Artificial Intelligence Applied to Battery Research: Hype or Reality?

This is a critical review of artificial intelligence/machine learning (AI/ML) methods applied to battery research. It aims at providing a comprehensive, authoritative, and critical, yet easily understandable, review of general interest to the battery community. It addresses the concepts, approaches, tools, outcomes, and challenges of using AI/ML as an accelerator for the design and optimization of the next generation of batteries─a current hot topic. It intends to create both accessibility of these tools to the chemistry and electrochemical energy sciences communities and completeness in terms of the different battery R&D aspects covered.

Carbonyl-Based Redox-Active Compounds as Organic Electrodes for Batteries: Escape from Middle-High Redox Potentials and Further Improvement?

Extracting─from the vast space of organic compounds─the best electrode candidates for achieving energy material breakthrough requires the identification of the microscopic causes and origins of various macroscopic features, including notably electrochemical and conduction properties. As a first guess of their capabilities, molecular DFT calculations and quantum theory of atoms in molecules (QTAIM)-derived indicators were applied to explore the family of pyrano[3,2-b]pyran-2,6-dione (PPD, i.e., A0) compounds, expanded to A0 fused with various kinds of rings (benzene, fluorinated benzene, thiophene, and merged thiophene/benzene). A glimpse of up-to-now elusive key incidences of introducing oxygen in vicinity to the carbonyl redox center within 6MRs─as embedded in the A0 core central unit common to all A-type compounds─has been gained. Furthermore, the main driving force toward achieving modulated low redox potential/band gaps thanks to fusing the aromatic rings for the A compound series was discovered.

Gaining Insight into the Electrochemical Interface Dynamics in an Organic Redox Flow Battery with a Kinetic Monte Carlo Approach.

Finding low-cost and nontoxic redox couples for organic redox flow batteries is challenging due to unrevealed reaction mechanisms and side reactions. In this study, a 3D kinetic Monte Carlo model to study the electrode-anolyte interface of a methyl viologen-based organic redox flow battery is presented. This model captures various electrode processes, such as ionic displacement and degradation of active materials. The workflow consists of input parameters obtained from density functional theory calculations, a kinetic Monte Carlo algorithm to simulate the discharging process, and an electric double layer model to account for the electric field distribution near the electrode surface. Galvanostatic discharge is simulated at different anolyte concentrations and input current densities, which demonstrate that the model captured the formation of the electrical double layer due to ionic transport. The simulated electrochemical kinetics (potential, charge density) are found to be in agreement with the Nernst equation and the obtained EDL structure corresponded with published molecular dynamics results. The model's flexibility allows further applications of simulating the behavior of different redox couples and makes it possible to consider other molecular-scale phenomena. This study paves the way for computational screening of active species by assessing their potential kinetics in electrochemical environments.

CHAMPION: Chalmers hierarchical atomic, molecular, polymeric and ionic analysis toolkit.

We present CHAMPION (Chalmers hierarchical atomic, molecular, polymeric, and ionic analysis toolkit): a software developed to automatically detect time-dependent bonds between atoms based on their dynamics, classify the local graph topology around them, and analyze the physicochemical properties of these topologies by statistical physics. In stark contrast to methodologies where bonds are detected based on static conditions such as cut-off distances, CHAMPION considers pairs of atoms to be bound only if they move together and act as a bound pair over time. Furthermore, the time-dependent global bond graph is possible to split into dynamically shifting connected components or subgraphs around a certain chemical motif and thereby allow the physicochemical properties of each such topology to be analyzed by statistical physics. Applicable to condensed matter and liquids in general, and electrolytes in particular, this allows both quantitative and qualitative descriptions of local structure, as well as dynamical processes such as speciation and diffusion. We present here a detailed overview of CHAMPION, including its underlying methodology, implementation, and capabilities.

Boosting Rechargeable Batteries R&D by Multiscale Modeling: Myth or Reality?

This review addresses concepts, approaches, tools, and outcomes of multiscale modeling used to design and optimize the current and next generation rechargeable battery cells. Different kinds of multiscale models are discussed and demystified with a particular emphasis on methodological aspects. The outcome is compared both to results of other modeling strategies as well as to the vast pool of experimental data available. Finally, the main challenges remaining and future developments are discussed.

Structural and surface coverage effects on CO oxidation reaction over carbon-supported Pt nanoparticles studied by quadrupole mass spectrometry and diffuse reflectance FTIR spectroscopy.

