19 = R(1216).
Apart from this sporadic example, there is also a sequence of examples: for m = 2h 2 and n = m(m + 2) = 2h m, R(m) = R(n)
2
because n + 1 = (m + 1) . A generalization of the ErdsWoods Problem to arithmetic progressions has o been suggested by T. N. Shorey. Question. Does there exist a positive integer k such that, for any non-zero integers m, n, d and d satisfying gcd(m, d) = gcd(n, d ) = 1, the conditions R(m + id) = R(n + id ) imply m = n and d = d ? On the one hand, if the answer is positive, k is at least 4, as shown by several examples of quadruples (m, n, d, d ), like (2, 2, 1, 7), (2, 8, 79, 1) or (4, 8, 23, 1): R(2) = R(2), R(3) = R(2 + 7), R(4) = R(2 + 2 7), R(2) = R(4) = R(8), R(2 + 79) = R(4 + 23) = R(9), (i = 0, . . . , k 1)
R(2 + 2 79) = R(4 + 2 23) = R(10). On the other hand, under the abc Conjecture, Shoreys question has a positive answer for k = 5 (see [Lan3]). Another related problem of T. S. Motzkin and E. G. Straus ([Guy, B19]) is to determine the pairs of integers m, n such that m and n + 1 have the same prime divisors, and also n and m + 1 have the same set of prime divisors. The known examples are m = 2k + 1, n = m2 1 (k 0) and the sporadic example m = 35 = 5 7, n = 4374 = 2 37 , which yields m + 1 = 22 32 and n + 1 = 54 7. We also quote another related conjecture attributed to P. Erds in [Lan2] and o to R. E. Dressler in [Ni].
257
Conjecture 2.3 (ErdsDressler). If a and b are two positive integers with a < b o and R(a) = R(b) then there is a prime p with a < p < b. The rst estimates in the direction of the abc Conjecture have been achieved by C. L. Stewart and R. Tijdeman, and then rened by C. L. Stewart and Yu Kunrui (see [SY1], [SY2]), using (p-adic) lower bounds for linear forms in logarithms: if a, b, c are relatively prime positive integers with a + b = c, then log c R1/3 (log R)3 where R = R(abc). An explicit version was worked out by Wong Chi Ho in 1999 [Wo] following an earlier version of [SY2]. For c > 2 the estimate log c R(1/3)+(15/ log log R) is valid. Further connexions between the abc Conjecture and measures of linear independence of logarithms of algebraic numbers have been pointed out by A. Baker [B2] and P. Philippon [P3] (see also [W6, exercise 1.11]). We reproduce here the main conjecture of the addendum of [P3]. For a rational number a/b with relatively prime integers a, b, we denote by h(a/b) the number log max{|a|, |b|}. Conjecture 2.4 (Philippon). There exist real numbers , and with 0 < < 1/2, 1 and 0, and a positive integer B, such that for any non-zero rational numbers x, y satisfying xy B = 1, if S denotes the set of prime numbers for which |xy B + 1|p < 1, then
pS
log p
The conclusion is a lower bound for the p-adic distance between xy B and 1; the main point is that several ps are involved. Conjecture 2.4 is telling us something about the prime decomposition of all numbers xy B +1 for some xed but unspecied value of B and it implies the abc Conjecture. Examples of optimistic Archimedean estimates related to measures of linear independence of logarithms of algebraic numbers are the LangWaldschmidt Conjectures in [La5] (introduction to Chap. X and XI, p. 212217). Here is a simple example. Conjecture 2.5 (LangWaldschmidt). For any > 0, there exists a constant C() > 0 such that, for any non-zero rational integers a1 , . . . , am , b1 , . . . , bm with ab1 abm = 1, m 1 |ab1 abm 1| m 1 where B = max1im |bi |. Similar questions related to Diophantine approximation on tori are discussed in [La8, Chap. IX, 7]. C()m B , (|b1 | |bm | |a1 | |am |)1+
258
M. WALDSCHMIDT
Conjecture 2.5 deals with rational integers; we shall consider algebraic numbers more generally in Section 4.3, once we have dened a notion of height. A very sharp conjectured lower bound for innitely many elements in a specic sequence |eb0 ab1 abm 1| m 1 with b0 arbitrary, and where all the exponents bi have the same sign (compare with Conjecture 2.14) has been shown by J. Sondow in [So] to yield the irrationality of Eulers constant. From either the abc Conjecture or Conjecture 2.5 one deduces a quantitative renement of Pillais Conjecture 1.3. Conjecture 2.6. For any > 0, there is a constant C() > 0 such that, for any positive integers x, y, p, q satisfying xp = y q , |xp y q | C() max{xp , y q }1(1/p)(1/q) . We consider two special cases of Conjecture 2.6: rst (p, q) = (2, 3), which gives rise to Halls Conjecture [H] (also [La8, Chap. II, 1]), Conjecture 2.7 (Hall). If x and y are positive integers with y 2 = x3 , then |y 2 x3 | C max{y 2 , x3 }1/6 . In this statement there is no . On the one hand, Conjecture 2.7 may be true by a sort of accident, but one may also expect that the estimate is too strong to be true. On the other hand, with the exponent (1/6) , the abc Conjecture provides a lower bound not only for |y 2 x3 |, but also for its radical [Lan3]: for x and y relatively prime positive integers with y 2 = x3 , R |y 2 x3 | C() max{y 2 , x3 }(1/6) . The exponent 1/6 in Conjecture 2.7 is optimal, as has been shown by L. V. Danilov and A. Schinzel. Indeed, using the polynomial identity (X 2 6X + 4)3 (X 2 + 1)(X 2 9X + 19)2 = 27(2X 11) which is related to Kleins identity for the icosahedron (cf. [Lan4, Th. 6]), they show that there exist innitely many pairs of positive integers (x, y) such that 54 0 < |y 2 x3 | < x. 25 5 2 3 The smallest known value for |y x |/ x (N. Elkies, 1998) is 0.0214. . . , with x = 3 7 211 38 791 6 975 841, y = 2 32 15 228 748 819 1 633 915 978 229, x3 y 2 = 33 72 17 73. The second special case of Conjecture 2.6 we consider is (x, y) = (3, 2). The question of how small 3n 2m can be in comparison with 2m has been raised by J. E. Littlewood [Guy, F23]. The example 312 7153 =1+ = 1.013. . . 19 2 524288 is related to music scales.
259
For further questions dealing with exponential Diophantine equations, we refer to Chap. 12 of the book of T. N. Shorey and R. Tijdeman [ST], as well as to the more recent surveys [Ti3] and [Sh1]. 2.2. ThueSiegelRothSchmidt. One of the main open problems in Diophantine approximation is to produce an eective version of the ThueSiegelRoth Theorem. For any > 0 and any irrational algebraic number , there is a positive constant C(, ) > 0 such that, for any rational number p/q, C(, ) p . > q q 2+ (2.8)
In connexion with the negative answer to Hilberts 10th Problem by Yu. Matiyasevich, it has been suggested by M. Mignotte that an eective version of Schmidts Subspace Theorem (which extends the ThueSiegelRoth Theorem to simultaneous Diophantine approximation) may be impossible. If this turns out to be the case also for the special case of the ThueSiegelRoth Theorem itself, then, according to E. Bombieri (see [Ni]), an eective version of the abc Conjecture would also be out of reach. M. Langevin noticed that the abc Conjecture yields a stronger inequality than Roths, p C() > . q R(pq)q So far, eective improvements are known only for Liouvilles bound, and to improve them is already a great challenge. Another goal would be to improve the estimate in Roths Theorem. In the lower bound (2.8) one would like to replace q 2 by, say, q 2 (log q)1 . It is expected that for any irrational real algebraic number of degree 3, the term q 2 cannot be replaced by q 2 in inequality (2.8), but the set of for which the answer is known is empty! This question is often asked for the special case of the number 3 2, but another interesting example (due to Stanislaw Ulam see for instance [Guy, F22]) is the real algebraic number dened by = 1 +y with y = 1 . 1+y
Essentially nothing is known about the continued fraction expansion of a real algebraic number of degree 3; one does not know the answer to any of the following two questions. Question 2.9. Does there exist a real algebraic number of degree 3 with bounded partial quotients? Question 2.10. Does there exist a real algebraic number of degree 3 with unbounded partial quotients? It is usually expected is that the continued fraction expansion of a real algebraic number of degree at least 3 always has unbounded partial quotients. More precisely one expects that real algebraic numbers of degree 3 behave like almost all real numbers (see Section 5.1).
260
M. WALDSCHMIDT
Let (q) be a continuous positive real valued function. Assume that the function q(q) is non-increasing. Consider the inequality p (q) > . q q (2.11)
Conjecture 2.12. Let be real algebraic number of degree at least 3. Then inequality (2.11) has innitely many solutions in integers p and q with q > 0 if and only if
(x)dx
1
diverges. A far-reaching generalization of the ThueSiegelRoth Theorem to simultaneous approximation is the Schmidt Subspace Theorem. Here are two special cases. Given real algebraic numbers 1 , . . . , n such that 1, 1 , . . . , n are linearly independent over Q, for any > 0,
1in
max i
pi 1 < 1+(1/n)+ q q
has only nitely many solutions (p1 , . . . , pn , q) in Zn+1 with q > 0. Given real algebraic numbers 1 , . . . , n such that 1, 1 , . . . , n are linearly independent over Q, for any > 0, |q1 1 + + qn n p| < 1 q n+
has only nitely many solutions (q1 , . . . , qn , p) in Zn+1 with q = max{|q1 |, . . . . . . , |qn |} > 0. These two types of Diophantine statements are parallel to the two types of Pad Approximants. It would be interesting to consider the analogue of Schmidts e Subspace Theorem in case of Pad Approximants, and also to study a corresponding e analogue of Khinchines transference principle [Ca]. One of the most important consequences of Schmidts Subspace Theorem is the niteness of nondegenerate solutions of the equation x1 + + xn = 1, where the unknowns take integer values (or S-integer values) in a number eld. Here, non-degenerate means that no proper sub-sum vanishes. One main open question is to prove an eective version of this result. Schmidts Theorem, which is a generalization of Roths Theorem, is not eective. Only for n = 2 does one know bounds for the solutions of the S-unit equation x1 +x2 = 1, thanks to Bakers method (see [B1, Chap. 5], [La5, Chap. VI], [ST, Chap 1], [Se] and [La8]). One would like to extend Bakers method (or any other eective method) to the higherdimensional case. A generalization of the Marko spectrum to simultaneous approximation is not yet available. Even the rst step is missing. Given a positive integer n and real
261
numbers (1 , . . . , n ), not all of which are rational, dene cn = cn (1 , . . . , n ) to be the inmum of all c in the range 0 < c 1 for which q|qi pi |n < c has innitely many solutions. Then dene the n-dimensional simultaneous Diophantine approximation constant n to be the supremum of cn over tuples (1 , . . . . . . , n ) as above. Following [Fi], here is a summary of what is known about the rst values of the approximation constants. 1 = 0.447. . . (Hurwitz) 1 = 5 2 64 0.285. . . = 2 = 0.378. . . (Cassels and Nowak) 7 169 1 2 0.120. . . = 3 = 0.437. . . (Cusick and Spohn) 2( 2) 5 11 The question remains open as to whether there are pairs with an approximation constant larger than 2/7 (see [Br]). We now illustrate with Warings Problem the importance of proving eective Roth-type inequalities for irrational algebraic numbers. In 1770, a few months before J. L. Lagrange proved that every positive integer is the sum of at most four squares of integers, E. Waring ([Wa, Chap. 5, Theorem 47 (9)]) wrote: Every integer is a cube or the sum of two, three, . . . nine cubes; every integer is also the square of a square, or the sum of up to nineteen such; and so forth. Similar laws may be armed for the correspondingly dened numbers of quantities of any like degree. See also Note 15 of the translator in [Wa]. For k 2 dene g(k) as the smallest positive integer g such that any integer is the sum of g elements of the form xk with x 0. In other terms, for each positive integer n n = xk + + xk 1 m has a solution if m = g(k), while there is a n which is not the sum of g(k) 1 such k-th powers. Lagranges Theorem, which solved a conjecture of Bachet and Fermat, is g(2) = 4. Following Chap. IV of [N1], here are the values of g(k) for the rst integers k, with the name(s) of the author(s) and the date.
g(2) = 4 g(3) = 9 g(4) = 19 g(5) = 37 g(6) = 73 g(7) = 143
R. Balasubramanian J-M. Deshouillers J. Chen S. S. Pillai L. E. Dickson F. Dress 1986 1964 1940 1936
262
M. WALDSCHMIDT
It is easy to show that g(k) I(k). Indeed, write 3k = 2k q + r and consider the integer N = 2k q 1 = (q 1)2k + (2k 1)1k . Since N < 3k , writing N as a sum of k-th powers can involve no term 3k , and since N < 2k q, it involves at most (q 1) terms 2k , all others being 1k ; hence it requires a total number of at least (q 1) + (2k 1) = I(k) terms. L. E. Dickson and S. S. Pillai (see for instance [HW] or [N1, Chap. IV]) proved independently in 1936 that g(k) = I(k), provided that r = 3k 2k q satises r 2k q 2. Otherwise there is another formula for g(k). It has been shown that the condition r 2k q 2 is satised for 3 k 471 600 000, and K. Mahler proved that it is also true for any suciently large k. Hence g(k) = I(k) for these values of k. The problem is that Mahlers proof relies on a p-adic version of the ThueSiegelRoth Theorem, and therefore is not eective. So there is a gap, of which we dont even know the size. The conjecture, dating back to 1853, is g(k) = I(k) for any k 2, and this is true as soon as 3 2
k
3 4
where denote the distance to the nearest integer. As remarked by S. David, such an estimate (for suciently large k) follows not only from Mahlers estimate, but also from the abc Conjecture! In [M1] K. Mahler dened a Z-number as a non-zero real number such that the fractional part rn of (3/2)n satises 0 rn < 1/2 for any positive integer n. It is not known whether Z-numbers exist (see [FLP]). A related remark by J. E. Littlewood ([Guy, E18]) is that we are not yet able to prove that the fractional part of en does not tend to 0 as n tends to innity (see also Conjecture 2.14 below). A well known conjecture of Littlewood ([B1, Chap. 10, 1] and [PV]) asserts that for any pair (x, y) of real numbers and any > 0, there exists a positive integer q such that q qx qy < . According to G. Margulis (communication of G. Lachaud), the proofs in a 1988 paper by B. F. Skubenko (see MR 94d:11047) are not correct and cannot be xed. There are several open questions known as view obstruction Problems. One of them is the following. Given n positive integers k1 , . . . , kn , there exists a real number x such that 1 for 1 i n. ki x n+1 It is known that 1/(n + 1) cannot be replaced by a larger number [CP].
263
2.3. Irrationality and Linear Independence Measures. Given a real number , the rst Diophantine question is to decide whether is rational or not. This is a qualitative question, and it is remarkable that an answer is provided by a quantitative property of . It depends ultimately on the quality of rational Diophantine approximations to . Indeed, on the one hand, if is rational, then there exists a positive constant c = c() such that c p > q q
for any p/q Q. An admissible value for c is 1/b when = a/b. On the other hand, if is irrational, then there are innitely many rational numbers p/q such that 1 p < 2. 0< q q Hence, in order to prove that is irrational, it suces to prove that for any > 0 there is a rational number p/q such that 0< p < . q q
This is a rather weak requirement. There are rational approximations in 1/q 2 , and we need only to produce rational approximations better than the trivial ones in c/q. Accordingly one should expect that it is rather easy to prove the irrationality of a given real number. In spite of that, the class of interesting real numbers which are known to be irrational is not as large as one would expect [KZ]. For instance no proof of irrationality has been given so far for Eulers constant = lim
n
1+
1 1 1 + + + log n 2 3 n
= 0.577215. . . ,
(2.13)
et t4/5 dt = 4.590843. . .
k (n) n!
