Novel machine learning-based prediction approach for nanoindentation load-deformation in a thin film: Applications to electronic industries. (December 2022)

• Record Type: Journal Article
• Authors: Laxmikant Vajire, Sujal; Prashant Singh, Abhishek; Kumar Saini, Dinesh; Kumar Mukhopadhyay, Anoop; Singh, Kulwant; Mishra, Dhaneshwar
• Is Part Of: Computers & industrial engineering. Volume 174(2022)
• Publication Date: 2022-12
• Subjects: Electronic devices -- Misfit strain -- Nanoindentation -- Thin film on a thick substrate -- Machine learning model -- Polynomial regression -- Deformation pattern; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2022.108824
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: Correct electronic device performance prediction by correctly predicting local deformation pattern. Lattice mismatch, and thermal expansion coefficient mismatch between the film, and the substrate. Nanoindentation-based load deformation investigations in presence of the misfit strain. Machine learning models to predict load deformation in a film deposited on a thick substrate. Can contribute to reduce the overall cost of electronic devices by reducing the requirements of experimental studies. Abstract: Electronics industries, notably device fabrication industries, deposit thin films on thick substrates to create various devices such as sensors, actuators, LED Multi-Quantum Well (MQW), and High Electron Mobility Transistors (HEMT). The lattice and thermal expansion coefficient mismatch between the thin film and the substrates causes misfit strain generation. The misfit strain can not only cause defect formation but can also change the local deformation pattern. The altered deformation pattern can have an impact on electronic device performance. Therefore, the load-deformation behaviour of thin films deposited on a thick substrate has been investigated using the nanoindentation modelling approach. Two alternative modelling strategies have been employed. The finite element analysis-based nanoindentation simulation was carried out to evaluate the load required for a fixed deformation in the thin film with changing misfit strain between the thin film and the substrate. The next stage was further predicting the load required for indentation deformation in the thin film at higher misfit strain using a machine learning-based linear regression model. Gallium nitride thin film layer on silicon and sapphire substrate were considered as case studies for this purpose. It was discovered that as the misfit strain between the substrate and the thin film increases, so does the thin film's elastic recovery and brittleness. Both the FEM and the machine learning model results were validated against published experimental findings available in the literature. The machine learning model presented in this work can be utilized to evaluate the performance of devices like MEMS, NEMS, LEDs, etc. without carrying out iterative nanoindentation experiments or FE simulations, which will aid in reducing the overall cost of these optoelectronic devices and increase the overall profit of the electronic industries.