Back to search

Boston University

Numerical simulation and data-driven analysis of infrared detector performance

Abstract

dc:description.abstract

Infrared detectors are a critical technology frequently used in commercial, military, and scientific settings. Research and development of modern infrared detectors is driven by finding new ways to reduce the system's size, weight, power, and cost while adding new functionalities without lowering performance. Physics-based numerical models can be instrumental in lowering the cost of developing such advances. This dissertation presents three main contributions to further the predictive modeling of infrared photodetectors. First, motivated by recent demonstrations of small-pitch focal plane arrays for infrared imaging---5 μm for SWIR and 10 to 15 μm for MWIR/LWIR---we use physics-based numerical simulations to assess the implications on dark current, quantum efficiency, specific detectivity, and modulation transfer function in SWIR InGaAs FPAs. From the results, we propose a new pixel sub-architecture aimed toward lowering dark current and improving MTF. Second, we present a methodology for simulating the capacitance-voltage (C-V) characteristics of nBn photodetectors. For junction-based semiconductor devices, C-V profiling is a common technique for non-destructively characterizing semiconductor layers in metal-insulator-semiconductor devices by using well-established analytical relations. However, this type of analysis cannot be directly applied to the barrier detector’s unique architecture and the formalism must be modified. To this end, we present a modified analytical formalism based on metal-oxide-semiconductor theory, and a methodology using the drift-diffusion method; both are used to explore the role of the device architectural properties on determining the C-V characteristics. Last, we present several cases of applying neural networks to nBn photodetector figures of merit. We use artificial neural networks as surrogate model for the capacitance-voltage, current-voltage, and quantum efficiency to explore the multi-dimensional parameter space to assess parameter-performance correlations and determine the global role of each feature in shaping each characteristic without the need for additional simulations. Moreover, using inspiration from image recognition, we demonstrate that a convolutional neural network can be trained to analyze a C-V characteristic to yield more information about a device than what would be possible from a conventional analysis.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Electrical & Computer Engineering
Grantor
Boston University
Year dc:date.issued
2020

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Glasmann, Andreu
Advisor dc:contributor.advisor
  • Bellotti, Enrico
ORCID dc:identifier.orcid

Subjects

dc:subject × 1

Rights

dc:rights
Statement dc:rights
  • Attribution 4.0 International
Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/2144/41492

Chain of custody

source
Harvested from
Boston University
Base URL
open.bu.edu/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Glasmann, Andreu. Numerical simulation and data-driven analysis of infrared detector performance. doctoral thesis, Boston University, 2020. https://hdl.handle.net/2144/41492