Research Projects

← Back to Homepage

1. DFT-Net: Specialized-Generalized DFT Foundation Model

Developing DFT-Net, an AI foundation model that intelligently constructs high-level density functionals.

Goal: Establish a new AI-DFT paradigm achieving chemical accuracy (~1 kcal/mol) with first-principles interpretability.

2. Precision Chemistry Databases

Leveraging the REST platform to generate high-fidelity datasets for complex chemical systems.

• 2.1 High-efficiency implementation of high-level wavefunction methods
• 2.2 Quasiparticle (BSE@GW) framework for IPs, EAs, and excited states
• 2.3 Natural language-driven intelligent computing platform & database agent

3. Physically Constrained High-Level Functional Descriptors

Designing novel descriptors incorporating unoccupied orbital and multi-configurational information (top rung of Jacob’s Ladder).

• Fifth-rung descriptors beyond PT2 (BGE2, osRPA)
• Multi-configurational descriptors from MC-PDFT (natural orbital entropy)
• QP-inspired descriptors from GW/BSE

4. Next-Generation Intelligent Electronic Structure Platform

Integrating Projects 1-3 into a unified ecosystem.

• Evolve REST into an intelligent electronic structure platform
• Cloud-based high-throughput computation
• Closed-loop “data → algorithm → theory” framework

← Back to Homepage