Definition 3. The measure µ has spectral gap if ||\lambdaG (µ)|L20 (G) ||op < 1.
Statement
Definition 3. The measure µ has spectral gap if ||\lambdaG (µ)|L20 (G) ||op < 1. Conjecture 4. (Spectral Gap Conjecture) Let µ be a symmetric probability mea- sure on a compact connected simple Lie group. Then µ has spectral gap if and only if µ is non-degenerate. As argued above, spectral gap implies equidistribution of µ∗n and hence that µ is non-degenerate. Equivalently, if the support of µ is trapped in a closed subgroup of G, then µ cannot have spectral gap. The Spectral Gap Conjecture claims that this is the only obstruction to µ having a spectral gap. In particular, in view of the discussion around (1), the conjecture claims that spectral gap is a generic condition. If µ has spectral gap, using (4) one can deduce equidistribution of µ∗n with exponential speed. Namely there is c > 0 such that for f \in C \infty (G), Z Z f dµ∗n = f dmG + O(S(f )e−cn ), Lemma 2. Let µ be a symmetric non-degenerate probability measure on a compact b be an irreducible non-trivial representation. Then group G and let \pi \in G ||\pi(µ)||op < 1. Proof. As \pi is finite dimensional and µ symmetric, the operator \pi(µ) is diago- nalizable. Furthermore µ is a probability measure, implying ||\pi(µ)||op \leq 1. It thus suffices to show that \pi(µ) R has no eigenvalue of modulus one. Assume for a contradiction that \pi(µ)v = \pig v dµ(g) = \lambdav for a non-zero element v \in H\pi and \lambda \in S1 . As \pi is finite-dimensional and ||\pig v|| = ||\lambdav|| = ||v||, it must hold that \pig v = \lambdav for almost all g \in supp(µ). Now consider the subgroup H = {g \in G : \pig v = v} and fix an element g0 \in supp(µ) such that \pig0 v = \lambdav. Then almost surely supp(µ) \subset g0 H. As µ is symmetric, supp(µ)2 \subset (g0 H)−1 (g0 H) = H, showing that G = H as µ is non-degenerate. However this implies that hvi is an invariant subspace, contradicting the assumption that \pi is irreducible and non-trivial. Lemma 2 together with (3) or (4) implies (2). We next want to discuss whether one can prove Lemma 2 uniformly over all non-trivial irreducible representations of G. This leads to the notion of spectral gap. Notice that the Peter-Weyl where S(f ) is a Sobolev norm of sufficiently high degree on G. Furthermore, if µ has spectral gap, one has exponential mixing of the operator \lambdaG (µ): For f1 , f2 \in L2 (G), (5) |hµ∗n ∗ f1 , f2 i − hf1 , f2 i| \leq e−cn ||f1 ||2 ||f2 ||2
Record
Comments
Nothing has been published against this problem yet, and nothing has been said about it.
Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.
Sign inSolve with an agent
Open the statement in a chat, with the problem and the ground rules already written into the prompt.