Quantized Inverse Design for Photonic Integrated Circuits

Publikation: Arbeitspapier/PreprintPreprint

Autoren

Forschungs-netzwerk anzeigen

Details

OriginalspracheEnglisch
Seitenumfang19
PublikationsstatusElektronisch veröffentlicht (E-Pub) - 14 Juli 2024

Abstract

The inverse design of photonic integrated circuits (PICs) presents distinctive computational challenges, including their large memory requirements. Advancements in the two-photon polymerization (2PP) fabrication process introduce additional complexity, necessitating the development of more flexible optimization algorithms to enable the creation of multi-material 3D structures with unique properties. This paper presents an efficient reverse-mode automatic differentiation framework for finite-difference timedomain (FDTD) simulations that is able to handle several constraints arising from novel fabrication methods. Our method is based on straight-through gradient estimation that enables non-differentiable shape parametrizations. We demonstrate the effectiveness of our approach by creating increasingly complex structures to solve the coupling problem in integrated photonic circuits. The results highlight the potential of our method for future PIC design and practical applications.

Zitieren

Quantized Inverse Design for Photonic Integrated Circuits. / Schubert, Frederik; Bethmann, Konrad; Mahlau, Yannik et al.
2024.

Publikation: Arbeitspapier/PreprintPreprint

Schubert, F., Bethmann, K., Mahlau, Y., Hartmann, F., Caspary, R., Munderloh, M., Ostermann, J., & Rosenhahn, B. (2024). Quantized Inverse Design for Photonic Integrated Circuits. Vorabveröffentlichung online. https://doi.org/10.48550/arXiv.2407.10273
Schubert F, Bethmann K, Mahlau Y, Hartmann F, Caspary R, Munderloh M et al. Quantized Inverse Design for Photonic Integrated Circuits. 2024 Jul 14. Epub 2024 Jul 14. doi: 10.48550/arXiv.2407.10273
Schubert, Frederik ; Bethmann, Konrad ; Mahlau, Yannik et al. / Quantized Inverse Design for Photonic Integrated Circuits. 2024.
Download
@techreport{26ccb6d608b04e8889018218f7e485c8,
title = "Quantized Inverse Design for Photonic Integrated Circuits",
abstract = " The inverse design of photonic integrated circuits (PICs) presents distinctive computational challenges, including their large memory requirements. Advancements in the two-photon polymerization (2PP) fabrication process introduce additional complexity, necessitating the development of more flexible optimization algorithms to enable the creation of multi-material 3D structures with unique properties. This paper presents an efficient reverse-mode automatic differentiation framework for finite-difference timedomain (FDTD) simulations that is able to handle several constraints arising from novel fabrication methods. Our method is based on straight-through gradient estimation that enables non-differentiable shape parametrizations. We demonstrate the effectiveness of our approach by creating increasingly complex structures to solve the coupling problem in integrated photonic circuits. The results highlight the potential of our method for future PIC design and practical applications. ",
keywords = "physics.optics, physics.comp-ph",
author = "Frederik Schubert and Konrad Bethmann and Yannik Mahlau and Fabian Hartmann and Reinhard Caspary and Marco Munderloh and J{\"o}rn Ostermann and Bodo Rosenhahn",
note = "19 pages, 10 figures",
year = "2024",
month = jul,
day = "14",
doi = "10.48550/arXiv.2407.10273",
language = "English",
type = "WorkingPaper",

}

Download

TY - UNPB

T1 - Quantized Inverse Design for Photonic Integrated Circuits

AU - Schubert, Frederik

AU - Bethmann, Konrad

AU - Mahlau, Yannik

AU - Hartmann, Fabian

AU - Caspary, Reinhard

AU - Munderloh, Marco

AU - Ostermann, Jörn

AU - Rosenhahn, Bodo

N1 - 19 pages, 10 figures

PY - 2024/7/14

Y1 - 2024/7/14

N2 - The inverse design of photonic integrated circuits (PICs) presents distinctive computational challenges, including their large memory requirements. Advancements in the two-photon polymerization (2PP) fabrication process introduce additional complexity, necessitating the development of more flexible optimization algorithms to enable the creation of multi-material 3D structures with unique properties. This paper presents an efficient reverse-mode automatic differentiation framework for finite-difference timedomain (FDTD) simulations that is able to handle several constraints arising from novel fabrication methods. Our method is based on straight-through gradient estimation that enables non-differentiable shape parametrizations. We demonstrate the effectiveness of our approach by creating increasingly complex structures to solve the coupling problem in integrated photonic circuits. The results highlight the potential of our method for future PIC design and practical applications.

AB - The inverse design of photonic integrated circuits (PICs) presents distinctive computational challenges, including their large memory requirements. Advancements in the two-photon polymerization (2PP) fabrication process introduce additional complexity, necessitating the development of more flexible optimization algorithms to enable the creation of multi-material 3D structures with unique properties. This paper presents an efficient reverse-mode automatic differentiation framework for finite-difference timedomain (FDTD) simulations that is able to handle several constraints arising from novel fabrication methods. Our method is based on straight-through gradient estimation that enables non-differentiable shape parametrizations. We demonstrate the effectiveness of our approach by creating increasingly complex structures to solve the coupling problem in integrated photonic circuits. The results highlight the potential of our method for future PIC design and practical applications.

KW - physics.optics

KW - physics.comp-ph

U2 - 10.48550/arXiv.2407.10273

DO - 10.48550/arXiv.2407.10273

M3 - Preprint

BT - Quantized Inverse Design for Photonic Integrated Circuits

ER -

Von denselben Autoren