< Back to previous page
Project
Chirplet-based framework for the efficent computation of numerical diffraction in complex systems (FWOTM1099)
Numerical diffraction algorithms are the backbone of many
applications in science and industry, ranging over the whole
electromagnetic spectrum. Maxwell’s equations can be simulated
using high precision methods, but do not efficiently scale to
macroscopic objects which are several orders of magnitude larger
than the wavelength. There, computer generated holography (CGH)
algorithms are typically used to efficiently model diffraction.
However, an inherent limitation to CGH algorithms is that they use
first-order approximations to Fresnel diffraction, ignoring the
Huygens–Fresnel principle, so no algorithm exists that can
accurately model diffraction for macroscopic scenes in a reasonable
time.
Solving this problem would enable accurate modeling of many
systems processing electromagnetic signals in engineering and
science, that presently take too long to compute: e.g. diffractive
optical elements and waveguides in complex systems; cascaded
optical systems, holographic measurement systems and provide
ground-truth data for approximate CGH methods.
This research proposal aims to address this problem with a chirpletbased
diffraction framework, utilizing propagating “wave packets”
analogously to ray bundles: (1) constructing a chirplet framework for
generic diffraction with new material interaction models, (2) designing
novel algorithms, acceleration structures and deep learning models,
and (3) optimizing the system and evaluating the system for multiple
target applications.
applications in science and industry, ranging over the whole
electromagnetic spectrum. Maxwell’s equations can be simulated
using high precision methods, but do not efficiently scale to
macroscopic objects which are several orders of magnitude larger
than the wavelength. There, computer generated holography (CGH)
algorithms are typically used to efficiently model diffraction.
However, an inherent limitation to CGH algorithms is that they use
first-order approximations to Fresnel diffraction, ignoring the
Huygens–Fresnel principle, so no algorithm exists that can
accurately model diffraction for macroscopic scenes in a reasonable
time.
Solving this problem would enable accurate modeling of many
systems processing electromagnetic signals in engineering and
science, that presently take too long to compute: e.g. diffractive
optical elements and waveguides in complex systems; cascaded
optical systems, holographic measurement systems and provide
ground-truth data for approximate CGH methods.
This research proposal aims to address this problem with a chirpletbased
diffraction framework, utilizing propagating “wave packets”
analogously to ray bundles: (1) constructing a chirplet framework for
generic diffraction with new material interaction models, (2) designing
novel algorithms, acceleration structures and deep learning models,
and (3) optimizing the system and evaluating the system for multiple
target applications.
Date:1 Oct 2022 → Today
Keywords:Numerical diffraction, Signal processing, High performance computing and algorithms
Disciplines:Engineering and technology, Social sciences, Natural sciences