Real Time Predictions of VGF-GaAs Growth Dynamics by LSTM Neural Networks

The aim of this study was to assess the aptitude of the recurrent Long Short-Term Memory (LSTM) neural networks for fast and accurate predictions of process dynamics in vertical-gradient-freeze growth of gallium arsenide crystals (VGF-GaAs) using datasets generated by numerical transient simulations. Real time predictions of the temperatures and solid–liquid interface position in GaAs are crucial for control applications and for process visualization, i.e., for generation of digital twins. In the reported study, an LSTM network was trained on 1950 datasets with 2 external inputs and 6 outputs. Based on network performance criteria and training results, LSTMs showed the very accurate predictions of the VGF-GaAs growth process with median root-mean-square-error (RMSE) values of 2 × 10−3. This deep learning method achieved a superior predictive accuracy and timeliness compared with more traditional Nonlinear AutoRegressive eXogenous (NARX) recurrent networks.

Application study of Mn-doped PIN-PMN-PT relaxor ferroelectric crystal grown by Vertical Gradient Freeze method

A 3-inch Mn:PIN-36%PMN-32%PT (Generation III) relaxor ferroelectric crystal was grown using the Vertical Gradient Freeze method. Crystals of [1, 11] and [111] orientations were prepared and studied, evaluating their dielectric constants, piezoelectric coefficients, electromechanical coupling coefficients, loss tangents and pyroelectric coefficients. It is shown that in pyroelectric applications, [111]-poled crystals are particularly suitable for high performance sensors due to a relatively high pyroelectric coefficient, low loss tangent. In addition, the specific heat is lower as compared to LiTaO3. It is also shown that the alternating current (AC) poling has a more significant effect on the [001]-poled crystal than on the [011]- and [111]-poled crystals due to the lattice distortion induced by a strong electric field along the [001] direction.