betaFeedback

WorksPhysics and Astronomy

ComDMFT v.2.0: Fully Self-Consistent ab initio GW+EDMFT for the Electronic Structure of Correlated Quantum Materials

Abstract

ComDMFT is a parallel computational package designed to study the electronic structure of correlated quantum materials from first principles. Our approach is based on the combination of first-principles methods and dynamical mean field theories. In version 2.0, we implemented fully-diagrammatic GW+EDMFT from first-principles. In this approach, correlated electrons are treated within full GW+EDMFT and the rest are treated within full-GW, seamlessly. This implementation enables the electronic structure calculation of quantum materials with weak, intermediate, and strong electron correlation without prior knowledge of the degree of electron correlation.

PROGRAM SUMMARY/NEW VERSION PROGRAM SUMMARY

Program Title: ComDMFT

Licensing provisions(please choose one): GPLv3

Programming language: fortran2008, C++, and Python3

Nature of problem:

In order to comprehend the electronic structure of correlated quantum materials, a non-perturbative and parameter-free ab initio method is required. One of the promising approaches is fully-diagrammatic GW+EDMFT; however, there is no available open-source package which implements the ab initio full GW+EDMFT methodology.

Solution method:

We implemented ab initio all electron full GW+EDMFT methodology for the electronic structure calculation of correlated quantum materials and compare the results with those of LQSGW+DMFT.

Additional comments:

ComDMFTv2 is built on top open source codes, ComDMFTv1 [1], Wannier90 [2], ComCTQMC[3] and LqsgwFlapw [4] codes. ComDMFTv2 underwent testing on Intel supercomputers with intel compilers and future revisions will ensure compatibility with other compilers.

Introduction

In materials with a large number of electrons, the quantum mechanical interactions between these electrons often dictate the materials properties. These so-called correlated quantum materials can display a range of exceptional emergent phenomena, including metal-insulator transition, colossal magnetoresistance, and high-temperature superconductivity. Recently, correlated quantum materials received renewed attentions due to their potential for quantum information science applications. An excellent example is the Fe-based superconductors, which have demonstrated potential as qubit hosts which are topologically-protected. [5, 6, 7, 8].

One of the grand challenges in condensed matter physics as well as materials science is to predict correlated quantum materials properties from first principles. Treating correlated quantum materials require new concepts, algorithms, as well as computer codes. To this end, various ideas have been realized and applied for the quantitative understanding of correlated quantum materials properties [9]. Among them, one of the most successful methodologies is dynamical mean-field theory (DMFT) [10, 11, 12] [13, 14, 15, 16, 17, 18, 19]. DMFT establishes a mapping from a quantum many-body problem onto a quantum impurity problem. Quantities obtained within quantum impurity problems are embedded into the Green’s function of the quantum many-body problems. This idea has been applied extensively to model Hamiltonians and has described correlated electron phenomena such as Mott physics, [16, 20] Hund metal physics [21] and heavy fermion physics [22].

The success of model-based DMFT approaches has led to the integration of DMFT with ab initio methodologies. This combination is based on the concept of quantum embedding, which entails separating electrons into two groups [23, 24, 25]: correlated electrons and itinerant electrons. Correlated electron phenomena are addressed by DMFT, while first-principles methods are used to describe itinerant electrons and to generate materials-specific quantum impurity problems. An example of this successful union is the DFT+DMFT methodology [26, 27, 28]. This approach can explain various phenomena specific to correlated quantum materials such as paramagnetic Mott transitions[29] and volume collapse transitions[30]. The development of the continuous-time Monte-Carlo approaches to quantum impurity problems [31] has enabled the analysis of broad classes of quantum impurity problems, making them the popular choice for DMFT-based ab initio calculations. Recent important advances in the DFT+DMFT methodology were introduced to compute total energies[32], free energies[33], and forces[34], as well as more accurate double-counting scheme corrections[35].

The DFT+DMFT methodology has attracted a lot of attention and is supported by the development of various software programs. Reusable libraries, such as TRIQS[36] and ALPS[37], have been created to aid DFT+DMFT implementation. Standalone impurity solvers, iQIST[38], W2dynamics[39] and ComCTQMC[3], are available. DFT+DMFT packages have been developed: ComDMFT[1]; EDMFTF[40] integrated with Wien2K[41]; TRIQS/DFTTools [42] built on top of TRIQS[36] and Wien2k[41]; DMFTwDFT[32] [43] bulit on top of EDMFTF package impurity solver[44]; D-Core [45] built on top of TRIQS[36] [42], ALPS[37], Quantum Espresso[46], and OpenMX[47]; AMULET[48] integrated with Quantum Espresso[46] and Elk[49]; and Wien2k+W2dynamics. Furthermore, DFT+DMFT functionality has been added to ab initio codes such as Abinit[50] and RSPt[51].

Similarly, the rotationally invariant slave bosons (RISB), which is equivalent to the Gutzwiller method, were combined with density functional theory in the LDA+RISB or LDA+G method [52, 53, 54, 55]. They were also cast as a quantum embedding method [56] [57] and open source implementations of this approach such as CyGutz and ComGutz are available [58, 59]. More recently, direct comparisons of LDA+DMFT and LDA+G, including total energies, have been carried out using the Portobello framework [60]. These approaches can also be combined with GW methods for the weakly correlated subsystems.

Density Functional Theory combined with Dynamical Mean Field Theory (DFT+DMFT) has been very successful in predicting the properties of static equilibrium properties of correlated quantum materials. Moreover, it has been proven that the predicted dynamics of electronic excitations agrees well with experimental data from many spectroscopic tools. It requires, however, the introduction of parameters in constructing quantum impurity problem. For examples, Coulomb interactions of the correlated subshell are required. For DFT+DMFT, the Coulomb interaction can be obtained by using cRPA [61, 62] as well as constrained DFT [63].

The success of DFT+DMFT has motivate the search for its justification and extension via diagrammatic Green’s function approaches based on free energy functionals [64, 65, 66]. By defining exchange-correlation functionals using Green’s function (GG) and the screened Coulomb interaction (WW)[67, 68], these approaches provide a systematic way to calculate the properties of materials properties. Diagrammatic approaches have proven successful in first-principles implementations. One good example is the GW approximation for studying itinerant electron systems. This method employs the first-order diagram in WW to represent the electron-boson coupling in the free-energy functional [67], which can be interpreted as an approximation of the bare three-point vertex[69]. The use of this ab initio method [70, 71, 72, 73, 74] has significantly improved the accuracy of quasiparticle spectra of semiconductors and insulators in comparison to DFT. The development of software programs such as ABINIT[50], BerkeleyGW[75], CP2K[76], Elk[49], FHI-aims[77], Fiesta[78], LqsgwFlapw[4], GPAW[79], PySCF[80], Questaal[81], VASP[82], West[83], and Yambo[84] has facilitated the usage of ab initio GW methodology.

For the correlated quantum materials, a combination of many-body perturbation theory (MBPT) and DMFT can be used through this functional approach. The simplest example of this functional approach is the GW+EDMFT approach [85, 86, 87]; by accounting for exchange and correlation effects within the GW approximation for itinerant electrons and extended DMFT[88, 89, 90] for correlated electrons, this approach merges these two theories. When GW methodology is merged with extended DMFT, this approach is advantageous as it is free of adjustable parameters; once a projector is chosen to describe the correlated orbitals, the method proceeds without any further input from the user and defines the quantum impurity problem for the correlated orbitals self-consistently. Similar approaches with the same goal include ab initio Density Matrix Embedding Theory (DMET) [91], full-cell DMFT [92], and Self-Energy Embedding Theory (SEET) [93]. Originally, full-cell DMFT and SEET were formulated using the Green’s function (GG) and the bare Coulomb interaction (VV), but they have recently been extended to incorporate the screened Coulomb interaction (WW).

Various earlier schemes of GW+EDMFT have been proposed. Two examples are the partial self-consistent scheme [94] and the multitier GW+EDMFT approach[87]. These combine GW and DMFT, but not in a fully self consistent manner. The former integrates DMFT with quasiparticle GW approaches such as one-shot GW[95, 96, 97], QSGW[98, 81], LQSGW[99] [1], and screened exchange[100]. Coulomb interactions of the correlated subshells are obtained within constrained random phase approximation[61]. In these approaches, DMFT is treated as a one-shot correction to the corresponding quasiparticle Hamiltonian. On the other hand, the multitier GW+EDMFT entails self-consistency within a subspace of the full Hilbert space, while the remaining Hilbert space is treated with a one-shot GW. These approaches have been applied to various correlated quantum materials [101, 102, 103, 104], exhibiting the capability of full implementation. Within the field of correlated quantum materials, there is a pressing requirement for a package that implements the ab initio full GW+EDMFT methodology. Currently, few such packages exist and none of them supports fully self-consistent GW+DMFT scheme. ComDMFT[1] supports the ab initio LQSGW+DMFT methodology [99] and Questaal[81] implements the QSGW+DMFT methodology.

This paper describes the ComDMFT v2.0 package, developed at Brookhaven National Lab. It is the first implementation of full GW+EDMFT. As an open source code for the community, it builds on earlier open source packages ComDMFT v1.0 [1], the LqsgwFlapw code[4] for ab initio GW calculations, the Wannier90 code[2] for the construction of correlated orbitals, and the Com-CTQMC code[3] for the solution of quantum impurity problems. The package is composed of five modules, written in FORTRAN2008 and Python3, namely: ComBasis, Comlocal, ComWeiss, ComDC, and ComEmbed. In this paper, we introduce full GW+EDMFT methodology and to provide benchmark results on SrVO3_3 and NiO. § 3 outlines the general theoretical framework, while § 4 provides a detailed description of our implementation. Benchmark results are presented in § 5 together with calculations with LQSGW+DMFT for the same materials for comparison.

Theoretical Background

A very useful approach in many body theory is provided by the functional formalism of Luttinger and Ward [65] and Baym and Kadanoff [64]. Here, the grand potential (Ω\Omega) is viewed as a functional of the electron Green’s function (GG) and electron self-energy (Σ\Sigma). Ω[G,Σ]\Omega[G,\Sigma] is stationary at the physical GG and the corresponding Σ\Sigma. Approximations to this exact functional result in approximate many body methods. This approach has been later extended to construct a new functional, Γ\Gamma: [67, 68]

Γ[G,Σ,W,Π]=−Tr⁡[log⁡(−GH−1)]−Tr⁡[log⁡(1−GHΣ)]−Tr⁡[GΣ]+12Tr⁡[log⁡(1−VΠ)]+12Tr⁡[ΠW]+Ψ[G,W].(1)\Gamma[G,\Sigma,W,\Pi] = -\operatorname{Tr}[\log(-G_H^{-1})] - \operatorname{Tr}[\log(1-G_H\Sigma)] - \operatorname{Tr}[G\Sigma] + \frac{1}{2}\operatorname{Tr}[\log(1-V\Pi)] + \frac{1}{2}\operatorname{Tr}[\Pi W] + \Psi[G,W]. \tag*{(1)}

Here, Tr is used to denote the sum over all relevant indices (crystal momentum, state indices, and spin indices) in addition to Matsubara frequency summation, with a prefactor of 1/β1/\beta. GHG_H and VV denote the Hartree Green’s function and bare Coulomb interaction, respectively. Ψ\Psi is a functional expressed as a summation of all two-particle irreducible diagrams constructed from GG and WW. Approximations to Ψ\Psi result in approximate many body methods. Within this construction, Γ\Gamma is regarded as the functional of Green’s function (GG), electron self-energy (Σ\Sigma), screened Coulomb interaction(WW), and polarizability (Π\Pi). These can be shown to be stationary at the physical GG and WW, along with the corresponding Σ\Sigma and Π\Pi. These conditions lead to the following equations:

δΓδG=0→Σ=δΨδG(2)\frac{\delta\Gamma}{\delta G}=0 \to\Sigma=\frac{\delta\Psi}{\delta G} \tag*{(2)}
δΓδΣ=0→G=(GH−1−Σ)−1(3)\frac{\delta\Gamma}{\delta\Sigma}=0 \to G=(G_H^{-1}-\Sigma)^{-1} \tag*{(3)}
δΓδW=0→Π=−2δΨδW(4)\frac{\delta\Gamma}{\delta W}=0 \to\Pi=-2\frac{\delta\Psi}{\delta W} \tag*{(4)}
δΓδΠ=0→W=(V−1−Π)−1(5)\frac{\delta\Gamma}{\delta\Pi}=0 \to W=(V^{-1}-\Pi)^{-1} \tag*{(5)}

This approach can be utilized to comprehend various diagrammatic approximations. For instance, the first-order diagram in WW (ΨGW(G,W)=−12Tr⁡GWG\Psi_{GW}(G,W)=-\frac{1}{2}\operatorname{Tr}GWG) corresponds to the renowned GW approximation[69, 67]. ΨGW(G,W)\Psi_{GW}(G,W) leads to the following self-energy and polarizability.

ΣGW(1,2)=δΨGWδG(2,1)=−G(1,2)W(1,2)\Sigma_{GW}(1,2)=\frac{\delta\Psi_{GW}}{\delta G(2,1)}=-G(1,2)W(1,2)
ΠGW(1,2)=−2δΨGWδW(2,1)=G(1,2)G(2,1)(6)\Pi_{GW}(1,2)=-2\frac{\delta\Psi_{GW}}{\delta W(2,1)}=G(1,2)G(2,1) \tag*{(6)}

Here, the notation 11 denotes (r1,σ1,τ1)=(x1,τ1)=x1(r_1,\sigma_1,\tau_1)=(x_1,\tau_1)=x_1, where rr, σ\sigma and τ\tau are the position vector, spin index, and imaginary time, respectively.

