5

Recently someone mentioned to me that there is a diophantine equation that looks very simple and innocent, but the smallest solution involves numbers of the order $10^{50}$ or something like this. The equation is probably in either 1,2, or 3 varaibles. It has low coefficients, probably all 1 or 2. And the degree is low also, probably 4 or less.

Is there such an equation?

Edit: I think the equation might have been studied by Fermat, but I'm not sure.

Joshua Benabou
  • 5,503
  • 17
  • 32
  • This would be what happens when people think there are no solutions at all. It is usually difficult to prove emptiness of a solution set in one paper, so people find partial results, often giving huge lower bounds on possible solutions. Other than that, we do not read minds. – Will Jagy May 30 '15 at 17:12
  • The claim seems quite plausible to me. I do not have such an example off-hand but the smallest solution of the "really simple" $x^2 -109 y^2=1$ is $(158070671986249,15140424455100)$. – quid May 30 '15 at 17:12
  • Now I am remembering better and he said the coefficients were all $1$ or $2$. The equation you proposed is much more difficult than the one my friend described. – Joshua Benabou May 30 '15 at 17:17
  • What would be relevant is the degree of the equation also. – quid May 30 '15 at 17:17
  • Do $x^3+y^3+z^3=30$ counts? The smallest solution is $(2220422932,-2218888517,-283059965)$ – SashaP May 30 '15 at 17:20
  • 1
    Contact your friend. I do not see how it is our responsibility to guess what your friend meant, especially if he was not entirely clear on the details. – Will Jagy May 30 '15 at 17:25
  • 4
    The smallest solution of the Pell equation $\ p^2-d q^2 \!= 1,\ \ d = 4\cdot 609\cdot 7766\cdot 4657^2 $ arising from the ancient [Archimedes Cattle Problem](http://en.wikipedia.org/wiki/Archimedes%27_cattle_problem) has over a **couple hundred thousand** decimal digits. – Bill Dubuque May 30 '15 at 17:31
  • 1
    @BillDubuque: I've seen the Archimedes cattle problem but it is not at all suprising that the solution will be very large, because it is a system with 10 or so equations. – Joshua Benabou May 30 '15 at 17:51
  • 1
    @Joshua The Archimedes cattle problem starts with seven tiny linear Diophantine equations and reduces to solving $\ x^2 - d y^2 = 1\ $ for $\ d = 410286423278424.\ $ You don't find it surprising that the smallest solution has $\,206544\,$ decimal digits? – Bill Dubuque May 30 '15 at 17:58
  • But $d$ has 15 digits, so it's concievable that the solution will be very large, especially with 7 contraints to satisfy. I suppose $20,6544$ is a bit suprising. But the equation we are looking for is even simpler than the Archimedes problem. – Joshua Benabou May 30 '15 at 18:01
  • @Joshua The Pell equation is the only constraint (btw, I mxed up the two parts of the problem above, see the linked article for details; esp. Lenstra's exposition). In any case, I thought you would find it of interest. – Bill Dubuque May 30 '15 at 18:05
  • Well it is interesting that pell equations can give very large solutions relative to $d$. – Joshua Benabou May 30 '15 at 18:10
  • 1
    The most famous versions of Pell´s equation is the above mentioned Archimedes cattle problem and the case d=61 with x=1 766 319 049, y= 226 153 980. – Mikael Jensen May 30 '15 at 18:21

3 Answers3

7

The smallest (in terms of naive height) solution of $y^2=x^3+877x$ is

$$\left(\frac{375494528127162193105504069942092792346201}{6215987776871505425463220780697238044100},\frac{256256267988926809388776834045513089648669153204356603464786949}{490078023219787588959802933995928925096061616470779979261000}\right)$$

This is an example of Bremner and Cassels. Thus, the smallest solution of $ZY^2=X^3+877XZ^2$ is $$(29604565304828237474403861024284371796799791624792913256602210,256256267988926809388776834045513089648669153204356603464786949,490078023219787588959802933995928925096061616470779979261000).$$ The $X$ coordinate is $>2\cdot 10^{61}$.

Álvaro Lozano-Robledo
  • 14,768
  • 2
  • 26
  • 51
4

Here is a general comment about the mechanism behind this phenomenon. By Matiyasevich's theorem, the problem of determining whether a Diophantine equation has a solution is undecidable. This implies that it is not possible to give a computable a priori bound on the size of the solutions to a Diophantine equation (since, given such a bound, we could solve Diophantine equations by checking all solutions up to the bound), so it follows that the size of the smallest solution to a Diophantine equation eventually exceeds any computable function of the Diophantine equation.

Qiaochu Yuan
  • 359,788
  • 42
  • 777
  • 1,145
  • For small degrees and symmetric equations is always possible to write a formula for solving equations. Everything solved quadratic forms. – individ May 30 '15 at 19:15
3

Here's a monster. The smallest integer solution to,

$$(x + a)^7 + (x - a)^7 + (2x + b)^7 + (2x - b)^7 + \\(-x - c)^7 + (-x + c)^7 + (-2x - d)^7 + (-2x + d)^7 \\= 14^7(a^6 + 2b^6 - c^6 - 2d^6)^7$$

has $x$ with $\color{red}{1179\; \text{digits}}$. The variables $a,b,c,d$ are,

$$292565171139318137956759657471297,\\ 863420822620431936290192229011966,\\ 534407060429869176086407612538177,\\ 859793943610761912321826231621886$$

and

$$\small{x =481563304865430516682423843723465575123177045754683810551700\dots \approx \color{red}{4.81 \times 10^{1178}}}$$

Ajai Choudhry found this using an elliptic curve, which may explain the large values.

Tito Piezas III
  • 47,981
  • 5
  • 96
  • 237