First-Principles Modeling of Transport Mechanisms in Carbonate–Hydroxide Electrolytes

Abstract

We performed ab initio molecular dynamics simulations of a molten [Li 0.6 K 0.4 ] 3 CO 3 OH electrolyte containing dissolved CO 2 and confirmed the presence of pyrocarbonate, bicarbonate, and water along with the constituent ions and molecular CO 2 . Our calculations indicate kinetics-driven formation of pyrocarbonate whereas bicarbonate and water are thermodynamically favored. Our results also demonstrate the presence of water at higher concentrations (double or more) than that of CO 2 , which reinforces the conclusions in our earlier work [AIChE J. 2020, e16988] based on chemical reaction equilibrium simulations. Structural analysis indicates a larger distortion in water geometry, due to its higher polarizability compared to the nonpolar CO 2 , explaining the higher reactivity and smaller average lifetime of H 2 O in the melt. The computed lifetime distributions of the reaction products reveal that the bicarbonate ion lives the shortest among all the species present in the system. It initiates a sequence of successive proton exchange events; such sequences of exchanges along a hydrogen-bonded network gives the Grotthuss mechanism for proton transport in liquid water. The estimated proton diffusion, based on a random walk model, is about 30 times faster than the hydroxide diffusion obtained from classical molecular dynamics simulations. We believe that the presence of proton transfer events in the system has a large impact on the overall ion dynamics and electrical conductivity of the medium.

Modeling Chemical Reactions in Alkali Carbonate–Hydroxide Electrolytes with Deep Learning Potentials

“…Training a robust DNN model capable of capturing a range of local atomic environments requires a data set that sufficiently covers the breadth of configurations that may manifest during finite-temperature MD simulations. As such, data for molten false[ Li 0.6 normalK 0.4 false] 3 CO 3 OH electrolyte containing dissolved CO 2 configurations were acquired from previously obtained AIMD data with simulation details as described in ref . Briefly, the first-principles calculations were performed at the DFT level , as implemented in the CP2K simulation program using the Quickstep module .…”

“…The scalability of the DNN model offers the advantage to examine such chemical reactions on a much longer time scale and large system sizes over the conventional AIMD approach, which typically involves trajectories of a few hundreds of picoseconds using moderate-size systems. 25 Our 2.2.1) have been rectified through the help of targeted learning by incorporating more training data representing the potential energy surface of CO 2 dissociation. Moreover, one of the limiting factors of the active learning approach is the short time scale which prevents access to events that may appear at a longer time scale enabled by NNs.…”

“…(3) (4) (5) In principle, OH − ions participate in all of these reactions, and eventually, all the H−O covalent bonds originally present in the hydroxide ions are broken within 1.5 ns, which is once again following our previous prediction based on the AIMD results. 25 Furthermore, this observation substantiates the AIMD anticipated thermodynamic nature of water and bicarbonate (HCO 3 − ) ion formation in the electrolyte. We analyzed the trajectories obtained from DPMD simulations of Type II systems to investigate the equilibrium composition of the electrolyte.…”

“…This particular feature cements the fact that, in this kind of eutectic mixture, water is a more stable species compared to CO 2 , as concluded based on our previous results from chemical reaction equilibrium simulations 11 and also inferred from AIMD trajectories. 25 In summary, these quantitative estimations provide us with an excellent insight into such a complex chemical environment. Furthermore, it is only possible because of the DNN potential to study the chemical reactivity involving large system sizes and a longer time scale to reach an equilibrium.…”

“…For example, by employing AIMD simulations, Corradini et al reported the behavior of CO 2 in molten CaCO 3 where they observed the formation of a pyrocarbonate anion via a spontaneous reaction between CO 2 and a carbonate anion . Recently, we investigated an MCFC electrolyte consisting of CO 2 and molten carbonate-hydroxide via an AIMD approach . We observed formation of several ionic species, including pyrocarbonate, bicarbonate, etc., primarily driven by the gas-phase concentrations of CO 2 , H 2 O, and O 2 . , Our results demonstrated the presence of water at concentrations double or more than that of CO 2 .…”