DPG Phi
Verhandlungen
Verhandlungen
DPG

Mainz 2017 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

Q: Fachverband Quantenoptik und Photonik

Q 31: Poster: Quantum Optics and Photonics I

Q 31.79: Poster

Dienstag, 7. März 2017, 17:00–19:00, P OG1+2

The Gauss-Newton algorithm for light scattering on transparent cylinders — •Gunnar Claussen1,2, Werner Blohm1, and Armin Lechleiter21Fachbereich Ingenieurwissenschaften, Jade Hochschule Wilhelmshaven Oldenburg Elsfleth — 2Zentrum für Technomathematik, Universität Bremen

We aim to determine the diameter of a glass fiber under perpendicular incidence of plane-wave light and treat this question as an inverse scattering problem, which is solvable through an iteratively regularized Gauss-Newton algorithm. Within each step of the algorithm, the expression ||F′[qnδ]hn + F(qnδ) − uδ||L22 + αn||hn + qnδq0||s2 is minimized. This term includes the measured far-field pattern uδ, the data-to-pattern operator F(·), i.e. the formalism given by the Mie theory for cylinder scattering, and its Fréchet derivate F′(·), the parameter vector qnδ and its alteration hn within the current step of the iteration. The algorithm terminates once the residual ||F(qnδ)−uδ|| becomes sufficiently small. However, it turns out that for transparent cylinders the residual forms a complex “landscape” in the parameter-space that is characterized by a number of false minima, thereby hindering the correct execution of the algorithm. We introduce a novel variation of the Gauss-Newton algorithm which allows to skip these minima in order to reach the global minimum, allowing us to determine the cylinder diameter with a precision several magnitudes smaller than the incident wavelength λ. We will present the performance of this algorithm in terms of precision, running time, experimental applicability and tolerance towards variations of fixed parameters.

100% | Mobil-Ansicht | English Version | Kontakt/Impressum/Datenschutz
DPG-Physik > DPG-Verhandlungen > 2017 > Mainz