Beyond Ion Dynamics: Efficient Charge Transport Simulations including Electrons at Battery Scales
ORAL · Invited
Abstract
Polarons, i.e., localized excess charges stabilized by lattice distortions, have long been recognized as fundamental to charge transport in energy materials. Nonetheless, a quantitative modeling of polaron dynamics has, so far, remained largely elusive in such materials. On the one hand, the activated dynamics of polarons requires time and length scales that are inaccessible to first-principles methods. On the other hand, off-the-shelf force-field models and machine learned interatomic potentials (MLIPs) do not capture such electronic degrees of freedom.
In this work, we overcome this hurdle proposing an MLIP model that explicitly accounts for these electronic degrees of freedom. To this end, the excess charge is incorporated into the MLIP as an independent, semi-classical degree of freedom that adiabatically follows the nuclear motion. We benchmark and validate the approach by training the proposed MLIP architecture to density-functional theory data obtained with the hybrid HSE06 functional using bismuth vanadate (BiVO4) as example. Furthermore, we demonstrate the power of the approach for lithium titanium oxide (Li4Ti5O12 LTO), a prototypical anode material, for which polarons are known to play a key role [1]. For this challenging LTO case, we explore the PES for ionic diffusion using saddle-point search methods for different polaronic concentrations. By this means, we identify those barriers that are most affected by polarons. For these, we additionally perform accelerated free-energy calculations via umbrella sampling to shed light on the underlying mechanism that couples polaron and ion dynamics. This reveals that polarons in LTO do not merely serve as spectators, but lower the effective Li-diffusion barrier by thermodynamically adapting to the much slower ionic motion, resulting in an increase of conductivity by up to two orders of magnitude at room temperature. This is further substantiated by large-scale molecular-dynamics simulations, which also yield ionic and polaronic diffusion coefficients in line with experimental measurements. For the first time, this theoretically corroborates the occurrence of a correlated polaron-ion dynamics with profound implications for the design of energy materials.
[1] M. Kick, C. Scheurer, and H. Oberhofer, ACS Appl. Energy Mater. 4, 8583 (2021).
In this work, we overcome this hurdle proposing an MLIP model that explicitly accounts for these electronic degrees of freedom. To this end, the excess charge is incorporated into the MLIP as an independent, semi-classical degree of freedom that adiabatically follows the nuclear motion. We benchmark and validate the approach by training the proposed MLIP architecture to density-functional theory data obtained with the hybrid HSE06 functional using bismuth vanadate (BiVO4) as example. Furthermore, we demonstrate the power of the approach for lithium titanium oxide (Li4Ti5O12 LTO), a prototypical anode material, for which polarons are known to play a key role [1]. For this challenging LTO case, we explore the PES for ionic diffusion using saddle-point search methods for different polaronic concentrations. By this means, we identify those barriers that are most affected by polarons. For these, we additionally perform accelerated free-energy calculations via umbrella sampling to shed light on the underlying mechanism that couples polaron and ion dynamics. This reveals that polarons in LTO do not merely serve as spectators, but lower the effective Li-diffusion barrier by thermodynamically adapting to the much slower ionic motion, resulting in an increase of conductivity by up to two orders of magnitude at room temperature. This is further substantiated by large-scale molecular-dynamics simulations, which also yield ionic and polaronic diffusion coefficients in line with experimental measurements. For the first time, this theoretically corroborates the occurrence of a correlated polaron-ion dynamics with profound implications for the design of energy materials.
[1] M. Kick, C. Scheurer, and H. Oberhofer, ACS Appl. Energy Mater. 4, 8583 (2021).
–
Presenters
-
Christian Carbogno
- Fritz Haber Institute of the Max Planck Society