EUVIMEDEUVIMED entdecken
Crossref

On-demand inverse design of meta-atoms via spectrum-aligned latent diffusion with manufacturability constraints

Lu Zhu, Qi Liu, Jiajun Ji, Yuanyuan Liu

Journal of the Optical Society of America B · 01.06.2026 · 43 (6) · 1130

DOI: 10.1364/josab.589749 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of the Optical Society of America B, online erschienen am 07.05.2026. Verfasst von Lu Zhu und 3 weiteren Autor:innen. Die Arbeit nennt 41 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Jiangxi Provincial Natural Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The inverse design of metamaterials involves mapping (low-dimensional) spectral responses to (high-dimensional) geometries. In this design process, a “one-to-many” mapping problem may arise (similar spectra may correspond to multiple structures). Traditional generative models such as VAEs, GANs, and DDPMs have been widely applied to inverse design problems, but often fail to ensure tight spectrum–geometry consistency and may produce complex structures that are difficult to manufacture. To address these issues, we propose a spectrum-aligned latent diffusion framework (SA-LDM), augmented with a predictive neural network (PNN) and multiple manufacturability constraints, enabling on-demand intelligent design of meta-atoms with high degrees of freedom. A multi-scale cross-attention encoder for latent encoding and a spectrum-aware conditional gated fusion module in the latent diffusion process jointly enhance alignment between generated structures and target spectra. During inference, the model leverages the PNN to rapidly evaluate candidate spectra and applies manufacturability constraints, thereby directly selecting suitable and fabricable meta-atoms from a single-pass library. Moreover, the constraint set can be flexibly tailored to the application scenario, obviating the need for complex, repetitive, and time-consuming iterative optimization. Simulation results show that SA-LDM outperforms mainstream generative models in both spectral-matching accuracy and structural simplicity.

Förderung

  • National Natural Science Foundation of China
  • Jiangxi Provincial Natural Science Foundation

Zitieren

APA

Zhu, L., Liu, Q., Ji, J., & Liu, Y. (2026). On-demand inverse design of meta-atoms via spectrum-aligned latent diffusion with manufacturability constraints. Journal of the Optical Society of America B, 43(6), 1130. https://doi.org/10.1364/josab.589749

Vancouver

Zhu L, Liu Q, Ji J, Liu Y. On-demand inverse design of meta-atoms via spectrum-aligned latent diffusion with manufacturability constraints. Journal of the Optical Society of America B. 2026;43(6):1130. doi:10.1364/josab.589749

BibTeX
@article{Zhu2026,
  author = {Zhu, Lu and Liu, Qi and Ji, Jiajun and Liu, Yuanyuan},
  title = {On-demand inverse design of meta-atoms via spectrum-aligned latent diffusion with manufacturability constraints},
  journal = {Journal of the Optical Society of America B},
  year = {2026},
  volume = {43},
  number = {6},
  pages = {1130},
  doi = {10.1364/josab.589749}
}
Kostenlos starten →