# Is the set of integers of the form $a/(b+c)+b/(a+c)+c/(a+b)$ computable?

The starting point of this question is the observation that the smallest positive integers $$a,b,c$$ satisfying

$$\frac{a}{b+c} + \frac{b}{a+c} + \frac{c}{a+b} = 4$$

are absurdly high, namely $$(154476802108746166441951315019919837485664325669565431700026634898253202035277999,$$ $$36875131794129999827197811565225474825492979968971970996283137471637224634055579 ,$$ $$4373612677928697257861252602371390152816537558161613618621437993378423467772036) .$$ This leads to the following general question: Is the set $$C\subseteq {\mathbb N}$$ defined by $$C = \{n\in\mathbb{N}\setminus\{0\}: (\exists a,b,c \in\mathbb{N}\setminus\{0\}):\frac{a}{b+c} + \frac{b}{a+c} + \frac{c}{a+b} = n\}$$ computable? (As user Watson points out in the comment section below, $$C$$ contains no odd numbers. It would also be great to see an even number $$\geq 6$$ not contained in $$C$$.)

• I don't have the time to look into the details, but from a cursory glance at this quora answer it looks like the problem amounts to deciding whether a certain elliptic curve has solutions over $\mathbb{Q}$. Now IIRC there is an algorithm for that (find solutions locally and try to globalize) provided Ш is finite, which is conjecturally always the case. So I think conjecturally your set is indeed computable. – Gro-Tsen Aug 14 '17 at 19:28
• So we're half done! – Joel David Hamkins Sep 14 '17 at 12:28
• This is somewhat similar to the question which natural numbers $n$ are congruent numbers (i.e. the area of a right-angled triangle with rational sides), which comes down to asking whether the elliptic curve $E_n \colon y^2 = x^3-n^2x$ has positive rank over $\mathbb Q$. In this case, there is a conjectural answer (Tunnell's Theorem, conditional on the BSD conjecture), which relies on the fact that the $E_n$ are all quadratic twists of a fixed curve. The question asked here is likely to be harder, since the resulting curves are not twists, and there is the positivity condition. --> – Michael Stoll Oct 26 '17 at 15:29
• (continued) Note that there are odd $n$ for which the associated elliptic curve has positive rank, but does not have rational points on the component that contains the positive points. (IIRC, $n = 19$ is an example.) This gives an additional twist to the question. – Michael Stoll Oct 26 '17 at 15:31
• Cross-posted to cstheory.stackexchange.com/questions/39383/… – Emil Jeřábek Oct 26 '17 at 16:05