If a prime can be expressed as sum of two squares, then prove that the representation is unique.

My attempt:
If $a^2+b^2=p$, then it is obvious that $a,b$ of different parity.
Now, I assume the contraposition that the representation is not unique, $p=a^2+b^2=c^2+d^2$. Again, $c,d$ are of different parity.

Now, let $b,d$ be even and $a,c$ be odd.

So, $a^2+b^2=c^2+d^2 \implies a^2-c^2=d^2-b^2 \implies (a+c)(a-c)=(d-b)(d+b)$.

I cannot proceed any further. Please help.

  • $\begingroup$ Nitpick corner case: $2 = 1^2 + 1^2$, so not necessarily of different parity. This would most likely be taken care of in a different case. :) $\endgroup$ – apnorton Mar 20 '14 at 12:02
  • $\begingroup$ @anorton That is definitely something I didn't think about...thanks for identification $\endgroup$ – Hawk Mar 20 '14 at 12:02

The slickest way is via a little Algebraic Number Theory. If $p=a^2+b^2=c^2+d^2$ then $$p=(a+bi)(a-bi)=(c+di)(c-di)$$ Now ${\bf Z}[i]$ is a unique factorization domain, so these two factorizations of $p$ show that $a+bi$ can't be a prime in ${\bf Z}[i]$. We must have a non-trivial factorization $a+bi=(s+ti)(u+vi)$, whence $a-bi=(s-ti)(u-vi)$, and then $$p=(s^2+t^2)(u^2+v^2)$$ contradicting primality of $p$.

There are ways to answer your question without these advanced concepts, but I can never remember how it's done. I'm sure someone else will.

| cite | improve this answer | |
  • $\begingroup$ Thanks for the approach...I will try to do something with this. $\endgroup$ – Hawk Mar 20 '14 at 12:00
  • $\begingroup$ How do you deduce that $a+bi$ is not prime from the UFD property of $\mathbb{Z}[i]$? $\endgroup$ – Cary Sep 18 '16 at 15:59
  • $\begingroup$ @Cary, maybe it would be better to say that $a+bi$ and $c+di$ can't both be prime (in ${\bf Z}[i]$), since that would give us two different prime factorizations of $p$. Then without loss of generality we may assume $a+bi$ is not prime. $\endgroup$ – Gerry Myerson Sep 18 '16 at 22:59
  • $\begingroup$ @GerryMyerson: if $p$ is an ordinary prime, then $p$ is either a Gauss prime or the product of a Gauss prime and its conjugate. Clearly, we are in the latter case. Can't we just say that if $p=(a+bi)(a-bi)=(c+di)(c-di)$ are two prime factorizations, then $a+bi=u(c+di)$ or $a+bi=u(c-di)$ for $u$ a unit. In either case, we have $(a,b)=(\pm c, \pm d)$ or $(a,b) = (\pm d, \pm c)$, which implies $a^2+b^2=c^2+d^2$? $\endgroup$ – Cary Sep 19 '16 at 7:08
  • $\begingroup$ @Cary, I was trying to do it using just the UFD property, and not the result about how rational primes factor in the Gaussian integers. $\endgroup$ – Gerry Myerson Sep 19 '16 at 7:13

Here's an answer without Algebraic Number Theory. I found it in Shanks, Solved and Unsolved Problems in Number Theory.

Assume $$p=a^2+b^2=c^2+d^2\tag1$$ with all variables positive integers. Then $$p^2=(a^2+b^2)(c^2+d^2)=a^2c^2+a^2d^2+b^2c^2+b^2d^2$$ and you can verify by just multiplying everything out that $$p^2=(ac+bd)^2+(ad-bc)^2\tag2$$ and $$p^2=(ac-bd)^2+(ad+bc)^2\tag3$$ By (1), we have $$(p-a^2)d^2=(p-c^2)b^2$$ which implies $$p(d^2-b^2)=(ad-bc)(ad+bc)\tag4$$ From (4), $p$ divides $ad-bc$, or $p$ divides $ad+bc$. If $p$ divides $ad-bc$, then from (2) we get $ad-bc=0$, so $d^2-b^2=0$, so $b=d$. If $p$ divides $ad+bc$, then from (3) we get $ac=bd$. Now $a$ and $b$ are relatively prime, so $a$ divides $d$, and $b$ divides $c$. Then by (1) we have $a=d$, and we have proved that the two representations of $p$ are the same.

This is probably something like what @Konstantinos was getting at in his answer.

| cite | improve this answer | |
  • $\begingroup$ Just noting that $(a^2+b^2)(c^2+d^2)=(ac+bd)^2+(ad-bc)^2=(ac-bd)^2+(ad+bc)^2$ states that if a number is the product of two sums of squares,then the number can also be written as the sum of two squares.This is the sum of squares identity of Diophantus. $\endgroup$ – rah4927 Mar 21 '14 at 9:11

If $p=a^2+b^2=c^2+d^2$ then $$p=\frac{(ac+bd)(ac-bd)}{(a+d)(a-d)}$$

| cite | improve this answer | |
  • 3
    $\begingroup$ How is that supposed to help? $\endgroup$ – Hawk Mar 20 '14 at 17:48
  • $\begingroup$ It is a contradiction this shows that $p$ must be composite.Maybe i will edit it later when i will have time! $\endgroup$ – Konstantinos Gaitanas Mar 20 '14 at 18:01
  • $\begingroup$ Had any time, Konstantinos? $\endgroup$ – Gerry Myerson Oct 24 '19 at 21:36
  • $\begingroup$ @Gerry Myerson . What is the matter with you? It is not the first time you are trying to be ironic to me and others as well. $\endgroup$ – Konstantinos Gaitanas Oct 24 '19 at 21:48
  • $\begingroup$ Sorry, Konstantinos, I didn't know that trying to be ironic shows something is the matter with me. $\endgroup$ – Gerry Myerson Oct 24 '19 at 21:56

The problem is trivialized if you are given that $x^2+y^2=n$ has $r_2(n)=4\sum_{d|n} \sin(\pi d/2) $ solutions in the integers.

The number of ways $r_2(p)=4\sum_{d|p} \sin(\frac{\pi d}{2})=4(1+\sin(\pi p/2))$.

If $\sin(\pi p /2)=1$ then there are 8 solution in the integers. Namely, $(\pm a,\pm b)$ and $(\pm b, \pm a)$ satisfy $x^2+y^2=p$ where $a\neq b$. There is another way one might imagine $8$ solutions: $(0,\pm a),(\pm a,0)$ and $(\pm b,\pm b)$ but this cannot happen because it would mean that $p$ is a perfect square and therefore not prime.

If $\sin(\pi p /2)=-1$ then we cannot write $p$ as the sum of squares.

$If \sin(\pi p /2)=0$ then $p=2$.

| cite | improve this answer | |

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.