The symmetric functions catalog

An overview of symmetric functions and related topics

2020-08-06

Cylindric Schur polynomials

The family of (skew) cylindric Schur functions were introduced by A. Postnikov in [Pos05]. This is motivated by the fact that the coefficients in the Schur expansion of such cylindric Schur functions are given by 3-point Gromov–Witten invariants of the Grassmanian. See also [McN06] for an introduction to this family of symmetric functions.

The cylindric Schur functions are a subset of skew affine Schur functions, which was proved by T. Lam, in [Prop. 33, Lam06a].

Definition

A cylindric shape $\lambda$ is an infinite lattice path in $\setZ^2,$ invariant under shifts $(n-m,m)$ where $m \in [1,n-1]$ and let $C^{n,m}$ be the set of such paths. Suppose $\lambda,\mu \in C^{n,m}$ and that $\mu$ always lie weakly to the left of $\lambda,$ then we say that $\mu \subseteq \lambda$ and $\lambda / \mu$ is a cylindric skew shape.

A cylindric semi-standard tableau of shape $\lambda / \mu$ is a sequence

$\mu = \lambda^0 \subseteq \lambda^1 \subseteq \dotsb \subseteq \lambda^{\ell} = \lambda$

such that each shape $\lambda^i/\lambda^{i-1}$ contains at most one box in each column. Furthermore, we can think of placing the value $i$ in the boxes determined by $\lambda^i/\lambda^{i-1}.$ The weight $(a_1,a_2,\dotsc,a_{\ell})$ of the tableau is given by letting $a_i$ be the number of boxes in any $n-m$ consecutive columns of $\lambda^i/\lambda^{i-1}.$

The cylindric Schur function $\schurCylindric_{\lambda/\mu}(\xvec)$ is then defined as

$\schurCylindric_{\lambda/\mu}(\xvec) = \sum_{T \in \mathrm{CSSYT(\lambda/\mu)}} \xvec^T$

where the sum is over all cylindric semi-standard tableaux of shape $\lambda/\mu.$ One can show via a variant of the Bender–Knuth involution that all $\schurCylindric_{\lambda/\mu}(\xvec)$ are symmetric functions.

Example.

In the following examples, $n=7,$ $m=3.$ This implies invariance when shifting $3$ steps up and $4$ steps to the right. We show two cylindric semistandard tableaux, of two different shapes, where exactly one representative of boxes under the shift action has been assigned an integer. Note that the second example shows that the skew Schur functions is a proper subset of cylindric Schur functions.

The first tableau has weight $(3,3,2,1,1)$ and the second one $(2,3,1,1).$ The Schur expansion of the first shape gives

$\schurCylindric_{\lambda/\mu} = \schurS_{433} + \schurS_{4321} - \schurS_{33211} + \schurS_{2221111} - \schurS_{211111111}.$

Postnikov uses the notation $\lambda/d/\mu$ to describe a cylindric shape, where $\lambda$ and $\mu$ are partitions and $d\geq 0.$ The cylindric shape is constructed as follows: Draw the outline of the partition $\lambda$ in the plane, but shifted $d$ steps right and $d$ steps down, then the outline of $\mu$ without shift, and extend both outlines periodically.

In the example above, the first shape can be described as $(6,3,3)/1/(6,3,1)$ and the second shape is just $(4,4,4)/0/(3,2,0),$ so that when $d=0$ the shape is a regular skew shape.

Every cylindric Schur polynomial can be obtained as a skew affine Schur function.

Toric Schur polynomials

A shape $\lambda/\mu \in C^{n,m}$ is called toric if every column contains at most $m$ boxes. Cylindric Schur functions indexed by such shapes are referred to as toric Schur polynomials. All the usual skew shapes are toric.

The specialization

$\schurCylindric_{\lambda/d/\mu}(\xvec)(x_1,\dotsc,x_m) \coloneqq \schurCylindric_{\lambda/d/\mu}(x_1,\dotsc,x_m,0,\dotsc)$

is non-zero if and only of the shape $\lambda/d/\mu$ is toric. These polynomials, in a finite set of variables, are referred to as toric Schur polynomials.

Schur expansion

Postnikov proves that when $\lambda/d/\mu$ is a toric shape in $C^{n,m}$ the coefficients $C_{\mu,\nu}^{\lambda,d}$ in the Schur expansion

$\schurCylindric_{\lambda/d/\mu}(x_1,\dotsc,x_m) = \sum_\nu C_{\mu,\nu}^{\lambda,d} \schurS_\nu(x_1,\dotsc,x_m)$

are non-negative integers given by certain Gromov–Witten invariants.

Problem (Postnikov, [Pos05]).

It is an open problem to find a combinatorial rule for $C_{\mu,\nu}^{\lambda,d}.$ Note that such a rule would generalize the classical Littlewood–Richardson rule.

Postnikov also introduces a certain family of toric Specht modules and conjecture that they decompose into irreducible Specht modules with multiplicities given by the $C_{\mu,\nu}^{\lambda,d}.$

Cylindric Schur positivity

P. McNamara [McN06] asked for which cylindric shapes $\schurCylindric_{\lambda/d/\mu}(\xvec)$ are Schur-positive (as proper symmetric functions in an infinite alphabet) and conjectured that each cylindric skew Schur function can be positively expanded in certain non-skew cylindric Schur functions, with coefficients being Gromov–Witten invariants. P. NcNamara proves that only ordinary skew Schur functions are Schur-positive. In fact, these are the only cylindric Schur functions that are positive in the Gessel fundamental quasisymmetric basis, see [Thm. 5.7, McN06].

McNamara's conjecture was recently answered in [Lee17] in the following theorem:

\begin{equation*} \schurCylindric_{\lambda/d/\mu}(\xvec) = \sum_{\nu \in C^{n,m}, e\geq 0} c^{\lambda/d/\mu}_{\nu/e/\emptyset} \schurCylindric_{\nu/e/\emptyset}(\xvec). \end{equation*}

The coefficients are all non-negative, and satisfy $c^{\lambda/d/\mu}_{\nu/e/\emptyset} = c^{\lambda/d-1/\mu}_{\nu/e-1/\emptyset}$ whenever $e \gt 0.$ Furthermore, we have that $c^{\lambda/d/\mu}_{\nu/0/\emptyset} = C_{\mu,\nu}^{\lambda,d}.$

Jacobi–Trudi identity

A Jacobi–Trudi identity is presented in [Eq. (11), Pos05].

Cylindric complete homogeneous polynomials

Introduced in [KP20].

Cylindric elementary polynomials

Introduced in [KP20].