(k = 1, 2)
where k (n) =
(see [Guy, B14]). Here is another irrationality question raised by P. Erds and E. Straus in 1975 o (see [E2] and [Guy, E24]). Dene an irrationality sequence as an increasing sequence
264
M. WALDSCHMIDT
(nk )k1 of positive integers such that, for any sequence (tk )k1 of positive integers, the real number 1 n k tk
k1
is irrational. On the one hand, it has been proved by Erds that (22 )k1 is an o irrationality sequence. On the other hand, the sequence (k!)k1 is not, since 1 1 = . k!(k + 2) 2
k1
An open question is whether an irrationality sequence must increase very rapidly. 1/2k No irrationality sequence (nk )k1 is known for which nk tends to 1 as k tends to innity. Many further open irrationality questions are raised in [E2]. Another related example is Conjecture 5.4 below. Assume now that the rst step has been completed and that we know our number is irrational. Then there are (at least) two directions for further investigation. (1) Considering several real numbers 1 , . . . , n , a fundamental question is to decide whether or not they are linearly independent over Q. One main example is to start with the successive powers of one number, 1, , 2 , . . . , n1 . The goal is to decide whether is algebraic of degree < n. If n is not xed, the question is whether is transcendental. This question, which is relevant also for complex numbers, will be considered in the next section. Observe also that the problem of algebraic independence is included here. It amounts to the linear independence of monomials. (2) Another direction of research is to consider a quantitative renement of the irrationality statement, namely an irrationality measure. We wish to bound from below the non-zero number | (p/q)| when p/q is any rational number; this lower bound will depend on as well as the denominator q of the rational approximation. In case where a statement weaker than an irrationality result is known, namely if one can prove only that at least one of n numbers 1 , . . . , n is irrational, then a quantitative renement will be a lower bound (in terms of q) for max 1 p1 pn , . . . , n q q ,
when p1 /q, . . . , pn /q are n rational numbers and q > 0 a common denominator. On the one hand, the study of rational approximation of real numbers is achieved in a satisfactory way for numbers whose regular4 continued fraction expansion is
4A regular continued fraction expansion
a0 +
1 1 a1 + a2 + b1 b2 a1 + a2 +
265
known. This is the case for rational numbers (!), for quadratic numbers, as well as for a small set of transcendental numbers, like e = [2, 1, 2, 1, 1, 4, 1, 1, 6, 1, 1, . . . ] = [2, {1, 2m, 1}m1 ] e = [7, 2, 1, 1, 3, 18, 5, 1, 1, 6, 30, 8, 1, 1, 9, 42, 11, . . . ] = [7, {3m 1, 1, 1, 3m, 12m + 6}m1 ] and e1/n = [1, n 1, 1, 1, 3n 1, 1, 1, 5n 1, 1, 1, . . . ] = [{1, (2m 1)n 1, 1}m1 ] for n > 1. On the other hand, even for a real number x for which an irregular continued fraction expansion is known, like log 2 = or n2 1 1 4 9 1+ 1+ 1+ 1+ 1+
2
1 9 25 49 (2n + 1)2 = , 4 1+ 2+ 2+ 2+ 2+ one does not know how well x can be approximated by rational numbers. No regular pattern has been observed or can be expected from the regular continued fraction of , = [3, 7, 15, 1, 292, 1, 1, 1, 2, 1, 3, 1, 14, 2, 1, 1, 2, 2, 2, 2, 1, 84, 2, 1, 1, 15, 3, 13, 1, 4, 2, 6, 6, 6, 1, . . . ],
nor from any number easily related to . One expects that for any > 0 there are constants C() > 0 and C () > 0 such that p p C() C () log 2 > 2+ and > 2+ q q q q hold for any p/q Q, but this is known only with larger exponents, namely 3.8913. . . and 8.0161. . . respectively (Rukhadze and Hata). The sharpest known exponent for an irrationality measure of 1 (3) = = 1.202056. . . n3
n1
is 5.513891. . . , while for (or for (2) = 2 /6) it is 5.441243. . . (both results due to Rhin and Viola). For a number like (1/4), the existence of absolute positive constants C and for which p C (1/4) > q q has been proved only recently [P4]. The similar problem for e is not yet solved. In other terms there is no proof so far that e is not a Liouville number. Earlier we distinguished two directions for research once we know the irrationality of some given numbers. Either, on the qualitative side, one studies the linear dependence relations, or else, on the quantitative side, one studies the quality of rational approximation. One can combine both. A quantitative version of a result
266
M. WALDSCHMIDT
of Q-linear independence of n real numbers 1 , . . . , n , is a lower bound, in terms of max{|p1 |, . . . , |pn |}, for p1 1 + + pn n when (p1 , . . . , pn ) is in Zn \ {0}. For specic classes of transcendental numbers, A. I. Galochkin [G], A. N. Korobov (Th. 1.22 of [FN, Chap. 1, 7]) and more recently P. Ivankov proved extremely sharp measures of linear independence (see [FN, Chap. 2, 6.2 and 6.3]). A general and important problem is to improve the known measures of linear independence for logarithms of algebraic numbers, as well as elliptic logarithms, Abelian logarithms, and more generally logarithms of algebraic points on commutative algebraic groups. For instance the conjecture that e is not a Liouville number should follow from improvements of known linear independence measures for logarithms of algebraic numbers. The next step, which is to obtain sharp measures of algebraic independence for transcendental numbers, will be considered later (see Sections 4.2 and 4.3). The so-called Mahler Problem (see [W8, 4.1]) is related to linear combination of logarithms |b log a|. Conjecture 2.14 (Mahler). There exists an absolute constant c > 0 such that log a > ac for all integers a 2. Equivalently, |a eb | > ac for some absolute constant c > 0 for all integers a, b > 1. A stronger conjecture is suggested in [W8] (4.1), log a > (log a)c for some absolute constant c > 0 for all integers a 3, or equivalently |a eb | > bc for some absolute constant c > 0 for all integers a, b > 1. So far the best known estimate is |a eb | > ec(log a)(log b) , so the problem is to replace the product (log a)(log b) in the exponent by the sum log a + log b. Such explicit lower bounds have interest in theoretical computer science [MT]. Another topic which belongs to Diophantine approximation is the theory of equidistributed sequences. For a positive integer r 2, a normal number in base r is a real number such that the sequence (xrn )n1 is equidistributed modulo 1. Almost all real numbers for Lebesgue measure are normal (i. e., normal in basis r for any r > 1), but it is not known whether any irrational real algebraic number is normal to any integer basis, and it is also not known whether there is an integer r for which any number like e, , (3), (1/4), , G, e + , e is normal in basis r (see [Ra]). Further studies by D. H. Bailey and M. E. Crandall have recently been advanced by J. C. Lagarias in [L].
267
The digits of the expansion (in any basis 2) of an irrational, real, algebraic number should be equidistributed in particular any digit should appear innitely often. But even the following special case is unknown. Conjecture 2.15 (Mahler). Let (n )n0 be a sequence of elements in {0, 1}. Assume that the real number n 3n
n0
is irrational, then it is transcendental. 3. Transcendence When K is a eld and k a subeld, we denote by trdegk K the transcendence degree of the extension K/k. In the case k = Q we write simply trdeg K (see [La9, Chap. VIII, 1]). 3.1. Schanuels Conjecture. We concentrate here on problems related to transcendental number theory. To start with, we consider the classical exponential function ez = exp(z). A recent reference on this topic is [W6]. Schanuels Conjecture is a simple but far-reaching statement see the historical note to Chap. III of [La1]. Conjecture 3.1 (Schanuel). Let x1 , . . . , xn be Q-linearly independent complex numbers. Then the transcendence degree over Q of the eld Q x1 , . . . , xn , ex1 , . . . . . . , exn is at least n. According to S. Lang ([La1, p. 31]): From this statement, one would obtain most statements about algebraic independence of values of et and log t which one feels to be true. See also [La2, p. 638639] and [Ri2, 10.7.G]. For instance the following statements [Ge1] are consequences of Conjecture 3.1. Question. Let 1 , . . . , n be Q-linearly independent algebraic numbers and let log 1 , . . . , log m be Q-linearly independent logarithms of algebraic numbers. Then the numbers e1 , . . . , en , log 1 , . . . , log m are algebraically independent over Q. Question. Let 1 , . . . , n be algebraic numbers with 1 = 0 and let log 1 , . . . . . . , log m be logarithms of algebraic numbers with log 1 = 0 and log 2 = 0. Then the numbers e
1 e2 e .. .n1 en
and
.. 1 2
.m
are transcendental, and there is no nontrivial algebraic relation between such numbers. A quantitative renement of Conjecture 3.1 is suggested in [W4, Conjecture 1.4].
268
M. WALDSCHMIDT
A quite interesting approach to Schanuels Conjecture is given in [Ro5] where D. Roy states the next conjecture which he shows to be equivalent to Schanuels one. Let D denote the derivation D= + X1 X0 X1 over the ring C[X0 , X1 ]. The height of a polynomial P C[X0 , X1 ] is dened as the maximum of the absolute values of its coecients. Conjecture 3.2 (Roy). Let k be a positive integer, y1 , . . . , yk complex numbers which are linearly independent over Q, 1 , . . . , k non-zero complex numbers and s0 , s1 , t0 , t1 , u positive real numbers satisfying 1 max{1, t0 , 2t1 } < min{s0 , 2s1 } and max{s0 , s1 + t1 } < u < (1 + t0 + t1 ). 2 Assume that, for any suciently large positive integer N , there exists a non-zero polynomial PN Z[X0 , X1 ] with partial degree N t0 in X0 , partial degree N t1 in X1 and height eN which satises
k k
(Dk PN )
j=1
mj yj ,
j=1
mj
exp(N u )
for any non-negative integers k, m1 , . . . , mk with k N s0 and max{m1 , . . . , mk } N s1 . Then trdeg Q(y1 , . . . , yk , 1 , . . . , k ) k. This work of Roys also provides an interesting connexion with other open problems related to the Schwarz Lemma for complex functions of several variables (see [Ro8, Conjectures 6.1 and 6.3]). The most important special case of Schanuels Conjecture is the Conjecture of algebraic independence of logarithms of algebraic numbers. Conjecture 3.3 (Algebraic Independence of Logarithms of Algebraic Numbers). Let 1 , . . . , n be Q-linearly independent complex numbers. Assume that the numbers e1 , . . . , en are algebraic. Then the numbers 1 , . . . , n are algebraically independent over Q. An interesting reformulation of Conjecture 3.3 is due to D. Roy [Ro4]. Denote by L the set of complex numbers for which e is algebraic. Hence L is a Q-vector subspace of C. Roys statement is: Conjecture. For any algebraic subvariety V of Cn dened over the eld Q of algebraic numbers, the set V Ln is the union of the sets E Ln , where E ranges over the set of vector subspaces of Cn which are contained in V . Such a statement is reminiscent of several of Langs conjectures in Diophantine geometry (e. g., [La8, Chap. I, 6, Conjectures 6.1 and 6.3]). Not much is known about the algebraic independence of logarithms of algebraic numbers, apart from the work of D. Roy on the rank of matrices whose entries are either logarithms of algebraic numbers, or more generally linear combinations
269
of logarithms of algebraic numbers. We refer to [W6] for a detailed study of this question as well as related ones. Conjecture 3.3 has many consequences. The next three ones are suggested by the work of D. Roy ([Ro1] and [Ro2]) on matrices whose entries are linear combinations of logarithms of algebraic numbers (see also [W6, Conjecture 11.17, 12.4.3 and Exercise 12.12]). Consider the Q-vector space L spanned by 1 and L. In other words L is the set of complex numbers which can be written 0 + 1 log 1 + + n log n , where 0 , 1 , . . . , n are algebraic numbers, 1 , . . . , n are non-zero algebraic numbers, and nally log 1 , . . . , log n are logarithms of 1 , . . . , n respectively. Conjecture 3.4 (Strong Four Exponentials Conjecture). Let x1 , x2 be two Qlinearly independent complex numbers and y1 , y2 be also two Q-linearly independent complex numbers. Then at least one of the four numbers x1 y1 , x1 y2 , x2 y1 , x2 y2 does not belong to L. The following special case is also open. Conjecture 3.5 (Strong Five Exponentials Conjecture). Let x1 , x2 be two Qlinearly independent complex numbers, and y1 , y2 be also two Q-linearly independent complex numbers. Further, let ij (i = 1, 2, j = 1, 2), 1 and 2 be six algebraic numbers with 1 = 0. Assume that the ve numbers ex1 y1 11 , ex1 y2 12 , ex2 y1 21 , ex2 y2 22 , e(1 x1 /x2 )2 are algebraic. Then all ve exponents vanish, xi yj = ij (i = 1, 2, j = 1, 2) and 1 x1 = 2 x2 .
