We are interested in the behaviour of laplacian eigenfunctions on negatively curved manifolds, in the high frequency limit. The Quantum Unique Ergodicity conjecture predicts that they should become uniformly distributed over phase space, and the Shnirelman theorem states that this is true if we allow ourselves to possibly drop a ``negligible'' family of eigenfunctions. Nonnenmacher and I proved that, in any case, the eigenfunctions must in the high frequency regime have a large Kolmogorov-Sinai entropy : this prevents them, for instance, from concentrating on periodic geodesics. The proof uses notions from ergodic theory (such as entropy) mixed with techniques from linear PDE. Talk II : I will prove the technical estimate and finish the proof of the main theorem.