The censored stochastic six-vertex model and parabolic Kazhdan--Lusztig $R$-polynomials

This paper introduces a censored stochastic six-vertex model, demonstrating that its blocking measure stochastically dominates the system at all times to control second-class particles, a result established through connections to Iwahori--Hecke algebras and the use of parabolic Kazhdan--Lusztig $R$-polynomials as both explanatory tools and intertwining kernels.

The Gist

Imagine a vast, infinite grid representing a city where time flows diagonally. On this grid, we have a system of "particles" (think of them as people) and "holes" (empty spaces). These people move according to a specific set of rules: they can walk straight or turn corners, but they can never pass through each other. This is the Stochastic Six-Vertex Model. It's a mathematical way to describe how crowds, traffic, or fluids behave when they are crowded and moving in one direction.

In this paper, the authors introduce a special version of this model called the "Censored" model.

The "Censorship" Analogy

Imagine you are watching a movie of this crowd moving. In the normal movie, people can sometimes walk straight past a hole, or a hole can slide past a person.

In the Censored version, the director (the mathematician) decides to "censor" certain scenes. At specific intersections (vertices) on the grid, the director says: "No, you cannot go straight! You must turn!"

• If a person tries to walk straight, the rule forces them to turn.
• If a hole tries to slide straight, it must turn.

The authors ask a big question: If we force people to turn more often than usual, does the crowd become more chaotic, or does it stay under control?

The Main Discovery: The "Traffic Jam" Limit

The authors prove a surprising result: Even with these extra rules forcing turns, the crowd never gets "worse" than a specific, well-organized state called the "Blocking Measure."

Think of the "Blocking Measure" as the ultimate traffic jam. It's a state where the people are packed as tightly as possible in a specific pattern, and the holes are packed on the other side. It's the most "ordered" chaos possible for this system.

The paper shows that no matter how you censor the rules (forcing turns at random spots), if you start with an empty street on the left and a full street on the right, the crowd will always stay "below" or "less chaotic" than this ultimate traffic jam. They can never exceed this limit.

Why is this hard?

Usually, in math, if you add restrictions (like forcing turns), you expect the system to behave more predictably. However, this specific model is tricky. It lacks a simple "monotonicity" property (a fancy word meaning "if you push it one way, it always moves that way"). Because of this, standard math tools don't work.

To solve this, the authors had to use a very advanced, abstract tool from a different branch of mathematics called Kazhdan–Lusztig R-polynomials.

The Secret Weapon: The "Mathematical Translator"

The authors discovered that this crowd-moving problem is secretly connected to something called Hecke Algebras (a type of algebra used to study symmetries).

• The Analogy: Imagine the crowd movement is a song in a foreign language. The authors found a "translator" (the Kazhdan–Lusztig polynomials) that translates the song into a language they understand.
• In this translated language, the "censored" rules correspond to specific mathematical shapes called partitions (like stacking blocks in a pyramid).
• They proved that these translated shapes always fit inside a specific "box" (the Blocking Measure). Because the translation is accurate, this means the original crowd also stays inside its box.

What about "Second-Class Particles"?

The paper also mentions a practical use for this result: controlling "Second-Class Particles."

• Imagine a VIP line where some people are "First Class" (VIPs), some are "Second Class" (regular people), and some are "Third Class" (people with no ticket).
• The authors show that by using their "censorship" trick, they can predict exactly how the "Second Class" people will behave relative to the "Third Class" people, even if the VIPs are moving around chaotically. They can prove that the Second Class people won't get pushed too far out of line.

Summary

1. The Setup: A model of particles moving on a grid.
2. The Twist: The authors "censor" the model, forcing particles to turn instead of going straight at certain points.
3. The Result: Even with these forced turns, the system never becomes more chaotic than a specific, known "maximum jam" state.
4. The Method: They used a complex mathematical "translator" (Kazhdan–Lusztig polynomials) to turn the particle problem into a shape problem, where the solution was obvious.
5. The Application: This helps predict the behavior of different types of particles (classes) moving together in a crowd.

In short, the paper proves that even if you force a chaotic crowd to take detours, they will never break the rules of the "ultimate traffic jam."

Technical Summary

Technical Summary: The Censored Stochastic Six-Vertex Model and Parabolic Kazhdan–Lusztig R-Polynomials

Problem Statement
The stochastic six-vertex model is a fundamental interacting particle system within the KPZ universality class, connected to the asymmetric simple exclusion process (ASEP) and the KPZ equation. While monotonicity properties are well-understood for exclusion processes like ASEP, the stochastic six-vertex model presents subtler challenges. Specifically, the model lacks the standard monotonicity required to apply classical "censoring inequalities" (which state that censoring updates increases the distance to stationarity).