In order to incorporate strong correlation effects, the DMFT idea can be applied. This amounts to constraining the GG and WW in the Ψ\Psi functional to the correlated orbital space, and is referred to as extended DMFT (EDMFT)[88, 89, 90]. Rigorously, the correlated orbitals φαf(r)\varphi_\alpha^f(r) are introduced, where α\alpha is a composite index for the correlated orbital and spin, and ff denotes “fermion”. The projection (P^f\hat{P}^f) and embedding (E^f\hat{E}^f) of the fermionic quantities can then be defined accordingly.

G~αβ(τ1,τ2)=P^αβf(G)(τ1,τ2)=∫dx1dx2φαf∗(x1)G(x1,τ1,x2,τ2)φβf(x2)\widetilde{G}_{\alpha\beta}(\tau_1,\tau_2)=\hat{P}^f_{\alpha\beta}(G)(\tau_1,\tau_2)=\int dx_1dx_2\varphi_\alpha^{f*}(x_1)G(x_1,\tau_1,x_2,\tau_2)\varphi_\beta^f(x_2)
G~(1,2)=E^f(G~)(1,2)=∑αβφαf(x1)G~αβ(τ1,τ2)φβf∗(x2)(7)\widetilde{G}(1,2)=\hat{E}^f(\widetilde{G})(1,2)=\sum_{\alpha\beta}\varphi_\alpha^f(x_1)\widetilde{G}_{\alpha\beta}(\tau_1,\tau_2)\varphi_\beta^{f*}(x_2) \tag*{(7)}

Here,  ~\widetilde{\ } denotes a physical quantity projected onto the correlated orbital space and ∫dx1=∑σ1∫dr1\int dx_1=\sum_{\sigma_1}\int dr_1.

For the projection (P^b\hat{P}^b) and embedding (E^b\hat{E}^b) of bosonic quantities, it is convenient to define orthonormalized product basis functions. To develop these functions, we initially construct unorthonormalized product basis functions, ϑ\vartheta.

ϑαβ(x)=φαf∗(x)φβf(x)(8)\vartheta_{\alpha\beta}(x)=\varphi_\alpha^{f*}(x)\varphi_\beta^f(x) \tag*{(8)}

The orthonormalized product basis function, φαb\varphi_\alpha^b can be represented as a linear combination of ϑ\vartheta:

φαb(x)=∑βγϑβγ(x)Xβγ,α(9)\varphi_\alpha^b(x)=\sum_{\beta\gamma}\vartheta_{\beta\gamma}(x)X_{\beta\gamma,\alpha} \tag*{(9)}

For the construction of φαb\varphi_\alpha^b, we employed canonical orthonormalization. The overlap matrix (Sαβ;γδ=∫dxϑαβb∗(x)ϑγδ(x))(S_{\alpha\beta;\gamma\delta}=\int dx\vartheta_{\alpha\beta}^{b*}(x)\vartheta_{\gamma\delta}(x)) was diagonalized by a unitary matrix UU.

U†SU=D(10)U^\dagger S U=D \tag*{(10)}

where DD is the diagonal matrix of the eigenvalues of SS. Then XX is given by

Xβγ,α=Uβγ;αDα−1/2(11)X_{\beta\gamma,\alpha}=U_{\beta\gamma;\alpha}D_{\alpha}^{-1/2} \tag*{(11)}

By using the orthonormalized product basis, we can define the projection and embedding of the bosonic quantities in the following way.

W~αβ(τ1,τ2)=P^αβb(W)(τ1,τ2)=∫dx1dx2ϕαb∗(x1)W(x1,τ1,x2,τ2)ϕβb(x2)W~(1,2)=E^b(W~)(1,2)=∑αβϕαb(x1)W~αβ(τ1,τ2)ϕβb∗(x2)(12)\begin{aligned} \widetilde{W}_{\alpha\beta}(\tau_{1},\tau_{2})&=\hat{P}^{b}_{\alpha\beta}(W)(\tau_{1},\tau_{2})=\int dx_{1}dx_{2}\phi_{\alpha}^{b*}(x_{1})W(x_{1},\tau_{1},x_{2},\tau_{2})\phi_{\beta}^{b}(x_{2})\\ \widetilde{W}(1,2)&=\hat{E}^{b}(\widetilde{W})(1,2)=\sum_{\alpha\beta}\phi_{\alpha}^{b}(x_{1})\widetilde{W}_{\alpha\beta}(\tau_{1},\tau_{2})\phi_{\beta}^{b*}(x_{2}) \tag*{(12)} \end{aligned}

Alternatively, the projected bosonic quantities can be defined by using unorthonormalized product basis.

W~αβγδ=P^αβγδb(W)(τ1,τ2)=∫dx1dx2ϑαδ∗(x1)W(x1,τ1,x2τ2)ϑγβ(x2)(13)\widetilde{W}_{\alpha\beta\gamma\delta}=\hat{P}^{b}_{\alpha\beta\gamma\delta}(W)(\tau_{1},\tau_{2})=\int dx_{1}dx_{2}\vartheta_{\alpha\delta}^{*}(x_{1})W(x_{1},\tau_{1},x_{2}\tau_{2})\vartheta_{\gamma\beta}(x_{2}) \tag*{(13)}

The EDMFT Ψ\Psi functional (ΨEDMFT(G~,W~)\Psi_{EDMFT}(\widetilde{G},\widetilde{W})) is given by all two-particle irreducible diagrams constructed from G~\widetilde{G} and W~\widetilde{W}. The electron self-energy and polarizability within EDMFT is given by the following equations.

ΣEDMFT(1,2)=δΨEDMFTδG(2,1)=δΨEDMFTδG~(2,1)ΠEDMFT(1,2)=−2δΨEDMFTδW(2,1)=−2δΨEDMFTδW~(2,1)(14)\begin{aligned} \Sigma_{EDMFT}(1,2)&=\frac{\delta\Psi_{EDMFT}}{\delta G(2,1)}=\frac{\delta\Psi_{EDMFT}}{\delta\widetilde{G}(2,1)}\\ \Pi_{EDMFT}(1,2)&=-2\frac{\delta\Psi_{EDMFT}}{\delta W(2,1)}=-2\frac{\delta\Psi_{EDMFT}}{\delta\widetilde{W}(2,1)} \tag*{(14)} \end{aligned}

The functional approach provides a natural way to merge these two diagrammatic approaches [85, 86, 87]. When we merge ΨGW(G,W)\Psi_{GW}(G,W) and ΨEDMFT(G~,W~)\Psi_{EDMFT}(\widetilde{G},\widetilde{W}), the double-counted diagrams should be subtracted. In GW+EDMFT, the GW Ψ\Psi functional in the correlated space (ΨDC(G~,W~)=−12Tr⁡G~W~G~\Psi_{DC}(\widetilde{G},\widetilde{W})=-\frac{1}{2}\operatorname{Tr}\widetilde{G}\widetilde{W}\widetilde{G}) provides the double counted diagrams. In particular, the double-counting self-energy and polarizability are given by

ΣDC(1,2)=δΨDCδG(2,1)=δΨDCδG~(2,1)=−G~(1,2)W~(1,2)ΠDC(1,2)=−2δΨDCδW(2,1)=−2δΨDCδW~(2,1)=G~(1,2)G~(2,1)(15)\begin{aligned} \Sigma_{DC}(1,2) &= \frac{\delta\Psi_{DC}}{\delta G(2,1)} = \frac{\delta\Psi_{DC}}{\delta\widetilde{G}(2,1)} = -\widetilde{G}(1,2)\widetilde{W}(1,2) \\ \Pi_{DC}(1,2) &= -2\frac{\delta\Psi_{DC}}{\delta W(2,1)} = -2\frac{\delta\Psi_{DC}}{\delta\widetilde{W}(2,1)} = \widetilde{G}(1,2)\widetilde{G}(2,1) \tag*{(15)} \end{aligned}

Finally, we can derive the Ψ\Psi functional of GW+EDMFT (ΨGW+EDMFT(G,W)\Psi_{GW+EDMFT}(G,W)) illustrated in Figure 1.

Figure 1. Diagrammatic representations of the Ψ\Psi functional within full GW+EDMFT approach. (a) GW Ψ\Psi functional (b) EDMFT Ψ\Psi functional and (c) local-GW Ψ\Psi functional. Full and wiggly lines represent Green’s function (GG) and screened Coulomb interaction (WW), respectively. Black and gray colors represent quantities in the full Hilbert space and in the correlated subspace, respectively.

ΨGW+EDMFT(G,W)=−12Tr⁡GWG+ΨEDMFT(G~,W~)+12Tr⁡G~W~G~.(16)\Psi_{GW+EDMFT}(G,W)=-\frac{1}{2}\operatorname{Tr}GWG+\Psi_{EDMFT}(\widetilde{G},\widetilde{W})+\frac{1}{2}\operatorname{Tr}\widetilde{G}\widetilde{W}\widetilde{G}. \tag*{(16)}

Then, the electron self-energy within GW+EDMFT is given by the following equations. For the correlated states,

Σ(1,2)=−G(1,2)W(1,2)+δΨEDMFTδG~(2,1)+G~(1,2)W~(1,2).(17)\Sigma(1,2)=-G(1,2)W(1,2)+\frac{\delta\Psi_{EDMFT}}{\delta\widetilde{G}(2,1)}+\widetilde{G}(1,2)\widetilde{W}(1,2). \tag*{(17)}

For the itinerant states,

ΣGW(1,2)=−G(1,2)W(1,2).(18)\Sigma_{GW}(1,2)=-G(1,2)W(1,2). \tag*{(18)}

The polarizability within GW+EDMFT is given by the following equations. For the correlated states,

Π(1,2)=G(1,2)G(2,1)−2δΨEDMFTδW~(2,1)−G~(1,2)G~(2,1).(19)\Pi(1,2)=G(1,2)G(2,1)-2\frac{\delta\Psi_{EDMFT}}{\delta\widetilde{W}(2,1)}-\widetilde{G}(1,2)\widetilde{G}(2,1). \tag*{(19)}

For the itinerant states,

Π(1,2)=G(1,2)G(2,1).(20)\Pi(1,2)=G(1,2)G(2,1). \tag*{(20)}

The GW+EDMFT approximation provides a set of self-consistent equations, which can be solved to obtain GG, Σ\Sigma, WW, and Π\Pi at their stationary conditions. Central equations are the fermionic and bosonic Dyson’s equations, as shown in eqs. (3) and (5), respectively. In addition, three distinct self-consistency loops must be considered: the GW loop, EDMFT loop, and double counting loop. A summary of these equations and loops is provided in Figure 2. Here we note that Dyson’s equations for Green’s function can be written in terms of GGWG_{GW}

Figure 2. The GW+EDMFT approach involves three self-consistency loops: the GW loop (blue), the EDMFT loop (red), and the double counting loop (green). This diagram seeks to emphasize the need for self-consistent determination of three-different terms in both fermionic and bosonic self-energies. For more information on the practical implementation of the self-consistent loop, please refer to Figure 4.

GGW=(GH−1−ΣGW)−1(21)G_{GW}=\left(G_H^{-1}-\Sigma_{GW}\right)^{-1} \tag*{(21)}
G={GGW−1−E^f(Σ~EDMFT−Σ~DC)}−1.(22)G = \left\{G_{GW}^{-1} - \hat{E}^{f}\left(\widetilde{\Sigma}_{EDMFT} - \widetilde{\Sigma}_{DC}\right)\right\}^{-1}. \tag*{(22)}

In the same way, Dyson’s equation for screened Coulomb interaction can be written in terms of WGWW_{GW}

WGW=(V−1−ΠGW)−1(23)W_{GW} = \left(V^{-1} - \Pi_{GW}\right)^{-1} \tag*{(23)}
W={WGW−1−E^b(Π~EDMFT−Π~DC)}−1(24)W = \left\{W_{GW}^{-1} - \hat{E}^{b}\left(\widetilde{\Pi}_{EDMFT} - \widetilde{\Pi}_{DC}\right)\right\}^{-1} \tag*{(24)}

Ab initio fully self-consistent GW+EDMFT implementation in ComDMFT v2.0

ComDMFT v2.0 comprises eight components and a Python script: LqsgwFlapw[4], Wannier90[2], ComBasis, ComWeiss, ComLocal, ComCTQMC[3], ComDC, and ComEmbed. These components are executed in sequence by the Python script “gwdmft.py”.

All the other components are developed using object-oriented programming features in python3 and fortran 2008. The implemented resources (classes) are reused by several components and can easily be reused by future developers. All the generated data by individual components are saved in a single hdf5 file and can be accessed by other components based on a data-map upon inquiry with keys (file name, group name, data name).

For the GW+EDMFT implementation, we will consider three Hilbert spaces: the full space (F), the correlated space (C), and the low-energy space (L). The full space is the entire Hilbert space. To label the basis functions in the full space, we will use an upper-case Latin alphabet ∣φIf/b⟩\lvert\varphi_{I}^{f/b}\rangle or a position vector together with a spin index ∣x⟩=∣rσ⟩\lvert x\rangle= \lvert r\sigma\rangle. Here ff and bb in the superscript denote “Fermion” and “Boson” respectively. The correlated space is a subspace of the full space. They are spanned by correlated orbitals for fermionic quantities and orthonormalized product basis sets of the correlated orbitals for bosonic quantities. We will use a lower-case Greek alphabet to label the basis functions in the correlated space, e.g. ∣φαf/b⟩\lvert\varphi_{\alpha}^{f/b}\rangle. The low-energy space is a subspace of the full space, and a parent vector space of the correlated space. We construct this Hilbert space (L) for the crystal momentum (k) interpolation of GGWG^{GW} and WGWW^{GW}. These are two quantities obtained from ab initio GW calculations. Due to the heavy numerical cost, it is not feasible to acquire them at an arbitrary crystal momentum resolution. Therefore, we take advantage of Wannier interpolation to interpolate these two quantities under the constraint of a restricted Hilbert space. Maximally-localized Wannier functions are used as the localized basis sets for fermionic quantities. Typically, they are constructed to span in the low-energy window (typically chosen in EF±10eVE_F \pm10eV as a frozen energy window). The basis functions for the bosonic quantities are constructed by calculating orthonormalized product basis functions of the low-energy fermionic basis functions. We will use lower-case Latin alphabet to label the basis functions in the low-energy space, e.g. ∣φf/b⟩|\varphi^{f/b}\rangle. After the basis function construction, we project GGWG^{GW} and WGWW^{GW} into the low-energy space and then interpolate them in the momentum space.

