Scribd Logo
Upload

Welcome to Scribd!

  • Approximation of Diophantine Equations

    Uploaded by

    Alessandro
    6 views14 pages
    Approximation of diophantine equations

    Original Title

    Approximation of diophantine equations

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    Approximation of diophantine equations

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    6 views14 pages

    Approximation of Diophantine Equations

    Uploaded by

    Alessandro
    Approximation of diophantine equations

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 14
    MATH, SCAND. 39 (1976), 5-18 NEW APPLICATIONS OF DIOPHANTINE APPROXIMATIONS TO DIOPHANTINE EQUATIONS T.N. SHOREY and R. TISDEMAN, 1. In recent years several papers on diophantine equations have been published, in which the Gel’fond-Baker method on linear forms of lo- garithms of algebraic numbers formed the main tool. We might mention papers on Thue’s equation f(x,y)=m (e.g. Baker [3], Feldman [5]), on the equation y"=P(x) where P € Z[a] (Baker [2], Schinzel and Tijde- man [15]) and on Catalan’s equation «—y"=1 (Tijdeman [17]). In this paper we give such applications to the diophantine equation in integer variables x,y,m,n, ax™—by" =k, a,b,k fixed , and to the equation in integers n,9,2,y, my wT = byt, aand b fixed, a a—-1 (see Theorems 4 and 5). In both cases we give conditions under which there are at most finitely many solutions. For example, there are only finitely many solutions of the last equation if x is fixed. In Theorem 6 we state that this equation has no solution at all if a=b=1, z= 10, q< 23. To prove this result we need a theorem of Baker [1] on the rational ap- proximation of (a/b) and the results of Nagell and Ljunggren as stated by Oblath [11, Théoréme 5]. The theorems 1, 2 and 3 are of a different character, at least in shape. They all boil down to some assertion that if b and by*+1 are both composed of fixed primes, then J=0 or |I| is not very small. Corollary 2 of Theorem 3 asserts that if d is a non-zero fixed integer, then the greatest prime factor of x"+d tends to infinity uni- formly in integers «>1, as n tends to infinity. We note that this last result can also be obtained from a recent result of Van der Poorten, [18]. Received January 8, 1976. 6 TT. N. SHOREY AND R. TIJDEMAN 2. In this section we collect those results that we shall use from other sources. Let n>1 be an integer. Let o,...,«, be non-zero algebraic numbers with degrees at most d and let the heights of «1,...,a,-, and «, be at most A’ and A (22) respectively. We begin with the following result of Baker [3]. TueorEM A. There exists an effectively computable number C, depending only on n, d and A' such that, for any 6 with 0<6<}, the inequalities 0 < |b, logayt+... +b, logan| < (8/B’)°#4e~2 have no solution in rational integers by,...,by-1 and b,(+0) with absolute values at most B and B’ respectively. Tarorem B. Let f(x,y) be an irreducible form of degree = 3 with rational integer coefficients. Then there exist effectively computable constants C, and C,, depending only on the form f, such that all integral solutions (x,y) of the diophantine equation f(z,y) =m, where m is a non-zero rational integer, satisfy the inequality max (lz, lyl) < Cyn. This was proved, independently, by Feldman [5] and Baker [3]. Tuzormm C. Let m2 2 be an integer. Denote by P(x) a polynomial with rational coefficients and at least two simple roots if m= 3, and at least three simple roots if m=2. Then any integral solution (x,y) of the equation y™ = P(x) satisfies max (2, yl) < Cs where Cs is an effectively computable constant depending only on m and P. This is due to Baker [2]. He gives explicit bounds for polynomials with rational integer coefficients. Since m is fixed, it is sufficient to require rational coefficients in Theorem C. NEW APPLICATIONS OF DIOPHANTINE APPROXIMATIONS . . . 