This award renews support of Chemistry in Solution and at Interfaces (CSI), a Computational Chemical Sciences Center that unites scientists at four universities, i.e., Princeton, Temple, Rutgers-Newark, and Stony Brook, with the mission of developing, validating, and applying novel multi-scale methods for simulating chemical phenomena in fluid and interfacial environments. These processes, playing an essential role in life and energy applications, involve a vast range of size and time scales going from the local dynamics of the chemical bonds to the collective motions of the large assemblies of atoms behind macroscopic materials changes. To model these phenomena, CSI developed a suite of software tools based on artificial intelligence methods, such as machine learning, to learn the chemical interactions from costly quantum mechanical calculations on relatively small model systems, and deep neural networks to describe these interactions in large molecular environments. These tools make possible large-scale molecular dynamics simulations that retain the accuracy of quantum mechanics needed to model chemical phenomena. In the new three-year funding period, these tools will be further optimized and extended with new capabilities, notably for the modeling of electron transfer processes and for the development of reduced models of macromolecular systems. As parallel efforts, the accuracy of the quantum mechanical training data will be improved by using machine learning to build better approximations for the dependence of the quantum mechanical energy on the electron density, and by treating small atomic clusters embedded in extended media with high-level quantum chemistry methods. The codes developed at CSI are designed to run efficiently on graphic processing units (GPUs) and can harness the computational power of the huge arrays of GPUs available in the most advanced supercomputers of the DOE national laboratories, including future exascale platforms. The simulation toolkit will be used at CSI in application studies of interest to the DOE. These will include atomistic simulations of heterogeneous ice nucleation at conditions relevant to climate modeling, atomistic simulations of electrolyte solutions in contact with an electrode with or without applied bias, relevant to chemical processes in batteries, atomistic simulations of speciation and mineralization of CO₂ in salt water, relevant to modeling CO₂ sequestration processes, and atomistic and coarse-grained simulations of polymers in solution, relevant to water purification and/or fuel-cell membranes. The integrated set of software tools developed at CSI will be made available to the scientific community in the open-source repository GitHub, and the center will run annual hands-on-workshops to train students and postdocs on the use of the new molecular simulation tools. To contribute to the scientific training of students from underrepresented minorities, CSI will institute summer internships for undergraduate students of minority serving institutions.