7
$\begingroup$

I can't find a way to prove that the following equation has only one solution :

$$ X = \frac{2^Q - 1}{2^{P+Q} - 3^P} $$

with $X,P,Q$ integers $> 0$.

One trivial solution is $X = 1, P = 1, Q = 1$.

Does anyone has an idea ? Best regards

$\endgroup$
  • $\begingroup$ This is very near related to an (open) conjecture of a detail in the Waring-problem of sums of like powers. I cannot at the moment show the exact relation/transformation, but you might search for keyword "waring" in MSE where I've given answers datailing this problem, sometimes to questins of the user Fred Kline there. I can come back to this possibly in the evening (west European time) $\endgroup$ – Gottfried Helms Jun 25 '15 at 11:52
  • $\begingroup$ Thank you everybody for your answers. I've read a little about Ray Steiner proof, and as commented by Lagarias "it is surprising that such heavy weapons are needed for such a tiny result". I was thinking ther was an easier proof, something like Gersonides proof for 3^n-2^n=1 $\endgroup$ – BenLaz Jun 25 '15 at 13:05
7
$\begingroup$

I'm more used to the formulation in the following form: $$ X(2^{P+Q} - 3^P)=2^Q-1 \\ 2^Q(2^Px -1) = 3^Px -1 $$ and then $$ 2^Q = {3^P \cdot X - 1 \over 2^P \cdot X - 1} \tag 1 $$ and Ray Steiner has proved in 1976 in the context of the Collatz-problem (using Rhin's result given in the other answer), that there is only one solution (which you've already noticed). A bit more about this (and the references) can be found in the wikipedia-article on the Collatz-problem and also in a remark in Lagarias' survey about the research in the Collatz-problem.

Footnote: The Waring-problem (which I mentioned in my comment just a minute ago) leads to a small modification; there the lhs is only required to be integer (instead of a power of 2) and even this conjecture (that there are no solutions for $P \gt 6$ ) seems to hold.

| cite | improve this answer | |
$\endgroup$
4
$\begingroup$

Unless $3^P$ is very close to $2^{P+Q}$, the right hand side will be smaller than 1. Hence the linear form $(P+Q)\log 2 - P\log 3$ is exceptionally small, and you should be able to obtain effective upper bounds for $P$ and $Q$ by Baker's method. Looking at the continuous fraction of $\frac{\log 2}{\log 3}$ you can probably reduce the upper bound to a range where you can check everything using a computer.

| cite | improve this answer | |
$\endgroup$
  • 2
    $\begingroup$ Georges Rhin has a paper where he computes an explicit irrationality measure for $\log 2/\log 3$. He shows that $|a\log 2+b\log 3+c|\le \max(|a|,|b|,|c|)^{-13.3}$. That should make a very small search indeed. $\endgroup$ – Anthony Quas Jun 25 '15 at 11:46
  • $\begingroup$ @Anthony : didn't the argument of Rhin actually mean, that the approximation of the sum of logarithms to the nearest integer is actually bad instead of good so instead of "$\le$" in your formula you needed "$\ge$" ? (I don't have access to Rhin's paper, it seems just logical to me; also I found in a paper of J. Simons/B.de Weger the argument $ (K+l) \log2 - K \log3 \gt e^{-13.3(0.46057+ \log K)} \qquad $ with the reference to Rhin's result. ) $\endgroup$ – Gottfried Helms Jul 17 '15 at 20:33
  • $\begingroup$ @Gottfried: Yes exactly. Thanks for the correction. $\endgroup$ – Anthony Quas Jul 17 '15 at 20:50

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.