Artificial Intelligence Efficiently Promotes Development of Lithium Battery Electrolytes
- Mar 30, 2018 -

Artificial Intelligence Efficiently Promotes Development of Lithium Battery Electrolytes

28th, Mar

Fujitsu Ltd. and the Japan Institute of Physical Chemistry recently announced that their joint research team applied first-principle calculations and artificial intelligence techniques in material design to make prediction, synthesis, and evaluation tests of the solid electrolyte composition of all-solid-state lithium ion batteries. Actual verification was performed. The results show that, even with less data, by combining with artificial intelligence methods, the best material composition can still be efficiently found, and the material development speed can be greatly improved.

 

To date, the development of materials has to rely on the long-term experience and keen intuition that researchers have accumulated. Many lessons need to be accumulated to succeed. The first-principles calculation is that if the composition of the material is specified, based on the predictable characteristics of quantum mechanics, the best composition of the new high-function material can be predicted before the experiment, thereby greatly reducing the number of failed experiments. However, the load of the first-principles calculation is very large, and the various components of the material require multiple calculations and will take a very long time.

The research team hopes to solve problems in material development through close integration of material simulation, experiment, and artificial intelligence, which will greatly shorten the time for material development, so as to make it easier to discover unexpected compositions and crystal structures and create new high-functional materials.

 

The research team used a combination of Bayesian inference methods, one of the artificial intelligence methods, to control the number of calculations for the first-principle calculations and to predict the three lithium-containing oxo-synthetic compounds of solid-state lithium-ion battery solid electrolytes. The results confirm that this method can predict the best combination of high lithium ion conductivity within achievable time. At the same time, high lithium ion conductivity of other compositions was also found near the predicted composition.

 

Lithium ion conductivity is one of the important features of solid electrolyte materials and is a factor that dominates the charge and discharge rate of lithium batteries. The results of this study verified that the use of material simulation and artificial intelligence methods can efficiently develop liquid-tight, non-pyrogenic lithium-ion batteries, and in the future, it is expected to exert great potential in battery, semiconductor, and magnetic materials.