---
title: Higher Specht Basis in Algebraic Combinatorics
url: https://www.emergentmind.com/topics/higher-specht-basis
type: topic
---

# Higher Specht Basis in Algebraic Combinatorics

In algebraic combinatorics and the representation theory of symmetric groups, a higher Specht basis is a basis of an $S_n$-module that is organized constituentwise by Specht modules and is usually given by explicit polynomials, monomials, or diagrams. A precise formulation used in the Hessenberg-variety setting is the following: if an $S_n$-module $R$ decomposes as $R\cong\bigoplus_\lambda c_\lambda V_\lambda$, then a higher Specht basis is a basis $B=\bigsqcup_\lambda\bigsqcup_{j=1}^{c_\lambda} B_{\lambda,j}$ with each $B_{\lambda,j}$ a basis of the $j$th copy of $V_\lambda$ [2410.08366]. In the cited literature, the term appears in several closely related settings: the Ariki–Terasoma–Yamada polynomial basis for the coinvariant ring and its generalizations, web and jellyfish bases for flamingo Specht modules, bases for diagonal and Hessenberg cohomological models, and stable constructions for direct limits and infinite symmetric groups [2005.02110].

## 1. Foundational polynomial construction

The basic higher Specht construction is formulated in the polynomial ring $\mathbb Q[X_n]=\mathbb Q[x_1,\dots,x_n]$ with the natural $S_n$-action by permutation of variables. For a partition $\lambda\vdash n$, a standard Young tableau $T\in SYT(\lambda)$ has row-stabilizer $R(T)$ and column-stabilizer $C(T)$, and the Young symmetrizer is
$$
E_T:=\sum_{\sigma\in C(T)}\operatorname{sgn}(\sigma)\cdot \sum_{\tau\in R(T)}\tau.
$$
Given $S,T\in SYT(\lambda)$, let $h_i$ be the value of the cocharge tableau $ct(S)$ in the box of $T$ labeled $i$, define
$$
p_{S,T}:=\prod_{i=1}^n x_i^{h_i},
$$
and set
$$
F_{S,T}(x_1,\dots,x_n)=\frac{1}{s_{S,T}}\,E_T\bigl(p_{S,T}\bigr),
$$
where $s_{S,T}=|Stab_{R(T)}(p_{S,T})|$ [2505.07097]. The normalization by $s_{S,T}$ is explicit in the finite-variable theory and is used to make the $S_n\to S_{n+1}$ compatibility transparent. These polynomials are homogeneous of degree $cc(S)$, where $cc(S)$ is the total cocharge of $S$, and they transform under $C(T)$ by the sign character [2505.07097].

This construction recovers the classical Specht polynomial in a special case. When the auxiliary tableau is chosen in the minimal way, one obtains the ordinary Specht polynomial up to a scalar, and when $S=T$ one recovers the classical Specht polynomial in the normalization used by Zemel [2505.07097, 2505.07098]. In the coinvariant quotient
$$
R_n=\mathbb Q[X_n]/(e_1,\dots,e_n),
$$
the collection of higher Specht polynomials yields a basis compatible with the decomposition of the regular representation into irreducible Specht modules [2005.02110].

A further generalization replaces the cocharge tableau by an arbitrary semistandard tableau $M$ of shape $\lambda$ and total entry-sum $d$. Writing
$$
P_{M,T}(x):=\prod_{i=1}^n x_i^{\,M(v_T(i))}
$$
and
$$
F_{M,T}(x):=\frac{1}{S_{M,T}}\sum_{\sigma\in C(T)}\operatorname{sgn}(\sigma)\,\sigma\cdot P_{M,T}(x),
$$
one obtains, for fixed $M$, a basis $\{F_{M,T}:T\in SYT(\lambda)\}$ of an irreducible $S_n$-submodule $V_M\cong S^\lambda$ inside $\mathbb Q[X_n]_d$ [2505.07098]. As $\lambda$ and $M$ vary, the homogeneous degree-$d$ piece decomposes as
$$
\mathbb Q[X_n]_d \cong \bigoplus_{\lambda\vdash n}\;\bigoplus_{C\in SSYT_\lambda,\ |C|=d} V_C,
$$
so the multiplicity of $S^\lambda$ is the number of semistandard $\lambda$-tableaux of total weight $d$ [2505.07098].

