---
title: Signed Knapsack Numbers in GL(n)
url: https://www.emergentmind.com/topics/signed-knapsack-numbers
type: topic
---

# Signed Knapsack Numbers in GL(n)

Searching arXiv for the primary paper and closely related references on LR coefficients, Kostka numbers, and Kostant partition function.
Signed knapsack numbers, in the setting of $GL_n$, are alternating sums of knapsack counts indexed by the Weyl group $S_n$. In the formulation developed in "Littlewood-Richardson coefficients as a signed sum of Kostka numbers" [2211.10669], they arise through Steinberg’s formula for Littlewood–Richardson coefficients, where each summand is a value of Kostant’s partition function and therefore counts nonnegative integer solutions to a linear equation with columns the positive roots. The same work places signed knapsack numbers at the intersection of symmetric-function expansions, tableau combinatorics, and root-system methods by showing that Littlewood–Richardson coefficients are signed sums of Kostka numbers, and, via Kostant’s partition function, signed sums of knapsack numbers [2211.10669].

## 1. Definition and ambient structures

The relevant ambient objects are partitions, Schur polynomials, Littlewood–Richardson coefficients, and Kostka numbers. A partition, or highest weight, is a weakly decreasing $n$-tuple of nonnegative integers
$$
\lambda = (\lambda_1,\lambda_2,\dots,\lambda_n),
$$
with total weight $|\lambda|:=\sum_i \lambda_i$. The dominance order $\mu \le \lambda$ means
$$
\sum_{i=1}^k \mu_i \le \sum_{i=1}^k \lambda_i \quad \text{for all } k \ge 1,
$$
with $|\mu|=|\lambda|$ [2211.10669].

For $\lambda$ with at most $n$ parts, the Schur polynomial $S_\lambda(x)$ in variables $x=(x_1,\dots,x_n)$ is defined by
$$
S_\lambda(x)=\frac{\det\big(x_i^{\,\lambda_j+n-j}\big)_{1\le i,j\le n}}{\det\big(x_i^{\,n-j}\big)_{1\le i,j\le n}}.
$$
These form a basis for $\mathrm{Sym}_n=\mathbb{C}[x_1,\dots,x_n]^{S_n}$ and represent the characters of irreducible polynomial representations of $GL_n(\mathbb{C})$ [2211.10669].

Given partitions $\lambda,\mu,\nu$ with $|\nu|=|\lambda|+|\mu|$, the Littlewood–Richardson coefficients $c_{\lambda\,\mu}^{\nu}$ are the structure constants for multiplication in the Schur basis:
$$
S_\lambda(x)\,S_\mu(x)=\sum_\nu c_{\lambda\,\mu}^{\nu}\,S_\nu(x).
$$
Equivalently, if $(\cdot,\cdot)$ denotes the Hall inner product, then
$$
c_{\lambda\,\mu}^{\nu}=(S_\lambda S_\mu,S_\nu).
$$

For a partition $\lambda$ and a composition $\mu$ of the same total weight, the Kostka number $K_{\lambda,\mu}$ counts semistandard Young tableaux of shape $\lambda$ and content $\mu$. It is also the coefficient of $S_\lambda$ in a product of complete symmetric functions:
$$
H_{\mu_1}(x)\,H_{\mu_2}(x)\cdots H_{\mu_n}(x)=\sum_\lambda K_{\lambda,\mu}\,S_\lambda(x),
$$
where
$$
\sum_{k\ge 0} t^k\,H_k(x)=\prod_{i=1}^n \frac{1}{1-tx_i}.
$$
A standard nonvanishing criterion is
$$
K_{\lambda,\mu}\neq 0 \iff |\lambda|=|\mu| \text{ and } \mu \le \lambda.
$$

These definitions supply the basic representation-theoretic and combinatorial context in which signed knapsack numbers appear. The key transition is that multiplicities in Schur multiplication can be rewritten as alternating sums, first over Kostka numbers and then over knapsack counts [2211.10669].

