Resolving Overlapping EBSD Patterns by Experiment -- Simulation Residuals Analysis

TL;DR

Overlapping EBSD patterns from the SEM interaction volume hinder accurate phase mapping. The authors address this with a simulation driven residual analysis that iteratively fits and subtracts a scaled Kikuchi simulation per map point, where the scale is $\alpha^* = NCC(E,S')$ and the residual is $Residual = E - \alpha^* S'$, with optimization via $SSR = sum Residual^2$. This yields improved detection of weak secondary signals and minor phases, demonstrated across ferritic–pearlitic steel, bronze, dual-phase steel, H11 steel, and CuPt thin films, without requiring known orientation relationships or non-overlapping references. The method effectively extends the practical spatial resolution of EBSD and offers a transferable pixelwise framework for complex microstructures, with potential for hyperspectral EBSD concepts.

Abstract

In the technique of Electron Backscatter Diffraction (EBSD), the accurate detection and identification of different phases existing in a sample is often limited by overlapping Kikuchi diffraction patterns originating from the extended probing volume of the individual EBSD map points measured in the scanning electron microscope (SEM). We present an iterative approach that uses simulated Kikuchi patterns to resolve several overlapping diffraction signals. For each measured EBSD pattern, our method first identifies the best-fit simulated Kikuchi pattern using dynamic template matching. This simulated, ideal reference pattern is then further processed to optimally match the experimental image, uncovering any underlying weaker signals after subtraction. Repeatedly utilizing dynamic template matching and pattern subtraction on residual signals of subsequent steps enables the identification of minor phases that might otherwise be missed from the probing volume of the EBSD map point. This method significantly improves phase detection in complex materials, addressing a key limitation of conventional EBSD analysis that conventionally assigns a single phase to each map point. The present method does not require a known orientation relationship between the phases of the overlapping patterns or close neighbor experimental patterns like previously published approaches.

Figures (6)

• Figure 1: Residual patterns generated using different methods, a) Experimental pattern from cementite lamella in pearlite, b) Best fit of BCC simulation to a), c) Difference of a) and b), d) Difference of a) and b)*NCC(E,S), e) Best fit simulation blurred and multiplied by gain mask, f) Best fit simulation for overlapping cementite pattern, g) Fitted gain mask, h) Residual calculated according to Eq. 8.
• Figure 2: Ferritic-perlitic steel sample. Phase maps; as received (a) after BCC phase subtraction (b) NCC > 0.175
• Figure 3: Cu–Al–Ni–Fe two-phase bronze. Phase maps; as received (a) after FCC phase subtraction (b) NCC > 0.2
• Figure 4: Dual-phase steel microstructure. Phase maps; as received (a) after BCC phase subtraction (b) NCC > 0.2
• Figure 5: AISI H11 tool steel. Phase maps; as received (a) after BCC phase subtraction (b) NCC > 0.2