Session T10: Modeling the Electrochemical Interface I

Challenges in the first-principles description of electrochemical interfaces

Electrochemical interfaces play a crucial role in many technologically relevant and scientificaly interesting processes related to electrochemical energy conversion and storage. Still, it is fair to say that our atomistic understanding of the structure of such interfaces is still limited, in particular as far as the electric double layer (EDL) at such interfaces is concerned. In this contribution, I will in particular address the challenges associated with an appropriate quantum modeling of the EDL, both from a conceptual and from a computational point of view. I will show how the liquid nature of the electrolyte can be considered in ab initio molecular dynamics calculations [1], also taking the electrode potential into account. On the other hand, I will present examples in which the structure of the electrode surface has been faithfully reproduced as a function of the electrochemical control parameters within a grand-canonical approach [3], but without explicitly considering the electrolyte.  

[1] A. Groß and S. Sakong, Chem. Rev. 122, 10746 (2022).

[2] A. Groß, Curr. Opin. Electrochem. 40, 101345 (2023),

[3] A. Groß, J. Phys. Chem. C 126,  11439 (2023).   

Ca-dimers and solvent layering determine electrochemically active species in Ca(BH4)2 in THF

Divalent ions, such as Mg, Ca, and Zn, are being considered as competitive, safe, and earth-abundant alternatives to Li-ion electrochemistry. However, the challenge remains to match electrode and electrolyte materials that stably cycle with these new formulations, based primarily on controlling interfacial phenomena. We explore the formation of electroactive species in the electrolyte Ca(BH₄)₂ in THF through molecular dynamics simulation and continuum modeling. Free-energy analysis indicates that this electrolyte has a majority population of neutral Ca dimers and monomers, albeit with diverse molecular conformations as revealed by unsupervised learning techniques, but with an order of magnitude lower concentration of possibly electroactive charged species, such as the monocation, CaBH₄⁺, which we show is produced via disproportionation of neutral Ca(BH₄)₂ complexes. Dense layering of THF molecules within 1 nm of the electrode surface  (modeled here using graphite) hinders the approach of reducible species to within 0.6 nm and instead enhances the local concentration of species in a narrow intermediate-density layer from 0.7-0.9 nm. A dramatic increase in the population of charged species within 1 nm of the electrode is induced at negative bias, supplied by local dimer disproportionation, if the concentration is sufficiently high. The consequences for performance and alternative formulations are discussed in light of this molecular-scale insight.   

Proton transfer at the IrO2-water interface from machine learning potentials

Heterogeneous electrocatalysts for the oxygen evolution reaction (OER) operate via inner-sphere processes that are highly sensitive to the composition and structure of the catalyst’s aqueous interface.  We investigate the proton transfer mechanisms at the aqueous interface of rutile IrO2, one of the most efficient catalytic materials for the OER, using large scale molecular dynamics simulations with ab-initio based machine learning potentials. The intrinsic proton affinities of the different IrO2(110) surface sites are characterized by calculating their acid dissociation constants, which yield a point of zero-charge in good agreement with experiments. A large fraction (≈ 80%) of adsorbed water dissociation is predicted, together with a short lifetime (≈ 0.5 ns) of the resulting terminal hydroxyls, due to rapid proton exchanges between adsorbed H2O and OH species at adjacent surface Ir sites. This rapid surface proton transfer supports the suggestion that the rate-determining step in the OER may not involve proton transfer across the double layer into solution, but rather depend on the concentration of oxidized sites formed by the deprotonation of terminal and bridging hydroxyls, as indicated by recent experiments.

    

Title of your abstract: Efficacy of non-metallic cations in electrochemical CO2 reduction on Bi(111) electrode: a first principles study.

