Publications

The study of isolated defects in solids is a natural target for classical or quantum embedding methods that treat the defect at a high level of theory and the rest of the solid at a lower level of theory. Here, in the context of active-space-based quantum embeddings, we study the performance of three active-space orbital selection schemes based on canonical (energy-ordered) orbitals, local orbitals defined in the spirit of density matrix embedding theory, and approximate natural transition orbitals. Using equation-of-motion coupled-cluster theory with single and double excitations (CCSD), we apply these active space selection schemes to the calculation of the vertical singlet excitation energy of a substitutional carbon dimer defect in hexagonal boron nitride, an oxygen vacancy in magnesium oxide, and a carbon vacancy in diamond. Especially when used in combination with a simple composite correction, we find that the best performing schemes can predict the excitation energy to about 0.1-0.2 eV of its converged value using only a few hundred orbitals, even when the full supercell has thousands of orbitals, which amounts to many-orders-of-magnitude computational savings when using correlated electronic structure theories. When compared to assigned experimental spectra and accounting for vibrational corrections, we find that CCSD predicts excitation energies that are accurate to about 0.1-0.3 eV, which is comparable to its performance in molecules and bulk solids.

Understanding the properties of well-generalizing minima is at the heart of deep learning research. On the one hand, the generalization of neural networks has been connected to the decision boundary complexity, which is hard to study in the high-dimensional input space. Conversely, the flatness of a minimum has become a controversial proxy for generalization. In this work, we provide the missing link between the two approaches and show that the Hessian top eigenvectors characterize the decision boundary learned by the neural network. Notably, the number of outliers in the Hessian spectrum is proportional to the complexity of the decision boundary. Based on this finding, we provide a new and straightforward approach to studying the complexity of a high-dimensional decision boundary; show that this connection naturally inspires a new generalization measure; and finally, we develop a novel margin estimation technique which, in combination with the generalization measure, precisely identifies minima with simple wide-margin boundaries. Overall, this analysis establishes the connection between the Hessian and the decision boundary and provides a new method to identify minima with simple wide-margin decision boundaries.

We present the mean field solution of the quantum and classical Heisenberg spin glasses, using the combination of a high precision numerical solution of the Parisi full replica symmetry breaking equations and a continuous time Quantum Monte Carlo. We characterize the spin glass order and its low-energy excitations down to zero temperature. The Heisenberg spin glass has a rougher energy landscape than its Ising analogue, and exhibits a very slow temperature evolution of its dynamical properties. We extend our analysis to the doped, metallic Heisenberg spin glass, which displays an unexpectedly slow spin dynamics reflecting the proximity to the melting quantum critical point and its associated Sachdev-Ye-Kitaev Planckian dynamics.

Multipoint vertex functions, and the four-point vertex in particular, are crucial ingredients in many-body theory. Recent years have seen significant algorithmic progress toward numerically computing their dependence on multiple frequency arguments. However, such computations remain challenging and are prone to suffer from numerical artifacts, especially in the real-frequency domain. Here, we derive estimators for multipoint vertices that are numerically more robust than those previously available. We show that the two central steps for extracting vertices from correlators, namely the subtraction of disconnected contributions and the amputation of external legs, can be achieved accurately through repeated application of equations of motion, in a manner that is symmetric with respect to all frequency arguments and involves only fully renormalized objects. The symmetric estimators express the core part of the vertex and all asymptotic contributions through separate expressions that can be computed independently, without subtracting the large-frequency limits of various terms with different asymptotic behaviors. Our strategy is general and applies equally to the Matsubara formalism, the real-frequency zero-temperature formalism, and the Keldysh formalism. We demonstrate the advantages of the symmetric improved estimators by computing the Keldysh four-point vertex of the single-impurity Anderson model using the numerical renormalization group.

We investigate the possible superconducting instabilities of strongly correlated electron materials using a generalization of linear response theory to external pairing fields depending on frequency. We compute a pairing susceptibility depending on two times, allowing us to capture dynamical pairing and in particular odd-frequency solutions. We first benchmark this method on the attractive one-band Hubbard model and then consider the superconductivity of strontium ruthenate Sr

The intersection of electronic topology and strong correlations offers a rich platform to discover exotic quantum phases of matter and unusual materials. An overarching challenge that impedes the discovery is how to diagnose topology in strongly correlated settings, as exemplified by Mott insulators. Here, we develop a general framework to address this outstanding question and illustrate its power in the case of Mott insulators. The concept of Green's function Berry curvature -- which is frequency dependent -- is introduced. We apply this notion in a system that contains symmetry-protected nodes in its noninteracting bandstructure; strong correlations drive the system into a Mott insulating state, creating contours in frequency-momentum space where the Green's function vanishes. The Green's function Berry flux of such zeros is found to be quantized, and is as such direct probe of the system's topology. Our framework allows for a comprehensive search of strongly correlated topological materials with Green's function topology.

The Matsubara Green's function formalism stands as a powerful technique for computing the thermodynamic characteristics of interacting quantum many-particle systems at finite temperatures. In this manuscript, our focus centers on introducing MatsubaraFunctions.jl, a Julia library that implements data structures for generalized n-point Green's functions on Matsubara frequency grids. The package's architecture prioritizes user-friendliness without compromising the development of efficient solvers for quantum field theories in equilibrium. Following a comprehensive introduction of the fundamental types, we delve into a thorough examination of key facets of the interface. This encompasses avenues for accessing Green's functions, techniques for extrapolation and interpolation, as well as the incorporation of symmetries and a variety of parallelization strategies. Examples of increasing complexity serve to demonstrate the practical utility of the library, supplemented by discussions on strategies for sidestepping impediments to optimal performance.

A major challenge in the field of correlated electrons is the computation of dynamical correlation functions. For comparisons with experiment, one is interested in their real-frequency dependence. This is difficult to compute, as imaginary-frequency data from the Matsubara formalism require analytic continuation, a numerically ill-posed problem. Here, we apply quantum field theory to the single-impurity Anderson model (AM), using the Keldysh instead of the Matsubara formalism with direct access to the self-energy and dynamical susceptibilities on the real-frequency axis. We present results from the functional renormalization group (fRG) at one-loop level and from solving the self-consistent parquet equations in the parquet approximation. In contrast to previous Keldysh fRG works, we employ a parametrization of the four-point vertex which captures its full dependence on three real-frequency arguments. We compare our results to benchmark data obtained with the numerical renormalization group and to second-order perturbation theory. We find that capturing the full frequency dependence of the four-point vertex significantly improves the fRG results compared to previous implementations, and that solving the parquet equations yields the best agreement with the NRG benchmark data, but is only feasible up to moderate interaction strengths. Our methodical advances pave the way for treating more complicated models in the future.

Kohn's theorem places strong constraints on the cyclotron response of Fermi liquids. Recent observations of a doping dependence in the cyclotron mass of La

We present a method of calculating the energy gap of a charge-neutral excitation using only ground-state calculations. We report Quantum Monte Carlo calculations of Γ→Γ and Γ→X particle-hole excitation energies in diamond carbon. We analyze the finite-size effect and find the same 1/L decay rate as that in a charged excitation, where L is the linear extension of the supercell. This slow decay is attributed to the delocalized nature of the excitation in supercells too small to accommodate excitonic binding effects. At larger system sizes, the apparent 1/L decay crosses over to a 1/L