Data-driven doping reconstruction

To reconstruct doping profiles via opto-electronic techniques (e.g. LBIC and LPS), we formulate an inverse problem based on the van Roosbroeck system. To solve it, we use neural networks fed with data created from efficient implementations of the forward model. We discuss errors of the reconstructed...

Full description

Saved in:
Bibliographic Details
Main Authors: Piani, S., Lei, W., Heltai, L., Rotundo, N., Farrell, P.
Format: Conference Proceeding
Language:English
Subjects:
Data models
Doping profiles
Inverse problems
Neural networks
Numerical simulation
Optoelectronic devices
Robustness
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To reconstruct doping profiles via opto-electronic techniques (e.g. LBIC and LPS), we formulate an inverse problem based on the van Roosbroeck system. To solve it, we use neural networks fed with data created from efficient implementations of the forward model. We discuss errors of the reconstructed doping profiles as well as their robustness with respect to noise.
ISSN:2158-3234
DOI:10.1109/NUSOD54938.2022.9894774