Regular semisimple Hessenberg varieties with cohomology rings generated in degree two

Basic Information

• Paper ID: 2301.03762
• Title: Regular semisimple Hessenberg varieties with cohomology rings generated in degree two
• Authors: Mikiya Masuda, Takashi Sato
• Classification: math.AG (Algebraic Geometry), math.AT (Algebraic Topology), math.SG (Symplectic Geometry)
• Publication Date: November 11, 2025 (arXiv v4)
• Paper Link: https://arxiv.org/abs/2301.03762

Abstract

Regular semisimple Hessenberg varieties $\mathrm{Hess}(S,h)$ are smooth subvarieties of flag varieties, determined by a square matrix $S$ with distinct eigenvalues and a Hessenberg function $h$. The cohomology ring $H^*(\mathrm{Hess}(S,h))$ is independent of the choice of $S$, but has not been explicitly described except in rare cases. This paper characterizes the Hessenberg functions $h$ for which $H^*(\mathrm{Hess}(S,h))$ as a ring is generated in degree two, showing that such $h$ are precisely the so-called (double) lollipop-type functions.

Research Background and Motivation

1. Research Problem

This paper investigates the cohomology ring structure of regular semisimple Hessenberg varieties, particularly characterizing which Hessenberg functions $h$ make the cohomology ring $H^*(X(h))$ generated in degree two (i.e., all generators come from degree two cohomology).

2. Problem Significance

• Deep connections with combinatorics: Hessenberg varieties are related to the symmetric group $S_n$ through Tymoczko's point action. The Shareshian-Wachs conjecture, resolved by Brosnan-Chow and Guay-Paquet, connects $H^*(\mathrm{Hess}(S,h))$ as an $S_n$-module to chromatic symmetric functions of graphs.
• Stanley-Stembridge conjecture: Studying the geometry and topology of Hessenberg varieties has opened avenues for proving the famous Stanley-Stembridge conjecture in graph theory.
• Foundational nature of ring structure: Understanding when the cohomology ring is generated in degree two is fundamental to studying its more complex algebraic structures.

3. Limitations of Existing Methods

• For general Hessenberg functions $h$, the ring structure of $H^*(X(h))$ remains explicitly undescribed.
• Known results are limited to special cases:
  • $h = (n,...,n)$ corresponds to the flag variety $\mathrm{Fl}(n)$
  • $h = (2,3,...,n,n)$ corresponds to the permutohedron variety (a toric variety)
  • $h = (h(1),n,...,n)$: results in 2 show the ring is generated in degree two only when $h(1)=2$ or $n$

4. Research Motivation

Based on patterns in known special cases, a natural question arises: for general Hessenberg functions $h$, when is $H^*(X(h))$ generated in degree two? This paper provides a complete answer by synthesizing Morse-Bott theory, GKM theory, and fibration techniques.

Core Contributions

1. Complete Characterization Theorem (Main Theorem 1.1): For connected regular semisimple Hessenberg varieties (i.e., $h(j) \geq j+1$ for all $j \in [n-1]$), $H^*(X(h))$ is generated in degree two if and only if $h$ has the following "double lollipop" form:a+1 & (1 \leq j \leq a) \\ j+1 & (a < j < b) \\ n & (b \leq j \leq n) \end{cases}$$ where $1 \leq a < b \leq n$
2. Geometric proof of sufficiency: When $h$ is double lollipop-type, $X(h)$ is a fibration over a compact smooth toric variety with fibers being products of flag varieties, hence the cohomology ring is generated in degree two.
3. Topological proof of necessity: Using Morse-Bott functions and surjectivity of restriction maps, inductive arguments eliminate all non-double-lollipop cases.
4. Technical contributions:
   • Explicit generators of $H^2(X(h))$ via GKM theory
   • Fine calculations of Hilbert series proving non-generation in degree two for certain cases
   • Establishment of fibration relationships between Hessenberg varieties

Detailed Methods

Task Definition

Input: Hessenberg function $h: [n] \to [n]$, monotone non-decreasing with $h(j) \geq j$

Output: Determine whether $H^*(X(h))$ as a ring is generated in degree two

Constraints: Assume $h(j) \geq j+1$ for all $j \in [n-1]$ (ensuring $X(h)$ is connected)

Theoretical Framework

1. GKM Theory and Equivariant Cohomology

Regular semisimple Hessenberg varieties $X(h)$ are GKM manifolds with algebraic torus $T = (\mathbb{C}^*)^n$ acting on them. The fixed point set is: $X(h)^T = \mathrm{Fl}(n)^T \cong S_n$

Equivariant cohomology $H^*_T(X(h))$ embeds via localization: $H^*_T(X(h)) \hookrightarrow \bigoplus_{w \in S_n} \mathbb{Z}[t_1,...,t_n]$

The image is characterized by graph cohomology conditions: $\{f \in \mathrm{Map}(S_n, \mathbb{Z}[t_1,...,t_n]) \mid f(w) - f(w(i,j)) \in (t_{w(i)} - t_{w(j)}), \forall w \in S_n, j < i \leq h(j)\}$

2. Explicit Generators of Degree Two Cohomology

According to Ayzenberg-Masuda-Sato 4, $H^2(X(h))$ is generated by:

• $x_k$ ($k \in [n]$): $x_k(w) = t_{w(k)}$, corresponding to first Chern classes of tautological line bundles
• $y_{j,k}$ ($j \in \bot(h), k \in [n]$): $y_{j,k}(w) = \begin{cases} t_k - t_{w(j+1)} & k \in \{w(1),...,w(j)\} \\ 0 & \text{otherwise} \end{cases}$ where $\bot(h) = \{j \in [n-1] \mid h(j-1) = h(j) = j+1\}$
• $\tau_A$ ($|A| \in L(h)$): $\tau_A(w) = \begin{cases} t_{w(|A|)} - t_{w(|A|+1)} & \{w(1),...,w(|A|)\} = A \\ 0 & \text{otherwise} \end{cases}$ where $L(h) = \{j \in [n-1] \mid h(j-1) = j, h(j) = j+1\}$

Necessity Proof Strategy (Section 3)

Key Lemma 3.1 (Morse-Bott Function)

Construct the moment map $f_h: X(h) \to \mathbb{R}$ whose critical point sets are $X(h_1)$ and $X(h_n)$, where $h_j$ is the Hessenberg function obtained by deleting row and column $j$ from $h$'s configuration.

