Artificial intelligence will help in the creation of safe lithium-ion batteries.

Many research groups from different countries, using a heavy and costly trial and error, spent a lot of years in search of a more secure alternative to liquid electrolytes used in today’s lithium-ion batteries. And recently, researchers from Stanford University have narrowed the range of the search for a suitable composition of the solid electrolyte from a few tens of thousands to just two dozen. And the artificial intelligence system, that has passed the pre-learning process and the subsequent self-learning, helped them in this.

“Lithium ions in the electrolyte are constantly moving between positive and negative battery electrodes,” – says Austin Sendek, a leading researcher – “Liquid electrolytes are cheap to produce and have excellent ion conductivity, however, they may catch fire because of the battery overheating“. It should be noted that the high flammability of lithium-ion batteries has led to the recall of nearly two million smartphones Samsung Galaxy Note7, the most widescale case in the history related with the inflammation of the lithium-ion batteries.

Solid electrolytes have more advantages than liquid electrolytes. Their main feature is a high stability, because the probability that the solid will evaporate, ignite or explode, much lower. Additionally, solid electrolytes are more durable from a mechanical point of view and the battery would be more durable.

As mentioned above, the amount of variants of solid lithium electrolyte composition is tens of thousands and scientists are exploring each of these compounds. To simplify this process, researchers from Stanford University have used an artificial intelligence system that has been trained on the data set, collected by scientists in the experiments. The analysis of these data allowed the system of artificial intelligence to develop a series of fairly complex criteria.

The creation of artificial intelligence systems has been quite a challenge, but an even more difficult task is preparation of database for training. Austin Sendek group took over two years to collect all known scientific information on this subject and its systematization.

As a result of all this titanic work, the artificial intellect system produced the analyzes of 12 thousand compositions of the solid lithium-containing electrolytes and left for further study 21 compositions. “The system took only a few minutes to make a full analysis of this” – says Austin Sendek, – “Much more time was spent on collecting all available data and the development of criteria by which was determined the quality of the test composition.”

And in the near future the researchers plan to perform laboratory testing of selected variants in order to choose the most suitable composition for practical application.