Fig. 1: Systematically improvable quantum embedding (SIE) framework and our Graphics Processing Unit (GPU)-accelerated quantum embedding for surfaces. | Nature Communications

Fig. 1: Systematically improvable quantum embedding (SIE) framework and our Graphics Processing Unit (GPU)-accelerated quantum embedding for surfaces.

From: A multi-resolution systematically improvable quantum embedding scheme for large-scale surface chemistry calculations

Fig. 1: Systematically improvable quantum embedding (SIE) framework and our Graphics Processing Unit (GPU)-accelerated quantum embedding for surfaces.

a illustrates the general workflow for a controllable multi-scale resolution of correlations at different length scales, comprising SIE with Møller-Plesset second-order perturbation theory (SIE+MP2), SIE with Coupled Cluster Singles and Doubles (SIE+CCSD), and with extra perturbative Triples (SIE+CCSD(T)). Specifically, the energy output layer of SIE+MP2 and SIE+CCSD employs the partitioned wave function density matrix method (PWF-DM), while SIE+CCSD(T) further adopts the in-situ form for perturbative (T). Further details in the Methods Section and supplemental information (SI) Section S1. b presents the computational time as a function of the total number of orbitals, measured on a single A100 GPU for systems comprising water monomer adsorbed on graphene under open boundary condition. The annotations beside the curves show the scaling obtained by fitting the last three points up to the maximum size of the test system. The dashed line corresponds to a linear fit to the last three data points in each method. Further discussion on the observed linear scaling of SIE+CCSD(T) are provided in SI Section S1.6.

Back to article page