## 2. Coinvariant generalizations and orbitwise refinements

The higher Specht philosophy extends beyond the classical coinvariant ring. For
$$
I_{n,k}=\langle x_1^k,\dots,x_n^k,\ e_n,e_{n-1},\dots,e_{n-k+1}\rangle
\quad\text{and}\quad
R_{n,k}=\mathbb C[x_1,\dots,x_n]/I_{n,k},
$$
an extended higher Specht basis is given by
$$
F^S_T\cdot e_1^{i_1}e_2^{i_2}\cdots e_{n-k}^{i_{n-k}}
$$
with $(i_1,\dots,i_{n-k})$ a vector of nonnegative integers satisfying
$$
i_1+\cdots+i_{n-k}<k-\mathrm{maj}(S).
$$
These elements descend to a basis of $R_{n,k}$, and the graded Frobenius series is
$$
\mathrm{Frob}(R_{n,k};q)
=
\sum_{S\in SYT(n)}
q^{\mathrm{maj}(S)}
\binom{n-\mathrm{maj}(S)-1}{n-k}_q
\,s_{\operatorname{shape}(S)}
$$
[2005.02110].

The same paper formulates a semistandard higher Specht basis conjecture for the Garsia–Procesi modules $R_\mu$, where $\mu\vdash n$, and proves it when $\mu$ has at most two rows. In the two-row case $\mu=(n-k,k)$, the candidate basis is indexed by semistandard tableaux $S\in SSYT(\lambda,\mu)$ together with standard tableaux $T\in SYT(\lambda)$, and the proof uses the Garsia–Procesi recursion and a block-lower-triangular transition matrix argument [2005.02110]. For Griffin’s modules $R_{n,k,\mu}$, a full theorem is established in the one-row-plus-one case: when $\mu=(n-1)$, the elements
$$
F^S_T\cdot e_1^i \qquad (0\le i<k-cc(S))
$$
descend to a basis of $R_{n,k,(n-1)}$ [2005.02110].

A distinct refinement decomposes non-transitive ordered-partition actions into orbit modules. For a subset $I\subseteq[n-1]$, Zemel defines
$$
F_{w,I}(x)=F_w(x)\cdot \prod_{i\in(I\setminus D_{sl}(w))} e_i
$$
and
$$
R_{n,I}:=\operatorname{Span}\{F_{w,I}:w\in S_n,\ D_{sl}(w)\subseteq I\}.
$$
Then
$$
R_{n,I}\cong \mathbb Q[OP_{n,I}]
$$
and
$$
R_{n,I}\cong
\bigoplus_{\lambda\vdash n}
\Bigl(\bigoplus_{S\in SYT(\lambda),\ D_{si}(S)\subseteq I} S^\lambda\Bigr),
$$
so the higher Specht basis respects the orbit decomposition of the ordered-partition action [2505.07097]. The same framework lifts the classical branching rule by two explicit operations: bar-insertion gives
$$
R_{n+1,I\cup\{n\}}\cong \operatorname{Ind}_{S_n}^{S_{n+1}} R_{n,I},
$$
while star-insertion gives
$$
R_{n+1,I}\cong \operatorname{Ext}_{S_n}^{S_{n+1}} R_{n,I}
$$
[2505.07097].

## 3. Web and jellyfish models for flamingo Specht modules

A major diagrammatic development realizes certain higher Specht bases by webs. For $\lambda=(d,d,1^\ell)$, Patrias, Pechenik, and Striker construct an invariant-ring model in which the multidegree-$(1^n)$ piece realizes $S^\lambda$, and they index a basis by noncrossing set partitions of $[n]$ into $d$ blocks of size at least $2$ [2112.05781]. For such a partition $\pi$, the web invariant is
$$
[\pi](x)=\sum_{T\in J(\pi)}(-1)^{inv(T)}\,J(T),
$$
where $J(\pi)$ is the set of jellyfish tableaux of content $\pi$ and $J(T)$ is a product of top-justified minors. The basis is controlled by a five-term skein relation, and linear independence is proved by exhibiting distinct leading monomials [2112.05781]. This recovers the classical $\mathfrak{sl}_2$ web basis when $\ell=0$, and the $\ell=1$ case matches the skein basis for $S^{(d,d,1)}$.

