In trying to solve $a^3 - 2b^3 = 1$ over the integers I came across the need to answer the question: when does $(1+ \sqrt[3]{2} + \sqrt[3]{2}^2)^n$ have no $\sqrt[3]{2}^2$ term in it's expansion (in terms of the basis $\{1,\sqrt[3]{2}, \sqrt[3]{2}^2\}$ of $\mathbb{Z}[\sqrt[3]{2}]$).

Clearly this cannot happen for positive $n$ (create a set of recursions that generate the next coeffs from the previous ones and you can observe all three coeffs are always positive). Of course for $n=0$ it does happen and you get the trivial solution $(a,b)=(1,0)$.

I am struggling to sort out the negative $n$ case. This is the same as studying positive integer powers of $(\sqrt[3]{2} - 1)$.

I guess that for positive $m$ you only ever get a zero coefficient of $\sqrt[3]{2}^2$ in $(\sqrt[3]{2} - 1)^m$ whenever $m=1$ (giving another solution $(a,b)=(-1,-1)$). However I am struggling to prove this using the recursions alone.

Have I missed something easy?

  • $\begingroup$ Has anyone managed to exploit the recurrences yet? $\endgroup$ – fretty Dec 30 '13 at 13:26
  • 1
    $\begingroup$ $a^3-2b^3=1$ is an example of what's called a Thue equation. I'm not sure it can be solved along the elementary lines you are pursuing --- it may need something a bit more advanced. Anyway, I have given you a search term. $\endgroup$ – Gerry Myerson Dec 31 '13 at 17:42
  • 1
    $\begingroup$ By the way there are no other integer solutions: $a^3-2b^3=1$ is equivalent to an elliptic curve $E$. $E$ has rank $0$ and the torsion is $\cong \Bbb Z_2$, so you get only 1 point out of it. The torsion point corresponds to $(-1,-1)$. The infinity point $\mathcal O$ corresponds to the case when $a=1$, but in that case $b=0$, so you get $(1,0)$ and there are no other points. So if a solution corresponds to no $\sqrt 8$ in expansion then it must be true that $\sqrt 8$ is never $0$. $\endgroup$ – Yong Hao Ng Dec 31 '13 at 18:31
  • $\begingroup$ Thanks. I know how to solve the Diophantine using these methods but was really searching for a non-trivial use of Dirichlet's unit theorem (for a possible talk). Now I must confess this strategy was mentioned in an answer to a question on this site somewhere (and so I assumed it was doable this way) but cannot for the life of me find the question. $\endgroup$ – fretty Jan 1 '14 at 8:55
  • 1
    $\begingroup$ Maybe math.stackexchange.com/questions/61014/solve-x3-1-2y3 $\endgroup$ – Gerry Myerson Jan 2 '14 at 3:48

This is a classical problem first solved by Nagell (Solution complète de quelques équations cubiques à deux indéterminées, J. Math. Pures Appl. 4 (1925), 209–270); for an English version see LeVesque's Topics in Number Theory, vol. II, and I have given a brief account in German here (see the appendix).

The question whether elements in a pure cubic field can have squares in which a coefficient with respect to the basis $\{1, \sqrt[3]{m}, \sqrt[3]{m}^2\}$ is $0$ is related to elliptic curves; see this article.


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.