Random Lie bracket on 𝔰 ​ 𝔩 2 ​ ( 𝐅 p ) \mathfrak{sl}_{2}(\mathbf{F}_{p})

Urban Jezernik, Matevž Miščič · 2025

Read Random Lie bracket on 𝔰 ​ 𝔩 2 ​ ( 𝐅 p ) \mathfrak{sl}_{2}(\mathbf{F}_{p}) by Urban Jezernik, Matevž Miščič (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-16175.txt.

Introduction

Urban Jezernik and Matevž Miščič study a random walk inside a small Lie algebra in this 2025 paper. New elements arise by applying moves from two chosen generators at random. The algebra holds two-by-two matrices with zero trace over a finite field. The authors ask where the walk settles and how fast it gets there.

The problem reduces to a random walk on a cyclic group, which is easier to track. With high chance, the walk shows a sharp pre-cutoff after about p steps. Its resting pattern need not be even; it depends on the prime factors of p minus one. A small random twist added to the walk speeds full coverage greatly. With the twist, the whole algebra is covered in roughly log p steps in typical cases.

Catalog record

CREATORUrban Jezernik, Matevž Miščič
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-16175

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Random Lie bracket on 𝔰 ​ 𝔩 2 ​ ( 𝐅 p ) \mathfrak{sl}_{2}(\mathbf{F}_{p}) Thanks: UJ is supported by the Slovenian Research and Innovation Agency program P1-0222 and grants J1-50001, J1-4351, J1-3004, N1-0217. MM is supported by the Slovenian Research and Innovation Agency program P1-0222 and by the Institute of Mathematics, Physics, and Mechanics. Urban Jezernik, Matevž Miščič Abstract . We study a random walk on the Lie algebra 𝔰 ​ 𝔩 2 ⁡ ( 𝐅 p ) \slfrak_{2}(\mathbf{F}_{p}) where new elements are produced by randomly applying adjoint operators of two generators. Focusing on the generic case where the generators are selected at random, we analyze the limiting distribution of the random walk and the speed at which it converges to this distribution. These questions reduce to the study of a random walk on a cyclic group. We show that, with high probability, the walk exhibits a pre-cutoff phenomenon after roughly p p steps. Notably, the limiting distribution need not be uniform and it depends on the prime divisors of p − 1 p-1 . Furthermore, we prove that by incorporating a simple random twist into the walk, we can embed a well-known affine random walk on 𝐅 p \mathbf{F}_{p} into the modified random Lie bracket, allowing us to show that the entire Lie algebra is covered in roughly log ⁡ p \log p steps in the generic case. 1. Introduction Random walks on finite groups are a powerful tool for analyzing their structure and properties (see [ Dia88 , Hil05 ] and references therein). Of particular significance is their connection to growth and expansion in Cayley graphs of finite simple groups (see, for example, [ BG08 ] ). However, extending these techniques, and ultimately the results, to the closely related setting of finite simple Lie algebras is not straightforward. Unlike groups, Lie algebras have two operations: addition and the Lie bracket. The fact that the Lie bracket is non-invertible introduces additional challenges. In this paper, we introduce and study a random walk on the Lie algebra \slfrak_{2}(\mathbf{F}_{p}) that focuses solely on the Lie bracket structure. This walk is defined as follows. Let A and B be a generating pair of the Lie algebra \slfrak_{2}(\mathbf{F}_{p}) with p\geq 3 . Let Z_{k} , for k\in\mathbf{N} , be independent random variables uniformly distributed in the set \{A,B\} . Starting with X_{0}=[A,B] , the random walk evolves according to X_{k}=[Z_{k},X_{k-1}]\qquad(k\in\mathbf{N}). This process generates a sequence (X_{k})_{k\in\mathbf{N}_{0}} that forms a Markov chain on the set \slfrak_{2}(\mathbf{F}_{p}) . New elements are produced by iteratively applying the adjoint operators associated with the generators, a process we refer to as the random Lie bracket . Example . Let us look at a simulation of the random Lie bracket. Suppose p=11,\quad A=\begin{pmatrix}7&5\\ 2&4\end{pmatrix},\quad B=\begin{pmatrix}8&8\\ 10&3\end{pmatrix},\quad X_{0}=[A,B]=\begin{pmatrix}1&10\\ 2&10\\ \end{pmatrix}. Randomly sample the sequence (B,A,B,B,B) from \{A,B\} to obtain the values X_{1}=\begin{pmatrix}4&1\\ 10&7\\ \end{pmatrix},\ X_{2}=\begin{pmatrix}4&7\\ 8&7\\ \end{pmatrix},\ X_{3}=\begin{pmatrix}5&4\\ 7&6\\ \end{pmatrix},\ X_{4}=\begin{pmatrix}5&6\\ 10&6\\ \end{pmatrix},\ X_{5}=\begin{pmatrix}9&5\\ 6&2\\ \end{pmatrix}. Observe that X_{2}=4X_{0} , X_{4}=5X_{0} . Our goal is to analyze the limiting distribution of this random walk and to study the rate at which it converges to that distribution, particularly when the elements A and B are chosen uniformly at random. We prove the following somewhat surprising results. The random walk (X_{k})_{k\in\mathbf{N}_{0}} exhibits different behavior depending on whether the number of steps is even or odd (as hinted at in the example above). In fact, as we will explain, it suffices to focus on what happens after an even number of steps. We first show that the distribution of X_{2k} is supported on the line \lin\{[A,B]\} and converges (in total variation distance 1 1 1 The total variation distance between probability measures \mu and \nu is \dist(\mu,\nu)=\frac{1}{2}\mathinner{\!