25
$\begingroup$

From time to time, I run into the finite product $\prod_{j=1}^n(1+q^j)$. And, the more it happens, the more fascinated I've become. So, herein, I wish to get help in collecting such results. To give some perspective into what I look for, check out the below examples. First, some nomenclature: $(q)_k=(1-q)(1-q^2)\cdots(1-q^k)$ and $\binom{n}k_q=\frac{(q)_n}{(q)_k(q)_{n-k}}$.

(0) It's almost silly, but the sum of elementary functions of the specialization $\pmb{q}=(q,q^2,\dots,q^n)$: $$e_0(\pmb{q})+e_1(\pmb{q})+\cdots+e_n(\pmb{q})=\prod_{j=1}^n(1+q^j).$$

(1) The classical $q$-binomial theorem, which results from counting restricted distinct partitions or number of weighted tilings: $$\sum_{k=0}^nq^{\binom{k+1}2}\binom{n}k_q=\prod_{j=1}^n(1+q^j).$$

(2) I can't remember where I saw this (do you?) but $$\sum_{k=0}^nq^k\binom{n}k_{q^2}=\prod_{j=1}^n(1+q^j).$$

(3) The $H$-polynomial of a symplectic monoid $MSp_n$ (see this paper, page 13): $$\sum_{k=0}^n(-1)^kq^{k^2}\binom{n}k_{q^2}^2\prod_{i=1}^k(1-q^{2i})\prod_{j=1}^{n-k}(1+q^j)^2=\prod_{j=1}^{2n}(1+q^j),$$ although the authors did not seem to be aware of the RHS.

QUESTION. Can you provide such formulas (in any field) with the same RHS (always a finite product) as in above, together with resources or references?

Thank you.

$\endgroup$
6
  • 6
    $\begingroup$ Big list question? Traditionally, such are made Community Wiki. $\endgroup$ Commented Dec 16, 2016 at 18:27
  • 3
    $\begingroup$ perhaps we could wait and see if this does actually generate a "big list" ... $\endgroup$ Commented Dec 16, 2016 at 18:28
  • $\begingroup$ We don't know if this will be "big". I'm with Carlo, if you don't mind Todd. $\endgroup$ Commented Dec 16, 2016 at 18:37
  • $\begingroup$ There are several questions about this or close topics on MO, under the tag "q-analogs": mathoverflow.net/questions/tagged/q-analogs $\endgroup$ Commented Dec 16, 2016 at 19:41
  • $\begingroup$ @PietroMajer: Thanks. Sure, there are millions of $q$-series, but I'm asking of a very specific target, preferable with interpretations as seen in my examples. None I checked under the tag reveals answers to our question here. $\endgroup$ Commented Dec 16, 2016 at 19:47

3 Answers 3

22
$\begingroup$

Up to scaling, $\prod_{j=1}^n(1+q^j)$ is the character of the principal specialization of the spinor representation of $\mathfrak{so}(2n+1)$. This was first explicitly stated by J. W. B. Hughes, Lie algebraic proofs of some theorems on partitions, in Number Theory and Algebra, Academic Press, 1977, pp. 135--155.

$\endgroup$
6
  • $\begingroup$ Cool, up-voted. $\endgroup$ Commented Dec 17, 2016 at 2:08
  • 2
    $\begingroup$ What is this parasite on the left of the formula $\mathfrak{so}(2n+1)$ ? $\endgroup$ Commented Dec 17, 2016 at 6:11
  • 1
    $\begingroup$ It is the orthogonal Lie algebra in $2n+1$ dimensions. $\endgroup$ Commented Dec 17, 2016 at 7:05
  • 1
    $\begingroup$ In Richard's expository paper on log-concave and unimodal sequences (Ann. New York Acad. Sci., 1989), he mentions that by considering the principal $\mathfrak{sl}(2)$ inside $\mathfrak{so}(2n+1)$ we can prove the unimodality of the coefficients of $\prod_{j=1}^n (1+q^j)$. See Equation (23). It seems to be very difficult to prove this unimodality result otherwise. $\endgroup$ Commented Dec 17, 2016 at 17:13
  • $\begingroup$ Proctor (Amer. Math. Monthly 89 (1982), 721-734) did give a more elementary proof of unimodality but his argument was "secretly" motivated by the Lie-algebraic proof. He phrased his result in terms of the rank-unimodality of a certain poset $M(n)$ whose rank generating function is $\prod_{j=1}^n (1+q^j)$. $\endgroup$ Commented Dec 17, 2016 at 17:47
8
$\begingroup$

In a $2n$ dimensional vector space equipped with a symplectic form, you can look at all the Lagrangian subspaces, and this forms a subvariety of the Grassmannian $G(n,2n)$ called the Lagrangian Grassmannian, $\Lambda(n)$. It is a smooth projective variety which inherits a schubert cell decomposition from the Grassmannian, and lots of other Schubert calculus goodness. It's cells are indexed by strict partitions with largest part at most $n$. The number of points of $\Lambda(n)$ over finite fields is given by $\prod_{i=1}^n(1+q^i)$, so unimodality of the coefficients can be obtained from the Hard Lefschetz theorem. (Not like this is a simpler way of looking at it, but perhaps a geometric point of view can be useful for certain applications, for example the classical q-binomial theorem follows from a certain stratification of $\Lambda(n)$.)

A reference for an explicit presentation of the cohomology ring using Schur Q-polynomials can be found in "Algebro—Geometric applications of Schur S- and Q-polynomials" by Piotr Pragacz, Topics in Invariant Theory ,M. P. Malliavin Ed., Springer Lecture Notes in Math., 1478, 1991.

$\endgroup$
0
2
$\begingroup$

Let $exp(z,q)=\sum_{k=0}^{\infty}z^k/[k]_q!$ be the usual $q-$analogue of the exponential function $e^x.$
The identity $\sum_{k=0}^n{q^k}{\binom{n}{k}_{q^2}}=\prod_{j=1}^n(1+q^j)$ can be obtained by comparing coefficients in $$exp(\frac{z}{1+q},q^2)exp(\frac{qz}{1+q},q^2)=exp(z,q),$$ which is a natural $q-$analogue of $e^\frac{x}{2} e^\frac{x}{2}=e^x.$ This identity occurs in Séminaire Lotharingien de Combinatoire, B05a (1981), but is perhaps older.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .