An integrated high-throughput robotic platform and active learning approach for accelerated discovery of optimal electrolyte formulations.

Solubility of redox-active molecules is an important determining factor of the energy density in redox flow batteries. However, the advancement of electrolyte materials discovery has been constrained by the absence of extensive experimental solubility datasets, which are crucial for leveraging data-driven methodologies. In this study, we design and investigate a highly automated workflow that synergizes a high-throughput experimentation platform with a state-of-the-art active learning algorithm to significantly enhance the solubility of redox-active molecules in organic solvents. Our platform identifies multiple solvents that achieve a remarkable solubility threshold exceeding 6.20 M for the archetype redox-active molecule, 2,1,3-benzothiadiazole, from a comprehensive library of more than 2000 potential solvents. Significantly, our integrated strategy necessitates solubility assessments for fewer than 10% of these candidates, underscoring the efficiency of our approach. Our results also show that binary solvent mixtures, particularly those incorporating 1,4-dioxane, are instrumental in boosting the solubility of 2,1,3-benzothiadiazole. Beyond designing an efficient workflow for developing high-performance redox flow batteries, our machine learning-guided high-throughput robotic platform presents a robust and general approach for expedited discovery of functional materials.

Emin: A First-Principles Thermochemical Descriptor for Predicting Molecular Synthesizability.

Predicting the synthesizability of a new molecule remains an unsolved challenge that chemists have long tackled with heuristic approaches. Here, we report a new method for predicting synthesizability using a simple yet accurate thermochemical descriptor. We introduce Emin, the energy difference between a molecule and its lowest energy constitutional isomer, as a synthesizability predictor that is accurate, physically meaningful, and first-principles based. We apply Emin to 134,000 molecules in the QM9 data set and find that Emin is accurate when used alone and reduces incorrect predictions of "synthesizable" by up to 52% when used to augment commonly used prediction methods. Our work illustrates how first-principles thermochemistry and heuristic approximations for molecular stability are complementary, opening a new direction for synthesizability prediction methods.

Mixed-Anion Contact Ion-Pair Formation Enabling Improved Performance of Halide-Free Mg-Ion Electrolytes.

Discovery of stable and efficient electrolytes that are compatible with magnesium metal anodes and high-voltage cathodes is crucial to enabling energy storage technologies that can move beyond existing Li-ion systems. Many promising electrolytes for magnesium anodes have been proposed with chloride-based systems at the forefront; however, Cl-containing electrolytes lack the oxidative stability required by high-voltage cathodes. In this work, we report magnesium trifluoromethanesulfonate (triflate) as a viable coanion for Cl-free, mixed-anion magnesium electrolytes. The addition of triflate to electrolytes containing bis(trifluoromethane sulfonyl) imide (TFSI-) anions yields significantly improved Coulombic efficiency, up to a 100 mV decrease in the plating/stripping overpotential, improved tolerance to trace H2O, and improved oxidative stability (0.35 V improvement compared to that of hybrid TFSI-Cl electrolytes). Based on 19F nuclear magnetic resonance and Raman spectroscopy measurements, we propose that these improvements in performance are driven by the formation of mixed-anion contact ion pairs, where both triflate and TFSI- are coordinated to Mg2+ in the electrolyte bulk. The formation of this mixed-anion magnesium complex is further predicted by the density functional theory to be thermodynamically driven. Collectively, this work outlines the guiding principles for the improved design of next-generation electrolytes for magnesium batteries.

Active Learning Guided Computational Discovery of Plant-Based Redoxmers for Organic Nonaqueous Redox Flow Batteries.

Organic nonaqueous redox flow batteries (O-NRFBs) are promising energy storage devices due to their scalability and reliance on sourceable materials. However, finding suitable redox-active organic molecules (redoxmers) for these batteries remains a challenge. Using plant-based compounds as precursors for these redoxmers can decrease their costs and environmental toxicity. In this computational study, flavonoid molecules have been examined as potential redoxmers for O-NRFBs. Flavone and isoflavone derivatives were selected as catholyte (positive charge carrier) and anolyte (negative charge carrier) molecules, respectively. To drive their redox potentials to the opposite extremes, in silico derivatization was performed using a novel algorithm to generate a library of > 40000 candidate molecules that penalizes overly complex structures. A multiobjective Bayesian optimization based active learning algorithm was then used to identify best redoxmer candidates in these search spaces. Our study provides methodologies for molecular design and optimization of natural scaffolds and highlights the need of incorporating expert chemistry awareness of the natural products and the basic rules of synthetic chemistry in machine learning.

