When I noticed that $3^3+4^3+5^3=6^3$, I wondered if there are any other times where $(a-1)^3+a^3+(a+1)^3$ equals another cube. That expression simplifies to $3a(a^2+2)$ and I'm still trying to find another value of $a$ that satisfies the condition (the only one found is $a=4$)

Is this impossible? (It doesn't happen for $3 \leq a \leq 10000$) Is it possible to prove?

  • 1
    $\begingroup$ I think you mean $a=4$, not $a=2$: $1^3 + 2^3 + 3^3 = 36$ is not a cube, but $3^3 + 4^3 + 5^3 = 6^3$ $\endgroup$ – Robert Israel Mar 14 '12 at 22:07
  • $\begingroup$ @RobertIsrael Yup, I changed that :) $\endgroup$ – itdoesntwork Mar 14 '12 at 22:10
  • 1
    $\begingroup$ Have you read this article? It seems that that is the only three consecutive integers whose cubes sum to a cube. $\endgroup$ – Jack Mar 14 '12 at 22:11
  • 1
    $\begingroup$ It is also known as a Plato's number en.wikipedia.org/wiki/Plato%27s_number $\endgroup$ – Jeremy Carlos Mar 14 '12 at 22:12
  • $\begingroup$ @Jack is it difficult to prove? I think I've read that but still tried to find one. $\endgroup$ – itdoesntwork Mar 14 '12 at 22:16

How about $$(-1)^3+0^3+1^3=0^3?$$


Here's a start. If you notice, $\gcd(a,a^2+2)=\gcd(a,2)$, so by the unique factorization theorem, $$ \eqalign{ & 3a(a^2+2)=z^3 \\ & \text{is a perfect cube} } ~\iff~ \matrix{ a &=& a_1^3\cdot2^{a_2}\cdot3^{a_3} \\ a^2+2 &=& b_1^3\cdot2^{b_2}\cdot3^{b_3} } \quad \text{with} \quad \eqalign{ & 2,3\not\mid a_1,b_1 \\ & a_2+b_2\equiv0\pmod3 \\ & a_3+b_3\equiv2\pmod3, } $$ with all $a_i,b_i$ nonnegative except for $a_1$, which has the same sign as $a$. This has known solutions $$a=0,\quad\text{for which}\quad \mathbb{a} =\left[\matrix{a_1\cr a_2\cr a_3}\right] =\left[\matrix{0\cr0\cr0}\right],\quad \mathbb{b} =\left[\matrix{b_1\cr b_2\cr b_3}\right] =\left[\matrix{1\cr1\cr0}\right], \tag{solution 0} $$ and $$a=\pm4,\quad\text{for which}\quad \mathbb{a}=\left[\matrix{\pm1\cr2\cr0}\right],\quad \mathbb{b}=\left[\matrix{1\cr1\cr2}\right]. \tag{solution 1} $$ But then for $a\ne0$, assume without loss of generality that $a>0$ (so $a_1,b_1\ge1$), and note that, looking at the powers of two and three in its prime factorization, $$ \eqalign{ b_1^3\cdot2^{b_2}\cdot3^{b_3} & = a^2+2 = a_1^6\cdot2^{2a_2}\cdot3^{2a_3}+2 \\ & = 2 \left( a_1^6\cdot2^{2a_2-1}\cdot3^{2a_3} + 1\right) } $$ implies that either $a_2\ge1=b_2$ (case 2.1) or $a_2=b_2=0$ (case 2.0), and by similar reasoning, either $a_3>0=b_3$ (case 3.1), or else $a_3=0$ and $b_3\equiv2\pmod3$, which I will call case 3.0. This leaves us with four combinations to check, and hopefully either rule out or simplify: $$ \matrix{ \\\text{case }0=2.0\wedge3.0:&\quad a_2=a_3=b_2=0 ,\quad b_3\equiv2\pmod3 \\\text{case }1=2.1\wedge3.0:&\quad a_2\ge1=b_2,\quad a_3=0,\quad b_3\equiv2\pmod3 \\\text{case }2=2.0\wedge3.1:&\quad a_3>0=a_2=b_2=b_3 \\\text{case }3=2.1\wedge3.1:&\quad a_2,a_3\ge1=b_2,\quad b_3=0. } $$ These four cases comprise all nontrivial (i.e. other than solution 0), (WLOG) postivie solutions, and we hope to deduce that only case 1 with solution 1 is viable. Case 0 simplifies to $$ b_1^3 \cdot 3^{b_3} = a_1^6+2 \qquad\text{for}\qquad a_1,b_1\not\equiv0\pmod{2,3} \quad\text{and}\quad b_3\equiv2\pmod3. $$ Since $b_3\ge2$, we have $a_1^6\equiv-2\pmod9$, which has no solutions $a_1$ (sixth powers are always $0$ or $1$ modulo $9$), ruling this case out. Case 2 simplifies to $$ b_1^3 = a_1^6 \cdot 3^{2a_3} + 2 \qquad\text{for}\qquad a_1,b_1\not\equiv0\pmod{2,3} \quad\text{and}\quad a_3 > 0, $$ so that $b_1^3\equiv2\pmod9$, which we can also rule out, since cubes are always $0$ or $\pm1$ modulo $9$. Case 3 simplifies to $$ 2b_1^3 = 2^{2a_2 } 3^{2a_3} a_1^6 + 2 = 2\left( 2^{2a_2-1} 3^{2a_3} a_1^6 + 1 \right) \qquad\text{for}\qquad a_1,b_1\not\equiv0\pmod{2,3} \quad\text{and}\quad a_2,a_3 > 1, $$ so that $b_1^3\equiv1\pmod9\implies b_1$ is even, a contradiction. Lastly, case 1 simplifies to $$ 3^{b_3} \cdot b_1^3 = 2^{2a_2-1} a_1^6 + 1 \qquad\text{for}\qquad a_1,b_1\not\equiv0\pmod{2,3}, \quad a_2 \ge 1 \quad\text{and}\quad b_3\equiv2\pmod3. $$ so that $b_3\ge2\implies$ $2^{2a_2-1} a_1^6\equiv-1\pmod9$ $\implies a_1^6 \equiv-2^{1-2a_2} \equiv 2^{-2(a_2+1)} \pmod 9$ since $2$ is a primitive root modulo $9$, i.e. $\text{ord}_9(2)=6=\phi(9)$, and $2^{\pm3}\equiv-1\pmod9$. But then $a_1\equiv2^{a_0}\pmod9$ for some $a_0$ with $-2(a_2+1)\equiv6a_0\equiv0\pmod6$, so that $a_2+1\equiv3~\implies~a_2\equiv2\pmod6$. In particular, we see that $a_2\ge2$.

