Multiobjective Electrolyte Solvent Designs via Physics-Based Modeling and RL

A dataset of 200 candidate molecules for electrolyte solvents, generated and validated via a reinforcement learning framework. The molecules were designed to optimize multiple properties, including viscosity, boiling point, and HOMO energy, using molecular dynamics and density functional theory simulations. The dataset was created by Mohammad Atif Faiz Afzal and last updated on 2026-04-23.

Use Cases

• Training surrogate machine learning models for molecular property prediction based on SMILES strings.
• Benchmarking multiobjective optimization algorithms for materials design.
• Identifying candidate molecules with low viscosity and high boiling point for battery electrolytes.
• Exploring the trade-off between oxidative stability (HOMO energy) and fluidic properties in solvent design.

Strengths

• Contains 200 top-performing candidate molecules validated with physics-based simulations.
• Molecules were generated via a reinforcement learning framework explicitly designed for multi-property optimization.
• Properties were computed using established methods: molecular dynamics for viscosity/boiling point and density functional theory for HOMO energy.

Limitations

• The dataset is very small at 32.7 KB, indicating limited scope.
• Row count and specific column-level documentation are unknown, requiring manual inspection after download.
• The description does not specify the chemical space or diversity of the 200 generated molecules.

Provenance

Source

figshare

Collection Method

Molecules were generated de novo using a reinforcement learning framework incorporating trained ML models for viscosity, boiling point, and HOMO energy.

Freshness

Last updated 2026-04-23 19:43:51.