7 Tueorem D. If a polynomial P(x) with rational coefficients has at least two distinct zeros, then the equation y™ = P(x) in integers x,y with |y|>1 implies that m1,q>1,2>1,y>1. Effective bounds for the solutions p,q,x,y can be given. This is due to Tijdeman [17]. The next result is due to Baker [1]. Its proof is not based on the Gel’fond-Baker method of estimating linear forms in the logarithms of algebraic numbers, but on hypergeometric functions. Tuxzorem F, Let m and n be integers such that n=3 and 1\Sm1. Then «=(a/b)™/™ satisfies la—p/al > ofa” for all integers p,q (q>0), where x and ¢ are given by Ad = Qu,(a+b), cmt = 242+). The last theorem of this section can be found in a paper of Obl4th, [11, Théoréme 5]. These results are immediate consequences of results by Nagell and Ljunggren. See also Inkeri [6]. Turormm G. If the equation 10°-1=y* is fulfilled by the integers n>1,q>1,y>1, then 2{n,3tn and 3tq. 3. In this section we show that if b and by?+1 (y>1,qg>1) are composed. of fixed primes, then either J=0 of |J| is not very small. Let M(22) be a real number. In this section we shall denote by ¢,¢s,... effectively computable positive constants depending only on M. 8 'T. N. SHOREY AND R, TISDEMAN Tueorem 1, Let § be the set of all positive integers composed of primes not exceeding M. Then there exists an effectively computable constant c, depending only on M such that if beS, a=byt+leS, with l,qy € Z, 140, q>1, y>1, then I > (byes toeate , Proor. Put z=by%. Without loss of generality, we assume that (1) ll Sd. Let a=p,"...p,% and b=p,"t...p,’ be prime factorisations of a and b. Hence 8S M and p,< M for i=1,...,8. Notice that, by (1), a, S 2loga S$ 3logz and 0b, S 2logb < 3 logz for i=1,...,8. From 0<|a—z|=|l| S }z, it follows that 0 < [(a,—b,) logp, +... +(4,—5,) logp,—g logy| < 2IH\/z. We apply Theorem A to the linear form in logarithms with B=3 logz, B'=q, A’=M, 6=(39)4, n=8+15M+1 and d=1, Hence there exists @ constant c, such that I@.—2,) logp, + ... + (4,—,) logp,—q logy| > exp(—c, logg logy—q- logz) . It follows that |l| 2 zexp(—g- logz—c, logg logy —1) 2 zexp(— (qt + cg logg + 29+)logz) > gmlosaig for some constant ¢,. Of course, the assertion of Theorem 1 is trivial if g0. NEW APPLICATIONS OF DIOPHANTINE APPROXIMATIONS .. . 9 TueormM 2. Under the assumptions of Theorem 1, there exists effectively computable numbers c, and c, such that P IU] > eg(by2)** , for q23. Proor. It is no loss of generality to assume that q is prime, (a,b) =1 and that a/b is not a g-th power of a rational number. Moreover, by Theorem 1, we can restrict ourselves to the finitely many values of q with ¢< 2c, logg. We write a=a'xt and b=b'rt where a’,b’,x,r are positive integers and a’ and B’ are g-free. We note (a’,b’)=1, that a’ and b’ are bounded by some constant c, and that a’xt—b' (yr)? = 1. We further observe that (a’/b’)7 can not be a rational number. Thus the binary form f(X,Y) = a'Xe—v'ye is irreducible over Z and of degree at least 3. This follows for example from the irreducibility criterion of Dumas. See [19, §27]. On applying Theorem B, we obtain constants c, and ¢, such that max(x,yr) S c,|l|*. Hence for certain constants ¢4,c, | eq Myr)e* = og-Nrdya)* = o4(bys)" . 4, In this section, it is proved that the greatest prime factor of the ele- ments of certain sequences of integers tend to infinity. We denote by P{a] the greatest prime factor of the integer a. TuroreM 3. Let b be a positive integer and Ny>0. Let E={E,(x)}P1 be a sequence of functions with E,,(x) ¢Z for all veN, neN. Then there exists an effectively computable constant Cy=C9(b) > 0 such that if (2) 0 < |B, (a)| < ater for all z>1 and n= No, then lim,,,..P[b2" +E q(2)] = 0 uniformly in xe N, x>1. 