Polarisation microscopy and its application to optical data storage: design, informational analysis, and ML-based decoding
File(s)
Author(s)
Feng, Zhonghe
Type
Thesis or dissertation
Abstract
Optical data storage (ODS) can encode information as birefringent nanostructures in glass, but the resulting polarisation features are weak and can be degraded by photon noise, aberrations, and crosstalk. This thesis develops microscopy and processing techniques for weak polarisation signals, spanning system design and polarisation control, modelling and simulation, practical implementation, calibration, and quantitative evaluation.
Single-shot polarimetry with a polarisation camera is studied first, enabling measurement of the linear components of polarisation. Polarisation-sensitive phase contrast (PPC) microscopy is proposed by combining Zernike's phase contrast setup with polarisation control to selectively attenuate the surround wave and enhance polarisation contrast. An analytical model is derived under a weak-object assumption and complemented by a partially coherent simulation pipeline based on coherent-mode decomposition of Gaussian Schell-model sources, enabling accurate simulation of the imaging process with a flexible system definition. Experimental implementation is presented, with software for automation, calibration, and real-time visualisation.
Voxel readout is formulated as a communication channel, and mutual information (MI) is used as a decoder-agnostic metric to quantify read performance under noise, aberrations, and intra- and inter-layer crosstalk. A processing pipeline enables MI estimation from experimental and simulated images, and voxel reading is experimentally demonstrated in single-layer and multi-layer settings. By comparing axial MI profiles from simulations and experiments, a thick-voxel model is found to reproduce crosstalk characteristics, and the framework also studies MI degradation from write-side artefacts, intra-layer crosstalk, and photon shot noise. Finally, explicit symbol recovery is demonstrated with a machine-learning decoder, completing an end-to-end demonstration of ODS.
To overcome limitations inherited from spatial filtering in PPC, we introduce a diattenuative module to attenuate the surround wave using a custom partial polariser, while retaining bright-field transfer functions and avoiding azimuthal bias. Beyond voxel imaging, the diattenuative system reveals weak birefringence in other samples such as glass fibres and mycobacteria.
Single-shot polarimetry with a polarisation camera is studied first, enabling measurement of the linear components of polarisation. Polarisation-sensitive phase contrast (PPC) microscopy is proposed by combining Zernike's phase contrast setup with polarisation control to selectively attenuate the surround wave and enhance polarisation contrast. An analytical model is derived under a weak-object assumption and complemented by a partially coherent simulation pipeline based on coherent-mode decomposition of Gaussian Schell-model sources, enabling accurate simulation of the imaging process with a flexible system definition. Experimental implementation is presented, with software for automation, calibration, and real-time visualisation.
Voxel readout is formulated as a communication channel, and mutual information (MI) is used as a decoder-agnostic metric to quantify read performance under noise, aberrations, and intra- and inter-layer crosstalk. A processing pipeline enables MI estimation from experimental and simulated images, and voxel reading is experimentally demonstrated in single-layer and multi-layer settings. By comparing axial MI profiles from simulations and experiments, a thick-voxel model is found to reproduce crosstalk characteristics, and the framework also studies MI degradation from write-side artefacts, intra-layer crosstalk, and photon shot noise. Finally, explicit symbol recovery is demonstrated with a machine-learning decoder, completing an end-to-end demonstration of ODS.
To overcome limitations inherited from spatial filtering in PPC, we introduce a diattenuative module to attenuate the surround wave using a custom partial polariser, while retaining bright-field transfer functions and avoiding azimuthal bias. Beyond voxel imaging, the diattenuative system reveals weak birefringence in other samples such as glass fibres and mycobacteria.
Version
Open Access
Date Issued
2026-02-20
Date Awarded
2026-06-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Neil, Mark A A
Foreman, Matthew R
Clegg, James
Sponsor
Microsoft Research
Grant Number
MRL 2020-029
Publisher Department
Department of Physics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)