🔢 mathematics

GIT quotient of minimal dimensional Schubert variety modulo a subtorus

This paper proves that the GIT quotient of the unique minimal dimensional Schubert variety in the Grassmannian $G_{r,n}$ (where $n=rq+1$) modulo a specific subtorus is isomorphic to the total space of the $r$ -th stage of an iterated projective space bundle over $\mathbb{P}^{q-1}$ .

Original authors: Arkadev Ghosh, S. S. Kannan

Published 2026-04-29

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Arkadev Ghosh, S. S. Kannan

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are an architect trying to build a massive, intricate sculpture out of geometric shapes. This paper is about a specific type of sculpture made from "Schubert varieties"—which are essentially special, highly structured rooms inside a giant mathematical building called a Grassmannian.

Here is the story of what the authors, Arkadev Ghosh and S. S. Kannan, discovered, explained in plain English.

The Setting: A Giant Dance Floor

Think of the Grassmannian ( $Gr_{r,n}$ ) as a giant dance floor where many different groups of dancers (mathematical points) can move.

The Dancers: These are points representing specific subspaces (like flat sheets floating in space).
The Music (The Torus): There is a group of musicians called a Torus ( $T$ ) playing a specific rhythm. This music dictates how the dancers can move.
The "Good" Dancers (Semistable Points): Not every dancer can keep up with the music. Some dancers get dizzy and fall off the floor. The authors are interested in the specific group of dancers who can stay on the floor and dance in harmony with the music. They call this the "semistable locus."

The Problem: Finding the Smallest Room

In a previous study, the authors found that among all the possible rooms (Schubert varieties) on this dance floor, there is exactly one smallest room that still contains enough "good dancers" to make a meaningful performance. Let's call this the Minimal Room ( $X(w_{r,n})$ ).

The paper asks: What does this specific room look like if we view it through a special lens?

The Lens: A Subgroup of Musicians

The authors decide to zoom in. Instead of listening to the whole orchestra (the full Torus $T$ ), they only listen to a smaller section of the musicians, a Subtorus ( $T_{J_r}$ ).

Imagine the orchestra has many instruments. The authors pick out a specific set of "peak" instruments (the peaks of the mathematical pattern) and ignore the rest.
They ask: If we only care about the rhythm played by these specific instruments, what does the "Minimal Room" look like?

The Discovery: A Tower of Projective Spaces

The main result of the paper is a beautiful description of what happens when you take this Minimal Room and "quotient" it (essentially, you squash the room down by identifying points that move together under the rhythm of the sub-musicians).

They found that the resulting shape is not a messy blob. Instead, it is a perfectly structured tower.

Here is the analogy:

The Base: The bottom of the tower is a simple shape called a projective space (think of it as a generalized version of a sphere or a flat plane with extra dimensions).
The Layers: On top of that base, they build a second layer. This layer is a "bundle" of projective spaces. Imagine taking the base and attaching a stack of smaller projective spaces to every single point on it.
The Stack: They keep doing this. They build a third layer on the second, a fourth on the third, and so on, up to the $r$ -th layer.

The authors prove that the final shape is an iterated projective space bundle.

Simple Metaphor: Imagine a Russian nesting doll, but instead of dolls inside dolls, you have a "bundle of shapes" sitting on top of another "bundle of shapes."
The Structure: They show that this tower is built in stages. The $k$ -th stage is built by attaching a specific bundle of projective spaces to the $(k-1)$ -th stage.
The Result: The final object is a "Generalized Bott Tower." It's a very specific, clean, and predictable mathematical structure, rather than a chaotic mess.

The "How" (The Mechanics)

To prove this, the authors had to do some heavy lifting:

Mapping the Floor: They figured out exactly which points in the Minimal Room are "stable" (safe) under the sub-musicians' rhythm. They found that a point is safe if certain specific coordinates (numbers in a matrix) are not zero.
The Projection: They created a map that takes a point from the big room ( $Y_r$ ) and projects it down to the smaller room ( $Y_{r-1}$ ).
The Bundle: They proved that for every point in the smaller room, the collection of points in the big room that map to it forms a perfect projective space (like a higher-dimensional version of a circle or sphere).
Splitting the Bundle: They showed that this bundle of shapes is actually made of simpler, straight-line bundles (line bundles) stuck together. This is like showing that a complex rope is actually just several simple strings woven together.

The Conclusion

The paper concludes that the complex geometric object formed by this specific Schubert variety, when viewed through the lens of this specific subgroup, is actually a Generalized Bott Tower.

In everyday terms:
If you take a very specific, minimal geometric shape defined by complex rules, and you "filter" it through a specific set of symmetries, the result isn't a random shape. It turns out to be a beautifully organized, multi-story building where each floor is a bundle of projective spaces stacked neatly on top of the one below it.

The authors didn't just say "it's a tower"; they gave the exact blueprint (the vector bundle $E$ ) showing exactly how the floors are connected and what materials (line bundles) they are made of. This transforms a complicated algebraic problem into a clear, structural description.

1. Problem Statement and Context

Mathematical Context:
The paper operates within the realm of Geometric Invariant Theory (GIT) applied to algebraic groups, specifically $G = PSL(n, \mathbb{C})$ . It focuses on the action of a maximal torus $T$ (and its subgroups) on the Grassmannian $Gr, n$ and its Schubert varieties.

