Machine Learning Frustrated Electronic correlations

Understanding the emergent phases of frustrated, strongly correlated electrons remains a major challenge in condensed-matter physics, with direct relevance to organic superconductors and layered quantum materials. This project leverages advanced Machine Learning techniques, specifically Neural-Network Quantum States  optimized via natural-gradient methods and neural importance sampling, to resolve the ground-state and low-energy properties of the anisotropic triangular-lattice Hubbard model. By combining Pfaffian-Transformer wavefunctions with large-scale GPU-accelerated variational Monte Carlo, we aim to obtain unbiased, high-accuracy characterizations of the ground-state properties of the Hubbard model on geometrically frustrated lattices. The project will also compute linear-response susceptibilities, providing direct links to experimentally observable signatures.