10 T. N. SHOREY AND R. TIJDEMAN Proor. Suppose that there exists a constant M=M(b)26 such that for every N > 0, there exist integers x= 2 and n2N with Plba"+£#,(x)] 3 M. Then we have an infinite sequence of pairs (x;,n,) with x,>1, n,2¢ such that Plbu"+#,,(z)] 3M. On applying Theorem 1 with y=2, q=n, and k=H,,(x), we find a con- stant c,=c,(M)=c,(b) such that [Bpg(ee)| > xghertonns Without loss of generality we may assume that cy >c,. Hence, in view of (2), mg$Np for all i. However, this is impossible since n,2i for all i. This completes the proof of Theorem 3. Remark. It follows from the proof that the limit is also uniform in the sequence of functions Z. Theorem 3 can easily be generalised to the case that is not a constant itself, but composed of fixed primes. We give two corollaries of Theorem 3. The first is an improvement of a recent result of Langevin [7]. Corotuary. Let {a,} be a sequence of integers with a,>1 for all n. Let {b,} be a sequence of positive integers tending to infinity with n. Let {d,} be a sequence of non-zero integers. Then there exist absolute constants Cy >0 and Ny >0 such that if [dgl < a,ba-e10108mn for n= N,, then limy,5..Pla_!™-+6,] = 00 uniformly in the sequence {a,}. Corotiary. Let d be a non-zero fixed integer. Then lim,_,..P[z"+d] = 00 uniformly in «¢N, e>1. 5. It has been conjectured by Pillai [13] that if a, 6 and & are fixed in- tegers, k+0, then the diophantine equation az™—by"=k has only fini- NEW APPLICATIONS OF DIOPHANTINE APPROXIMATIONS . . . ll tely many solutions in integers m>1, n>1, >1, y>1 with mn26. Many special results of this type have been obtained. We mention some general ones. The assertion of Pillai’s conjecture is true if m and n are fixed (Siegel [16]), «and y are fixed (Pillai[12]),a=b=k=1 (Tijdeman(17]). The more general result for a=b=1 has been announced by Cudnovskit [4. p. 52]. Mahler [9] proved that Plax™—by"] -» co if max(z,y) > 00, (x,y)=1 and m, n are fixed. Theorem 4 gives some conditions under which Pillai’s assertion is valid. TurorEM 4, Let a, b and k+0 be integers. The diophantine equation (3) ax” — by” = k has only finitely many solutions in integers m>1, m>1, z>1, y>1 with mn z 6, if at least one of the following conditions is satisfied: (i) m is fixed, (ii) mn, (iii) a is composed of fixed primes. Proor. We may assume that a and 5 are positive. (i) Suppose that m is fixed. Observe that all zeros of the polynomial ax™—k in « are distinct. If follows from Theorem D that there exists a constant C,=C,(a,b,k,m) such that n0. In view of (i), it suffices to prove that n is bounded. We have 0 < log (a/b) +n log (a/y) < ab-Ya/y)"—1 < k[by” . On applying Theorem A we obtain a constant ¢,, such that log (a/b) +n log(a/y) > exp(—¢y, logn logH) , where H=max(z,y). We can assume that H <2y, otherwise it follows from (3) that n is bounded. Hence, for some constants ¢,,=¢,9(a,b,4) and ¢,3=¢9(4,0, i), y S eyy*", Since y>1, it follows that n is bounded. This completes the proof of (ii). (iii) In view of (i) it suffices to prove that n is bounded. Let P[x]< M. Put M*=max(a,b,M) and let S* be the set of all positive integers 12 T. N, SHOREY AND R. TIJDEMAN composed of primes not exceeding M*. Since be S*,by"+keS*, it follows from Theorem 1 that y"~%°8" is bounded. Hence, x is bounded. 6. The rest of the paper will be devoted to the diophantine equation a(x" —1) = by(e—1) in integers a,b,n,q,2,y with a,b fixed, n>2, ¢>1, #>1, y>1. This equa- tion arises from the problem which perfect powers have digits that are all identical. In certain special cases all solutions of the diophantine equation have been determined, for example, if (i) a=b=1, 4{n (Nagell [10]), (ii) a=b=1, g=2 (Ljunggren [8]), (iii) a=b=1, 3|n (Ljunggren [8]), (iv) 12, q>1, e>1, y>1 has only finitely many solutions, if at least one of the following conditions is satisfied: (i) & is fixed, (ii) » has fixed divisor d> 2 with dq> 6, (iii) 2 is even and q is fixed, (iv) n is even anda=b= (v) n is even, a> land x+1is (g/2l-free, (vi) n is odd, by has a fixed prime divisor and q> 2, (vii) n is odd, ab>1 and q>2. Proor. (i) We have ax"—b(a—1)y2=a. Hence the assertion is an immediate consequence of Theorem 4 (iii). Note that in this case n>1, q>1 suffices. (ii) Put am —1 am —1 grid] =~ 2 a aB. a-l amd—] g—1 It is easily seen that the greatest prime factor dy of A and B divides d. Indeed. A = 1+and+...4+a"-"/4 = d (modd,). NEW APPLICATIONS OF DIOPHANTINE APPROXIMATIONS . . . 