1

We know that $p=x^2+xy+y^2$ if and only if $p \equiv 1 \pmod {3}$.

But I need $a^2+ab+b^2 \equiv 0 \pmod{p} $ if and only if $p \equiv 1 \pmod {3}$, more generalized theorem.

I think that i should proof below :

$\exists a,b$ such that $a^2 + ab + b^2 \equiv 0 \pmod{p} \iff \exists x,y $ such taht $x^2 + xy + y^2 = p$

Easy to prove ($\Longleftarrow$), but not ($\implies$).

Image 1

These are some examples.

$2^2 + 2 \cdot 4 + 4^2 = 4( 1^2 + 1 \cdot 2 + 2^2) = 4 \cdot 7$

$3^2 + 3 \cdot 9 + 9^2 = 9( 1^2 + 1 \cdot 3 + 3^2) = 9 \cdot 13$

$7^2 + 7 \cdot 11 + 11^2 = 13( 2^2 + 2 \cdot 3 + 3^2) = 13 \cdot 19$

$5^2 + 5 \cdot 25 + 25^2 = 25( 1^2 + 1 \cdot 5 + 5^2) = 25 \cdot 31$

$10^2 + 10 \cdot 26 + 26^2 = 28( 3^2 + 3 \cdot 4 + 4^2) = 28 \cdot 37$

$6^2 + 6 \cdot 36 + 36^2 = 36( 1^2 + 1 \cdot 6 + 6^2) = 36 \cdot 43$

$13^2 + 13 \cdot 47 + 47^2 = 49( 4^2 + 4 \cdot 5 + 5^2) = 49 \cdot 61$

$29^2 + 29 \cdot 37 + 37^2 = 49( 2^2 + 2 \cdot 7 + 7^2) = 49 \cdot 67$

Wolgwang
  • 1,563
  • Short answer: $\mathbb{Z}[\omega]$ is a Euclidean domain, hence a $UFD$. The results about (prime) numbers of the form $a^2+b^2$ translates very directly. – Jack D'Aurizio Jul 09 '17 at 14:42
  • In other terms: there is a unique reduced binary quadratic form of discriminant $-3$ (this is an instance of the class number problem), hence if $a,b$ are coprime every prime divisor of $a^2+ab+b^2$ can be written in a similar fashion. – Jack D'Aurizio Jul 09 '17 at 14:46
  • Have a look at this answer, too: https://math.stackexchange.com/a/685982/44121 – Jack D'Aurizio Jul 09 '17 at 14:46
  • Worth remarking: you need to exclude the trivial case. $a=0=b$ always satisfies your congruence. – lulu Jul 09 '17 at 14:49
  • see yesterday's https://math.stackexchange.com/questions/2351978/prove-there-are-no-integer-solutions/2351981#2351981 – Will Jagy Jul 09 '17 at 18:00
  • We can probably combine the following identities$$(a^2 + 3b^2)(c^2 + 3d^2) = (ac + 3bd)^2 + 3(bc - ad)^2$$ and $$a^2 + 3b^2 = (b+a)^2 + (a+b)(b-a) + (b-a)^2$$

    to get this third identity and take it from here.

    $$\left((b+a)^2 + (b+a)(b-a) + (b-a)^2)((d+c)^2 + (d+c)(d-c) + (d-c)^2)\right) = \left((bc - ad + ac + 3bd)^2 + (bc-ad + ac+3bd)(bc -ad -ac - 3bd) + (bc -ad - ac - 3bd)^2\right)$$

    – sku Jul 10 '17 at 03:07
  • Thank you guys but i don't get it. How could your advices prove that $p, \exists a,b$ such that $a^2 + ab + b^2 \equiv 0 \pmod{p} \implies \exists x,y $ such taht $x^2 + xy + y^2 = p$? – Ryu Dae Sick Jul 12 '17 at 04:53
  • 1
    Jack D'Aurizio, how about $p=19$? The prime divisor of $7^2 + 7 \cdot 11 + 11^2$ are 13 and 19, two primes. How about $p=37$? That $a=10$ and $b=26$ are not comprime. Then how i guarantee 37 can be written in a form $a^2+ab+b^2$? I'm so confused more. Please help me, please... Your reference is saying about equal, not congruent, too... – Ryu Dae Sick Jul 12 '17 at 05:03
  • Thanks for other comment, I learn $a^2+ab+b^2$ can be expressed by product of same forms. But i don't understand why your answer guarantees $\exists x,y $ exactly $x^2+xy+y^2=p$ for odd prime p. – Ryu Dae Sick Jul 12 '17 at 05:09
  • @JackD'Aurizio Please answer. – Takahiro Waki Jul 12 '17 at 07:38

1 Answers1

1

Given any sum with a particular prime factor $p$ we can isolate the factor $p$ via Eisenstein integer multiplication.

Take your example with $10, 26$ having the sum $28×37$. How to isolate the factor $37$?

First get rid of the common factor $2^2$ which is totally useless:

$5^2+(5×13)+13^2=7×37$

Next we associate thus quadratic form with the Eisenstein integer $5-13\omega$ whose squared norm matches the value of the quadratic form. Generally: $|a-b\omega|^2=a^2+ab+b^2$.

Now $7$ has several sums that similarly match up with Eisenstein integers having that squared norm:

$7=1^2+(1×2)+2^2 \rightarrow 1-2\omega$ $7=2^2+(2×1)+1^2 \rightarrow 2-\omega$ $7=1^2-(1×3)+3^2 \rightarrow 1+3\omega$ $7=3^2-(3×1)+1^2 \rightarrow 3+\omega$ $7=2^2-(2×3)+3^2 \rightarrow 2+3\omega$ $7=3^2-(3×2)+2^2 \rightarrow 3+2\omega$

All we need do is multiply our Eisenstein integer for $7×37$ above, $5-13\omega$, by the above multipliers, and look for a product that has a factor of $7$ which we drop out. We will then have a combination equalling $37$ with integer inputs.

$(5-13\omega)×(1-2\omega)=(5×1-13×2)-(5×2+13×1+13×2)\omega=-21-49\omega=-7×(3+7\omega)$

So $|3+7\omega|^2=37$ meaning $3^2-(3×7)+7^2=37$.

To get a solution with positive integers we can multiply $3+7\omega$ by $\omega$ to get a product with mixed signs:

$(3+7\omega)×\omega=-7-4\omega$

$(-7-4\omega)×\omega=4-3\omega$

Then $37=4^2+(4×3)+3^2$.

All these manipulations turn on the Euclidean domain property of the Eisenstein integers so we are sure they work.

Oscar Lanzi
  • 39,403
  • First, thanks for your attention. You make this problem more easier. You said If $a^2+ab+b^2=Mp$ then $M=m^2qp$ for some $q \equiv 1 \pmod{3}$ , so $q=(x+y\omega)^2 $ for some $x,y$ and we can find $\alpha^2+\alpha \beta+\beta ^ 2=p$ exactly, right? But still i don't get why $M=m^2 q$ for some $q \equiv 1 \pmod{3}$ hold. – Ryu Dae Sick Jul 22 '17 at 00:44
  • Oh! I get it! Thank you! – Ryu Dae Sick Jul 22 '17 at 00:47