Monotonicity of Lyapunov exponents

Let $A$ be a square matrix with non-negative entries, such that the entries of $A^TA$ are positive.
A nice example to start with is

$A=\begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix} \,.$

Given $p \in (0,1)$ , consider a random matrix product

$Y_n=M_1\cdots M_n$

where the $M_k$ are independent, and
for each $k$ ,

$\mathbb P_p(M_k=A)=p \quad \text{and} \quad \mathbb P_p(M_k=A^T)=1-p \,.$

Fix a
submultiplicative norm $\| \cdot\|$ on $d \times d$ matrices, e.g. the operator norm

$\|M\|=\sup_{\|x\|_2=1} \|Mx\|_2 \,,$

where the supremum is over vectors $x$ in the unit sphere of
${\mathbb R}^d$ . The sequence $E_p(\|Y_n\|)$ is submultiplicative, i.e.

$\forall m,n \quad E_p(\|Y_{m+n}\|) \le E_p(\|Y_m\|) \cdot E_p(\|Y_n\|)\,,$

so by Fekete’s lemma, the limit

$L(p):=\lim_{n \to \infty} \frac1n\log E_p(\|Y_n\|)$

exists; it is called the top Lyapunov exponent for this matrix product, and was
first studied in [1]. In [2], it was shown that $L(p)$ is a real analytic function of $p$ .

(a) Does $L(p)$ attain its maximum at $p=1/2$ ?
(b) Is $L(p)$ nondecreasing in $(0,1/2]$ ?

[1] Furstenberg, Harry, and Harry Kesten. "Products of random
matrices." The Annals of Mathematical Statistics 31, no. 2 (1960):
457-469.

[2] Peres, Yuval Analytic dependence of Lyapunov exponents on
transition probabilities. Lyapunov exponents (Oberwolfach, 1990), 64–80,
Lecture Notes in Math., 1486, Springer, Berlin, 1991.

Leave a Reply Cancel reply