The computational infrastructure of Kristin Persson’s ‘Electrolyte Genome’ project. Candidate molecules (top left) are mapped to workflows (top right) and computed automatically using supercomputers. The results are processed, collected, and shared on the web.Finding a new and better battery electrolyte usually takes time, expertise, and a bit of luck. Now high-performance computing and sophisticated chemical modeling software is removing the guesswork from materials discovery by screening multiple candidates in one go.
A new high-throughput system devised by researchers at the Lawrence Berkeley and Argonne National Laboratories, University of Notre Dame, and the University of California, San Diego has calculated of the key properties of nearly over 16,000 potential electrolyte molecules for a new generation of batteries [Qu et al., Computational Materials Science 103 (2015) 56, http://dx.doi.org/10.1016/j.commatsci.2015.02.050].
The new approach developed by Kristin A. Persson and her colleagues follows the successful model of the Materials Project, which she set up with Gerbrand Ceder at Massachusetts Institute of Technology to calculate the properties of all known materials. Based largely on open-source code, the system uses individual modules for materials analysis, workflow management, and error correction that can be tacked together for a flexible approach.
The approach enables tens of thousands of quantum mechanical ab initio – or first principle – calculations to be performed. The results are then transformed into easy to analyze electrochemical data in a post-processing step.
“In ab initio simulations, no empirical or fitting parameters are involved,” explains Persson. “So the errors, such as they are, are systematic and consistent for a class of compounds, no matter if the structures are well-known or completely new.”
This means that the approach offers greater accuracy in the calculation of key properties such as ionization potential (IP) – the energy required to oxidize a molecule – and electron affinity (EA) – the energy to reduce a molecule. These two properties define the performance of batteries – creating an ‘electrochemical’ window within which it operates.
“We are confident that predictions of novel materials are fairly trustworthy,” she says. “The amount of data is large and gives unprecedented power to couple molecular properties to chemical features and structures.”
Instead of trial and error, the approach provides a rapid and comprehensive means of searching for new electrolytes with improved stability. New design rules for electrolyte components are starting to emerge, enabling the rational construction of electrolyte combinations and molecules.
Ultimately, the ‘Electrolyte Genome’, as researchers dub the effort, will encompass all the chemical components of an electrolyte, as well as more complex properties such as solvation structure, solubility, and chemical stability. The results (and code) are also available as a searchable database with a web interface, which can be readily used by other researchers.
“Our approach can be described in three words,” says Persson, “automatic, massive and user-friendly.”
To download the article related to this news story, please click here.