Despite recent advances, systematic quantitative treatment of the electron correlation problem in extended systems remains a formidable task. Systematically improvable Green's function methods capable of quantitatively describing weak and at least qualitatively strong correlations appear as promising candidates for computational treatment of periodic systems. We present a periodic implementation of temperature-dependent self-consistent 2nd-order Green's function (GF2) method, where the self-energy is evaluated in the basis of atomic orbitals. Evaluating the real-space self-energy in atomic orbitals and solving the Dyson equation in k-space are the key components of a computationally feasible algorithm. We apply this technique to the one-dimensional hydrogen lattice--a prototypical crystalline system with a realistic Hamiltonian. By analyzing the behavior of the spectral functions, natural occupations, and self-energies, we claim that GF2 is able to recover metallic, band insulating, and at least qualitatively Mott regimes. We observe that the iterative nature of GF2 is essential to the emergence of the metallic and Mott phases.
Despite recent advances, systematic quantitative treatment of the electron correlation problem in extended systems remains a formidable task. Systematically improvable Green's function methods capable of quantitatively describing weak and at least qualitatively strong correlations appear as promising candidates for computational treatment of periodic systems. We present a periodic implementation of temperature-dependent self-consistent 2nd-order Green's function (GF2) method, where the self-energy is evaluated in the basis of atomic orbitals. Evaluating the real-space self-energy in atomic orbitals and solving the Dyson equation in k-space are the key components of a computationally feasible algorithm. We apply this technique to the one-dimensional hydrogen lattice--a prototypical crystalline system with a realistic Hamiltonian. By analyzing the behavior of the spectral functions, natural occupations, and self-energies, we claim that GF2 is able to recover metallic, band insulating, and at least qualitatively Mott regimes. We observe that the iterative nature of GF2 is essential to the emergence of the metallic and Mott phases.
We report an implementation of self-consistent Green's function many-body theory within a second-order approximation (GF2) for application with molecular systems. This is done by iterative solution of the Dyson equation expressed in matrix form in an atomic orbital basis, where the Green's function and self-energy are built on the imaginary frequency and imaginary time domain, respectively, and fast Fourier transform is used to efficiently transform these quantities as needed. We apply this method to several archetypical examples of strong correlation, such as a H32 finite lattice that displays a highly multireference electronic ground state even at equilibrium lattice spacing. In all cases, GF2 gives a physically meaningful description of the metal to insulator transition in these systems, without resorting to spin-symmetry breaking. Our results show that self-consistent Green's function many-body theory offers a viable route to describing strong correlations while remaining within a computationally tractable single-particle formalism.
The second-order Green's function method (GF2) was shown recently to be an accurate self-consistent approach for electronic structure of correlated systems since the self-energy accounts for both the weak and some of the strong correlation. The numerical scaling of GF2 is quite steep however, O(N5) (where the pre-factor is often hundreds), effectively preventing its application to large systems. Here, we develop a stochastic approach to GF2 (sGF2) where the self-energy is evaluated by a random-vector decomposition of Green's functions so that the dominant part of the calculation scales quasi linearly with system size. A study of hydrogen chains shows that the resulting approach is numerically efficient and accurate, as the stochastic errors are very small, 0.05% of the correlation energy for large systems with only a moderate computational effort. The method also yields automatically efficient MP2 energies and is automatically temperature dependent.
The second-order Matsubara Green's function method (GF2) is a robust temperature-dependent quantum chemistry approach, extending beyond the random-phase approximation. However, until now the scope of GF2 applications was quite limited as they require computer resources that rise steeply with system size. In each step of the self-consistent GF2 calculation there are two parts: estimating of the self-energy from the previous step's Green's function and updating the Green's function from the self-energy. The first part formally scales as the fifth power of the system size, while the second has a much gentler cubic scaling. Here, we develop a stochastic approach to GF2 (sGF2), which reduces the fifth power scaling of the first step to merely quadratic, leaving the overall sGF2 scaling as cubic. We apply the method to linear hydrogen chains with up to 1000 electrons, showing that the approach is numerically stable, efficient, and accurate. The stochastic errors are very small, on the order of 0.1% or less of the correlation energy for large systems, with only a moderate computational effort. The first iteration of GF2 is an MP2 calculation that is done in linear scaling; hence we obtain an extremely fast stochastic MP2 (sMP2) method as a byproduct. While here we consider finite systems with large band gaps where at low temperatures effects are negligible, the sGF2 formalism is temperature dependent and general and can be applied to finite or periodic systems with small gaps at finite temperatures.
