To mitigate the worst impacts of climate change, society must adopt renewable energy quickly, and on a global scale. One critical element of addressing this challenge is the need for more affordable, abundant, and higher performing energy storage capabilities.
At Sepion Technologies, we are tackling an important area of the energy storage problem by developing better batteries for electric mobility. Our goal is to enable rapid automotive electrification, including private and fleet battery electric vehicles (EVs) such as cars, buses, and delivery vans.
A major challenge with current lithium-ion batteries for EVs is that they contain a graphite anode that occupies a lot of space in the battery and adds additional weight, restricting the feasible size of the battery pack and limiting overall energy storage capacity. To solve this problem, we are applying next-gen materials chemistry to build a much lighter lithium metal battery (LMB) that does away with the graphite altogether.
Sepion is using a custom membrane technology to stabilize the lithium metal anode, which is notoriously reactive and has historically suffered from a short lifespan. Our approach enables lithium metal anodes to be employed with liquid electrolytes and current state-of-the-art high capacity cathodes, offering up to a 40 percent increase in EV range over incumbent lithium ion batteries. We’re developing this new LMB technology to be entirely drop-in compatible with established lithium-ion manufacturing infrastructure, incentivizing Gigafactories (factories producing e-mobility batteries at scale) to adopt this new technology with minimal hurdles and expense.
Using Schrödinger’s digital molecular simulation platform, we’ve explored thousands of new materials in silico and used that exploration to select the most likely candidates to improve LMB cell performance and stability. This approach has led to a 10-fold improvement in our battery performance over the past two years.
Innovation is often born from necessity. The battery industry has matured to the point where it has nearly reached the theoretical limit for lithium-ion energy density in commercial applications. Recognizing this, we knew it was time to think outside the box and deliver a solution for e-mobility customers that extends range and improves cost simultaneously. Using Schrödinger’s digital molecular simulation platform, we’ve explored thousands of new materials in silico and used that exploration to select the most likely candidates to improve LMB cell performance and stability. This approach has led to a 10-fold improvement in our battery performance over the past two years. Our innovative battery technology recently landed us on C&EN’s 2022 list of 10 Startups to Watch.
I first started using digital simulation in 2014 as part of my research to predict how materials might work in a solar cell. I applied density functional theory (DFT) to run geometry optimizations and energy calculations on single molecules and used those insights to select target compounds on which to focus my experimental efforts. I tried a few different modeling platforms, but as an experimentalist with no background in computational chemistry and limited coding experience, I found Schrödinger’s software the easiest to use.
When I joined Sepion, we were a small team and each of us spent most of our time in the lab focused on improving different aspects of our LMB prototype cell design. This left little internal capacity for computational chemistry. Digital simulation can be daunting, especially among synthetic organic chemists who may instinctively—and fairly—be skeptical over the validity of results. At Sepion, our team worked to embrace the tool, approaching it the same way we approach bench chemistry: We can’t expect to run a simulation and get an instantaneous answer. Instead, we develop a method, iteratively validate and refine that method until we’re confident in its ability to predict real-world outcomes, and then exploit it.
There are talented engineers all over the world working hard to extract incremental benefits from conventional technologies. But to ensure a sustainable future, we need step changes in battery performance that enable rapid adoption of mass-scale automotive electrification.
We started using DFT for small molecule calculations to build a fundamental understanding of certain properties, including electrochemical stability, for specific materials. This insight allowed us to sift through enormous candidate libraries of materials that are not conventionally applied to batteries. For a startup like Sepion, this strategy has been pivotal in saving us precious time and money. We have a team of talented scientists who can approach a whiteboard session, jot down a new library of molecular structures that have never been described before, use DFT to predict the properties of that library, and then synthesize the structures that are predicted to perform the best based on quantitative descriptors for electrochemical stability.
Digital simulation allows us to focus our synthetic efforts on high-value compounds, rather than spending lots of time developing new syntheses to generate libraries of compounds that need to be screened experimentally. In the startup world, the ability to innovate quickly is an incredible asset, and we have been able to embrace digital simulation as a tool to help us do that effectively.
Over time, our use of digital simulation has expanded. We now take huge libraries of materials and run DFT at a much larger scale, link that data to actual experimental results in battery cells, and use DFT properties to generate machine learning (ML) models for battery cell performance. Adding ML to our toolbox closed the loop on molecular design, providing insights into which structural and electrochemical properties derived from DFT are most strongly correlated with LMB performance. Our team now leverages DFT and the first principles knowledge contained in Schrödinger’s physics-based platform—combined with experimental, physical property data and empirical ML-based modeling—to simulate electrolytes based on any of the solvents, additives, and lithium salts contained in our data libraries. We can predict how a theoretical molecule would affect over a dozen properties relevant to performance in batteries, such as density, conductivity, cycle life, rate capability and more. Our team of scientists and engineers can synthesize and formulate actual battery electrolytes with the new components we design, build and test the battery, and then use the performance data to refine our ML models iteratively. This approach has shattered the ceiling for LMB performance using conventional materials.
Most clean energy technologies are held back by the current state-of-the-art material sets. There are talented engineers all over the world working hard to extract incremental benefits from conventional technologies. But to ensure a sustainable future, we need step changes in battery performance that enable rapid adoption of mass-scale automotive electrification. Consumers won’t switch to electric vehicles just because they’re good for the environment, and it would be unfair to ask them to do so. The technology also needs to be affordable, and to perform at least as well as traditional gasoline cars. Sepion is working to make that vision a reality.