AI has been used to create find and create a new material that could revolutionise not just the humble battery, but the time required to find new breakthroughs   

On January 9, this year, the Pacific Northwest National Laboratory (PNNL) and Microsoft announced a breakthrough in the quest for a new material to reduce the role of lithium in battery technology.

Artificial Intelligence

The involvement of Microsoft points to the most interesting factor in this announcement – the new material was not discovered after decades of meticulous research, but over the space of 80 hours. It took only nine months from project launch to building a prototype battery and using it to power a light bulb. The key was the use of generative artificial intelligence, provided by Microsoft’s Azure Quantum team.

Firebug

The market for the soft, silvery metal (sometimes referred to as ‘white gold’ because of its colour and value) is a valuable one. The market for lithium-ion batteries was valued at $46 billion in 2022.  Forecasts put that value at $189 billion by 2032. 

While it may be useful in batteries, lithium does bring problems and there is an effort to find a replacement. The metal is rare and, as a vital resource, is a geopolitical hot potato.

Demand for lithium is expected to grow rapidly, leading the International Energy Agency to warn of a shortage by 2025.  Indeed, the US Department for Energy expects our hunger for lithium to grow tenfold by 2030.

As a battery, it takes a long time to charge and does not hold that charge for very long.  Alarmingly, modern lithium-ion batteries are a notorious fire hazard and can unexpectedly burst into flames.  In 2023, the New York fire department reported that 18 people died in fires linked to electric-vehicle batteries.

Mining the metal is time consuming, taking years to develop a mine, and requires large amounts of water and power to extract. The process can scar and pollute the landscape.

The process of how

Microsoft launched its Azure Quantum Elements (AQE) platform, intended for use in chemistry and materials science, in June last year. It uses high performance computing (roughly equivalent to 5,000 laptops) to speed up the discovery process.  

The task Microsoft’s team set themselves was to find materials that can be used in a sold state battery.

The process went as follows; researchers at Microsoft asked the AI to list battery materials that use less lithium.  That produced 32 million results.  The search was filtered to looking for materials that are stable enough to be useful, which produced a list of 500,000 possible candidates. Extra search filters for conducting energy, simulation of the movement of atoms within the materials.  Cost and material availability were other filters.

This whittled the prospects down to 23 candidates.  All in all, the process lasted 80 hours, something not possible without AI and Microsoft’s AQE.

When they had a result, the team consulted the Pacific Northwest National Laboratory (PNNL), to prove the validity of the idea that computers can give research answers.

We need a better name

One promising prospect was chosen from the shortlist and synthesised before being used to power a light bulb and a clock.  It is hoped the new material, currently called N2116, will reduce the lithium content of a battery by 70%.  The next hope is to ditch the liquid electrolyte and use N2116 to create a solid-state battery.

While solid state electrolytes are not as energy conductive as their liquid counterparts, there are other candidate materials that could be tried, and the AI can radically improve the discovery time.  The contradiction and the challenge here is that artificial intelligence itself can consume an enormous amount of energy to carry out its work – something that makes the increased efficiency of data centres, batteries and the switch to cleaner energy all the more important.