The switch from fossil fuels to renewable energy is currently being driven by improving lithium-ion battery (LIB) performance and cycle life.
Up to now, however, traditional battery design has proven slow and expensive. Computational modelling on the other hand, offers a smart track forward for novel battery designs for both electric vehicles (EV) as well as grid energy storage.
Computational modelling techniques such as molecular dynamics, electrochemical modelling techniques, and machine learning/AI techniques are giving researchers new tools to study systems at different length scales (depending on the system under examination), at the same time enabling the screening, optimizing, and testing of novel materials.
Taken together, these techniques promise a significant revamp of the present generation of lithium-ion batteries and the fast-track development of a host of emerging chemistries to supercharge a burgeoning global e-mobility ecosystem and a fossil fuel-free future.
Computational modelling techniques are giving researchers new tools to study and optimize the testing of novel materials in battery design. These techniques promise a significant revamp of lithium-ion batteries to supercharge the global e-mobility ecosystem.
The lithium-ion battery (LIB) is the present gold standard in energy storage technology.
It offers both high energy density, which is one of the main requirements of electric vehicles (EV), as well as grid energy storage. LIB-based electric vehicles (EVs) are barely a decade old, but the LIBs that power them have demonstrated significant flaws. As demand for LIBs in consumer electronics, EVs, and grid storage applications grows, the study and development of innovative battery chemistries is becoming increasingly important.
Compared to their internal combustion engine (ICE)-based counterparts, the current generation of LIB-based EVs fall short in various ways, not the least of these being the amount of time it takes to charge them compared to the time taken at the fuel pump.
Cycle life, defined by a typical calendar life of ten years, is presently 2,000-4,000 cycles. This is also factored into the cost of a typical battery, which accounts for a significant amount in the overall price of an EV. This last point also means that the upfront cost of an EV is significantly higher than an equivalent ICE-powered vehicle.
Another thing to consider is the range limitation of present battery technology leading to range anxiety between charges and the impact of external temperatures. The ideal operating range for a battery is 20 to 55 °C. Operating temperatures beyond these limits are detrimental to the battery. Also, owing to their high operating voltages, lithium-ion batteries have proven to be highly flammable.
A final consideration is the issue of sustainability. In all instances, it is important to optimize existing battery devices for improved performance across the board. This suggests the need to explore lithium-based electrolytes that can lead to better novel batteries.
To an extent, battery performance and safety challenges can be addressed with appropriate sensors and robust battery management systems.
Equally, the development of innovative battery materials answers the shortcomings of current generation batteries, such as decreased cycle life, range, thermal stability, operating voltages, and longer charging times. Traditional pathways for discovery of innovative battery materials are, however, resource-intensive, and time-consuming.
Furthermore, the quest for sustainable alternatives makes this procedure even more complex.
A depiction of the various length scales involved in modeling a lithium-ion battery cell versus their computational complexity. Along with the traditional physics-based models, ML/AI based models too have been shown to be effective at different scales.
Molecular dynamics (MD) enables the evaluation of the transport properties of lithium in the electrolytes.
Molecular dynamics (MD) is used to study the interactions between a group of molecules. The MD method mimics particle motion i.e., atoms and molecules, so that their trajectories describing the system's spatio-temporal evolution can be used to compute various attributes. MD simulations are commonly used to investigate the impact of various factors on the performance of lithium-ion battery components. The transport properties of the electrolyte, such as lithium-ion diffusivity and transference number, and ionic conductivity of the electrolyte, influence numerous performance metrics of the battery, such as rate performance, amount of heat generated, cycle life, and so on. Ion-ion and ion-solvent interactions alter these properties at the molecular level.
Electrolytes based on certain mixtures of solvents and salt concentration perform better. The reason for this is not fully understood based on experiments alone. Molecular simulations provide insights into molecular level phenomena responsible for observed phenomena. Molecular dynamics simulations suggest that the way solvent molecules surround and interact with lithium ions is a key factor. These simulations help identify the right combinations of solvents and additives etc. that improve the performance of electrolytes, and consequently enhance overall performance of the battery.