Well anyway, that's a start. If we could show that $a_2,b_3\le2$, then we could reduce case 1 to $9y^3-8x^6=1$, and our goal would be to show that its only positive solution is $(1,1)$, and a start at that might be to note that $9y^3\equiv8x^6+1\equiv9\pmod{72}$.


How about $$\left(-\frac{1}{2}\right)^3 + \left(\frac{1}{2}\right)^3 + \left(\frac{3}{2}\right)^3 = \left(\frac{3}{2}\right)^3 ?$$ After all, the OP didn't specify where $a$ lives... (by the way, there are infinitely many distinct rational solutions of this form!).

Now for a more enlightened answer: no, there are no other integral solutions with $a\in\mathbb{Z}$, other than $a=0$ and $a=4$. Here is why (what follows is a sketch of the argument, several details would be too lengthy to write fully).

Suppose $(a-1)^3+a^3+(a+1)^3=b^3$. Then $3a^3+6a=b^3$. Hence $(a:b:1)$ is a point on the elliptic curve $E:3Z^3+6ZY^2=X^3$ with origin at $(0:1:0)$. In particular, a theorem of Siegel tells us that there are at most finitely many integral solutions of $3a^3+6a=b^3$ with $a,b\in\mathbb{Z}$. Now the hard part is to prove that there are exactly $2$ integral solutions.

With a change of variables $U=X/Z$ and $V=Y/Z$ followed by a change $U=x/6$ and $V=y/36$, we can look instead at the curve $E':y^2=x^3-648$. This curve has a trivial torsion subgroup and rank $2$, with generators $(18,72)$ and $(9,9)$. Moreover each point $(x,y)$ in $E'$ corresponds to a (projective) point $(x/6:y/36:1)$ on $E$, and a point $(X:Y:Z)$ on $E$ corresponds to a solution $a=Z/Y$ and $b=X/Y$. This means that $E$ is generated by $P_1=(3:2:1)$ and $P_2=(18:3:12)$ which correspond respectively to $(a,b)=(1/2,3/2)$ and $(4,6)$. The origin $(0:1:0)$ corresponds to $(a,b)=(0,0)$.

Now it is a matter of looking through all $\mathbb{Z}$-linear combinations of $P_1$ and $P_2$ to see if any gives another $(a,b)$ integral. However, this is a finite search, because of the way heights of points work, and one can calculate a bound on the height for a point $(a,b)$ to have both integral coordinates. Once this bound is found, a search among a few small linear combinations of $P_1$ and $P_2$ shows that $(0,0)$ and $(4,6)$ are actually the only two possible integral solutions.

Here is another rational solution, not as trivial as the first one I offered, that appears from $P_1-P_2$: $$\left(-\frac{10}{11}\right)^3 + \left(\frac{1}{11}\right)^3 + \left(\frac{12}{11}\right)^3 = \left(\frac{9}{11}\right)^3 $$

  • $\begingroup$ Shouldn't it be $6a$ instead of $6a^2$? $\endgroup$ – joriki Mar 15 '12 at 16:01
  • 2
    $\begingroup$ About that last solution: That's nice; it contains the cubes for the Hardy–Ramanujan number :-) $\endgroup$ – joriki Mar 15 '12 at 16:04
  • $\begingroup$ Thanks! Fixed it. Luckily, the projective equation $3Z^3+6ZY^2=X^3$ was correctly written, so the rest is OK. $\endgroup$ – Álvaro Lozano-Robledo Mar 15 '12 at 16:04

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.