7
$\begingroup$

Let $(q;q)_n=(1-q)(1-q^2)\cdots(1-q^n)$ with $(q;q)_0:=1$. Define a $q$-exponential by $$e(z;q)=\sum_{n\geq0}\frac{z^n}{(q;q)_n}.$$ There is a notion of $q$-Eulerian polynomials, see the reference. I like to introduce $q$-Eulerian polynomial of type B via the generating function $$\sum_{n\geq1}B_n(t,q)\frac{z^n}{(q;q)_n}=\frac{(e(z;q)-e(tz;q))(e(tz;q)+te(z;q))}{e(2tz;q)-te(2z;q)}.$$ Now, expand $B_n(t,q)$ as a polynomial $$B_n(t,q)=\sum_{k=0}^nB_{n,k}(q)t^k$$ and call $B_{n,k}(q)$ $q$-Eulerian numbers type B. Here are the first few terms: \begin{align} B_1(t,q)&=1+t, \\ B_2(t,q)&=1+(2q+4)t+t^2, \\ B_3(t,q)&=1+(7q^2+7q+9)t+(7q^2+7q+9)t^2+t^3. \end{align}

Claim. if $a, b\in\Bbb{N}$ and $\alpha=a+b+1$, then the symmetric relation holds: $$\binom{\alpha}a_q+\sum_k\binom{\alpha}k_q2^{\alpha-k}B_{k,b}(q)= \binom{\alpha}b_q +\sum_k\binom{\alpha}k_q2^{\alpha-k}B_{k,a}(q).$$

QUESTIONS:

(a) I don't have a proof for my claim which seems very true though. Do you?

(b) Is there a combinatorial interpretation for these polynomials $B_n(t,q)$ or the Eulerian numbers $B_{n,k}(q)$? You might be inspired by the reference.

$\endgroup$
  • $\begingroup$ I proved (a), but I don't know anything about (b). $\endgroup$ – user82588 Oct 24 '16 at 15:26
  • $\begingroup$ @Nemo: do you like to share your proof? $\endgroup$ – T. Amdeberhan Oct 25 '16 at 11:56
  • $\begingroup$ Why type B? Is it somehow related to the ABCDEF nomenclature? $\endgroup$ – მამუკა ჯიბლაძე Oct 25 '16 at 13:59
  • $\begingroup$ Type B because there are types A & B in the classical literature ($q=1$). $\endgroup$ – T. Amdeberhan Oct 25 '16 at 15:38
  • $\begingroup$ Could you please provide a link? $\endgroup$ – მამუკა ჯიბლაძე Oct 26 '16 at 7:57
3
$\begingroup$

$\bf{Step~1}.$ $B_{n,a}(q)=B_{n,n-a}(q)$.

$\it{Proof}$. Write $$ \sum_{n\geq1}\dfrac{B_n(t,q)}{t^{n/2}}\frac{z^n}{(q;q)_n}=\frac{e(z/\sqrt{t};q)-e(z\sqrt{t};q)}{\dfrac{e(2z\sqrt{t};q)}{\sqrt{t}}-\sqrt{t}~e(2z/\sqrt{t};q)}\left(\dfrac{e(z\sqrt{t};q)}{\sqrt{t}}+\sqrt{t}~e(z/\sqrt{t};q)\right). $$ The rhs is symmetric under $t\to 1/t$, so $\dfrac{B_n(t,q)}{t^{n/2}}$ is also symmetric: $$ \dfrac{B_n(t,q)}{t^{n/2}}=t^{n/2}{B_n(1/t,q)}.\tag{1} $$ We know that $B_n(t,q)=B_{n,0}t^0+...$ starts with the power $t^0$. Due to $(1)$ this means that the highest power of $t$ in the polynomial $B_n(t,q)$ is $n$, and $B_{n,a}(q)=B_{n,n-a}(q)$ as required.

$\bf{Step~2}.$ $\displaystyle{\binom{\alpha}a_q+\sum_k\binom{\alpha}k_q2^{\alpha-k}B_{k,b}(q)= \binom{\alpha}b_q +\sum_k\binom{\alpha}k_q2^{\alpha-k}B_{k,a}(q)},~ a,b,\in \mathbb{N},~a+b+1=\alpha.$

$\it{Proof}$. According to q-binomial theorem $e(z;q)=\sum_{n\geq0}\frac{z^n}{(q;q)_n}=\frac{1}{(z;q)_\infty}$. Now we equate the coefficients of $z^n$ in both sides of \begin{align} \left(\sum_{l\geq0}\frac{(2tz)^l}{(q;q)_l}-t\sum_{l\geq0}\frac{(2z)^l}{(q;q)_l}\right)\cdot\sum_{k\geq1}B_k(t,q)\frac{z^k}{(q;q)_k}={(e(z;q)-e(tz;q))(e(tz;q)+te(z;q))}\\ =\left(\sum_{l\geq0}\frac{z^l}{(q;q)_l}-\sum_{l\geq0}\frac{(tz)^l}{(q;q)_l}\right)\left(\sum_{l\geq0}\frac{(tz)^l}{(q;q)_l}+t\sum_{l\geq0}\frac{z^l}{(q;q)_l}\right) \end{align} and obtain $$ \sum _{k=1}^n \binom{n}{k}_q\frac{B_k(t,q) \left(t^{n-k}-t\right) }{2^{k-n}}-\sum _{k=1}^n \left(1-t^k\right) \left(t^{n-k}+t\right) \binom{n}{k}_q=0. $$ The coefficient of $t^{a+1}$ in the above formula is (note that for $a\in \mathbb{N}$ we have $a+1\neq1$ and for $b=n-a-1\in \mathbb{N}$ we have $a+1\neq n$, so the 'boundary' terms with $t$ and $t^n$ don't make any contribution in the formula below) $$ -\sum _{k=1}^n\binom{n}{k}_q \frac{B_{k,a}(q) }{2^{k-n}}+\sum _{k=1}^n \binom{n}{k}_q\frac{B_{k,a+1+k-n}(q)}{2^{k-n}}+\binom{n}{a}_q-\binom{n}{b}_q=0. $$ Finally, taking into account $B_{k,a+1+k-n}(q)=B_{k,k-b}(q)=B_{k,b}(q)$ we have $$ \binom{n}{b}_q+\sum _{k=1}^n\binom{n}{k}_q \frac{B_{k,a}(q) }{2^{k-n}}=\binom{n}{a}_q+\sum _{k=1}^n \binom{n}{k}_q\frac{B_{k,b}(q)}{2^{k-n}}. $$

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Step 1 was proved in arxiv.org/pdf/1207.4045.pdf as well. I'll look into the rest. $\endgroup$ – T. Amdeberhan Oct 25 '16 at 15:42
  • $\begingroup$ @T.Amdeberhan , what logic did lead you to this conjecture? $\endgroup$ – user82588 Oct 25 '16 at 15:52
  • $\begingroup$ I tried generalizing classical symmetrical relations. There's already one such q-analogue for type A by Guo-Niu Han. $\endgroup$ – T. Amdeberhan Oct 25 '16 at 16:33
  • $\begingroup$ I checked the proof. Correct. $\endgroup$ – T. Amdeberhan Nov 2 '16 at 0:44

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.