Electrolytes control how ions move between battery electrodes and strongly influence energy density, charging speed, safety, and lifetime. Yet electrolyte design remains difficult because small changes in chemical composition, structure, concentration, or operating conditions can produce large changes in performance. Across four recent studies, Dr. Zhilong Wang in our our group is developing a connected approach that combines artificial intelligence, molecular and materials modeling, chemical insight, and experimental validation to move electrolyte discovery beyond trial and error.
Building an AI Ecosystem for Solid-State Batteries
Our Science Advances article, “Toward AI ecosystems for electrolyte and interface engineering in solid-state batteries”, presents a roadmap for integrating AI across the solid-state battery research cycle. The study reviews advances in materials screening, machine-learning force fields, generative design, and electrolyte–electrode interface engineering, while highlighting the need for multiscale and multimodal models that incorporate physical constraints.
The goal is not a single stand-alone model, but an intelligent ecosystem in which data, simulations, experiments, and domain knowledge continuously inform one another.
Interpretable AI for Liquid Electrolytes
In Nature Computational Science, we introduced SCAN in the article “A dynamic routing-guided interpretable framework for salt–solvent chemistry”. SCAN is an interpretable AI framework for modeling salt–solvent chemistry in non-aqueous lithium-ion battery electrolytes. It treats salts, solvents, and operating conditions as distinct but interacting contributors to ionic conductivity.
SCAN reduced predictive error by more than 65% relative to leading machine-learning baselines and was used to map more than 11.5 million possible salt–solvent systems. Beyond prediction, the framework provides chemical insight into how molecular flexibility and ion–solvent interactions influence conductivity, helping translate AI results into practical electrolyte design principles.
IonNet Screens Solid Electrolytes from Composition Alone
Our latest Science Advances study, “Decoding the chemical space of fast-ion conductors via a descriptor-guided transfer learning framework”, introduces IonNet. Most AI tools for solid materials require reliable crystal structures, which may be unavailable for newly reported or hypothetical compounds. IonNet instead predicts lithium-ion mobility using chemical composition alone, enabling candidate materials to be evaluated much earlier in the discovery process.
IonNet identified 87 potential fast-ion conductors among approximately 4,500 stable compounds and nearly 63,000 candidates among about 5 million substituted compositions. Physics-based simulations evaluated 20 selected predictions and confirmed 13 as fast-ion conductors. The framework also extracts chemically meaningful design rules, allowing it to serve as a rapid front-end for prioritizing candidates before researchers commit substantial experimental or computational resources.
Reimagining Concentration Batteries through Electrolyte Chemistry
A complementary Nature Communications study, “Manipulating electrolyte solvation structures to build high-voltage concentration batteries for efficient energy storage”, demonstrates how electrolyte solvation structures can expand the design space of concentration batteries. Conducted with collaborators at the University of Puerto Rico–Río Piedras, the study examines batteries that use the same redox couple at both electrodes and generate voltage through differences in electrolyte composition—an approach traditionally associated with very low voltages.
By combining a highly concentrated positive-side electrolyte, or catholyte, with a strongly complexing negative-side electrolyte, or anolyte, the team increased the voltage of zinc- and copper-based concentration batteries to 0.6–0.7 V. When the electrolyte strategy was paired with conventional positive electrodes, it enabled aqueous full cells operating at 2.2–2.5 V. The results show that electrolytes can actively regulate electrode potential rather than functioning only as passive ion-transport media.
Together, these four studies advance a common vision: rational battery design guided by AI, chemistry, and physical understanding.
Selected Coverage
- Cornell Chronicle: AI plus chemistry can expand battery electrolyte design
- Cornell Chronicle: AI reveals chemistry behind high-performance battery electrolytes
- Tech Xplore: AI and electrolyte engineering open new paths for better batteries
- Phys.org: New AI framework reveals chemistry driving high-conductivity lithium-ion electrolytes