A consequence of Conjecture 3.5 is the solution of the open problem of the 2 2 transcendence of the number e , and more generally of log = e when is a non-zero algebraic number and = log a non-zero logarithm of . The next conjecture is proposed in [Ro4]. Conjecture 3.6 (Roy). For any 4 4 skew-symmetric matrix M with entries in L and rank 2, either the rows of M are linearly dependent over Q, or the column space of M contains a non-zero element of Q4 . Finally, a special case of Conjecture 3.6 is the well known Four Exponentials Conjecture due to Schneider ([Schn, Chap. V, end of 4, Problem 1]), S. Lang ([La1, Chap. II, 1], [La2, p. 638]) and K. Ramachandra ([R, II, 4]). Conjecture 3.7 (Four Exponentials Conjecture). Let x1 , x2 be two Q-linearly independent complex numbers and y1 , y2 also be two Q-linearly independent complex numbers. Then at least one of the four numbers exp(xi yj ) is transcendental. (i = 1, 2, j = 1, 2)
270
M. WALDSCHMIDT
The four exponentials Conjecture can be stated as follows: consider a 2 2 matrix whose entries are logarithms of algebraic numbers, M= log 11 log 21 log 12 ; log 22
assume that the two rows of this matrix are linearly independent over Q (in C2 ), and also that the two columns are linearly independent over Q; then the rank of this matrix is 2. We refer to [W6] for a detailed discussion of this topic, including the notion of structural rank of a matrix and the result, due to D. Roy, that Conjecture 3.3 is equivalent to a conjecture on the rank of matrices whose entries are logarithms of algebraic numbers. A classical problem on algebraic independence of algebraic powers of algebraic numbers has been raised by A. O. Gelfond [Ge2] and Th. Schneider [Schn, Chap. V, end of 4, Problem 7]. The data are an irrational algebraic number of degree d, and a non-zero algebraic number with a non-zero logarithm log . We write z in place of exp{z log }. Gelfonds problem is Conjecture 3.8 (Gelfond). The two numbers log are algebraically independent over Q. Schneiders question is Conjecture 3.9 (Schneider). The d 1 numbers , , . . . , are algebraically independent over Q. The rst partial results towards a proof of Conjecture 3.9 are due to A. O. Gelfond [Ge3]. For the more recent ones, see [NP, Chap. 13 and 14]. Combining both questions 3.8 and 3.9 yields a stronger conjecture. Conjecture 3.10 (GelfondSchneider). The d numbers log , , , . . . , are algebraically independent over Q. Partial results are known. They deal, more generally, with the values of the usual exponential function at products xi yj , when x1 , . . . , xd and y1 , . . . , y are Q-linearly independent complex (or p-adic) numbers. The six exponentials Theorem states that, in these circumstances, the d numbers exi yj (1 i d, 1 j ) cannot all be algebraic if d > d + . Assuming stronger conditions on d and , namely d 2(d + ), one deduces that two at least of these d numbers exi yj are algebraically independent over Q. Other results are available involving, in addition to exi yj , either the numbers x1 , . . . , xd themselves, or y1 , . . . , y , or both. But an interesting point is that, if we wish to obtain a higher transcendence degree, say to obtain that three at least of the numbers exi yj are algebraically independent over Q, one needs a further assumption, which is a measure of linear independence over Q
2 d1 2 d1
and
271
for the tuple x1 , . . . , xd as well as for the tuple y1 , . . . , y . To remove this so-called technical hypothesis does not seem to be an easy challenge (see [NP, Chap. 14, 2.2 and 2.3]). The need for such a technical hypothesis seems to be connected with the fact that the actual transcendence methods produce not only a qualititative statement (lower bound for the transcendence degree), but also quantitative statements (transcendence measures and measures of algebraic independence). Several complex results have not yet been established in the ultrametric situation. Two noticeable instances are Conjecture 3.11 (p-adic analogue of LindemannWeierstrasss Theorem). Let 1 , . . . , n be p-adic algebraic numbers in the domain of convergence of the padic exponential function expp . Then the n numbers expp 1 , . . . , expp n are algebraically independent over Q. Conjecture 3.12 (p-adic analogue of an algebraic independence result of Gelfond). Let be a non-zero algebraic number in the domain of convergence of the p-adic logarithm logp , and let be a p-adic cubic algebraic number, such that logp is in the domain of convergence of the p-adic exponential function expp . Then = expp ( logp ) are algebraically independent over Q. The p-adic analogue of Conjecture 3.3 would solve Leopoldts Conjecture on the p-adic rank of the units of an algebraic number eld [Le] (see also [N2] and [Gra]), by proving the nonvanishing of the p-adic regulator. Algebraic independence results for the values of the exponential function (or more generally for analytic subgroups of algebraic groups) in several variables have already been established, but they are not yet satisfactory. The conjectures stated in [W2, p. 292293] as well as those of [NP, Chap. 14, 2] are not yet proved. One of the main obstacles is the above-mentioned open problem with the technical hypothesis. The problem of extending the LindemannWeierstrass Theorem to commutative algebraic groups is not yet completely solved (see conjectures by P. Philippon in [P1]). Algebraic independence proofs use elimination theory. Several methods are available; one of them, developed by Masser, Wstholz and Brownawell, relies on the u Hilbert Nulstellensatz. In this context we quote the following conjecture of Blum, Cucker, Shub and Smale (see [Sm] and [NP, Chap. 16, 6.2]), related to the open problem P = N P ? [J]. Conjecture 3.13 (Blum, Cucker, Shub and Smale). Given an absolute constant c and polynomials P1 , . . . , Pm with a total of N coecients and no common complex zeros, there is no program to nd, in at most N c step, the coecients of polynomials Ai satisfying Bzouts relation, e A1 P1 + + Am Pm = 1. and = expp ( 2 logp )
2
272
M. WALDSCHMIDT
In connexion with complexity in theoretical computer science, W. D. Brownawell suggests investigating Diophantine approximation from a new point of view in [NP, Chap. 16, 6.3]. Complexity theory may be related to a question raised by M. Kontsevich and D. Zagier in [KZ]. They dened a period as a complex number whose real and imaginary part are values of absolutely convergent integrals of rational functions with rational coecients over domains of Rn given by polynomial (in)equalities with rational coecients. Problem 3 in [KZ] is to produce at least one number which is not a period. This is the analogue for periods of Liouvilles Theorem for algebraic numbers. A more dicult question is to prove that specic numbers like e, 1/,
(where is Eulers constant) are not periods. Since every algebraic number is a period, a number which is not a period is transcendental. Another important tool missing for transcendence proofs in higher dimension is a Schwarz Lemma in several variables. The following conjecture is suggested in [W1, 5]. For a nite subset of Cn and a positive integer t, denote by t () the least total degree of a non-zero polynomial P in C[z1 , . . . , zn ] which vanishes on with multiplicity at least t, z1
1
zn
P (z) = 0,
for any z and = (1 , . . . , n ) Nn with 1 + + n < t. Further, when f is an analytic function in an open neighborhood of a closed polydisc |zi | r (1 i n) in Cn , denote by f (r) the average mass of the set of zeroes of f in that polydisc (see [BL]). Conjecture 3.14. Let be a nite subset of Cn , and be a positive number. There exists a positive number r0 (, ) such that, for any positive integer t and any entire function f in Cn which vanishes on with multiplicity t, f (r) t () t for r r0 (, ).
The next question is to compute r0 (, ). One may expect that for a chunk of a nitely generated subgroup of Cn , say = s1 y1 + + s y : (s1 , . . . , s ) Z , |sj | S (1 j ) Cn , an admissible value for the number r0 (, ) will depend only on , y1 , . . . , y , but not on S. This would have interesting applications, especially in the special case = n + 1. Finally we refer to [Chu] for a connexion between the numbers t (S) and Nagatas work on Hilberts 14th Problem. 3.2. Multiple Zeta Values. Many recent papers (see for instance [C]) are devoted to the study of algebraic relations among multiple zeta values,
s s n1 1 nk k , n1 >>nk 1
273
where (s1 , . . . , sk ) is a k-tuple of positive integers with s1 2. The main Diophantine conjecture, suggested by the work of D. Zagier, A. B. Goncharov, M. Kontsevich, M. Petitot, Minh Hoang Ngoc, K. Ihara, M. Kaneko and others (see [Z], [C] and [Zu]), is that all such relations can be deduced from the linear and quadratic ones arising in the shue and stue products (including the relations arising from the study of divergent series see [W7] for instance). For p 2, let Zp denote the Q-vector subspace of R spanned by the real numbers (s) satisfying s = (s1 , . . . , sk ) and s1 + + sk = p. Set Z0 = Q and Z1 = {0}. Then the Q-subspace Z spanned by all Zp , p 0, is a subalgebra of R, and part of the Diophantine conjecture states Conjecture 3.15 (Goncharov). As a Q-algebra, Z is the direct sum of Zp for p 0. In other terms, all algebraic relations should be consequences of homogeneous ones, involving values (s) with dierent s but with the same weight s1 + + sk . Assuming conjecture 3.15, the question of algebraic independence of the numbers (s) is reduced to the question of linear independence of the same numbers. The conjectural situation is described by the next conjecture of Zagier [Z] on the dimension dp of the Q-vector space Zp . Conjecture 3.16 (Zagier). For p 3, dp = dp2 + dp3 , with d0 = 1, d1 = 0, d2 = 1. That the actual dimensions of the spaces Zp are bounded above by the integers which are dened inductively in Conjecture 3.16 has been proved by T. Terasoma in [T], who expresses multiple zeta values as periods of relative cohomologies and uses mixed Tate Hodge structures (see also the work of A. G. Goncharov referred to in [T]). Further work on Conjectures 3.15 and 3.16 is due to J. Ecalle. In case k = 1 (values of the Riemann zeta function) the conjecture is Conjecture 3.17. The numbers , (3), (5), . . . , (2n + 1), . . . are algebraically independent over Q. So far the only known results on this topic [Fis] are: (2n) is transcendental for n1 (because is transcendental and (2n) 2n Q), (3) is irrational (Apry, 1978), e and For any > 0 the Q-vector space spanned by the n + 1 numbers 1, (3), (5), . . . , (2n + 1) has dimension 1 log n 1 + log 2
for n n0 () (see [Riv] and [BR]). For instance innitely many of these numbers (2n + 1) (n 1) are irrational. W. Zudilin proved that at least one of the four numbers (5), (7), (9), (11) is irrational.
274
M. WALDSCHMIDT
Further, more recent results are due to T. Rivoal and W. Zudilin. For instance, in a joint paper they have proved that innitely many numbers among (1)n (2n + 1)2s (s Z, s 1)
n1
are irrational, but, as pointed out earlier, the irrationality of Catalans constant G see (2.13) is still an open problem. It may turn out to be more ecient to work with a larger set of numbers, including special values of multiple polylogarithms,
n n z1 1 zk k . n s1 n sk 1 k
n1 >>nk 1
An interesting set of points z = (z1 , . . . , zk ) to consider is the set of k-tuples consisting of roots of unity. The function of a single variable, Lis (z) = ns 1 n1 >>nk 1 1 z n1 , n sk k
is worth of study from a Diophantine point of view. For instance, Catalans constant mentioned above is the imaginary part of Li2 (i), Li2 (i) =
n1
1 in = (2) + iG. n2 8
1 n4 k 2 1
n1
= (3)2
4 6 , 2835
n2 2n
2 1 (log 2)2 12 2
1
nk1
2n n2 k
= (3)
1 2 log 2, 12
is known so far. According to P. Bundschuh [Bun], the transcendence of the numbers 1 ns 1 n=2 for even s 4 is a consequence of Schanuels Conjecture 3.1. For s = 2 the sum is 3/4, and for s = 4 the value is (7/8) (/4) coth , which is a transcendental number since and e are algebraically independent over Q (Yu. V. Nesterenko [NP]). Nothing is known about the arithmetic nature of the values of the Riemann zeta function at rational or algebraic points which are not integers.
275
3.3. Gamma, Elliptic, Modular, G and E-Functions. On the one hand, the transcendence problem of the values of the Euler beta function at rational points was solved as early as 1940, by Th. Schneider. For any rational numbers a and b which are not integers and such that a + b is not an integer, the number (a, b) = (a)(b) (a + b)
is transcendental. On the other hand, transcendence results for the values of the gamma function itself are not so precise: apart from G. V. Chudnovskys results, which imply the transcendence of (1/3) and (1/4) (and Lindemanns result on the transcendence of which implies that (1/2) = is also transcendental), not much is known. For instance, as we said earlier, there is no proof so far that (1/5) is transcendental. This is because the Fermat curve of exponent 5, viz. x5 + y 5 = 1, has genus 2. Its Jacobian is an Abelian surface, and the algebraic independence results known for elliptic curves like x3 + y 3 = 1 and x4 + y 4 = 1 which were sucient for dealing with (1/3) and (1/4), are not yet known for Abelian varieties (see [Grin]). One might expect that Nesterenkos results (see [NP, Chap. 3]) on the algebraic independence of , (1/4), e and of , (1/3), e 3 should be extended as follows. Conjecture 3.18. At least three of the four numbers , (1/5), (2/5), e are algebraically independent over Q. So the challenge is to extend Nesterenkos results on modular functions in one variable (and elliptic curves) to several variables (and Abelian varieties). This may be one of the easiest questions to answer on this topic (but it is still open). But one may ask for a general statement which would produce all algebraic relations between gamma values at rational points. Here is a conjecture of Rohrlich [La4]. Dene 1 G(z) = (z). 2 According to the multiplication theorem of Gauss and Legendre [WW, 12.15], for each positive integer N and for each complex number x such that N x 0 (mod Z),
N 1 5
G x+
i=0
i N
The gamma function has no zero and denes a map from C \ Z to C . We restrict that function to Q \ Z and we compose it with the canonical map C C /Q which amounts to considering its values modulo the algebraic numbers. The composite map has period 1, and the resulting mapping, G: Q C \ {0} , Z Q
276
M. WALDSCHMIDT
G x+
i=0
i N
= G(N x)
for x
Rohrlichs Conjecture ([La4], [La6, Chap. II, Appendix, p. 66]) asserts that Conjecture 3.19 (Rohrlich). G is a universal odd distribution with values in groups where multiplication by 2 is invertible. In other terms, any multiplicative relation between gamma values at rational points b/2
aQ
(a)ma Q
with b and ma in Z can be derived for the standard relations satised by the gamma function. This leads to the question whether the distribution relations, the oddness relation and the functional equations of the gamma function generate the ideal over Q of all algebraic relations among the values of G(x) for x Q. In [NP] (Chap. 3, 1, Conjecture 1.11) Yu. V. Nesterenko proposed another conjectural extension of his algebraic independence result on Eisenstein series of weight 2, 4 and 6: P (q) = 1 24 nq n = 1 24 1 (n)q n , 1 qn n=1 n=1 n3 q n = 1 + 240 3 (n)q n , 1 qn n=1 n=1 n5 q n = 1 504 5 (n)q n . 1 qn n=1 n=1
Conjecture 3.20 (Nesterenko). Let C have positive imaginary part. Assume that is not quadratic. Set q = e2i . Then at least 4 of the 5 numbers , q, P (q), Q(q), R(q) are algebraically independent. Finally we remark that essentially nothing is known about the arithmetic nature of the values of either the beta or the gamma function at algebraic irrational points. A wide range of open problems in transcendental number theory, including not only Schanuels Conjecture 3.1 and Rohrlichs Conjecture 3.19 on the values of the gamma function, but also a conjecture of Grothendieck on the periods of an algebraic variety (see [La1, Chap. IV, Historical Note], [La2, p. 650], [A1, p. 6] e and [Ch, 3]), are special cases of very general conjectures due to Y. Andr [A2], which deal with periods of mixed motives. A discussion of Andrs conjectures for e certain 1-motives related to the products of elliptic curves and their connexions with elliptic and modular functions is given in [Ber]. Here is a special case of the elliptico-toric Conjecture in [Ber].
