6
$\begingroup$

I have to validate the following identity which is defined: $$ \sum_{k=1}^n (-1)^{k-1} \cdot q^{\frac{k(k-1)}{2}} \cdot \frac{\prod_{i=n-k+1}^n(1-q^i)}{\prod_{i=1}^k(1-q^i)} = 1 $$ where $0<q<1$.

I verified the expression numerically with R/Matlab by choosing different values for $q$ and so far it is correct. I also verified trivial cases where $n=2,3,4$ and it is also correct. But I couldn't figure out how one can prove it for all of $n$.

I hope someone could give me a hint to proceed or share some experiences in proving this kind of identities. Thanks in advance!

Edit: I thought giving a context might be interesting. The original identity that I am working is: $$ \sum_{k=1}^n \frac{1}{(1-q^k)(q^{-1},q^{-1})_{k-1}(q,q)_{n-k}} = \frac{1}{(q^{-1},q^{-1})_0 (q,q)_n} = \frac{1}{(q,q)_n} $$ where $(q,q)_n = \prod_{i=1}^n (1-q^i) $ is the q_Pochhamer symbol with $(q,q)_0 = 1$. These two identities are equivalent, I need to prove at least one :P

Update: I have demonstrated the following identity, in case of somebody is interested: $$ \sum_{k=1}^n \frac{1}{(q^{-1},q^{-1})_{k-1}(q,q)_{n-k}} = 1 $$

Even though this and the original are surprisingly close to each other, I couldn't find away to demonstrate the first identity.

$\endgroup$
  • $\begingroup$ By induction maybe? $\endgroup$ – Avitus Dec 11 '13 at 14:12
  • $\begingroup$ I did try induction but it didn't lead to anywhere. As I see, the $n$ step doesn't bring any useful information for the $n+1$ step. Nevertheless, I keep digging the induction way 'cause it's the only 'obvious' method for the time being. Thanks anyway! $\endgroup$ – Tuan Dec 11 '13 at 14:35
5
$\begingroup$

You're asking why $$ \sum(-1)^kq^{k(k-1)/2}\genfrac[]{0pt}{}nk=0. $$ That's a $q$-deformed version of the classical identity $\sum(-1)^k\binom nk=0$ (which corresponds to $q=1$). At least some proofs of the classical identity can be generalized to the «quantum» case. For instance, there is (as I've mentioned in comments) $q$-binomial theorem (that can be prove either purely combinatorially or by induction), $$ (1+x)(1+xq)(1+xq^2)\ldots(1+xq^{n-1})=\sum x^kq^{k(k-1)/2}\genfrac[]{0pt}{}nk $$ — now plug $x=-1$.

$\endgroup$
  • $\begingroup$ Thank you very much! It is even proven in 2 lines! I've learned new notion about the q-binomial coefficients thanks to you! $\endgroup$ – Tuan Jan 14 '15 at 14:25

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.