Advanced Photonics Congress

26 July 2021 – 30 July 2021
Optica Virtual Event - Eastern Daylight/Summer Time (UTC - 04:00)

Symposium: Machine Learning for Materials Discovery (JW3D)

Presider: Sarvagya Dwivedi, IMEC
Authentication Required

Access to this content is only available to registered attendees. Please login now to access the live content.

16:00 - 16:30
(UTC - 04:00)

Generalized nano-optics fields predictions and inverse design of complex transmission matrices enabled by deep learning (JW3D.1)
Presenter: Peter Wiecha, Laboratoire d'Analyse et d'Architecture des Systemes du CNRS

We present a generalized nano-optics deep learning (DL) model, capable to predict arbitrary near- and far-field effects without re-training. We furthermore discuss DL inverse design of complex transmission matrices in multimode interference devices.

Authors:Peter Wiecha, Laboratoire d'Analyse et d'Architecture des Systemes du CNRS / Otto Muskens, University of Southampton

16:30 - 17:00
(UTC - 04:00)

New photocathode materials identified by data driven discovery (JW3D.2)
Presenter: Evan Reed, Stanford University

We identify 11 new photocathodes with emittance up to 4x lower than current K2CsSb photocathodes. This is accomplished by screening over 70,000 candidates using machine learning and electronic structure calculations.

Authors:Evan Reed, Stanford University

17:00 - 17:30
(UTC - 04:00)

Thin-Film Metamaterial Optical Diode Designed Using Machine Learning (JW3D.3)
Presenter: Tengfei Luo, University of Notre Dame

In this work, we use a machine learning model, Factorization Machine (FM), to help design thin film metamaterial optical diode with high transmission and isolation. The identified optimal structures are validated using RCWA calculations.

Authors:Eungkyu Lee, Kyung Hee University / Seongmin Kim, University of Notre Dame / Tengfei Luo, University of Notre Dame

17:30 - 17:45
(UTC - 04:00)

High-speed analysis of spectroscopic ellipsometry data using deep learning methods (JW3D.4)
Presenter: Yifei Li, Massachusetts Institute of Technology

We develop deep-learning methods for rapid analysis of spectroscopic ellipsometry data, which speeds analysis by thousand-fold compared to traditional methods. We demonstrate the usefulness for a high-throughput study of phase-change alloys.

Authors:Yifei Li, Massachusetts Institute of Technology / Yifeng Wu, Independent research / Heshan Yu, University of Maryland at College Park / Ichiro Takeuchi, University of Maryland at College Park / Rafael Jaramillo, Massachusetts Institute of Technology

  Paper