Accurate Prediction of Adiabatic Ionization Potentials of Organic Molecules using Quantum Chemistry Assisted Machine Learning.

In previous work (Dandu et al., J. Phys. Chem. A, 2022, 126, 4528-4536), we were successful in predicting accurate atomization energies of organic molecules using machine learning (ML) models, obtaining an accuracy as low as 0.1 kcal/mol compared to the G4MP2 method. In this work, we extend the use of these ML models to adiabatic ionization potentials on data sets of energies generated using quantum chemical calculations. Atomic specific corrections that were found to improve atomization energies from quantum chemical calculations have also been used in this study to improve ionization potentials. The quantum chemical calculations were performed on 3405 molecules containing eight or fewer non-hydrogen atoms derived from the QM9 data set, using the B3LYP functional with the 6-31G(2df,p) basis set for optimization. Low-fidelity IPs for these structures were obtained using two density functional methods: B3LYP/6-31+G(2df,p) and ωB97XD/6-311+G(3df,2p). Highly accurate G4MP2 calculations were performed on these optimized structures to obtain high-fidelity IPs to use in ML models based on the low-fidelity IPs. Our best performing ML methods gave IPs of organic molecules within a mean absolute deviation of 0.035 eV from the G4MP2 IPs for the whole data set. This work demonstrates that ML predictions assisted by quantum chemical calculations can be used to successfully predict IPs of organic molecules for use in high throughput screening.

Coordination-Dependent Chemical Reactivity of TFSI Anions at a Mg Metal Interface.

Charge transfer across the electrode-electrolyte interface is a highly complex and convoluted process involving diverse solvated species with varying structures and compositions. Despite recent advances in in situ and operando interfacial analysis, molecular specific reactivity of solvated species is inaccessible due to a lack of precise control over the interfacial constituents and/or an unclear understanding of their spectroscopic fingerprints. However, such molecular-specific understanding is critical to the rational design of energy-efficient solid-electrolyte interphase layers. We have employed ion soft landing, a versatile and highly controlled method, to prepare well-defined interfaces assembled with selected ions, either as solvated species or as bare ions, with distinguishing molecular precision. Equipped with precise control over interfacial composition, we employed in situ multimodal spectroscopic characterization to unravel the molecular specific reactivity of Mg solvated species comprising (i.e., bis(trifluoromethanesulfonyl)imide, TFSI-) anions and solvent molecules (i.e., dimethoxyethane, DME/G1) on a Mg metal surface relevant to multivalent Mg batteries. In situ multimodal spectroscopic characterization revealed higher reactivity of the undercoordinated solvated species [Mg-TFSI-G1]+ compared to the fully coordinated [Mg-TFSI-(G1)2]+ species or even the bare TFSI-. These results were corroborated by the computed reaction pathways and energy barriers for decomposition of the TFSI- within Mg solvated species relative to bare TFSI-. Finally, we evaluated the TFSI reactivity under electrochemical conditions using Mg(TFSI)2-DME-based phase-separated electrolytes representing different solvated constituents. Based on our multimodal study, we report a detailed understanding of TFSI- decomposition processes as part of coordinated solvated species at a Mg-metal anode that will aid the rational design of improved sustainable electrochemical energy technologies.

An SECM-Based Spot Analysis for Redoxmer-Electrode Kinetics: Identifying Redox Asymmetries on Model Graphitic Carbon Interfaces.

The fundamental process in non-aqueous redox flow battery (NRFB) operation revolves around electron transfer (ET) between a current collector electrode and redox-active organic molecules (redoxmers) in solution. Here, we present an approach utilizing scanning electrochemical microscopy (SECM) to evaluate interfacial ET kinetics between redoxmers and various electrode materials of interest at desired locations. This spot-analysis method relies on the measurement of heterogeneous electron transfer rate constants (kf or kb ) as a function of applied potential (E-E0 '). As demonstrated by COMSOL simulations, this method enables the quantification of Butler-Volmer kinetic parameters, the standard heterogeneous rate constant, k0 , and the transfer coefficient, α. Our method enabled the identification of inherent asymmetries in the ET kinetics arising during the reduction of ferrocene-based redoxmers, compared to their oxidation which displayed faster rate constants. Similar behavior was observed on a wide variety of carbon electrodes such as multi-layer graphene, highly ordered pyrolytic graphite, glassy carbon, and chemical vapor deposition-grown graphite films. However, aqueous systems and Pt do not exhibit such kinetic effects. Our analysis suggests that differential adsorption of the redoxmers is insufficient to account for our observations. Displaying a greater versatility than conventional electroanalytical methods, we demonstrate the operation of our spot analysis at concentrations up to 100 mM of redoxmer over graphite films. Looking forward, our method can be used to assess non-idealities in a variety of redoxmer/electrode/solvent systems with quantitative evaluation of kinetics for applications in redox-flow battery research.

A KMnO4-Generated Colloidal Electrolyte for Redox Mediation and Anode Protection in a Li-Air Battery.

The rechargeable lithium-oxygen (Li-O2) battery has the highest theoretical specific energy density of any rechargeable batteries and could transform energy storage systems if a practical device could be attained. However, among numerous challenges, which are all interconnected, are polarization due to sluggish kinetics, low cycle life, small capacity, and slow rates. In this study, we report on use of KMnO4 to generate a colloidal electrolyte made up of MnO2 nanoparticles. The resulting electrolyte provides a redox mediator for reducing the charge potential and lithium anode protection to increase cycle life. This electrolyte in combination with a stable binary transition metal dichalcogenide alloy, Nb0.5Ta0.5S2, as the cathode enables the operation of a Li-O2 battery at a current density of 1 mA·cm-2 and specific capacity ranging from 1000 to 10 000 mA·h·g-1 (corresponding to 0.1-1 mA·h·cm-2) in a dry air environment with a cycle life of up to 150. This colloidal electrolyte provides a robust approach for advancing Li-air batteries.

Improving the Accuracy of Composite Methods: A G4MP2 Method with G4-like Accuracy and Implications for Machine Learning.

G4MP2 theory has proven to be a reliable and accurate quantum chemical composite method for the calculation of molecular energies using an approximation based on second-order perturbation theory to lower computational costs compared to G4 theory. However, it has been found to have significantly increased errors when applied to larger organic molecules with 10 or more nonhydrogen atoms. We report here on an investigation of the cause of the failure of G4MP2 theory for such larger molecules. One source of error is found to be the "higher-level correction (HLC)", which is meant to correct for deficiencies in correlation contributions to the calculated energies. This is because the HLC assumes that the contribution is independent of the element and the type of bonding involved, both of which become more important with larger molecules. We address this problem by adding an atom-specific correction, dependent on atom type but not bond type, to the higher-level correction. We find that a G4MP2 method that incorporates this modification of the higher-level correction, referred to as G4MP2A, becomes as accurate as G4 theory (for computing enthalpies of formation) for a test set of molecules with less than 10 nonhydrogen atoms as well as a set with 10-14 such atoms, the set of molecules considered here, with a much lower computational cost. The G4MP2A method is also found to significantly improve ionization potentials and electron affinities. Finally, we implemented the G4MP2A energies in a machine learning method to predict molecular energies.

Interactions of CO2 Anion Radicals with Electrolyte Environments from First-Principles Simulations.

Successful transformation of carbon dioxide (CO2) into value-added products is of great interest, as it contributes in part to the circular carbon economy. Understanding chemical interactions that stabilize crucial reaction intermediates of CO2 is important, and in this contribution, we employ atom centered density matrix propagation (ADMP) molecular dynamics simulations to investigate interactions between CO2 - anion radicals with surrounding solvent molecules and electrolyte cations in both aqueous and nonaqueous environments. We show how different cations and solvents affect the stability of the CO2 - anion radical by examining its angle and distance to a coordinating cation in molecular dynamics simulations. We identify that the strength of CO2 - interactions can be tailored through choosing an appropriate cation and solvent combination. We anticipate that this fundamental understanding of cation/solvent interactions can facilitate the optimization of a chemical pathway that results from selective stabilization of a crucial reaction intermediate.

Fluorination Enables Simultaneous Improvements of a Dialkoxybenzene-Based Redoxmer for Nonaqueous Redox Flow Batteries.

Redoxmers or redox-active organic materials, are one critical component for nonaqueous redox flow batteries (RFBs), which hold high promise in enabling the time domain of the grid. While tuning redox potentials of redoxmers is a very effective way to enhance energy densities of NRFBs, those improvements often accompany accelerated kinetics of the charged species, undermining stability and cycling performance. Herein, a strategy for designing redoxmers with simultaneous improvements in redox potential and stability is proposed. Specifically, the redoxmer 1,4-di-tert-butyl-2,5-bis(2,2,2-trifluoroethoxy)benzene (ANL-C46) is developed by incorporating fluorinated substitutions into the dialkoxybenzene-based platform. Compared to the non-fluorinated analogue, ANL-C46 demonstrates not only an increased (∼0.41 V) redox potential but also much enhanced stability (1.6 times) and cyclability (4 times) evidenced by electron paramagnetic resonance kinetic study, H-cell and flow cell cycling. In fact, the cycling performance of ANL-C46 is among the best of high potential (>1.0 V vs Ag/Ag+) redoxmers ever reported. Density functional theory calculations suggest that while the introduced fluorine substitutions elevate the redox potentials, they also help to depress the decomposition reactions of the charged redoxmers, affording excellent stability. The findings represent an interesting strategy for simultaneously improving energy density and stability, which could further prompt the development of high-performance redoxmers.

Insights into Spontaneous Solid Electrolyte Interphase Formation at Magnesium Metal Anode Surface from Ab Initio Molecular Dynamics Simulations.

Spontaneous chemical reactivity at multivalent (Mg, Ca, Zn, Al) electrode surfaces is critical to solid electrolyte interphase (SEI) formation, and hence, directly affects the longevity of batteries. Here, we report an investigation of the reactivity of 0.5 M Mg(TFSI)2 in 1,2-dimethoxyethane (DME) solvent at a Mg(0001) surface using ab initio molecular dynamics (AIMD) simulations and detailed Bader charge analysis. Based on the simulations, the initial degradation reactions of the electrolyte strongly depend on the structure of the Mg(TFSI)2 species near the anode surface. At the surface, the dissociation of Mg(TFSI)2 species occurs via cleavage of the N-S bond for the solvent separated ion pair (SSIP) and via cleavage of the C-S bond for the contact ion pair (CIP) configuration. In the case of the CIP, both TFSI anions undergo spontaneous bond dissociation reactions to form atomic O, C, S, F, and N species adsorbed on the surface of the Mg anode. These products indicate that the initial SEI layer formed on the surface of the pristine Mg anode consists of a complex mixture of multiple components such as oxides, carbides, sulfides, fluorides, and nitrides. We believe that the atomic-level insights gained from these simulations will lay the groundwork for the rational design of tailored and functional interphases that are critical for the success of multivalent battery technology.

Graph-Based Approaches for Predicting Solvation Energy in Multiple Solvents: Open Datasets and Machine Learning Models.

The solvation properties of molecules, often estimated using quantum chemical simulations, are important in the synthesis of energy storage materials, drugs, and industrial chemicals. Here, we develop machine learning models of solvation energies to replace expensive quantum chemistry calculations with inexpensive-to-compute message-passing neural network models that require only the molecular graph as inputs. Our models are trained on a new database of solvation energies for 130,258 molecules taken from the QM9 dataset computed in five solvents (acetone, ethanol, acetonitrile, dimethyl sulfoxide, and water) via an implicit solvent model. Our best model achieves a mean absolute error of 0.5 kcal/mol for molecules with nine or fewer non-hydrogen atoms and 1 kcal/mol for molecules with between 10 and 14 non-hydrogen atoms. We make the entire dataset of 651,290 computed entries openly available and provide simple web and programmatic interfaces to enable others to run our solvation energy model on new molecules. This model calculates the solvation energies for molecules using only the SMILES string and also provides an estimate of whether each molecule is within the domain of applicability of our model. We envision that the dataset and models will provide the functionality needed for the rapid screening of large chemical spaces to discover improved molecules for many applications.