Quantitative structure-property relationships (QSPR) is a widely used ML-based technique for determining material properties based on physical or chemical properties obtained using as either simulation techniques such as MD/ DFT, or experiments involving ionic conductivity or impedance measurements.
The benefit of QSPR is that it paves the way for quick estimation and for screening of large material databases. At a system level, physics-based continuum models are used to estimate the variables that govern the overall performance of the battery during charge and discharge - parameters like voltage, current, state of charge, state of health temperature, cycle life etc.
However, the system level performance is inherently coupled with the processes occurring at lower length scales. Thus, multi-scale modelling, which utilizes modelling techniques at different length-scales, is the key to understanding at a holistic level, the influence of novel battery materials on cell performance.
Electrochemical models are physics-based models that can simulate the charge and discharge process of a battery at the system length-scale.
The governing equations comprise electrochemical-kinetic expressions and transport equations. The battery voltage, current, temperature, state of charge (SoC), capacity fade, and state of health (SoH), required by battery management systems (BMS) to track battery performance and safety, can all be estimated under different operating conditions using these models.
Furthermore, the battery material properties derived from experiments or lower length-scale simulations can be tested and optimized using electrochemical models. For instance, the properties of an electrolyte can be used in electrochemical models and the influence of electrolyte on system level performance can also be evaluated.
These models are, therefore, highly suitable for virtual testing of new battery materials. Electrochemical models are invaluable for the future of battery design. For example, modelling thermal effects in a battery is critical and electrochemical models offer a fast bridge to safer battery design under different operating conditions and the selection of battery materials that are less prone to thermal runaways / battery overheating.
A typical lithium-ion battery cell consists of porous electrodes isolated by a separator and filled with an electrolyte, which acts as a conducting medium for the ions. Electrochemical models offer a means to simulate the charge and discharge process.
In addition, when simulating battery lifetime, electrochemical modelling enables the simulation of both SEI (solid electrolyte interface) formation at higher temperatures and lithium plating at lower temperatures during the last few cycles, which are the most common ageing mechanisms in LIBs.
Electrochemical models enable researchers to simulate such conditions in just hours without the need for excessive experimentation. This accelerates the process of material discovery. Thus, when combined with lower length-scale models, electrochemical models significantly augment the scope of material exploration.
Design and testing
Access to huge amounts of data from experimental and computational databases has driven the use of ML and deep learning techniques for the design and discovery of novel battery materials.
This is at both continuum and lower length scales. The benefit of ML/AI models is their lower computational cost compared to traditional physics-based modelling techniques, and, with reasonable accuracy. In this way, ML/DL-based material informatics can complement high-fidelity physics-based models and accelerate the process of material design and discovery.
Thus, multiscale modelling, including both physics-based and ML/AI based approaches, can enable a new level of high-fidelity and rapid screening of multiple battery materials and the exploration of novel battery materials.
At this critical juncture, when we must both meet the world’s growing energy demands and do so sustainably, the multiscale modelling approach promises an accelerated research path to lithium-ion battery design.
The authors would like to thank Kaustubh Badwekar for his contributions to the whitepaper and Dr Beena Rai for her guidance and support.
 Chade Lv, Xin Zhou, Lixiang Zhong, Chunshuang Yan, Madhavi Srinivasan, Zhi Wei Seh. "Machine Learning: An Advanced Platform for Materials Development and State Prediction in Lithium-Ion Batteries." Advanced Materials (2021).
 Alejandro A. Franco, Alexis Rucci, Daniel Brandell, Christine Frayret, Miran Gaberscek, Piotr Jankowski, and Patrik Johansson. "Boosting Rechargeable Batteries R&D by Multiscale Modeling: Myth or Reality?" Chemical Reviews 7.119 (2019): 4569-4627.
 Yue Yang, Emenike G. Okonkwo, Guoyong Huang, Shengming Xu, Wei Sun, Yinghe He. "On the sustainability of lithium-ion battery industry – A review and perspective." Energy Storage Materials 36 (2021): 186-212.