# What is the value of this sum involving q-binomials?

Let $n\ge 2r$ be positive integers.
Is there a closed form for following finite summation involving in q-binomial coefficients $$\sum_{s=0}^r(-1)^sq^{\frac{s(s+1)}{2}}{n-2r+s\brack n-2r}_q{n\brack r-s}_q\,\,\,\,\,\, ?$$ I found this while studying q-Fibonacci/ Lucas polynomials.
What is the general approach for evaluating this type of series?
Any suggestion would be appreciated.

• What is your definition of $\big[..\big]_q$ ? – Konstantinos Kanakoglou Jan 23 '17 at 1:10
• I use the definition $[n]_q=1+q+q^2+\cdots+q^{n-1}$ for q-integers . – Bumblebee Jan 23 '17 at 1:15
• Out of curiosity: where did you find this sum? I mean did you find it mentioned in some paper or is it a byproduct of some calculations of yours? It would be helpful to add some more context. – Konstantinos Kanakoglou Jan 23 '17 at 2:04
• I an also very curious about the sum - it looks like the major index statistic on some family of lattice paths with some restriction (using inclusion-exclusion)... – Per Alexandersson Jan 26 at 22:33

Doron Zeilberger has written a Maple code for checking and proving ordinary binomial identities and their $q$-analogues. What you need in the present case is the package called qEKHAD.
I just tested your sum and it leads to a quadratic recurrence in $r$, so you may not expect a closed form as an answer. The lucky cases are linear recurrences in the "external" parameter (here, it is $r$). The latter situation fits the so-called Wilf-Zeilberger pair.
I also sought for other factors $q^{\mu(s)}$ instead of $q^{\binom{s+1}2}$, none of the usual suspects lead to a closed form. Sorry.
• Thank you for your help. I am tring to simplify this summation bit further. Numerical values suggest that above sum equal to $\sum_{s=0}^{r-1}(-1)^{s+1}q^{\frac{s(s+1)}{2}}{n-2r+s\brack n-2r}_q{n\brack r-s-1}_q.$ And also I fund that this equality holds for ordinary binomial coefficients (in which $q=1$). Do you think that they are actually equals for q-binomials? – Bumblebee Jan 23 '17 at 3:10
• I just checked: the two sums satisfy the same quadratic recurrence (in $n$) and matching initial conditions. They are indeed equal. – T. Amdeberhan Jan 23 '17 at 3:56