The authors investigate a censored version of the stochastic six-vertex model, where at a specific set of vertices $C$, particles and holes are forbidden from passing each other (configurations III and V in the standard six-vertex weight table are disallowed). The central question is whether the law of this censored process, starting from the minimal configuration (particles to the right of the origin), remains stochastically dominated by the stationary "blocking measure" at any fixed time $t$.

Methodology
The paper employs a novel algebraic and probabilistic framework linking the stochastic six-vertex model to Iwahori–Hecke algebras of symmetric groups.

1. Operator Formalism: The dynamics of the censored process are represented as a product of propagator operators $P_k$. These operators act on particle configurations by swapping adjacent values with probabilities determined by parameters $b_1$ and $b_2$. The censored process corresponds to a specific sequence of these operators where some are replaced by identity maps.
2. Parabolic Kazhdan–Lusztig R-measures: The core technical innovation is the introduction of a family of probability measures, denoted $v_\lambda$, indexed by integer partitions $\lambda$. These measures are defined combinatorially using "rim decompositions" of Young diagrams and are shown to be normalized versions of parabolic Kazhdan–Lusztig R-polynomials.
3. Convex Decomposition: The authors prove that the action of the propagator operators $P_k$ on the measures $v_\lambda$ results in a convex combination of $v_\lambda$ and $v_{\lambda_k}$ (where $\lambda_k$ is a partition obtained by flipping a corner in the $k$-th diagonal). This establishes that the law of the censored process at any time $t$ can be expressed as a convex combination of these $v_\lambda$ measures.
4. Stochastic Domination: The proof relies on demonstrating that for any partition $\lambda$, the measure $v_\lambda$ is stochastically dominated by the blocking measure $\pi$. This is achieved by showing that as partitions grow, $v_\lambda$ converges to $\pi$, and utilizing the monotonicity of the measures with respect to the containment ordering of partitions.

Key Contributions and Results

• Theorem 1.8 (Main Result): For parameters $0 < b_1 < b_2 < 1$ and any set of censored vertices $C$, the censored stochastic six-vertex process started from the minimal configuration $\mathbb{1}_{x>0}$ is stochastically dominated by the blocking measure $\pi$ at any time $t$.
  • Significance: This serves as a "partial censoring inequality." Unlike the full censoring inequality for monotone spin systems (where censoring increases the distance to stationarity), the authors note that the first part of that inequality (distance increase) does not hold here. However, the second part (domination by the stationary measure) is preserved.
• Intertwining Relation (Theorem 4.1): The proof reveals an intertwining relation between the censored process with parameters $(b_1, b_2)$ and a process with swapped parameters $(b_2, b_1)$. Specifically, the law of the process with parameters $(b_1, b_2)$ can be expressed as an expectation of the measures $v_\lambda$ under the law of the process with swapped parameters.
• Control of Multi-Class Particles (Theorem 4.7): The main result is applied to the multi-class (colored) stochastic six-vertex model. By censoring vertices corresponding to first-class particles and holes, the authors derive a stochastic domination result for the behavior of second-class particles relative to third-class particles. This allows for the control of higher-class particles starting from non-stationary initial conditions.
• Algebraic Connection (Section 5): The paper establishes a rigorous connection between the probabilistic measures $v_\lambda$ and parabolic Kazhdan–Lusztig R-polynomials (defined by Deodhar). The authors show that $v_\lambda$ corresponds to the projection of these polynomials onto a parabolic quotient of the symmetric group.
• Counterexample to General Monotonicity (Example 5.11): The authors demonstrate that the stochastic monotonicity observed for the parabolic measures $v_\lambda$ does not lift to the full Kazhdan–Lusztig R-polynomials on the symmetric group. Specifically, they show that for $S_3$, the measure associated with a larger element in the Bruhat order does not necessarily stochastically dominate the measure of a smaller element.

Significance and Claims
The paper claims to provide a new avenue for interpreting properties of Kazhdan–Lusztig R-polynomials probabilistically. By identifying the evolution of the censored six-vertex model with a random walk on the Hecke algebra, the authors bridge the gap between integrable probability and representation theory.

The primary significance lies in extending domination results to a model that lacks the standard monotonicity required for traditional coupling arguments. This result is particularly motivated by the need to control the behavior of second-class particles in multi-class systems, a problem for which such domination techniques were previously unavailable in the context of the six-vertex model. The authors emphasize that while the full censoring inequality (regarding distance to stationarity) fails for this model, the domination by the stationary measure holds, offering a robust tool for analyzing the system's long-term behavior and fluctuations.