NuHAG :: TALKS

Talks given at NuHAG events

Generalization of the Weyl-Moyal calculus and computational homogenization of wave propagation


  Agissilaos Athanassoulis

  given at  strobl07 (22.06.07 11:30)
  id:  620
  length:  25min
  status:  accepted
  type:  talk
  www:  http://www.math.princeton.edu/~aathanas
  LINK-Preprint:  http://arxiv.org/ftp/math/papers/0611/0611564.pdf
  LINK-Presentation:  http://www.math.princeton.edu/~aathanas/strobl_agis.pdf
  ABSTRACT:
In the study of wave propagation over distances much longer than the typical wavelength phase-space methods are often used, i.e. spectral densities are used as a homogenized representation of the wavefield, and kinetic equations are constructed for their evolution.

The Wigner Transform (WT) is a nonlinear, nonparametric spectral density, first introduced by E. Wigner in the context of quantum mechanics. Recently it has been extensively used in the formulation of phase-space models for a variety of problems, such as geometrical optics limits, periodic problems, nonlinear and/or random waves. Physical areas of application include semiconductors, linear and nonlinear optics, water waves and more.

However, the WT features counterintuitive ‘interference components’, which make computation and interpretation problematic. To face this, variants such as Wigner measures are typically used in practice.

We use smoothed Wigner Transforms (SWT) to study wave propagation. We present the derivation of new, to the best of our knowledge, exact equations of motion for the SWT covering a broad class of wave propagation problems. As a special case we get exact equations of motion for spectrograms of wave fields. These equations are typically pseudodifferential.

The new equations are used for the construction of a reliable, efficient, homogenized (slow-scale) numerical solver for the Schrodinger equation. The new solver successfully captures the spatial structure of caustics, while clearly outperforming conventional, full (i.e. not homogenized) solvers. A subtle issue which we discuss is the construction of an appropriate `slow-scale error', i.e. a quantitative measure of how good an approximation our SWT solution offers; exact solutions are used to test our 'slow-scale error', with good results.


Enter here the CODE for editing this talk:
If you have forgotten the CODE for your talk click here to send an email to the Webmaster!
NOTICE: In [EDIT-MODUS] you can also UPLOAD a presentation"