## Columbia Mathematics Department Colloquium

Randomization and long time dynamics in nonlinear evolution PDEs

by

# Andrea Nahmod

# (U. Mass.)

Abstract:

The field of nonlinear dispersive equations has undergone significant progress

in the last twenty years thanks to the influx of tools and ideas from nonlinear Fourier and harmonic analysis, geometry and analytic number theory, to the existing functional analytic methods.

This body of work has primarily focused on deterministic aspects of wave phenomena and answered questions such as existence and long time behavior of solutions, in various regimes. Yet there remain some important obstacles and open questions.

A natural approach to tackle some of them, and one which has recently seen a growing interest, is to consider certain evolution equations from a non-deterministic point of view (e.g. the random data Cauchy problem, invariant measures, etc) and incorporate to the deterministic toolbox, powerful but still classical tools from probability as well.

Such approach goes back to seminal work by Bourgain in the mid 90’s where global

well-posedness of certain periodic Hamiltonian PDEs was studied in the almost sure sense via the existence and invariance of their associated Gibbs measures.

In this talk we will explain these ideas, describe some recent work and future directions with an emphasis on the interplay of deterministic and probabilistic approaches.

in the last twenty years thanks to the influx of tools and ideas from nonlinear Fourier and harmonic analysis, geometry and analytic number theory, to the existing functional analytic methods.

This body of work has primarily focused on deterministic aspects of wave phenomena and answered questions such as existence and long time behavior of solutions, in various regimes. Yet there remain some important obstacles and open questions.

A natural approach to tackle some of them, and one which has recently seen a growing interest, is to consider certain evolution equations from a non-deterministic point of view (e.g. the random data Cauchy problem, invariant measures, etc) and incorporate to the deterministic toolbox, powerful but still classical tools from probability as well.

Such approach goes back to seminal work by Bourgain in the mid 90’s where global

well-posedness of certain periodic Hamiltonian PDEs was studied in the almost sure sense via the existence and invariance of their associated Gibbs measures.

In this talk we will explain these ideas, describe some recent work and future directions with an emphasis on the interplay of deterministic and probabilistic approaches.