Figure 3 displays the three distinct Hilbert spaces (C, L, and F) and their projection onto NiO DFT eigenstates. For NiO, the correlated orbitals consist of Ni-d orbitals. The low-energy itinerant orbitals, found in the low-energy subspace but not the correlated subspace, include Ni-s, Ni-p, and O-p orbitals. The high-energy itinerant states are those in the full Hilbert space, but not in the low-energy subspace. For demonstration, we computed three quantities for each DFT eigenstate (nk⟩n\mathbf{k}\rangle) and represented them in a band structure plot using RGB colors: ∑M∈C∣⟨φM∣nk⟩∣2\sum_{M\in C}|\langle\varphi_M|n\mathbf{k}\rangle|^2 (red), ∑M∉C,M∈L∣⟨φM∣nk⟩∣2\sum_{M\notin C,M\in L}|\langle\varphi_M|n\mathbf{k}\rangle|^2 (green), and ∑M∉L,M∈F∣⟨φM∣nk⟩∣2\sum_{M\notin L,M\in F}|\langle\varphi_M|n\mathbf{k}\rangle|^2 (blue). Orbitals near the Fermi energy are primarily correlated orbitals, whereas in the 5-10 eV excitation range, they consist mainly of low-energy itinerant orbitals. For excitations above 10 eV, the orbitals are predominantly high-energy itinerant states.

Three distinct Hilbert spaces and their projection onto NiO DFT eigenstates

Figure 3. Three distinct Hilbert spaces (C, L, and F) and their projection onto NiO DFT eigenstates.

We also interpolate the Green’s function, GGWG^{GW}, and the screened Coulomb interaction, WGWW^{GW}, onto a finer Matsubara frequency grid. The simulation temperature determines the frequency grid for both GGWG^{GW} and WGWW^{GW}. For the ab initio GW calculation, we employ the LqsgwFlapw package; however, GW calculations at temperatures lower than 1000K can be computationally costly. As an alternative, we approximate GGWG^{GW} and WGWW^{GW} at temperatures lower than 1000K via interpolation of the quantities obtained at 1000K. For the frequency interpolation of Green’s function, we are assuming that G‾GW−1(kf,ω=0)=(G‾GW−1(kf,iωn=1c)+(G‾GW−1(kf,iωn=1c))†)/2\overline{G}^{-1}_{GW}(\mathbf{k}^{f},\omega=0)=\left(\overline{G}^{-1}_{GW}(\mathbf{k}^{f},i\omega^{c}_{n=1})+\left(\overline{G}^{-1}_{GW}(\mathbf{k}^{f},i\omega^{c}_{n=1})\right)^{\dagger}\right)/2, where iωn=1ci\omega^{c}_{n=1} is the lowest Matsubara frequency in the coarse frequency grid. This is based on the fact that G‾GW−1\overline{G}^{-1}_{GW} is determined by the GW self-energy diagram and thus expected to exhibit smooth behavior near zero frequency. This has been confirmed in the cases of NiO and SrVO3_3.

For a function (A) in the low-energy space and correlated space, we utilize A‾\overline{A} and A~\widetilde{A} notations, respectively. To exemplify, AA represents the function AA in the full space, which is calculated on a coarse crystal momentum and frequency grid. A‾\overline{A} denotes the same function in the low-energy space, calculated on a fine crystal momentum and frequency grid using interpolation techniques. Lastly, A~\widetilde{A} is denoting the function in the correlated space, calculated on a fine frequency grid.

For the projection of a parent space onto a subspace, we employ the notation P^\hat{P}, while E^\hat{E} is used for the embedding of a subspace quantity. Subscripts on the E^\hat{E} and P^\hat{P} denote the initial and destination vector spaces, and FF, LL, and CC are used to represent full space, low-energy space, and correlated space, respectively. Additionally, superscripts “f” and “b” signify “Fermionic” and “Bosonic”. To illustrate, P^F→Lf(A)\hat{P}^{f}_{F\to L}(A) denotes the fermionic projection of AA from the full space to the low-energy space.

Figure 4 provides the complete GW+EDMFT flowchart. To begin, LqsgwFlapw computes the GW Green’s function (GGWG_{GW}) and GW screened Coulomb interaction (WGWW_{GW}). Subsequently, ComBasis constructs localized orbitals φif(r)\varphi^{f}_{i}(\mathbf{r}) and φαf(r)\varphi^{f}_{\alpha}(\mathbf{r}) as well as bosonic basis functions φib(r)\varphi^{b}_{i}(\mathbf{r}) and φαb(r)\varphi^{b}_{\alpha}(\mathbf{r}). These basis functions in the low-energy space and correlated space facilitate fermionic projection (P^f\hat{P}^{f}), fermionic embedding (E^f\hat{E}^{f}), bosonic projection (P^b\hat{P}^{b}) and bosonic embedding (E^b\hat{E}^{b}). ComLocal then calculates the local Green’s function (G~loc\widetilde{G}_{loc}) and local screened Coulomb interaction W~loc\widetilde{W}_{loc} under the GW+EDMFT approximation. Following this, ComDC obtains the double-counted self-energy (ΣDC\Sigma_{DC}) and double-counting polarizability (ΠDC\Pi_{DC}) within the local GW approximation. ComCTQMC and ComWeiss together evaluate the fermionic and bosonic Weiss fields, and compute the EDMFT self-energy (ΣEDMFT\Sigma_{EDMFT}) and EDMFT polarizability (ΠEDMFT\Pi_{EDMFT}). Finally, ComEmbed embeds the EDMFT and double-counting objects into GGWG_{GW} and WGWW_{GW} in the full space. This loop is iteratively solved until a converged solution is achieved.

ComDMFT v2.0 components and the flowchart for full GW+EDMFT calculation

Figure 4. ComDMFT v2.0 components and the flowchart for full GW+EDMFT calculation. Inputs, outputs and important equations each component solves are also shown. Colored equations show GW (blue), EDMFT (red), and double counting (green) calculations shown in Figure 2

LqsgwFlapw

full GW calculation

We conducted a first-principles study to obtain the GW Green’s Function, GGWG_{GW}, and the GW screened Coulomb Interaction, WGWW_{GW}, using LqsgwFlapw [4]. To this end, we modified LqsgwFlapw to calculate the GW self-energy, ΣGW\Sigma_{GW} and GW polarizability, ΠGW\Pi_{GW}, from GG and WW within the framework of GW+EDMFT.

ΣGW(kc,iωnc)=−∑Rc∫0βcdτG(Rc,τ)∘W(−Rc,−τ)e−i(kc⋅Rc−ωncτ)PGW(kc,iνnc)=∑Rc∫0βcdτG(Rc,τ)∘G(−Rc,−τ)e−i(kc⋅Rc−νncτ)(25)\begin{aligned} \Sigma_{GW}(\mathbf{k}^{c},i\omega_{n}^{c}) &= -\sum_{\mathbf{R}^{c}}\int_{0}^{\beta^{c}} d\tau G(\mathbf{R}^{c},\tau)\circ W(-\mathbf{R}^{c},-\tau)e^{-i(\mathbf{k}^{c}\cdot\mathbf{R}^{c}-\omega_{n}^{c}\tau)} \\ P_{GW}(\mathbf{k}^{c},i\nu_{n}^{c}) &= \sum_{\mathbf{R}^{c}}\int_{0}^{\beta^{c}} d\tau G(\mathbf{R}^{c},\tau)\circ G(-\mathbf{R}^{c},-\tau)e^{-i(\mathbf{k}^{c}\cdot\mathbf{R}^{c}-\nu_{n}^{c}\tau)} \tag*{(25)} \end{aligned}

Here, superscript “c” in kc\mathbf{k}^{c}, Rc\mathbf{R}^{c}, iωnci\omega_{n}^{c}, iνnci\nu_{n}^{c} and βc\beta^{c} denotes “coarse” frequency and momentum grids, and ∘\circ denotes element-wise product in the position and spin

space. Then LqsgwFlapw calculates GGWG_{GW} and WGWW_{GW} in the following way.

GGW(kc,iωnc)=(GH−1(kc,iωnc)−ΣGW(kc,iωnc))−1WGW(kc,iωnc)=(V−1(kc)−PGW(kc,iωnc))−1(26)\begin{aligned} G_{GW}(\mathbf{k}^{c},i\omega_{n}^{c}) &= \left(G_{H}^{-1}(\mathbf{k}^{c},i\omega_{n}^{c})-\Sigma_{GW}(\mathbf{k}^{c},i\omega_{n}^{c})\right)^{-1} \\ W_{GW}(\mathbf{k}^{c},i\omega_{n}^{c}) &= \left(V^{-1}(\mathbf{k}^{c})-P_{GW}(\mathbf{k}^{c},i\omega_{n}^{c})\right)^{-1} \tag*{(26)} \end{aligned}

Hartree energy matrix element in the position representation is calculated by using GG within GW+EDMFT

ΣH(x,x′,kc)=δ(x−x′)1Nk∑k1c∫dx′′G(x′′,x′′,k1c,τ=0−)V(x,x′′,k2c=0).(27)\Sigma_{H}(\mathbf{x},\mathbf{x}',\mathbf{k}^{c})=\delta(\mathbf{x}-\mathbf{x}')\frac{1}{N_{\mathbf{k}}}\sum_{\mathbf{k}_{1}^{c}}\int d\mathbf{x}''G(\mathbf{x}'',\mathbf{x}'',\mathbf{k}_{1}^{c},\tau=0^{-})V(\mathbf{x},\mathbf{x}'',\mathbf{k}_{2}^{c}=0). \tag*{(27)}

Quasiparticle Hamiltonian construction

For the purpose of Wannier interpolation of GGWG_{GW} and WGWW_{GW} in the low-energy space, localized basis sets are required. We construct Wannier functions spanning the low-energy space (Hilbert space in the energy window of −10 eV<E−EF<10 eV-10\ eV < E-E_{F} < 10\ eV) by using the Wannier90 package. The Wannier90 package requires the best mean-field Hamiltonian to reproduce the fully interacting GW Green’s function, GGWG^{GW}. By linearizing the GW self-energy ΣGW\Sigma^{GW} in the frequency, we construct a GW quasiparticle Hamiltonian at each iteration in the following manner.

HGWQP(kc)=ZGW(kc)(HH(kc)+ΣGW(kc,ω=0))ZGW(kc)(28)H_{GW}^{QP}(\mathbf{k}^{c})=\sqrt{Z_{GW}(\mathbf{k}^{c})}\left(H_{H}(\mathbf{k}^{c})+\Sigma_{GW}(\mathbf{k}^{c},\omega=0)\right)\sqrt{Z_{GW}(\mathbf{k}^{c})} \tag*{(28)}
ZGW(kc)=(1−∂ΣGW(kc,ω)∂ω)∣ω=0−1Z_{GW}(\mathbf{k}^{c})=\left(1-\left.\frac{\partial\Sigma_{GW}(\mathbf{k}^{c},\omega)}{\partial\omega}\right)\right|_{\omega=0}^{-1}

Here, HHH_{H} denotes Hartree Hamiltonian. By diagonalizing HGWQP(kc)H_{GW}^{QP}(\mathbf{k}^{c}), we obtain energy eigenvectors ∣nkc⟩|n\mathbf{k}^{c}\rangle, where nn is the band index. These eigenvectors are then utilized in order to calculate the inputs for the Wannier90 package. It is important to note that HGWQPH_{GW}^{QP} is solely constructed in order to build Wannier functions. The comparison between HGWQPH_{GW}^{QP} bandstructure and the spectral function of GGWG^{GW} show excellent agreement within EF±10eVE_{F}\pm10eV. Please see the supplementary materials.

ComBasis

Calculating local quantities such as G~loc\widetilde{G}_{loc} and W~loc\widetilde{W}_{loc} presents a challenge due to the need to calculate GGWG_{GW} and WGWW_{GW} on a fine k-point grid. The accuracy of G~loc\widetilde{G}_{loc} and W~loc\widetilde{W}_{loc} is dependent on the momentum resolution of the aforementioned quantities; however, obtaining them from first principles is too computationally expensive. Wannier interpolation can be employed to interpolate GGW(kc,iωn)G_{GW}(\mathbf{k}^{c},i\omega_{n}) and WGW(kc,iωn)W_{GW}(\mathbf{k}^{c},i\omega_{n}) into those on a fine k grid instead.

For the Wannier interpolation of fermionic quantities in low-energy space, Maximally Localized Wannier Functions (MLWFs) spanning the low-energy space are constructed. To find orbitals with maximum localizations (or minimum of the total orbital spreads), quasiparticle wavefunction information is needed for Wannier90. For this purpose, we utilized the eigenfunctions (∣ψnk⟩|\psi_{nk}\rangle) of HGWQP(kc)H^{QP}_{GW}(\mathbf{k}^{c}) as stated in Eq. (28).