13 Hence, by aAB=by, there exist positive integers a,,b,,y,, such that a, and b, are bounded and a,4 =byy,%. Put 2=2"/¢, We obtain 2a] ey It is, therefore, sufficient to prove assertion (ii) when n is fixed, n23. It follows from Theorem D, applied to the polynomial (a/b)(x"— 1)/(a—1) that q is bounded. The full assertion then follows from Theorem (. (iii) We have A a—-1 1 (4) a (w"/2-41) = byt. Since (x"/?— 1,24 1)|2, we obtain bounded positive integers a, and by and an integer y, such that ay(er2+1) = bunt. Hence (b,/a;)y,2-1=2". On applying Theorem D to the polynomial (b,/a,)y,2—1, we find that n is bounded. The assertion now follows from (ii). (iv) In this case, the formula (4) reads ne a1 i: (m4) = yt, Suppose ((x"/®— 1)/(x—1),2"/+1)=2. Then a is odd and hence 4|a"/—1, If n is an even multiple of 2, then there are finitely many solutions in view of (ii) with d=4, If n is an odd multiple of 2, then it follows that @=1 (mod4). We have ar—1 a1 a—1 ¢—-1 Since «+1 contains only one factor 2 and g > 1, we see that («" — 1)/(x*—1) is even. On the other hand n = l+a?+...+2"%* = — (mod2) a 2 as x is odd. It follows that 4|n which is a contradiction. We now suppose that 14 T. N. SHOREY AND R. TISDEMAN Then there exists a positive integer y, such that anrat] = yt, Now we apply Theorem E to this equation and (iv) follows immediately. (vi) Suppose that by? has a fixed prime divisor p. Since (a,b)=1 and ais q-free, we see that p|(x*— 1)/(x— 1). We distinguish the cases pf (x —1) and p|(«—1). If pt(w—1) then put t=ord,(«). Since t|n and t|(p—1), t is a bounded divisor of n. Furthermore t> 2, as n is odd. In this case the assertion follows from (ii). On the other hand, if p|(z—1), then : s =l+zt+...¢2%1=n (modp). Since p|(x*—1)/(x—1), we again obtain that n has a fixed prime divisor p( > 2). Now we can apply (ii). (vii) It is a direct consequence of (vi). (v) In view of (ii) we assume that 4{n. Let p be a prime factor of a. Since (a,b) =1, we have ply. Hence (a2) 1 eat (+1). (5) pila Tf p|((2?)"*—1)/(z?—1), then we can apply the argument of (vi). We conclude that }n has a fixed divisor >1 and then we apply (ii). We now assume that pf (x"—1)/(2*—1). The number of factors p in a is bounded, while the number of factors p in +1 is at most [4g]. Hence, by (5), 7 is bounded. We now apply (iii) to complete the proof of (v). 7. It follows from Theorem 5 (i) that there are only finitely many perfect powers whose digits in the decimal scale are identical. In fact, Obl4th [11] has shown that a number of m digits a, ie. a+10a+...+10"-la, is never a perfect power if n> 2 and 11, is a perfect power yf, then (i) 2tn, (ii) 3tn, (iii) 8tq (iv) 5tn if lg, (v) Ttn if 7\g. Furthermore it is easily seen that 2tq and 5tq, since every odd square is =1(mod4) and every fifth power is =0,+1 or +7 (mod25). We shall prove that if a number 11...1 with n digits 1 in the base 10 is a perfect power y? with g>1, y>1, then q2 23. NEW APPLICATIONS OF DIOPHANTINE APPROXIMATIONS . . . 15 TuEorem 6. Suppose that the integers n>1,q>1,y>1 satisfy the equation -9yt=1. Then q= 23. Of course it is no loss of generality to assume that q is a prime. In view of the remarks at the beginning of this section we further assume that qg27. In the proof we have to distinguish the cases n=1 (modgq) and n+ 1 (modg). In section 8, we prove that there are no solutions with n=1(modg) by applying Theorem F of Baker on the rational approxi- mation of numbers (r/s)"/". In section 9, we prove that there is no solu- tion with n#1(modg) using congruences modulo p for primes p with p=1 (modg). At least in principle, both methods are applicable for other bases. For example, Baker’s theorem is applicable for bases x with 4loga 2. 