The CO oxidation reaction on carbon-supported Pt nanoparticles (average size of 2.8 to 7.7 nm) was studied under flowing conditions at atmospheric pressure and temperatures between 300 and 353 K by coupling quadrupole mass spectrometry (QMS) and diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS). The Pt loading was varied between 20 and 60 wt%. Gases diluted in He (0.5 mol%) were used together with Ar as a tracer. Reactions with CO and O2 introduced separately onto the samples were studied by QMS, applying successive step changes of the reaction mixtures. Variations in the rate of the reactions were observed and correlated with changes of the calculated coverage of the Pt surface by CO and/or O adspecies at varying steps of the experiment. The transient reaction of CO(g) with adsorbed O (Oad) was fast and mass transport-limited while that of O2(g) with adsorbed CO (COad) was sluggish. Following the same experimental procedures, FTIR spectra of adsorbed CO after varying steps were recorded, confirming the variations of COad and Oad as determined by QMS and indicating changes in the CO distribution over varying types of Pt surface sites. The influence of the adlayer composition (co-adsorption of COad and Oad), the particle size/structure and some possible surface reconstruction effects on the CO oxidation rate were evidenced and discussed. The structure of the Pt nanoparticles supported on carbon appears as an important factor for the efficiency of the so-called O2 bleeding as a CO mitigation strategy in polymer electrolyte membrane fuel cells.

Coverage-dependent thermodynamic analysis of the formation of water and hydrogen peroxide on a platinum model catalyst.

Understanding the selectivity of the oxygen reduction reaction, especially the formation of water versus hydrogen peroxide in fuel cells, is an ongoing challenge in electrochemistry, surface science and catalysis. In this study, we propose a comprehensive thermodynamic analysis of the reaction intermediates for the formation of water on Pt(111). Density functional theory calculations of all the elementary steps linking hydroxyl and hydroperoxyl surface species with water and hydrogen peroxide have been performed at low (1/12 ML, ML = monolayer) and high (1/4 ML) coverages. The reaction energy variation for the two competing elementary events (molecular oxygen dissociation and hydroperoxyl formation) is strongly coverage-dependent. For the direct dissociation, an increase is observed at low coverage with respect to the usual high coverage picture. The stability of the reaction intermediates is investigated from thermodynamic diagrams. At 353 K and a total pressure of 1 atm, water and hydroxyl surface species are expected to compete for adsorption on Pt(111).

Binder-Free Cnt Cathodes for Li-O2 Batteries with More Than One Life.

Li-O2 batteries (LOB) performance degradation ultimately occurs through the accumulation of discharge products and irreversible clogging of the porous electrode during the cycling. Electrode binder degradation in the presence of reduced oxygen species can result in additional coating of the conductive surface, exacerbating capacity fading. Herein, a facile method to fabricate free-standing is established, binder-free electrodes for LOBs in which multi-wall carbon nanotubes form cross-linked networks exhibiting high porosity, conductivity, and flexibility. These electrodes demonstrate high reproducibility upon cycling in LOBs. After cell death, efficient and inexpensive methods to wash away the accumulated discharge products are demonstrated, as reconditioning method. The second life usage of these electrodes is validated, without noticeable loss of performance. These findings aim to assist in the development of greener high energy density batteries while reducing manufacturing and recycling costs.

A multiscale physical model for the transient analysis of PEM water electrolyzer anodes.

Polymer electrolyte membrane water electrolyzers (PEMWEs) are electrochemical devices that can be used for the production of hydrogen. In a PEMWE the anode is the most complex electrode to study due to the high overpotential of the oxygen evolution reaction (OER), not widely understood. A physical bottom-up multi-scale transient model describing the operation of a PEMWE anode is proposed here. This model includes a detailed description of the elementary OER kinetics in the anode, a description of the non-equilibrium behavior of the nanoscale catalyst-electrolyte interface, and a microstructural-resolved description of the transport of charges and O(2) at the micro and mesoscales along the whole anode. The impact of different catalyst materials on the performance of the PEMWE anode, and a study of sensitivity to the operation conditions are evaluated from numerical simulations and the results are discussed in comparison with experimental data.

Bridging nano- and microscale X-ray tomography for battery research by leveraging artificial intelligence.

X-ray computed tomography (CT) is a non-destructive imaging technique in which contrast originates from the materials' absorption coefficient. The recent development of laboratory nanoscale CT (nano-CT) systems has pushed the spatial resolution for battery material imaging to voxel sizes of 50 nm, a limit previously achievable only with synchrotron facilities. Given the non-destructive nature of CT, in situ and operando studies have emerged as powerful methods to quantify morphological parameters, such as tortuosity factor, porosity, surface area and volume expansion, during battery operation or cycling. Combined with artificial intelligence and machine learning analysis techniques, nano-CT has enabled the development of predictive models to analyse the impact of the electrode microstructure on cell performances or the influence of material heterogeneities on electrochemical responses. In this Review, we discuss the role of X-ray CT and nano-CT experimentation in the battery field, discuss the incorporation of artificial intelligence and machine learning analyses and provide a perspective on how the combination of multiscale CT imaging techniques can expand the development of predictive multiscale battery behavioural models.