Using Morse-Bott theory (Tolman-Weitsman 19), prove that the restriction maps: $H^*(X(h);\mathbb{Q}) \to H^*(X(h_1);\mathbb{Q}), \quad H^*(X(h);\mathbb{Q}) \to H^*(X(h_n);\mathbb{Q})$ are both surjective.

Corollary 3.2: If $h'$ is a minor of $h$ (obtained by repeatedly taking $h_1$ or $h_n$) and $H^*(X(h);\mathbb{Q})$ is generated in degree two, then $H^*(X(h');\mathbb{Q})$ is also generated in degree two.

Proposition 3.3 (Combinatorial Characterization)

$h$ is double lollipop-type if and only if $h$ contains no minors of the following types:

• $(\alpha, \beta,...,\beta)$ or $(\beta-1,...,\beta-1, \beta,...,\beta)$ ($\alpha$ times), where $3 \leq \alpha < \beta$
• $(2, \gamma-1,...,\gamma-1, \gamma, \gamma)$, where $\gamma \geq 5$

The first two cases were already excluded in 2; the key is handling the third case.

Proof of Proposition 3.4 (Technical Core)

For $h = (2, n-1,...,n-1, n, n)$ ($n \geq 5$), prove that $H^*(X(h);\mathbb{Q})$ is not generated in degree two through:

Step 1: Use Theorem 2.1(4) to compute the recursion formula for Poincaré polynomials (Lemma 3.7): $P_n(q) = (1+q)^2[n-2]_q! + (n-2)(q+q^2)[n-3]_q[n-3]_q! + \cdots$

Step 2: Prove that the coefficient of $P_n(q)$ at degree $n-3$ is (Lemma 3.8): $\binom{n}{2} + n(n-3)/2$

Step 3: Calculate the upper bound for the Hilbert series of the subring $R(h)$ generated by $H^2(X(h))$:

• Introduce auxiliary elements $\rho_k$, rewrite generators as $x_k, \tau_k, \rho_k$
• Decompose $R(h)$ as direct sum $A(h) \oplus B(h) \oplus C(h) \oplus D(h)$ (as $A(h)$-modules)
• Compute Hilbert series bounds for each part separately (formulas 3.23-3.27)

Step 4: Prove that the coefficient of $\mathrm{Hilb}(R(h), \sqrt{q})$ at degree $n-3$ is: $\binom{n}{2}$ strictly less than the corresponding coefficient of $P_n(q)$, a contradiction!

Sufficiency Proof Strategy (Section 4)

Fibration Construction

For double lollipop-type $h$, define the map: $\pi_{[a,b]}: \mathrm{Fl}(n) \to \mathrm{Fl}_{[a,b]}(n)$$\pi_{[a,b]}(V_1 \subset \cdots \subset V_n) = (V_a \subset V_{a+1} \subset \cdots \subset V_b)$

Key observation: $V_\bullet \in X(h)$ if and only if $\pi_{[a,b]}(V_\bullet)$ satisfies $SV_k \subset V_{k+1}$ ($a \leq k \leq b-1$)

Define: $Y_{[a,b]} = \{(V_a \subset \cdots \subset V_b) \mid SV_k \subset V_{k+1}, a \leq k \leq b-1\}$

Then $\pi_{[a,b]}$ restricted to $X(h)$ gives a fibration: $\mathrm{Fl}(a) \times \mathrm{Fl}(n-b) \to X(h) \xrightarrow{\pi_{[a,b]}} Y_{[a,b]}$

Lemma 4.1 (Toric Variety Structure)

Prove that $Y_{[a,b]}$ is a compact smooth toric variety of dimension $n-1$:

• Construct $(\mathbb{C}^*)^n$-orbits by choosing special vectors $g = (g_1,...,g_n) \in (\mathbb{C}^*)^n$
• Use Vandermonde determinants to prove the orbit has dimension $n-1$

Conclusion: Since both the base space $Y_{[a,b]}$ (toric variety) and fiber $\mathrm{Fl}(a) \times \mathrm{Fl}(n-b)$ (product of flag varieties) have cohomology rings generated in degree two, the Serre spectral sequence of the fibration implies $H^*(X(h))$ is also generated in degree two.

Experimental Setup

This is a pure mathematics theoretical paper with no numerical experiments. All results are derived through rigorous mathematical proofs.

Computational Verification

• Explicit Poincaré polynomial calculations for small values of $n$ (such as $n=3,4,5$)
• Concrete computations of degree two cohomology generators via GKM theory (Examples 2.1, 2.2)

Experimental Results

Main Results

Theorem 1.1 (Main Theorem): Assume $h(j) \geq j+1$ for all $j \in [n-1]$. Then $H^*(X(h))$ is generated in degree two if and only if there exist $1 \leq a < b \leq n$ such that: