One of the biggest problems facing society today is climate change. 60% of emissions accelerating climate change come from energy generation and the transportation system, both of which need better batteries to be able to cut their emissions. Lithium-ion is the industry standard for both, and while it has great properties, it’s too expensive to scale for long duration energy storage for the grid, a crucial step towards renewables, and it does not have very high energy density, which is vital for the transition to electric vehicles with driving ranges similar to combustion-engine-based vehicles today.
However, the way we try to discover these battery advancements is terribly inefficient. It is typically done through lab experiments, which are complex, expensive, and time consuming.
This leads us to more technologically advanced solutions: simulations and machine learning. Instead of testing every compound and every situation manually, we can use compute to simulate and predict how different materials and conditions will affect the battery. This speeds up the process, and instead of taking around a month it can be done in mere seconds - 0.5 to be exact [1]. It also dramatically reduces the costs, as physical materials are not necessary for testing.
Unfortunately, this pre-built simulation software only exists for lithium based battery chemistries. The only way that researchers can virtually simulate their lithium alternative battery chemistries is by building their own models, which entails learning special programming languages and building complex AI models that take incredible investments in time and knowledge.
Because current software can only simulate lithium based batteries, the gap between lithium-based and non-lithium based batteries is rapidly growing. Sodium ion batteries are one of the most well known non-lithium based chemistries, yet for every publication on sodium ion batteries there are twenty on lithium based batteries, because of a lack of fast research through simulation software.
There is so much investment in lithium-based batteries that within the next 5 years, it’s projected to hit approximately $130 billion, with a CAGR of 18%. [3] Lithium-ion batteries became the industry standard in the early 2000s, driven by their high energy density and suitability for portable electronics, and because of their pre-built supply chain and manufacturing market, they became the most viable battery for electric vehicles and renewable energy storage as both industries began to surge. Now, lithium-ion battery companies dominate the space and fund a large amount of battery research, meaning that the incentive for fast and efficient battery simulation systems comes from companies that benefit from the advancement of lithium-ion batteries.
As seen in the graph above, each different kind of battery excels in a different way, so the development of each one to solve smaller problems in the EV and grid storage industries is incredibly important to creating a multifaceted solution to a globally impactful problem.
To give a concrete example of someone who may face this situation, take for example a PhD researcher who thinks sodium ion batteries have high potential, but wants to look specifically at the anodes and how modifying them will impact the battery’s performance in higher temperature applications. Because there are no available simulation software, that researcher with either need to hard code their own software or perform all experiments manually. This will take additional time and money, and slow down the advancement of this technology. Now scale this up to the thousands of researchers who are working on sodium ion batteries. And then to the hundreds of thousands working on non-lithium based batteries, all of which are spending months on simulations that with better software could take them literal seconds.
Our solution is to develop a simulation software for all battery chemistries, not only lithium based ones. This will open up the battery industry and make space for novel chemistries to gain more traction, and have to potential to solve problems we are currently facing in less time, with less money wasted.
The user first inputs the material parameters for the battery they are building
Using machine learning, we can classify the kinds of chemical reactions happening, and how they would work in the constraints of a battery system
The user defines the parameters they want the battery to be used for/with