Machine Learning Analysis of the Thermodynamic Responses of In Situ Dielectric Spectroscopy Data in Amino Acids and Inorganic Electrolytes.

Dielectric spectroscopy (DS) can be a robust in situ technique for geochemical applications. In this study, we applied deep-learning techniques to DS measurement data to enable rapid science interrogation and identification of electrolyte solutions containing salts and amino acids over a wide temperature range (20 to -60 °C). For the purpose of searching for signs of life, detecting amino acids is a fundamental high priority for field and planetary instruments as amino acids are one of the building blocks for life as we know it. A convolutional neural network (CNN) with channel-wise one-dimensional filters is proposed to fulfill the task, using the DS data of amino acid and inorganic salt solutions. Experimental results show that the CNN with two convolutional layers and one fully connected layer can effectively differentiate solutions containing amino acids from those containing salts in both the liquid and solid (water ice) states. To complement the experimental measurements and CNN analysis, the diffusive behaviors of ions (K+, Cl-, and OH-) were further discussed with atomistic molecular dynamics simulations performed in this work as well as the quantum simulation published in the literature. Combining DS with machine-learning techniques and simulations will greatly facilitate more real-time decision-making of mobility systems for future exploratory endeavors in other worlds beyond Earth.