Speaker
Description
Development of a new material with required properties is a very complex task. Theoretical
models for growth procedure are usually not available, at least at the beginning. Therefore,
many attempts are made to achieve required properties of the material and many
characterization datasets are obtained. However, the way how the physical properties of the
material are affected by the growth conditions does not have to be straightforwardly evident.
For such a case, machine learning can be very helpful. In this contribution, applications of
several simple machine learning approaches are applied to the development process of the
InGaN/GaN scintillator structure. A properly trained neural network is capable to predict
luminescence properties from the growth parameters of the structure. This enables
optimization of the growth parameters from empirical data only. On the other hand,
understanding of underlying physics is not guaranteed but the predictions of the model can
give a clue.
