Let $p$ be prime. How many integer pairs $(n,m)$ exist such that $$ m^p-p^n=1? $$

When $p=2$, I have shown that the only integer pair that satisfies our condition is $(3,2)$ and when $p=3$ I have shown there are no integer pairs that satisfy our condition.

I suspect that there are no integer pairs for any prime $p>2$, but my attempts at generalizing have failed. I have observed that we can rearrange our condition a bit: $$ m^p-p^n=1 \iff p^n = m^p-1 = (m-1)(m^{p-1}+m^{p-2}+\cdots+m+1), $$ Which implies that $m-1=p^k$ for some $1\le k <n$ and that $$ p^n=\left(p^k+1\right)^p-1. $$ Not sure what to do next. I have manipulated and massaged this relationship, but I do not see how to get anything useful from what I have. Any ideas or references about this?

  • 2
    $\begingroup$ This is Catalan's conjecture, now Mihăilescu's theorem as it has been proved that $(3,2)$ is the only solution $\endgroup$ – Ross Millikan Oct 21 '17 at 4:26
  • 2
    $\begingroup$ @RossMillikan As $p$ occurs twice, this is only a special case of Mihailescu's theorem. I don't know whether it has a simpler proof... $\endgroup$ – Lord Shark the Unknown Oct 21 '17 at 4:42
  • $\begingroup$ Thanks for the references! I suspect the special case presented here has a simpler proof as it appeared as a contest problem. $\endgroup$ – Laars Helenius Oct 21 '17 at 4:55

As you've written, setting $m=p^k+1$ where $k$ is a non-negative integer gives, by the binomial theorem, $$p^n=(p^k+1)^p-1=\binom p1p^k+\binom p2p^{2k}+\binom p3p^{3k}+\cdots +\binom{p}{p-1}p^{(p-1)k}+\binom ppp^{pk}$$ Dividing the both sides by $p^{k+1}$ gives $$p^{n-k-1}=1+\binom p2p^{k-1}+\binom p3p^{2k-1}+\cdots +p^{(p-2)k}+p^{pk-k-1}$$ Here, if $k-1\ge 1$, then the LHS is divisible by $p$ while the RHS isn't.

So, we have to have $k-1\lt 1$, i.e. $k=0,1$.

  • Case 1 : $k=0$$$p^n=2^p-1\implies 2^p\equiv 1\pmod p$$Since $p$ has to be an odd prime with $\gcd(p,2)=1$, by Fermat's little theorem, we have $2^{p-1}\equiv 1\pmod p\implies 2^p\equiv 2\pmod p$. There are no odd primes satisfying both $2^p\equiv 1\pmod p$ and $2^p\equiv 2\pmod p$.

  • Case 2 : $k=1$$$p^{n-2}=1+\frac{p(p-1)}{2}+\binom p3p+\cdots +p^{p-2}+p^{p-2}$$Here, if $p$ is an odd prime, then the LHS is divisible by $p$ while the RHS isn't. So, we have to have $p=2$ to have $$2^{n}=(2+1)^2-1=2^3\implies n=3$$

It follows from the two cases that $\color{red}{(n,m,p)=(3,3,2)}$ is the only solution.


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.