6
$\begingroup$

I am seeking all positive integer solutions to the equation $x^3+y^3=3^z$.

After doing number crunching, I think there are no solutions. But I am unable to prove it.

Attempt

If $x$ and $y$ have common divisor $d$, we have $d^3(m^3+n^3)=3^z$. So $d$ must be a power of $3$, and we are back to where we started. So we assume $x$ and $y$ are coprime.

Testing the parity, we have sum of 2 cubes to be odd. WLOG, we can assume $x$ is even and $y$ is odd.

Trying mod $3$, we have $x+y=0 \pmod 3$. Since $x$ and $y$ are coprime, $x$ and $y$ must be congruent to $1$ and $-1$ or vice-versa.

If I assume $x=3m+1$ and $y=3n-1$, expand out and simplify, I get $27(m^3+n^3)+27(m^2-n^2)+9(m+n)=3^z$. If I assume $z \geq 3$, this gives $(m^3+n^3)+(m^2-n^2)+\frac{m+n}{3}=3^{z-3}$. But I don't see how to proceed.

I also tried mod $9$ but didn't get anywhere, it didn't cut down the possibilities by much.

I also tried letting $y=x+r$. Then \begin{align*} x^3+y^3 &= x^3+(x+r)^3 \\ &= x^3 + (x^3+3x^2r+3xr^2+r^3) \\ &= 2x^3+3x^2r+3xr^2+r^3 \\ &= 3^z \end{align*}

Then $3\mid 2x^3+3x^2r+3xr^2+r^3$, and $3\mid 3x^2r+3xr^2$, so this implies $3 \mid 2x^3+r^3$. But this doesn't yield any contradiction.

Can anyone supply a proof? Or if my hypothesis is wrong, how to derive all the integer solutions?

Thank you.

$\endgroup$
5
  • $\begingroup$ Maybe concentrate on $z$? It can’t be a multiple of $3$, due to Fermat. $\endgroup$ Sep 9, 2018 at 9:20
  • $\begingroup$ Are you familiar with the Lifting Exponent Lemma? Thus is a classic example of this. I think that there are an article in www.mathlinks.ro $\endgroup$ Sep 9, 2018 at 9:37
  • $\begingroup$ @RicardoLargaespada Can we apply it here? the LTE requires $p \mid x-y$ $\endgroup$
    – eatfood
    Sep 9, 2018 at 10:28
  • $\begingroup$ Also you have to read Zsigmondy Theorem, the solution by this technique is awesome $\endgroup$ Sep 9, 2018 at 15:52
  • $\begingroup$ I'm confused; isn't $(x,y,z)=(1,2,2)$ a solution? $\endgroup$ Nov 19, 2020 at 13:35

3 Answers 3

3
$\begingroup$

$27(m^3+n^3) +27(m^2-n^2) + 9(m+n) = 3^z$

$(m+n)(3m^2 +3n^2-3mn+3m-3n+1) = 3^{z-2}$

since $z \ne 2$, we have $3(m^2 +n^2-mn+m-n)+1$ divides a power of $3$ or is equal to $1$, both of which is not possible.

$\endgroup$
3
  • $\begingroup$ Now that you write it out, it seems really obvious XD. Thanks! $\endgroup$
    – eatfood
    Sep 9, 2018 at 12:10
  • $\begingroup$ You are welcome :) $\endgroup$
    – Shobhit
    Sep 10, 2018 at 17:49
  • $\begingroup$ But it is equal to $1$ for $(m,n)=(0,1)$. And why "...since $z\neq2$..."? It would make more sense to infer that $$3(m^2 +n^2-mn+m-n)+1,$$ must be a power of $3$, which happens only if it equals $1$, and from here conclude that $z=2$. $\endgroup$ Nov 19, 2020 at 13:39
3
$\begingroup$

You already proved that we can assume that $x,y$ are relatively prime.

We can find easly solution for $n\leq 2$. Let $n\geq 3$.

So $x+y = 3^a$ and $x^2-xy+y^2=3^b$ for some nonegative integers $a+b=n$.

If $b\geq 2$ then $9\mid x^2-xy+y^2$ and $3\mid x+y$, so $$9\mid (x+y)^2-(x^2-xy+y^2)=3xy\implies 3\mid xy$$ A contradiction. So $b\leq 2$ and this should be easly done.

$\endgroup$
1
  • $\begingroup$ that is really an ingenious way to show it. Thanks! $\endgroup$
    – eatfood
    Sep 9, 2018 at 12:10
1
$\begingroup$

Solution by LTE, let $d=(x,y)>1$, $x=dx_1$, $y=dy_1$, $(x_1,y_q)=1$ then $$d^3\mid x^3+y^3=3^z$$ then $d\mid 3^a$ for a positive integer $a<z/3$. Dividing by $d^3$ we obtain $x^3_1+y^3_1=3^b$, for a positive integer $b=z-a$. This is a equation similar to the original, then we can assume $(x,y)=1$. If $3\nmid x+y$ then $x+y=1$ a contradiction. Then we have all the conditions to use LTE: $$v_3(x^3+y^3)=v_3(x+y)+v_3(3)=v_3(x+y) +1=v_3(3^z)=z$$ From here we conclude that $x+y=3^{z-1}$ this implies that $x^2-xy+y^2=3$, $(x-y)^2+xy=3$, if $|x-y|\ge 2$ we have a contradiction, then there are two cases: Case 1. $|x-y|=0$ this is not possible due $x$ and $y$ are coprimes and $x+y>2$. Case 2. $|x-y|=1$ implying $xy=2$, with solutions $(x,y)=(2,1)$ or $(1,2)$, implying $3^{z-1}=2+1\Rightarrow z=2$.

Then the solutions to the original equation are: $(x,y,z)=(3^k,2\cdot 3^k,3k+2)$ or $(2\cdot 3^k, 3^k,3k+2)\forall k\in \mathbb{N}$.

$\endgroup$

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.