Competitive Pi-Stacking and H-Bond Piling Increase Solubility of Heterocyclic Redoxmers.

Redoxmers are organic molecules that carry electric charge in flow batteries. In many instances, they consist of heteroaromatic moieties modified with appended groups to prevent stacking of the planar cores and increase solubility in liquid electrolytes. This higher solubility is desired as it potentially allows achieving greater energy density in the battery. However, the present synthetic strategies often yield bulky molecules with low molarity even when they are neat and still lower molarity in liquid solutions. Fortunately, there are exceptions to this rule. Here, we examine one well-studied redoxmer, 2,1,3-benzothiadiazole, which has solubility ∼5.7 M in acetonitrile at 25 °C. We show computationally and prove experimentally that the competition between two packing motifs, face-to-face π-stacking and random N-H bond piling, introduces frustration that confounds nucleation in crowded solutions. Our findings and examples from related systems suggest a complementary strategy for the molecular design of redoxmers for high energy density redox flow cells.

Self-Assembled Solute Networks in Crowded Electrolyte Solutions and Nanoconfinement of Charged Redoxmer Molecules.

Redoxmers are electrochemically active organic molecules storing charge and energy in electrolyte fluids circulating through redox flow batteries (RFBs). Such molecules typically have solvent-repelling cores and solvent-attracting pendant groups introduced to increase solubility in liquid electrolytes. These two features can facilitate nanoscale aggregation of the redoxmer molecules in crowded solutions. In some cases, this aggregation leads to the emergence of continuous networks of solute molecules in contact, and the solution becomes microscopically heterogeneous. Here, we use small-angle X-ray scattering (SAXS) and molecular dynamics modeling to demonstrate formation of such networks and examine structural factors controlling this self-assembly. We also show that salt ions become excluded from these solute aggregates into small pockets of electrolytes, where these ions strongly associate. This confinement by exclusion is also likely to occur to charged redoxmer molecules in a "sea" of neutral precursors coexisting in the same solution. Here, we demonstrate that the decay lifetime of the confined charged molecules in such solutions can increase several fold compared to dilute solutions. We attribute this behavior to a "microreactor effect" on reverse reactions of the confined species during their decomposition.

Anion Association Strength as a Unifying Descriptor for the Reversibility of Divalent Metal Deposition in Nonaqueous Electrolytes.

Developing next-generation battery chemistries that move beyond traditional Li-ion systems is critical to enabling transformative advances in electrified transportation and grid-level energy storage. In this work, we provide the first evidence for common descriptors for improved reversibility of metal plating/stripping in nonaqueous electrolytes for multivalent ion batteries. Focusing first on the specific role of chloride (Cl-) in promoting electrochemical reversibility in multivalent systems, rotating disk (RDE) and ring-disk electrode (RRDE) investigations were performed utilizing a variety of divalent cations (Mg2+, Zn2+, and Cu2+) and the bis-(trifluoromethane sulfonyl) imide (TFSI-) anion. By introducing varying concentrations of Cl-, a cooperative effect is observed between TFSI- and Cl- that yields the more reversible behavior of mixed electrolytes relative to electrolytes containing only TFSI-. This effect is shown to be general for Mg, Zn, and Cu electrodeposition, and mechanistic understanding of the role of Cl- in improving reversibility of TFSI-based electrolytes is obtained through the combination of R(R)DE experimental results and density functional theory (DFT) evaluation of the redox activity and thermodynamic stability of various TFSI- and Cl-based solution complexes of metal ions. The cooperative anion effect is further generalized to other mixed-anion systems, where systematic variations in anion association strength predicted from DFT (i.e., Cl- > OTf- ≈ TFSI- > BF4- > PF6-) yield corresponding trends in redox potentials and improvements of ≥200 mV in the reversibility of metal deposition/dissolution. These results identify anion association strength as a common descriptor for the reversibility of divalent metal anodes and suggest a set of general design principles for developing new electrolytes with improved activity and stability.

