Microsoft's AI found a new material to replace li-ion batteries

Alfonso Maruccia

Posts: 819   +265
Staff
Forward-looking: Modern lithium-ion rechargeable batteries rely on lithium and other rare earth metals. While they offer an efficient energy source with a longer cycle life, they can also pose environmental concerns and prove challenging to recycle.

Microsoft collaborated with the Department of Energy's Pacific Northwest National Laboratory (PNNL) to discover a novel material that could alleviate the global dependence on lithium in the production of rechargeable batteries. Leveraging AI and the Azure Quantum Elements cloud platform, the Microsoft team significantly expedited the time-consuming research on materials not found in nature.

With Redmond's high-performance computing (HPC) systems, the AI algorithms were able to predict the characteristics of new materials – such as energy, force, stress, electronic band gap, and mechanical properties. Microsoft and PNNL trained the AI models using millions of data points from simulations, resulting in a 1,500 times faster prediction of materials properties compared to traditional density functional theory calculations.

The selection process began with 32.6 million candidate materials and AI algorithms identifying 500,000 predicted stable materials. After screening for functional properties, the pool was further narrowed down to 800 potential candidates. Utilizing "AI-accelerated" simulations to explore dynamic properties like ionic diffusivity, the Microsoft Quantum team narrowed down the selection to 150 materials.

With the synergy of AI, the Azure Quantum Elements platform, and established scientific expertise, Microsoft claims that modern technology can condense the next 250 years of chemistry and materials science innovation "into the next 25."

Practical considerations, including novelty, mechanics, and the availability of elements, were then factored in to identify a group of 18 top candidates. Leveraging PNNL's expertise and insights for additional screening parameters, Microsoft researchers pinpointed the final candidate: an electrolyte material that uses approximately 70 percent less lithium than existing li-ion batteries, substituting some lithium with sodium.

The new material has already been synthesized by PNNL, and additional tests are planned to verify its stability and efficiency. Brian Abrahamson, chief digital officer at PNNL, said developing novel batteries is a crucial global challenge. Abrahamson stated that synthesizing and testing materials at a human scale is a labor-intensive and fundamentally limiting process.

With the synergy of AI, the Azure Quantum Elements platform, and established scientific expertise, Microsoft claims that modern technology can condense the next 250 years of chemistry and materials science innovation "into the next 25." According to the Redmond corporation, AI will revolutionize every industry, unlocking a new era of scientific discovery.

Permalink to story.

 
The solution in search of a problem found another problem to tackle. If the verification pans out, then AI will have proved its worth in another area besides medical. PNNL has been big in the PHEV/EV arena for a long time.
 
Let's see what 8 billion unemployed and starving monkeys will have to say about AI...
They'll be too busy with AI generated porn to remember they're starving to death. But, to be fair, if AI could replace all jobs we could have a star trek like utopia if we could figure out the politics around it. The biggest problem with communism is that it requires slaves. If we could have AI communism without slaves then it might have a chance of working
 
This is pretty std AI , happens in the making of antibiotics etc

Material science is not even close to maturity. If it was you could work out the best alloys , the best super conductors etc. The computing power needed to calculate properties of materials is woefully inadequate

Just look at Water H2O, it has so many usual states and different pressures and temperatures. It's not just steam , liquid and ice. Add in the isotopes of H in different percentages to boot.

I think where it will get interesting is AI suggesting real experiments in fuzzy knowledge areas . Something seems to happen here.

Distillation , cracking etc must of seemed an art at one time , same as baking the perfect sponge , or making Damascus steel etc . yet most of the is well understood and exact science.
 
And...this will probably be "the next best battery" technology. We are still waiting for the last "best technology" when it comes to batteries, that we see stories on every month or so. ;)
 
This may seem a bit of cold water, but did anyone notice that sodium, by coincidence, just happens to be the next item up on the alkali metals list if the periodic table? Who would have thought it might behave similar to lithium? Chemistry anyone? My question isn't about AI finding it, my question is how did we avoid trying this for so long? Could it be they oxidize quickly and sodium gives off hydrogen gas if it gets wet?
 
The solution in search of a problem found another problem to tackle. If the verification pans out, then AI will have proved its worth in another area besides medical. PNNL has been big in the PHEV/EV arena for a long time.

