Energy Determining the age of lithium batteries with face recognition

Lithium-ion batteries are becoming increasingly powerful but are susceptible to ageing processes. These processes must be better understood at the atomic level to be able to use the batteries more sustainably. To this end, BAM has developed an innovative method based on face recognition algorithms.

Charging or discharging an electric battery follows a simple mechanism: lithium ions move back and forth between two electrodes releasing or accepting electrons in the process. This redox reaction, as it is called, supplies a mobile phone, a laptop or an electric vehicles with energy. The lithium ions are deposited in the grid-like structure of the anode and cathode like the cavities of a porous sponge. The more of them the “sponge” can absorb, the more powerful the battery cell becomes. Over time, however, tiny fractures and cracks appear in the delicate structures of the electrodes. As a result, an increasing number of lithium ions can no longer fit into the cavities and, instead, they accumulate around the electrodes. These accumulations in turn become an obstacle for other lithium ions which impairs the performance of the battery so that it has to be recharged at ever shorter intervals.

Ageing processes in fast motion

The accumulations of ions show characteristic patterns in the process. This is because lithium occurs in nature in two different isotopes: the lighter lithium-6 can move through obstacles easier than the heavier lithium-7 and enters the cavities of the anode more easily during charging and those of the cathode during discharge. ”We now know that the distribution of lithium isotopes in a cell is directly related to its age condition,” explains BAM chemist Carlos Abad. “The number and speed of charging cycles probably play a role, but so does the ‘active material’, a conductive electrolyte through which the lithium ions migrate from electrode to electrode.” Abad and his colleague Dalia Morcillo have analysed the ageing processes of a lithium-ion battery in more detail in the laboratory. They have performed many charge and discharge cycles over a very short time and then determined the distribution of the lithium ions.

Algorithms can interpret images faster

The BAM scientists use spectral analysis for this purpose. Lithium ions are excited by light energy that is absorbed by the two isotopes lithium-6 and lithium-7 at different intensities. “The images of this isotope distribution are extremely difficult to interpret with the naked eye on a computer,” explains Dalia Morcillo. “They resemble each other like the faces of twins or those of very closely related people. Differences are virtually undetectable.” This gave the team the idea of using face recognition algorithms to evaluate the measurements. They first “trained” their measuring instrument with reference materials: mixtures of lithium isotopes whose ratio was precisely known. Then they programmed the device using machine learning methods until it was able to recognise the differences between various quantity ratios better and faster. “As a next step, we would like to provide companies developing lithium-ion batteries with a faster and more cost-effective method for assessing and evaluating the ageing behaviour of their batteries at an early stage in the laboratory,” says Dalia Morcillo. “This should make it possible to produce longer-lasting and thus ultimately more sustainable batteries more quickly in the future.”