Find all triples ($a$,$b$,$n$) of positive integers such that $$a^3=b^2+2^n$$
Problem
Source: 2023 Olympic Revenge problem 2
Tags: number theory, Diophantine equation
28.03.2023 20:07
Note that if $a^3-b^2=2^n$ then $(4a)^3-(8b)^2=2^{n+6}$ so we have infinitely many solutions.
28.03.2023 20:11
the problem requires you to find all natural solutions of the equation, so simply saying that there are infinitely many of them (if there is one) is not enough!
28.03.2023 20:17
mathitoo wrote: the problem requires you to find all natural solutions of the equation, so simply saying that there are infinitely many of them (if there is one) is not enough! Of course. I’m still working on it. And by the way there is a solution with $n=1$ $$3^3-5^2=2$$
28.03.2023 21:43
Seems like the solution is not elementary https://artofproblemsolving.com/community/c6h1180420p5709966
28.03.2023 22:04
For now assume that $a$ and $b$ are coprime, we'll deal with the other case later. Particularly, they're odd. First, let $n=2k$ is even. Then, working in $\mathbb{Z}[i]$ which is an UFD and note the norm is $\text{N}(a+bi)=a^2+b^2$, \[a^3=\underbrace{(b+i2^k)}_A\underbrace{(b-i2^k)}_B \text{ and assume }p\mid \text{gcd}(A,B)\text{ is a prime}.\]We get that \[p\mid A-B=2^{k+1}i\Longrightarrow \text{N}(p)\mid \text{N}(2^{k+1}i)=2^{2k+2}.\]On the other hand, $p\mid a\Longrightarrow \text{N}(p)\mid a^2$ and because $p$ isn't a unit, we get $2\mid a$, a contradiction.
, and we get $u=1$, $v=2$ and $k=1$ because of parity. This gives $a=5, b=11$ and $n=2$. The case where $n=2k+1$ is odd is similar, only we'll need to work in $\mathbb{Z}[\sqrt{-2}]$. \[a^3=(b+2^{k}\sqrt{-2})(b-2^{k}\sqrt{-2})\]and as before, the gcd of the RHS is $1$. Then $b+2^{k}\sqrt{-2}=(u+v\sqrt{-2})^3$ and $b-2^{k}\sqrt{-2}=(u-v\sqrt{-2})^3$.
and by parity we get $a=3$, $b=5$ and $n=1$. Now let $a=2^\alpha x$ and $b=2^\beta y$, so \[2^{3\alpha}x^3=2^{2\beta}y^2+2^n\]and look at $v_2$ on both sides. If $3\alpha=2\beta$ we reduce to the case from before so $(a,b,n)=(5\cdot2^{2k},11\cdot 2^{3k},1+6k)$ or $(3\cdot2^{2k},5\cdot 2^{3k},1+6k)$. If $2\beta=n$ \[2^{3\alpha-n}x^3=y^2+1\]and modulo $4$ gives $(2^{2k+1},2^{3k+1},1+6k)$. And if $3\alpha=n$ \[a^3=2^{2\beta-n}b^2+1.\]This, again, is solved by factoring in the above rings and there are no solutions.
28.03.2023 22:22
@above unfair, you missed out details and now forever you'll be the first solver despite not actually finishing it, oh well on an unrelated note, rip my sanity