Mar 27 2020
Nanostructured layers exhibit unlimited promising characteristics, but how should we identify the most appropriate one without using any time-intensive experiments?
A look into the sputtering system where nanostructured layers are generated. Image Credit: © Lars Banko.
A team of researchers from the Materials Discovery Department at Ruhr-Universität Bochum (RUB) has identified a shortcut method—with the help of a machine learning algorithm. Using this method, the team was able to consistently estimate the characteristics of a nanostructured layer. The researchers’ report was published in the new Communications Materials journal on March 26th, 2020.
Porous or Dense, Columns or Fibers
During the development of thin films, the condition of the surface and, thus, its properties are established by various control variables. Pertinent factors include the layer’s composition and also the process conditions at the time of its formation, for example, temperature.
The combination of all these elements leads to the formation of either a dense or porous layer at the time of the coating process, while atoms integrate to create fibers or columns.
In order to find the optimal parameters for an application, it used to be necessary to conduct countless experiments under different conditions and with different compositions; this is an incredibly complex process.
Alfred Ludwig, Professor and Head, Materials Discovery and Interfaces, Ruhr-University Bochum
Such experiments led to findings otherwise called structure zone diagrams. These diagrams make it possible to read the surface of a specific composition ensuing from specific process parameters.
Experienced researchers can subsequently use such a diagram to identify the most suitable location for an application and derive the parameters necessary for producing the suitable layer. The entire process requires an enormous effort and is highly time consuming.
Alfred Ludwig, Professor and Head, Materials Discovery and Interfaces, Ruhr-University Bochum
Algorithm Predicts Surface
In an attempt to identify a shortcut approach to realize the optimal material, the researchers leveraged artificial intelligence—to be more specific, machine learning. To this end, Lars Banko, a PhD researcher, along with collaborators from the Interdisciplinary Centre for Advanced Materials Simulation (ICAMS) at RUB, altered a so-called generative model.
Then using certain process parameters, Banko trained this algorithm to produce images of the surface of a fully researched model layer of chromium, aluminum, and nitrogen to predict how the layer would appear under the respective conditions.
We fed the algorithm with a sufficient amount of experimental data in order to train it, but not with all known data.
Lars Banko, PhD Researcher, Materials Discovery and Interfaces, Ruhr-University Bochum
The scientists were thus able to compare the outcomes of the calculations against those obtained from the experiments and checked how reliable its prediction would be.
There were conclusive results: “We combined five parameters and were able to look in five directions simultaneously using the algorithm—without having to conduct any experiments at all,” emphasized Alfred Ludwig. “We have thus shown that machine learning methods can be transferred to materials research and can help to develop new materials for specific purposes.”
The study was financially supported by the German Research Foundation under the umbrella of Collaborative Research Centre/Transregio 87 “Pulsed high power plasmas for the synthesis of nanostructural functional layers,” subproject C2.