1
$\begingroup$

This should probably be not that hard, but I would like to see a nifty way of proving it.

Consider the double-indexed sequence given by $$f(n,k)=\binom{2n + 2k}{n + k}\binom{n + k}{n - k}3^k.$$

QUESTION. For $1\leq k\leq n$, does this hold true for the $3$-adic valuations? $$\nu_3(f(n,k))>\nu_3(f(n,0)).$$

$\endgroup$
2
  • $\begingroup$ In general, Kummer's theorem is a useful tool for these sorts of calculations: en.wikipedia.org/wiki/Kummer%27s_theorem $\endgroup$
    – Terry Tao
    Jun 26, 2021 at 15:38
  • $\begingroup$ This is true, although in the present case it is a bit subtle to decide $2\nu_3(f(n,k))=k+s_3(n+k)+s_3(n-k)+s_3(2k)-s_3(2n+2k)>2s_3(n)-s_3(2n)$ unless one proceeds with a case-by-case analysis. Tedious? $\endgroup$ Jun 26, 2021 at 16:14

1 Answer 1

5
$\begingroup$

Notice that $$ 3^{-k-1}f(n,k+1)/3^{-k}f(n,k)=\frac{{2n+2k+2\choose n+k+1}}{{2n+2k\choose n+k}}\frac{{n+k+1\choose n-k-1}}{{n+k \choose n-k}}. $$ The first factor here equals $$ \frac{(2n+2k+2)!}{(2n+2k)!}\frac{(n+k)!^2}{(n+k+1)!^2}=\frac{2(2n+2k+1)}{n+k+1}. $$ The second is $$ \frac{(n+k+1)!}{(n+k)!}\frac{(n-k)!}{(n-k-1)!}\frac{(2k)!}{(2k+2)!}=(n+k+1)(n-k)\frac{(2k)!}{(2k+2)!}. $$ For instance, $$ \frac{3^{-k-1}(2k+2)!f(n,k+1)}{3^{-k}(2k)!f(n,k)}=2(n-k)(2n+2k+1) $$ is an integer. Multiplying several fractions of this type together, we see that for any $k\leq n$ we have $$ \frac{3^{-k}(2k)!f(n,k)}{3^{-0}0!f(n,0)}\in\mathbb Z. $$ Hence $$ \nu_3(3^{-k}(2k)!f(n,k))\geq \nu_3(f(n,0)). $$ Next, for $k>0$ $$ \nu_3((2k)!)=\left[\frac{2k}{3}\right]+\left[\frac{2k}{9}\right]+\ldots< $$ $$ <\frac{2k}{3}+\frac{2k}{9}+\ldots=k, $$ which means that $$ \nu_3(f(n,k))-\nu_3(f(n,0))\geq k-\nu_3((2k)!)>0, $$ as needed.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.