XiEff Representation for Near-Field Optics

Vasyl Vasylenko, Ihor Tymchyshyn, Vitalii Tymchyshyn

Keywords: neural fields, PINNs, near-field optics

TL;DR: We reproduce matter distribution by electric field distortion

Abstract:

Near-field optics, or near-field electrodynamics, is a field that studies the interaction between materials and light at spatial scales smaller than the wavelength. At these extremely small scales, below the diffraction limit, the interaction between materials and electromagnetic fields can exhibit unique behaviors and properties not observed in conventional optics. This area of research is crucial for understanding the optical characteristics of nanotechnical systems and nanoscale biological objects. One of the primary tools used in near-field optics research is scanning near-field optical microscopy (SNOM), which allows researchers to measure near-field optical images (NFI). However, these images often lack visual clarity and interpretability, hindering a comprehensive understanding of the properties of the probed particles.

The main goal of this paper is to introduce a novel approach that addresses these challenges. Inspired by the prominent progress in Neural Radiance Fields (NeRFs) from computer vision and ideas from physics-informed neural networks (PINNs). We propose an unsupervised method that introduces the XiEff representation – a neural field-based reparameterization of the effective susceptibility tensor. By integrating XiEff into the Lippmann-Schwinger integral equation framework for near-field optics we develop an optimization strategy to reconstruct the effective susceptibility distribution directly from NFI data.

The optimized XiEff representation provides an interpretable and explainable model of the particle's shape. Extensive evaluations on a synthetically generated NFI dataset demonstrate the effectiveness of the method, achieving high intersection-over-union scores between XiEff and ground truth shapes, even for complex geometries. Furthermore, the approach exhibits desirable robustness to measurement noise, a crucial property for practical applications. The XiEff representation, combined with the proposed optimization framework, potentially introduces a valuable tool for enabling explainable near-field optics imaging and enhancing the understanding of particle characteristics through interpretable representations

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 7623