{orders}ms/98711/cohn.3d -17.11.00 - 12:37
THE DIOPHANTINE SYSTEM x
2ÿ 6y
2 ÿ5; x 2z
2ÿ 1
J. H. E. COHN*
1. Introduction.
In a recent paper [5] showing that the system of the title has only the solu- tions in integers given by z0;1;2;3;6 and 91, the introduction men- tioned in passing that the elementary algebraic methods developed for ex- ample in [2] do not appear su¤cient to solve this problem. The authors then proceeded to prove their result using the high powered analytical methods for which they, Tzanakis, de Weger, Steiner and others have become re- nowned. We present below a short simple proof of their result.
It is clearly su¤cient to assume thatx ÿ1;y1 and z0 we shall do so without further mention. The equation v2ÿ6u21 has fundamental so- lution 52
p6
, and then it is easily shown that the general solution of x2ÿ6y2 ÿ5 is given by the two classes xy
p6
1
p6
n with n0. Let 5ÿ2
p6
and de¢ne the sequences un nÿn= ÿ, vn nn=2. Then x vn12un, and so our problem reduces to proving that the only solutions of the equation
2z21vn12un 1
are given byn0;1 and 4 of the equation 2z21ÿvn12un
2
are given byn0;1 and 2.
Here bothunandvnsatisfy the recurrence relationwn210wn1ÿwnand the ¢rst few values are given in the following table:
MATH. SCAND. 82 (1998), 161^164
Received January 9, 1996.
{orders}ms/98711/cohn.3d -17.11.00 - 12:37
n un vn vn12un ÿvn12un
0 0 1 1 í-1
1 1 5 17 7
2 10 49 169 71
3 99 485 1673 703
4 980 4801 16561 6959
We shall quote several standard identities involving the sequencesun and vn and results concerning the periodicity of these sequences modulo certain moduli which are easily veri¢ed, without writing them all out in detail.
2. Preliminaries.
Lemma 1. The equation 3x4ÿ2y21 has only the solutions in non-negative integers given by x1;3.
This is shown in [1].
Lemma2. The equation 6y2x4ÿ1has only the solutions in non-negative integers given by x1;7.
This is shown in [3].
Lemma3. The equation y2 24x41has only the solutions in non-negative integers given by x0;1.
For, y1 yÿ1 24x4 and here y1;yÿ1 2. Thus withxx1x2
either y12x41;y112x42 in which case1x41ÿ6x42; here the lower sign is impossible modulo 3, whereas the upper one gives onlyx11;x20 by Lemma 2;
or y16x41;y14x42and now13x41ÿ2x42. Again the lower sign is impossible modulo 3, and the upper sign gives only x1x21 in view of Lemma 2.
Lemma4. The equation3y22x41has only the solution in non-negative integers given by x1.
This is proved in [4], and is the only result that we use which has not been proved by technically elementary methods.
162 j. h. e. cohn
{orders}ms/98711/cohn.3d -17.11.00 - 12:38
3. Proof of result.
There are four cases.
(A) The only solutions of (1) withn even aren0 and 4.
For with n2k;2z21v2k12u2k2vk vk12uk, and so since vk;vk12uk 1 we must havevkz21. Then 1v2kÿ24u2kz41ÿ6 2uk2, and so by Lemma 2,z1 1 or 7 whencek0 or 2 as required.
(B) The only solution of (1) withnodd isn1.
For withn2k1,
2z21v2k112u2k1
1 5v2k24u2k 12 v2k5u2k
v2kÿ24u2k 17 v2k24u2k 168ukvk
18v2k168ukvk384u2k
6 vk4uk 3vk16uk:
Here the ¢nal two factors on the right have no common factor, and since 3vk16uk3 (mod 4, we must havevk4ukz21;3vk16uk3z22. Thus 4uk3z22ÿ3z21;vk4z21ÿ3z22, and so
1v2kÿ24u2k 4z21ÿ3z222ÿ272 z21ÿz222; whence 25z416z21z22ÿ9z426z41ÿ z21ÿ3z222; or 13z41ÿ2 12 z21ÿ3z222;
and so by Lemma 1, the only possibilities arez1 1 or 3. The former gives k0 whencen1 and the latter no solution.
(C) The only solutions of (2) withn even aren0 and 2.
For with n2k;2z21ÿv2k12u2k24uk ÿ2ukvk, and so since vkÿ2ukis odd and has no factor in common withukwe must have
either ukz21;vkÿ2uk3z22; in this case 1v2kÿ24u2kv2kÿ24z41 implies z1 0 or 1 in view of Lemma 3; the latter yields n2 whilst the former gives no solution;
or uk3z21;vkÿ2ukz22. Sinceuk0 (mod 3), it follows that 3jk, and with k3m we ¢nd that z2113u3mum 32u2m1. This implies that um is a square, and just as above this is possible onlyumif equals 0 or 1. The former givesn0 and the latter no solution.
(D) The only solution of (2) withnodd isn1.
For withn2k1,
the diophantine system ... 163
{orders}ms/98711/cohn.3d -17.11.00 - 12:38
2z21ÿv2k112u2k1
1ÿ 5v2k24u2k 12 v2k5u2k
v2kÿ24u2k 7 v2k24u2k 72ukvk
8 vk3uk vk6uk and this is possible only ifz21vk6uk. But then
2z4112 vk6uk21
2v2k24ukvk72u2kv2kÿ24u2k
3 vk4uk2
and by Lemma 4, this is possible only fork0.
This concludes the proof.
Note added in proof. By an extraordinary coincidence this same pro- blem was considered by R. J. Stroeker & B. M. M. de Weger in ``On a quartic Diophantine Equation'', Proc. Edinburgh Math. Soc. 39 (1996), 97^
114. Their method is as long and similar in conception to that of [5] although the details are totally di¡erent.
REFERENCES
1. Richard T. Bumby,The Diophantine equation3x4ÿ2y21, Math. Scand. 21 (1967), 144^
2. J. H. E. Cohn,148. Some quartic Diophantine equations, Paci¢c J. Math. 26 (1968), 233^243.
3. J. H. E. Cohn,The Diophantine equation x4ÿDy21, Quart. J. Math. Oxford (2) 26 (1975), 279^281.
4. Wilhelm Ljunggren,U«ber die unbestimmte Gleichung Ax2ÿBy4C, Arch. for Mathematik og Naturvidenskab (41) Nr. 10.
5. Maurice Mignotte & Attila Pethoë, On the system of Diophantine Equations x2ÿ6y2 ÿ5;x2z2ÿ1, Math. Scand 76 (1995), 50^60.
DEPARTMENT OF MATHEMATICS
ROYAL HOLLOWAY UNIVERSITY OF LONDON EGHAM, SURREY TW20 0EX
U.K.
e-Mail J.Cohn@rhbnc.ac.uk
164 j. h. e. cohn