277
Conjecture 3.21 (Bertolin). Let E1 , . . . , En be pairwise non isogeneous elliptic curves with modular invariants j(Eh ). For h = 1, . . . , n, let 1h , 2h be a pair of fundamental periods of h where 1h , 2h are the associated quasi-periods, Pih points on Eh (C) and pih (resp. dih ) elliptic integrals of the rst (resp. second ) kind associated to Pih . Dene h = [kh : Q] and let dh be the dimension of the kh subspace of C/(kh 1h + kh 2h ) spanned by p1h , . . . , prh h . Then the transcendence degree of the eld Q is at least 2
h=1
1irh 1hn
dh + 4
h=1
1 n + 1. h
A new approach to Grothendiecks Conjecture via Siegels G-functions was ine troduced in [A1, Chap. IX]. A development of this method led Y. Andr to his conjecture on the special points on Shimura varieties [A1, Chap. X, 4], which gave rise to the AndrOort Conjecture [O] (for a discussion of this topic, including a e precise denition of Hodge type, together with relevant references, see [Co]). Conjecture 3.22 (AndrOort). Let Ag (C) denote the moduli space of principally e polarized complex Abelian varieties of dimension g. Let Z be an irreducible algebraic subvariety of Ag (C) such that the complex multiplication points on Z are dense for the Zariski topology. Then Z is a subvariety of Ag (C) of Hodge type. Conjecture 3.22 is a far-reaching generalization of Schneiders Theorem on the transcendence of j( ), where j is the modular invariant and an algebraic point in the Poincar upper half plane H, which is not imaginary quadratic ([Schn, Chap. II, e 4, Th. 17]). We also mention a related conjecture of D. Bertrand (see [NP, Chap. 1, 4, Conjecture 4.3]) which may be viewed as a nonholomorphic analogue of Schneiders result and which would answer the following question raised by N. Katz. Question. Assume that a lattice L = Z1 + Z2 in C has algebraic invariants g2 (L) and g3 (L) and no complex multiplication. Does this implies that the number G (L) = lim 2 is transcendental ? Many open transcendence problems dealing with elliptic functions are consequences of Andrs conjectures (see [Ber]), most of which are likely to be very hard. e The next one, which is still open, may be easier, since a number of partial results are already known, as a result of the work of G. V. Chudnovsky and others (see [Grin]). Conjecture 3.23. Given an elliptic curve with Weierstrass equation y 2 = 4x3 g2 x g3 , a non-zero period , the associated quasi-period of the zeta function and a complex number u which is not a pole of , trdeg Q g2 , g3 , /, (u), (u) (/)u 2.
s0 L\{0}
2 ||s
278
M. WALDSCHMIDT
Given a lattice L = Z1 + Z2 in C with invariants g2 (L) and g3 (L), denote by i = L (z + i ) L (z) (i = 1, 2) the corresponding fundamental quasi-periods of the Weierstrass zeta function. Conjecture 3.23 implies that the transcendence degree over Q of the eld Q g2 (L), g3 (L), 1 , 2 , 1 , 2 is at least 2.This would be optimal in the CM case, while in the non CM case, we expect it to be 4. These lower bounds are given by the period conjecture of Grothendieck applied to an elliptic curve. According to [Di2, Conjectures 1 and 2, p. 187], the following special case of Conjecture 3.23 can be stated in two equivalent ways: either in terms of values of elliptic functions, or in terms of values of Eisenstein series E2 , E4 and E6 (which are P , Q and R in Ramanujans notation). Conjecture. For any lattice L in C without complex multiplication and for any non-zero period of L, trdeg Q g2 (L), g3 (L), / , / 2. Conjecture. For any H which is not imaginary quadratic, trdeg Q E2 ( ), 2 E4 ( ), 3 E6 ( ) 2. Moreover, each of these two statements implies the following one, which is stronger than one of Langs conjectures ([La2, p. 652]). Conjecture. For any H which is not imaginary quadratic, trdeg Q j( ), j ( ), j ( ) 2. Further related open problems are proposed by G. Diaz in [Di1] and [Di2], in connexion with conjectures due to D. Bertrand on the values of the modular function J(q), where j( ) = J(e2i ) (see [Bert2] as well as [NP, Chap. 1, 4 and Chap. 2, 4]). Conjecture 3.24 (Bertrand). Let q1 , . . . , qn be non-zero algebraic numbers in the unit open disc such that the 3n numbers J(qi ), DJ(qi ), D2 J(qi ) (i = 1, . . . , n) are algebraically dependent over Q. Then there exist two indices i = j (1 i n, 1 j n) such that qi and qj are multiplicatively dependent. Conjecture 3.25 (Bertrand). Let q1 and q2 be two non-zero algebraic numbers in the unit open disc. Suppose that there is an irreducible element P Q[X, Y ] such that P J(q1 ), J(q2 ) = 0. Then there exist a constant c and a positive integer s such that P = cs , where s is the modular polynomial of level s. Moreover q1 and q2 are multiplicatively dependent. Among Siegels G-functions are the algebraic functions. Transcendence methods produce some information, in particular in connexion with Hilberts Irreducibility Theorem. Let f Z[X, Y ] be a polynomial which is irreducible in Q(X)[Y ]. According to Hilberts Irreducibility Theorem, the set of positive integers n such that
279
P (n, Y ) is irreducible in Q[Y ] is innite. Eective upper bounds for an admissible value for n have been studied (especially by M. Fried, P. D`bes and U. Zannier), e but do not yet answer the next question. Question 3.26. Is there such a bound depending polynomially on the degree and height of P ? Such questions are also related to the Galois inverse Problem [Se]. Also the polylogarithms zm , Lis (z) = ns
n1
where s is a positive integer, are G-functions; unfortunately no way has yet been found to use the SiegelShidlovskii method to prove the irrationality of the values of the Riemann zeta function ([FN, Chap. 5, 7, p. 247]). With G-functions, the other class of analytic functions introduced by C. L. Siegel in 1929 is the class of E-functions, which includes the hypergeometric ones. One main open question is the arithmetic nature of the values at algebraic points of hypergeometric functions with algebraic parameters,
2 F1
, z =
n0
dened for |z| < 1 and {0, 1, 2, . . . }. In 1949, C. L. Siegel ([Si2, Chap. 2, 9, p. 54 and 58]; see also [FS, p. 62] and [FN, Chap. 5, 1.2]) asked whether any E-function satisfying a linear dierential equation with coecients in C(z) can be expressed as a polynomial in z and a nite number of hypergeometric E-functions or functions obtained from them by a change of variables of the form z z with algebraic s? Finally, we quote from [W4]: a folklore conjecture is that the zeroes of the Riemann zeta function (say their imaginary parts, assuming it > 0) are algebraically independent. As suggested by J-P. Serre, one might also be tempted to consider The eigenvalues of the zeroes of the hyperbolic Laplacian in the upper half plane modulo SL2 (Z) (i. e., to study the algebraic independence of the zeroes of the Selberg zeta function). The eigenvalues of the Hecke operators acting on the corresponding eigenfunctions (Maass forms). 3.4. Fibonacci and Miscellanea. Many further open problems arise in transcendental number theory. An intriguing question is to study the arithmetic nature of real numbers given in terms of power series involving the Fibonacci sequence Fn+2 = Fn+1 + Fn , F0 = 0, F1 = 1.
Several results are due to P. Erds, R. Andr-Jeannin, C. Badea, J. Sndor, P. Bundo e a schuh, A. Peth, P. G. Becker, T. Tpfer, D. Duverney, Ku. et Ke. Nishioka, I. Sho o iokawa and T. Tanaka. It is known that the number 1 =1 F F n=1 n n+2
280
M. WALDSCHMIDT
is rational, while
7 5 1 = , Fn 2 n=0 2
and
5 2
and
n1
1 F1 F2 Fn
is irrational, but it is not known whether they are algebraic or transcendental. The numbers
1 F2n1
n=1
1 , 2 Fn n=1
n , F2n n=1
are all transcendental (further results of algebraic independence are known). The rst challenge here is to formulate a conjectural statement which would give a satisfactory description of the situation. There is a similar situation for innite sums n f (n) where f is a rational function [Ti4]. While 1 =1 n(n + 1) n=1 and
n=0
1 3 1 1 + + 4n + 1 4n + 2 4n + 3 4n + 4
=0
are rational numbers, the sums 1 = log 2, (2n + 1)(2n + 2) n=0 1 2 = , n2 6 n=1 and 1 (6n + 1)(6n + 2)(6n + 3)(6n + 4)(6n + 5)(6n + 6) n=0 1 (192 log 2 81 log 3 7 3) 4320 s (see Secare transcendental. The simplest example of the Euler sums nn tion 3.2) illustrates the diculty of the question. Here again, even a suciently =
1 1 e + e = + , n2 + 1 2 2 e e n=0
281
general conjecture is missing. One may remark that there is no known algebraic irrational number of the form P (n) , Q(n)
n0 Q(n)=0
where P and Q are non-zero polynomials having rational coecients and deg Q 2 + deg P . The arithmetic study of the values of power series suggests many open problems. We shall only mention a few of them. The next question is due to K. Mahler [M3]. Question 3.27 (Mahler). Are there entire transcendental functions f (z) such that if x is a Liouville number then so is f (x)? The study of integral valued entire functions gives rise to several open problems; we quote only one of them which arose in the work of D. W. Masser and F. Gramain on entire functions f of one complex variable which map the ring of Gaussian integers Z[i] into itself. The initial question (namely to derive an analogue of Plyas Theorem in this setting) has been solved by F. Gramain in [Gr] (following o previous work of Fukasawa, Gelfond, Gruman and Masser). If f is not a polynomial, then 1 lim sup 2 log |f |r 2e r r Here, |f |r = max |f (z)|.
|z|=r
Preliminary works on this estimate gave rise to the following problem, which is still unsolved. For each integer k 2, let Ak be the minimal area of a closed disk in R2 containing at least k points of Z2 , and for n 2 dene
n
n = log n +
k=2
1 . Ak
The limit = limn n exists (it is an analogue in dimension 2 of the Euler constant), and the best known estimates for it are [GW] 1.811. . . < < 1.897. . . (see also [Fi]). F. Gramain conjectures that =1+ where is Eulers constant and L(s) =
n0
4 L(1) + L (1) ,
282
M. WALDSCHMIDT
is the L function of the quadratic eld Q(i) (Dirichlet beta function). Since L(1) = and 4 log(2n + 1) L (1) = (1)n+1 = 3 log + 2 log 2 + 4 log (1/4) , 2n + 1 4
n0
Gramains conjecture is equivalent to = 1 + 3 log + 2 log 2 + 2 4 log (1/4) = 1.822825. . . Other problems related to the lattice Z[i] are described in the section On the borders of geometry and arithmetic of [Sie1]. 4. Heights For a non-zero polynomial f C[X] of degree d,
d
f (X) = a0 X d + a1 X d1 + + ad1 X + ad = a0
i=1
(X i ),
dene its usual height by H(f ) = max{|a0 |, . . . , |ad |} and its Mahlers measure by
d 1
M (f ) = |a0 |
i=1
max{1, |i |} = exp
0
log |f (e2it )| dt .
The equality between these two formulae follows from Jensens formula (see [M2, Chap. I, 7], as well as [W6, Chap. 3] and [S]; the latter includes an extension to several variables). When is an algebraic number with minimal polynomial f Z[X], dene its Mahlers measure by M () = M (f ) and its usual height by H() = H(f ). Further, if has degree d, dene its logarithmic height as 1 h() = log M (). d Furthermore, if 1 , . . . , d are the complex roots of f (also called the complex conjugates of ), then the house of is = max{|1 |, . . . , |d |}. The height of an algebraic number is the prototype of a whole collection of height, like the height of projective point (1 : 1 : : n ) Pn which is denoted by h(1 : 1 : : n ) (see for instance [W6, 3.2]) and the height of a subvariety h(V ) (see for instance [D1] and [D2]). Further, if = (1 , . . . , n ) is a n-tuple of multiplicatively independent algebraic numbers, () denotes the minimum degree of a non-zero polynomial in Q[X1 , . . . , Xn ] which vanishes at . A side remark is that Mahlers measure of a polynomial in a single variable with algebraic coecients is an algebraic number. The situation is much more intricate
283
for polynomials in several variables and suggests to further very interesting open problems [Boy1], [Boy2]. 4.1. Lehmers Problem. The smallest known value for dh(), which was found in 1933 by D. H. Lehmer, is log 0 = 0.162357. . . , where 0 = 1.176280. . . is the real root5 of the degree 10 polynomial X 10 + X 9 X 7 X 6 X 5 X 4 X 3 + X + 1. D. H. Lehmer asked whether it is true that for every positive there exists an algebraic integer for which 1 < M () < 1 + ? Conjecture 4.1 (Lehmers Problem). There exists a positive absolute constant c such that, for any non-zero algebraic number which is not a root of unity, M () 1 + c. Equivalently, there exists a positive absolute constant c such that, for any non-zero algebraic number of degree at most d which is not a root of unity, c h() . d Since h() log , the following statement [SZ] is a weaker assertion than Conjecture 4.1. Conjecture 4.2 (SchinzelZassenhaus). There exists an absolute constant c > 0 such that, for any non-zero algebraic integer of degree d which is not a root of unity, c 1+ . d Lehmers Problem is related to the multiplicative group Gm . Generalizations to Gn have been considered by many authors (see for instance [Bert1] and [Sch2]). In m [AD1, Conjecture 1.4], F. Amoroso and S. David extend Lehmers Problem 4.1 to simultaneous approximation. Conjecture 4.3 (AmorosoDavid). For each positive integer n 1 there exists a positive number c(n) having the following property. Let 1 , . . . , n be multiplicatively independent algebraic numbers. Dene D = [Q(1 , . . . , n ) : Q]. Then
n
h(i )
i=1
c(n) . D
The next statement ([AD1, Conjecture 1.3] and [AD2, Conjecture 1.3]) is stronger. Conjecture 4.4 (AmorosoDavid). For each positive integer n 1 there exists a positive number c(n) such that, if = (1 , . . . , n ) is a n-tuple of multiplicatively independent algebraic numbers, then c(n) h(1 : 1 : : n ) . ()
5Further properties of this smallest known Salem number are described by D. Zagier in his paper Special values and functional equations of polylogarithms, Appendix A of Structural properties of polylogarithms, ed. L. Lewin, Mathematical Surveys and Monographs, vol. 37, Amer. Math. Soc. 1991, pp. 377400.
284
M. WALDSCHMIDT
Many open questions are related to the height of subvarieties [D1], [D2]. The next one, dealing with the height of subvarieties of Gn and proposed by F. Amoroso m and S. David in [AD2, Conjecture 1.4] (see also Conjecture 1.5 of [AD2], which is due to S. David and P. Philippon [DP]), is more general than Conjecture 4.4. Conjecture 4.5 (AmorosoDavid). For each integer n 1 there exists a positive constant c(n) such that, for any algebraic subvariety V of Gn which is dened m over Q, which is Q-irreducible, and which is not a union of translates of algebraic subgroups by torsion points, h(V ) c(n) deg(V )(sdim V 1)/(sdim V ) , where s is the dimension of the smallest algebraic subgroup of Gn containing V . m Let V be an open subset of C. The LehmerLangevin constant of V is dened as L(V ) = inf M ()1/[Q():Q] , where ranges over the set of non-zero and non-cyclotomic algebraic numbers, , lying with all their conjugates outside of V . It was proved by M. Langevin in 1985 that L(V ) > 1 as soon as V contains a point on the unit circle |z| = 1. Problem 4.6. For (0, ), dene V = {reit : r > 0, |t| > }. Compute L(V ) in terms of . The solution is only known for a very few values of . In 1995 G. Rhin and C. Smyth [RS] computed L(V ) for nine values of , including L(V/2 ) = 1.12. . . In a dierent direction, an analogue of Lehmers Problem has been proposed for elliptic curves, and more generally for Abelian varieties. Here is Conjecture 1.4 of [DH]. Let A be an Abelian variety dened over a number eld K and equipped with a symmetric ample line bundle L. For any P A(Q), dene (Q) = min deg(V )1/ codim(V ) , where V ranges over the proper subvarieties of A, dened over K, K-irreducible and containing Q, while deg(V ) is the degree of V with respect to L. Also denote by hL the NronTate canonical height on A(Q) associated to L. e Conjecture 4.7 (DavidHindry). There exists a positive constant c depending only on A and L, such that for any P A(Q) which has innite order modulo any Abelian subvariety, hL (P ) c(P )1 . An extension of Conjecture 4.7 to linearly independent tuples is also stated in [DH, Conjecture 1.6]. The dependence on A of these constants also suggests interesting questions. Take an elliptic curve E and consider the NronTate height h(P ) of a nontorsion e rational point on a number eld K. Several invariants are related to E: the modular
285
invariant jE , the discriminant E and Faltings height h(E). S. Lang conjectured that h(P ) c(K) max{1, h(E)}, while S. Lang ([La5, p. 92]) and J. Silverman ([Sil, Chap. VIII, 10, Conjecture 9.9]) conjecture that h(P ) c(K) max{log |NK/Q (E )|, h(jE )}. Partial results are known (J. Silverman, M. Hindry and J. Silverman, S. David), but the conjecture is not yet proved. There is another Abelian question related to Mahlers measure. According to D. A. Lind, Lehmers Problem is known to be equivalent to the existence of a continuous endomorphism of the innite torus (R/Z)Z with nite entropy. A similar question has been asked by P. DAmbros, G. Everest, R. Miles and T. Ward [AEMW] for elliptic curves, and it can be extended to Abelian varieties, and more generally to commutative algebraic groups. 4.2. WirsingSchmidt Conjecture. According to Dirichlets box principle, for any irrational, real number there is an innite set of rational numbers p/q with q > 0 such that p 1 2. (4.8) q q There are several extensions of this result. For the rst one, we write (4.8) as |q p| < 1/q and we replace q p by P () for some polynomial P . Lemma 4.9. Let be a real number, d and H be positive integers. There exists a non-zero polynomial P Z[X], of degree less than or equal to d and usual height less than or equal to H such that |P ()| cH d , where c = 1 + || + + ||d . There is no assumption on , but if is algebraic of degree d, then there is a trivial solution! A similar result applies to complex numbers, and more generally when is replaced by a m-tuple (1 , . . . , m ) Cm (see for instance [W6] Lemma 15.11). For simplicity, we only deal here with the easiest case. Another extension of (4.8) that is interesting to consider is where p/q is replaced by an algebraic number of degree d. If the polynomial P given by (4.9) has a single simple root close to , then | | c H d where c depends only on and d. However, the root of P which is nearest may be a multiple root, and may be not unique. This occurs precisely when the rst derivative P of P has a small absolute value at . Dirichlets box principle does not allow us to construct a polynomial P as in (4.8) with a lower bound for |P ()|.
