0
$\begingroup$

I want to have a proof of the fact that each prime number is the sum of a square and three times a square (Euler).

Context

I read the answer to my former question about the number of points on an elliptic curve over $\mathbb F_p$, and in the answer $p=a^2+3b^2$ is used.

I have looked in a Latin version of Gauss' Disquisitiones but I'm not familiar with Latin. And there are a lot of relations between (prime) numbers and squares but I can't see which one will lead to a prove of this question.

$\endgroup$
  • $\begingroup$ I read the answer to my former question about the number of points on an elliptic curve over Fp, and in the answer is used: p=a^2+3b^2 $\endgroup$ – user148584 Jul 29 '14 at 21:25
  • $\begingroup$ I have looked in a latin version of Gauss' Disquisitiones but I'm not familiar with latin. And there are a lot of relations between (prime) numbers and squares but I cant see which one will lead to a prove of this question. Overmore I do not understand the relation of this property for prime numbers congruent 1 mod 3 and the frobenius map, norm and determinant. $\endgroup$ – user148584 Jul 29 '14 at 21:29
  • $\begingroup$ A standard way is to go to $\mathbb{Z}[\omega]$, use the fact that $-3$ is a QR of $p$ for these primes, which means these primes factor. Same as the standard $\mathbb{Z}[i]$ proof for primes of the form $4k+1$. $\endgroup$ – André Nicolas Jul 29 '14 at 21:38
0
$\begingroup$

You might like David A. Cox, Primes of The Form $x^2 + n y^2$

Anyway, if you believe in quadratic reciprocity, $(-3|p) = (p|3)$ which you are saying is $1.$ The equation uses $3 \equiv 3 \pmod 4.$ That is, $(-7|p) = (p|7),$ but don't replce the $7$ by $5.$ So we can solve $$ \beta^2 \equiv -12 \pmod p. $$ Now, if $\beta$ is odd, replce it by $b=p-\beta$ which is even, so now $$ b^2 \equiv -12 \pmod {4p}. $$ So $$ b^2 = -12 + 4pc $$ and $$ b^2 - 4pc = -12. $$ There is a (positive) binary quadratic form $$ \langle p,b,c \rangle $$ This reduces to $$ \langle 1,0,3 \rangle $$ The inverse of the 2 by 2 matrix that accomplishes the reduction gives a representation of $p=u^2 + 3 v^2,$ as the left column.

$\endgroup$
0
$\begingroup$

An elementary proof. The ideas here can quickly be generalized to other cases such as $a^2+b^2$ or $a^2+7b^2$. Let $a^2\equiv -3\pmod p$, $q=\lfloor \sqrt p\rfloor$ and consider all numbers $f(x,y)=ax-y$ with $0\leq x, y\leq q$.

There are $(q+1)^2$ such numbers. Since $(q+1)^2\gt p$, there are two numbers $f(x_1,y_1)$ and $f(x_2, y_2)$ that are congruent $\pmod p$. This means $f(x',y')=f(x_1-x_2, y_1-y_2)$ is divisible by $p$. The pairs $(x_1,y_1)$ and $(x_2,y_2)$ are different solutions, so $(x,y)\neq(0,0)$.

Note that this implies that $(a|x_1-x_2|-|y_1-y_2|)(a|x_1-x_2|+|y_1-y_2|)$ is divisible by $p$, so there is a solution $(x,y)$ to $a^2x^2\equiv y^2\pmod p$ that respects $0\lt x,y\leq q$.

This shows a solution to $a^2x^2\equiv-3x^2\equiv y^2\pmod p$ with $0\lt x,y\leq q$. Using this to bound $3x^2+y^2$ gives $0 \lt 3x^2+y^2\leq 3p$. Let us consider three cases:

  1. $3x^2+y^2=3p$. Since $3\mid y^2$ we know $9\mid y^2$. Diving by three gives a solution $u^2+3v^2=p$.

  2. $3x^2+y^2=2p$. If this is true, $y$ and $x$ must have the same parity. If they were even, $4\mid 2p$, absurd - this shows they are odd. But if they are odd, $x^2\equiv y^2\equiv 1\pmod 8$, so $2p\equiv 4\pmod 8$. Since $p\equiv \pm 1\, \mathrm{or}\pm3\pmod 8$, this is also impossible.

  3. $3x^2+y^2=p$. This case is easy, we have already gotten our desired representation.

$\endgroup$

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.