Towards quantitative mapping of physical and chemical properties of materials using unsupervised data clustering methods

T. De Muijlder, P. Leclère

Laboratory for Physics of Nanomaterials and Energy (LPNE), Research Institute for Materials Science and Engineering
University of Mons (UMONS), B 7000 Mons, Belgique.
thomas.demuijlder@umons.ac.be

It is now possible to characterize the mechanical, electrical, chemical, and thermal properties of a sample at the nanoscale, with increasingly short measurement times. Indeed, the development of Scanning Probe Microscopy (SPM) techniques and the numerous acquisition modes, gives us access to very large amounts of data. Once acquired, these data must be analyzed and described and given their number, using statistical tools is needed.

In our research, we have written a program (coded in Python) to adapt known statistical tools to SPM data. We focused on unsupervised data clustering using KMeans or Gaussian Mixed Model (GMM) algorithms to identify the different populations present in SPM mappings. We also developed tools to reconstruct mapping of a sample mechanical properties by using the best suited contact mechanics model [1] for each pixel of the dataset.

The program has been designed to provide the user with a simple and fast analysis procedure for multidimensional detection of the different areas of the samples in a user independent manner. In order to facilitate the use of the program for people who are not familiar with programming, we have added a graphical interface to the program.

The program is currently used for the analysis of our data in the laboratory on different types of materials (polymer blends (Figure 1), nanocomposites, nanodielectrics, hydrogels, piezoelectric materials, or adhesives of biological origin [2]).

We are also working on a supervised machine learning algorithm that can sort the data depending on its quality and the post-treatment needed before the analysis.


 

[1] Piétrement, O.; Troyon, M. Journal of Colloid and Interface Science 2000, 226 (1), 166–171.

[2] Lefevre, M.; Tran, T. Q.; De Muijlder, T.; Pittenger, B.; Flammang, P.; Hennebert, E.; Leclère, Ph., Front. Mech Eng. 2021, 7.