Simulation-based Super-Resolution EBSD for Measurements of Relative Deformation Gradient Tensors

TL;DR

This paper introduces a simulation-based super-resolution method for electron backscatter diffraction (EBSD) that significantly improves the precision of deformation gradient tensor measurements from low-resolution Kikuchi patterns, enabling high-resolution strain and orientation analysis with reduced data size.

Contribution

The study presents a novel simulation-based supersampling approach that enhances EBSD data resolution and accuracy, reducing required pattern data size by about two orders of magnitude.

Findings

Achieved noise levels near 1e-4 in strain and rotation measurements.

Demonstrated high-resolution analysis on silicon wafer indentations.

Reduced pattern data size needed for EBSD analysis by approximately 100 times.

Abstract

We summarize a data analysis approach for electron backscatter diffraction (EBSD) which uses high-resolution Kikuchi pattern simulations to measure isochoric relative deformation gradient tensors from experimentally measured Kikuchi patterns of relatively low resolution. Simulation-based supersampling of the theoretical test diffraction patterns enables a significant precision improvement of tensor parameters obtained in best-fit determinations of strains and orientations from low-resolution experimental patterns. As an application, we demonstrate high-resolution orientation and strain analysis for the model case of hardness test indents on a Si(100) wafer, using Kikuchi patterns of variable resolution. The approach shows noise levels near $1 \times 10^{-4}$ in the relative deviatoric strain norm and in the relative rotation angles on nominally strain-free regions of the silicon wafer. The strain and rotation measurements are interpreted by finite element simulations. While confirming the basic findings of previously published studies, the present approach enables a potential reduction in the necessary pattern data size by about two orders of magnitude. We estimate that pattern resolutions in the order of $256\times256$ pixels should be enough to solve a majority of EBSD analysis tasks using pattern matching techniques.