Could You give your ideas, your comment, or a referen for a conjecture as follows:

Consider $A, B, C$ be three positive integers numbers. By Fundamental theorem of arithmetic we write:




Let $N=\min\{x_i, y_j, z_h \}$ where $1 \le i \le n, 1\ \le j \le m, 1\le h \le k$

For a positive integer $N_0 > 3$, there exist only finitely many triples $(A, B, C)$ of coprime positive integers, with $A+B=C$, such that: $N \ge N_0$

Edited: I repalced $N_0 \ge 3$ by $N_0 > 3$


This conjecture is false, though only barely.

It should be true for $N_0 \geq 4$ by the ABC conjecture. But for $N_0 = 3$ there are infinitely many counterexamples such as $$ 191114642^3 + 4309182809^3 = 7^4 \, 321817873^3. $$ I obtained this by starting from the point $P_1 = (2:1:1)$ on the elliptic curve $x^3+y^3=7z^3$ (with origin $P_0 = (1:-1:0)$), and looking for a multiple $P_n = nP_1$ whose $z$-coordinate is divisible by $7$; such $n$ must exist because $7|z(P_n)$ iff $P_n$ reduces mod $7$ to $P_0$, and here it turned out that $n=7$ was the first example. Then any $n=7m$ works, so we get an infinite sequence of counterexamples.

The same construction works for any elliptic curve $x^3+y^3=Dz^3$ of positive rank.

P.S. I see that a simpler counterexample$$ 271^3 + 3^5 146^3 = 919^3 $$ was given by user459832 to the parallel question on Math Stackexchange. This corresponds to the third multiple of the generator $(2:1:1)$ on $x^3+y^3=9z^3$; here the first case of $3|z$ is small enough that we can still comfortably exhibit the next one, $$ 415280564497^3 + 676702467503^3 = 3^5 116223894220^3. $$

  • $\begingroup$ Can You help me improving the conjecture? $\endgroup$ Jul 4 '17 at 5:01
  • $\begingroup$ Dear Dr. @NoamD.Elkies if $N_0 \ge 4$ the conjecture can easily prove by abc conjecture? Or can using this conjecture to prove the abc conjecture? $\endgroup$ Jul 4 '17 at 5:26
  • 5
    $\begingroup$ Your conjecture for $N_0 \geq 4$ is a special case of the ABC conjecture; that's what I was getting at in the second sentence of my answer. The ABC conjecture is much stronger, so I don't think that a proof of the $N_0 \geq 4$ part of your conjecture could be used to prove the full ABC conjecture. Still I don't think any techniques are known for solving even your more modest conjecture for any value of $N_0$, however large (except of course if Mochizuki's approach to the full ABC conjecture turns out to work). $\endgroup$ Jul 4 '17 at 5:31
  • $\begingroup$ I repalced $N_0 \ge 3$ by $N_0 >2$ $\endgroup$ Jul 7 '17 at 8:14
  • $\begingroup$ Dear @NoamD.Elkies Since your answer, and comment of Yaakov Baruch. The conjecture here mathoverflow.net/questions/303141/… true up to $10^{18}$. So we can state that: $h(A), h(B), h(C) \le 3$ there are infinity solution, and $h(A), h(B), h(C) >3$ there are no solution for equation $A+B=C$, $gcd(A,B)=1$ $\endgroup$ Oct 8 '19 at 15:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.