EDIT: Please see EDIT(2) below, thanks very much.

I want to prove by infinite descent that the positive divisors of integers of the form $a^2+3b^2$ have the same form. For example, $1^2+3\cdot 4^2=49=7^2$, and indeed $7,49$ can both be written in the same form: $49$ is already shown, and $7$ is $2^2+3\cdot 1^2$.

I'm looking for a general proof of this using infinite descent. I tried to let $a^2+3b^2=xy$, where $x,y$ are positive integer factors of the LHS, but couldn't continue from here I'm afraid. However, I have managed to prove that integers of the form $a^2+3b^2$ are closed under multiplication, i.e. the product of two of them is of the same form. I was hoping this would help.

I also had another question if you have time: Is the representation in the form $a^2+3b^2$ unique? For example, is it possible for an integer, say $N$, to be $N=a^2+3b^2=x^2+3y^2$ with $a\ne x$ and $b\ne y$?

EDIT: Sorry, here is the "closed under multiplication" proof I found:

Using Diophantus' Identity we have $(a^2+3b^2)(x^2+3y^2)=(a^2+(\sqrt{3}b)^2)(x^2+(\sqrt{3}y)^2=(ax+\sqrt{3}\cdot \sqrt{3} y)^2)+(a\cdot \sqrt{3} y - \sqrt{3} b\cdot x)^2=(ax+3by)^2+3(ay-bx)^2$ as required.

EDIT (2): I am very sorry. The question should be: If $\gcd(a,b)=1$, then every odd factor of $a^2 + 3b^2$ has this same form.

  • $\begingroup$ Note that if $p$ is not of this form then using $pa$ and $pb$ gives a number with a factor $p$, so you have to have some conditions on $a$ and $b$ to exclude such cases. $\endgroup$ – Mark Bennet Aug 5 '14 at 16:19
  • $\begingroup$ But $a=b=1$ is still a counterexample, $(a,b)=1$ and $2$ cannot be expressed as $x^2 + 3y^2$ $\endgroup$ – Darth Geek Aug 5 '14 at 21:33

EDIT: added second answer proving the main question.

Since we've proven that the family of numbers in the form of $a^2+3b^2$ is closed unfer product, we just jave to prove that any prime divisor $d$ of $a^2 + 3b^2$ can be represented as $r^2 + 3s^2$ for some integers $r,s$.

Let $d\ \vert \ (a^2 + 3b^2)$ then, $dk = a^2 +3b^2$ for a certain $k$.

If $d=1$ or $k = 1$ then the representation is obvious.

($d = 1^2+3·0^2$ for the first case and $d = a^2 + 3b^2$ for the second)

By integer division:

$$a = md + r \hspace{2cm} b=nd+s$$

If $r >d/2$ then $a = (m+1)d + (r-d) = m'd + r'$ where $m' = m+1$ and $r' = r-d$ (yes, $r'$ is negative, that's fine, we just want that either $r < d/2$ or $-r' < d/2$

(we use strict inequalities since $d$ is odd and therefore $d/2 \notin \mathbb{Z}$)

Similarly, if $s > d/2$ we construct $n'$ and $s'$ acordingly.

(use $m', r'$ and/or $n', s'$ from here on if you constructed them)

So $$a^2 +3b^2 = (md+r)^2 + (nd+s)^2 = $$

$$m^2d^2+2mdr+r^2+3n^2d^2+6nds+3s^2 = d(m^2d+2mr+3n^2d+6ns)+r^2+3s^2$$

But $d\ \vert\ (a^2 + 3b^2)$, therefore $d\ \vert\ (r^2 + 3s^2)$

But $r < d/2$ and $s < d/2$ therefore $r^2 + 3s^2 < \displaystyle\frac{d^2}{4} + \frac{3d^2}{4} = d^2$, so $r^2 + 3s^2 = d$

(remember that $d$ is prime, so the only multiples of $d$ lower than $d^2$ are $d$ and $0$ and $r^2 + 3s^2\neq 0$ because otherwise $r = s = 0$ what would mean that $d$ divides both $a$ and $b$)

Thus any odd divisors, $d$ of $a^2 + 3b^2$ can be written as $r^2 + 3s^2$ for some integers $r$ and $s$.

NOTE: If you don't asume that $d$ is prime, then $d$ is a divisor of $r^2 + 3s^2 < a^2 + 3b^2$ and you can create the descent argument you wanted.

  • $\begingroup$ Hi, this is an old answer but I wonder if I could seek some clarification - (remember that..) Why does d being prime mean that multiples of d other than d and 0 aren't allowed as a value for the expression in question? $\endgroup$ – AnotherJohnDoe Aug 6 '17 at 15:27
  • 1
    $\begingroup$ @AnotherJohnDoe I should proofread my proofs before posting them. This was from three years ago (to the day!). Just glancing over it I can see that there are some typos and I wouldn't be surprised to find some mistakes (how embarrasing!). Let me check it a bit further and I'll edit the answer with either a detailed explanation or an alternative reasoning. $\endgroup$ – Darth Geek Aug 6 '17 at 17:36

Note that: $$(ax-3by)^2 + 3(ay+bx)^2 = (a^2+3b^2)(x^2+3y^2) = (ax+3by)^2 + 3(ay-bx)^2$$

So if all divisors of numbers of that form are also of that form, then only numbers of that form that are primes have unique representation.


This is false as stated. $4$ has an expression as $a^2 + 3 b^2,$ with $a=b=1,$ but $2$ does not. $25$ has such an expression, with $a=5,b=0,$ but $5$ itself does not. The same applies to any $q^2,$ where $q \equiv 2 \pmod 3$ is prime.


For your second question. Not unique. $$28=4^2+3\cdot 2^2=5^2+3\cdot 1^2=1^2+3\cdot 3^2.$$ $$84=6^2+3\cdot 4^2=3^2+3\cdot 5^2.$$

  • $\begingroup$ @EulerianAdventurer: I found another example:) $\endgroup$ – mathlove Aug 5 '14 at 16:25
  • $\begingroup$ @EulerianAdventurer: haha, I hope so:) $\endgroup$ – mathlove Aug 5 '14 at 16:49

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.