Quantum-Chemically Informed Machine Learning: Prediction of Energies of Organic Molecules with 10 to 14 Non-hydrogen Atoms.

High-fidelity quantum-chemical calculations can provide accurate predictions of molecular energies, but their high computational costs limit their utility, especially for larger molecules. We have shown in previous work that machine learning models trained on high-level quantum-chemical calculations (G4MP2) for organic molecules with one to nine non-hydrogen atoms can provide accurate predictions for other molecules of comparable size at much lower costs. Here we demonstrate that such models can also be used to effectively predict energies of molecules larger than those in the training set. To implement this strategy, we first established a set of 191 molecules with 10-14 non-hydrogen atoms having reliable experimental enthalpies of formation. We then assessed the accuracy of computed G4MP2 enthalpies of formation for these 191 molecules. The error in the G4MP2 results was somewhat larger than that for smaller molecules, and the reason for this increase is discussed. Two density functional methods, B3LYP and ωB97X-D, were also used on this set of molecules, with ωB97X-D found to perform better than B3LYP at predicting energies. The G4MP2 energies for the 191 molecules were then predicted using these two functionals with two machine learning methods, the FCHL-Δ and SchNet-Δ models, with the learning done on calculated energies of the one to nine non-hydrogen atom molecules. The better-performing model, FCHL-Δ, gave atomization energies of the 191 organic molecules with 10-14 non-hydrogen atoms within 0.4 kcal/mol of their G4MP2 energies. Thus, this work demonstrates that quantum-chemically informed machine learning can be used to successfully predict the energies of large organic molecules whose size is beyond that in the training set.

In Situ Formed Ir3Li Nanoparticles as Active Cathode Material in Li-Oxygen Batteries.

Lithium-oxygen (Li-O2) batteries are a promising class of rechargeable Li batteries with a potentially very high achievable energy density. One of the major challenges for Li-O2 batteries is the high charge overpotential, which results in a low energy efficiency. In this work size-selected subnanometer Ir clusters are used to investigate cathode materials that can help control lithium superoxide formation during discharge, which has good electronic conductivity needed for low charge potentials. It is found that Ir particles can lead to lithium superoxide formation as the discharge product with Ir particle sizes of ∼1.5 nm giving the lowest charge potentials. During discharge these 1.5 nm Ir nanoparticles surprisingly evolve to larger ones while incorporating Li to form core-shell structures with Ir3Li shells, which probably act as templates for growth of lithium superoxide during discharge. Various characterization techniques including DEMS, Raman, titration, and HRTEM are used to characterize the LiO2 discharge product and the evolution of the Ir nanoparticles. Density functional calculations are used to provide insight into the mechanism for formation of the core-shell Ir3Li particles. The in situ formed Ir3Li core-shell nanoparticles discovered here provide a new direction for active cathode materials that can reduce charge overpotentials in Li-O2 batteries.

Accurate quantum chemical energies for 133 000 organic molecules.

The energies of the 133 000 molecules in the GDB-9 database have been calculated at the G4MP2 level of theory and then were used to calculate their enthalpies of formation. This database contains organic molecules having nine or less atoms of carbon, nitrogen, oxygen, and fluorine, as well as hydrogen atoms. The accuracy of the G4MP2 energies was investigated on a subset of 459 of the molecules having experimental enthalpies of formation with small uncertainties. On this subset the G4MP2 enthalpies of formation have an accuracy of 0.79 kcal mol-1, which is similar to its accuracy previously reported for the smaller G3/05 test set. An error analysis of the theoretical enthalpies of formation of the 459 molecules is presented in terms of the size and type of the molecules. Three different density functionals (B3LYP, ωB97X-D, M06-2X) were also assessed on 459 molecules of accurate enthalpy data for comparison with the G4MP2 results. The G4MP2 energies for the 133 K molecules provide a database that can be used to calculate accurate reaction energies as well as to assess new or existing experimental enthalpies of formation. Several examples are given of types of reactions that can be predicted using the G4MP2 database of energies. The G4MP2 energies of the GDB-9 molecules will also be useful in future investigations of applications of machine learning to quantum chemical data.