EUVIMEDEUVIMED entdecken
Crossref

Optimizing Heat-Assisted Magnetic Recording Plasmonic Structures via Deep Generative Models

Zulfidin Khodzhaev, Mike Seigler, Olle Heinonen

IEEE Transactions on Magnetics · 2026-07 · 62 (7) · 3200507-3200507

DOI: 10.1109/tmag.2026.3698384 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Magnetics, 2026-07. Verfasst von Zulfidin Khodzhaev und 2 weiteren Autor:innen. Die Arbeit nennt 11 Literaturangaben. Gefördert von: Seagate Technology, Bloomington, MN, USA.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Designing efficient plasmonic components for heat-assisted magnetic recording (HAMR) involves optimizing complex geometries under competing physical objectives, a task challenging for traditional methods. We propose and evaluate two deep generative approaches—a denoising diffusion probabilistic model (DDPM) pipeline and a variational autoencoder (VAE) pipeline coupled with latent space optimization—to optimize the geometry of a critical HAMR component: the mini solid immersion mirror (miniSIM). Our objective is to minimize the COMSOL power input (CPI), the laser power needed to achieve a 750 K peak recording temperature, while maintaining the effective width of a written track on the magnetic storage medium, the write-plus-erasure (WPE), measured at the 684 K isotherm. Experiments show both methods offer a promising methodological framework that improves upon manual optimization baselines in this demonstration case, with the potential for greater advantages in higher dimensional parameter spaces. Notably, the VAE combined with a multi-objective genetic algorithm (MOGA) in the learned latent space achieves a 20.1% reduction in CPI from 28.59 to 22.83 (arbitrary units) while maintaining comparable WPE performance. We also demonstrate that the miniSIM opening (MSO), often treated as fixed, is a geometry-dependent parameter important for optimal performance. Our work presents effective deep learning frameworks for complex, physics-based nanophotonic design optimization with implications for accelerating next-generation storage technology development.

Förderung

  • Seagate Technology, Bloomington, MN, USA

Zitieren

APA

Khodzhaev, Z., Seigler, M., & Heinonen, O. (2026). Optimizing Heat-Assisted Magnetic Recording Plasmonic Structures via Deep Generative Models. IEEE Transactions on Magnetics, 62(7), 3200507-3200507. https://doi.org/10.1109/tmag.2026.3698384

Vancouver

Khodzhaev Z, Seigler M, Heinonen O. Optimizing Heat-Assisted Magnetic Recording Plasmonic Structures via Deep Generative Models. IEEE Transactions on Magnetics. 2026;62(7):3200507-3200507. doi:10.1109/tmag.2026.3698384

BibTeX
@article{Khodzhaev2026,
  author = {Khodzhaev, Zulfidin and Seigler, Mike and Heinonen, Olle},
  title = {Optimizing Heat-Assisted Magnetic Recording Plasmonic Structures via Deep Generative Models},
  journal = {IEEE Transactions on Magnetics},
  year = {2026},
  volume = {62},
  number = {7},
  pages = {3200507--3200507},
  doi = {10.1109/tmag.2026.3698384}
}
Kostenlos starten →