\left\lVert\mu-\nu\right\rVert}_{1} . ) to a distribution \nu_{2k} on this line in a pre-cutoff manner at k\approx p . Theorem 1.1 . For every \delta,\epsilon>0 , there are constants C,c>0 such that the following holds. Let p>C , let A,B\in\slfrak_{2}(\mathbf{F}_{p}) be uniformly random, and let (X_{k})_{k\in\mathbf{N}_{0}} be the associated random Lie bracket. Let \mu_{2k} be the distribution of X_{2k} . There is a distribution \nu_{2k} that is uniform on a coset of a subgroup of \mathbf{F}_{p}^{\ast}=\lin\{[A,B]\}\setminus\{0\} such that with probability at least 1-\delta : 1. For k<cp , we have \dist(\mu_{2k},\nu_{2k})>1-\epsilon . 2. For k>Cp , we have \dist(\mu_{2k},\nu_{2k})<\epsilon . After 2k+1 of steps, the distribution \nu_{2k} equally splits into two distributions, one supported on the line \lin\{[A,[A,B]]\} and the other on the line \lin\{[B,[A,B]]\} . Example . Consider the example above. Take k=50 and generate 1000 independent samples of the random Lie bracket X_{100} . The following table shows the sample distribution: \begin{pmatrix}[A,B]&5[A,B]&4[A,B]&3[A,B]&9[A,B]\\ 189&185&193&209&224\end{pmatrix} In this case, the distribution \nu_{100} (indeed, any \nu_{2k} with k>0 ) is uniform over the five listed multiples of [A,B] . The total variation distance between the sample distribution and \nu_{100} is 0.033 . For randomly chosen A,B , the distribution \nu_{2k} is uniform on the whole of \lin\{[A,B]\}\setminus\{0\} with a certain probability that depends on the prime divisors of p-1 . Theorem 1.2 . Let (s,t) be a subinterval of \mathcal{I}=\left[\frac{1}{\zeta(2)},\frac{2}{3}\right]\cup\left[\frac{9}{8\zeta(2)},\frac{3}{4}\right]\approx[0.6079,0.6667]\cup[0.6839,0.7500]. Then there is a set of primes p of positive lower density with the following property. Let A,B be uniformly random in \slfrak_{2}(\mathbf{F}_{p}) , and let \nu be the uniform distribution on \lin\{[A,B]\}\setminus\{0\} . Then \mathbf{P}_{A,B}(\nu_{2k}=\nu)\in(s,t) for all k\geq 0 . 2 2 2 The probability \mathbf{P}_{A,B}(\nu_{2k}=\nu) is independent of k , as \nu_{2k} is supported on cosets of the same subgroup of \lin\{[A,B]\}\setminus\{0\} . Let \nu(p) denote this probability. The theorem and its proof imply that the set of limit points of the sequence \nu(p) as p varies is exactly \mathcal{I} . Based on our inspection of the random Lie bracket on \slfrak_{2}(\mathbf{F}_{p}) driven by random generators A and B , we can prove an additional result in this setting. Although the random walk requires roughly p steps to approach its limiting distribution, the two elements A and B generate the entire Lie algebra \slfrak_{2}(\mathbf{F}_{p}) much more quickly. In fact, we will show that, with high probability, every element of \slfrak_{2}(\mathbf{F}_{p}) can be reached after only O(\log p) steps. We measure steps in a weighted way using the notion of the diameter of a Lie algebra as defined in [ Don23 ] . Let S be a subset of a Lie algebra \gfrak . The set S is called symmetric if 0\in S and S=-S . For a symmetric set S , we inductively define the set of weighted k -balls as S^{1}=S,\qquad S^{k}=\bigcup_{0<j<k}\left((S^{j}+S^{k-j})\cup[S^{j},S^{k-j}]\right)\quad(k\geq 2), where we have denoted X+Y=\{x+y\mid x\in X,y\in Y\} and [X,Y]=\{[x,y]\mid x\in X,y\in Y\} for subsets X,Y\subseteq\gfrak . If S is a symmetric generating set of a Lie algebra \gfrak over \mathbf{F}_{p} , then the diameter \diam(\gfrak,S) of \gfrak with respect to S is the smallest k such that S^{k}=\gfrak . For any classical Lie algebra \gfrak over \mathbf{F}_{p} , Dona proves in [ Don23 ] that there is a constant C=O(\dim(\gfrak)^{2}\log\dim(\gfrak)) such that \diam(\gfrak,S)\leq\left(\log\mathinner{\!\left\lvert\gfrak\right\rvert}\right)^{C} for any symmetric generating set S of \gfrak . For the case of \slfrak_{2}(\mathbf{F}_{p}) , this gives a bound of the form (\log p)^{O(1)} . We will now improve this bound to linear in \log p for the case when the generating set is chosen uniformly at random. Theorem 1.3

Read the complete text