Alessandra Serva

Machine-learned interatomic potentials hold the promise to enable the modeling of highly concentrated liquids over meaningful time scales, far from reach for current ab initio electronic structure methods. Here we evaluate the performances of various MACE potentials in modeling a 21 m water-in-salt electrolyte based on lithium bis(trifluoromethanesulfonyl)imide. We test out-of-the-box foundation models, as well as both fine-tuning and from-scratch training strategies. Our simulations demonstrate that surrogate models allow us to overcome sampling limitations of ab initio molecular dynamics, reaching an excellent agreement with experimental observables such as the structure factor. We also demonstrate the benefit of fine-tuning a foundation model over training from scratch, in terms of data efficiency, but most importantly as a means to provide information regarding configurations hard to sample, such as short Li+–Li+ distances. Finally, we show that depending on the reference exchange-correlation functional, empirical dispersion correction schemes can be detrimental. All in all, our work shows that machine-learned interatomic potentials are a good fit for the modeling of highly concentrated electrolytes over long time scales.

Many natural phenomena involve processes that happen simultaneously at different characteristic length- and timescales. Typically, the region where the process of interest happens is affected by fluctuations in its surroundings. Modeling these systems requires an effective combination of different simulation resolutions. The Hamiltonian-adaptive resolution simulation (H-AdResS) method allows us to model dual-resolution systems in length- and timescales compatible with molecular diffusion, by combining atomistic and particle-based coarse-grained models in the same simulation box. In this work, a new implementation of H-AdResS is provided in LAMMPS 2023. The new features extend the usage to more diverse interaction potentials and simplify the preparation of input files via dedicated LAMMPS input commands, while keeping the efficiency gain of the basis method. The implementation is benchmarked by reproducing water properties from a reference atomistic simulation. Importantly, the new implementation includes changes in compensation routines allowing simulation of systems with fluctuating density. As an example, the method in its new implementation is applied to model a porous metal–organic framework and its gas adsorption structure and transport properties. We demonstrate that structural and dynamic properties in the atomistic region of the dual-resolution scheme are unaffected and remain those of the fully atomistic system, while increasing simulation efficiency.

Localized high concentration electrolytes (LHCEs) have emerged as a transformative choice of electrolytes, simultaneously exhibiting good electrochemical stability, sufficiently fast ionic transport, and improved rate capability. The design principle of LHCEs is to dilute the conventional electrolyte with a non-solvating diluent, thereby forming extended ionic networks while keeping the salt concentration sufficiently low. Several fluorinated cyclic ethers have recently been reported as particularly promising diluents. To better understand the success of these diluents, we perform ab initio molecular dynamics simulations of LHCEs based on 2,2-bis(trifluoromethyl)-1,3-dioxolane (BTFMD) diluent, 1,2-dimethoxyethane (DME) solvent, and lithium bis(fluorosulfonyl)imide (LiFSI) salt. The Li ⁺ ions are found to be coordinated soley by FSI ^– anions and DME molecules, whereas the concentration of the diluent dictates the competition between the two coordinating species in the Li ⁺ first solvation shell. The Li ⁺ ion solvation environment is linked to the electronic structure of the electrolyte: a higher concentration of the diluent decreases the density of electronic states in the vicinity of the valence band maxima, which arises from the depletion of non-solvating DME molecules, while no significant differences are observed in the conduction band minima. This observation is conceptually related to the improved oxidative stability measured experimentally and the chemical bonding in these systems. Our findings rationalize the performance of BTFMD-based LHCEs and suggest the tunability of electrochemical responses by controlling the local Li ⁺ structure.

Tuning the ratio between weakly (TFSI ⁻ ) and strongly (NO ₃ ⁻ ) associating anions in mixed-anion battery electrolytes modulates the transport properties, driven by significant changes in Li ⁺ solvation dynamics and the electrolyte structure.

