Shanhui Fan: Computation for Classical and Quantum Light Using Optical Structures (October 22, 2026)

Shanhui Fan: Computation for Classical and Quantum Light Using Optical Structures (October 22, 2026)

Title history
  • Shanhui Fan: Computation for Classical and Quantum Light Using Optical Structures (October 22, 2026) — observed on 20261117-0331 UTC (YouTube)
Applied Sciences & Engineering Physics PHPhysicsPHJOptical physics
🎙 Shanhui Fan 👥 62K 📅 November 16, 2026 ⏱ 53 min 👁 193 📄 original study 🧭 2026-11-17
Available in: English (current) Français

Keywords

non-Hermitian photonicsedge detectionwaveguide meshSchmidt decompositionoptical neural network

Summary

Shanhui Fan presents a research overview on using optical structures for both classical and quantum computation. He begins by framing the energy cost of AI and the potential of passive optical matrix-vector multiplication. The talk then details three lines of work: metasurface-based image processing, programmable waveguide interferometer meshes, and the extension to quantum photonics. For classical light, he demonstrates a nonlocal metasurface that performs isotropic edge detection, and a topologically robust operator with broad bandwidth. For on-chip systems, he explains the use of waveguide meshes for programmable unitary transformations and introduces an in-situ backpropagation protocol for training optical neural networks without a computer model. He also discusses electro-optic activation for nonlinearity. In the quantum part, he shows how waveguide meshes can perform Schmidt decomposition of two-photon states, enabling entanglement characterization and potential coherence mitigation. The talk emphasizes experimental implementations and theoretical insights, highlighting the role of symmetries and topology in photonic computing.

152 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides high value by integrating theoretical foundations with concrete experimental demonstrations, addressing the pressing issue of computing energy consumption. The argumentation is solid: each concept is built from fundamental physics and mathematically justified, and claims are backed by experimental results, such as the broadband edge detection and the in-situ training of a waveguide mesh. The speaker effectively explains complex ideas like the Riemann surface perspective and the adjoint variable method, making the underlying logic accessible without oversimplifying. The progression from simple to complex systems is logical, and the inclusion of quantum photonics adds depth and novelty.

Scientific Rigor, Source Quality, Title Accuracy

The speaker demonstrates scientific rigor by clearly stating the limits and contexts of each approach, using precise mathematical formulations, and citing prior work (e.g., Capasso, Alù, Miller, Englund, and Soljačić). Although detailed references are not provided in the talk, the assertions align with established research. The title accurately matches the content, which systematically explores computation with optical structures. The only provided link is to the Simons Foundation event page, which serves as a valuable reference for the talk’s context, though specific papers are not explicitly linked.

199 words

Title / Content Match

The title accurately reflects the content: the talk covers computational applications of optical structures for both classical and quantum light, including metasurfaces, waveguide meshes, and Schmidt decomposition.

Quality & Reliability

9/10

The speaker, a leading researcher in photonics, presents original research with experimental validations, rigorous mathematical reasoning, and references to established work. The talk is technically dense and credible, demonstrating a deep understanding of both theoretical and practical aspects.

Key Moments

Cited Sources

Contribution & Novelties

This talk synthesizes original advances in using photonic structures for computation, particularly the demonstration of passive optical edge detection with broad bandwidth, the in-situ training method for programmable waveguide meshes, and the application of waveguide meshes for Schmidt decomposition of quantum states. The novelty lies in the direct implementation of topology and symmetries to achieve robust, programmable optical processors that could reduce energy costs in AI and quantum information processing.

Pour aller plus loin :

  • Metasurface — Provides foundational concepts on spatial light manipulation used in the talk.
  • Backpropagation — The standard training algorithm for neural networks, which the in-situ protocol adapts to optical hardware.
  • Schmidt decomposition — Mathematical tool central to quantum state characterization and entanglement analysis.
  • Photonic crystal — Underpins the guided resonance physics exploited for edge detection.

130 words

Radar Profile

The radar profile is strongly balanced, with high scores across quantity, quality, technical depth, and reliability, indicating a comprehensive and technically rigorous presentation that excels in delivering dense, trustworthy content.

Reliability 9/10