Evaluation of thin film material properties using a deep nanoindentation and ANN

Due to the substrate effect, there are several difficulties to evaluate the material properties of thin films via nanoindentation. In this study, an inverse analysis method based on an artificial neural network (ANN) is proposed to obtain free-volume-model (FVM) parameters of thin film metallic glass (TFMG) via nanoindentation. Unlike conventional nanoindentation procedures for thin films, a deeper indentation depth (asymptotic to 30% of film thickness) is adopted to accurately identify the film properties even with significant substrate deformations. Both spheroconical and Berkovich tips are employed to ensure unique solutions. A complex mapping function of inverse analysis is replaced by establishing ANN between nanoindentation (features) and material (targets) parameters. The established ANN model is trained with the database generated via systematic finite element analyses (FEA). The trained ANN model is experimentally validated by estimating the material properties of Zr55Cu30Ag15 TFMG deposited on two different substrates (Si and soda lime glass). The maximum difference of plastic indentation energy between experiments and FEA values using the estimated film material properties was within 3%. (C) 2022 The Author(s). Published by Elsevier Ltd.

Automated summary

In this study, an inverse analysis method based on an artificial neural network (ANN) is proposed to evaluate the material properties of thin film metallic glass (TFMG) via nanoindentation. By using a deeper indentation depth and employing different indenters, the proposed method can accurately identify the film properties and ensure unique solutions. The established ANN model, trained with a systematic finite element analysis (FEA) database, is experimentally validated and shows good accuracy in estimating the material properties.