9 Proor of THEOREM 7 IN CASE n+ 1 (modq). We recall that by Theorem G both 2tn and 3tn. We use this in the proof without reference. We work out the argument for the case g=7. For the other cases we give only some essential data. Here ord, (10) denotes the smallest positive integer ¢ such that 10°=1 (mod). . We consider p= 29 and p=43. 9. Suppose 29|y. Then 10"=1(mod 29). Since n is odd and ordgy 8, this is impossible. Hence 29¢y and ((1o"—1)/9)* = 8 = 1 (mod29) . The 4-th roots of unity mod29 are +1, +12. It follows that 10"=—8, —7, 9, 10 (mod 29) and, hence n=1, 6, 19, 26 (mod28). In particular, since 7 is odd, (9) n = lor 5 (mod7). p=43. We have ord,(10)=21. Since 43ly implies 43|(10"—1) and 3|n, we obtain 43{n. Hence, ((10"— 1)/9)° = 48 = 1 (mod43) . The 6-th roots of unity mod43 are +1, +6, +7. Proceeding as in case p=29, we find that n=1,4,6,12 (mod 21). Hence, (10) n = lor4(mod7). In view of (9) and (10) there are no solutions with n+ 1 (mod7). q=H1. Let p=67. We have ord,,(10)=33. Hence, 67ty. The 6-th roots of unity mod67 are +1, +29, +30. It follows that n=1 or 27 (mod 33). Since 317, there are no solutions with m1 (mod11). NEW APPLICATIONS OF DIOPHANTINE APPROXIMATIONS . . . 7 q=13. We consider p= 157 and p=53. p=157, We have ord,;,(10)=156. Hence, 157+y. The 12-th roots of unity mod 157 are +1, +12, +13, +22, +28, +50. This implies (11) n = 1, 5or7 (mod13). p=53. We have ord,.(10)=13. We have, by (11), that 534y. The 4-th roots of unity mod 57 are +1, +23. It follows that (12) n = 1or6 (mod13). The combination of (11) and (12) excludes solutions with n+ 1 (mod 13). q=11. We consider p=103 and p= 409. p=103. We have ordyo3(10) = 34. Hence, 103¢y. The 6-th roots of unity mod103 are +1, +46, +47. It follows that (18) n = 1,17,19,29 (mod 34) . p=409. We have ordgo9(10)= 204. Hence, 409ty. The 24-th roots of unity mod409 are +1, +7, +31, +38, +49, +53, +54, +66, +117, +143, +183, +192. It follows that (14) n = 1,8 (mod34). The combination of (13) and (14) excludes solutions with n#1 (mod17), q=19. We consider p= 191 and p=229. p=229. We have ordgoe(10) = 228. Hence, 229ty. The 12-th roots of unity mod 229 are +1, +18, +89, +94, +95, +107. It follows that (15) = —5,1,6,7 (mod19). p=191. We have ordy9,(10)=95. We have, by (15), that 191ty. The 10-th roots of unity mod191 are +1, +7, +39 +49, +82. It fol- Jows that (16) n= -7, 6,—1,1,2,8,9 (mod 19) . The combination of (15) and (16) excludes solutions with m1 (mod 19). This completes the proof of Theorem 7. REFERENCES 2 1. A. Baker, Rational approximations to V2 and other algebraic numbers, Quart. J. Math, Oxford 15 (1964), 375-383. 2. A. Baker, Bounds for the solutions of the hyperelliptic equation, Proc. Cambridge Philos. Soc. 65 (1969), 439-444, 3. A, Baker, A sharpening of the bounds for linear forme in logarithms II, Acta Arith. 24 (1973), 33-36. Math, Scand. 89 — 2 18 T, N. SHOREY AND R. TISDEMAN 4, G. V. Cudnovskil, Some analytic methods in the theory of diophantine approximations [Russian], Preprint 74-9, Math, Institute Acad. Sci. Ukrainian 8.8.R., Kiev, 1974. 5. N. I. Feldman, An effective power sharpening of a theorem of Liouville, Math. USSR Izv. 5 (1972), 985-1002. 6, K. Inkeri, On the diophantine equation a(z®—1)/(e—1)=y™, Acta Arith. 21 (1972), 299-311, 7, M. Langevin, Plus grand facteur premier d'entiers consécutife, C.R. Acad. Sci., Paris, Sér. A 280 (1975), 1667-1570. 8. W. Ljunggren, Noen setninger om ubestemte likninger av formen (2"—1)/(z—1)=y2, Norek. Mat. Tidskr, 1, 25 (1943), 17-20. 9, K.Mahler, On the greatest prime factor of ax +by", Niouw Arch. Wiskunde (3) I (1953), 113-122. 10. T. Nagell, Note sur U'équation indéterminée (x” —1)/(e—1)=y%, Norsk. Mat. Tidskr. (1920), 75-78. 11. R. Oblith, Une propriété des puissances parfaites, Mathosia 65 (1956), 356-364. 12, 8.8. Pillai, On the inequality “0

    You might also like