Abstract
My research is motivated by a question arising in first-principles calculations and dynamical mean-field theory: how can we represent and compute Green's functions and response functions efficiently without losing the underlying physics? This challenge becomes particularly acute in strongly correlated systems at low temperatures, in two-particle quantities, and in nonequilibrium dynamics.
This research grew out of materials studies of correlated electron systems, including pyrochlore oxides and LiV2O4, together with the development of practical tools for DFT+DMFT and quantum Monte Carlo. These efforts led to the intermediate representation (IR) basis, which provides a compact and system-independent description of imaginary-time Green's functions, and more recently to quantics tensor train (QTT) representations and tensor cross interpolation (TCI), which extend these ideas beyond imaginary time and provide a scalable route to multiscale structure in two-particle and nonequilibrium problems.
A central aspect of my work is that methods become most useful when embedded in shared software and a collaborative community. This perspective has driven the development of open-source ecosystems such as sparse-ir and DCore, as well as broader community-building efforts including tensor4all, workshops, schools, and hackathons.
At HKUST, I hope to build a research program at the interface of condensed matter theory, applied mathematics, and scientific computing, developing compact and scalable methods for quantum many-body problems through open-source software, AI-assisted development, and broad collaborations.
Hiroshi Shinaoka is a Professor of Physics at Saitama University, Japan, where he has been based since 2015. He received his PhD from the University of Tokyo in 2009 and held postdoctoral positions, including in Switzerland from 2012, working on quantum Monte Carlo and dynamical mean-field theory with Matthias Troyer and Philipp Werner. His research has expanded from first-principles studies of correlated materials such as pyrochlore oxides toward computational methods, emphasizing compact representations of imaginary-time response functions. He pioneered the intermediate representation (IR) basis and quantics tensor train (QTT) approaches and leads the open-source ecosystem including sparse-ir and DCore. He co-founded the tensor4all collaboration, an international effort to develop tensor-network methods based on quantics tensor trains and cross interpolation. He is also promoting AI-assisted open-source software development for computational physics.