The other input to the Wannier90 package is trial orbitals ⟨r∣TiR⟩\langle\mathbf{r}|T_{iR}\rangle, which provide the initial condition for the total spread minimization. We employed muffin-tin orbitals of the LAPW basis set for our trial orbitals. The radial function of an initial trial orbital can be any linear combination of any muffin-tin orbitals with the same angular momentum character and that of the centered atom. ComBasis takes the linear combination which maximizes ∑Efrozmin⁡<En,k<Efrozmax⁡∣⟨ψnk∣Tik⟩∣2\sum_{E^{\min}_{\mathrm{froz}}<E_{n,k}<E^{\max}_{\mathrm{froz}}}|\langle\psi_{nk}|T_{ik}\rangle|^2. Here, Efrozmin⁡/max⁡E^{\min/\max}_{\mathrm{froz}} is the lower/upper bound of the frozen energy window, and ∣Tik⟩|T_{ik}\rangle is the Bloch sum of ∣TiR⟩|T_{iR}\rangle. In ComBasis, the angular momentum part of the trial orbitals is approximately defined as the cubic-harmonics YlmY_{lm} by default when there is no spin-orbit coupling present.

Ylm={i2(Yl−∣m∣−(−1)mYl∣m∣),m<0,Yl0,m=0,12(Yl−∣m∣+(−1)mYl∣m∣),m>0.(29)Y_{lm} = \begin{cases} \frac{i}{\sqrt{2}}\left(Y_l^{-|m|}-(-1)^mY_l^{|m|}\right), & m<0,\\ Y_l^0, & m=0,\\ \frac{1}{\sqrt{2}}\left(Y_l^{-|m|}+(-1)^mY_l^{|m|}\right), & m>0. \end{cases} \tag*{(29)}

where YlmY_l^m are the spherical harmonics. That is, a subscript mm corresponds to the cubic harmonics, and a superscript mm denotes spherical harmonics.

For a system with spin-orbit coupling, the angular part is approximately the spin-angular function (Ωl,a,m)(\Omega_{l,a,m}):

Ωl,a=±12,m=∑s=±12Ca,sl,mYlm−s(r^)us.(30)\Omega_{l,a=\pm\frac{1}{2},m}=\sum_{s=\pm\frac{1}{2}}C_{a,s}^{l,m}Y_l^{m-s}(\hat{r})u_s. \tag*{(30)}

Here, usu_s is a spinor and Ca,sl,m=⟨l,m−s,12,s∣l+a,m⟩C_{a,s}^{l,m}=\langle l,m-s,\frac{1}{2},s|l+a,m\rangle. Then, following the procedure outlined in § 3, we constructed orthonormalized product basis from the MLWFs (∣ψi⟩|\psi_i\rangle) spanning low-energy space and obtained a basis set for the bosonic quantities in that space.

We select a subset of orbitals with desired center positions and angular momentum characters from MLWFs and label them as correlated orbitals, which constitute the basis set of fermionic quantities in the correlated space. Subsequently, we construct an orthonormalized product basis of these correlated orbitals, thereby obtaining a basis set for the bosonic quantities in the correlated space, following the procedure detailed in § 3.

Now, we possess all the necessary components for the projection and embedding between three distinct spaces. The projection and embedding between Full space and low-energy space can be accomplished via the following approach.

P^F→L,ijf/b(A)=∑MN∈F⟨φif/b∣φMf/b⟩AMN⟨φNf/b∣φjf/b⟩\widehat{P}^{f/b}_{F\to L,ij}(A)=\sum_{M N\in F}\langle\varphi_i^{f/b}|\varphi_M^{f/b}\rangle A_{MN}\langle\varphi_N^{f/b}|\varphi_j^{f/b}\rangle
E^L→F,MNf/b(Aˉ)=∑ij∈L⟨φMf/b∣φif/b⟩Aˉij⟨φjf/b∣φNf/b⟩(31)\widehat{E}^{f/b}_{L\to F,MN}(\bar{A})=\sum_{ij\in L}\langle\varphi_M^{f/b}|\varphi_i^{f/b}\rangle\bar{A}_{ij}\langle\varphi_j^{f/b}|\varphi_N^{f/b}\rangle \tag*{(31)}

Projection and embedding between low-energy space and correlated space can be achieved in the following way.

P^L→C,αβf/b(A‾)=∑ij∈L⟨φαf/b∣φif/b⟩A‾ij⟨φjf/b∣φβf/b⟩\hat{P}^{f/b}_{L\to C,\alpha\beta}(\overline{A})=\sum_{ij\in L}\langle\varphi^{f/b}_{\alpha}|\varphi^{f/b}_{i}\rangle\overline{A}_{ij}\langle\varphi^{f/b}_{j}|\varphi^{f/b}_{\beta}\rangle
E^C→L,ijf/b(A~)=∑αβ∈C⟨φif/b∣φαf/b⟩A~αβ⟨φβf/b∣φjf/b⟩(32)\hat{E}^{f/b}_{C\to L,ij}(\widetilde{A})=\sum_{\alpha\beta\in C}\langle\varphi^{f/b}_{i}|\varphi^{f/b}_{\alpha}\rangle\widetilde{A}_{\alpha\beta}\langle\varphi^{f/b}_{\beta}|\varphi^{f/b}_{j}\rangle \tag*{(32)}

Finally, a mapping between the full space and the correlated-energy space can be constructed through the following projection and embedding.

P^F→C,αβf/b(A)=∑MN∈F⟨φαf/b∣φMf/b⟩AMN⟨φNf/b∣φβf/b⟩\hat{P}^{f/b}_{F\to C,\alpha\beta}(A)=\sum_{MN\in F}\langle\varphi^{f/b}_{\alpha}|\varphi^{f/b}_{M}\rangle A_{MN}\langle\varphi^{f/b}_{N}|\varphi^{f/b}_{\beta}\rangle
E^C→F,MNf/b(A~)=∑αβ∈C⟨φMf/b∣φαf/b⟩A~αβ⟨φβf/b∣φNf/b⟩(33)\hat{E}^{f/b}_{C\to F,MN}(\widetilde{A})=\sum_{\alpha\beta\in C}\langle\varphi^{f/b}_{M}|\varphi^{f/b}_{\alpha}\rangle\widetilde{A}_{\alpha\beta}\langle\varphi^{f/b}_{\beta}|\varphi^{f/b}_{N}\rangle \tag*{(33)}

ComLocal

ComLocal calculates the local Green’s function (G~loc\widetilde{G}_{\mathrm{loc}}) and local screened Coulomb interaction (W~loc\widetilde{W}_{\mathrm{loc}}) in the following way. First, with the input of G‾GW(kc,iωnc)\overline{G}_{GW}(\mathbf{k}^{c},i\omega_{n}^{c}) and W‾GW(kc,iνnc)\overline{W}_{GW}(\mathbf{k}^{c},i\nu_{n}^{c}) of LqsgwFlapw, ComLocal projects them into the low-energy space.

G‾GW(kc,iωnc)=P^F→Lf(GGW(kc,iωnc))(34)\overline{G}_{GW}(\mathbf{k}^{c},i\omega_{n}^{c})=\hat{P}^{f}_{F\to L}(G_{GW}(\mathbf{k}^{c},i\omega_{n}^{c})) \tag*{(34)}
W‾GW(kc,iνnc)=P^F→Lb(WGW(kc,iνnc))(35)\overline{W}_{GW}(\mathbf{k}^{c},i\nu_{n}^{c})=\hat{P}^{b}_{F\to L}(W_{GW}(\mathbf{k}^{c},i\nu_{n}^{c})) \tag*{(35)}

Next, ComLocal Wannier-interpolates them from the coarse k-grid to the fine k-grid, to obtain G‾GW(kf,iωnc)\overline{G}_{GW}(\mathbf{k}^{f},i\omega_{n}^{c}) and W‾GW(kf,iνnc)\overline{W}_{GW}(\mathbf{k}^{f},i\nu_{n}^{c}). Subsequently, the interpolation from the coarse Matsubara grid to fine-Matsubara grid is carried out to acquire G‾GW(kf,iωnf)\overline{G}_{GW}(\mathbf{k}^{f},i\omega_{n}^{f}) and W‾GW(kf,iνnf)\overline{W}_{GW}(\mathbf{k}^{f},i\nu_{n}^{f}). Then, EDMFT and double-counting quantities are embedded into G‾GW(kf,iωnf)\overline{G}_{GW}(\mathbf{k}^{f},i\omega_{n}^{f}) and W‾GW(kf,iνnf)\overline{W}_{GW}(\mathbf{k}^{f},i\nu_{n}^{f}).

G‾(kf,iωnf)={G‾GW−1(kf,iωnf)−E^C→Lf(Σ~EDMFT(iωnf)−Σ~DC(iωnf))}−1(36)\overline{G}(\mathbf{k}^{f},i\omega_{n}^{f})=\left\{\overline{G}_{GW}^{-1}(\mathbf{k}^{f},i\omega_{n}^{f})-\hat{E}^{f}_{C\to L}\left(\widetilde{\Sigma}_{EDMFT}(i\omega_{n}^{f})-\widetilde{\Sigma}_{DC}(i\omega_{n}^{f})\right)\right\}^{-1} \tag*{(36)}
W‾(kf,iνnf)={W‾GW−1(kf,iνnf)−E^C→Lb(Π~EDMFT(iνnf)−Π~DC(iνnf))}−1(37)\overline{W}(\mathbf{k}^{f},i\nu_{n}^{f})=\left\{\overline{W}_{GW}^{-1}(\mathbf{k}^{f},i\nu_{n}^{f})-\hat{E}^{b}_{C\to L}\left(\widetilde{\Pi}_{EDMFT}(i\nu_{n}^{f})-\widetilde{\Pi}_{DC}(i\nu_{n}^{f})\right)\right\}^{-1} \tag*{(37)}

Finally, we project G‾\overline{G} and W‾\overline{W} into the correlated subspace and obtain local quantities.

G~loc(iωnf)=P^L→Cf(G‾(kf,iωnf))(38)\widetilde{G}_{\mathrm{loc}}(i\omega_{n}^{f})=\hat{P}^{f}_{L\to C}\left(\overline{G}(\mathbf{k}^{f},i\omega_{n}^{f})\right) \tag*{(38)}
W~loc(iνnf)=P^L→Cb(W‾(kf,iνnf))(39)\widetilde{W}_{\mathrm{loc}}(i\nu_{n}^{f})=\hat{P}^{b}_{L\to C}\left(\overline{W}(\mathbf{k}^{f},i\nu_{n}^{f})\right) \tag*{(39)}

ComDC

ComDC is the module to calculate the double-counted electron self-energy and polarizability. Local GW self-energy has been calculated in the following way.

Σ~DC,α,β(iωn)=−∑γδ∫dτ G~loc,δγ(τ)W~loc,αγβδ(τ)eiωnτ.(40)\widetilde{\Sigma}_{DC,\alpha,\beta}(i\omega_{n})=-\sum_{\gamma\delta}\int d\tau\,\widetilde{G}_{\mathrm{loc},\delta\gamma}(\tau)\widetilde{W}_{\mathrm{loc},\alpha\gamma\beta\delta}(\tau)e^{i\omega_{n}\tau}. \tag*{(40)}

where α,β,γ\alpha,\beta,\gamma and δ\delta are composite indices for both orbital and spin (spin-angular function index) in the system without (with) spin-orbit coupling. Double-counting polarizability is calculated in the following way.

P~αβγδ(iνn)=∫dτG~δβ(τ)G~γα(−τ)eiνnτ.(41)\widetilde{P}_{\alpha\beta\gamma\delta}(i\nu_n)=\int d\tau\widetilde{G}_{\delta\beta}(\tau)\widetilde{G}_{\gamma\alpha}(-\tau)e^{i\nu_n\tau}. \tag*{(41)}

ComWeiss

ComWeiss calculates bosonic as well as fermionic Weiss field in the following way.

G~=(G~loc−1+Σ~EDMFT)−1U~=(W~loc−1+Π~EDMFT)−1(42)\begin{aligned} \widetilde{G} &= \left(\widetilde{G}_{\mathrm{loc}}^{-1}+\widetilde{\Sigma}_{EDMFT}\right)^{-1}\\ \widetilde{U} &= \left(\widetilde{W}_{\mathrm{loc}}^{-1}+\widetilde{\Pi}_{EDMFT}\right)^{-1} \tag*{(42)} \end{aligned}

ComCTQMC

In order to solve the impurity model action as required by ComDMFT, hybridization-expansion continuous time quantum Monte Carlo (CTQMC) is employed. CTQMC is an established stochastic approach for obtaining exact numerical solutions of an impurity model, namely a small interacting system embedded in an electron bath. The EDMFT action of the impurity model, required for GW+DMFT, is

S=−∑αχ∫∫0βdτdτ′cα†(τ)G~αχ−1(τ−τ′)cχ(τ′)+12∑αχγδ∫∫0βdτdτ′cα†(τ)cχ†(τ′)U~αχγδ(τ−τ′)cγ(τ′)cδ(τ),(43)\begin{aligned} S={}&-\sum_{\alpha\chi}\int\int_{0}^{\beta}d\tau d\tau' c_{\alpha}^{\dagger}(\tau)\widetilde{G}_{\alpha\chi}^{-1}(\tau-\tau')c_{\chi}(\tau')\\ &+\frac{1}{2}\sum_{\alpha\chi\gamma\delta}\int\int_{0}^{\beta}d\tau d\tau' c_{\alpha}^{\dagger}(\tau)c_{\chi}^{\dagger}(\tau')\widetilde{U}_{\alpha\chi\gamma\delta}(\tau-\tau')c_{\gamma}(\tau')c_{\delta}(\tau), \tag*{(43)} \end{aligned}

where cα†c_{\alpha}^{\dagger} creates an electron in the generalized orbital α\alpha (which includes both spin and orbital degrees of freedom), β\beta is the inverse temperature.

The present study postulates that the frequency-dependent interaction is assumed to take the following form.

