AI discovers new organic lithium-ion battery cathode
The development of environmentally friendly and energy efficient technologies is one of the most pressing needs of this century. In addition, energy consumption on Earth is expected to increase dramatically in the future, which will lead to increased demand for new sources of energy that must be safe, clean and sustainable. An article in the journal Energy storage materials contemplates the use of organic electrode materials.
Study: discovery in silico by artificial intelligence of new cathodes of organic lithium-ion batteries. Image Credit: nevodka / Shutterstock
In this context, organic electrode materials (OEM) combine key properties of durability and versatility with the potential to realize the next generation of truly green battery technologies. Organics offer a combination of attractive characteristics such as low cost and light weight and versatile synthetic methods, customizable properties and production from renewable sources.
Therefore, the proper design of new organic materials with improved properties is highly crucial for sustainable development. However, for OEMs to become a competitive alternative, the difficult issues of energy density, throughput capacity, and cycle stability must be overcome.
This study details the development of an efficient and elegant workflow combining density functional theory (DFT) and machine learning to accelerate the discovery of new organic electroactive materials.
Flowchart illustrating the entire workflow of the developed framework and how the AI kernel allows quick access to the world of organic materials after the learning stage. OMEAD stands for “Organic Materials Database for Energy Applications”. Image credit: Carvalho et al., 2021.
The framework is divided into three key stages. First, the crystal structures of a limited set of 28 candidate electrodes and their corresponding lithiated phases were solved by combining DFT and an evolutionary algorithm.
Second, a database containing structural information and properties of 26,218 organic molecules extracted from high level DFT calculations was developed. Most of the organic fractions recently proposed for energy conversion and storage applications have been included.
Third, models have been developed based on machine learning methodologies to dramatically speed up the assessment of electrochemical properties of OEMs. By combining the data from the first and second stages, an efficient AI kernel with good statistical fidelity was designed, which relies only on the knowledge of molecular structure as input to predict the open circuit voltages of the battery, in completely bypassing the time-consuming ab -Initio calculations.
Results and discussion
The crystal structure of the molecules has been predicted for their respective first two lithiated phases, and the mean lithiation voltage (VCO) for a two-step reaction was calculated. Several molecules in this dataset are based on dicarboxylates because they initially form stable crystals. In addition, dicarboxylate building blocks can be further customized by different mechanisms, thus providing adjustable thermodynamic properties.
A common signature of these crystals is the formation of a layer of salt intercalated by their organic counterpart. The Li ions in this layer are usually surrounded by four oxygen carboxylates, which form a tetrahedral coordination. This characteristic adds considerably to the general stability of these types of organic electrodes, a favorable property for lithium-ion batteries (LIB).
A neural model was built by comparing different combinations of fingerprints and network architectures to generate the most efficient model. Neural networks for all combinations of Coulomb Matrix (CM) and Multibody Tensor Representation (MBTR) were coded on the TensorFlow framework, while the Simplified Molecular-Input Line-Entry System (SMILES ) was developed on top of PyTorch.
The mean absolute error (MAE) was chosen as a learning criterion for networks when analyzing the general performance of different fingerprints and architectures. The training was carried out in part of the Organic Materials Database for Energy Applications (OMEAD) molecular database with 18,528 samples, while 2,290 were reserved for testing.
The SMILES representation achieved performance similar to that of the MBTR, a footprint considerably more powerful and capable of encoding more structural information. Thanks to its conceptual elegance and simplicity, the SMILES architecture was the final choice.
With the trained neural model, the AI core was installed. The next step was to apply the framework in production to explore the organic universe and identify potential new electrodes for LIBs using a high throughput screening approach.
To select potential candidates, a simple voltage filter was applied to identify cathode compounds with VCO greater than 2.9 V (vs Li / Li+) and anode compounds with VCO between 0.0 V and 0.5 V (vs Li / Li+).
The overall result shows a good agreement between DFT and AI, reaffirming the performance of the model. Small deviations are mainly due to outliers mainly from molecules that have undergone major structural changes during the redox process in DFT calculations.
Achieving these improved materials could place organic-based batteries in a desirable position as a next-generation technology for power-hungry applications where the combination of high gravity energy density and battery durability is crucial. .
The AI core discussed in this study enabled high throughput screening of a huge library of organic molecules, leading to the discovery of 459 potential new OEMs, with candidates offering the potential to achieve theoretical energy densities at- over 1000 W h kg-1.
Additionally, the machines accurately identified the common molecular functionalities that result in such higher voltage electrodes and identified an interesting donor-acceptor-like effect that could fuel the future design of active cathode OEMs.
Carvalho, RP, et al. (2021) In silico discovery by artificial intelligence of new organic lithium-ion battery cathodes. Energy storage materials. https://www.sciencedirect.com/science/article/pii/S240582972100489X