Google's AI has been making amazing materials science discoveries for several years. It's helping scientists drastically slash the time and effort required to test potentially millions of theoretical compounds and helps them identify those materials with the desired properties much faster and. In one case it culled a list of a few million down to about 77 candidates.
 
This may seem a bit of cold water, but did anyone notice that sodium, by coincidence, just happens to be the next item up on the alkali metals list if the periodic table? Who would have thought it might behave similar to lithium? Chemistry anyone? My question isn't about AI finding it, my question is how did we avoid trying this for so long? Could it be they oxidize quickly and sodium gives off hydrogen gas if it gets wet?


Huh, Sodium Ion batteries have been worked on for a long time and are already in production. They have long had issues with energy density though as Sodium ions are much larger than Li ions. A Swedish company has largely overcome those issues but competing products are being produced where the energy density isn't as important.

It's hilarious think these material scientists don't know the properties of all the materials in the periodic table and need to be reminded by techspot people.
 
Huh, Sodium Ion batteries have been worked on for a long time and are already in production. They have long had issues with energy density though as Sodium ions are much larger than Li ions. A Swedish company has largely overcome those issues but competing products are being produced where the energy density isn't as important.

It's hilarious think these material scientists don't know the properties of all the materials in the periodic table and need to be reminded by techspot people.

Well - I know the sea rising is all fake as when floating ice floats , sea level doesn't rise .Ancient greeks know more than climate scientists , it's embarrassing how they know simple stuff like this .../s

Yes I saw people posting that many times in all earnesty.
 
The solution in search of a problem found another problem to tackle. If the verification pans out, then AI will have proved its worth in another area besides medical. PNNL has been big in the PHEV/EV arena for a long time.
I think you're confusing BlockChain with AI, AI has so many successfully and fundamentally valuable applications and have been applied for years on end. BlockChain on the other hand is the solution that has yet to find enough solutions for it to be widespread.

Let's see what 8 billion unemployed and starving monkeys will have to say about AI...
The only unemployed people, will be the ones not learning how to use AI to be better at their jobs or the ones not teaching AIs. AIs only gets rid of the useless jobs like data entry and all day meeting managers. Someone still has to teach the AI the correct way, that requires subject matter experts, engineers, data analysts. For every job replaced 4 more jobs will open up. Unfortunately what people don't realize is that technology is the future, virtually every job will require some kind of programming literacy eventually. People just haven't realized it yet.
 
I think you're confusing BlockChain with AI, AI has so many successfully and fundamentally valuable applications and have been applied for years on end. BlockChain on the other hand is the solution that has yet to find enough solutions for it to be widespread.


The only unemployed people, will be the ones not learning how to use AI to be better at their jobs or the ones not teaching AIs. AIs only gets rid of the useless jobs like data entry and all day meeting managers. Someone still has to teach the AI the correct way, that requires subject matter experts, engineers, data analysts. For every job replaced 4 more jobs will open up. Unfortunately what people don't realize is that technology is the future, virtually every job will require some kind of programming literacy eventually. People just haven't realized it yet.
Programming is a specialized field that not many are well suited for. You're suggesting the vast majority will go hungry for the technocracy of the future.

Hint: this has traditionally not gone down well throughout history.
They'll be too busy with AI generated porn to remember they're starving to death. But, to be fair, if AI could replace all jobs we could have a star trek like utopia if we could figure out the politics around it. The biggest problem with communism is that it requires slaves. If we could have AI communism without slaves then it might have a chance of working
The biggest issue with communism isnt the need of slaves, it's human nature. Communism doesnt work when greed, jealousy, and power dynamics exist. What would result would be a copy of the soviet union, only with even worse inequality and more tyrannical control. A lack of labor jobs is directly opposed to communism, without that labor those people become "useless". Either they get fed to the blood gods, or you end up with a soviet style bureaucratic nightmare where 5 people have to sign off on your grocery purchases, like moscow in the 80s.
 
I've gotten somewhat annoyed at hype science breakthrough articles with no follow-up (because of failure to perform or whatnot). Please be responsible and write a follow-up whether the material succeeds in moving to production and market, or not, and why.
 
Back