Fast Forward Prediction of Metasurface Transmission Spectra Using Deep Learning

Md Imran Hossain; Linzhi Yu; Humeyra Caglayan

doi:10.1109/Metamaterials65622.2025.11174168

Fast Forward Prediction of Metasurface Transmission Spectra Using Deep Learning

Md Imran Hossain^*
, Linzhi Yu
, Humeyra Caglayan

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

Abstract

Metasurface, ultra-thin, 2D, subwavelength-engineered structure with unique electromagnetic properties and exceptional light-matter interactions. In this work, we are presenting a neural network miming a conventional numerical simulator like Ansys Lumerical FDTD. We have introduced a neural network based on ResNet-18 to predict the transmission spectra of gold metasurfaces within the 1200 to 1700 nm region. The model is trained with 10,000 simulated data of metasurfaces with varied geometries. The model can accurately predict the transmission spectra within a couple of milliseconds. Evaluation over 150 datasets shows that the model has an average prediction accuracy of 85.27%. The model aims to present a time-efficient method for investigating polarization conversion and vector holography through gradient metasurfaces.

Original language	English
Title of host publication	19th International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials
Publisher	IEEE
Pages	150-152
Number of pages	3
ISBN (Electronic)	9798331536565
DOIs	https://doi.org/10.1109/Metamaterials65622.2025.11174168
Publication status	Published - 2025
Publication type	A4 Article in conference proceedings
Event	International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials - Amsterdam, Netherlands Duration: 1 Sept 2025 → 6 Sept 2025 Conference number: 19

Publication series

Name	International congress on advanced electromagnetic materials in microwaves and optics
Publisher	IEEE
ISSN (Print)	2573-2684
ISSN (Electronic)	2573-2706

Conference

Conference	International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials
Country/Territory	Netherlands
City	Amsterdam
Period	1/09/25 → 6/09/25

Publication forum classification

Publication forum level 0

ASJC Scopus subject areas

Computer Networks and Communications
Electronic, Optical and Magnetic Materials
Surfaces, Coatings and Films
Instrumentation

Access to Document

10.1109/Metamaterials65622.2025.11174168

Cite this

Hossain, M. I., Yu, L., & Caglayan, H. (2025). Fast Forward Prediction of Metasurface Transmission Spectra Using Deep Learning. In 19th International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials (pp. 150-152). (International congress on advanced electromagnetic materials in microwaves and optics). IEEE. https://doi.org/10.1109/Metamaterials65622.2025.11174168

@inproceedings{b899a9fdb95c4dd2b67b6284bf73015a,

title = "Fast Forward Prediction of Metasurface Transmission Spectra Using Deep Learning",

abstract = "Metasurface, ultra-thin, 2D, subwavelength-engineered structure with unique electromagnetic properties and exceptional light-matter interactions. In this work, we are presenting a neural network miming a conventional numerical simulator like Ansys Lumerical FDTD. We have introduced a neural network based on ResNet-18 to predict the transmission spectra of gold metasurfaces within the 1200 to 1700 nm region. The model is trained with 10,000 simulated data of metasurfaces with varied geometries. The model can accurately predict the transmission spectra within a couple of milliseconds. Evaluation over 150 datasets shows that the model has an average prediction accuracy of 85.27\%. The model aims to present a time-efficient method for investigating polarization conversion and vector holography through gradient metasurfaces.",

author = "Hossain, \{Md Imran\} and Linzhi Yu and Humeyra Caglayan",

note = "Publisher Copyright: {\textcopyright} 2025 IEEE.; International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials ; Conference date: 01-09-2025 Through 06-09-2025",

year = "2025",

doi = "10.1109/Metamaterials65622.2025.11174168",

language = "English",

series = "International congress on advanced electromagnetic materials in microwaves and optics",

publisher = "IEEE",

pages = "150--152",

booktitle = "19th International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials",

address = "United States",

}

Hossain, MI, Yu, L & Caglayan, H 2025, Fast Forward Prediction of Metasurface Transmission Spectra Using Deep Learning. in 19th International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials. International congress on advanced electromagnetic materials in microwaves and optics, IEEE, pp. 150-152, International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials, Amsterdam, Netherlands, 1/09/25. https://doi.org/10.1109/Metamaterials65622.2025.11174168

Fast Forward Prediction of Metasurface Transmission Spectra Using Deep Learning. / Hossain, Md Imran; Yu, Linzhi ; Caglayan, Humeyra.
19th International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials. IEEE, 2025. p. 150-152 (International congress on advanced electromagnetic materials in microwaves and optics).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

TY - GEN

T1 - Fast Forward Prediction of Metasurface Transmission Spectra Using Deep Learning

AU - Hossain, Md Imran

AU - Yu, Linzhi

AU - Caglayan, Humeyra

N1 - Conference code: 19

PY - 2025

Y1 - 2025

N2 - Metasurface, ultra-thin, 2D, subwavelength-engineered structure with unique electromagnetic properties and exceptional light-matter interactions. In this work, we are presenting a neural network miming a conventional numerical simulator like Ansys Lumerical FDTD. We have introduced a neural network based on ResNet-18 to predict the transmission spectra of gold metasurfaces within the 1200 to 1700 nm region. The model is trained with 10,000 simulated data of metasurfaces with varied geometries. The model can accurately predict the transmission spectra within a couple of milliseconds. Evaluation over 150 datasets shows that the model has an average prediction accuracy of 85.27%. The model aims to present a time-efficient method for investigating polarization conversion and vector holography through gradient metasurfaces.

AB - Metasurface, ultra-thin, 2D, subwavelength-engineered structure with unique electromagnetic properties and exceptional light-matter interactions. In this work, we are presenting a neural network miming a conventional numerical simulator like Ansys Lumerical FDTD. We have introduced a neural network based on ResNet-18 to predict the transmission spectra of gold metasurfaces within the 1200 to 1700 nm region. The model is trained with 10,000 simulated data of metasurfaces with varied geometries. The model can accurately predict the transmission spectra within a couple of milliseconds. Evaluation over 150 datasets shows that the model has an average prediction accuracy of 85.27%. The model aims to present a time-efficient method for investigating polarization conversion and vector holography through gradient metasurfaces.

U2 - 10.1109/Metamaterials65622.2025.11174168

DO - 10.1109/Metamaterials65622.2025.11174168

M3 - Conference contribution

AN - SCOPUS:105019040510

T3 - International congress on advanced electromagnetic materials in microwaves and optics

SP - 150

EP - 152

BT - 19th International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials

PB - IEEE

T2 - International Congress on Artificial Materials for Novel Wave Phenomena, Metamaterials

Y2 - 1 September 2025 through 6 September 2025

ER -

Fast Forward Prediction of Metasurface Transmission Spectra Using Deep Learning

Abstract

Publication series

Conference

Publication forum classification

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this