5
$\begingroup$

Does the system of congruence equations \begin{eqnarray} A_{17k}&\equiv& 0 \pmod {17^2}, \nonumber \\ A_{17k+1}&\equiv& 0 \pmod {17^2}, \tag{1} \end{eqnarray} has solutions other than $k=3$? Here $A_n$ are Apery numbers: $$A_n=\sum\limits_{k=0}^n\binom{n}{k}^2\binom{n+k}{k}^2.$$ Thanks to the recurrence relation $$n^3 A_n = (34n^3-51n^2+27n-5)A_{n-1}- (n-1)^3 A_{n-2},$$ if $k$ is a solution of (1), then all seventeen consecutive numbers beginning from $A_{17k}$ are congruent to zero modulo $17^2$. In particular, $$A_{51},A_{52},A_{53},\ldots,A_{66},A_{67}\equiv 0 \pmod {17^2}.$$

$\endgroup$
5
$\begingroup$

Results of Gessel can be used to solve this question. Gessel showed, in Theorem 1 here, that for any prime $p$, if $n=\sum d_i p^i$ is the base-$p$ expansion of $n$, then $$(*) A_n \equiv \prod A_{d_i} \bmod p.$$ In Theorem 4 here he essentially refined this as follows. He proved that there is an (explicit) sequence $\{b_n\}_{n \ge 0}$ such that $$(**) A_{pn+k} \equiv A_{n} (A_{k} + pn b_{k}) \bmod {p^2},$$ where $p$ is a prime, $n \ge 0$ and $0 \le k < p$. The first values of $b_n$ are given by $b_0=0,b_1=12$.


Specializing $(*)$ to $p=17$, we see that $A_n$ is divisible by $17$ iff one of the digits $d_i$ in the base-17 expansion of $n$ satisfies $A_{d_i} \equiv 0 \bmod 17$. A finite check shows that the solutions to $A_d \equiv 0 \bmod 17$ in the range $0 \le d < 17$ are $d=3,13$.

Specializing $(**)$ to $p=17$ and $k \in \{0,1\}$, we find that $$A_{17n} \equiv A_{n} \bmod {17^2},$$ $$A_{17n+1} \equiv A_{n} (A_1 + 17\cdot 12\cdot n) \bmod {17^2}.$$ This immediately shows that the first equation in your system implies the second one, and that your system is equivalent to $$(***)A_n \equiv 0 \bmod {17^2}.$$ We now proceed to solve $(***)$. Let $S$ be the set of solutions to $(***)$. Writing $n$ as $17m+m'$ ($0 \le m' < 17$), we find that $(***)$ is equivalent to $$A_{m} (A_{m'}+17mb_{m'}) \equiv 0 \bmod {17^2}.$$ There are three possible cases.

  1. $A_m \equiv 0 \bmod {17^2}$.
  2. $A_m \equiv 0 \bmod {17}$, $A_{m'}+17mb_{m'} \equiv 0 \bmod 17$.
  3. $A_{m'} + 17mb_{m'} \equiv 0 \bmod {17^2}.$

The first case says: if $n \in S$ then $17n+n' \in S$ ($0 \le n' < 17$).

The second case says: $m$ has $3$ or $13$ as one of its base-17 digits, and $m' \in \{3,13\}$.

The third case says: Either $m'=3$ and $m \equiv 0 \bmod 17$, or $m'=13$ and $m \equiv -1 \bmod 17$.

(Here I've used the values $A_3=1445=17^2\cdot 5$, $A_{13} = 364713572395983725$, $b_{3}=4438$, $b_{13}=7475161875743183448469/6006$.)

This characterizes the set $S$.

$\endgroup$
2
$\begingroup$

According to a computation there are other solutions.

Solutions $k$ start: $3 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 , 63 , 64 , 65 , 66 , 67,????, 51\cdot 17, 52 \cdot 17$


Jeremy Rouse suggests missing $k=224, k=234, k=285, k=292, k=343$ and I didn't search the range $????$.

$\endgroup$
  • 1
    $\begingroup$ This is $3$, then $3\cdot 17,\ldots, 3\cdot 17+ 16$, then $3\cdot 17^2$,\ldots$, which might be suggestive. $\endgroup$ – Jeff Strom May 30 '14 at 11:26
  • $\begingroup$ @JeffStrom indeed. $\endgroup$ – joro May 30 '14 at 11:30
  • $\begingroup$ That is it can be conjected that $k=3*17^m$ are all solutions? Or only for prime m? $\endgroup$ – Zurab Silagadze May 30 '14 at 11:45
  • $\begingroup$ @ZurabSilagadze maybe for all $m$, don't know. Don't have efficient computation of binomial coefficients modulo prime powers so far. $\endgroup$ – joro May 30 '14 at 12:03
  • 1
    $\begingroup$ @joro - You're missing some. For example, $k = 224$, $k = 234$, $k = 285$, $k = 292$, $k = 343$, etc. $\endgroup$ – Jeremy Rouse May 30 '14 at 12:52

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.