Ab-initio and Two-layer Graph Atomic Cluster Expansion (GRACE-2L) Foundational Machine Learning Interatomic Potential Modeling of Wurtzite AlScN and AlBN Piezoelectric Systems for Predicting their Temperature-dependent Electronic Properties in Thin Film Bulk Acoustic Wave Devices
Presented by Dr. Ivo Koutsaroff
In the present study is focused on the properties of acoustic wave devices. We employ ab-initio simulations to calculate the full set of piezoelectric, mechanical, and acoustic properties of Al1−xScxN (0≤x≤0.5, AlScN) and Al1−xBxN (0≤x≤0.5, AlBN), incorporating temperature affects from 200K to 500K [1, 2]. The measured phase velocities and piezoelectric coupling coefficients, kt2 from Al1−xScxN (0.31≤x≤0.38, AlScN) based Thin Film Bulk Acoustic Wave Devices at 5-7 GHz allowed good consistency when comparing with ab-initio simulated piezoelectric and stiffness tensors. Two-layer Graph Atomic Cluster Expansion (GRACE-2L) Foundational Machine Learning Interatomic Potential trained on the Meta Open Materials 2024 (OMat24) dataset, was utilized in predicting e33, k2, C33 and lattice densities from 200K to 500K temperature range for Al1−xBxN (0≤x≤0.5, AlBN).
3531098543052973660
This Session Will Cover
Ab-initio modeling of AlScN and AlBN using DFT and VASP 6.5.1.
Temperature-dependent piezoelectric, mechanical, and acoustic properties.
Comparison of simulated results with experimental measurements from Thin Film Bulk Acoustic Wave (BAW) devices operating at 5–7 GHz.
Application of GRACE-2L machine learning interatomic potentials to predict key AlBN material properties from 200 K to 500 K.
Integrating first-principles simulations, experimental validation, and machine learning for advanced materials research.
Who Should Attend?
This webinar is designed for materials scientists, computational physicists, electrical engineers, and researchers working in piezoelectric materials, semiconductors, RF technologies, acoustic devices, and AI-driven materials modeling.
Join us to explore how first-principles simulations and machine learning can support the prediction and understanding of advanced piezoelectric materials for next-generation acoustic devices.