The Core Problem:
Previous work (specifically [13]) established that for the line bundle $L(n\omega_r)$ on the Grassmannian $Gr, n$ (where $n = rq + 1$), there exists a unique minimal dimensional Schubert variety, denoted $X(w_{r,n})$ , that admits semistable points under the action of $T$ .
The authors investigate the structure of the GIT quotient of this specific variety $X(w_{r,n})$ when the acting group is reduced from the full maximal torus $T$ to a specific subtorus $T_{J_r}$ . This subtorus is generated by one-parameter subgroups corresponding to the "peaks" of the Weyl group element $w_{r,n}$ .

Key Question:
What is the geometric structure of the quotient space $Y_r := T_{J_r} \backslash\backslash X(w_{r,n})^{ss}_{T_{J_r}}(L(n\omega_r))$ ? Specifically, does it possess a recursive or bundle-like structure?

2. Methodology

The authors employ a combination of combinatorial group theory, standard monomial theory, and explicit construction of morphisms.

Combinatorial Description of $w_{r,n}$ : They utilize the reduced expression of the Weyl group element $w_{r,n}$ :
$w_{r,n} = (s_q \cdots s_1)(s_{2q} \cdots s_2) \cdots (s_{rq} \cdots s_r)$
They identify the "peaks" of this element as the simple roots $\{\alpha_{iq} \mid 1 \le i \le r\}$ . The subtorus $T_{J_r}$ is generated by the one-parameter subgroups $\lambda_{iq}(\mathbb{G}_m)$ .
Characterization of the Semistable Locus:
Using the Hilbert-Mumford criterion and properties of the coordinate ring of the Schubert cell, they explicitly describe the semistable locus $X(w_{r,n})^{ss}_{T_{J_r}}$ . They prove that a point is semistable if and only if specific coordinate functions (associated with the peaks) are non-zero. Crucially, they show that for this specific setup, semistability coincides with stability, and the stabilizer of any semistable point in $T_{J_r}$ is trivial. This ensures the quotient is a geometric quotient and is smooth.
Recursive Construction (The "Peeling" Process):
The authors define a sequence of varieties $Y_k$ for $1 \le k \le r$ , where $Y_k$ is the GIT quotient of the minimal Schubert variety in $G_{k, kq+1}$ modulo the corresponding subtorus.
They construct a morphism $\phi_r: Y_r \to Y_{r-1}$ by projecting the matrix coordinates of the Schubert cell of $X(w_{r,n})$ to those of $X(w_{r-1, n-1})$ .
Bundle Structure Analysis:
To determine the structure of the fibers of $\phi_r$ , they construct an open cover of $Y_{r-1}$ and define transition functions. They analyze the action of the torus on the coordinates to show that the transition functions are diagonal matrices. This allows them to prove that the vector bundle associated with the projective bundle splits into a direct sum of line bundles.

3. Key Contributions and Results

The paper establishes three main theorems describing the geometry of the quotient:

Theorem 1: Bundle Structure
The quotient $Y_r$ is a projective space bundle over $Y_{r-1}$ . Specifically, there exists a morphism $\phi_r: Y_r \to Y_{r-1}$ such that $Y_r$ is isomorphic to $\mathbb{P}(E)$ , where $E$ is a vector bundle of rank $r(q-1) + 1$ over $Y_{r-1}$ .

Theorem 2: Splitting of the Vector Bundle
The vector bundle $E$ is not just any bundle; it splits into a direct sum of line bundles. Specifically:
$E \cong L_0^{\oplus q} \oplus \left( \bigoplus_{j=1}^{r-1} L_j^{\oplus (q-1)} \right)$
where $L_0, \dots, L_{r-1}$ are line bundles on $Y_{r-1}$ , and $L_0$ is the trivial line bundle. This splitting is derived from the fact that the transition functions lie in the maximal torus of $GL(r(q-1)+1, \mathbb{C})$ .

Theorem 3: Generalized Bott Tower
The sequence of quotients forms a Generalized Bott Tower of height $r$ :
$Y_r \xrightarrow{\phi_r} Y_{r-1} \xrightarrow{\phi_{r-1}} \cdots \xrightarrow{\phi_2} Y_1 \xrightarrow{\phi_1} Y_0 = \{pt\}$
where $Y_1 \cong \mathbb{P}^{q-1}$ .
This means $Y_r$ is constructed iteratively by taking projective bundles over the previous stage.

4. Significance and Implications

Geometric Classification: The paper provides a complete geometric classification of the GIT quotients of minimal dimensional Schubert varieties under specific subtorus actions. It reveals that these quotients are not arbitrary varieties but possess a highly structured, recursive nature (Generalized Bott towers).
Connection to Polytopes and Toric Geometry: Generalized Bott towers are significant in toric geometry and the study of polytopes (specifically, they correspond to certain generalized permutahedra or related structures). This result links the combinatorics of Weyl groups and Schubert calculus directly to the topology of these iterated bundles.
Smoothness and Stability: The proof that the semistable locus coincides with the stable locus and that the quotient is smooth is a strong result. It implies that the moduli space is well-behaved (no singularities arising from non-free group actions), which is not always guaranteed in GIT quotients.
Generalization: The methodology extends previous results on $G_{2,n}$ (related to polygon moduli spaces) and $G_{3,7}$ (rational normal scrolls) to the general case of $Gr, n$ with $n = rq+1$.

5. Conclusion

Ghosh and Kannan successfully demonstrate that the GIT quotient of the minimal dimensional Schubert variety $X(w_{r,n})$ modulo the subtorus $T_{J_r}$ is a Generalized Bott Tower. By explicitly constructing the morphisms and analyzing the transition functions, they prove that the total space is an iterated projective bundle over $\mathbb{P}^{q-1}$ , with the vector bundles splitting into direct sums of line bundles. This work bridges the gap between the combinatorial description of Schubert varieties and the explicit geometric structure of their moduli spaces.