## 2. Littlewood–Richardson coefficients as signed sums of Kostka numbers

The central structural result is a product expansion expressing $S_\lambda S_\mu$ in terms of Schur functions indexed by chambered permutations of $\lambda+\xi$. Let
$$
\rho=(n-1,n-2,\dots,1,0),
$$
the half-sum of positive roots for $GL_n$, and let $G(\xi)\le S_n$ denote the stabilizer of $\xi$ under coordinate permutations. For each coset representative $\sigma\in S_n/G(\xi)$, let $\eta_{\lambda,\sigma(\xi)}\in S_n$ be the unique permutation sending
$$
\lambda+\sigma(\xi)+\rho
$$
to the dominant Weyl chamber. Then the theorem is [2211.10669]:
$$
S_\lambda(x)\, S_\mu(x)
\;=\;
\sum_{\xi} K_{\mu,\xi}\,
\sum_{\sigma \in S_n/G(\xi)}
(-1)^{\ell\!\big(\eta_{\lambda,\sigma(\xi)}\big)}
\;
S_{\;\eta_{\lambda,\sigma(\xi)}\big(\lambda + \sigma(\xi) + \rho\big)-\rho}(x).
$$

Equating coefficients of $S_\nu$ yields the signed-sum formula
$$
c_{\lambda\,\mu}^{\nu}
\;=\;
\sum_{w \in S_n}
(-1)^{\ell(w)}\;
K_{\mu,\; T_w\big(w(\nu+\rho)-(\lambda+\rho)\big)}.
$$
Here $T_w\in S_n$ is the unique permutation that reorders the tuple
$$
w(\nu+\rho)-(\lambda+\rho)
$$
into weakly decreasing order, so that the Kostka number is well defined. Any term for which this chambered tuple is not a partition contributes zero, because Kostka numbers vanish outside the dominance constraints [2211.10669].

This identity is the immediate precursor of the signed knapsack interpretation. It expresses Littlewood–Richardson multiplicities not as direct tableau counts, but as alternating sums of tableau-counting quantities. The alternation is governed by the Coxeter length $\ell(w)$, and the chambering operator $T_w$ enforces dominance by moving weights into the positive Weyl chamber.

## 3. Kostant’s partition function and the knapsack interpretation

The knapsack component enters through Kostant’s partition function. For type $A_{n-1}$, the positive roots of $GL_n$ are
$$
\Phi^+=\{\,e_i-e_j \mid 1\le i<j\le n\,\},
$$
where $e_i$ are the standard basis vectors of $\mathbb{Z}^n$. Kostant’s partition function $\mathcal{P}$ is defined by the generating series
$$
\prod_{\alpha \in \Phi^+} \frac{1}{1-e^{-\alpha}}
=
\sum_{\beta\in\mathbb{Z}^n} \mathcal{P}(\beta)\,e^{-\beta}.
$$
Equivalently,
$$
\mathcal{P}(\beta)
=
\#\left\{
(m_{ij})_{1\le i<j\le n}\in \mathbb{Z}_{\ge 0}^{\binom{n}{2}}
\;\middle|\;
\sum_{1\le i<j\le n} m_{ij}(e_i-e_j)=\beta
\right\}.
$$
Thus $\mathcal{P}(\beta)$ counts the number of nonnegative integer solutions to a linear equation whose columns are the positive roots; this is precisely a knapsack-type counting problem [2211.10669].

The same exposition records a signed-sum representation of Kostka numbers in terms of $\mathcal{P}$:
$$
K_{\lambda,\mu}
=
\sum_{w\in S_n}
(-1)^{\ell(w)}\;
\mathcal{P}\big(w(\mu+\rho)-(\lambda+\rho)\big).
$$
Conceptually, this is an inclusion–exclusion formula over the Weyl group. The change of basis between complete symmetric functions and Schur functions, together with chambering into the dominant region, produces the alternating sum. The sign is the parity of the permutation length, and the partition function $\mathcal{P}$ records the underlying knapsack counts [2211.10669].

In this sense, a signed knapsack number is not merely a counting function but an alternating count obtained by Weyl-group correction. The “signed” aspect is intrinsic: it compensates for overcounting across chambers.

## 4. Steinberg’s formula and the precise meaning of “signed knapsack number”

Substituting the signed-sum formula for Kostka numbers into the signed-sum formula for Littlewood–Richardson coefficients yields Steinberg’s formula:
$$
c_{\lambda\,\mu}^{\nu}
=
\sum_{w\in S_n}
(-1)^{\ell(w)}\;
\mathcal{P}\big(w(\nu+\rho)-(\lambda+\mu+\rho)\big).
$$
This is the standard form for $GL_n$, with $W=S_n$ and $\rho=(n-1,\dots,0)$ [2211.10669].

This identity gives the exact meaning of signed knapsack numbers in this context. Each summand
$$
\mathcal{P}\big(w(\nu+\rho)-(\lambda+\mu+\rho)\big)
$$
is a knapsack number, because it counts nonnegative integer solutions to a linear system with columns the positive roots. The Littlewood–Richardson coefficient is then an alternating sum of these knapsack counts over the Weyl group. The exposition makes the terminology explicit: LR multiplicities are “signed knapsack numbers” [2211.10669].

The proof strategy recorded there proceeds in three steps. First, start from the corollary expressing $c_{\lambda\,\mu}^{\nu}$ as a signed sum of Kostka numbers indexed by chambered differences. Second, write each Kostka number as an alternating sum of partition-function values. Third, reorganize the sum using carefully chosen permutations, after which the chambering signs and nonvanishing constraints force the expression into Steinberg’s form. A key observation is that terms outside the positive cone vanish, while the permutation sending a weight into the positive chamber contributes a sign equal to the parity of its length [2211.10669].

A plausible implication is that signed knapsack numbers synthesize three perspectives that are often treated separately: Schur-basis structure constants, semistandard tableau enumeration, and positive-root partition counts. The source material states this unification explicitly at the level of formulas.

## 5. Algorithmic formulation and complexity-theoretic status

The signed-sum Kostka formulation yields an explicit algorithm for computing $c_{\lambda\,\mu}^{\nu}$, assuming access to a black box for Kostka numbers in polynomial time. The steps stated in the source are as follows [2211.10669].

1. Input $n$ and partitions $\lambda,\mu,\nu$ as $n$-tuples, and initialize $c:=0$.

2. Enumerate all compositions $\xi$ with $\xi\le \mu$ in dominance order. The exposition states that this can be done in time
$$
O(\mathrm{size}(\mu)^n),
$$
where $\mathrm{size}(\cdot)$ denotes the bit-length of the input tuple.

3. For each such $\xi$, check whether there exist permutations $w,T\in S_n$ such that
$$
\xi=T\big(w(\nu+\rho)-(\lambda+\rho)\big),
$$
that is, whether $\xi$ is a permutation of the relevant chambered difference. The paper provides an $O(n^2)$ procedure, described as iteratively matching and deleting entries, to find such $w$ and $T$ or conclude that none exist.

4. If a match exists, add
$$
c \leftarrow c + (-1)^{\ell(w)} K_{\mu,\xi}.
$$

5. Output $c$.

The complexity discussion states that Step 2 dominates enumeration, Step 3 requires $O(n^2)$ per $\xi$, and Step 1 and final output are linear in input size. Under the assumption that $K_{\mu,\xi}$ is available in polynomial time, $c_{\lambda\,\mu}^{\nu}$ is computable in polynomial time in the input size [2211.10669].

The same exposition further states that Kostka numbers are special Littlewood–Richardson coefficients and that computing either Littlewood–Richardson coefficients or Kostka numbers is #P-complete. Combined with the polynomial-time reduction from Littlewood–Richardson coefficients to Kostka numbers and known reductions in the opposite direction, this places the two problems in the same class of decision and counting problems under Turing reductions [2211.10669].

This computational perspective is central to the significance of signed knapsack numbers. They are not only formal identities; they provide a mechanism for transferring complexity-theoretic and algorithmic information between tableau counts, structure constants, and root-theoretic partition counts.

## 6. Illustrative $GL_3$ cases and interpretive remarks

Two small $GL_3$ examples recorded in the exposition illustrate how the signed formulas collapse in simple cases [2211.10669].