286
M. WALDSCHMIDT
However E. Wirsing [Wi] succeeded in proving the following theorem. Theorem 4.10. There exist two positive absolute constants, c and , such that, for any transcendental real number and any positive integer n, there are innitely many algebraic numbers of degree n for which | | cH()n . Wirsing himself obtained his estimate in 1960 with c replaced by c(n, n , ) and n by (n/2) + 2 n , where n 0 as n . He conjectured that the exponent n can be replaced by n + 1 [Wi]. For n = 2, H. Davenport and W. M. Schmidt in 1967 reached the exponent 3 without . For any transcendental real number , there exists a positive real number c() such that the inequality | | c()H()3 has innitely many solutions with [Q() : Q] 2. A conjecture of Schmidt ([Sch1, Chap. VIII, 3]; see also [Bu1] and [Bu2]) asserts that (4.10) is valid when n is replaced by n + 1. Conjecture 4.11 (Wirsing and Schmidt). For any positive integer n and any real number which is either transcendental or else is algebraic of degree greater than n, there exists a positive constant c = c(n, ) with the following property: there exist innitely many algebraic numbers of degree n for which 0 < | | < cH()n1 . A third extension of (4.8) is the study of the simultaneous rational approximation of successive powers of a real number. Let n 2 be an integer; denote by En the set of real numbers which are not algebraic of degree n. For En , let n () be the inmum of the set of real numbers such that, for any suciently large real number X, there exists (x0 , x1 , . . . , xn ) Zn satisfying 0 < x0 X and
1jn
max |x0 j xj | X 1/ .
From Dirichlets box principle one deduces n () n for any En and any n 2. Moreover, for any n 2, the set of En for which n () < n has Lebesgue measure zero. H. Davenport and W. M. Schmidt proved in [DS] that 2 () for any E2 , where = (1 + 5)/2 = 1.618. . . It was expected that 2 () would be equal to 2 for any E2 , but D. Roy [Ro9] has produced a E2 for which 2 () = , showing that the result of Davenport and Schmidt is optimal. This raises a number of open problems and suggests that we study the set An = {n () : En }. Recent results concerning the set A2 , by Y. Bugeaud and M. Laurent, S. Fischler, indicate a structure like the Marko spectrum. For further references on this topic, see [Bu3]. In Section 3 we considered problems of algebraic independence. In Section 2 we discussed questions related to measures of linear independence of logarithms of algebraic numbers. In Section 4 we introduced a notion of height. Connexions between these three topics arise from the study of simultaneous approximation of
287
complex numbers by algebraic numbers (see for instance [W6, Chap. 15]). For a m-tuple = (1 , . . . , m ) of algebraic numbers, we dene () = [Q() : Q] max h(j ),
1jm
so that for m = 1 and Q, () = log M (). So far, relations between simultaneous approximation and algebraic independence have only been established for small transcendence degrees. The missing link for large transcendence degrees is given by the next statement (see [W6, Conjecture 15.31], [Lau1, 4.2, Conjecture 5], [Lau2, Conjecture 1], [W5, Conjecture 2], as well as [Ro6, Conjectures 1 and 2]). Conjecture 4.12. Let = (1 , . . . , m ) be a m-tuple of complex numbers. Dene t = trdeg Q() and assume t 1. There exist positive constants c1 and c2 with the following property. Let (D )0 and ( )0 be sequences of real numbers satisfying c1 D , D D+1 2D , +1 2 ( 0). Assume also that the sequence ( )0 is unbounded. Then for innitely many there exists a m-tuple (1 , . . . , m ) of algebraic numbers satisfying [Q() : Q] D , and
1im 1/t max |i i | exp{c2 D }.
()
There are two dierent, but related quantitative renements to a transcendence result: for a transcendental number , either one proves a transcendence measure, which is a lower bound for |P ()| when P is a non-zero polynomial with integer coecients, or else one proves a measure of algebraic approximation for , which is a lower bound for | | when is an algebraic number. In both cases such a lower bound will usually depend on the degree (of the polynomial P , or on the algebraic number ), and on the height of the same. Next, given several transcendental numbers 1 , . . . , n , one may consider either a measure of simultaneous approximation by algebraic numbers, namely a lower bound for max{|i i |} when 1 , . . . , n are algebraic numbers, or a measure of algebraic independence, which is a lower bound for |P (1 , . . . , n )| when P is a non-zero polynomial with integer coecients. The rst estimate deals with algebraic points (algebraic sets of zero dimension), the second with hypersurfaces (algebraic sets of codimension 1). There is a set of intermediate possibilities which have been studied by Yu. V. Nesterenko and P. Philippon, and are closely connected. For instance, Conjecture 4.12 deals with simultaneous approximation by algebraic points; M. Laurent and D. Roy asked general questions about the approximation by algebraic subsets of Cm , dened over Q. For instance Conjecture 2 in
288
M. WALDSCHMIDT
[Lau2] as well as the conjecture in 9 of [Ro7] deal with the more general problem of approximation of points in Cn by points located on Q-varieties of a given dimension. For an algebraic subset Z of Cm , dened over Q, denote by t(Z) the size of a Chow form of Z. Conjecture 4.13 (LaurentRoy). Let Cm . There is a positive constant c, depending only on and m, with the following property. Let k be an integer with 0 k m. For innitely many integers T 1, there exists an algebraic set Z Cm , dened over Q, of dimension k, and a point Z, such that t(Z) T mk and | | exp{cT m+1 }.
Further far-reaching, open problems in this direction have been proposed by P. Philippon as Probl`mes 7, 8 and 10 in [P2, 5]. e 4.3. Logarithms of Algebraic Numbers. We have already suggested several questions related to linear independence measures over the eld of rational numbers for logarithms of rational numbers (see Conjectures 2.4, 2.5 and 2.14). Now that we have a notion of height for algebraic numbers at our disposal, we can extend our study to linear independence measures over the eld of algebraic numbers for the logarithms of algebraic numbers. The next statement is Conjecture 14.25 of [W6]. Conjecture 4.14. There exist two positive absolute constants c1 and c2 with the following property. Let 1 , . . . , m be logarithms of algebraic numbers with i = ei (1 i m), let 0 , . . . , m be algebraic numbers, D the degree of the number eld Q(1 , . . . , m , 0 , . . . , m ), and, nally, let h 1/D satisfy h max h(i ),
1im
1 max |i | D 1im
and
h max h(j ).
0jm
(1) Assume that the number = 0 + 1 1 + + m m is non-zero. Then || exp c1 mD2 h . (2) Assume that 1 , . . . , m are linearly independent over Q. Then
m
|i i | exp c2 mD1+(1/m) h .
i=1
Assuming both Conjecture 4.12 and part 2 of Conjecture 4.14, one deduces not only Conjecture 3.3, but also further special cases of Conjecture 3.1 (these connexions are described in [W5] as well as [W6, Chap. 15]). As far as part 1 of Conjecture 4.14 is concerned, weaker estimates are available (see [W6, 10.4]). Here is a much weaker (but still open) statement than either Conjecture 2.5 or part 1 of Conjecture 4.14.
289
Conjecture 4.15. There exists a positive absolute constant C with the following property. Let 1 , . . . , n be non-zero algebraic numbers and log 1 , . . . , log n logarithms of 1 , . . . , n respectively. Assume that the numbers log 1 , . . . , log n are Q-linearly independent. Let 0 , 1 , . . . , n be algebraic numbers, not all of which are zero. Denote by D the degree of the number eld Q(1 , . . . , n , 0 , 1 , . . . , n ) over Q. Further, let A1 , . . . , An and B be positive real numbers, each e, such that | log j | 1 , (1 j n), log Aj max h(j ), D D B max h(j ).
1jn1
Then the number = 0 + 1 log 1 + + n log n satises || > exp{C n Dn+2 (log A1 ) (log An )(log B + log D)(log D)}. One is rather close to such an estimate (see [W8, 5 and 6], as well as [Matv]). The result is proved now in the so-called rational case, where 0 = 0 and i Q for 1 i n. B max log Ai .
1in
Removing this extra condition would enable one to prove that numbers like e or 2 2 are not Liouville numbers. These questions are the rst and simplest ones concerning transcendence measures,measures of Diophantine approximation, measures of linear independence and measures of algebraic independence. One may ask many further questions on this topic, including an eective version of Schanuels conjecture. It is interesting to notice that in this case a technical condition cannot be omitted ([W4, Conjecture 1.4]). Recall that the rank of a prime ideal P Q[T1 , . . . , Tm ] is the largest integer r 0 such that there exists an increasing chain of prime ideals (0) = P0 P1 Pr = P. The rank of an ideal I Q[T1 , . . . , Tm ] is the minimum rank of a prime ideal containing I. Conjecture 4.16 (Quantitative Renement of Schanuels Conjecture). Let x1 , . . . . . . , xn be Q-linearly independent complex numbers. Assume that for any > 0, there exists a positive number H0 such that, for any H H0 and n-tuple (h1 , . . . , hn ) of rational integers satisfying 0 < max{|h1 |, . . . , |hn |} H, the inequality |h1 x1 + + hn xn | exp H
290
M. WALDSCHMIDT
is valid. Let d be a positive integer. Then there exists a positive number C = C(x1 , . . . , xn , d) with the following property: for any integer H 2 and any n + 1 tuple P1 , . . . , Pn+1 of polynomials in Z[X1 , . . . , Xn , Y1 , . . . , Yn ] with degrees d and usual heights H, which generate an ideal of Q[X1 , . . . , Xn , Y1 , . . . , Yn ] of rank n + 1,
n+1
A consequence of Conjecture 4.16 is a quantitative renement to Conjecture 3.3 on the algebraic independence of logarithms of algebraic numbers [W4]. Conjecture. If log 1 , . . . , log n are Q-linearly independent logarithms of algebraic numbers and d a positive integer, there exists a constant C > 0 such that, for any non-zero polynomial P Z[X1 , . . . , Xn ] of degree d and height H, with H 2, |P (log 1 , . . . , log n )| H C . 4.4. Density: Mazurs Problem. Let K be a number eld with a given real embedding. Let V be a smooth variety over K. Denote by Z the closure, for the real topology, of V (K) in V (R). In his paper [Maz1] on the topology of rational points, Mazur asks, Question 4.17 (Mazur). Assume that K = Q and that V (Q) is Zariski dense; is Z a union of connected components of V (R)? An interesting fact is that Mazur asks this question in connexion with the rational version of Hilberts tenth Problem (see [Maz2] and [Maz3]). The answer to question 4.17 is negative. An example is given in [CSS] by J.-L. Colliot-Thl`ne, A. N. Skorobogatov and P. Swinnerton-Dyer of a smooth ee surface V over Q, whose Q-rational points are Zariski-dense, but such that the closure Z in V (R) of the set of Q-points is not a union of connected components. However for the special case of Abelian varieties, there are good reasons to believe that the answer to question 4.17 is positive. Indeed for this special case a reformulation of question 4.17 is the following. Conjecture. Let A be a simple Abelian variety over Q. Assume that the Mordell Weil group A(Q) has rank 1. Then A(Q) A(R)0 is dense in the neutral component, A(R)0 of A(R). This statement is equivalent to the next one. Conjecture 4.18. Let A be a simple Abelian variety over Q, expA : Rg A(R)0 the exponential map of the Lie group A(R)0 , and = Z1 + + Zg its kernel. Let u = u1 1 + + ug g Rg satisfy expA (u) A(Q). Then 1, u1 , . . . , ug are linearly independent over Q. The following quantitative renement of Conjecture 4.18 is suggested in [W3, Conjecture 1.1].
291
max |i j j i | .
0jN
max |i | max |j |
Conjecture 4.19. Let A be a simple Abelian variety of dimension g over a number eld K embedded in R. Denote by the rank over Z of the MordellWeil group A(K). For any > 0, there exists h0 > 0 (which depends only on the Abelian variety A, the real number eld K and ) such that, for any h h0 and any A(R)0 , there is a point A(K) with NronTate height h such that e dist(, ) h(
/2g)+
Similar problems arise for commutative algebraic groups. Let us consider the easiest case, a torus Gn over the eld of real algebraic numbers. We replace the simple m Abelian variety A of dimension g by the torus Gn of dimension n, the Mordell m Weil group A(K) by a nitely generated multiplicative subgroup of (Q )n , and the connected component A(R)0 of the origin in A(R) by (R )n . The corresponding + problem is then, given positive algebraic numbers ij (1 i n, 1 j m), to consider the approximation of a tuple (1 , . . . , n ) (R )n by tuples of algebraic + numbers of the form s1 sm s1 sm 11 1m , . . . , n1 nm with s = (s1 , . . . , sm ) Zm . Recently D. Prasad [Pr] studied this question in terms of toric varieties. The qualitative density question is solved by the following statement, which is a consequence of Conjecture 3.3. Conjecture 4.20. Let m, n, k be positive integers and aij rational integers (1 i n, 1 j m, 1 k). For x = (x1 , . . . , xk ) (R )k denote + by (x) the following nitely generated subgroup of (R )n , +
m k
(x) =
j=1 =1
xkij
sj 1in
Assume that there exists x (R )k such that + any = (1 , . . . , k ) in (R )k with 1 , . . . , k + pendent, the subgroup () is dense in (R )n . +
If there is a x in (R )k such that (x) is dense in (R )n , then the set of such + + x is dense in (R )k . Hence again, loosely speaking, Conjecture 4.20 means that + logarithms of algebraic numbers should behave like almost all numbers (see also [La8, Chap. IX, 7, p. 235]). Conjecture 4.20 would provide an eective solution to the question raised by J.-L. Colliot-Thl`ne and J.-J. Sansuc and solved by D. Roy (see [Ro3]). ee Theorem. Let k be a number eld of degree d = r1 + 2r2 , where r1 is the number of real embeddings and r2 the number of pairwise non-conjugate embeddings of k. Then there exists a nitely generated subgroup of k , with rank r1 + r2 + 1, whose image in Rr1 Cr2 is dense.