Supercapacitors are energy storage devices with high power density and long cycle life. Combined with spectroscopy and electrochemistry, molecular simulations and theory have allowed to characterize their charging mechanisms, and we now have an excellent understanding of the effect of properties such as nanopore size or structural disorder on the supercapacitor performances. However, the influence of electrode flexibility remains to be addressed as state-of-the-art models focused on the polarization effects of the electrode, but enforced its structural rigidity, as an approximation. Here we overcome this limitation by integrating a constant-potential molecular dynamics scheme with a state-of-the-art machine-learning potential for carbon, while controlling the applied potential. Using nanoporous sp²/sp³ carbon electrodes filled with an ionic liquid electrolyte, we compare the behavior of the rigidified and flexible frameworks; the latter allowing for local atomic relaxation, breathing modes, etc. We demonstrate that flexibility significantly enhances in-pore ionic diffusivity, thus shortening the characteristic charging time by a factor of 3 relative to the rigid analogue, while the specific capacitance remains in the experimental range (≈140 F·g^-1) for both cases. More specifically, our analyses demonstrate that flexibility accelerates co-ion expulsion, mitigates pore overcrowding, and promotes a homogeneous induced-charge profile that penetrates deeply from the very start of the applied voltage onset, explaining the improved kinetics.

Liquid battery electrolytes based on mixtures of salts with weakly and strongly associating anions have emerged as a promising route toward high-performance, sustainable battery technologies. Their success is primarily attributed to the unique influence of salt composition on the solvation structure. Here, we employ classical molecular dynamics simulations, corroborated by experimental data, to study mixed lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) / lithium nitrate (LiNO3) in diglyme electrolytes, a formulation of particular interest for lithium–sulfur and lithium-oxygen batteries. We investigate how the ratio of weakly associating anions (TFSI–) to strongly associating anions (NO3–) affects ion transport within the electrolyte. Our findings reveal that the anion ratio significantly impacts both the solvation structure and the solvation dynamics, which together contribute to the distinct transport behavior observed in these systems. These findings underscore the tunability of battery electrolyte transport properties through careful mixing of anions.

Supercapacitors are energy storage devices with high power density and long cycle life. Combined with spectroscopy and electrochemistry, molecular simulations and theory have allowed to characterize their charging mechanisms, and we now have an excellent understanding of the effect of properties such as nanopore size or structural disorder on the supercapacitor performances. However, the influence of electrode flexibility remains to be addressed as state-of-the-art models focused on the polarization effects of the electrode, but enforced its structural rigidity, as an approximation. Here we overcome this limitation by integrating a constant‑potential molecular dynamics scheme with a state‑of‑the‑art machine‑learning potential for carbon, while controlling the applied potential. Using nanoporous sp$^2$/sp$^3$ carbon electrodes filled with an ionic liquid electrolyte, we compare the behavior of the rigidified and flexible frameworks; the latter allowing for local atomic relaxation, breathing modes, etc. We demonstrate that flexibility significantly enhances in-pore ionic diffusivity, thus shortening the characteristic charging time by a factor of three relative to the rigid analogue, while the specific capacitance remains in the experimental range ($\approx 140~ \text{F} \cdot \text{g}^{-1}$) for both cases. More specifically, our analyses demonstrate that flexibility accelerates co‑ion expulsion, mitigates pore overcrowding, and promotes a homogeneous induced‑charge profile that penetrates deeply from the very start of the bias onset, explaining the improved kinetics. By explicitly linking the electrochemical response to mechanical and vibrational constraints, this work paves the way toward rational and optimal energy storage device design.

Protic ionic liquids, PILs, are promising materials for energy storage applications, in part due to their ability to decouple proton transport from ion diffusion. In this work, the proton transfer mechanism in 1-ethylimidazolium bis(trifluoromethanesulfonyl)imide ([HEIM][TFSI]) IL was studied by means of three different computational approaches. Classical polarizable molecular dynamics simulations were used to explore the structure and dynamics of the fully ionized system, while Density Functional Theory calculations were carried out to estimate the energy barriers for the different proton transfer reactions. Finally, the proton transfer was explicitly studied by means of Neural Network Force Field simulations. Our results show that this reaction is indeed possible when doping the IL with an excess of deprotonated cations, and highlight the importance of the formation of dimers between donor and acceptor species for the proton transfer to occur. The main driving factor for the reaction was found to be the energy cost for reaching a suitable coordination environment and form such dimers, which is higher than that for the transfer reaction.