14
$\begingroup$

A plane partition is a subset of $\mathbb Z_{\geqslant0}^3$ s.t. if it contains $(i+1,j,k)$ or $(i,j+1,k)$ or $(i,j,k+1)$ it also contains $(i,j,k)$.

What is the generating function $R(q)$ of (volumes of) plane partitions not containing the cell $(1,1,1)$?

Since a plane partition containing $(1,1,1)$ contains a $2\times2\times2$ cube, mod $q^8$ this g.f. coincides with MacMahon’s $$ M(q)=\frac1{\prod(1-q^i)^i}=1+q+3q^2+6q^3+13q^4+24q^5+48q^6+86q^7+160q^8+282q^9+500q^{10}+\ldots $$ and computing a couple more terms $$ R(q)=1+q+3q^2+6q^3+13q^4+24q^5+48q^6+86q^7+159q^8+279q^9+488q^{10}+\ldots $$ OEIS doesn’t know this sequence, but in the form $$ R(q)=1+\frac{q-3q^3+6q^6-10q^{10}+\ldots}{[(1-q)(1-q^2)(1-q^3)\ldots]^3}. $$ a pattern is evident.

So actually the question is

How to prove that $R(q)=1+\frac1{(q)_\infty^3}\sum\limits_i(-1)^i\frac{i(i-1)}2q^{\tfrac{i(i-1)}2}$?

AFAIK the statement can be found somewhere in the literature on representation theory of quantum toroidal something. But surely there is a well-known combinatorial (in the broad sense, not necessarily bijective) proof?

Context 1. G.f. for plane partitions not containing $(1,1,0)$ is $1/(q)_\infty^2\cdot\sum(-1)^iq^{\frac{i(i+1)}2}$. See the subsection about «V-shaped partitions» in Stanley's Enumerative Combinatorics for a (simple) combinatorial proof. Perhaps there is something like that proof in «(1,1,1)-case».

Context 2. G.f. for plane partitions not containing $(1,1,1)$ but intersecting axes at $(i,0,0)$, $(0,j,0)$ and $(0,0,k)$ resp. is $\genfrac[]{0pt}{}{i+j}i\genfrac[]{0pt}{}{j+k}j\genfrac[]{0pt}{}{k+i}kq^{i+j+k+1}$. So maybe it’s possible to compute $R(q)$ as $1+\sum_n q^{n+1}\sum_{i+j+k=n}\genfrac[]{0pt}{}{i+j}i\genfrac[]{0pt}{}{j+k}j\genfrac[]{0pt}{}{k+j}k$. And for the internal sum I know the answer at least at $q=1$: https://math.stackexchange.com/q/177209/

$\endgroup$
4
  • $\begingroup$ Am I correct that you mean $R(q)=1+\sum_{n=0}^\infty q^{n+1}\sum_{i+j+k=n}\genfrac[]{0pt}{}{i+j}i\genfrac[]{0pt}{}{j+k}k\genfrac[]{0pt}{}{k+i}k$ $\endgroup$ – Hjalmar Rosengren Jan 9 '15 at 8:11
  • $\begingroup$ @Hjalmar Yes, of course, thank you (I'll fix the typo) $\endgroup$ – Grigory M Jan 9 '15 at 10:39
  • $\begingroup$ Regard a plane partition as a two-dimensional array $(\pi_{ij})_{i,j\geq 1}$ in the usual way. The present question asks to count plane partition $n$ with $\pi_{22}\leq 2$. If $v(n)$ denotes the number of plane partitions of $n$ with $\pi_{22}=0$, then by EC1, Prop. 2.5.1, $\sum_{n\geq 0} v(n)q^n = \left( \sum_{m\geq 0} (-1)^mq^{{m+1\choose 2}}\right)\prod_{i\geq 1}(1-q^i)^{-2}$. This suggests counting plane partitions of $n$ with $\pi_{22}\leq k$ for $k\geq 2$ (and similar problems). $\endgroup$ – Richard Stanley Sep 14 '20 at 18:13
  • 1
    $\begingroup$ @RichardStanley We also have some formulas for $\pi_{nm}=0$ in arXiv:1512.08779. In that case an answer is given by Brion's theorem. What really puzzles me is that the formula for $R(q)-1$ also looks like some version of Brion's theorem (each triple (x,y,z) in the answer below should correspond to a vertex etc)… $\endgroup$ – Grigory M Sep 14 '20 at 21:47
8
$\begingroup$

It is indeed possible to get the result from your "Context 2".

Correcting typos, we have $$R(q)=1+\sum_{i,j,k=0}^\infty\frac{(q)_{i+j}(q)_{i+k}(q)_{j+k}}{(q)_i^2(q)_j^2(q)_k^2}\,q^{i+j+k+1} $$ (where $(q)_k=\prod_{j=1}^k(1-q^j)$). By the $q$-binomial theorem, $$\frac{(q)_{i+j}}{(q)_i(q)_j}=\sum_{x=0}^i\frac{1}{(q)_x(q)_{i-x}}(-1)^xq^{\binom x2+(j+1)x}. $$ Using this, and the expansions obtained by cyclically permuting $(i,j,k)$, we get \begin{align*}R(q)-1&=\sum_{i,j,k=0}^\infty\sum_{x,y,z=0}^{i,j,k}\frac{(-1)^{x+y+z}q^{\binom x2+\binom y2+\binom z3+(j+1)x+(k+1)y+(i+1)z+i+j+k+1}}{(q)_x(q)_{i-x}(q)_y(q)_{j-y}(q)_z(q)_{k-z}}\\ &=\sum_{x,y,z=0}^\infty \frac{(-1)^{x+y+z}q^{\binom x2+\binom y2+\binom z2+(y+2)x+(z+2)y+(x+2)z+1}}{(q)_x(q)_y(q)_z}\\ &\quad\times\sum_{i=0}^\infty\frac{q^{(1+z)i}}{(q)_i}\sum_{j=0}^\infty\frac{q^{(1+x)j}}{(q)_j}\sum_{k=0}^\infty\frac{q^{(1+y)k}}{(q)_k}\\ &=\frac 1{(q)_\infty^3}\sum_{x,y,z=0}^\infty {(-1)^{x+y+z}q^{\binom x2+\binom y2+\binom z2+(y+2)x+(z+2)y+(x+2)z+1}} ,\end{align*} where we first replaced $(i,j,k)$ by $(i+x,j+y,k+z)$ and then computed the inner sums using again a special case of the $q$-binomial theorem. We now observe that $$\binom x2+\binom y2+\binom z2+(y+2)x+(z+2)y+(x+2)z+1=\binom{x+y+z+2}2. $$ Since there are $\binom n2$ solutions $(x,y,z)$ to $x+y+z+2=n$, we get indeed $$R(q)=1+\frac 1{(q)_\infty^3}\sum_{n=0}^\infty(-1)^n\binom n2q^{\binom n2}. $$

$\endgroup$

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.