Hybrid First-Principle/Neural Network Correlations for Thermoelectric Transport Coefficients in Gold-Silver Solutions from Bulk to Nanometer Scale

State of art estimation methods were revisited to build hybrid first-principle/artificial neural network correlations to capture the impact of solute concentration, specimen sizes down to nanometer scale, and electron and phonon temperatures in (non)equilibrium for the electric and thermal transport coefficients in gold-silver mixtures at temperatures above the metals Debye temperatures. Deviations with respect to Matthiessen’s additivity rule of both electric and electronic thermal transport coefficients were approximated by means of two neural network correlations as a function of silver atom fraction and temperature. The hybrid approach was confronted and validated against a large repository of data recommended for gold-silver transport properties encompassing pure metals and the full binary-solution composition range. Sensitivity of electric and thermal conductivities in gold-silver mixtures to electron and phonon temperatures, nanoparticle sizes and silver contamination was also discussed in the developed frame. The developed correlations will be useful for estimation of transport properties in areas as diverse as catalysis, electrochemical dissolution and gold nanomaterial synthesis