Microsoft’s AI found a new material to replace li-ion batteries
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
Why is this time consuming? Why does it take lot of scientists and long time to find new batteries and material.
Why do they need AI and computers to speed this up? Why does it take so long to find new batteries and material?
Why is chemistry so hard that it takes very long time to find new chemicals and batteries?
I’ve seen a number of articles on using sodium instead of lithium in batteries.
The problem with any material is it has to beat the current tech in several aspects - cost, reliability/durability, practicality, weight, safety, charge capacity, number of lifetime charge cycles. There have been frequent news stories about assorted breakthroughs that have yet to appear. Typically high capacity seems to come at the expense of lifetime or reliability. Figuring out how to mitigate these is an ongoing focus of research.
It depends how much chemistry is understood and scientists know most chemical combinations. If there is trillions of different chemical combinations it is going to take army of scientists a very long time well on other hand if there is only million different chemical combinations it is not going to take to long.
Well most chemical in nature is well known so most likely any new battery or materials is going to be made in lab.
Plus, it ideally doesn’t require raw materials which are rare and/or difficult to process (which affects cost), and isn’t environmentally problematic, either for production or disposal.
The problem is that there is no single property of a chemical that defines “good battery”. It’s a combination of many different properties, and there are many such combinations, and it’s impossible to predict exactly which one is right.
Of course we can rule out obvious things. Helium won’t make a good battery, alcohol won’t make a good battery. A potato can be a battery, but it’s not optimal for most consumer electronics. So that helps us limit the possibilities, but there are still thousands of years of possibilities that would need to be tested to find an exact fit.
However the AI is a fantastic tool for going through many different combinations that have never been tested, considering all their different properties, rejecting all the potatoes and pointing out some new candidates that might work.
It’s not really the potato that’s the battery-- It’s the pieces of copper and zinc that you poke into it. Which make a much better battery if you instead, say, poke them into sulfuric acid.
Well, yes, but the potato isn’t even the important part of the battery. What makes a battery a battery is the different metals of the electrodes. Put the electrodes in almost anything, and it’ll produce a voltage.
In retrospect, I DO find it remarkable how little practical advance we have achieved in battery tech in the past 10 or so years, given that this is probably the single hottest scientific and commercial area on the planet (well along with AI)
.
In order to build EV you need to assure batteries … and if your batteries are 20% better than other’s, your car will be faster, lighter and have more range … so batteries make you win in the EV market-space
however, your avg. 18650 (or 21700) top battery still has the same capacity as 10 years ago …
the “past me in 2014” would have thought that battery-capacity for LiIons is now 50% higher …
possibly and industry with lots of leading factors - slow to ramp up …
As an example of how long things take to develop, here’s a quote from Wikipedia about lithium-ion batteries, “after their market introduction in 1991: within the next 30 years, their volumetric energy density increased threefold while their cost dropped tenfold.”
Why do electric utilities use lithium ion batteries? Surely there’s a battery technology that performs better on cost and worse on weight.
A couple of the many nondefinitive articles on the subject:
The answer is going to be, “It’s easier to take something off the shelf than develop something new,” a variant of, “first mover advantage”, but I’m still somewhat surprised.
Maybe, but that doesn’t mean they’re also available at commercial grid-scale quantities. Also, what may be an acceptable risk at the scale of a laptop or an e-bike may be very much not-acceptable when you fill a building with them.
I think the fundamental problem with developing a battery for something like an EV, or any battery for that matter, is that you want one that can charge quickly, provide a lot of power, and doesn’t have a propensity to explode. Anything with those three properties (or frankly, even just one of them) tends to want to blow up. The Goldilocks zone is quite small and comes with many tradeoffs.
thats prob. rooted in the fact that the first gen of Lit-Ions were really very poorly (easy to improve upon) … fact is that the mentioned 18650/21700 did not increase their vol. energy in the past years … possibly an indicator of being at a dead end / peak for the current technology.
To that point, Li-ion batteries were discovered about twenty years earlier than that. It took until 1991 to commercialize them. Materials discovery is hard. AI can be a useful screening pool to winnow down the candidates from possible combinations, but it’s still hard. The original proof of concept was unstable and dangerous. It took decades to figure out which cathode, anode, and electrolyte similar to the original discovery were stable enough, compatible enough, manufacturable enough, cheap enough… to make a viable battery.
Part of it is that it’s not just “chemistry”. Any given chemical formation may have many different structural configurations that have very different properties. Then, even if you knew you wanted a particular formula in a particular configuration, you have to figure out how to make it. That’s processing and there are essentially an infinite number of processing pathways, most of which will not take you to the material you want.
I love it when non-scientists and non-technical people wonder why science takes a long time…
FYI - science doesn’t work like it does in the movies, where a couple of plucky high schoolers can invent the cure for the zombie apocalypse in the chem lab in a couple of hours.
As stated above, the easy battery technologies have been discovered already. We’re now at the bleeding edge. This is difficult.
My lab has been trying to crack a particular biology problem for over 8 years now. We’re getting closer but we’re not there yet. Science is hard.
Indeed. And then you get the popular science communicators who seem to provide the illusion that science begins and ends with theoretical physics. Just sort of out quantum gravity and science is done.
Using AI to sift through the huge space of structural chemistry is an obvious use. The basic idea gets traction in a lot of places where the chemistry is otherwise intractable. You do of course need good data to work against. Which is a whole other problem.