While it is known that CO₂ electroreduction necessitates the involvement of metal cations, questions have arisen regarding the potential efficacy of non-metallic cations. To address this inquiry, we have delved into a comparative analysis of the influence of Na⁺, and NH₄⁺ in the conversion of CO₂ into formate (HCOO^-) and CO on a Bi(111) electrode, employing grand canonical density functional theory. We find that the hydrogenation process, in which a proton (H⁺) migrates from the electrode toward CO₂ primarily drives the reduction of CO₂ into formate (HCOO^-) in the presence of cations. Notably, we find that the activation energy barrier for HCOO- formation in the presence of Na⁺ is 0.03 eV and rises to 0.08 eV in the presence of NH₄^⁺. Furthermore, we discover that CO₂ can also preferentially undergo reduction to CO through a proton shuttling process. This process involves the transfer of a proton (H^⁺ ion) from the electrode to the adsorbed CO₂* species through a water molecule, resulting in the formation of the intermediate COOH*. The activation energy barrier for this step is determined to be 1.17 eV and 1.04 eV in the presence of Na⁺ and NH₄⁺, respectively. Finally, our investigation finds that the activation energy barrier for COOH* dissociation is 0.48 eV and 0.04 eV in the presence of Na^⁺ and NH_₄^⁺, respectively. These outcomes suggest that non-metallic cations can be as effective as metallic cations in facilitating the CO₂ electroreduction process, in agreement with experimental findings [1]. 

[1] K. Shi, D. Le, T. Panagiotakopoulos, T. S. Rahman, and X. Feng, Effect of Quaternary Ammonium Cations on CO2 Electroreduction (Submitted, 2023).   

Electrified Gold/Ice interface

Hexagonal ice (ice Ih), also known as ordinary ice, displays a variety of fascinating properties that are significant for the existence of life and the control of Earth's climate. Ice XI is the proton ordered phase of hexagonal ice Ih and it is often referred to as ferroelectric ice. Experimentally, it is difficult to observe the phase transition from ice Ih to ice XI, and  the presence of ionic impurities seems to be a prerequisite to catalyze the order-disorder phase transition. However, the presence of a surface can enhance the symmetry breaking and promote the growth of proton ordered ice layers (ice iX layers) without the need for doping. In this work we analyze the structure and electronic properties of ice Ih and XI at the interface of a gold (Au) electrode, as a function of an external bias potential applied to the electrodes. This is accomplished using a combination of Density Functional Theory (DFT) and non-equilibrium Green's functions methods (NEGF).   

Surface Ferroelectricity in Au-Ice Capacitors

At atmospheric pressure and below 272 K water crystallizes into an hexagonal (wurtzite) proton-disordered phase called ice Ih. At around 72 K, bulk ice Ih undergoes a phase transition to a ferroelectric (proton-ordered) phase known as ice XI. Using density functional theory we show that when hexagonal ice is placed in contact with a metallic interface, the relative surface free energy of the XI versus the Ih phase decreases, inducing a surface ferroelectric phase transition with a skin layer depth-dependent transition temperature higher than that of the bulk. We explain the origin of this enhanced stability and show that our results explain the recent experimental observations by X. Wen et al., who have measured this phase transition to occur at temperatures as high as 163 K for different metal capacitors.   

Picking a Paradigm: Insight vs. Active Machine Learning for NaCl Solvation in Water

It’s no secret that machine learning has exploded in popularity – for use in both the scientific disciplines as well as industry. With the explosion in these methods, separate groups approaching things with their own methodology have already yielded a new kind of “zoo” of options of machine learning frameworks. Several studies have shown the capabilities of these methodologies, demonstrating their usefulness for a broad range of systems in efficiently achieving reference-level accuracy.

Despite these achievements, universality of the trained models is often left as a second thought. This leaves others to test the transferability of published networks on their own systems of interest, often necessitating the creation of their own problem-specific network. The fundamental questions then become i) how can we maximize the representation in one's training data, and ii) how small of a training set can one use to still yield sufficient accuracy? Here, we frame these questions in the context of a choice between insight learning, where one needs to capture the behavior of transition state regions and must seed the training to roughly cover the bounds of the relevant potential energy surface, as opposed to active learning, where one regularly recalculates outliers in generated, small subsets as new reference targets during training.   

Screening of Point charges in dielectrics: Disentangling Coulombic from Entropic effects.

The  free energy of ion solvation of different solvent models can be decomposed  into enthalpic and entropic contributions. This can help our understanding of the dielectric properties of the solvent and their connection to the physical properties of the  underlying model. In an effort to model the effects of charge screening on ion pair stability, we present a linear response model that modifies the Coulomb potential and gives an alternate understanding of solvation transition barriers. Moreover, it predicts stable Coulombic pairings between like-charged ions. We show that this is a universal behavior associated to the non local response function of any dielectric or metallic system.