So Wiles' proof showed that no three positive integers $a$, $b$, and $c$ can solve the equation $a^n+b^n=c^n$ for any integer value of $n$ greater than $2$. Now what about the opposite?

What does this mean for any $a$ greater than $2$, and $x$, $y$ and $z$ are positive integers in the equation $a^x+a^y=a^z$. Is there any relation? Is it solvable?

  • $\begingroup$ Is it necessary that $x, y$ and $z$ are different? $\endgroup$ – Aldon Sep 2 '15 at 23:32
  • 49
    $\begingroup$ "Opposite of Fermat's Last Theorem" ... Fermat's first theorem? $\endgroup$ – Au101 Sep 3 '15 at 0:47
  • 11
    $\begingroup$ @Au101 Tamref's first theorem? $\endgroup$ – Michael Sep 3 '15 at 5:16
  • 2
    $\begingroup$ Did you see this in the book "Gödel, Escher, Bach"? $\endgroup$ – Stig Hemmer Sep 3 '15 at 10:10
  • 7
    $\begingroup$ @Lucas He calls it Fourmi's Well-Tested Conjecture (that no nontrivial solutions exist), in the chapter Ant Fugue: "I have discovered a truly marvelous proof of this statement, which, unfortunately is so small that it would be well-nigh invisible if written in the margin." $\endgroup$ – Peter Sep 3 '15 at 16:23

Clearly we must have $x, y<z$. Suppose $x\not=y$; WLOG, suppose $x<y$. Then this can be rewritten as $$a^x(1+a^{y-x})=a^xa^{z-x},$$ with $z-x, y-x>0$. This gives $$1+ a^{y-x}=a^{z-x};$$ but the right hand side is divisible by $a$, while the left hand side is not.

So we must have $x=y$. But then this yields $$2a^x=a^z,$$ which in turn yields $$2=a^{z-x}$$ with $z-x>0$. So there are no nontrivial solutions.

  • 1
    $\begingroup$ Except in the case of $a=2$, in which case $z=x+1$. This is simply the fact that twice a power of two is the next power of two. $\endgroup$ – R.. Sep 3 '15 at 4:21
  • 7
    $\begingroup$ I consider $a=2$ a trivial solution. $\endgroup$ – Noah Schweber Sep 3 '15 at 4:21
  • 6
    $\begingroup$ The question explicitly asks "What does this mean for any a greater than 2". $\endgroup$ – user81060 Sep 3 '15 at 12:42
  • 13
    $\begingroup$ @MarkBannister Which is why I consider $a=2$ a trivial solution. $\endgroup$ – Noah Schweber Sep 3 '15 at 15:49
  • $\begingroup$ @MarkBannister What if we include real (or complex) numbers as basis? Ie, which $x\in \mathbb R \ [\mathbb C]$ are solutions to $x^m - x^n = 1$ for some integer $m$, $n >2$? Does it seem like a reasonable question? $\endgroup$ – eudes Sep 4 '15 at 11:56

Here's a less algebraic way to interpret your question: Look at things in base $a$. Then the equation $a^x+a^y=a^z$ looks like $1 \ldots 0 + 1 \ldots 0 = 1\ldots0$. (where the number of zeros is $x$, $y$, or $z$) This can only work if we're in base $2$ and the addition carries to produce a $1$ in the next digit.

  • 4
    $\begingroup$ Brilliant method. $\endgroup$ – nbubis Sep 3 '15 at 21:43

The equation $a^x+a^y=a^z$ has no solutions in positive integers with $a\gt 2$. For suppose the relation holds. Without loss of generality we may assume $x\le y$. Then $$1+a^{y-x}=a^{z-x}.$$ This is only possible if $y=x$, $a=2$, and $z=x+1$.

  • $\begingroup$ It doesn't even have a solution if $a = 2$, you'd always have $1 + 1 = 1$ which is obviously wrong. Only possible solution would be the trivial case $a = 0$. $\endgroup$ – Aldon Sep 2 '15 at 23:36
  • 4
    $\begingroup$ @Aldon Careful: $2+2=4$ :P (Or, more generally, if $x=y$, $z=x+1$, and $a=2$, then $a^x+a^y=a^z$.) $\endgroup$ – Noah Schweber Sep 2 '15 at 23:37
  • $\begingroup$ It works if a=x=y=2 and z=3 $\endgroup$ – Lucas Sep 2 '15 at 23:38
  • 1
    $\begingroup$ No sorry I was meant to say $a = 1$ that's why I wrote down $1+1=1$. Sorry for the typo. $\endgroup$ – Aldon Sep 2 '15 at 23:44

Informally, but intuitively:

Let's say $a = 3$. Then, we're looking at the sequence $\{1, 3, 9, 27, 81\}$. The numbers are clearly too far apart for adding two of them to reach the next one.

This leads us to a slightly more formal proof:

Let's say that $x \leq y$. Then, $$a^x + a^y \leq a^y + a^y = 2a^y < a(a^y) = a^{y+1}$$ So $z$ must be between $y$ and $y + 1$, so there's no possible solution.

We can also note that $$\sum_{k=1}^{k=z-1}{a^k} = {a^z-1\over a-1} < a^z$$

Meaning, even if we added all the exponents of $a$ less than $z$, we still wouldn't reach $a^z$.


It is clear that $1+a^{y-x}=a^{z-x}$ where $a=1$,$x=y=z$ is an obvious solution. On the other hand the minimun difference of powers of integers is of the forme $2k+1$ which corresponds to squares.So, (making abstraction of the previous division by zero!) this would gives $a=0$.

  • 3
    $\begingroup$ $a=1$ doesn't work, since $1+1\not=1$. $\endgroup$ – Noah Schweber Sep 3 '15 at 2:36

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.