11
$\begingroup$

It is known that the Somos-$k$ sequences for $k\ge 8$ do not give integers. But the first terms of Somos-8 sequence $s_n=a_n/b_n$ $$1, 1, 1, 1, 1, 1, 1, 1, 4, 7, 13, 25, 61, 187, 775, 5827, 14815,\frac{420514}{7}, \frac{28670773}{91}$$ defined by $s_1=s_2=s_3=s_4=s_5=s_6=s_7=s_8=1$, $$s_{n+8}s_n=s_{n+7}s_{n+1}+s_{n+6}s_{n+2}+s_{n+5}s_{n+3}+s_{n+4}^2\qquad(n\ge1)$$ have only odd denominators $b_n$. Morever $s_n$ has even numerator $a_n$ only for $n=9k$. It was checked for $n\le 67$. First terms of $s_n\mod 8:=a_n\cdot b_n^{-1}\mod 8$ are $$\begin{array}{l} 1, 1, 1, 1, 1, 1, 1, 1, {\bf 4},\\ 7, 5, 1, 5, 3, 7, 3, 7, {\bf 6},\\ 7, 7, 5, 7, 5, 3, 1, 1, {\bf 6},\\ 5, 1, 5, 5, 3, 3, 7, 1, {\bf 4},\\ 7, 7, 3, 3, 7, 7, 3, 1, {\bf 2},\\ 1, 3, 1, 3, 7, 5, 3, 5, {\bf 2},\\ 5, 7, 3, 7, 7, 3, 7, 3, {\bf 0},\\ 1, 3, 7, 3,\ldots\end{array}$$

Is it possible to prove that the sequence $s_n\mod 2$ is periodic?

EDT. It was found by მამუკა ჯიბლაძე that numerator of $s_{71}$ is even while that of $s_{72}$ is odd. The last line of the table above is $$1, 3, 7, 3, 1, 5, 1, {\bf 2}, 5$$ So the conjecture about numerators of $s_n$ is wrong. But the qustion about periodicity is still valid.

$\endgroup$
  • $\begingroup$ There might be something in Robinson, Periodicity of Somos sequences, Proc Amer Math Soc 116 (1992) no. 3, 613-619, MR1140672 (93a:11012). The review by Peter Kiss says, in part, "For some sequences related to the Somos($k$) sequences, the author shows that they are periodic modulo $m$ for every $m$, although the periodicity problem remains open for the original sequences. Some observations are also made concerning the prime divisors of the terms and the rate of growth of certain sequences. $\endgroup$ – Gerry Myerson Jun 9 '17 at 6:54
  • 2
    $\begingroup$ Believe it or not but actually the numerator of $s_{71}$ is even while that of $s_{72}$ is odd (unless there is a bug in Mathematica 11, that is). $\endgroup$ – მამუკა ჯიბლაძე Jun 9 '17 at 7:51
  • 1
    $\begingroup$ Doing some further calculations with 2-adics; although I am not sure I am doing it right but seems like $s_{95}$ has even denominator... $\endgroup$ – მამუკა ჯიბლაძე Jun 9 '17 at 8:51
  • 2
    $\begingroup$ Still have to re-check it but I switched to Pari and computed in 2-adics (with $2^{100000}$ precision), seemingly $s_{103}$ has even denominator. Up to that, results with Mathematica and Pari are identical, and also coincide with results for exact rationals (these I am currently running, they are at 82 so far). $\endgroup$ – მამუკა ჯიბლაძე Jun 9 '17 at 11:17
  • 1
    $\begingroup$ Alas, run out of memory on 85 with exact rationals. There might be other ways to find out about parity of the denominator in $s_{103}$ though... $\endgroup$ – მამუკა ჯიბლაძე Jun 10 '17 at 9:31
8
$\begingroup$

I confirm the observation of @მამუკაჯიბლაძე that $\nu_2(s_{103})=-1$. That is, $s_n\bmod 2$ is not well-defined at first place, which invalidates the question.

Moreover, for $n\geq 133$, $\nu_2(s_n)$ seems to form a strictly decreasing function, i.e., $s_n$ accumulates larger and larger powers of $2$ in the denominators.

| cite | improve this answer | |
$\endgroup$

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.