U~αχγδ(iνn)=U~αχγδ+F0(iνn)δαδδχγ,(44)\widetilde{U}_{\alpha\chi\gamma\delta}(i\nu_n)=\widetilde{U}_{\alpha\chi\gamma\delta}+F^{0}(i\nu_n)\delta_{\alpha\delta}\delta_{\chi\gamma}, \tag*{(44)}

We take into account only the dynamical screening of the Slater-Condon parameter F0F^{0} for the sake of convenience in the numerical algorithm based on hybridization-expansion CTQMC. U~αχγδ\widetilde{U}_{\alpha\chi\gamma\delta} are frequency-independent and are approximated by their values at iνn=∞i\nu_n=\infty. The fermionic Weiss field is divided into two parts:

G~αχ−1(iωn)=(iωn+μ)δαχ−t~αχ−Δ~αχ(iωn),(45)\widetilde{G}_{\alpha\chi}^{-1}(i\omega_n)=(i\omega_n+\mu)\delta_{\alpha\chi}-\widetilde{t}_{\alpha\chi}-\widetilde{\Delta}_{\alpha\chi}(i\omega_n), \tag*{(45)}

Here, t~\widetilde{t} is the impurity-level matrix and Δ~\widetilde{\Delta} is the hybridization function matrix. By solving the quantum impurity action, we obtain the self-energy (Σ~EDMFT\widetilde{\Sigma}_{EDMFT}) and the polarizability(Π~EDMFT\widetilde{\Pi}_{EDMFT}).

ComEmbed

In the final step, EDMFT and double-counting quantities were incorporated into the GGW(kc,iωnc)G_{GW}(\mathbf{k}^{c},i\omega_{n}^{c}) and WGW(kc,iνnc)W_{GW}(\mathbf{k}^{c},i\nu_{n}^{c}) of LqsgwFlapw.

G={GGW−1−E^C→Ff(Σ~EDMFT−Σ~DC)}−1(46)G=\left\{G_{GW}^{-1}-\hat{E}_{C\to F}^{f}\left(\widetilde{\Sigma}_{EDMFT}-\widetilde{\Sigma}_{DC}\right)\right\}^{-1} \tag*{(46)}
W={WGW−1−E^C→Fc(Π~EDMFT−Π~DC)}−1W=\left\{W_{GW}^{-1}-\hat{E}_{C\to F}^{c}\left(\widetilde{\Pi}_{EDMFT}-\widetilde{\Pi}_{DC}\right)\right\}^{-1}

The obtained Green’s function, G(kc,iωnc)G(\mathbf{k}^{c},i\omega_{n}^{c}), and the screened Coulomb interaction, W(kc,iνnc)W(\mathbf{k}^{c},i\nu_{n}^{c}), will be utilized to calculate the self-energy, ΣGW(kc,iωnc)\Sigma_{GW}(\mathbf{k}^{c},i\omega_{n}^{c}), and the polarization function, PGW(kc,iνnc)P_{GW}(\mathbf{k}^{c},i\nu_{n}^{c}) in eq. (25).

Implementation details

Causal Optimization

Truncations in the Ψ\Psi functional of the Luttinger-Ward (or Baym-Kadanoff) functional[64, 65, 66] provides a roadmap to build GW+EDMFT theory. Recently, it has been shown that the free energy functional can be multivalued [105, 106, 107, 108, 109]. Moreover, it has been shown that this multivaluedness can bring about noncausality of the Green’s functions and self energies[105]. Noncausal Green’s functions can also arise from non-local extension of dynamical mean field theory[110, 111, 112]. To address this problem, we used the causal optimization method[113] which replaces non-causal objects with best-possible causal objects. In this implementation, we used the method to find the closest causal ΠEDMFT\Pi_{EDMFT}, U\mathcal{U}, and WlocW_{\mathrm{loc}}, if these quantities are non-causal.

Database architecture

In ComDMFT v2.0, data exchange between components is facilitated by a database. This version supports writing and reading data with the parallel HDF5 IO library. Each component stores data in a singular HDF5 file called “global.dat”. Keys are provided to ensure the data is easily retrieved when needed. There are four HDF5 groups included in the file: ”Full Space”, ”Low energy space”, ”Correlated Space”, and ”Correlated Problem” which are denoted as different Hilbert spaces in Sec. 3. Data in each space is stored in its corresponding HDF5 group. Additionally, the “Correlated Problem” group is designed to store local quantities of each inequivalent impurity problem. Figure 5 provides an overview of the groups and important data stored in the “global.dat”. For example, GG and WW obtained from the LqsgwFlapw calculation are stored in the “full space group”, while P^F→Lf^/b\hat{P}_{F\rightarrow L}^{\hat{f}/b} and E^L→Ff^/b\hat{E}_{L\rightarrow F}^{\hat{f}/b} are stored in the “Low-energy space” group.

Grouped data structure within GW+EDMFT

Figure 5. Grouped data structure within GW+EDMFT.

Results

To evaluate the performance of fully self-consistent GW+EDMFT, we calculated the electronic structure of several quantum materials, with different degrees of correlation, NiO and SrVO3_3 at a temperature of 1000 K. We compared the full GW+EDMFT results with experiments, as well as with those of LQSGW+DMFT. We find that the intermediate quantities such as the Weiss fields are quite different in the two methods, but the final physical results are in rather good agreement.

NiO

NiO is a paradigmatic Mott insulator and it has been intensively studied [114, 115, 116, 117, 118, 119, 120, 121]. Experimentally, NiO is a paramagnetic insulator; however, traditional ab initio electronic structure methodologies based on one-particle picture predict it to be metallic when it is not ordered magnetically. We use NiO to illustrate the GW+EDMFT method and compare the results to LQSGW+DMFT. Both calculations treat the Ni-3dd orbitals as correlated orbitals, and the simulation temperature is 1000 K.

GW self-consistency

Figure 6 (a) and (Figure 6b) show the Wannier-interpolated GW bandstructures (HGWQPH_{GW}^{QP} in eq. (28)), obtained at the first and last iterations of the GW+EDMFT self-consistent loop in Figure 4, respectively. The GW self-energy, as expressed in equation 25, does not exhibit any divergence near the chemical potential.

Wannier-interpolated GW bandstructures of NiO at the first and last iterations of the self-consistent loop

Figure 6. (a) and (b) show the Wannier-interpolated GW bandstructures (HGWQPH_{GW}^{QP} in eq. (28)) of NiO at the first and last iterations of the self-consistent loop in Figure 4, respectively.

Consequently, ΣGW\Sigma_{GW} can be linearized safely even in this charge transfer insulator. This implies that the EDMFT self-energy, described in eq. (14), must have a divergence near the chemical potential to open a Mott insulating gap. Furthermore, the importance of GW self-consistency is evident when comparing Figure 6 (a) and (Figure 6b). The band width of two low-lying bands (largely originated from Ni-ege_g and O-pp) increased by 50%. A strong downshift of the occupied Ni-t2gt_{2g} bands is also observed. Note that this is not the GW+EDMFT bandstructure, and that the spectral function within GW+EDMFT can be obtained by embedding EDMFT and double counting self-energy to the GW Green’s Function.

Hybridization functions

Figure 7 illustrates the imaginary part of the hybridization functions associated with Ni-d orbitals on the imaginary frequency axis. Results from both LQSGW+DMFT and GW+EDMFT are displayed in blue and orange, respectively. It is observed that Ni-ege_g orbitals have a stronger hybridization near the chemical potential compared to Ni-t2gt_{2g} orbitals in both methodologies. Additionally, GW+EDMFT reveals a substantially enhanced hybridization function when compared to the LQSGW+DMFT result. Within LQSGW+DMFT, it is found to be near-zero at ∼200\sim200 eV. However, a peak is present even at an energy of ∼400\sim400 eV in the GW+EDMFT hybridization function.

Imaginary parts of hybridization functions associated with Ni-$t_{2g}$ and Ni-$e_g$ orbitals

Figure 7. Imaginary parts of hybridization functions associated with (a) Ni-t2gt_{2g} and (b) Ni-ege_g of NiO. (c) and (d) are zoom in view of (a) and (b), respectively. Orange and blue colors show the results within full GW+EDMFT and LQSGW+DMFT, respectively

One possible explanation for this surprising feature of the extended tail in the hybridization function is the incoherent spectral weight introduced by ab initio GW self-energy diagrams. These diagrams introduce significant incoherence to the Green’s function, resulting in a greater deviation from the 1/iωn1/i\omega_n asymptotic limit in the correlated subspace. Consequently, this leads to an extended tail of the hybridization function, in contrast to DFT+DMFT or LQSGW+DMFT. In the supplementary materials, we provide a numerical demonstration of this notable feature, emphasizing the role of incoherent spectral weight from ab initio GW self-energy diagrams in extending the hybridization function’s tail.

We note that the high energy features which show up in GW+EDMFT (but no other existing method), and therefore look shocking, have little effect on the local self-energy and polarizability. Indeed, CTQMC finds identical self-energies when we dramatically adjusted these hybridization functions in order to test the consequences of these high energy features.2

Bosonic Weiss field

Figure 8 (a) presents the monopole part (F0F^0) of the bosonic Weiss field associated with Ni-dd orbitals. Figure 8 (b) displays a zoomed-in view of the same figure. The static value of F0F^0 is 10 eVeV within full GW+EDMFT, which is 40% greater than that acquired from LQSGW+DMFT. Moreover, this value further increases to 15 eVeV at a frequency of 10 eVeV, indicating a strong frequency dependence in the low energy region. The dynamical screening captured by the Lang-Firsov approximation is suggested to be a pathway for renormalization of the hybridization function[122]. Discussion on renormalization of the hybridization function due to the dynamical screening is in the following section.

Monopole part of the bosonic Weiss field and zoom-in view

Figure 8. (a) The monopole part (F0F^0 of Slater integrals) of the bosonic Weiss field F0F^0 associated with Ni-dd orbitals in NiO. (b) zoom-in view of (a). Orange and blue colors show the results within full GW+EDMFT and LQSGW+DMFT, respectively

EDMFT self-energies

Figure 9 presents the imaginary parts of the EDMFT self-energies for the Ni-t2gt_{2g} and Ni-ege_g orbitals on the imaginary frequency axis. The blue and orange lines refer to LQSGW+DMFT and GW+EDMFT, respectively. The Ni-t2gt_{2g} orbital self-energy displays a linear dependence on the imaginary frequency near the chemical potential; this is a characteristic of Fermi-liquid behavior. In contrast, the imaginary part of the Ni-ege_g self-energy reveals the onset of the divergence near the chemical potential, indicating a Mott-insulating phase. This is supported by the impurity Green’s function in the static limit; on the Matsubara axis, Ni-ege_g impurity Green’s function approaches to zero values. Interestingly, the Ni-t2gt_{2g} orbital quasi-particle weights within the two different theories are found to be comparable, with values of 0.71 and 0.47 within full GW+EDMFT and LQSGW+DMFT, respectively. This is surprisingly similar, given the significant differences between the fermionic and bosonic Weiss fields in the two theories. This implies a compensation of the enhanced fermionic Weiss field by the enhanced bosonic Weiss fields.

Imaginary parts of EDMFT self-energies associated with (a) Ni-$t_{2g}$ and (b) Ni-$e_g$ in NiO

Figure 9. Imaginary parts of EDMFT self-energies associated with (a) Ni-t2gt_{2g} and (b) Ni-ege_g in NiO. Orange and blue colors show the results within full GW+EDMFT and LQSGW+DMFT, respectively

Double-counting self-energies

Fig. 10 displays the double-counting self-energies associated with Ni-dd orbitals on the imaginary frequency axis within the local-GW approximation. The results obtained from the LQSGW+DMFT and GW+EDMFT methodologies are shown in blue and orange lines, respectively. Similarly, it is observed that neither the Ni-ege_g nor the Ni-t2gt_{2g} orbitals exhibit any divergence in their double-counting self-energies near zero frequency. However, there are quantitative differences in the quasiparticle-weights between the two methodologies; the Ni-t2gt_{2g} orbitals possess quasiparticle-weights of 0.88 and 0.60 in GW+EDMFT and LQSGW+DMFT, respectively, whereas for the Ni-ege_g orbitals, the quasiparticle-weights are 0.88 and 0.47.

Imaginary parts of the double-counting self-energies associated with (a) Ni-$t_{2g}$ and (b) Ni-$e_g$ in NiO

Figure 10. Imaginary parts of the double-counting self-energies associated with (a) Ni-t2gt_{2g} and (b) Ni-ege_g in NiO. Orange and blue colors show the results within full GW+EDMFT and LQSGW+DMFT, respectively

Spectral functions

Figure 11 presents the electronic structures of NiO without magnetic order with various approaches: LDA, LQSGW, LQSGW+DMFT and GW+EDMFT. GW+EDMFT spectral function are obtained by the the analytical continuation of the momentum-resolved Greens function on the imaginary frequency axis. Results from the non spin-polarized LDA and non spin-polarized LQSGW approaches indicate that NiO is a metal when it is not magnetically ordered, as evidenced by the presence of electronic states near the chemical potential. In contrast, GW+EDMFT and LQSGW+DMFT predict that NiO is an insulator.

NiO electronic structures within (a) non spin-polarized LDA, (b) non spin-polarized LQSGW, (c) paramagnetic LQSGW+DMFT, and (d) paramagnetic full GW+EDMFT

Figure 11. NiO electronic structures within (a) non spin-polarized LDA and (b) non spin-polarized LQSGW, (c) paramagnetic LQSGW+DMFT and (d) paramagnetic full GW+EDMFT, respectively

Projected density of states

Fig. 12 demonstrates projected density of states (DOS) from LQSGW+DMFT and GW+EDMFT. GW+EDMFT data is obtained through analytical continuation of the orbital-resolved Greens function on the imaginary frequency axis. This data reveals that for Ni-ege_g orbitals, both methodologies open up a well-defined Mott insulating gap, as indicated by the strong divergence of the EDMFT self-energy. The Ni-t2gt_{2g} orbitals are fully-filled as expected. The separation between peaks in XPS and BIS is approximately 6eV [123]. The GW+EDMFT method slightly overestimates this value. The LQSGW+DMFT method underestimates the separation.

