NOMATEN scientists show how to link the material’s structure with the properties thanks to materials informatics tools

• Materials Informatics – Structure and Function (MASIF) research group at NOMATEN investigates the connection between the material’s structure and properties
• The data produced during the experiment and simulations are being processed with the materials informatics tools
• The up to date scientific papers review  has been recently published by the MASIF group: Materials Informatics for Mechanical Deformation: A Review of Applications and Challenges (Materials 2021, 14(19), 5764).
• This research group is managed by Stefanos Panikolaou PhD

Material structure studies, such as microscopic measurements, provide enormous amounts of data that can be used to reconstruct the microstructure of a material and become the basis for computer simulations on a molecular scale or larger. Understanding big data requires the use of statistical methods and machine learning, and simulations require efficient techniques to reconstruct the microstructure. Materials subjected to extreme conditions - such as irradiation or high temperatures - experience changes that are difficult to understand using traditionally used models. In such cases, artificial intelligence methods turn out to be irreplaceable in order to capture these changes and relate them to specific processes and physical properties taking place.

Experimental data (e.g. stress-strain curves, electron microscopy images of microstructures, strain maps from digital image correlation) are acquired by other research groups. In the MASIF group, on the other hand, simulations are performed at very different scales of space and time using techniques such as:

• density functional theory (DFT), based on a number of quantum-mechanical methods for modeling the structure of crystals and chemical particles,
• molecular dynamics (MD - molecular dynamics), a computer simulation method that enables the study of the structure of materials, their properties and physical processes taking place in them (thermal conductivity, diffusion, radiation damage, etc.),
• simulations in the micro- and millimeter scale using the Fast Fourier transform (FFT) and the finite element method.

The obtained data sets are then processed using statistical methods (e.g. principal component analysis, PCA - principal component analysis or discrete wavelet transform, DWT - discrete wavelet transform) and artificial intelligence (machine learning, deep learning).

In this way, a lot of useful information can be obtained from existing datasets, which would otherwise be lost - Karol Frydrych PhD  says - For example, using PCA or DWT, on the basis of the deformation maps, it is possible to determine the moment when the material reaches the plastic state, which was described[1] by prof. Mikko Alava, NOMATEN director,, and Stefanos Papanikolaou PhD within the paper„Direct detection of plasticity onset through total-strain profile evolution”.

Thanks to deep learning, it is possible, for example, to find defects in photos taken from an electron microscope or to classify the microstructure of material. In this aspect, we can use materials informatics tools in cooperation with other research groups at NOMATEN – for example materials characterization group managed by Iwona Jóźwik PhD or functional properties group by prof. Łukasz Kurpaska PhD DSc– adds Karol Frydrych PhD.

The up to date scientific papers review  has been recently published by the MASIF group  here: Materials Informatics for Mechanical Deformation: A Review of Applications and Challenges w periodyku Materials (2021, 14(19), 5764)[2].