While the structure of a lithium-ion battery can be simplified as shown in the diagram below, in reality, the cathode and anode—along with components such as separators, electrolytes, and current collectors—have complex geometries and heterogeneous microstructures. Evaluating their performance requires considering their multiscale characteristics as well as the manufacturing process, and applying a wide range of methods and technologies.
With JSOL’s simulation software suite, you can run simulations tailored to each scale: from nanometers, capturing electronic states and molecular structures, to micrometers, modeling phase separation and composite materials. By linking these tools, it is possible to analyze the mechanisms behind the performance of advanced materials directly through simulation.
Today, there is growing demand for data-driven materials development using AI technologies, under concepts such as Materials Informatics and Process Informatics. JSOL’s AI-enabled software allows you to predict properties based on molecular and crystal structures, composition ratios, and processing environments—and even perform inverse design. When experimental data is scarce, high-throughput simulations can be executed to supplement datasets and accelerate discovery.

In materials design, it is essential to determine which scale—from nanometers to micrometers—has the greatest impact on material properties.
No single simulation method can cover all scales and phenomena. That’s why we offer multiscale-ready software, combining multiple engines (solvers) into a single materials design solution.
This enables comprehensive simulation capabilities to analyze the multiscale characteristics of a wide range of materials.
Learn more on each product’s dedicated page.

J-OCTA MI-Suite includes machine learning capabilities for predicting material properties using molecular/crystal structures and various conditions as explanatory variables, along with related features such as molecular descriptor calculations, access to public databases, and analysis of each variable’s contribution.
When experimental data are insufficient, the software can supplement them with simulation results. It also provides API functionality to support database construction using physical simulations.
Data selection and learning condition settings for machine learning can be performed via dialogs, without the need for programming. Multiple pre-trained models are included, allowing users to start Materials Informatics projects even without their own data.