SrVO3_3

SrVO3_3 has been recognized to be one of the simplest correlated transition metal oxides. It is a cubic perovskite with a three V-d bands crossing the Fermi level. Photoemission and inverse photoemission spectra both exhibit satellite peaks[124], making it a topical subject for DMFT methodologies[95, 125, 126, 127, 87, 96, 128, 129, 103, 130, 131] [43]. In this paper, we apply GW+EDMFT to this metallic phase of SrVO3_3 and compare the results with other methodologies. For our calculations, the V-3d orbitals were treated as correlated orbitals, and the simulation temperature was set to T=1000K.

GW self-consistency

Figure 13 (a) and (b) display the Wannier-interpolated GW bandstructures (HGWQPH_{GW}^{QP}) in eq. (28), obtained at the first and last iterations of the self-consistent loop in Figure 4 respectively. As the GW self-energy in eq. (25) does not exhibit any divergence near the chemical potential, a linearization of the self-energy can be safely applied to construct HGWQPH_{GW}^{QP} in this correlated metallic phase. Notably, the GW bandstructure at the first and last iteration is almost identical, indicating that the EDMFT and double counting quantities compensate each other.

Image not extracted

Figure 13. (a) and (b) show the Wannier-interpolated GW bandstructures (HGWQPH_{GW}^{QP} in eq. (28)) of SrVO3_3 at the first and last iterations of the self-consistent loop in Figure 4, respectively.

Hybridization function

Figure 14 displays the imaginary part of the hybridization functions for the VV-t2gt_{2g} and VV-ege_g orbitals on the imaginary frequency axis. Blue and orange colors in the figure correspond to the results obtained from LQSGW+DMFT and GW+EDMFT, respectively. As in NiO, the magnitude of this hybridization is very different for the same reasons discussed in Sec. 4.1.

Image not extracted

Figure 14. Imaginary parts of hybridization functions associated with (a) V-t2gt_{2g} and (b) Ni-ege_g of SrVO3_3. (c) and (d) are zoom-in view of (a) and (b), respectively. Orange and blue colors show the results within full GW+EDMFT and LQSGW+DMFT, respectively

Bosonic Weiss field

Figure 15 (a) presents the monopole part of the bosonic Weiss field associated with V-d orbitals (F0F^0 of Slater integrals). Figure 15 (b) is a zoomed-in view. The displayed blue and orange lines represent the results obtained by means of LQSGW+DMFT and GW+EDMFT, respectively. The static value of the monopole terms is 4 eV in both theories. As is the case for NiO, GW+EDMFT exhibits a stronger frequency dependence in the low-energy region, with the monopole term reaching 15 eV at a frequency of 10 eV. As discussed in 4.1, this renormalizes the hybridization.

Image not extracted

Figure 15. (a) The monopole part (F0F^0 of Slater integrals) of the bosonic Weiss field F0F^0 associated with V-d orbitals in SrVO3_3. (b) zoom-in view of (a). Orange and blue colors show the results within full GW+EDMFT and LQSGW+DMFT, respectively

Image not extracted

Figure 12. Projected density of states of NiO within (a) LQSGW+DMFT and (b) GW+EDMFT. (c) Comparison of the Ni-dd density of states with experimental result [123].

EDMFT self-energies

Figure 16 illustrates the imaginary parts of the EDMFT self-energies for both V-t2gt_{2g} and V-ege_g orbitals as represented by the blue and orange lines, respectively, on the imaginary frequency axis. These results confirm the presence of sizable electron correlations for both orbitals. The quasiparticle weights of V-t2gt_{2g} were determined to be 0.58 and 0.53 within LQSGW+DMFT and GW+EDMFT, respectively, while the quasiparticle weights of V-ege_g were 0.73 and 0.70 within LQSGW+DMFT and GW+EDMFT, respectively. These results suggest the cancellation between the enhanced hybridization and Weiss field, as indicated by the nearly identical quasiparticle weights in both approaches.

Image not extracted

Figure 16. Imaginary parts of EDMFT self-energies associated with (a) V-t2gt_{2g} and (b) V-ege_g in SrVO3\mathrm{SrVO}_3. Orange and blue colors show the results within full GW+EDMFT and LQSGW+DMFT, respectively

Double-counting self-energy

Figure 17 depicts the double-counted self-energies of V-t2gt_{2g} and V-ege_g orbitals in the local-GW approximation on the imaginary frequency axis. Blue and orange lines indicate the results of LQSGW+DMFT and GW+EDMFT, respectively. It is observed that both of these approaches exhibit qualitatively similar outcomes, indicating that the V-t2gt_{2g} orbital is more correlated than V-dz2d_{z^2}. Additionally, while the self-energies provided by LQSGW+DMFT and

Image not extracted

Figure 17. Imaginary parts of the double-counting self-energies associated with (a) V-t2gt_{2g} and (b) V-ege_g in SrVO3\mathrm{SrVO}_3. Orange and blue colors show the results within full GW+EDMFT and LQSGW+DMFT, respectively

GW+EDMFT are almost identical, the quasiparticle weight in the local GW is slightly greater than that of EDMFT in GW+EDMFT.

Spectral functions

Figure 18 displays the electronic structures of SrVO3\mathrm{SrVO}_3 computed with various ab initio approaches: non spin-polarized Local Density Approximation (LDA), non spin-polarized LQSGW method, LQSGW+DMFT, and full GW+EDMFT. Momentum-resolved spectral function data on the real frequency axis is derived from the analytical continuation of the momentum-resolved Greens function on the imaginary frequency axis. All approaches lead to qualitatively similar band structures in the vicinity of the Fermi level. Three bands cross the Fermi level and are distinct from the other bands. The hole excitation energies at the Γ\Gamma point obtained from LDA, LQSGW, LQSGW+DMFT and GW+EDMFT are 1.0, 0.5, 0.7, and 0.7 eV, respectively, compared to the experimental value of approximately 0.5 eV[132, 133, 134, 133].

Image not extracted

Figure 18. SrVO3\mathrm{SrVO}_3 electronic structures within (a) non spin-polarized LDA and (b) non spin-polarized LQSGW, (c) paramagnetic LQSGW+DMFT and (d) paramagnetic full GW+EDMFT, respectively

Projected density of states

Fig. 19 illustrates the projected density of states of both LQSGW+DMFT and GW+EDMFT. Results from both theories present similar features, with three peak structures located at -4 eV, 0 eV, and 2 eV matching well with each other. With both theories, we did not observe strong satellite peaks on either side of the Fermi level, which agrees with the recent claim that defect states are the source of the observed satellite peak at -1.5 eV[132]. Additionally, only a VV-ege_g peak located at 2 eV was observed above the Fermi level.

Conclusions

We presented the implementation of fully self-consistent ab initio GW+EDMFT for the electronic structure of correlated quantum materials in ComDMFT. This implementation enables the electronic structure calculation of quantum materials within weak, intermediate, and strong electron correlation without prior knowledge of the degree of electron correlation. We tested our implementation by calculating the electronic structures of NiO and SrVO3\mathrm{SrVO}_3 and obtained qualitatively similar results to LQSGW+DMFT.

Image not extracted

Figure 19. Projected density of states of SrVO3\mathrm{SrVO}_3 within (a) LQSGW+DMFT and (b) GW+EDMFT.

Acknowledgments

This paper was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences as a part of the Computational Materials Science Program. S.C. was supported by a KIAS individual Grant (No. CG090601) at Korea Institute for Advanced Study. M.H. was supported by a KIAS individual Grant (No. CG091301) at Korea Institute for Advanced Study. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of Science User Facility operated under Contract No. DE-AC02-05CH11231.