How Machine Learning Will Revolutionize Electrochemical Sciences.

Electrochemical systems function via interconversion of electric charge and chemical species and represent promising technologies for our cleaner, more sustainable future. However, their development time is fundamentally limited by our ability to identify new materials and understand their electrochemical response. To shorten this time frame, we need to switch from the trial-and-error approach of finding useful materials to a more selective process by leveraging model predictions. Machine learning (ML) offers data-driven predictions and can be helpful. Herein we ask if ML can revolutionize the development cycle from decades to a few years. We outline the necessary characteristics of such ML implementations. Instead of enumerating various ML algorithms, we discuss scientific questions about the electrochemical systems to which ML can contribute.

Probing and Interpreting the Porosity and Tortuosity Evolution of Li-O2 Cathodes on Discharge through a Combined Experimental and Theoretical Approach.

Li-O2 batteries offer a high theoretical discharge capacity due to the formation of light discharged species such as Li2O2, which fill the porous positive electrode. However, in practice, it is challenging to reach the theoretical capacity and completely utilize the full electrode pore volume during discharge. With the formation of discharge products, the porous medium evolves, and the porosity and tortuosity factor of the positive electrode are altered through shrinkage and clogging of pores. A pore shrinks as solid discharge products accumulate, the pore clogging when it is filled (or when access is blocked). In this study, we investigate the structural evolution of the positive electrode through a combination of experimental and computational techniques. Pulsed field gradient nuclear magnetic resonance results show that the electrode tortuosity factor changes much faster than suggested by the Bruggeman relation (an equation that empirically links the tortuosity factor to the porosity) and that the electrolyte solvent affects the tortuosity factor evolution. The latter is ascribed to the different abilities of solvents to dissolve reaction intermediates, which leads to different discharge product particle sizes: on discharging using 0.5 M LiTFSI in dimethoxyethane, the tortuosity factor increases much faster than for discharging in 0.5 M LiTFSI in tetraglyme. The correlation between a discharge product size and tortuosity factor is studied using a pore network model, which shows that larger discharge products generate more pore clogging. The Knudsen diffusion effect, where collisions of diffusing molecules with pore walls reduce the effective diffusion coefficients, is investigated using a kinetic Monte Carlo model and is found to have an insignificant impact on the effective diffusion coefficient for molecules in pores with diameters above 5 nm, i.e., most of the pores present in the materials investigated here. As a consequence, pore clogging is thought to be the main origin of tortuosity factor evolution.

Metallofullerenes as fuel cell electrocatalysts: a theoretical investigation of adsorbates on C59Pt.

Nano-structured electrode degradation in state-of-the-art polymer electrolyte membrane fuel cells (PEMFCs) is one of the main shortcomings that limit the large-scale development and commercialization of this technology. During normal operating conditions of the fuel cell, the PEMFC lifetime tends to be limited by coarsening of the cathode's Pt-based catalyst and by corrosion of the cathode's carbon black support. Because of their chemical properties, metallofullerenes such as C(59)Pt may be more electrochemically stable than the Pt/C mixture. In this paper we investigate, by theoretical methods, the stability of oxygen reduction reaction (ORR) adsorbates on the metallofullerene C(59)Pt and evaluate its potential as a PEMFC fuel cell catalyst.

New insights in the electrocatalytic proton reduction and hydrogen oxidation by bioinspired catalysts: a DFT investigation.

In this paper, we present a DFT study of the proton reduction mechanism catalyzed by the complex [Ni(P2(H)N2(H))2](2+), bioinspired from the hydrogenases. A detailed analysis of the reactive isomers is discussed together with the localizations of the transitions states and energy minima. The reactive catalytic species is a biprotonated Ni(0) complex that can show different conformations and that can be protonated on different sites. The energies of the different conformations and biprotonated species have been calculated and discussed. Energy barriers for two different reaction mechanisms have been identified in solvent and in gas phase. Frequency calculations have been performed to check the nature of the energy minima and for the calculations of entropic energetic terms and zero point energies. We show that only one conformation is mostly reactive. All the others species are nonreactive in their original form, and they have to pass through conformational barriers in order to transform in the reactive species.

Mesoscale Effects in the Extraction of the Solid-State Lithium Diffusion Coefficient Values of Battery Active Materials: Physical Insights from 3D Modeling.

