Can anyone help me find the number of solutions to the equation:

$$ x^2 + y^2 + xy = (xy)^2 $$

Let me give a brief account of what I've tried to proceed with:

Case 1: One of $x$ and $y$ is odd. This results into a contradiction where the parity of LHS and RHS differs.

Case 2: Both of $x$ and $y$ are even. I have shown that no solution apart from $x=y=0$ exists and I've proved that using Infinite Descent.

Now I'm stuck in case $3$ where both $x$ and $y$ are odd.

  • $\begingroup$ Did you prove Case 2 with infinite descent? $\endgroup$ – TheRandomGuy Mar 23 '16 at 12:19
  • $\begingroup$ @Dhruv Yes, I did it using Fermat's method of Infinite Descent only. That's quite tedious to type. $\endgroup$ – Mathejunior Mar 23 '16 at 12:24
  • 2
    $\begingroup$ Try writing it a $(x^2-1)(y^2-1)=xy+1$. $\endgroup$ – Thomas Andrews Mar 23 '16 at 12:27
  • 3
    $\begingroup$ I thought of writing it as $$(x+y)^2 = xy(xy+1)$$ Now we find that in the RHS the two numbers are relatively prime and hence do not share a common multiple other than 1 and since 2 consecutive numbers can not be both squares (other than 1 and 0) and thus $(xy(xy+1))$ is not a perfect square and we reach a contradiction. The exception is of $x=y=0$. $\endgroup$ – TheRandomGuy Mar 23 '16 at 12:40

First, write it as $(x^2-1)(y^2-1)=xy+1$.

In the narrow ranges where $|x|\leq 1$ or $|y|\leq 1$, you can solve yourself.

Then, when $x>1$ and $x>1$ you have $(x^2-1)(y^2-1)>=2(x^2+y^2-2)=2(x-y)^2+4xy-4\geq 4xy-4$.

But $xy\geq 4$, so $4xy -4> xy +3\cdot 4 -4 \geq xy-1$.

The case when $x<-1$ and $y<-1$ is the same.

There are no cases when $x<-1$ and $y>1$ because then $xy+1<0$.

Finally, if $x=0$, $y=0$. If $x=-1$ then $y=1$, and if $x=1,$ $y=-1$.

The key is that in most cases, it is "obvious" that $(x^2-1)(y^2-1)$ is a lot bigger than $xy$.

Another approach. Let $d=xy$. Then assume $d\neq 0$ and you have:



$$x^4-(d^2-d)x^2 + d^2=0$$

So by the quadratic formula:

$$x^2=\frac{d^2-d \pm \sqrt{d^2(d-3)(d+1)}}{2}$$

When is $(d-3)(d+1)=(d-1)^2-4$ a perfect square? The only case of perfect squares that differ by four is $0$ and $4$. That means $d=3$ or $d=-1$.

The only cases left to handle then are $d=0$ and $d=3$.


You can write your equation (leaving the trivial $x=0,y=0$) as $$\frac xy+ \frac yx +1 =xy$$ now let $x/y=k$ and rearrange: $$k^2(1-y^2)+k+1=0$$ which has rational solutions only if $4y^2-3=p^2$. This implies that $y=\pm 1$. Hence the only integer solutions are $(-1,1)$ and $(1,-1)$.


I'm assuming you're looking for integer solutions, even though you didn't say os.

If you rewrite the LHS as $(x+y)^2 - xy$, your equations becomes $$(x+y)^2 = xy(xy + 1)$$

For positive $x$ and $y$:

Suppose (by swapping names if necessary) that $x \ge y$. Then the left hand side is no more than $4x^2$, while the right hand side is at least $y^2 x^2$.
We have $$ LHS = (x+y)^2 \le (x+x)^2 = 4x^2 $$ and $$ RHS = xy(xy + 1) > (xy)^2 = y^2 x^2. $$ If $y \ge 2$, then $$ RHS > y^2 x^2 \ge 2^2 x^2 = 4 x^2 $$ but RHS and LHS are equal, so we have a number that's both $ \le 4x^2$ and $> 4x^2$, and that's a contraduction.

That means that $y$ is at most one. That should get you on your way.

For the case where one of $x$ and $y$ is negative, you still have some work to do.

  • $\begingroup$ Why is $y$ at most 1? $\endgroup$ – TheRandomGuy Mar 23 '16 at 12:27
  • $\begingroup$ Yes..Why so John Hughes? Pl elaborate your point. $\endgroup$ – Mathejunior Mar 23 '16 at 12:30
  • $\begingroup$ Expanded details to show contradiction if $y \ge 2$. $\endgroup$ – John Hughes Mar 23 '16 at 14:48

We need $$x^2(1-y^2)+xy+y^2=0$$

The discriminant is $$y^2-4y^2(1-y^2)=y^2(4y^2-3)$$

So, $4y^2-3$ needs to be perfect square




Your answer is three, and in my opinion, a basic algebraic approach most clearly shows the solution. This equation is comprised of symmetric polynomials $x^2+y^2$ and $xy$, so finding one solution, $(x,y)=(p,q) \ | \ p \neq q$ is equivalent to finding two $(x,y)=(p,q),(q,p)$. If we let algebra drive:

$x^2 + y^2 + xy = (xy)^2 \to \\ (x + y)^2 = (xy)^2+(xy) \to \\ (2x + 2y)^2 = 4(xy)^2+4(xy) \to \\ (2x + 2y)^2 = 4(xy)^2+4(xy) +1-1\to \\ (2x + 2y)^2 = (2xy+1)^2-1 \implies$


Then we arrive at a difference of two squares being $1$. It can only imply that $$ \begin{cases} 2xy+1&=(\pm 1)\\ 2x+2y&=0 \end{cases} $$ From the second $x=-y$. Putting this into the first yields $$-2y^2+1=(\pm 1) \to \\ y^2=\frac{1-(\pm 1)}{2}$$ So $(x,y)=(0,0),(1,-1),(-1,1)$


Note that $$t^2-1\geq |t|+1>0$$ for all real numbers $t$ such that $|t|\geq 2$ (this is because $\big(|t|-2\big)\big(|t|+1\big)\geq 0$). Therefore, if $|x|\geq 2$ and $|y|\geq 2$, then $$\left(x^2-1\right)\left(y^2-1\right)\geq \big(|x|+1\big)\big(|y|+1\big)>|xy|+1\geq xy+1\,.$$ This means $x^2+y^2+xy<(xy)^2$. Thus, any solution $(x,y)\in\mathbb{Z}\times\mathbb{Z}$ to $x^2+y^2+xy=(xy)^2$ must come from the case where $|x|\leq 1$ or $|y|\leq 1$.

Clearly, $x=0$ if and only if $y=0$. We now exclude the solution $(x,y)=(0,0)$, so we are left with the case $|x|=|y|=1$. This means $$xy+1=\left(x^2-1\right)\left(y^2-1\right)=0\,,$$ so $xy=-1$. This implies $(x,y)=(-1,+1),(+1,-1)$.


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.