A Crystal Ball for Chemistry

Early attempts to make a lithium-ion battery were a misfire — the first ones created by Exxon chemist Stanley Whittingham in the 1970s short-circuited and burst into flames. After years of experimentation, the end result was a success. If you’ve got a phone in your pocket, a laptop or tablet on your desk, a cordless tool in your garage, or an electric car in your driveway, you benefit from lithium-ion batteries.

Lithium-ion batteries are our best option for powering our devices, and their inventors received the Nobel Prize in 2019.

But they have their downsides. They are expensive, short-lived, and fragile, requiring protection from overcharging. Mining lithium leads to soil degradation, ecosystem failure, water shortages, biodiversity loss, and climate change. Nearly 600,000 gallons of water are needed to produce one ton of lithium.

In the next few years, better options may emerge. And UNC-Chapel Hill PhD student Jack Sundberg is giving a huge assist.

Energized by innovation

A few years ago, UNC-Chapel Hill chemist Scott Warren and his lab began working with Honda, Caltech, and NASA’s Jet Propulsion Laboratory to develop a better battery — one powered by fluoride instead of lithium.

The first fluoride-ion battery was introduced in 2011, but with a lot of kinks that still need to be worked out, like that it required operating temperatures of 150 degrees Celsius — that’s over 300 degrees Fahrenheit. When scientists developed a successful liquid electrolyte in 2018, a commercial fluoride-ion battery seemed truly possible.

Fluoride is incredibly promising. Not only is it more abundant than lithium in Earth’s crust, but it offers higher theoretical voltages. Plus, even if it doesn’t outperform lithium in every way, it will ease the burden on lithium supplies.

For fluoride-ion batteries to compete with lithium, finding better conductors is key.

To understand how batteries work, imagine a landscape with a hill, a valley, and a ball — where the hill is the anode, the valley the cathode, and the ball is an ion. When the ball is on the top of the hill, the battery is fully charged. As the ball rolls down the hill, the battery powers a device. And to move the ball between the two, you need an electrolyte.

While electrolytes connect the two electrodes, conductors are the materials the electrolyte needs to flow through to make the battery function. To produce a successful battery, the electrolyte must be stable, powerful, and function at room temperature.

Enter Jack Sundberg, a PhD student in Warren’s lab. While working on a project to find better conductors for fluoride, Sundberg grew frustrated with the trial and error of sifting through more than 10,000 potential options, resulting in an archaic guessing game that siphoned away hours and hours of time.

Armed with a fascination for computational chemistry, Sundberg began developing software that could help whittle down the list of the best conductors for fluoride. What he developed allowed him to randomly select 300 of those 10,000 materials, complete an accurate calculation of each one’s ability to transport fluoride, and then use those calculations as a benchmark.

Now, he and Warren are working with the UNC Office of Technology Commercialization to patent a collection of promising fluoride conductors — some of which, on paper, could outperform all known conductors.

Perhaps more importantly, Sundberg’s software — called Simmate — can be used by chemists to predict properties for any material in a way that’s time-sensitive and cost-effective.

“The one part I’ve never liked about lab work is the trial and error,” Sundberg says. “Even the best chemists experience an element of luck when exploring new materials. I wanted to know for sure if a material would work before starting an experiment. And just obsessing over that is how I created Simmate.”

Powering materials chemistry

Time and cost are huge considerations when chemists are looking for new materials to use in their experiments. They do this by running calculations via existing software programs. Small calculations on a laptop take about 20 minutes. On a supercomputer, those same calculations may take seconds — but cost more money to run. Even with help from a supercomputer, larger calculations often take months.

“At the start of this research, I found this same problem over and over again,” Sundberg shares. “I wanted to be absolutely sure a material would work before dedicating several months to it. That slowly turned into an obsession of guiding real syntheses with theoretical predictions.”

The existing software to run these experiments was over 10 years old, and Sundberg began to quickly identify the pain points.

“A lot of the software is tiny little codes or programs that you need to fit together on your own to achieve a goal,” he says. “It wasn’t easy for a beginner to use.”

And Sundberg was a beginner. He didn’t know anything about computational chemistry before entering the lab.

“I had never been exposed to it at all,” he admits. “My only computational background was a computer science 101 class in undergrad, without any chemistry overlap. So until grad school, my entire chemistry research experience was at a lab bench.”

But that didn’t stop him. These problems launched Sundberg into action, and he began building a software that scientists of all levels can navigate. By the end of his second year in the Warren Lab, Sundberg had fully transitioned into computational research, guiding his lab mates’ experiments rather than his own. Simmate is now the focus of his dissertation.

“To be honest, I would never have recommended such a project to Jack,” Warren admits. “As someone deep in the chemistry side of things, it was truly outside of my wheelhouse — but Jack saw a major opportunity and he went for it. After about six months of independent work, Jack had a working version of his software and was able to apply it to the problem of finding improved materials for a fluoride-ion battery.”

Sundberg’s software has helped the Warren Lab discover a handful of materials for moving fluoride ions within a battery. Some were already known and widely used, while others were theoretical structures that hadn’t been synthesized before.

“Because our predictions suggest these materials could outperform known conductors, we were confident that our findings would have a big impact on the field and, in the long term, have a key role in the technology as it moves from proof-of-concept to an industrial product,” Sundberg says. “Even though our predictions were all theoretical, the results were robust and exciting enough to move forward with patenting.”

Warren’s lab has one year from the submission of the patent to turn Sundberg’s theoretical predictions into real-world results in the lab. If they can make the materials and prove that they are viable fluoride conductors, it could greatly propel production of an effective battery.

“It’s a pretty crazy feeling to be playing around in the lab one day and then submitting legal patents the next,” Sundberg says. “It makes you reflect on the real-world impact of daily research.”