Random walks on linear groups satisfying a Schubert condition

Weikun He

TL;DR

This paper relaxes a key assumption in the theory of random walks on linear groups, showing that a "proximality" condition can be replaced by a weaker "Schubert condition." This generalization allows the theory to apply to a broader class of random walks, with implications for understanding their long-term behavior and the distribution of their stationary measures.

What Problem Does This Paper Address?

The study of random walks on linear groups, particularly their asymptotic behavior and stationary measures, has been a rich area of research. A central concept in this field is "proximality," a condition that essentially means the random walk "contracts" distances in a certain way. This condition has been crucial for proving many fundamental results, such as the existence and properties of stationary measures on spaces like the Grassmannian. However, proximality is a strong condition, and researchers have long suspected that many of these results might hold under weaker assumptions. This paper tackles the question of whether the proximality assumption can be relaxed, specifically by replacing it with a "Schubert condition," which relates to how the limit set of the random walk interacts with certain geometric structures (Schubert varieties) on the Grassmannian.

Methods

The paper introduces and studies a "Schubert condition" (property (S)) for subsemigroups of linear groups. The core of the paper involves:

  • Defining and characterizing property (S): The authors formally define property (S) and establish its equivalence with conditions on the limit set of the group's action on the Grassmannian. They also show that this property is preserved under taking the Zariski closure of the group.
  • Large Deviation Estimates: A key technical tool is a large deviation inequality (Proposition 3.1) that quantifies how unlikely it is for a random walk on the Grassmannian to fall into a proper Schubert variety, given property (S). This is a generalization of existing results for proximal groups.
  • Discretized Projection Theorem: The paper leverages a generalized discretized projection theorem (Theorem 4.2) which is crucial for extending results from proximal groups to groups satisfying property (S). This theorem provides a way to "discretize" projections and is a significant technical component.
  • Adaptation of Existing Proofs: The authors adapt the proof of the Bourgain-Furman-Lindenstrauss-Mozes theorem (Theorem 1.3) to show that it holds when the proximality assumption is replaced by the Schubert condition, provided the proximal dimension divides the dimension of the ambient space. This involves carefully re-examining where proximality is used in the original proof and showing that property (S) suffices.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Schubert Condition as a Viable Alternative to Proximality: The most significant insight is demonstrating that the geometric "Schubert condition" is a powerful enough substitute for the analytical "proximality" condition in many contexts of random walks on linear groups. This broadens the applicability of existing theories.
  • Grassmannian as a Key Space: The paper highlights the importance of the Grassmannian (the space of linear subspaces) as a natural space to study the behavior of random walks on linear groups, especially when dealing with geometric conditions.
  • Connection Between Group Theory and Probability: The work beautifully illustrates the interplay between algebraic group theory (Zariski closure, irreducible representations) and probability theory (random walks, Lyapunov exponents, large deviations).
  • Generalization of a Major Theorem: The paper successfully generalizes a significant result (Bourgain-Furman-Lindenstrauss-Mozes theorem) concerning random walks on tori, showing it holds under a weaker assumption.

Key Results

  • Theorem 1.5: The main result states that the Bourgain-Furman-Lindenstrauss-Mozes theorem holds when the proximality assumption on the group is replaced by the Schubert condition (property (S)), provided the proximal dimension divides the dimension of the ambient space.
  • Proposition 3.1: A key large deviation estimate is proven, showing that random walks on the Grassmannian satisfying property (S) avoid proper Schubert varieties with exponentially decaying probability.
  • Examples and Non-Examples: The paper provides examples of groups satisfying property (S) (e.g., certain groups arising from restriction of scalars) and non-examples (e.g., SO(1,d) acting on a specific space), illustrating the scope of the condition.
  • Uniqueness of Stationary Measures: Under property (S) and strong irreducibility, the paper proves the uniqueness of a stationary Borel probability measure on the Grassmannian (Proposition 3.4).

Limitations and Future Work

  • Technical Condition on Proximal Dimension: The generalization of Theorem 1.3 requires the proximal dimension $r$ to divide the dimension $d$ of the ambient space. The authors note that removing this assumption would significantly lengthen the paper without adding conceptual novelty, suggesting it's a technical convenience rather than a fundamental barrier.
  • Optimality of Discretized Projection Theorem: The paper mentions that the discretized projection theorem used is "far from being optimal." This suggests that proving a stronger version could potentially cover an even larger class of non-proximal groups, pointing towards future research directions.
  • Broader Applicability: While the paper successfully relaxes the proximality condition, future work could explore whether property (S) is the "best possible" weakening or if other geometric conditions could also suffice.

Summary updated

Abstract

We study random walks on $\mathrm{GL}_d(\mathbb{R})$ whose proximal dimension $r$ is larger than $1$ and whose limit set in the Grassmannian $\mathrm{Gr}_{r,d}(\mathbb{R})$ is not contained any Schubert variety. These random walks, without being proximal, behave in many ways like proximal ones. Among other results, we establish a Hölder-type regularity for the stationary measure on the Grassmannian associated to these random walks. Using this and a generalization of Bourgain's discretized projection theorem, we prove that the proximality assumption in the Bourgain-Furman-Lindenstrauss-Mozes theorem can be relaxed to this Schubert condition.
BibTeX
@article{he2020random,
  author = {Weikun He},
  title = {{Random walks on linear groups satisfying a Schubert condition}},
  journal = {Israel Journal of Mathematics},
  year = {2020},
  doi = {10.1007/s11856-020-2032-x},
}
Figure