During the screening of active materials (AMs) for lithium-ion batteries, the solid-state lithium diffusion coefficient (DLi) is one of the most relevant descriptors used to evaluate the relevance of an AM candidate. However, for a given compound, the DLi values reported in literature span over several orders of magnitude. Therefore, through the case study of LiNi1/3Mn1/3Co1/3O2 cathode AM, new physical insights are provided to explain the dispersion of DLi values obtained through galvanostatic intermittent titration technique (GITT) . For the first time, a 3D electrochemical model (accounting for the carbon-binder domain) fed with experimental inputs is capable of highlighting the limitations of the most widely used equation for deriving DLi. Through our model, we show that these limitations arise from the influence of the carbon-binder domain location throughout the electrode and the non-homogeneous AM phasedistribution and particle size.

A Versatile and Efficient Voxelization-Based Meshing Algorithm of Multiple Phases.

This paper presents a new algorithm (INNOV) capable of generating a mesh of three-dimensional objects containing multiple phases. This mesh can later be imported into commercial or open-source software to perform multiphysics-based simulations based on partial differential equations. While the range of application is large, this algorithm is designed as a post-processing tool of micro/nanotomography images and electrode mesostructures predicted from CGMD (coarse-grained molecular dynamics) simulations of the electrode fabrication process carried out in LAMMPS software. With INNOV, it becomes possible to import the predicted multiparticle electrode mesostructures into COMSOL Multiphysics in order to simulate electrochemistry and transport in operating lithium-ion batteries.

Stochasticity of Pores Interconnectivity in Li-O2 Batteries and its Impact on the Variations in Electrochemical Performance.

While large dispersions in electrochemical performance have been reported for lithium oxygen batteries in the literature, they have not been investigated in any depth. The variability in the results is often assumed to arise from differences in cell design, electrode structure, handling and cell preparation at different times. An accurate theoretical framework turns out to be needed to get a better insight into the mechanisms underneath and to interpret experimental results. Here, we develop and use a pore network model to simulate the electrochemical performance of three-dimensionally resolved lithium-oxygen cathode mesostructures obtained from TXM nanocomputed tomography. We apply this model to the 3D reconstructed object of a Super P carbon electrode and calculate discharge curves, using identical conditions, for four different zones in the electrode and their reversed configurations. The resulting galvanostatic discharge curves show some dispersion, (both in terms of capacity and overpotential) which we attribute to the way pores are connected with each other. Based on these results, we propose that the stochastic nature of pores interconnectivity and the microscopic arrangement of pores can lead, at least partially, to the variations in electrochemical results observed experimentally.

Compactness of the Lithium Peroxide Thin Film Formed in Li-O2 Batteries and Its Link to the Charge Transport Mechanism: Insights from Stochastic Simulations.

We simulated the discharge process of Li-O2 batteries and the growth of Li2O2 thin films at the mesoscale with a novel kinetic Monte Carlo model, which combined a stochastic description of mass transport and detailed elementary reaction kinetics. The simulation results show that the ordering of the Li2O2 thin film is determined by the interplay between diffusion and reaction kinetics. Due to the fast reaction kinetics on the catalyst, the Li2O2 formed in the presence of catalyst (cat-CNF) shows a low degree of ordering and is more likely to be amorphous. Moreover, the mobility of the LiO2 ion pair, which depends largely on the nature of the electrolyte, also impacts the homogeneity of the compactness of the Li2O2 thin film. These results are of high importance for understanding the role of the catalyst and reaction kinetics in Li-O2 batteries.

Self-Organization of Electroactive Suspensions in Discharging Slurry Batteries: A Mesoscale Modeling Investigation.

We report a comprehensive modeling-based study of electroactive suspensions in slurry redox flow batteries undergoing discharge. A three-dimensional kinetic Monte Carlo model based on the variable step size method is used to describe the electrochemical discharge of a silicon/carbon slurry electrode in static mode (i.e., no fluid flow conditions). The model accounts for Brownian motion of particles, volume expansion of silicon upon lithium insertion, and formation and destruction of conducting carbon networks. Coupled to an electrochemical model, this study explores the impact of carbon fraction in the slurry and applied c-rate on the specific capacity. The trends obtained are analyzed by following the behavior of parameters such as number of contacts between electroactive particles and the percentage of electroactive silicon particles. Furthermore, instead of studying the bulk behavior of the slurry, here the focus is given to the slurry/current collector interface in order to illustrate its importance. Hereby, it is demonstrated how this modeling tool can lead to deeper understanding and optimization of electroactive particle suspensions in redox flow batteries.