References

  1. [4]A. L. Kutepov, V. S. Oudovenko, G. Kotliar, Linearized self-consistent quasiparticle GW method: Application to semiconductors and simple metals 219 407–414. doi:10.1016/j.cpc.2017.06.012. URL https://www.sciencedirect.com/science/article/pii/S0010465517301947DOI
  2. [5]G. Xu, B. Lian, P. Tang, X.-L. Qi, S.-C. Zhang, Topological Superconductivity on the Surface of Fe-Based Superconductors 117 (4) 047001. doi:10.1103/PhysRevLett.117.047001. URL https://link.aps.org/doi/10.1103/PhysRevLett.117.047001DOI
  3. [6]D. Wang, L. Kong, P. Fan, H. Chen, S. Zhu, W. Liu, L. Cao, Y. Sun, S. Du, J. Schneeloch, R. Zhong, G. Gu, L. Fu, H. Ding, H.-J. Gao, Evidence for Majorana bound states in an iron-based superconductor 362 (6412) 333–335. arXiv:30115743, doi:10.1126/science.aao1797. URL https://science.sciencemag.org/content/362/6412/333DOI
  4. [7]P. Zhang, K. Yaji, T. Hashimoto, Y. Ota, T. Kondo, K. Okazaki, Z. Wang, J. Wen, G. D. Gu, H. Ding, S. Shin, Observation of topological superconductivity on the surface of an iron-based superconductor 360 (6385) 182–186. arXiv:29519917, doi:10.1126/science.aan4596. URL https://science.sciencemag.org/content/360/6385/182DOI
  5. [8]R. M. Fernandes, A. I. Coldea, H. Ding, I. R. Fisher, P. J. Hirschfeld, G. Kotliar, Iron pnictides and chalcogenides: A new paradigm for superconductivity 601 (7891) 35–44. doi:10.1038/s41586-021-04073-2. URL https://www.nature.com/articles/s41586-021-04073-2
  6. [9]P. R. C. Kent, G. Kotliar, Toward a predictive theory of correlated materials 361 (6400) 348–354. arXiv:30049873, doi:10.1126/science.aat5975. URL http://science.sciencemag.org/content/361/6400/348DOI
  7. [10]W. Metzner, D. Vollhardt, Correlated Lattice Fermions in $d=\ensuremath{\infty}$ Dimensions 62 (3) 324–327. doi:10.1103/PhysRevLett.62.324. URL https://link.aps.org/doi/10.1103/PhysRevLett.62.324DOI
  8. [11]E. Müller-Hartmann, Correlated fermions on a lattice in high dimensions 74 (4) 507–512. doi:10.1007/BF01311397. URL https://doi.org/10.1007/BF01311397
  9. [12]U. Brandt, C. Mielsch, Thermodynamics and correlation functions of the Falicov-Kimball model in large dimensions 75 (3) 365–370. doi:10.1007/BF01321824. URL https://doi.org/10.1007/BF01321824
  10. [23]V. I. Anisimov, J. Zaanen, O. K. Andersen, Band theory and Mott insulators: Hubbard \textit{U} instead of Stoner \textit{I} 44 (3) 943–954. doi:10.1103/PhysRevB.44.943. URL http://link.aps.org/doi/10.1103/PhysRevB.44.943DOI
  11. [24]V. I. Anisimov, I. V. Solovyev, M. A. Korotin, M. T. Czyżyk, G. A. Sawatzky, Density-functional theory and NiO photoemission spectra 48 (23) 16929. URL http://journals.aps.org/prb/abstract/10.1103/PhysRevB.48.16929pubmed.ncbi.nlm.nih.gov/10008291
  12. [25]A. I. Liechtenstein, V. I. Anisimov, J. Zaanen, Density-functional theory and strong interactions: Orbital ordering in Mott-Hubbard insulators 52 (8) R5467–R5470. doi:10.1103/PhysRevB.52.R5467. URL http://link.aps.org/doi/10.1103/PhysRevB.52.R5467DOI
  13. [26]V. I. Anisimov, A. I. Poteryaev, M. A. Korotin, A. O. Anokhin, G. Kotliar, First-principles calculations of the electronic structure and spectra of strongly correlated systems: Dynamical mean-field theory 9 (35) 7359. doi:10.1088/0953-8984/9/35/010. URL http://stacks.iop.org/0953-8984/9/i=35/a=010DOI
  14. [27]A. I. Liechtenstein, M. I. Katsnelson, Ab initio calculations of quasiparticle band structure in correlated systems: LDA++ approach 57 (12) 6884–6895. doi:10.1103/PhysRevB.57.6884. URL https://link.aps.org/doi/10.1103/PhysRevB.57.6884
  15. [28]G. Kotliar, S. Y. Savrasov, K. Haule, V. S. Oudovenko, O. Parcollet, C. A. Marianetti, Electronic structure calculations with dynamical mean-field theory 78 (3) 865–951. doi:10.1103/RevModPhys.78.865. URL https://link.aps.org/doi/10.1103/RevModPhys.78.865DOI
  16. [29]K. Held, G. Keller, V. Eyert, D. Vollhardt, V. I. Anisimov, Mott-Hubbard Metal-Insulator Transition in Paramagnetic $V_{1−x}O_{3}$: An $LDA+DMFT$\(\mathrm{QMC}\)$ Study 86 (23) 5345–5348. doi:10.1103/PhysRevLett.86.5345. URL https://link.aps.org/doi/10.1103/PhysRevLett.86.5345
  17. [30]S. Y. Savrasov, G. Kotliar, E. Abrahams, Correlated electrons in $\ensuremath{\delta}$-plutonium within a dynamical mean-field picture 410 (6830) 793–795. doi:10.1038/35071035. URL https://www.nature.com/articles/35071035DOI
  18. [31]P. Werner, A. Comanac, p. u. family=Medici, given=Luca, M. Troyer, A. J. Millis, Continuous-Time Solver for Quantum Impurity Models 97 (7) 076405. doi:10.1103/PhysRevLett.97.076405. URL http://link.aps.org/doi/10.1103/PhysRevLett.97.076405DOI
  19. [32]H. Park, A. J. Millis, C. A. Marianetti, Computing total energies in complex materials using charge self-consistent DFT + DMFT 90 (23) 235103. doi:10.1103/PhysRevB.90.235103. URL https://link.aps.org/doi/10.1103/PhysRevB.90.235103DOI
  20. [33]K. Haule, T. Birol, Free Energy from Stationary Implementation of the $\mathrm{DFT}+\mathrm{DMFT}$ Functional 115 (25) 256402. doi: 10.1103/PhysRevLett.115.256402. URL https://link.aps.org/doi/10.1103/PhysRevLett.115.256402DOI
  21. [34]K. Haule, G. L. Pascut, Forces for structural optimizations in correlated materials within a DFT+embedded DMFT functional approach 94 (19) 195146. doi:10.1103/PhysRevB.94.195146. URL https://link.aps.org/doi/10.1103/PhysRevB.94.195146DOI
  22. [35]K. Haule, Exact Double Counting in Combining the Dynamical Mean Field Theory and the Density Functional Theory 115 (19) 196403. doi: 10.1103/PhysRevLett.115.196403. URL https://link.aps.org/doi/10.1103/PhysRevLett.115.196403DOI
  23. [36]O. Parcollet, M. Ferrero, T. Ayral, H. Hafermann, I. Krivenko, L. Messio, P. Seth, TRIQS: A toolbox for research on interacting quantum systems 196 398–415. doi:10.1016/j.cpc.2015.04.023. URL http://www.sciencedirect.com/science/article/pii/S0010465515001666DOI
  24. [37]A. Gaenko, A. E. Antipov, G. Carcassi, T. Chen, X. Chen, Q. Dong, L. Gamper, J. Gukelberger, R. Igarashi, S. Iskakov, M. Könz, J. P. F. LeBlanc, R. Levy, P. N. Ma, J. E. Paki, H. Shinaoka, S. Todo, M. Troyer, E. Gull, Updated core libraries of the ALPS project 213 235–251. doi:10.1016/j.cpc.2016.12.009. URL http://www.sciencedirect.com/science/article/pii/S0010465516303885DOI
  25. [38]L. Huang, Y. Wang, Z. Y. Meng, L. Du, P. Werner, X. Dai, iQIST: An open source continuous-time quantum Monte Carlo impurity solver toolkit 195 140–160. doi:10.1016/j.cpc.2015.04.020. URL http://www.sciencedirect.com/science/article/pii/S0010465515001629DOI
  26. [39]M. Wallerberger, A. Hausoel, P. Gunacker, A. Kowalski, N. Parragh, F. Goth, K. Held, G. Sangiovanni, W2dynamics: Local one- and two-particle quantities from dynamical mean field theory 235 388–399. doi:10.1016/j.cpc.2018.09.007. URL https://www.sciencedirect.com/science/article/pii/S0010465518303217DOI
  27. [40]K. Haule, C.-H. Yee, K. Kim, Dynamical mean-field theory within the full-potential methods: Electronic structure of CeIrIn_{5}, CeCoIn_{5},
  28. [49]The Elk Code. URL http://elk.sourceforge.net/
  29. [50]X. Gonze, F. Jollet, F. Abreu Araujo, D. Adams, B. Amadon, T. Applencourt, C. Audouze, J. M. Beuken, J. Bieder, A. Bokhanchuk, E. Bousquet, F. Bruneval, D. Caliste, M. Côté, F. Dahm, F. Da Pieve, M. Delaveau, M. Di Gennaro, B. Dorado, C. Espejo, G. Geneste, L. Genovese, A. Gerossier, M. Giantomassi, Y. Gillet, D. R. Hamann, L. He, G. Jomard, J. Laflamme Janssen, S. Le Roux, A. Levitt, A. Lherbier, F. Liu, I. Lukaˇcević, A. Martin, C. Martins, M. J. T. Oliveira, S. Poncé, Y. Pouillon, T. Rangel, G. M. Rignanese, A. H. Romero, B. Rousseau, O. Rubel, A. A. Shukri, M. Stankovski, M. Torrent, M. J. Van Setten, B. Van Troeye, M. J. Verstraete, D. Waroquiers, J. Wiktor, B. Xu, A. Zhou, J. W. Zwanziger, Recent developments in the ABINIT software package 205 106–131. doi:10.1016/j.cpc.2016.04.003. URL http://www.sciencedirect.com/science/article/pii/S0010465516300923DOI
  30. [51]J. M. Wills, M. Alouani, P. Andersson, A. Delin, O. Eriksson, O. Grechnyev, The Full-Potential Electronic Structure Problem and RSPt, in: Full-Potential Electronic Structure Method, Springer Series in Solid-State Sciences, Springer, Berlin, Heidelberg, pp. 47–73. doi:10.1007/978-3-642-15144-6_6. URL https://link.springer.com/chapter/10.1007/978-3-642-15144-6_6
  31. [52]X. Deng, L. Wang, X. Dai, Z. Fang, Local density approximation combined with Gutzwiller method for correlated electron systems: Formalism and applications 79 (7) 075114. doi:10.1103/PhysRevB.79.075114. URL https://link.aps.org/doi/10.1103/PhysRevB.79.075114
  32. [53]Y. X. Yao, C. Z. Wang, K. M. Ho, Including many-body screening into self-consistent calculations: Tight-binding model studies with the Gutzwiller approximation 83 (24) 245139. doi:10.1103/PhysRevB.83.245139. URL https://link.aps.org/doi/10.1103/PhysRevB.83.245139
  33. [54]K. M. Ho, J. Schmalian, C. Z. Wang, Gutzwiller density functional theory for correlated electron systems 77 (7) 073101. doi:10.1103/PhysRevB.77.073101. URL https://link.aps.org/doi/10.1103/PhysRevB.77.073101
  34. [55]T. Schickling, J. Bünemann, F. Gebhard, W. Weber, Gutzwiller density functional theory: A formal derivation and application to ferromagnetic nickel 16 (9) 093034. doi:10.1088/1367-2630/16/9/093034. URL https://doi.org/10.1088%2F1367-2630%2F16%2F9%2F093034
  35. [56]N. Lanatà, Y. Yao, C.-Z. Wang, K.-M. Ho, G. Kotliar, Phase Diagram and Electronic Structure of Praseodymium and Plutonium 5 (1) 011008.
  36. [77]V. Blum, R. Gehrke, F. Hanke, P. Havu, V. Havu, X. Ren, K. Reuter, M. Scheffler, Ab Initio molecular simulations with numeric atom-centered orbitals 180 (11) 2175–2196. doi:10.1016/j.cpc.2009.06.022. URL https://www.sciencedirect.com/science/article/pii/S0010465509002033DOI
  37. [78]X. Blase, C. Attaccalite, V. Olevano, First-principles GW calculations for fullerenes, porphyrins, phthalocyanine, and other molecules of interest for organic photovoltaic applications 83 (11) 115103. doi:10.1103/PhysRevB.83.115103. URL https://link.aps.org/doi/10.1103/PhysRevB.83.115103DOI
  38. [79]F. Hüser, T. Olsen, K. S. Thygesen, Quasiparticle GW calculations for solids, molecules, and two-dimensional materials 87 (23) 235132. doi:10.1103/PhysRevB.87.235132. URL https://link.aps.org/doi/10.1103/PhysRevB.87.235132DOI
  39. [80]Q. Sun, X. Zhang, S. Banerjee, P. Bao, M. Barbry, N. S. Blunt, N. A. Bogdanov, G. H. Booth, J. Chen, Z.-H. Cui, J. J. Eriksen, Y. Gao, S. Guo, J. Hermann, M. R. Hermes, K. Koh, P. Kovál, S. Lehtola, Z. Li, J. Liu, N. M. Richardsson, J. D. McClain, M. Motta, B. Mussard, H. Q. Pham, A. Pulkin, W. Purwanto, P. J. Robinson, E. Ronca, E. R. Sayfutyarova, M. Scheurer, H. F. Schurkus, J. E. T. Smith, C. Sun, S.-N. Sun, S. Upadhyay, L. K. Wagner, X. Wang, A. White, J. D. Whitfield, M. J. Williamson, S. Wouters, J. Yang, J. M. Yu, T. Zhu, T. C. Berkelbach, S. Sharma, A. Y. Sokolov, G. K.-L. Chan, Recent developments in the PySCF program package 153 (2) 024109. doi:10.1063/5.0006074. URL https://doi.org/10.1063/5.0006074
  40. [81]D. Pashov, S. Acharya, W. R. L. Lambrecht, J. Jackson, K. D. Belashchenko, A. Chantis, F. Jamet, p. u. family=Schilgaarde, given=Mark, QuestaaI: A package of electronic structure methods based on the linear muffin-tin orbital technique 249 107065. doi:10.1016/j.cpc.2019.107065. URL https://www.sciencedirect.com/science/article/pii/S0010465519303868DOI
  41. [82]VASP - Vienna Ab initio Simulation Package. URL https://www.vasp.at/
  42. [83]M. Govoni, G. Galli, Large Scale GW Calculations 11 (6) 2680–2696. doi:10.1021/ct500958p. URL https://doi.org/10.1021/ct500958p
  43. [84]D. Sangalli, A. Ferretti, H. Miranda, C. Attaccalite, I. Marri, E. Cannuccia, P. Melo, M. Marsili, F. Paleari, A. Marrazzo, G. Prandini, P. Bonfà, M. O. Atambo, F. Affinito, M. Palummo, A. Molina-Sánchez, C. Hogan, M. Grüning, D. Varsano, A. Marini, Many-body perturbation theory calculations using the yambo code 31 (32) 325902. doi:10.1088/1361-648X/ab15d0. URL https://dx.doi.org/10.1088/1361-648X/ab15d0
  44. [85]P. Sun, G. Kotliar, Extended dynamical mean-field theory and GW method 66 (8) 085120. doi:10.1103/PhysRevB.66.085120. URL http://link.aps.org/doi/10.1103/PhysRevB.66.085120DOI
  45. [86]S. Biermann, F. Aryasetiawan, A. Georges, First-Principles Approach to the Electronic Structure of Strongly Correlated Systems: Combining the GW Approximation and Dynamical Mean-Field Theory 90 (8) 086402. doi:10.1103/PhysRevLett.90.086402. URL http://link.aps.org/doi/10.1103/PhysRevLett.90.086402DOI
  46. [87]F. Nilsson, L. Boehnke, P. Werner, F. Aryasetiawan, Multitier self-consistent $GW+\text{EDMFT}$ 1 (4) 043803. doi:10.1103/PhysRevMaterials.1.043803. URL https://link.aps.org/doi/10.1103/PhysRevMaterials.1.043803DOI