The flamingo generalization treats partitions
$$
\lambda=(d^r,1^{n-rd}).
$$
For an ordered set partition $\pi=(\pi_1|\cdots|\pi_d)$ of $[n]$ into $d$ blocks each of size at least $r$, the associated jellyfish polynomial is
$$
[\pi]_r=\sum_{T\in J_r(\pi)} \operatorname{sgn}(T)\,J(T),
$$
where $J_r(\pi)$ is the set of $r$-jellyfish tableaux and $J(T)$ is a product of minors of an $n\times n$ matrix $M$ [2308.07256]. Via the map $\Phi:M\mapsto (M_0|M)$ with the sign-twisted diagonal matrix $M_0$, these invariants lie in the homogeneous coordinate ring of $Gr(n,2n)$, and they can also be realized as tensor invariants of explicit bipartite tensor diagrams $W_{\pi,r}$ [2308.07256].

The behavior depends sharply on $r$. For $r=2$, the noncrossing-partition family is a basis of $S^{(d,d,1^{n-2d})}$, recovering the pennant web basis. For $r=1$, one can choose exactly one noncrossing ordered partition $\pi^P$ for each $P\subset\{2,\dots,n\}$ of size $d-1$, and the family $\{[\pi^P]_1\}$ is a basis of the hook Specht module $S^{(d,1^{n-d})}$. For $r>2$, the noncrossing family is linearly independent and gives a basis of a well-behaved subspace, but not of the entire module [2308.07256]. This directly rules out the common extrapolation from the hook and pennant cases to all flamingo shapes. The same paper records a $3$-term skein in the hook case and, for general $r$, a $(2^r+1)$-term higher Plücker recurrence, while conjecturing that an $r$-weakly noncrossing enlargement should produce a full basis [2308.07256].

## 4. Three-row bases: M-diagrams and rotation-invariant webs

For the three-row rectangular shape $(n,n,n)$, Zhu proves that M-diagrams form a basis of the Specht module $S^{(n,n,n)}$ [2305.04731]. An M-diagram is a fork diagram on $3n$ boundary points in which the left arcs among all forks are pairwise non-crossing and the right arcs are pairwise non-crossing. The resulting basis sits between the classical polytabloid basis and the non-elliptic $\mathfrak{sl}_3$ web basis, and all three transition matrices are unitriangular with respect to the left-weak Bruhat order [2305.04731]. The paper states that, from the point of view of higher Specht theory, these diagrams realize a parabolic Kazhdan–Lusztig-type basis for the Hecke module underlying $S^{(n,n,n)}$.

A further refinement, specialized to the three-row flamingo shape $(d^3,1^{n-3d})$, is the rotation-invariant web basis indexed by augmented $\mathrm{SL}_3$-webs $AW(n,d)$ [2402.11994]. These are planar bipartite normal plabic graphs embedded in a disk with boundary vertices $1,\dots,n$, white vertices of degree exactly $3$, black vertices of degree at least $3$, no faces of degree $<6$, and exceedance
$$
\#(\text{black vertices})-\#(\text{white vertices})=d
$$
[2402.11994]. There is a natural bijection
$$
AW(n,d)\;\longleftrightarrow\;WNC(n,d,3)\;\longleftrightarrow\;SYT(d^3,1^{n-3d}),
$$
where $WNC(n,d,3)$ denotes $3$-weakly noncrossing set partitions [2402.11994].

The $S_n$-action is described combinatorially by local skein rules once a perfect orientation is chosen. These rules include crossing reduction at adjacent boundary legs, square-face removal, double-edge collapse, and leaf and $2$-valent reductions [2402.11994]. The long cycle acts by boundary rotation, yielding cyclic sieving with the fake-degree polynomial
$$
X_{n,d}(q)=q^{b(\lambda)}\frac{[n]!_q}{\prod_{(i,j)\in\lambda}[h_{ij}]_q}.
$$
If $n$ is odd, $(AW(n,d),C,X_{n,d}(q))$ exhibits genuine cyclic sieving; if $n$ is even, the result is signed sieving [2402.11994]. The construction extends the jellyfish invariants of Fraser, Patrias, Pechenik, and Striker, and it is closely related to Lam’s weblike subgraph expansion into classical $\mathrm{SL}_3$ web invariants [2402.11994].