292
M. WALDSCHMIDT
The existence of is known, but the proof by D. Roy does not yield an explicit example. Density questions are closely related to transcendence questions. For instance the multiplicative subgroup of R generated by e and is dense if and only if log + is irrational (which is an open question). The simplest case of Conjecture 4.20 is obtained with n = 2 and m = 3. It reads as follows. Conjecture. Let 1 , 2 , 3 , 1 , 2 , 3 be non-zero positive algebraic numbers. Assume that for any (a, b) Z2 \ {(0, 0)}, two at least of the three numbers
a b a b a b 1 1 , 2 2 , 3 3
are multiplicatively independent. Then the subgroup = of (R )2 is dense. + It is easy to deduce this statement from the four exponentials Conjecture 3.7. The next question is to consider a quantitative renement. Let be a nitely generated subgroup of (Q R )n which is dense in (R )n . Fix a set of generators + + 1 , . . . , m of . For s = (s1 , . . . , sm ) Zm and 1 i n dene
m s s s s s s 11 22 33 , 11 22 33 : (s1 , s2 , s3 ) Z3
i (s) =
j=1
ijj Q .
The density assumption means that for any = (1 , . . . , n ) (R )n and any + > 0, there exists s Zm such that
1in
max |i (s) i | .
We wish to bound |s| = max1jm |sj | in terms of . We x a compact neighborhood K of the origin (1, . . . , 1) in (R )n . For instance + K = { (R )n : 1/2 |i | 2 (1 i n)} + would do. Conjecture 4.21. For any > 0 there exists S0 > 0 (depending on , 1 , . . . , m and K) such that, for any S S0 and any K, there exists s Zm with |s| S and max |i (s) i | S 1(1/n)+ .
1in
These questions suggest a new kind of Diophantine approximation problem. 5. Further Topics 5.1. Metric Problems. Among the motivations for studying metric problems in Diophantine analysis (not to mention secular perturbations in astronomy and the statistical mechanics of a gas see [Ha]), one would like to be able to guess the behavior of certain classes of numbers (such as algebraic numbers, logarithms
293
of algebraic numbers, and numbers given as values of classical functions, suitably normalized [La2, p. 658 and 664]). A rst example is related to the WirsingSchmidt Conjecture. V. G. Sprindzuk showed in 1965 that the conjecture 4.11 is true for almost all (for Lebesgue measure). A second example is the question of rening Roths Theorem. Conjecture 2.12 is motivated by Khinchines Theorem ([Sp, Chap. I, 1, Th. 1, p. 1]) which answers the question of rational Diophantine approximation for almost all real numbers. In 1926 A. Khinchine himself extended his result to the simultaneous Diophantine rational approximation max |qi pi |
1in
([Sp, Chap. I, 4, Th. 8, p. 28]), and in 1938 A. V. Groshev proved the rst very general theorem of Khinchine type for systems of linear forms,
1in
max |q1 i1 + + qm im pi |
([Sp, Chap. I, 5, Th. 12, p. 33]). Using the same heuristic arguments, one may extend Conjecture 2.12 to the context of simultaneous linear combinations of algebraic numbers. In Conjecture 2.12 (as well as in Khinchines result for almost all real numbers) the function q(q) is assumed to be non-increasing. A conjecture of Dun and Schaeer (see [Sp, Chap. 1, 2, p. 17] and [Ha]) would enable one to work without such a restriction. Denote by (n) Eulers function (n) =
1kn gcd(k,n)=1
1.
Conjecture 5.1 (Dun and Schaeer). Let be a positive real valued function. Then, for almost all R, inequality (2.11) has an innite number of solutions in integers p and q with q > 0 and gcd(p, q) = 1 if and only if the series
q=1
1 (q)(q) q
diverges. The KhinchineGoshev Theorem has been extended to certain manifolds (see [Sp], [BD], as well as more recent papers by V. Bernick, M. Dodson, D Kleinbock and G. Margulis). Further, connexions between the metrical theory of Diophantine approximation on one hand, hyperbolic geometry, ergodic theory and dynamics of ows on homogeneous spaces of Lie groups on the other, have been studied by several mathematicians, including D. Sullivan, S. J. Dani, G. Margulis and D. Kleinbock. Also S. Hersonsky and F. Paulin [HP] have recently studied the Diophantine approximation properties of geodesic lines on the Heisenberg group, which suggests new, open questions, for instance to study
1in
max |qi pi |i
294
M. WALDSCHMIDT
The set of real numbers with bounded partial quotients is countable. This is the set of real numbers which are badly approximable by rational numbers. Y. Bugeaud asks a similar question for numbers which are badly approximable by algebraic numbers of bounded degree. Question 5.2 (Bugeaud). Let n 2. Denote by n the set of real numbers with the following property: there exists c1 () > 0 and c2 () > 0 such that for algebraic number of degree n, | | c2 ()H()n1 , and such that there are innitely many algebraic numbers of degree n with Does the set | | c1 ()H()n1 ,
n
In connexion with the algebraic independence problems of Section 3.1, one would like to understand better the behavior of real (or complex) numbers with respect to Diophantine approximation by algebraic numbers of large degree (see Conjecture 4.12). A natural question is to consider this question from a metrical point of view. Roughly speaking, what is expected is that for almost all real numbers , the quality of approximation by algebraic numbers of degree d and measure t be edt . This is the precise suggestion of Y. Bugeaud [Bu2]. For a real number > 0, denote by F the set of real numbers with the following property: for any with 0 < < and any d0 1, there exists a real number h0 1 such that, for any d d0 and any t h0 d, the inequality | | e dt has a solution Q where [Q() : Q] d and () t. Also, denote by F the set of real numbers with the following property: for any > there exist d0 1 and h0 1 such that, for any d d0 and any t h0 d, the inequality | | e dt has no solution Q where [Q() : Q] d and () t. These denition are given more concisely in [Bu2]: for t d 1 denote by Q(d, t) the set of real algebraic numbers of degree d and measure t. Then F =
< d0 1 h0 1 dd0 th0 d Q(d,t)R
] e dt , + e dt [, ] e dt , + e dt [c
> d0 1 h0 1 dd0 th0 d Q(d,t)R
F =
c
where ]a, b[ denotes the complement of the intervall ]a, b[. According to Theorem 4 of [Bu2], there exist two positive constants and such that, for almost all R, max{ > 0 : F } = Further, 1 1. 850 and min{ > 0 : F } = .
295
Bugeauds conjecture is = = 1. It is an important open question to study the simultaneous approximation of almost all tuples in Rn by algebraic tuples = (1 , . . . , n ) in terms of the degree [Q() : Q] and the measure (). Most authors have devoted much attention to the dependence on the height, but now it is necessary to study more thoroughly the behavior of the approximation for large degree. Further problems which we considered in the previous sections deserve to be studied from the metrical point of view. Our next example is a strong quantitative form of Schanuels Conjecture for almost all tuples ([W5, Conjecture 4]). Conjecture 5.3. Let n be a positive integer. For almost all n-tuples (x1 , . . . , xn ), there are positive constants c and D0 (depending on n, x1 , . . . , xn and ), with the following property. For any integer D D0 , any real number D and any 2n-tuple 1 , . . . , n , 1 , . . . , n of algebraic numbers satisfying [Q(1 , . . . , n , 1 , . . . , n ) : Q] D and [Q(1 , . . . , n , 1 , . . . , n ) : Q] max {h(i ), h(i ) : 1 i n} , max {|xi i |, |exi i | : 1 i n} exp{cD1/(2n) }. One may also expect that c does not depend on x1 , . . . , xn . An open metrical problem of uniform distribution was suggested by P. Erds to o R. C. Baker in 1973 (see [Ha] Chap. 5, p. 163). It is a counterpart to the conjecture of Khinchine which was disproved by J. M. Marstrand in 1970. Question. Let f be a bounded measurable function with period 1. Is it true that 1 N log N lim for almost all R? 5.2. Function Fields. Let K be a eld and C = K((T 1 )) be the eld of Laurent series on K. The eld C has similar properties to the real number eld, when Z is replaced by K[T ] and Q by K(T ). An absolute value on C is dened by selecting |T | > 1. We set || = |T |k if = nZ an T n is a non-zero element of C, where k = deg() denotes the least index such that ak = 0. Hence C is the completion of K(T ) for this absolute value. A theory of Diophantine approximation has been developed on C in analogy to the classical one. If K has zero characteristic, the results are very similar to the classical ones. But if K has nite characteristic, the situation is completely dierent (see [dML] and [Sch3]). It is not yet even clear how to describe the situation from a conjectural point of view. A conjectural description of the set of algebraic numbers for which a Roth type inequality is valid is still missing. Some algebraic elements satisfy a Roth type inequality, while for some others, Liouvilles estimate is optimal. However, from a certain point of view, much more is known about the function eld case, since the exact approximation exponent is known for several classes of algebraic numbers. 1 f (n) = n n=1
N 1
f (x) dx
0
296
M. WALDSCHMIDT
There is also a transcendence theory over function elds. The starting point is a paper by Carlitz in the 40s. He denes functions on C which behave like analogues of the exponential function (Carlitz module). A generalization is due to V. G. Drinfeld (Drinfeld modules), and a number of results on the transcendence of numbers related to these objects are known, going much further than their classical (complex) counterpart. For example, the number (1 p1 )1
p
(in a suitable extension of a nite eld), where p runs over the monic irreducible polynomials over the given nite eld, is known to be transcendental (over the eld of rational functions on the nite eld) [AT]; it may be considered to be an analogue of Eulers constant since 1 . = lim (s) s1 s1 However the theory is far from being complete. An analogue of Schanuels Conjecture for Drinfeld modules was proposed by W. D. Brownawell in [Brow], together with many further related problems, including large transcendence degree, Diophantine geometry, values of CarlitzBessel functions and values of gamma functions. For the study of Diophantine approximation, an important tool (which is not available in the classical number theoretic case) is the derivation d/dT . In the transcendence theory this suggests new questions which started to be studied by L. Denis. Also in the function eld case, interesting new questions are suggested by considering several characteristics. So Diophantine analysis for function elds involve dierent aspects, some which are reminiscent of the classical theory, and some which have no counterpart. For the related transcendence theory involving automata theory, we refer to the paper by D. Thakur [Th] (especially on p. 389390) and to [AS] for the state of the art concerning the following open problem, Conjecture 5.4 (Loxton and van der Poorten). Let (ni )i0 be an increasing sequence of positive integers. Assume that there is a prime number p such that the power series z ni Fp [[z]]
i0
is algebraic over Fp (z) and irrational (i. e., does not belong to Fp (z)). Then the real number 10ni
i0
is transcendental. Acknowledgements. This survey grew out of lectures given in several places including India (October 2000 and September 2002), Ivory Coast (February 2001), Italy (April 2001), Canada (May 2001), Lebanon (November 2002) and France. The author is grateful to all the colleagues who gave him the opportunity to speak on this topic, and also to all those who contributed by relevant remarks and suggestions.
297
References
J.-P. Allouche and J. Shallit, Automatic sequences, Cambridge University Press, Cambridge, 2003. MR 1 997 038 [AEMW] P. DAmbros, G. Everest, R. Miles, and T. Ward, Dynamical systems arising from elliptic curves, Colloq. Math. 84/85 (2000), no. 1, 95107. MR 2001m:11093 [AD1] F. Amoroso and S. David, Le probl`me de Lehmer en dimension suprieure, J. Reine e e Angew. Math. 513 (1999), 145179. MR 2001a:11116 [AD2] F. Amoroso and S. David, Minoration de la hauteur normalise des hypersurfaces, Acta e Arith. 92 (2000), no. 4, 339366. MR 2001f:11102 [AT] G. W. Anderson and D. S. Thakur, Tensor powers of the Carlitz module and zeta values, Ann. of Math. (2) 132 (1990), no. 1, 159191. MR 91h:11046 [A1] Y. Andr, G-functions and geometry, Aspects of Mathematics, E13, Friedr. Vieweg & e Sohn, Braunschweig, 1989. MR 90k:11087 [A2] Y. Andr, Quelques conjectures de transcendance issues de la gomtrie algbrique, e e e e Prpublication de lInstitut de Mathmatiques de Jussieu, 121, 1997; available at http: e e //www.institut.math.jussieu.fr/~preprints/index-1997.html. [B1] A. Baker, Transcendental number theory, Cambridge University Press, Cambridge, 1990. MR 91f:11049 [B2] A. Baker, Logarithmic forms and the abc-conjecture, Number theory (Eger, 1996), de Gruyter, Berlin, 1998, pp. 3744. MR 99e:11101 [BLSW] R. Balasubramanian, M. Langevin, T. N. Shorey, and M. Waldschmidt, On the maximal length of two sequences of integers in arithmetic progressions with the same prime divisors, Monatsh. Math. 121 (1996), no. 4, 295307. MR 98a:11124 [BR] K. Ball and T. Rivoal, Irrationalit dune innit de valeurs de la fonction zta aux e e e entiers impairs, Invent. Math. 146 (2001), no. 1, 193207. MR 2003a:11086 [BD] V. I. Bernik and M. M. Dodson, Metric Diophantine approximation on manifolds, Cambridge Tracts in Mathematics, vol. 137, Cambridge University Press, Cambridge, 1999. MR 2001h:11091 [Ber] C. Bertolin, Priodes de 1-motifs et transcendance, J. Number Theory 97 (2002), no. 2, e 204221. MR 2003i:11104 [Bert1] D. Bertrand, Duality on tori and multiplicative dependence relations, J. Austral. Math. Soc. Ser. A 62 (1997), no. 2, 198216. MR 98f:11070 [Bert2] D. Bertrand, Theta functions and transcendence, Ramanujan J. 1 (1997), no. 4, 339 350. MR 99b:11080 [Bi] Y. F. Bilu, Catalans conjecture (after Mihilescu), Sminar Bourbaki, 2002/2003, vol. a e 909, pp. 122. Preprint version available at http://www.math.u-bordeaux.fr/~yuri. [BL] E. Bombieri and S. Lang, Analytic subgroups of group varieties, Invent. Math. 11 (1970), 114. MR 45 #5089 [Boy1] D. W. Boyd, Mahlers measure and special values of L-functions, Experiment. Math. 7 (1998), no. 1, 3782. MR 99d:11070 [Boy2] D. W. Boyd, Mahlers measure and special values of L-functionssome conjectures, Number theory in progress, vol. 1 (Zakopane-Kocielisko, 1997), de Gruyter, Berlin, s 1999, pp. 2734. MR 2000f:11139 [Br] K. Briggs, Some explicit badly approximable pairs, J. Number Theory 103 (2003), no. 1, 7176. MR 2 008 066. Preprint vesrion available at http://arxiv.org/abs/math.NT/ 0211143. [Bro] J. Browkin, The abc-conjecture, Number theory, Trends Math., Birkhuser, Basel, 2000, a pp. 75105. MR 2001f:11053 [Brow] W. D. Brownawell, Transcendence in positive characteristic, Number theory (Tiruchirapalli, 1996), Contemp. Math., vol. 210, Amer. Math. Soc., Providence, RI, 1998, pp. 317332. MR 98h:11092 [Bu1] Y. Bugeaud, On the approximation by algebraic numbers with bounded degree, Algebraic number theory and Diophantine analysis (Graz, 1998), de Gruyter, Berlin, 2000, pp. 47 53. MR 2001e:11075 [AS]
