1: \begin{abstract}

2: 

3: Let $V$ be an $n$ dimensional vector space over the finite field $\FF_q$.  

4: We study the random walk on $GL_n(V)$ generated by repeated random transvections.

5: Any matrix $M\in GL_n(V)$ can be redueced to $LU$ form by Gaussian elimination:

6: $M= LwU$, with $L$ lower triangular, $w$ a permutation matrix (representing the `pivots'

7: in the elimination process) and $U$ upper triangular.  A random

8: walk on $GL_n$ induces a walk on permutations.  This walk has the Mallows distribution

9: $\pi(w) = z^{-1}q^{-I(w)}$ as stationary distribution.

10: 

11: Our first contribution is a description of ``how the $S_n$ walk goes'' in terms of the Hecke algebra

12: $B\backslash GL_n/B$.  This is indexed by permutations with generators

13: $\{T_w\}_{w\in S_n}$.  We show that the walk is equivalent to repeated multiplication by 

14: $$D = c_{nq}I+ (q-1) \sum_{i<j} q^{n-1-(j-i)}T_{(ij)},

15: \qquad\hbox{where}\quad

16: c_{nq} = (n-1)q^{n-1}-[n-1].

17: $$

18: 

19: The transvections walk on $GL_n$ is studied by Hildebrand \cite{Hil} who shows that $n+c$ steps are 

20: necessary and sufficient for convergence to stationarity.  The walk on $S_n$ is shown to mix faster.

21: 

22: The mathematical details involve symmetric function theory, the character theory of $GL_n$

23: and the Hecke algebra.  The results give new complexity bounds for basic linear algebra,

24: new combinatorial results and a substantial illustration of the `mapping to double coset theory'

25: in \cite{Sim}

26: \end{abstract}

27: