4

In the course of answering another question here (link) I struggled quite hard at a rather unsatisfactory argument to show that $$x^6-y^6=z^2,$$ has no nontrivial integral solutions. For that question it was sufficient to assume $y$ even. From there is not much more work to prove it in full, though. It did leave me wondering whether there is a simpler argument. For completeness' sake I repeat my argument here:

Proposition: The Diophantine equation $x^6-y^6=z^2$ has no primitive integral solutions with $xyz\neq0$.

Proof. Let $x$, $y$ and $z$ be pairwise coprime integers such that $x^6-y^6=z^2$, with $|y|>0$ minimal. Then $(u,v,w)=(x^2,-y^2,z)$ satisfies $$u^3+v^3=w^2,$$ and it is a classical result by Euler that any primitive solution is of one of the following three forms: \begin{align} (1)&\begin{cases} u&=s(s+2t)(s^2-2st+4t^2)\\ v&=-4t(s-t)(s^2+st+t^2)\\ w&=\pm(s^2-2st-2t^2)(s^4+2s^3t+6s^2t^2-4st^3 + 4t^4) \end{cases} & 2\nmid s,\ 3\nmid(s-t),\\ \\ (2)&\begin{cases} u&=s^4-4s^3t-6s^2t^2-4st^3 + t^4\\ v&=2(s^4+2s^3t+2st^3+t^4)\\ w&=3(s-t)(s+t)(s^4+2s^3t+6s^2t^2+2st^3+t^4) \end{cases} &2\nmid(s-t),\ 3\nmid(s-t),\\ \\ (3)&\begin{cases} u&=-3s^4 + 6s^2t^2 + t^4\\ v&=3s^4+6s^2t^2-t^4\\ w&=6st(3s^4+t^4) \end{cases} &2\nmid(s-t),\ 3\nmid t, \end{align} after swapping $u$ and $v$ if necessary. In all three cases $s$ and $t$ are coprime integers. For a proof, see $\S 14.3.1$ of Henri's Cohen's Number Theory, Vol II, Analytic and Modern Tools.

If $y$ is even then $w=z$ is odd, so the solution $(u,v,w)=(x^2,-y^2,z)$ cannot be of the form $(3)$. It cannot be of the form $(2)$ because $-y^2\equiv0\pmod{4}$, whereas $u\equiv1\pmod{4}$ and $v\equiv2\pmod{4}$. It folows that it is of the form $(1)$, and so \begin{eqnarray} x^2&=&s(s+2t)(s^2-2st+4t^4)&=&s(s^3+8t^3),\\ y^2&=&4t(s-t)(s^2+st+t^2)&=&4t(s^3-t^3) \end{eqnarray} for two coprime integers $s$ and $t$ with $s$ odd and $3\nmid(s-t)$. The factors in each product are pairwise coprime, and hence they are all perfect squares, say $$s=S^2,\qquad t=T^2,\qquad s+2t=q^2,\qquad s^3-t^3=p^2,$$ for some integers $S$, $T$ , $p$ and $q$, not necessarily prime. Reducing mod $4$ shows that $t$ is even, and so we find $$S^6-T^6=p^2,$$ with $S$, $T$ and $p$ pairwise coprime and $T$ even, where clearly $|S|<|x|$ and $|T|<|y|$. This contradicts the minimality of $|y|$, and so we conclude that there does not exist any primitive solution with $y\neq0$ even.

If $y$ is odd then $z$ is even, and we have $$y^6+z^2=x^6,$$ which shows that $(y^3,z,x^3)$ is a primitive Pythagorean triple. This means there exist coprime integers $m$ and $n$ such that $$y^3=m^2-n^2,\qquad z=2mn,\qquad x^3=m^2+n^2,$$ because $z$ is even. Then $m+n=s^3$ and $m-n=t^3$ for two coprime integers $s$ and $t$, because $$y^3=m^2-n^2=(m+n)(m-n),$$ where $m+n$ and $m-n$ are coprime. It follows that $$2x^3=2(m^2+n^2)=(m+n)^2+(m-n)^2=(s^3)^2+(t^3)^2=(s^2)^3+(t^2)^3,$$ where $s$, $t$ and $x$ are pairwise coprime. But by the Main Theorem of this article no such solution exists, a contradiction.$\qquad\square$

Servaes
  • 63,261
  • 7
  • 75
  • 163

1 Answers1

1

There is a less complicated method than the one given by "OP" to show that a numerical solution is not possible to the below equation:

Consider the equation:

$w^2=a^6-b^6=(a^2-b^2)(a^4+a^2b^2+b^4)$

we impose, $(a^2-b^2)=p^2$ &

$(a^4+a^2b^2+b^4)=q^2$

We make 'p' a square by taking:

$a=m^2+n^2$,$b=m^2-n^2$ & $p=2mn$

Hence after substituting value of $(a,b)$ we get:

$(a^4+a^2b^2+b^4)$=$(m^4+3n^4)(3m^4+n^4)$

Let:

$(m^4+3n^4)=u^2$ ---(1)

$(3m^4+n^4)=v^2$ ---(2)

Above simultaneous only has numerical solution:

$(u,v,n)=(6,10,1)$ &

$(m^4)=(33)$

Since, the numerical value of (m) is irrational there cannot be a numerical (integral) solution to the equation proposed by "OP".

Addendum:

$(a^6−b^6)=w^2$ ----(1)

We take:

$w^2=(p^2+pq+q^2)−(pq)-2q^2$

Where, $(p,q)=(a^3,b^3)$

We make, $(p^2+pq+q^2)$ a square by taking,

$(p,q)=[(u^2−v^2),(2uv+v^2)]$

Hence,

$(a^6−b^6)=(u^2+uv+v^2)^2−v(2u+v)(u^2+4uv+v^2)$

Above (RHS) is a square at,

$(u,v)=(−1,2)$

Above value for, $(u,v)$ gives trivial solution for equation (1). Hence equation (1) has no numerical integral solution.

David
  • 71
  • -1 What if $a^2-b^2$ and $a^4+a^2b^2+b^4$ are not coprime? What if $p$ is odd? And once you get those expressions for $u^2$ and $v^2$, how do you know that there are no other solutions? But how do you know that $m4+3n^4$ and $3m^4+n^4$ are coprime in the first place? And how does this fit with the fact that there are integral solutions such as $(x,y,z)=(1,1,0)$? – Servaes Mar 29 '24 at 18:52
  • @Servaes. See the addendum I added to my answer today. – David Mar 30 '24 at 16:19
  • The addendum does not address any of the issues I point out in my comment above. Also the addendum has many similar problems. None of this is anywhere near a coherent argument. – Servaes Mar 31 '24 at 15:26