298
M. WALDSCHMIDT
[Bun] [C]
[Ca]
[Cat] [Ch]
[Chu]
[Co] [Coh]
[CSS]
[DMR]
[dML]
Y. Bugeaud, Approximation par des nombres algbriques, J. Number Theory 84 (2000), e no. 1, 1533. MR 2001g:11113 Y. Bugeaud, Approximation by algebraic numbers, Cambridge Tracts in Mathematics, Cambridge University Press, CambridgeNew York, 2004. Y. Bugeaud and M. Mignotte, Sur lquation diophantienne (xn 1)/(x 1) = y q . II, e C. R. Acad. Sci. Paris Sr. I Math. 328 (1999), no. 9, 741744. MR 2001f:11050 e Y. Bugeaud and T. N. Shorey, On the number of solutions of the generalized RamanujanNagell equation, J. Reine Angew. Math. 539 (2001), 5574. MR 2002k:11041 P. Bundschuh, Zwei Bemerkungen uber transzendente Zahlen, Monatsh. Math. 88 (1979), no. 4, 293304. MR 81f:10046 P. Cartier, Fonctions polylogarithmes, nombres polyztas et groupes pro-unipotents, e Sminar Bourbaki, 2001/2002, Exp. 885, Astrisque (2002), no. 282, 137173. MR 1 e e 975 178 J. W. S. Cassels, An introduction to Diophantine approximation, Cambridge Tracts in Mathematics and Mathematical Physics, No. 45, Cambridge University Press, New York, 1957. MR 19,396h E. Catalan, Note extraite dune lettre adresse ` lditeur, J. Reine Angew. Math. 27 e a e (1844), 192. A. Chambert-Loir, Thor`mes dalgbricit en gomtrie diophantienne (dapr`s e e e e e e e J.-B. Bost, Y. Andr, D. & G. Chudnovsky), Sminar Bourbaki, 2001/2002, Exp. e e 886, Astrisque (2002), no. 282, 175209. MR 1 975 179 e G. V. Chudnovsky, Singular points on complex hypersurfaces and multidimensional Schwarz lemma, Seminar on Number Theory, Paris 197980, Progr. Math., vol. 12, Birkhuser Boston, Mass., 1981, pp. 2969. MR 83m:32002 a P. B. Cohen, Perspectives de lapproximation Diophantienne et la transcendence, Ramanujan J. 7 (2003), no. 13, 367384. H. Cohn, Marko geodesics in matrix theory, Number theory with an emphasis on the Marko spectrum (Provo, UT, 1991), Lecture Notes in Pure and Appl. Math., vol. 147, Dekker, New York, 1993, pp. 6982. MR 94d:11049 J.-L. Colliot-Thl`ne, A. N. Skorobogatov, and P. Swinnerton-Dyer, Double bres and ee double covers: paucity of rational points, Acta Arith. 79 (1997), no. 2, 113135. MR 98a:11081 T. W. Cusick and M. E. Flahive, The Marko and Lagrange spectra, Mathematical Surveys and Monographs, vol. 30, Amer. Math. Soc., Providence, RI, 1989. MR 90i:11069 T. W. Cusick and C. Pomerance, View-obstruction problems. III, J. Number Theory 19 (1984), no. 2, 131139. MR 86b:11046 H. Davenport and W. M. Schmidt, Approximation to real numbers by algebraic integers, Acta Arith. 15 (1968/1969), 393416. MR 40 #91 S. David, Petits points, points rationnels, XXII`mes Journes Arithmtiques de Lille, e e e 2001. S. David, On the height of subvarieties of group varieties, to appear in Ramanujan J. S. David and M. Hindry, Minoration de la hauteur de NronTate sur les varits e ee abliennes de type C. M, J. Reine Angew. Math. 529 (2000), 174. MR 2001j:11054 e S. David and P. Philippon, Minorations des hauteurs normalises des sous-varits e ee des tores, Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4) 28 (1999), no. 3, 489543. MR 2001a:11109 M. Davis, Y. Matiyasevich, and J. Robinson, Hilberts tenth problem: Diophantine equations: positive aspects of a negative solution, Mathematical developments arising from Hilbert problems (Proc. Sympos. Pure Math., vol. XXVIII, Northern Illinois Univ., De Kalb, Ill., 1974), Amer. Math. Soc., Providence, RI, 1976, pp. 323378 (loose erratum). MR 55 #5522 B. de Mathan and A. Lasjaunias, Dierential equations and Diophantine approximation in positive characteristic, Monatsh. Math. 128 (1999), no. 1, 16. MR 2000f:11091
299
[Di1]
[Di2] [Du]
[E1] [E2]
[Ge1] [Ge2]
[Ge3]
[Gr] [GW]
G. Diaz, La conjecture des quatre exponentielles et les conjectures de D. Bertrand sur la fonction modulaire, J. Thor. Nombres Bordeaux 9 (1997), no. 1, 229245. e MR 98h:11096 G. Diaz, Transcendance et indpendance algbrique: liens entre les points de vue ellipe e tique et modulaire, Ramanujan J. 4 (2000), no. 2, 157199. MR 2001g:11120 A. Dujella, There are only nitely many Diophantine quintuples, to appear in J. Reine Angew. Math.; preprint version available at http://www.math.hr/~duje/dvi/finit. dvi. P. Erds, How many pairs of products of consecutive integers have the same prime o factors?, Amer. Math. Monthly 87 (1980), 391392. P. Erds, On the irrationality of certain series: problems and results, New advances o in transcendence theory (Durham, 1986), Cambridge Univ. Press, Cambridge, 1988, pp. 102109. MR 89k:11057 G. Faltings, Diophantine equations, Mathematics unlimited2001 and beyond, Springer, Berlin, 2001, pp. 449454. MR 1 852 170 N. I. Feldman and Yu. V. Nesterenko, Transcendental numbers, Number theory, IV, Encyclopaedia Math. Sci., vol. 44, Springer, Berlin, 1998, pp. 1345. MR 99a:11088b N. I. Feldman and A. B. Shidlovskii, The development and present state of the theory of transcendental numbers, Uspehi Mat. Nauk 22 (1967), no. 3 (135), 381 (Russian). MR 35 #5400. English translation in: Russian Math. Surveys, 22 (1967) no. 3, 179. S. R. Finch, Mathematical constants, Encyclopedia of Mathematics and its Applications, vol. 94, Cambridge University Press, Cambridge, 2003. MR 2 003 519 S. Fischler, Irrationalit de valeurs de zta [dapr`s Apry, Rivoal,. . . ], Bourbaki e e e e Sminar, 2002/2003, vol. 910, pp. 135. e L. Flatto, J. C. Lagarias, and A. D. Pollington, On the range of fractional parts {(p/q)n }, Acta Arith. 70 (1995), no. 2, 125147. MR 96a:11073 A. I. Galochkin, Sur des estimations prcises par rapport ` la hauteur de certaines e a formes linaires, Approximations Diophantiennes et Nombres Transcendants (Luminy, e 1982), Progr. Math., vol. 31, Birkhuser, Boston, Mass., 1983, pp. 9598. a A. O. Gelfond, Sur quelques rsultats nouveaux dans la thorie des nombres transcene e dants, C. R. Acad. Sci. Paris Sr. A 199 (1934), 259. e A. O. Gelfond, The approximation of algebraic numbers by algebraic numbers and the theory of transcendental numbers, Uspehi Matem. Nauk (N. S.) 4 (1949), no. 4(32), 1949 (Russian). MR 11,231c. English translation in: Amer. Math. Soc. Translation, 65 (1952), 81124. A. O. Gelfond, Transcendental and algebraic numbers, Gosudarstv. Izdat. Tehn.-Teor. Lit., Moscow, 1952 (Russian). MR 15,292e. English translation in: Dover Publications, Inc, New York, 1960. F. Gramain, Sur le thor`me de FukasawaGelfond, Invent. Math. 63 (1981), no. 3, e e 495506. MR 83g:30028 F. Gramain and M. Weber, Computing an arithmetic constant related to the ring of Gaussian integers, Math. Comp. 44 (1985), no. 169, 241250. MR 87a:11028. Corrigendum: Math. Comp. 48 (1987), no. 178, 854. MR 88a:11034. G. Gras, Class eld theory, Springer Monographs in Mathematics, Springer-Verlag, Berlin, 2003. MR 2003j:11138 I. Grattan-Guinness, A sideways look at Hilberts twenty-three problems of 1900, Notices Amer. Math. Soc. 47 (2000), no. 7, 752757. MR 2001f:01036 C. A. Grimm, A conjecture on consecutive composite numbers, Amer. Math. Monthly 76 (1969), 11261128. MR 41 #5292 P. Grinspan, Measures of simultaneous approximation for quasi-periods of Abelian varieties, J. Number Theory 94 (2002), no. 1, 136176. MR 2003k:11122 R. K. Guy, Unsolved problems in number theory, Unsolved Problems in Intuitive Mathematics, vol. 1, Springer-Verlag, New York, 1994. MR 96e:11002 K. Gyarmati, On a problem of Diophantus, Acta Arith. 97 (2001), no. 1, 5365. MR 2002d:11031
300
M. WALDSCHMIDT
[H]
[HW] [Ha]
[HP] [Hi]
[La1]
[La2]
[La3]
[La4]
[La5]
[La6] [La7]
[Lan1]
[Lan2]
M. Hall, Jr., The Diophantine equation x3 y 2 = k, Computers in number theory (Proc. Sci. Res. Council Atlas Sympos, no. 2, Oxford, 1969), Academic Press, London, 1971, pp. 173198. MR 48 #2061 G. H. Hardy and E. M. Wright, An introduction to the theory of numbers, The Clarendon Press, Oxford University Press, New York, 1979. MR 81i:10002 G. Harman, Metric number theory, London Mathematical Society Monographs. New Series, vol. 18, The Clarendon Press, Oxford University Press, New York, 1998. MR 99k:11112 S. Hersonsky and F. Paulin, Hausdor dimension of Diophantine geodesics in negatively curved manifolds, J. Reine Angew. Math. 539 (2001), 2943. MR 2002i:53053 D. Hilbert, Mathematical problems, Bull. Amer. Math. Soc. (N. S.) 37 (2000), no. 4, 407436 (electronic). MR 1 779 412. Reprinted from: Bull. Amer. Math. Soc. 8 (1902), 437479. A. Jackson, Million-dollar mathematics prizes announced, Notices Amer. Math. Soc. 47 (2000), no. 8, 877879. See also http://www.claymath.org/. M. Kontsevich and D. Zagier, Periods, Mathematics unlimited2001 and beyond, Springer, Berlin, 2001, pp. 771808. MR 2002i:11002 A. Kraus, On the equation xp + y q = z r : a survey, Ramanujan J. 3 (1999), no. 3, 315333. MR 2001f:11046 J. C. Lagarias, On the normality of arithmetical constants, Experiment. Math. 10 (2001), no. 3, 355368. MR 2003f:11107. Preprint version available at http://arXiv. org/abs/math.NT/0101055. S. Lang, Introduction to transcendental numbers, Addison-Wesley Publishing Co., Reading, Mass.LondonDon Mills, Ont., 1966. MR 35 #5397. Reprinted in: Collected papers, vol. I, Springer-Verlag, New York, 2000, 396506. MR 2002d:01026a. S. Lang, Transcendental numbers and Diophantine approximations, Bull. Amer. Math. Soc. 77 (1971), 635677. MR 44 #6615. Reprinted in: Collected papers, vol. II, Springer-Verlag, New York, 2000, 143. MR 2002d:01026b. S. Lang, Higher dimensional Diophantine problems, Bull. Amer. Math. Soc. 80 (1974), 779787. MR 50 #12914. Reprinted in: Collected papers, vol. II, Springer-Verlag, New York, 2000, 102110. MR 2002d:01026b. S. Lang, Relations de distributions et exemples classiques, Sminaire Delange-Pisote Poitou, 19e anne: 1977/78, Thorie des nombres, Fasc. 2, Secrtariat Math., Paris, e e e 1978, pp. Exp. No. 40, 6. MR 80c:10033. Reprinted in: Collected papers, vol. III, Springer-Verlag, New York, 2000, 5965. MR 2002d:01026c. S. Lang, Elliptic curves: Diophantine analysis, Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], vol. 231, SpringerVerlag, Berlin, 1978. MR 81b:10009 S. Lang, Cyclotomic elds, Graduate Texts in Mathematics, vol. 59, Springer-Verlag, New York, 1978. MR 58 #5578 S. Lang, Old and new conjectured Diophantine inequalities, Bull. Amer. Math. Soc. (N. S.) 23 (1990), no. 1, 3775. MR 90k:11032. Reprinted in: Collected papers, vol. III, Springer-Verlag, New York, 2000, 355393. MR 2002d:01026c. S. Lang, Number theory. III, Encyclopaedia of Mathematical Sciences, vol. 60, SpringerVerlag, Berlin, 1991. MR 93a:11048. S. Lang, Algebra, Addison-Wesley Publishing Co., Inc., Reading, Mass., 1993. S. Lang, La conjecture de BatemanHorn, Gaz. Math. (1996), no. 67, 8284. MR 97b:11114. Reprinted in: Collected papers, vol. IV, Springer-Verlag, New York, 2000, 213216. MR 2002d:01026d. M. Langevin, Plus grand facteur premier dentiers en progression arithmtique, e Sminaire DelangePisotPoitou, 1976/1977, Exp. 3, Thorie des nombres, Fasc. 1, e e Secrtariat Math., Paris, 1977. MR 81a:10011 e M. Langevin, Partie sans facteur carr dun produit dentiers voisins, Approximations e diophantiennes et nombres transcendants (Luminy, 1990), de Gruyter, Berlin, 1992, pp. 203214. MR 93g:11081
301
