Implementing photometric stereo for scanning helium microscopy (SHeM) to reconstruct true-to-size 3D surfaces

TL;DR

This paper introduces a novel 3D surface reconstruction method called heliometric stereo for scanning helium microscopy, utilizing adapted photometric stereo algorithms and multi-detector setups to achieve true-to-size imaging and advanced material analysis.

Contribution

The work develops a new heliometric stereo imaging mode for SHeM, integrates 3D reconstruction with multi-detector systems, and accelerates image analysis through GPU-based Monte Carlo simulations.

Findings

Successful implementation of heliometric stereo for 3D surface reconstruction.

Development of a rapid prototyping process for vacuum-compatible SHeM components.

GPU acceleration achieved up to 380 times faster simulations.

Abstract

Scanning Helium Microscopy (SHeM) offers a combination of spatial and angular resolution via a pinhole-collimated beam of thermal energy, neutral helium-4 atoms for non-destructive imaging. This thesis introduces a novel 3D imaging mode, "heliometric stereo", enabling true-to-size 3D surface reconstruction using an adapted photometric stereo algorithm. Stereolithography (SLA) 3D printed plastics are explored for SHeM pinhole plates due to limitations in traditional machining. FormLabs "Clear Resin" via SLA printing proves ideal for rapid prototyping of vacuum components, with a developed baking protocol ensuring vacuum compatibility. The study indicates re-wetting of such plastics is a surface process over weeks. Developing 3D image reconstruction for both single and multi-detector setups required a real-space point tracking method. The point tracking method facilitates facet angle measurement in various materials, including technological and biological crystals. It has since become integral to SHeM imaging protocols for sample manipulator debugging. The thesis also details a multi-detector SHeM instrument, referred to as B-SHeM. While primarily designed to perform heliometric stereo reconstructions, the instrument also enables the range of novel SHeM experiments such as mixed-species beams to investigate inelastic scattering. The heliometric stereo methods implemented in the work have motivated the development of a GPU accelerated version of the in-house Monte-Carlo based ray tracing framework, which is the de-facto standard for SHeM image analysis. GPU parallelisation was explored as a method for decreasing simulation time and enabling previously inaccessible simulations involving complex scattering distributions and high resolution, realistic sample geometries. Preliminary testing on an analogous problem yielded a potential performance increase of up to 380 times.