Excited States of Crystalline Point Defects with Multireference Density Matrix Embedding Theory

Abstract

Accurate and affordable methods to characterize the electronic structure of solids are important for targeted materials design. Embedding-based methods provide an appealing balance in the trade-off between cost and accuracy -particularly when studying localized phenomena. Here, we use the density matrix embedding theory (DMET) algorithm to study the electronic excitations in solid-state defects with a

Green’s Function Formulation of Quantum Defect Embedding Theory

“…In Table , we compare our best converged values obtained with the EDC@ G 0 W 0 correction with those obtained with the HFDC correction, as well as with available experimental and theoretical data. In general, the energies predicted using EDC@ G 0 W 0 are higher than those obtained with HFDC, and in better agreement with those of quantum chemical cluster calculations . We note that the experimental zero-phonon line (ZPL) corresponding to the 3 E u level is 1.31 eV, but the contribution from the dynamical Jahn–Teller effect is unknown.…”

“…We then apply QDET with the EDC@ G 0 W 0 scheme to several spin defects in solids, and we present a strategy to systematically converge the results as a function of the composition and size of the active space. Finally, we show that, using the EDC@ G 0 W 0 , we obtain results for the electronic structure of spin defects consistent with experiments and in good agreement with results obtained with other embedding theories …”

Periodic Density Matrix Embedding for CO Adsorption on the MgO(001) Surface

“…For the test case of the CO + Mg 5 O 4 fragment at the equilibrium geometry with more than 200 orbitals, the 4c–2e calculation requires 200 Gb of memory on a AMD EPYC 7502 32-core processor, while the DF integral calculation requires 30 Gb of memory. This is because the previous implementation in refs ( 56 and 62 ) required storage of two-electron integrals, whereas the new implementation requires storage of decomposed intermediates, where N imp is the number of impurity fragment orbitals and N aux is the number of auxiliary density-fitting orbitals. For large impurity fragments, and the storage saving becomes significant.…”

“…The DMET calculations are performed with the periodic DMET (pDMET) code, , which uses the electron integrals and quantum chemical solvers from the PySCF package. , Similar to the workflow in ref , we first perform a HF calculation to obtain the mean-field wave function. Next, we define the impurity region using a set of localized orbitals in real space.…”

“…Therefore, we implemented an efficient way to store and manipulate the memory-intensive ERIs within the periodic DMET algorithm. We envision that, with our recent implementation of the multireference solvers, such as complete active space self-consistent field (CASSCF) − and n -electron valence state second-order perturbation theory (NEVPT2), − this approach can allow us to study bond-breaking phenomena of multireference systems on surfaces, at an affordable cost, which would be otherwise non-trivial for mean-field and single reference methods.…”