A study of acousto-electron effects in semiconductor quantum dots and bionanocomplexes based on them using deep machine learning

An artificial neural network architecture has been developed that is capable of predicting changes in the energy spectrum of semiconductor quantum dots and their bionanocomplexes under the influence of an acoustic wave and interaction with human serum albumin molecules based on its specified geometric sizes, elastic constants and deformation potential constants of the allowed energy bands of quantum dot materials, as well as the frequency and amplitude of the ultrasonic wave and the surface concentration of human serum albumin. Two artificial neural network architectures have been implemented and optimized. Both models contain three hidden layers; however, the second architecture involves expanding the input space by adding harmonic functions. Both approaches to neural network modeling demonstrate good agreement with the results of mathematical modelling, provided that the input parameters lie within the training range. In the case when the input parameters lie outside the training sample, the model using additional harmonic functions at the input demonstrates a much better result. Within the framework of the developed model for the CdSe/ZnS/CdS/ZnS QD–human serum albumin bionanocomplex, the dependence of the energy shift of the radiation quantum on the frequency of the acoustic wave was investigated for different values ​​of the surface concentration of albumin and different geometric sizes of the core and shell of the quantum dot.