By Konstantinos Andrikos, Nanomechanic Engineer

Testing materials’ mechanical properties has always been a matter of great importance for the evaluation of their robustness, wear resistance, but also stability index, failure prediction and materials’design, to meet the demanding requirements of manufacturing and product trends; reliability, performance, durability, quality, to name a few. In order to have a fine resolution of a material’s properties, researchers and scientists have been always trying to characterize materials at the lowest scale possible, not without challenges though.

One way to extract mechanical properties at the nanoscale regime, is to poke a material’s surface with a few nanometers wide sharp tip. During this “nanoindentation technique” the applied load is being measured as a function of the tip’s penetration depth. Based on the acquired force-displacement data, we are able to directly calculate the modulus of elasticity (E) and the hardness (H) of the indented point. From a model that started all, various modifications have now been introduced to take into account phenomena that appear on non-ideal, pragmatic cases of e.g. complex shapes, non-ambient environments etc..

The idea is to apply this methodology to a grid of indentation-points in order to quantitatively map the nanomechanical properties of the selected area.  This area may include points of different mechanical properties, such as reinforcements or constituent phases, thus we end up with a distribution of the measured quantities (H, E). The next step is the analysis of these results in order to gain some insights about the nanomechanical properties of the material under test.

The statistical character of nanoindentation is a key factor to supply information on heterogeneity of the indented material. This is where machine learning clustering algorithms are being utilized as a tool to process information and extract relations inherent in data. More specifically, clustering algorithms are used for the identification of groups of data with similar properties. The goal of this approach is to create groups of indents which we can later link each one to a specific cluster (material phase, reinforcement, matrix, interphase, etc.).

Nevertheless, due to the nature of the experiment many challenges arise such as phenomena associated with “pile-up” or “sink-in”. These terms refer to dislocations taking place during the indentation, in which material is concentrated and move upwards (pile-up) or downward (sink-in) around the indenter. At the one hand these phenomena result to false estimation of hardness and elastic modulus, but at the other hand it provides additional material information. The quantification of pile-up or sink-in height by means of microscopy methods (SEM or AFM), is a valuable information which can be utilized by predictive models.

Nanoindentation is a powerful tool in our disposal for materials characterization at the nanoscale regime. Its capabilities are further pushed when it is combined with statistical methods for extracting insights from data (machine learning algorithms) and with microscopy methods which will provide additional information about the topology of the indented area.

Thanks to Nanomecommons project (supported by the European Union under the HORIZON2020 Framework Programme Grant Agreement no. 952869) IRES is collaborating with partners involved for coupling nanoindentation and machine learning towards the establishment of automated methods for materials nanomechanical characterization. It is a challenging yet extremely useful tool since it could enhance our understanding for material properties and thus, provide means for the design of better materials with tailored properties.

More to come, with use cases this time..