  47. [88]Q. Si, J. L. Smith, Kosterlitz-Thouless Transition and Short Range Spatial Correlations in an Extended Hubbard Model 77 (16) 3391–3394. doi:10.1103/PhysRevLett.77.3391. URL https://link.aps.org/doi/10.1103/PhysRevLett.77.3391DOI
  48. [89]A. M. Sengupta, A. Georges, Non-Fermi-liquid behavior near a T=0 spin-glass transition 52 (14) 10295–10302. doi:10.1103/PhysRevB.52.10295. URL https://link.aps.org/doi/10.1103/PhysRevB.52.10295DOI
  49. [90]Henrik Kajueter, Interpolating perturbation scheme for correlated electron systems.
  50. [91]Z.-H. Cui, T. Zhu, G. K.-L. Chan, Efficient Implementation of Ab Initio Quantum Embedding in Periodic Systems: Density Matrix Embedding Theory 16 (1) 119–129. doi:10.1021/acs.jctc.9b00933. URL https://doi.org/10.1021/acs.jctc.9b00933
  51. [92]T. Zhu, G. K.-L. Chan, Ab Initio Full Cell $GW+\mathrm{DMFT}$ for Correlated Materials 11 (2) 021006. doi:10.1103/PhysRevX.11.021006. URL https://link.aps.org/doi/10.1103/PhysRevX.11.021006DOI
  52. [93]T. N. Lan, A. Shee, J. Li, E. Gull, D. Zgid, Testing self-energy embedding theory in combination with GW 96 (15) 155106. doi:10.1103/PhysRevB.96.155106. URL https://link.aps.org/doi/10.1103/PhysRevB.96.155106DOI
  53. [94]J. M. Tomczak, QSGW +DMFT: An electronic structure scheme for the iron pnictides and beyond 592 (1) 012055. doi:10.1088/1742-6596/592/1/012055. URL http://stacks.iop.org/1742-6596/592/i=1/a=012055DOI
  54. [95]J. M. Tomczak, M. Casula, T. Miyake, S. Biermann, Asymmetry in band widening and quasiparticle lifetimes in SrVO3: Competition between screened exchange and local correlations from combined GW and dynamical mean-field theory GW + DMFT 90 (16) 165138. doi:10.1103/PhysRevB.90.165138. URL https://link.aps.org/doi/10.1103/PhysRevB.90.165138DOI
  55. [96]C. Taranto, M. Kaltak, N. Parragh, G. Sangiovanni, G. Kresse, A. Toschi, K. Held, Comparing quasiparticle $GW$+DMFT and LDA+DMFT for the test bed material SrVO${}_{3}$ 88 (16) 165119. doi:10.1103/PhysRevB.88.165119. URL http://link.aps.org/doi/10.1103/PhysRevB.88.165119DOI
  56. [97]P. Werner, M. Casula, T. Miyake, F. Aryasetiawan, A. J. Millis, S. Biermann, Satellites and large doping and temperature dependence of electronic properties in hole-doped BaFe2As2 8 (4) 331–337. doi:10.1038/nphys2250. URL http://www.nature.com/nphys/journal/v8/n4/abs/nphys2250.htmlDOI
  57. [98]L. Sponza, P. Pisanti, A. Vishina, D. Pashov, C. Weber, p.-u. family=Schilfgaarde, given=Mark, S. Acharya, J. Vidal, G. Kotliar, Self-energies in itinerant magnets: A focus on Fe and Ni 95 (4) 041112. doi:10.1103/PhysRevB.95.041112. URL https://link.aps.org/doi/10.1103/PhysRevB.95.041112DOI
  58. [99]S. Choi, A. Kutepov, K. Haule, p.-u. family=Schilfgaarde, given=Mark, G. Kotliar, First-principles treatment of Mott insulators: Linearized QSGW+DMFT approach 1 16001. doi:10.1038/npjquantmats.2016.1. URL https://www.nature.com/articles/npjquantmats20161DOI
  59. [100]p.-u. family=Roekeghem, given=Ambroise, T. Ayral, J. M. Tomczak, M. Casula, N. Xu, H. Ding, M. Ferrero, O. Parcollet, H. Jiang, S. Biermann, Dynamical Correlations and Screened Exchange on the Experimental Bench: Spectral Properties of the Cobalt Pnictide ${\mathrm{BaCo}}_{2}{\mathrm{As}}_{2}$ 113 (26) 266403. doi:10.1103/PhysRevLett.113.266403. URL https://link.aps.org/doi/10.1103/PhysRevLett.113.266403DOI
  60. [101]F. Petocchi, V. Christiansson, P. Werner, Fully ab initio electronic structure of Ca 2 RuO 4 104 (19) 195146. doi:10.1103/PhysRevB.104.195146. URL https://link.aps.org/doi/10.1103/PhysRevB.104.195146
  61. [102]F. Petocchi, V. Christiansson, F. Nilsson, F. Aryasetiawan, P. Werner, Normal State of ${\mathrm{Nd}}_{\{1\ensuremath{-}x\}{\mathrm{Sr}}_{\{x\}}{\mathrm{NiO}}_{\{2\}}$ from Self-Consistent $GW+\mathrm{EDMFT}$ 10 (4) 041047. doi:10.1103/PhysRevX.10.041047. URL https://link.aps.org/doi/10.1103/PhysRevX.10.041047
  62. [103]F. Petocchi, F. Nilsson, F. Aryasetiawan, P. Werner, Screening from ${e}_{\{g\}}$ states and antiferromagnetic correlations in ${d}^{\{(1,2,3)\}}$ perovskites: A $GW+\text{EDMFT}$ investigation 2 (1) 013191. doi:10.1103/PhysRevResearch.2.013191. URL https://link.aps.org/doi/10.1103/PhysRevResearch.2.013191DOI
  63. [104]L. Boehnke, F. Nilsson, F. Aryasetiawan, P. Werner, When strong correlations become weak: Consistent merging of G W and DMFT 94 (20) 201106. doi:10.1103/PhysRevB.94.201106. URL https://link.aps.org/doi/10.1103/PhysRevB.94.201106DOI
  64. [105]J. Vuˇ cičević, N. Wentzell, M. Ferrero, O. Parcollet, Practical consequences of the Luttinger-Ward functional multivaluedness for cluster DMFT methods 97 (12) 125411. doi:10.1103/PhysRevB.97.125411. URL https://link.aps.org/doi/10.1103/PhysRevB.97.125411
  65. [106]R. Rossi, F. Werner, Skeleton series and multivaluedness of the self-energy functional in zero space-time dimensions 48 (48) 485202. doi:10.1088/1751-8113/48/48/485202. URL https://dx.doi.org/10.1088/1751-8113/48/48/485202
  66. [107]A. J. Kim, V. Sacksteder, Multivaluedness of the Luttinger-Ward functional in the fermionic and bosonic system with replicas 101 (11) 115146. doi:10.1103/PhysRevB.101.115146. URL https://link.aps.org/doi/10.1103/PhysRevB.101.115146DOI
  67. [108]O. Gunnarsson, G. Rohringer, T. Schäfer, G. Sangiovanni, A. Toschi, Breakdown of Traditional Many-Body Theories for Correlated Electrons 119 (5) 056402. doi:10.1103/PhysRevLett.119.056402. URL https://link.aps.org/doi/10.1103/PhysRevLett.119.056402DOI
  68. [109]E. Kozik, M. Ferrero, A. Georges, Nonexistence of the Luttinger-Ward Functional and Misleading Convergence of Skeleton Diagrammatic Series for Hubbard-Like Models 114 (15) 156402. doi:10.1103/PhysRevLett.114.156402. URL https://link.aps.org/doi/10.1103/PhysRevLett.114.156402DOI
  69. [110]S. Backes, J.-H. Sim, S. Biermann, Nonlocal correlation effects in fermionic many-body systems: Overcoming the noncausality problem 105 (24) 245115. doi:10.1103/PhysRevB.105.245115. URL https://link.aps.org/doi/10.1103/PhysRevB.105.245115DOI
  70. [111]G. Biroli, O. Parcollet, G. Kotliar, Cluster dynamical mean-field theories: Causality and classical limit 69 (20) 205108. doi:10.1103/PhysRevB.69.205108. URL https://link.aps.org/doi/10.1103/PhysRevB.69.205108DOI
  71. [112]J. Lee, K. Haule, Diatomic molecule as a testbed for combining DMFT with electronic structure methods such as $GW$ and DFT 95 (15) 155104, comment: 13pages, 9 figures. arXiv:1611.07090, doi:10.1103/PhysRevB.95.155104. URL http://arxiv.org/abs/1611.07090
  72. [113]M. Han, H. J. Choi, Causal optimization method for imaginary-time Green’s functions in interacting electron systems 104 (11) 115112. doi:10.1103/PhysRevB.104.115112. URL https://link.aps.org/doi/10.1103/PhysRevB.104.115112DOI
  73. [114]J. Kuně`s, V. I. Anisimov, A. V. Lukoyanov, D. Vollhardt, Local correlations and hole doping in NiO: A dynamical mean-field study 75 (16). doi:10.1103/PhysRevB.75.165115. URL http://link.aps.org/doi/10.1103/PhysRevB.75.165115DOI
  74. [115]X. Ren, I. Leonov, G. Keller, M. Kollar, I. Nekrasov, D. Vollhardt, LDA + DMFT computation of the electronic spectrum of NiO 74 (19). doi:10.1103/PhysRevB.74.195114. URL http://link.aps.org/doi/10.1103/PhysRevB.74.195114DOI
  75. [116]M. Karolak, G. Ulm, T. Wehling, V. Mazurenko, A. Poteryaev, A. Lichtenstein, Double counting in LDA+DMFT—The example of NiO 181 (1) 11–15. doi:10.1016/j.elspec.2010.05.021. URL https://www.sciencedirect.com/science/article/pii/S0368204810001222DOI
  76. [117]L. Zhang, P. Staar, A. Kozhevnikov, Y.-P. Wang, J. Trinastic, T. Schulthess, H.-P. Cheng, $\mathrm{DFT}+\mathrm{DMFT}$ calculations of the complex band and tunneling behavior for the transition metal monoxides MnO, FeO, CoO, and NiO 100 (3) 035104. doi:10.1103/PhysRevB.100.035104. URL https://link.aps.org/doi/10.1103/PhysRevB.100.035104DOI
  77. [118]J. Kuně`s, V. I. Anisimov, S. L. Skornyakov, A. V. Lukoyanov, D. Vollhardt, NiO: Correlated Band Structure of a Charge-Transfer Insulator 99 (15). doi:10.1103/PhysRevLett.99.156404. URL https://link.aps.org/doi/10.1103/PhysRevLett.99.156404DOI
  78. [119]P. Thunström, I. Di Marco, O. Eriksson, Electronic Entanglement in Late Transition Metal Oxides 109 (18). doi:10.1103/PhysRevLett.109.186401. URL https://link.aps.org/doi/10.1103/PhysRevLett.109.186401DOI
  79. [120]Q. Yin, A. Gordienko, X. Wan, S. Y. Savrasov, Calculated Momentum Dependence of Zhang-Rice States in Transition Metal Oxides 100 (6). doi:10.1103/PhysRevLett.100.066406. URL https://link.aps.org/doi/10.1103/PhysRevLett.100.066406DOI
  80. [121]F. Lechermann, W. Körner, D. F. Urban, C. Elsässer, Interplay of charge-transfer and Mott-Hubbard physics approached by an efficient combination of self-interaction correction and dynamical mean-field theory. arXiv:1902.07000. URL http://arxiv.org/abs/1902.07000
  81. [122]M. Casula, g.-i. family=Werner, given=Ph., L. Vaugier, F. Aryasetiawan, T. Miyake, A. J. Millis, S. Biermann, Low-Energy Models for Correlated Materials: Bandwidth Renormalization from Coulombic Screening 109 (12) 126408. doi:10.1103/PhysRevLett.109.126408. URL https://link.aps.org/doi/10.1103/PhysRevLett.109.126408DOI
  82. [123]G. A. Sawatzky, J. W. Allen, Magnitude and Origin of the Band Gap in NiO 53 (24) 2339–2342. doi:10.1103/PhysRevLett.53.2339. URL http://link.aps.org/doi/10.1103/PhysRevLett.53.2339DOI
  83. [124]K. Morikawa, T. Mizokawa, K. Kobayashi, A. Fujimori, H. Eisaki, S. Uchida, F. Iga, Y. Nishihara, Spectral weight transfer and mass renormalization in Mott-Hubbard systems SrVO 3 and CaVO 3 : Influence of long-range Coulomb interaction 52 (19) 13711–13714. doi:10.1103/PhysRevB.52.13711. URL https://link.aps.org/doi/10.1103/PhysRevB.52.13711
  84. [125]R. Sakuma, g.-i. family=Werner, given=Ph., F. Aryasetiawan, Electronic structure of SrVO3 within GW+DMFT 88 (23) 235110. doi:10.1103/PhysRevB.88.235110. URL http://link.aps.org/doi/10.1103/PhysRevB.88.235110
  85. [126]p.-u. family=Roekegehem, given=Ambroise, S. Biermann, Screened exchange dynamical mean-field theory and its relation to density functional theory: SrVO3 and SrTiO3 108 (5) 57003. doi:10.1209/0295-5075/108/57003. URL http://iopscience.iop.org/0295-5075/108/5/57003
  86. [127]K. Haule, T. Birol, G. Kotliar, Covalency in transition-metal oxides within all-electron dynamical mean-field theory 90 (7) 075136. doi:10.1103/PhysRevB.90.075136. URL http://link.aps.org/doi/10.1103/PhysRevB.90.075136DOI
  87. [128]J. M. Tomczak, M. Casula, T. Miyake, F. Aryasetiawan, S. Biermann, Combined GW and dynamical mean-field theory: Dynamical screening effects in transition metal oxides 100 (6) 67001. doi:10.1209/0295-5075/100/67001. URL http://iopscience.iop.org/0295-5075/100/6/67001DOI
  88. [129]G. Trimarchi, I. Leonov, N. Binggeli, D. Korotin, V. I. Anisimov, LDA+DMFT implemented with the pseudopotential plane-wave approach 20 (13) 135227. doi:10.1088/0953-8984/20/13/135227. URL http://stacks.iop.org/0953-8984/20/i=13/a=135227DOI
  89. [130]H. Chen, H. Park, A. J. Millis, C. A. Marianetti, Charge transfer across transition-metal oxide interfaces: Emergent conductance and electronic structure 90 (24) 245138. doi:10.1103/PhysRevB.90.245138. URL https://link.aps.org/doi/10.1103/PhysRevB.90.245138DOI
  90. [131]I. A. Nekrasov, K. Held, G. Keller, D. E. Kondakov, g.-i. family=Pruschke, given=Th., M. Kollar, O. K. Andersen, V. I. Anisimov, D. Vollhardt, Momentum-resolved spectral functions of ${\mathrm{SrVO}}_{3}$ calculated by $\mathrm{LDA}+\mathrm{DMFT}$ 73 (15) 155112. doi:10.1103/PhysRevB.73.155112. URL https://link.aps.org/doi/10.1103/PhysRevB.73.155112DOI
  91. [132]S. Backes, T. C. Rödel, F. Fortuna, E. Frantzeskakis, P. Le Févre, F. Bertran, M. Kobayashi, R. Yukawa, T. Mitsuhashi, M. Kitamura, K. Horiba, H. Kumigashira, R. Saint-Martin, A. Fouchet, B. Berini, Y. Dumont, A. J. Kim, F. Lechermann, H. O. Jeschke, M. J. Rozenberg, R. Valentí, A. F. Santander-Syro, Hubbard band versus oxygen vacancy states in the correlated electron metal ${\mathrm{SrVO}}_{3}$ 94 (24) 241110. doi:10.1103/PhysRevB.94.241110. URL https://link.aps.org/doi/10.1103/PhysRevB.94.241110DOI
  92. [133]T. Yoshida, M. Kobayashi, K. Yoshimatsu, H. Kumigashira, A. Fujimori, Correlated electronic states of SrVO3 revealed by angle-resolved photoemission spectroscopy 208 11–16. doi:10.1016/j.elspec.2015.11.012. URL https://www.sciencedirect.com/science/article/pii/S0368204815002868DOI
  93. [134]T. Mitsuhashi, M. Minohara, R. Yukawa, M. Kitamura, K. Horiba, M. Kobayashi, H. Kumigashira, Influence of $k_{\ensuremath{\perp}}$ broadening on ARPES spectra of the (110) and (001) surfaces of ${\mathbf{SrVO}}_{3}$ films 94 (12) 125148. doi:10.1103/PhysRevB.94.125148. URL https://link.aps.org/doi/10.1103/PhysRevB.94.125148DOI

Paper details

vlnc:W_yq547wng
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents