Lithium-rich oxides and oxyfluorides are an attractive class of materials for next-generation cathodes for lithium-ion batteries, due to their high theoretical energy density, low cost and potential to be composed of Earth-abundant elements. These cathode materials harness anion redox (oxygen redox) as well as conventional transition metal redox to access very high capacities. Despite their attractive theoretical capacities, Li-rich cathodes also display structural degradation and a loss of energy density, as a byproduct of the oxygen redox. The rapid deterioration in performance means that these materials are not yet used in practical devices. Understanding the oxygen redox mechanisms and preventing the energy density loss are critical to the further development of these materials. The mechanisms of oxygen redox, however, are elusive, requiring complex experimental methods to characterise; consequently, there is extensive debate amongst experimental researchers as to the true redox mechanisms in these materials.

 First-principles modelling circumvents many of the problems that affect experimental investigation and can offer unique insights into oxygen-redox chemistry. However, electrochemical cycling of these cathode materials causes atomic-scale disordering and nanoscale structural changes, which make it difficult to determine a priori which structures should be used in any computational modelling study. To compound the computational challenge, anion redox mechanisms in cathode materials require a high level of computational theory to investigate, making simulations highly computationally intensive.

Our team has been able to tackle this complex problem through the provision of resources from the Materials Chemistry Consortium (MCC). Through our allocation of compute time on the Archer2 HPC, we have implemented a thorough and systematic computational search of structures using DFT calculations [1,2,3], that we used to parameterise computationally efficient atomistic models using the cluster expansion approach. This approach allows us to perform large scale (10,000s of atoms) Monte Carlo simulations of these Li-rich cathode materials , that have revealed the atomic-to-nanoscale structural degradation mechanisms and how these arise from oxygen redox chemistry [4]. This work, made possible due to outstanding computational resources from the MCC, has provided clear design directions for the improvement of Li-rich cathodes. 

Electrochemical energy storage devices (batteries) are a crucial component of the transition to Net Zero. Yet, the battery sector is heavily dependent on positive electrode (cathode) materials composed primarily from oxides of critical, supply-chain limited elements cobalt and nickel. Projected increasing demand for batteries, due to factors such as the continued growth of electric vehicles (EVs), presents a major problem for long-term scalability, where the supply of Co and Ni become a bottleneck for battery manufacturing, and the extraction, processing and use of these elements becomes a major geopolitical flashpoint. As such, there is an urgent need for new high-performance battery cathode materials composed of low-cost, Earth-abundant elements like iron and manganese.

From a technical standpoint, nickel and cobalt-based cathodes, such as NCA (nickel cobalt aluminium) and NMC (nickel manganese cobalt), deliver high energy densities that are needed for EVs and portable electronics. Iron-based polyanion lithium iron phosphate can be used effectively as a cathode material and has excellent safety characteristics, but LFP cannot match the energy density of NCA and NMC cathode and is therefore not viable for high-energy density applications. Few other options exist for new cathode chemistries. The most promising class of materials are the lithium-rich manganese-based cathodes (oxides and oxyfluorides). These cathodes use Earth-abundant and low-cost manganese as the primary transition metal, significantly reducing or eliminating the need for cobalt and nickel. Lithium-rich manganese cathodes have the potential to deliver comparable or even superior energy densities to traditional NMC cathodes. Their high capacities are a result of their Li-rich compositions, which allow a large quantity of Li to be extracted upon charge, activating a combination of both cationic and anionic (oxygen) redox reactions. Li-rich manganese-based cathodes, however, display voltage fade over repeated cycling, which impacts their long-term energy density. The voltage fade is thought to be intimately linked to the oxygen redox mechanisms. Unravelling how the two are linked, and whether oxygen redox can be harnessed without voltage fade, is the major challenge in the field.

Experimental studies that attempt to resolve O-redox degradation mechanisms suffer from the problem that the high-energy X-ray synchrotron techniques used to characterise the mechanisms are claimed to induce degradation themselves. As such, there is major debate within experimental circles as to the true mechanisms. First principles modelling is, therefore, a crucial tool in understanding oxygen redox mechanisms, as modelling circumvents the problems associated with experimental probes. Modelling, however, presents its own challenges. O-redox cathodes become disordered and nanostructured with extended use and feature complex dynamic processes of possible oxygen dimerisation and O–O bond-breaking each time they are cycled. From a computational perspective, this means that there is an enormous possible configurational space of candidate structures that the cathodes could take. Many of these structures are high-energy metastable states and not likely to form in real materials. The challenge for first principles modelling is therefore to assess this vast configurational space of site disorder, and establish which structures are realistic, based on a thorough assessment of thermodynamics and kinetics, whilst also being able to capture nanoscale structural changes in the materials. Further compounding the problem is the significant computational cost of the high-level DFT calculations (metaGGA or hybrid-exchange DFT) that are required to accurately capture both transition metal and oxygen redox properties.

Our group has been able to tackle this challenging problem, through the provision of substantial resources from the Materials Chemistry Consortium (MCC), in the form of a ‘Grand Challenge’ allocation. Our work involved applying different computational methods to a series of different materials so that we could establish a consistent picture of O-redox chemistry across a range of different cathode compositions and structure types.

Initially, we studied a class of disordered rocksalt-structured cathode materials [1]. We were able to establish a clear picture of the thermodynamically and kinetically stable redox species present in these materials at the top of change, using a comprehensive quasi-random structure search. Investigating materials with an initially disordered structure was challenging, however, due to the size of the configurational space. We decided to switch directions to study a class of layered Li-rich cathodes which only become disordered after cycling [2,3], revealing new insights into their kinetic stability and susceptibility to oxygen loss at different states of charge. Nevertheless, the obstacle of simulating nanoscale structural changes in these materials remained.

Figure 1. Thermodynamics and structural properties of O-redox cathode after cycling, from first-principles calculations and cluster-expansion driven Monte Carlo simulations. (a) The DFT-calculated convex hull of formation energies used to fit a cluster expansion for compositions along the O₂–MnO₂–MnO tie line, showing the position, above the ground state hull, of structures with a delithiated composition of Mn_0.8O₂. (b) A supercell of 48,000 atoms obtained from canonical lattice Monte Carlo simulated annealing of Mn_0.8O₂. (c) The detailed lattice structure of a section from b, showing void regions which are filled with O₂ molecules, arising from clustered O–MnX₅ and O–X₆ sites. The inset shows the relaxed structure of a box of confined O₂ molecules.

With insight from our previous investigations and an initial structural database, we were able to develop a new modelling strategy, using the lattice-based cluster expansion method. We overcame a conceptual challenge, by approximating regions of O₂ molecules in the bulk as a lattice-like state, which allowed us to fit and train a successful cluster expansion for a cathode with a composition Li_1.2Mn_0.8O₂. Using the cluster expansion, we could run lattice Monte Carlo simulations of the material in its delithiated state, in large structural cells of ~50,000 atoms. These simulations revealed the nanostructural changes that occur in Li-rich cathodes as they are cycled, which take the form of nanovoids (Figure 1). Nanovoid formation is a direct outcome of the oxygen redox chemistry, and is the major contributor to voltage fade in these materials. Engineering the cathode structure to prevent the voids from forming is therefore a key goal in limiting voltage fade and retaining energy density in these materials. Finally, our work to date has allowed us to develop a new computational strategy, wherein we can predict the voltage fade of a given cathode material, based on the as-synthesised structure. This new approach will provide direct design criteria for the development of new Li-rich cathodes with better energy density retention over long-term cycling.

Outcomes:

• We have built a clear atomistic understanding of O-redox chemistry in a range of commercially relevant cathode materials and compositions.
• We have shown that the stable form of oxidised oxygen–the main contributor to the high capacity in these materials–is molecular O₂, which is formed and trapped within the cathode bulk crystal.
• We have established a protocol for modelling O-redox cathodes using a modified cluster-expansion approach, allowing efficient Monte Carlo simulations that can assess both the disorder and nanoscale structuration in the cathodes over long-term cycling.
• Our database of DFT structures and new protocol has allowed us to develop a model that predicts voltage fade from cathode starting structure, which is a computational first for the field of atomistic modelling of battery cathode materials.

References:

[1] K. McColl, R.A. House, et al., Nature Communications, 13, 5275 (2022)

[2] P. Csernica, K. McColl et al., Nature Materials, 24, 92–100 (2025)

[3] R.A. House, G.J. Rees, K. McColl, et al., Nature Energy, 8, 351–360 (2023)

[4] K. McColl, S.W. Coles, et al., Nature Materials, 23, 826–833 (2024)

Collaborators:

• Robert A. House, Gregory J. Rees, John-Joseph Marie, Simon Cassidy, Peter G. Bruce (University of Oxford)
• Mirian Garcia-Fernandez, Abhishek Nag, Ke-Jin Zhou (Diamond Light Source)
• Peter M. Csernica, Grace M. Busse, Kipil Lim, Diego F. Rivera, David A. Shapiro, William C. Chueh (Stanford University)

Funding:

The Faraday Institution CATMAT project (EP/S003053/1, FIRG016)

Royal Society University Research Fellowship (URF/R/191006) (BJM)

HEC Materials Chemistry Consortium (EP/R029431)

Contact:

Prof M. Saiful Islam: m.s.islam@ox.ac.uk

Dr Benjamin J. Morgan: b.j.morgan@bath.ac.uk