Let us use the standard notation for $q$-integers, $q$-binomials, and the $q$-analog $$ \operatorname{Cat}_q(n) := \frac{1}{[n+1]_q} \left[\matrix{2n \\ n}\right]_q. $$ I want to prove that for all integers $n\geq 0$, we have \begin{equation} \operatorname{Cat}_q(n+2) = \sum_{0\leq j,k \leq n} q^{k(k+2) + j(n+2)} \left[\matrix{n \\ 2k}\right]_q \operatorname{Cat}_q(k) \frac{[n+4]_q}{[k+2]_q} \left[\matrix{n-2k \\ j}\right]_q. \end{equation}

I have tried quite a bit, but not succeeded. Using $q$-hypergeometric series, this is equivalent with proving $$ \sum_{\substack{k\geq 0 \\ j \geq 0}} q^{k(k+2)+j(n+2)} \frac{ (q;q)_{n+4} }{ (q^{n+3};q)_{n+2} (q;q)_{j} } \frac{ (q^{n-2k+1};q)_{2k} (q^{n-2k-j+1};q)_{j} }{ (q;q)_{k} (q;q)_{k+2} } =1 $$ which I have also not managed to prove. I believe that some WZ-method could solve this easily, but a human-friendly proof would be preferrable. Note that the identity above is very similar to a theorem by Andrews (see reference below). It states that \begin{equation} \operatorname{Cat}_q(n+1) = \sum_{k \geq 0} q^{k(k+2)} \left[\matrix{n \\ 2k}\right]_q \operatorname{Cat}_q(k) \frac{(-q^{k+2};q)_{n-k}}{(-q;q)_k}. \end{equation}

UPDATE: I have managed to find a more general conjecture, which would imply the one above. It states that for integer $n \geq 0$, and general $a,c$, we have $$ \sum_{s} \frac{ (-a q^n)^{s} q^{-\binom{s}{2}} (q^{-n};q)_{s} }{ (q;q)_{s} } {}_{2}\phi_{1}(cq^{s-1}/a,q^{-s};c;q,q) = \frac{ (ac ;q)_{n} }{ (c;q)_{n} }. $$

I use This book as my main reference for notation and identities.

Andrews, George E., (q)-Catalan identities, Alladi, Krishnaswami (ed.) et al., The legacy of Alladi Ramakrishnan in the mathematical sciences. New York, NY: Springer (ISBN 978-1-4419-6262-1/hbk; 978-1-4419-6263-8/ebook). 183-190 (2010). ZBL1322.11018.

  • 2
    $\begingroup$ Is there a human-friendly proof for the non-$q$ analog of this identity? That might be a start. $\endgroup$ Jun 6, 2020 at 8:26
  • $\begingroup$ Yes, there is a combinatorial interpretation, tusing triangulations. I'll get the reference later today. $\endgroup$ Jun 6, 2020 at 9:50
  • 2
    $\begingroup$ Here is the combinatorial interpretation Per is talking about: core.ac.uk/download/pdf/82091698.pdf $\endgroup$ Jun 6, 2020 at 14:29
  • 2
    $\begingroup$ If one takes $q=1$, makes the change of variables $n \to n-4$ and $k \to k-2$ and sum over all $j$, one obtains the formula in the article. The summands the number of triangulations of regular $(n+2)$-gons with $k$ ears (an ear is a triangle on the vertices $i$, $i+1$ and $i+2$, where the indices are taken modulo $n+2$. $\endgroup$ Jun 6, 2020 at 14:44
  • 1
    $\begingroup$ Is it possible that the triangulation interpretation of the $q=1$ identity is compatible with the statistic on triangulations discussed here?: mathoverflow.net/questions/93136/… $\endgroup$ Jun 6, 2020 at 20:21

1 Answer 1


I managed to solve the problem, in the last general conjecture, one can apply the $q$-Chu-Vandermonde theorem. After some simplification, the resulting expression can be expressed as a ${}_2\phi_1$ q-hypergeometric series, where one again can apply the $q$-Chu-Vandermonde theorem.

Skipping lots of details the proof still requires a few pages - it can now be found in this paper (Thm. 57)


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.