Electrolyte development plays a key role in the design of current and future battery chemistries. Almost all of the advanced battery technologies such as lithium metal, high voltage cathodes, dual ion batteries, metal air batteries and other beyond lithium ion battery technologies are bottlenecked by the performance of the electrolytes.^{[i],[ii],[iii]} There is an increasing need for electrolytes with large electrochemical windows, wide temperature operating windows, very good ionic conductivity for operation at high rates of charge and discharge, low viscosity, good solubility and low flammability and toxicity from a safety standpoint. Thus, there are a large number of constraints on the electrolytes and they are becoming increasingly stringent. To add to the complexity of electrolyte design, there exist multiple trade-offs between all these properties as shown in previous literature.^[iv] Although various solid electrolytes, ionic liquids and polymer electrolytes are being researched, organic liquid electrolytes still outperform the other electrolytes when we consider all the properties required. The current approach of new electrolyte design is largely intuition based using certain structural features in the organic molecules. Currently only a few hundred solvents and salts have been demonstrated in all the battery technologies while there are over 20 million organic molecules discovered. The richness of liquid organic compounds is multiplied even further if we start considering mixtures as most of the properties vary non-linearly as you mix two organic compounds. In this work, we suggest a much better data-driven systematic exploration of new electrolytes. The methodology employs determination of various electronic structure features calculated from simple and scalable density functional theory (DFT) calculations and topological features of the molecules. We will identify unique parameters which are non-linear functions of the above features for each of the properties above along with features that govern mixtures. We currently have structures for over 1 million molecules with over 100,000 data points for each property. This gives a unique capability to identify parameters accurately, explore multiple trade-offs and different methods of overcoming these trade-offs.

The cathodic stability of the electrolyte can be described in terms of the highest occupied molecular orbital (HOMO) energy level of each of the components. The HOMO level will be estimated using GGA level DFT, quantitative structure-property relationships (QSPR) and higher order theories and comparisons will be shown. The solubility of a salt is governed by the solvation of the individual ions and the dielectric constant of the solvent as shown by the extended Debye- Hückel and other models. We have developed a modified Ising model to describe solvation shell around the Li⁺ ions, and the model has been solved under mean field approximations to derive the fraction of all species in the Li⁺ solvation shell.^[v] The model can also be extended to describe other cations and anions and is applicable to mixtures. The Ising model coefficients used for the model are the ion-solvent interactions, ion-ion interactions and solvent-solvent interactions can be calculated from DFT. The dielectric constant, important for solubility is correlated to the dipole moment of molecule. This relationship is quite qualitative as in a liquid, the molecules exist as multiple conformers. To provide more quantitative data, we have identified the volume of molecule calculated using DFT along with number of rotatable bonds as additional geometric features to determine the dielectric constant. The dielectric constant along with the solvation energies will be used to give a feature for solubility. The viscosity of an electrolyte is dependent on the solvation shell structure of ions and the viscosity of the solvent which is related to intermolecular interactions and will be explored. We will also show the relationship between the Li⁺ solvation energy and charge transfer resistance.

We believe this electrolyte exploration method is much more systematic, quantitative, scalable and also able to identify outliers to certain trends. This method and database will create a new frontier for electrolyte discovery for most electrochemical systems and also other areas of science. In the future, we will add features for other physical and chemical properties and also extend these to inorganic compounds.

References:

[i] Bruce, Peter G., et al. Nature materials 11.1 (2012): 19.

[ii] Tarascon, J-M., Michel Armand. Materials For Sustainable Energy: A Collection of Peer-Reviewed Research and Review Articles from Nature Publishing Group. 2011. 171-179.

[iii] Choi, Jang Wook, Doron Aurbach. Nature Reviews Materials 1 (2016): 16013.

[iv] Khetan, Abhishek, Alan Luntz, Venkatasubramanian Viswanathan. The Journal of Physical Chemistry Letters 6.7 (2015): 1254-1259.

[v] Burke, Colin M., et al. Proceedings of the National Academy of Sciences 112.30 (2015): 9293-9298.