Recent progress in materials synthesis combined with advances in computational science have opened up an opportunity to design new materials with specific properties. Semiconductors, heterostructures, and thin layered films are used for energy harvesting and solar cells. Transition metal oxides and metallic surfaces with deposited magnetic molecules are used for signal conversion, nonvolatile memory modules and spintronics. Due to growing complexity of the new materials the most efficient approach to their design requires combined experimental and theoretical effort. In order to advance our understanding of the complicated physics of the modern materials new theoretical methods based on controlled, reliable, computationally efficient and systematically improvable approximations describing correlation and finite-temperature effects are needed. Density functional theory (DFT) remains a method of choice for materials science calculations due to its low computational cost. However, despite all recent efforts, the majority DFT applications are limited to weakly correlated systems and zero temperature. The lack of systematic improvability, inaccurate treatment of strongly correlated systems will continue to restrict the domain of DFT applications. The multi-reference wave-function methods can provide an accurate description of the strongly correlated systems. Unfortunately, these methods are computationally expensive. In spite of recent improvement the computational complexity will likely to hinder the applications of wave-function based methods to materials science in foreseeable future. Finite-temperature Green's function methods offer several advantages such as systematic improvability, access to the excitation spectrum, thermodynamic properties and straightforward implementation of embedding frameworks. It has been realized long time ago and has been successfully used in condensed matter physics to study model systems. The applications of such methods to realistic systems are impeded due to lack of theoretical approaches providing chemical accuracy and efficient computational algorithms. In this dissertation, these shortcomings are addressed by developing self-energy embedding theory and its extensive comparison to established quantum chemistry methods. New numerical algorithms for efficient application of the finite-temperature Green's function methods to molecular and extended systems were developed. Properties of the second-order Green's function theory (GF2) in the context of quantum chemistry were investigated. Additionally, a new range-separated hybrid functional combining GF2 and DFT was developed and benchmarked.
Electronic structure of materials from first principles is essential for one-to-one comparisons between theories and experiments. However, the many-body effects resulting from electron correlation and their interplay with other degrees of freedom pose a great challenge to design an ab initio approximate method with good accuracy and efficiency at the same time. In this thesis, we present our recent developments of ab initio many-body theory for solids based on the finite-temperature Green’s function formalism. We begin by reviewing the basic concepts and definitions of realistic electronic Hamiltonians (Ch. 2) as well as Green’s functions at finite temperature (Ch. 3). Overviews of self-consistent many-body perturbation theories and quantum embedding theories are then presented in Chs. 4 and 5 as complementary materials for the following chapters. In Ch. 6, we present algorithms and implementation details for the finite-temperature self-consistent GW (scGW) method in Gaussian-type orbitals for solids. Our implementation is based on the finite-temperature Green’s function formalism in which all equations are solved on the imaginary axis, without resorting to analytical continuation during the self-consistency. No quasiparticle approximation is employed and all matrix elements of self-energies are explicitly evaluated. We show agreement with other, differently formulated finite-temperature scGW implementations. By migrating computationally intensive calculations to GPUs, we obtain scalable results on large supercomputers with nearly optimal performance. Inclusion of relativistic effects in scGW is discussed in Ch. 7. Our formulation of relativistic scGW is based on the exact two-component formalism with one-electron approximation (X2C1e) and non-relativistic Coulomb interactions. Our theory allows us to study scalar relativistic effects, spin-orbit coupling, and their interplay with electron correlation without adjustable parameters. We examine the effect of the X2C1e by comparison to the established four-component formalism and reach excellent agreement. The simplicity of X2C1e enables construction of higher order theories, such as embedding theories, on top of perturbative calculations. In Ch. 8, we use self-energy embedding theory (SEET) to study the spectra of the prototypical undistorted cubic perovskites SrVO3 and SrMnO3. In the strongly correlated metallic SrVO3, we find that the usual attribution of the satellite peaks at -1.8 eV to Hund or Hubbard physics in the t2g orbitals is inconsistent with our calculations. In the strongly correlated insulator SrMnO3 we recover insulating behavior due to a feedback effect between the strongly correlated orbitals and the weakly correlated environment. Our calculation shows a systematic convergence of spectral features as the space of strongly correlated orbitals is enlarged, paving the way to a systematic parameter-free study of correlated perovskites. Lastly, in Ch. 9, we apply the Green’s function coupled cluster singles and doubles (GFCCSD) impurity solver to realistic impurity problems in strongly correlated solids within the framework of SEET. We describe the implementation details of our Green’s function coupled cluster (GFCC) solver, investigate its performance, and highlight potential advantages and problems on examples of impurities created for antiferromagnetic MnO and paramagnetic SrMnO3. GFCCSD provides satisfactory descriptions for weakly and moderately correlated impurities with sizes that are intractable by existing accurate impurity solvers such as exact diagonalization. However, when correlations become strong, the singles and doubles approximation used in GFCC could lead to instabilities in searching for the particle number present in impurity problems.
We present an implementation of a fully self-consistent finite temperature second order Green's function perturbation theory (GF2) within the diagrammatic Monte Carlo framework. In contrast to the previous implementations of stochastic GF2 ({\it J. Chem. Phys.},{\bf 151}, 044144 (2019)), the current self-consistent stochastic GF2 does not introduce a systematic bias of the resulting electronic energies. Instead, the introduced implementation accounts for the stochastic errors appearing during the solution of the Dyson equation. We present an extensive discussion of the error handling necessary in a self-consistent procedure resulting in dressed Green's function lines. We test our method on a series of simple molecular examples.