Computational modelling of modern battery material interfaces

Abstract

Material interface simulations often require electronic-level accuracy to deliver meaningful results. \cite{sutton_interfaces_1996}. However, interfaces for the complex systems observed in batteries often require larger simulation sizes than can be simulated to this level of accuracy. This work aims to provide electron-level chemical information, through leveraging the recent innovations of linear-scaling density functional theory over the last decade with the ever-increasing scale of computational resources available to us, about previously inaccessibly large battery interfaces. Li intercalation is one of the most fundamental mechanisms that occur in a Li-ion battery. Providing an electronic-level resolution of this process could significantly enhance the rational design of electrodes. We present an electrostatically-driven workflow that seeks to emulate the charging process for an anode. This workflow proceeds by selecting intercalation sites within a material using the global electrostatic minimum calculated using density functional theory (DFT). This methodology is applied to a high surface area to volume ratio structure in order to better investigate the anodic interface. The structure we select is a 592-atom graphite nanoparticle. We are able to simulate the charging of this large structure with electronic-level accuracy by using linear-scaling DFT methods which allow for the calculation of far larger systems than conventional DFT. The results of our application showed how several experimentally observed features of graphitic intercalation are retained in an unparameterised manner, lending to the quality of our selected methodologies. We observe Li staging, Li to C charge transfer, and the formation of an OCV-like voltage step profile using a convex-hull formalism. We also demonstrate the importance of the role of the interface in large structural shifts. Specifically, how Li accumulation at the edges can hamper the typical graphite transition from the AB to AA polymorphs with increasing Li content. Moving towards the other side of this interface, we present work on the solid electrolyte material Li$_7$La$_3$Zr$_2$O$_{12}$ (LLZO). cubic LLZO's (c-LLZO) exact structure has not been able to be established through experiment or simulation. This problem is essential to resolve as the ground state structure of c-LLZO is required before performing more complex atomistic investigations. By implementing a three-stage methodology of generation, symmetry checking, and energetic ordering we have been able to reduce the vast configuration space of c-LLZO (\(7.4\times 10^{34}\) structures) to just four crystallographically predicted structures. This is achieved by implementing a restriction that no Li atom may be within 1.7 \si{\angstrom} of each other. By considering this geometrical constraint we also determine that a large portion of previously reported structures may not be feasible or stable. We further reduce our configuration space by developing a symmetry checking technique capable of handling over \(1\times 10^8\) structures. Finally, we conclude by energetically ordering our symmetrically unique structure with a machine learning-based energetic calculator that was able to reproduce DFT energetic ordering with 99.96\% accuracy. The data set of our produced structures is freely accessible to all with the aim of improving the accuracy and reproducibility of future LLZO research. To expand the utility of our efforts made in our c-LLZO research we move towards simulation of the entire Li|LLZO interface, featuring a tetragonal LLZO (t-LLZO) interphase. One of the major obstacles in including the solid-state electrolyte LLZO in modern battery packs is the formation of dendrites at the Li|LLZO interface. Experimentally, such interfaces are challenging to probe. This provides a niche in which theory can deliver insights. In this work, we present the progress we have made towards simulating the Li|t-LLZO interface using linear-scaling DFT and high-accuracy surface energies. We also present a methodology to compute the configuration space structures necessary for thoroughly sampling solid-state interfaces which we apply to the system in question and find a minimum, well-sampled, configuration space of \(1.04\times10^7\) structures. To get to this number we find all possible Miller plane combinations, find all possible t-LLZO cuts, create a scheme for sampling inter-slab translations, provide a methodology for finding low-strain commensurate interfaces, and present the results of large-scale DFT calculations on the Li|t-LLZO interface that find a reasonable inter-slab distance value. Finally, having established our configuration space, we present our efforts towards the creation of density functional tight-binding Slater-Koster libraries specific to LLZO. During the process, we demonstrate a general workflow for solid-state systems. We find that our Slater-Koster libraries are able to reproduce DFT values for the total electronic density of states at the Fermi level of bulk t-LLZO. However, further parameterisation is required in order to apply these libraries to Li|LLZO interfaces, particularly the Li-O element-pair interaction.

More information

Identifiers

ORCID for Julian Oliver Holland: orcid.org/0000-0001-8959-0112

ORCID for Chris Skylaris: orcid.org/0000-0003-0258-3433

Catalogue record

Export record