M. Langevin, Sur quelques consquences de la conjecture (abc) en arithmtique et en e e logique, Rocky Mountain J. Math. 26 (1996), no. 3, 10311042. MR 97k:11052 M. Langevin, Equations diophantiennes polynomiales ` hautes multiplicits, J. Thor. a e e Nombres Bordeaux 13 (2001), no. 1, 211226. MR 2002f:11035 M. Laurent, New methods in algebraic independence, Number theory (Eger, 1996), de Gruyter, Berlin, 1998, pp. 311330. MR 99e:11100 M. Laurent, Diophantine approximation and algebraic independence, Colloque FrancoJaponais: Thorie des Nombres Transcendants (Tokyo, 1998), Sem. Math. Sci., vol. 27, e Keio Univ., Yokohama, 1999, pp. 7589. MR 2000m:11061 H.-W. Leopoldt, Zur Arithmetik in abelschen Zahlkrpern, J. Reine Angew. Math. 209 o (1962), 5471. MR 25 #3034 K. Mahler, An unsolved problem on the powers of 3/2, J. Austral. Math. Soc. 8 (1968), 313321. MR 37 #2694 K. Mahler, Lectures on transcendental numbers, Lecture Notes in Mathematics, vol. 546, Springer-Verlag, Berlin, 1976. MR 58 #10772 K. Mahler, Some suggestions for further research, Bull. Austral. Math. Soc. 29 (1984), no. 1, 101108. MR 85e:11002 Yu. I. Manin, A course in mathematical logic, Graduate Texts in Mathematics, vol. 53, Springer-Verlag, New York, 1977. MR 56 #15345 D. W. Masser, Note on a conjecture of Szpiro, Sminaire sur les Pinceaux de Courbes e Elliptiques (` la recherche de Mordell eectif), Paris, 1988, Astrisque (1990), no. 183, a e 1923. MR 91g:11058 Yu. Matiyasevich, Le dixi`me probl`me de Hilbert: que peut-on faire avec les quations e e e diophantiennes?, La recherche de la vrit, Ecrit. Math., ACL-Ed. Kangourou, Paris, e e 1999, pp. 281305. MR 2001f:03081. See also http://logic.pdmi.ras.ru/Hilbert10. E. M. Matveev, An explicit lower bound for a homogeneous rational linear form in logarithms of algebraic numbers. II, Izv. Ross. Akad. Nauk Ser. Mat. 64 (2000), no. 6, 125180 (Russian). MR 2002e:11091. English translation in: Izv. Math. 64 (2000), no. 6, 12171269. R. D. Mauldin, A generalization of Fermats last theorem: the Beal conjecture and prize problem, Notices Amer. Math. Soc. 44 (1997), no. 11, 14361437. MR 98j:11020 B. Mazur, The topology of rational points, Experiment. Math. 1 (1992), no. 1, 3545. MR 93j:14020 B. Mazur, Questions of decidability and undecidability in number theory, J. Symbolic Logic 59 (1994), no. 2, 353371. MR 96c:03091 B. Mazur, Speculations about the topology of rational points: an update, Columbia University Number Theory Seminar (New York, 1992), Astrisque (1995), no. 228 (4), e 165182. MR 96c:11068 B. Mazur, Questions about powers of numbers, Notices Amer. Math. Soc. 47 (2000), no. 2, 195202. MR 2000k:11111 T. Metsnkyl, Catalans conjecture: another old Diophantine problem solved, Bull. a a Amer. Math. Soc. (N. S.) 41 (2004), no. 1, 4357 (electronic). MR 2 015 449 P. Mihilescu, Primary units and a proof of Catalans conjecture, submitted to Crelle a Journal; preprint version available at http://www-math.uni-paderborn.de/~preda/ papers/zips/catcrelle.ps.gz. J.-M. Muller and A. Tisserand, Towards exact rounding of the elementary functions, Scientic computing and validated numerics (Proceedings of the International Symposium on Scientic Computing, Computer Arithmetic and Validated Numerics (SCAN95) held at the University of Wuppertal, Wuppertal, September 2629, 1995) (Berlin), Mathematical Research, vol. 90, Akademie Verlag, 1996, p. 340. MR 97a:65005 W. Narkiewicz, Classical problems in number theory, Monograe Matematyczne [Mathematical Monographs], vol. 62, PWNPolish Scientic Publishers, Warsaw, 1986. MR 90e:11002 W. Narkiewicz, Elementary and analytic theory of algebraic numbers, Springer-Verlag, Berlin; PWNPolish Scientic Publishers, Warsaw, 1990. MR 91h:11107
[Mat]
[Matv]
[MT]
[N1]
[N2]
302
M. WALDSCHMIDT
[NP]
[Ni] []
[O]
[P1]
[P2]
[P3]
[Pr]
[Ro1]
[Ro2]
[Ro3]
Yu. V. Nesterenko and P. Philippon (eds.), Introduction to algebraic independence theory, Lecture Notes in Mathematics, vol. 1752, Springer-Verlag, Berlin, 2001. MR 2002g:11104 A. Nitaj, web cite http://www.math.unicaen.fr/~nitaj/abc.html. J. sterl, Nouvelles approches du thor`me de Fermat, Sminar Bourbaki, e e e e 1987/1988, Exp. 694, Astrisque (1988), no. 161162 (4), 165186 (1989). e MR 90g:11038 F. Oort, Canonical liftings and dense sets of CM-points, Arithmetic geometry (Cortona, 1994), Sympos. Math., XXXVII, Cambridge Univ. Press, Cambridge, 1997, pp. 228234. MR 98g:14056 P. Philippon, Indpendance et groupes algbriques, Number theory (Montreal, Que., e e 1985), CMS Conf. Proc., vol. 7, Amer. Math. Soc., Providence, RI, 1987, pp. 279284. MR 894 328 P. Philippon, Une approche mthodique pour la transcendance et lindpendance e e algbrique de valeurs de fonctions analytiques, J. Number Theory 64 (1997), no. 2, e 291338. MR 98i:11052 P. Philippon, Quelques remarques sur des questions dapproximation Diophantienne, Bull. Austral. Math. Soc., 59 (1999), 323334. MR 2000m:11057. Addendum: Ibid., 61 (2000), no. 1, 167169. MR 1 819 326 P. Philippon, Mesures dapproximation de valeurs de fonctions analytiques, Acta Arith. 88 (1999), no. 2, 113127. MR 2000d:11098 S. S. Pillai, On the equation 2x 3y = 2X + 3Y , Bull. Calcutta Math. Soc. 37 (1945), 1520. MR 7,145i A. D. Pollington and S. L. Velani, On a problem in simultaneous Diophantine approximation: Littlewoods conjecture, Acta Math. 185 (2000), no. 2, 287306. MR 2002a:11076 D. Prasad, An analogue of a conjecture of Mazur: a question in Diophantine approximation on tori, To appear in: Shalika Festschrift, a supplemental volume of the Amer. J. Math.; preprint version available at http://www.mri.ernet.in/~mathweb/ PB440-0662G-23-699-710.pdf. K. Ramachandra, Contributions to the theory of transcendental numbers. I, II, Acta Arith. 14 (1967/68), 6572; ibid. 14 (1967/1968), 7388. MR 37 #165 G. Rauzy, Proprits statistiques de suites arithmtiques, Le Mathmaticien, vol. 15, ee e e Collection SUP, Presses Universitaires de France, Paris, 1976. MR 53 #13152 G. Rmond, Ingalit de Vojta en dimension suprieure, Ann. Scuola Norm. Sup. Pisa e e e e Cl. Sci. (4) 29 (2000), no. 1, 101151. MR 2001g:11123 G. Rhin and C. Smyth, On the absolute Mahler measure of polynomials having all zeros in a sector, Math. Comp. 64 (1995), no. 209, 295304. MR 95c:11123 P. Ribenboim, Catalans conjecture, Academic Press Inc., Boston, MA, 1994. MR 95a:11029 P. Ribenboim, My numbers, my friends, Springer-Verlag, New York, 2000. MR 2002d:11001 T. Rivoal, La fonction zta de Riemann prend une innit de valeurs irrationnelles e e aux entiers impairs, C. R. Acad. Sci. Paris Sr. I Math. 331 (2000), no. 4, 267 e 270. MR 2001k:11138. Preprint version available at http://arXiv.org/abs/math.NT/ 0008051. D. Roy, Sur la conjecture de Schanuel pour les logarithmes de nombres algbriques, e Groupe dEtudes sur les Probl`mes Diophantiens 19881989, Publ. Math. Univ. P. et e M. Curie (Paris VI), 90 (1989), no. 6. D. Roy, Matrices dont les coecients sont des formes linaires, Sminaire de Thorie e e e des Nombres, Paris 198788, Progr. Math., vol. 81, Birkhuser Boston, Boston, MA, a 1990, pp. 273281. MR 91c:11038 D. Roy, Simultaneous approximation in number elds, Invent. Math. 109 (1992), no. 3, 547556. MR 93f:11087
303
[Sil] [Sm]
[So]
D. Roy, Points whose coordinates are logarithms of algebraic numbers on algebraic varieties, Acta Math. 175 (1995), no. 1, 4973. MR 96i:11079 D. Roy, An arithmetic criterion for the values of the exponential function, Acta Arith. 97 (2001), no. 2, 183194. MR 2001m:11120 D. Roy, Approximation algbrique simultane de nombres de Liouville, Canad. Math. e e Bull. 44 (2001), no. 1, 115120. MR 2001j:11060 D. Roy, Results and methods of transcendental number theory, Notes de la SMC 33 (2001), no. 2, 914. D. Roy, Interpolation formulas and auxiliary functions, J. Number Theory 94 (2002), no. 2, 248285. MR 2003e:11079 D. Roy, Approximation simultane dun nombre et de son carr, C. R. Math. Acad. e e Sci. Paris 336 (2003), no. 1, 16. MR 2003m:11103 A. Schinzel, The Mahler measure of polynomials, Number theory and its applications (Ankara, 1996), Lecture Notes in Pure and Appl. Math., vol. 204, Dekker, New York, 1999, pp. 171183. MR 99k:11041 A. Schinzel and H. Zassenhaus, A renement of two theorems of Kronecker, Michigan Math. J. 12 (1965), 8185. MR 31 #158 H. P. Schlickewei and C. Viola, Generalized Vandermonde determinants, Acta Arith. 95 (2000), no. 2, 123137. MR 2002a:11028 W. M. Schmidt, Diophantine approximation, Lecture Notes in Mathematics, vol. 785, Springer, Berlin, 1980. MR 81j:10038 W. M. Schmidt, Heights of algebraic points, Number theory and its applications (Ankara, 1996), Lecture Notes in Pure and Appl. Math., vol. 204, Dekker, New York, 1999, pp. 185225. MR 2000b:11076 W. M. Schmidt, On continued fractions and Diophantine approximation in power series elds, Acta Arith. 95 (2000), no. 2, 139166. MR 2001j:11063 T. Schneider, Einfhrung in die transzendenten Zahlen, Springer-Verlag, Berlin, 1957. u MR 19,252f. J.-P. Serre, Lectures on the MordellWeil theorem, Aspects of Mathematics, E15, Friedr. Vieweg & Sohn, Braunschweig, 1989. MR 90e:11086 T. N. Shorey, Exponential Diophantine equations involving products of consecutive integers and related equations, Number theory, Trends Math., Birkhuser, Basel, 2000, a pp. 463495. MR 2001g:11045 T. N. Shorey, Some conjectures in the theory of exponential Diophantine equations, Publ. Math. Debrecen 56 (2000), no. 34, 631641. MR 2001i:11038 T. N. Shorey and R. Tijdeman, Exponential Diophantine equations, Cambridge Tracts in Mathematics, vol. 87, Cambridge University Press, Cambridge, 1986. MR 88h:11002 C. L. Siegel, Uber einige Anwendungen diophantischer Approximationen, Abh. Preuss. Akad. Wiss., Phys.-Math. 1 (1929), 170. Reprinted in: Gesammelte Abhandlungen, Bnde I, Springer-Verlag, Berlin, 1966. MR 33 #5441. a C. L. Siegel, Transcendental numbers, Annals of Mathematics Studies, no. 16, Princeton University Press, Princeton, N. J., 1949. MR 11,330c W. Sierpiski, A selection of problems in the theory of numbers, A Pergamon Press n Book, The Macmillan Co., New York, 1964. MR 30 #1078 W. Sierpi ski, 250 problems in elementary number theory, Modern Analytic and Comn putational Methods in Science and Mathematics, No. 26, American Elsevier Publishing Co., Inc., New York, 1970. MR 42 #4475. J. H. Silverman, The arithmetic of elliptic curves, Graduate Texts in Mathematics, vol. 106, Springer-Verlag, New York, 1986. MR 87g:11070 S. Smale, Mathematical problems for the next century, Math. Intelligencer 20 (1998), no. 2, 715. MR 99h:01033. Reprinted in: Mathematics: frontiers and perspectives, Amer. Math. Soc., Providence, RI, 2000, pp. 271294. MR 2001i:00003. J. Sondow, Criteria for irrationality of Eulers constant, Proc. Amer. Math. Soc. 131 (2003), no. 11, 33353344 (electronic). MR 2004b:11102. Preprint version available at http://arXiv.org/abs/math.NT/0209070.
304
M. WALDSCHMIDT
[Ti1]
[V] [W1]
[W6]
[W7] [W8]
V. G. Sprindzuk, Metric theory of Diophantine approximations, V. H. Winston & Sons, Washington, D.C., 1979. MR 80k:10048 C. L. Stewart and Kun Rui Yu, On the abc conjecture, Math. Ann. 291 (1991), no. 2, 225230. MR 92k:11037 C. L. Stewart and Kun Rui Yu, On the abc conjecture. II, Duke Math. J. 108 (2001), no. 1, 169181. MR 2002e:11046 T. Terasoma, Mixed Tate motives and multiple zeta values, Invent. Math. 149 (2002), no. 2, 339369. MR 2003h:11073 D. S. Thakur, Automata and transcendence, Number theory (Tiruchirapalli, 1996), Contemp. Math., vol. 210, Amer. Math. Soc., Providence, RI, 1998, pp. 387399. MR 98k:11105 R. Tijdeman, Hilberts seventh problem: on the GelfondBaker method and its applications, Mathematical developments arising from Hilbert problems (Proc. Sympos. Pure Math., Northern Illinois Univ., De Kalb, Ill., 1974), Amer. Math. Soc., Providence, RI, 1976, pp. 241268. MR 55 #7936 R. Tijdeman, On the equation of Catalan, Acta Arith. 29 (1976), no. 2, 197209. MR 53 #7941 R. Tijdeman, Exponential Diophantine equations 19861996, Number theory (Eger, 1996), de Gruyter, Berlin, 1998, pp. 523539. MR 99f:11046 R. Tijdeman, Some applications of Diophantine approximation, Number theory for the millennium, III (Urbana, IL, 2000), A K Peters, Natick, MA, 2002, pp. 261284. MR 2003j:11076. Preprint version: Math. Inst. Leiden, Report MI 2000-27, September 2000. P. Vojta, On the ABC conjecture and Diophantine approximation by rational points, Amer. J. Math. 122 (2000), no. 4, 843872. MR 2001i:11094 M. Waldschmidt, Proprits arithmtiques de fonctions de plusieurs variables. II, ee e Sminaire Pierre Lelong (Analyse), 1975/1976, Lecture Notes in Math., vol. 578, e Springer, Berlin, 1977, pp. 108135. MR 56 #11919 M. Waldschmidt, Groupes algbriques et grands degrs de transcendance, Acta Math. e e 156 (1986), no. 34, 253302. MR 87k:11076 M. Waldschmidt, Density measure of rational points on abelian varieties, Nagoya Math. J. 155 (1999), 2753. MR 2000h:11077 M. Waldschmidt, Algebraic independence of transcendental numbers: a survey, Number theory, Trends Math., Birkhuser, Basel, 2000, pp. 497527. MR 2001g:11118. a M. Waldschmidt, Conjectures for large transcendence degree, Algebraic number theory and Diophantine analysis (Graz, 1998), de Gruyter, Berlin, 2000, pp. 497520. MR 2001i:11092 M. Waldschmidt, Diophantine approximation on linear algebraic groups, Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], vol. 326, Springer-Verlag, Berlin, 2000. MR 2001c:11075 M. Waldschmidt, Valeurs zta multiples. Une introduction, J. Thor. Nombres Bordeaux e e 12 (2000), no. 2, 581595. MR 2002a:11106 M. Waldschmidt, Linear independence measures for logarithms of algebraic numbers, Diophantine approximation (Cetraro, 2000), Lecture Notes in Math., vol. 1819, Springer, Berlin, 2003, pp. 250344. MR 2 009 832. Preprint version available at http://www.math.jussieu.fr/~miw/articles/cetraro.html. E. Waring, Meditationes algebraic, Cambridge, 1770. English translation: Meditationes algebraic, Amer. Math. Soc., Providence, RI, 1991. MR 93a:01026. E. T. Whittaker and G. N. Watson, A course of modern analysis, Cambridge Univ. Press, 1927. E. Wirsing, Approximation mit algebraischen Zahlen beschrnkten Grades, J. Reine a Angew. Math. 206 (1960), no. 12, 6777. MR 26 #79 C. H. Wong, An explicit result related to the abc-conjecture, MPhil Thesis, Hong Kong University of Science & Technology, 1999. A. Woods, Some problems in logic and number theory, Ph.D. thesis, Manchester, 1986.
305
[Z]
[Zu]
D. Zagier, Values of zeta functions and their applications, First European Congress of Mathematics, vol. II (Paris, 1992), Progr. Math., vol. 120, Birkhuser, Basel, 1994, a pp. 497512. MR 96k:11110 V. V. Zudilin, Algebraic relations for multiple zeta values, Uspekhi Mat. Nauk 58 (2003), no. 1, 332 (Russian). MR 1 992 130
Univ. Paris VI, Institut de Mathematiques de Jussieu (CNRS UMR 7586), Case 247, 175 rue du Chevaleret, F-75013 Paris, France E-mail address: miw@math.jussieu.fr URL: http://www.math.jussieu.fr/~miw/