Phase Segmentation in Atom-Probe Tomography Using Deep Learning-Based Edge Detection

Sandeep Madireddy,David N Seidman,Prasanna Balaprakash,Ding-Wen Chung,Troy Loeffler,Olle Heinonen,Subramanian K R S Sankaranarayanan

doi:10.1038/s41598-019-56649-8

Sandeep Madireddy, David N Seidman + Show 5 more

Open Access

https://doi.org/10.1038/s41598-019-56649-8

Copy DOI

Abstract

Atom-probe tomography (APT) facilitates nano- and atomic-scale characterization and analysis of microstructural features. Specifically, APT is well suited to study the interfacial properties of granular or heterophase systems. Traditionally, the identification of the interface between, for precipitate and matrix phases, in APT data has been obtained either by extracting iso-concentration surfaces based on a user-supplied concentration value or by manually perturbing the concentration value until the iso-concentration surface qualitatively matches the interface. These approaches are subjective, not scalable, and may lead to inconsistencies due to local composition inhomogeneities. We introduce a digital image segmentation approach based on deep neural networks that transfer learned knowledge from natural images to automatically segment the data obtained from APT into different phases. This approach not only provides an efficient way to segment the data and extract interfacial properties but does so without the need for expensive interface labeling for training the segmentation model. We consider here a system with a precipitate phase in a matrix and with three different interface modalities—layered, isolated, and interconnected—that are obtained for different relative geometries of the precipitate phase. We demonstrate the accuracy of our segmentation approach through qualitative visualization of the interfaces, as well as through quantitative comparisons with proximity histograms obtained by using more traditional approaches.

Highlights

Advances in atom-probe tomography (APT) allow three-dimensional atomic reconstruction of materials with an unparalleled spatial and atomic resolution[1,2,3]
We demonstrate the qualitative and quantitative accuracy of our approach using a synthetic dataset constructed with molecular dynamics (MD) simulations, as well as experimental Atom-probe tomography (APT) datasets of Co and Al superalloys[24,25]
We evaluate the effectiveness and quantitative fidelity of the proposed supervised edge detection and transfer-learning-based digital image segmentation approach using three interface modalities: (1) layered interface, in which the precipitate and matrix are two different layers separated by a thin interface; (2) isolated interface, in which the precipitate is an ellipsoid embedded in the matrix; and (3) interconnected interface, which is a general case where the precipitate phase exhibits an irregular morphology

Summary

Introduction

Advances in atom-probe tomography (APT) allow three-dimensional atomic reconstruction of materials with an unparalleled spatial and atomic resolution[1,2,3]. We focus on identifying the interface in a precipitate-matrix system by phase segmentation This approach holds the potential to expedite and reduce inconsistencies in the process of identifying interfaces and study of interfacial properties and can be scaled up to high-performance computer platforms. Several supervised segmentation approaches exist that use a priori knowledge involving the ground truth of the segments to recognize and label the pixels in a new image according to one of the object classes on which the model is trained This approach is usually referred to as semantic segmentation[12] and tends to identify different objects present in an image as well as their location

Methods

Results

Conclusion