NLSTEM: Non-local denoising for enhanced 4D-STEM pattern indexing

TL;DR

This paper presents NLSTEM, a post-processing denoising method for 4D-STEM pattern indexing that improves accuracy and rate, especially in damaged samples, by averaging similar diffraction patterns.

Contribution

It introduces a non-local pattern averaging algorithm for 4D-STEM data, enhancing indexing rates without requiring beam precession.

Findings

Indexing rates significantly improved with the denoising method.

Highest improvements observed in ion-irradiated, damaged samples.

Enhanced signal-to-noise ratios lead to better pattern indexing.

Abstract

4D-STEM-based orientation and phase mapping has enabled rapid microstructure quantification that can be directly combined with standard TEM- and STEM-based imaging modes. Typically, orientation mapping is coupled with beam precession (i.e. precession electron diffraction) to achieve high indexing rates, adding to the cost and often decreasing the spatial resolution of the approach. This paper introduces a new post processing approach modeled after the non-local pattern averaging and reindexing algorithm developed for the electron backscatter diffraction community, wherein post-collection, patterns are averaged using a distance similarity parameter. Results from Ni and Au thin films show that indexing rates can be significantly improved using this post-processing technique due to improved signal-to-noise ratios in the diffraction patterns. Interestingly, the highest indexing rates are achieved in samples heavily damaged via ion irradiation, suggesting that averaging over curved lattices further improves indexing rates.