Speeding up crystal structure analysis

 

Researchers at the University of California San Diego have developed a computer-based method that could make it less labour-intensive to determine the crystal structures of various materials and molecules, including alloys, proteins and pharmaceuticals.

 

The method uses a machine learning algorithm, similar to the type used in facial recognition, to independently analyse electron diffraction patterns, and do so with at least 95% accuracy.

 

The algorithm, developed by UCSD nanoengineer,Professor Kenneth Vecchio and colleagues, works with a scanning electron microscope equipped with electron backscatter diffraction.

 

A Vecchio highlights, SEM-based EBSD can be performed on large samples and analysed at multiple length scales, providing local sub-micron information mapped to centimetre scales.

 

But while these systems can determine fine-scale grain structures, crystal orientations, relative residual stress or strain, software can't yet analyse the atomic structure of the crystalline lattices present within a material.

 

Given this, this approach demands structural guesses and user input that can be time-consuming and incorrect.

 

To solve this problem, Vecchio and colleagues developed a method that uses a convolutional neural network to automatically determine the crystal structure quickly and with high accuracy.

 

The deep neural network independently analyses each diffraction pattern from a material to determine the crystal lattice, out of all possible lattice structure types, with a high degree of accuracy (greater than 95%).

 

Crucially, the method eliminates a lot of the guesswork from crystal structure determination.

 

The researchers reckon a wide range of research areas including pharmacology, structural biology, and geology are expected to benefit from using similar automated algorithms to reduce the amount of time required for crystal structural identification.

 

Research is published in Science.