We have the following result:

If $a$, $b$ and $c$ are pairwise coprime integers, such that

$$\frac 1a+\frac 1b=\frac 1c$$

then $a+b$, $a-c$ and $b-c$ are perfect squares.

What I did.

I tried to prove first that $a+b$ is a perfect square.

Multiplying by $abc$ we get


Since $a$ and $c$ are coprime, we get by Gauss' lemma that

$$a\mid a+b.$$


$$b\mid a+b.$$

Since $a$ and $b$ are coprime,

$$ab\mid a+b.$$

But we obviously have $a+b\mid ab$, so


and $c=1$.

I think this would imply $a=b=2$ which is absurd since $a$ and $b$ are coprime...

The question. Is this going somewhere? What else could I do?

  • 1
    $\begingroup$ see this $\endgroup$ – lulu Feb 11 '17 at 16:35
  • 1
    $\begingroup$ Is there any triple satisfying the assumptions ? I did not find any yet. $\endgroup$ – Peter Feb 11 '17 at 16:39
  • 1
    $\begingroup$ @Peter I don't think there is... all the examples provided in lulu's link don't satisfy the condition that $a$, $b$ and $c$ are pairwise coprime. $\endgroup$ – E. Joseph Feb 11 '17 at 16:40

The equation $\frac1a+ \frac1b = \frac1c$ is equivalent to $ab=c(a+b) \implies (a−c)(b−c)=c^2$.

Since $a,b,c$ are positive integers, ${1\over{a}}<{1\over{c}}$ and ${1\over{b}}<{1\over{c}}$. Hence $a>c$ and $b>c$. For each pair of integers $x,y $ satisfying $xy=c^2$, we get $a−c=x$ , $b−c=y$.

$\gcd(a,b) = 1$ thus if a prime $p$ divides both $x,y$ we have $p$ dividing both $a,b$; thus $\gcd(x,y) = 1$. And since $xy=c^2$ we have individually $x = z_1^2$ and $y= z_2^2$.

$a−c=x$ , $b−c=y$. Thus $a+b = 2c +x+y = (z_1+z_2)^2$. Hence $a+b$ is a prefect square.

In addition we have also proved that $a-c$ and $b-c$ are perfect squares.

  • 1
    $\begingroup$ But if $xy=c^2$, $x=z_1^2$, $y=z_2^2$, then $c=z_1z_1$. But then $a-c=x$ implies $a = c + x = z_1z_2 + z_1^2 = z_1(z_1+z_2)$, and therefore $z_1 \mid \gcd(a,c)=1$ thus $z_1=1$. The same argument for $b-c$ gives $z_2=1$. Thus we get $c=1$, and thus $a=1+1=2$ and $b=1+1=2$, thus $\gcd(a,b)=2\ne 1$. In other words, the conditions cannot be fulfilled. $\endgroup$ – celtschk Feb 11 '17 at 17:11

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.