Billions of tiny particles in the electrodes of rechargeable lithium-ion batteries are responsible for storing charge and putting it to work when needed. X-ray movies of this process show that the particles absorb and release lithium ions as the battery charges and discharges. Now, researchers have used a machine learning technique called computer vision to dig deeper and analyze every pixel of these X-ray movies, uncovering previously unseen physical and chemical details of the battery cycle, a major step forward.
Researchers from the U.S. Department of Energy's SLAC National Accelerator Laboratory, Stanford University, MIT, and Toyota Research Institute report on September 13 in the journal Nature that the new approach has suggested a way to enable billions of nanoparticles in a lithium-ion battery electrode to store and release charge more efficiently.
"It's now possible to produce beautiful X-ray movies of battery nanoparticles in action, but these movies are so information-intensive that understanding the subtle details of how the particles function is a real challenge," said William Chueh, associate professor at Stanford University, SLAC department scientist and director of the SLAC-Stanford Battery Center, who co-led the research with MIT professor Martin Bazant.
"Now we can gain insights that weren't possible before," Chueh said. Our industry partners need this essential science-based information to develop better batteries faster. "
More broadly, the researchers say, this method of discovering the physics behind complex patterns in images could even provide unprecedented insights into other types of chemical and biological systems, such as cell division in developing embryos.
See-through battery reveals secrets
The battery particles studied by the research team are made of lithium iron phosphate, or LFP. They are packed into the positive electrodes of many lithium-ion batteries by the billions, each coated with a thin layer of carbon to improve the electrode's conductivity.
To see what's going on inside the battery as it works, Chueh's team created tiny transparent cells in which two electrodes are surrounded by an electrolyte solution filled with free-moving lithium ions.
When the battery discharges, lithium ions flow into the positive lithium-ion battery electrode and become lodged in its nanoparticles like a car in a crowded parking lot, a reaction known as intercalation. When the battery charges, the lithium ions flow out again, reaching the opposite negative electrode.
A team of researchers from SLAC, Stanford University, MIT, and Toyota Research Institute used machine learning techniques to reanalyze X-ray movies like this one pixel by pixel, discovering new physical and chemical details of battery cycling. This animation is based on X-ray images produced by the team in 2016. It shows how some of the billions of nanoparticles in a lithium-ion battery electrode charge (red to green) and discharge (green to red) as lithium ions flow in and out, and reveals how uneven the process is inside individual particles. Source: SLAC National Accelerator Laboratory
Brian Storey, senior director of energy and materials at Toyota Research Institute, said: "Lithium iron phosphate is an important battery material because of its low cost, good safety performance and abundant use of elements. We are seeing increasing use of LFP in the electric vehicle market, so the timing of this research could not be better."
Collaboration history and previous work
Chueh and Bazant began collaborating on battery research eight years ago. Bazant has done extensive mathematical modeling of the patterns formed by lithium ions as they move in and out of LFP particles. Chueh has been using an advanced X-ray microscope at Lawrence Berkeley National Laboratory's Advanced Light Source to take nanoscale movies of battery particles at work, with details as small as one billionth of a meter.
In 2016, their research team published groundbreaking nanoscale movies showing how lithium ions move in and out of individual LFP nanoparticles.
Then, with funding from Toyota Research Institute, the team began using machine learning tools developed at MIT to greatly speed up the process of testing the batteries and sifting through many possible charging methods to find the most efficient one. They also combined traditional machine learning, which looks for patterns in data, with knowledge gained from experiments and equations guided by physics to discover and explain the processes that shorten the life of fast-charging lithium-ion batteries.
Pixel-by-pixel analysis
In this latest study, Chueh and Bazant used computer vision, a subfield of machine learning, to mine more detailed information from 62 nanoscale X-ray movies they took in 2016 of lithium-ion battery particles charging or discharging. Each still image in these movies contains about 490 pixels - the smallest unit of information that can be obtained from an image, whether it's a detector shot with X-rays or visible light shot by a smartphone camera. This gives them approximately 180,000 pixels of information.
The team used these 180,000 pixels to train their computational model to generate equations that accurately describe how the lithium insertion reaction proceeds. They found that the motion of ions within the LFP particles matched Bazant's computer simulation predictions very well.
"Every little pixel inside is jumping from full to empty, full to empty," Bazant said. "We're mapping the entire process, using our equations to understand how this happens."
"The new technology revealed some previously unseen phenomena, including changes in the rate of lithium insertion reactions in different regions of a single LFP nanoparticle," Bazant said. "Some regions appear to react very quickly and others very slowly."
The most important practical finding of the paper is that changes in the thickness of the carbon coating of LFP particles directly control the entry and exit speed of lithium ions, which may lead to more efficient charging and discharging.
The scientists learned from this study that what controls the battery process is the interface between the liquid electrolyte and the solid electrode material - where intercalation reactions and changes in the thickness of the granular carbon coating interact in complex ways. This means that the next step should really be on engineering the interface.
Toyota Research Institute's Storey added: "The publication of this paper is the culmination of six years of hard work and collaboration. This technology allows us to uncover the inner workings of a battery in a way that has never been done before. Our next goal is to improve battery design by applying this new understanding."