## 5. Diagonal and geometric realizations

Under the diagonal action of $S_n$ on two sets of variables $x_1,\dots,x_n$ and $y_1,\dots,y_n$, Gillespie defines two-variable higher Specht polynomials
$$
F_T^{c,d}(x,y):=\varepsilon_T\cdot(x_T^c\,y_T^d),
$$
where $\varepsilon_T$ is the Young anti-symmetrizer and $c,d$ are exponent vectors [2402.05221]. If the set $\{F_T^{c,d}:T\in SYT(\lambda)\}$ is linearly independent, then its span is a copy of the irreducible $S_n$-module $V_\lambda$. The main application is the hook-shape Garsia–Haiman module $DR_\mu$ for $\mu=(n-k+1,1^{k-1})$, where modified cocharge labelings $ccTab_\mu(S)$ and $ccTab'_\mu(S)$ attached to a second tableau $S$ define
$$
F_T^S(x,y)=\varepsilon_T\cdot\bigl(x_T^{\,ccTab_\mu(S)}\,y_T^{\,ccTab'_\mu(S)}\bigr).
$$
As $(T,S)$ ranges over pairs of standard tableaux of the same shape, these polynomials form a higher Specht basis of $DR_\mu$ and refine the ordinary decomposition into irreducibles [2402.05221]. The same paper derives a doubly graded Frobenius formula in terms of the statistics $maj_{1,n-k+1}$ and $comaj_{n-k+1,n}$.

A different geometric realization arises in the cohomology of regular semisimple Hessenberg varieties of type $h=(h(1),n,\dots,n)$. Salois constructs two sets of monomials, $B_1$ and $B_3$, and proves that $B_1\cup B_3$ is a $\mathbb Z$-basis of $H^*(Hess(S,h))$ [2410.08366]. In this basis, the elements of $B_1$ are $S_n$-fixed, while each block
$$
\{x_n^{\ell_1}\cdots x_2^{\ell_{n-1}}(y_{k+1}-y_1)\}_{k=1}^{n-1}
$$
spans a copy of the Specht module $V_{(n-1,1)}$, so
$$
H^*(Hess(S,h))
\cong
h(1)(n-1)!\,V_{(n)}
\ \oplus\
(n-h(1))(n-2)!\,V_{(n-1,1)}.
$$
The construction also gives a permutation-basis refinement and weight-preserving bijections with $P_h$-tableaux and pairs $(S,T)$ of shape $(2,1^{n-2})$ [2410.08366]. Analogous statements hold for the transpose Hessenberg function $h'=((n-1)^{n-m},n^m)$.

## 6. Stable limits and terminological extensions

The finite-variable theory admits a stable limit. Zemel proves that if $w\in S_n$ and $w^+\in S_{n+1}$ fixes $n+1$, then
$$
F_{w^+}(x_1,\dots,x_n,x_{n+1})\big|_{x_{n+1}=0}=F_w(x_1,\dots,x_n),
$$
and from this obtains a stable higher Specht power series
$$
F_w^\infty(x_1,x_2,\dots)\in \mathbb Q[[x_1,x_2,\dots]]
$$
of degree $\mathrm{maj}(w)$ for each $w\in S_\infty$ [2505.07097]. A more general stable framework is developed in the ring of eventually symmetric functions
$$
\tilde\Lambda
=
\{F\in\mathbb Q[[x_1,x_2,\dots]]:\deg(F)<\infty,\ F\text{ fixed by }S_{>N}\text{ for some }N\},
$$
where one defines stable generalized higher Specht polynomials $\widetilde F_{\mathcal M,\mathcal T}$ indexed by infinite semi-standard and standard tableaux [2505.07099]. As $\mathcal T$ varies, these polynomials form a basis of an irreducible $S_\infty$-module $S^{\widehat A}$, the direct-limit analogue of a finite Specht module. The same paper introduces a filtration by the number of nonzero entries in the first row and proves that successive quotients are maximal completely reducible subrepresentations [2505.07099].

A separate, explicitly terminological extension appears in the completed ring of symmetric functions. In “Specht modules decompose as alternating sums of restrictions of Schur modules,” the basis elements $F_\nu$ are called the higher-Specht or stable-Specht basis and are defined by inverting stable restriction from Schur modules [1809.10125]. Their Schur expansions alternate in sign by degree, and their multiplication is governed by stable Kronecker coefficients:
$$
F_\alpha\cdot F_\beta=\sum_\gamma g^\gamma_{\alpha,\beta}\,F_\gamma.
$$
This usage is related by representation-theoretic motivation, but it is not the same construction as the ATY polynomial basis [1809.10125].

Taken together, these constructions show that “higher Specht basis” is best understood as a family of representation-compatible bases rather than a single object. In the polynomial, web, cohomological, and stable settings, the common feature is the same: an explicit basis that exhibits individual copies of Specht modules inside a larger $S_n$-module, often together with branching, skein, or grading data that is invisible in a generic basis.

Source: https://www.emergentmind.com/topics/higher-specht-basis