Speaker
Description
Computational many-body physics is governed by a persistent model–method dilemma. The finite resources available to a simulation must be divided between representing the physical system and solving its dynamics. A realistic Hamiltonian may require large basis sets, spatial inhomogeneity, and nonlocal interactions, whereas an accurate real-time propagation method must retain correlations, memory, and scattering while scaling favorably with system size and simulation time. Consequently, one typically chooses either sophisticated models combined with comparatively simple, often adiabatic dynamics, or advanced correlated solvers applied to reduced models. The regime in between—large, spatially complex, correlated systems driven far from equilibrium—has remained largely inaccessible.
Model downfolding, embedding, and low-rank compression techniques have steadily expanded the accessible range of many-body simulations, but mostly by shifting the compromise rather than removing it. To move beyond this limit, the two-particle correlations that carry memory and scattering must themselves be represented in a scalable form. δNEGF [1] provides such a representation: it reformulates nonequilibrium Green functions in terms of quantum fluctuations and replaces the explicit propagation of the rank-four two-particle correlation function by an ensemble of low-rank single-particle fluctuation trajectories. This retains dynamical self-energy effects and memory while substantially reducing runtime and memory requirements.
I will show how the resulting GW and particle-particle and particle-hole T-matrix schemes extend time-dependent dynamical NEGF simulations from approximately 10$^2$ to 10$^4$ basis states. Applications to diffusion in two-dimensional Hubbard systems and ultrafast dynamics in graphene-nanoribbon heterostructures illustrate regimes in which correlations and spatial complexity are simultaneously essential. The central implication is therefore not merely a computational speed-up: δNEGF changes the class of nonequilibrium many-body problems that can be addressed and can reveal phenomena previously hidden by computational limitations.
The talk will introduce δNEGF as a new route to scalable correlated real-time dynamics, emphasizing the computational idea behind the method, the physical regimes it opens, and future connections to embedding and first-principles modeling.
[1] E. Schroedter, M. Bonitz, and J.-P. Joost, arXiv:2606.10773