Hao Zeng, an N.E.T. master’s student, joined Professor Xiang Sun’s research group at NYU Shanghai in Fall 2023. His work has focused broadly on quantum dynamics, with each project building naturally on the progress and insights gained from the previous one. He began by studying machine-learning approaches for quantum dynamics in open quantum systems. As this work advanced, he gradually took on related projects involving generalized quantum master equations, transfer-tensor methods, and numerically exact tensor-train techniques.
Although these projects use different computational and theoretical tools, they are closely connected by a common scientific goal: understanding and efficiently simulating complex quantum dynamics, especially in systems where environmental effects and memory play important roles. This continuity has allowed Hao to develop a coherent research profile rather than a collection of isolated projects. It has also enabled him to move progressively from data-driven modeling to physics-informed methods and rigorous numerical benchmarks.
One major direction of his research has been the use of neural networks to model non-Markovian quantum dynamics. In this work, Hao examined how the effectiveness of different neural-network architectures depends on the memory characteristics of the physical system. His results provided practical guidance for choosing machine-learning models while also revealing how physical structure can inform machine-learning model design.
He later extended this line of research to quantum wavepacket dynamics and nonadiabatic semiclassical dynamics. His paper, “Data-Driven versus Physics-Informed Neural Networks for Nonadiabatic Semiclassical Mapping Dynamics,” received the 2025 Outstanding Paper Award from Communications in Computational Chemistry. This recognition reflects both the scientific quality of the work and the broader relevance of combining physical principles with modern machine-learning methods.
Another important milestone is the acceptance of his paper, “Estimating Memory Time within the Frameworks of Generalized Quantum Master Equation and Transfer Tensor Methods,” in the Journal of Chemical Physics, 164, 224117 (2026). The study addresses a central problem in non-Markovian quantum dynamics: how to estimate the timescale over which a system retains memory of its earlier evolution. By connecting generalized quantum master equation and transfer-tensor method frameworks, the work contributes to the development of more efficient and reliable approaches for long-time quantum dynamics simulations.
Hao has also worked on numerically exact tensor-train methods for multi-state harmonic models and on benchmarking approximate quantum dynamics approaches. These projects provide rigorous reference results that help assess the accuracy and limitations of more approximate computational methods. Together with his machine-learning and effective model studies, this work forms a closely integrated research program centered on the development, validation, and application of methods for quantum dynamics.
His progress has led to a strong publication record across journals including the Journal of Chemical Theory and Computation, Journal of Chemical Physics, and Communications in Computational Chemistry. He has also received the competitive N.E.T. Student Research Excellence Award in both 2025 and 2026.
Beyond individual research outputs, Hao has contributed to the development of the QCDyn software package for nonadiabatic dynamics simulations of large condensed-phase systems. This work has given him experience in collaborative scientific software development and in translating theoretical methods into tools that can be used by a broader research community.