For
$$
\lambda=(5,3,2),\quad \mu=(4,3,3),\quad \nu=(9,6,5),\quad n=3,\quad \rho=(2,1,0),
$$
one has
$$
\nu+\rho=(11,7,5),\qquad \lambda+\rho=(7,4,2).
$$
The tuples
$$
w(\nu+\rho)-(\lambda+\rho)
$$
range over permutations of the difference, and only $(4,3,3)$ is dominant and $\le \mu$. It occurs for $w=\mathrm{id}$ with $T=\mathrm{id}$, so
$$
c_{\lambda\,\mu}^{\nu}
=
(-1)^{\ell(\mathrm{id})}K_{\mu,(4,3,3)}
=
1\cdot 1
=
1.
$$
Using Steinberg’s formula gives the same result. Since $\lambda+\mu=\nu$, one has
$$
w(\nu+\rho)-(\lambda+\mu+\rho)=w(11,7,5)-(11,7,5).
$$
For $w=\mathrm{id}$ this is $0$, so $\mathcal{P}(0)=1$; for $w\ne \mathrm{id}$ the difference is not in the positive root cone, hence the partition function vanishes. Therefore $c_{\lambda\,\mu}^{\nu}=1$.

A second check uses
$$
\lambda=(2,1,0),\quad \mu=(2,1,0),\quad \nu=(4,2,0),\quad n=3,\quad \rho=(2,1,0).
$$
Again $\nu=\lambda+\mu$, so
$$
\nu+\rho=(6,3,0),\qquad \lambda+\mu+\rho=(6,3,0).
$$
The identity permutation contributes $\mathcal{P}(0)=1$, while the other permutations contribute zero because the corresponding weights lie outside the positive cone. Hence
$$
c_{\lambda\,\mu}^{\nu}=1.
$$
In the signed-sum Kostka identity, the only contributing Kostka number comes from
$$
\xi=T\big(w(\nu+\rho)-(\lambda+\rho)\big)=(0,0,0),
$$
for which $K_{\mu,\xi}=1$, again with positive sign.

These examples show how substantial cancellation can occur in signed knapsack formulas. The exposition notes that understanding which Weyl-group elements produce weights in the positive cone is crucial [2211.10669]. It also remarks that, for type $A$, $\mathcal{P}(\beta)$ admits a flow interpretation on the complete directed graph with capacities $m_{ij}$ along edges $i\to j$ for $i<j$ summing to $\beta$, and that efficient pseudopolynomial routines exist for special cases. This suggests that the utility of signed knapsack numbers is both conceptual and practical: they provide a common language for tableau combinatorics, symmetric-function identities, and integer-solution counting.

## 7. Conceptual synthesis

The key formulas may be organized as a progression from symmetric functions to signed knapsack numbers [2211.10669]:
$$
S_\lambda S_\mu=\sum_\nu c_{\lambda\,\mu}^{\nu} S_\nu,
$$
$$
H_{\mu_1}\cdots H_{\mu_n}=\sum_\lambda K_{\lambda,\mu} S_\lambda,
$$
$$
c_{\lambda\,\mu}^{\nu}
=
\sum_{w\in S_n}(-1)^{\ell(w)}
K_{\mu,\;T_w\big(w(\nu+\rho)-(\lambda+\rho)\big)},
$$
$$
K_{\lambda,\mu}
=
\sum_{w\in S_n}(-1)^{\ell(w)}
\mathcal{P}\big(w(\mu+\rho)-(\lambda+\rho)\big),
$$
and finally
$$
c_{\lambda\,\mu}^{\nu}
=
\sum_{w\in S_n}(-1)^{\ell(w)}
\mathcal{P}\big(w(\nu+\rho)-(\lambda+\mu+\rho)\big).
$$

Within this chain, signed knapsack numbers are the endpoint at which Littlewood–Richardson multiplicities become alternating sums of counts of nonnegative integer solutions to linear equations. The source material emphasizes two consequences. First, the viewpoint unifies combinatorial tableau definitions, symmetric-function expansions, and root-theoretic partition functions. Second, it yields polynomial-time reductions between the computation of Littlewood–Richardson coefficients and Kostka numbers, situating both in the same #P-complete framework [2211.10669].

The term therefore designates more than a reformulation of Steinberg’s formula. In the precise $GL_n$ context developed in [2211.10669], a signed knapsack number is an alternating Weyl-group sum of Kostant partition counts, and Littlewood–Richardson coefficients are exactly such quantities.

Source: https://www.emergentmind.com/topics/signed-knapsack-numbers