EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Quantum-enhanced physics-informed neural network framework for multi-modal optimization of diffractive optics in extreme laser environments

R. Hariharan, N. Prathap, M. Irshad Ahamed, M. Nuthal Srinivasan

Bulletin of the Polish Academy of Sciences Technical Sciences · 03.11.2026 · 74 (6) · 3058

DOI: 10.24425/bpasts.2026.3058 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Bulletin of the Polish Academy of Sciences Technical Sciences, online erschienen am 03.11.2026. Verfasst von R. Hariharan und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Q-PINN-opt creates a new framework by combining the concepts of quantum computing with Physics-Informed Neural Networks (PINNs) for multi-modal optimization of diffractive optics under laser conditions. In high-power laser applications, which involve the optimization of complex problems, the model employs quantum-enhanced optimization coupled with physics-informed learning. In contrast to traditional approaches based on PINNs orWiener filters, which are limited by their inability to scale, poor computational efficiency, and limited resilience to extreme conditions, Q-PINN-opt is a hybrid quantum-classical algorithm that accelerates convergence and maximizes computational efficiency. The major innovation of the Q-PINN-opt variant is that quantum algorithms (such as Quantum Gradient Descent and Quantum Approximate Optimization Algorithm (QAOA)) are implemented using classical PINNs. This enables faster, scalable, and multimodal optimization of problems, including diffraction efficiency, phase accuracy, and resistance to thermal, nonlinear, and quantum-induced effects. Current models such as PINNs may also exhibit slow convergence and computational inefficiency when applied to optimization problems of a complex, high-dimensional nature, especially when multiple performance criteria are to be optimized simultaneously. Q-PINN-opt addresses all these shortcomings; thus, it is specifically applicable to the optimization of diffractive optics for real-time adaptive learning. Q-PINN-opt addresses this problem by leveraging quantum computing to train more effectively and process large-scale data efficiently. Moreover, the hybrid quantum-classical model ensuresthe physical constraints of network solutions, enhancing the resilience and scalability of high-performance optical systems. The Q-INN-opt framework metric analysis reveals that it performs better than other models. MSE, SSIM, and SNR in the model are 6.84, 0.99, and 7.98, respectively, which implies that the model has high accuracy, diffraction efficiency, and resistance to noise. In contrast, the original model achieves an MSE of 9.55 and an SNR of 5.54. The improvements enhance both the training and the convergence of the optimization, making Q-PINN-opt well-constructed and efficient for the design of complicated diffractive optics in extreme environments.

Zitieren

APA

Hariharan, R., Prathap, N., Ahamed, M. I., & Srinivasan, M. N. (2026). Quantum-enhanced physics-informed neural network framework for multi-modal optimization of diffractive optics in extreme laser environments. Bulletin of the Polish Academy of Sciences Technical Sciences, 74(6), 3058. https://doi.org/10.24425/bpasts.2026.3058

Vancouver

Hariharan R, Prathap N, Ahamed MI, Srinivasan MN. Quantum-enhanced physics-informed neural network framework for multi-modal optimization of diffractive optics in extreme laser environments. Bulletin of the Polish Academy of Sciences Technical Sciences. 2026;74(6):3058. doi:10.24425/bpasts.2026.3058

BibTeX
@article{Hariharan2026,
  author = {Hariharan, R. and Prathap, N. and Ahamed, M. Irshad and Srinivasan, M. Nuthal},
  title = {Quantum-enhanced physics-informed neural network framework for multi-modal optimization of diffractive optics in extreme laser environments},
  journal = {Bulletin of the Polish Academy of Sciences Technical Sciences},
  year = {2026},
  volume = {74},
  number = {6},
  pages = {3058},
  doi = {10.24425/bpasts.2026.3058}
}
Kostenlos starten →