In this work, we present a relativistic quantum embedding formalism capable of variationally treating relativistic effects, including scalar-relativity and spin–orbit coupling. We extend density functional theory (DFT)-in-DFT projection-based quantum embedding to a relativistic two-component formalism, where the full spin magnetization vector form is retained throughout the embedding treatment. To benchmark various relativistic embedding schemes, spin–orbit splitting of the nominally t2g valence manifold of W(CO)6, exchange coupling of [(H3N)4Cr(OH)2Cr(NH3)4]4+, and the dissociation potential curve of WF6 are investigated. The relativistic embedding formalism introduced in this work is well suited for efficient modeling of open-shell systems containing late transition metal, lanthanide, and actinide molecular complexes.

As the number of electrons in a system increases, one must make a compromise between cost and accuracy for the electronic structure method used, opening the door for quantum embedding methods. Quantum embedding requires the separation of the total system into a number of subsystems where the subsystems of interest are treated with a more accurate, oftentimes more expensive, method. Developments of quantum embedding in the past 20 years have been numerous and excellent reviews are available.1–6 

Embedding is often used as a tool to improve the electron–electron correlation of a subsystem by exploiting the locality of the interaction. Since relativistic effects, such as the scalar relativity and spin–orbit coupling, are more local than electron–electron correlation,7–9 embedding should be an effective tool for introducing new physics within the Dirac framework into the Hamiltonian of a quantum subsystem. Relativistic embedding has been previously explored;10–15 however, the handling of open-shell fragments in a relativistic framework is yet to be studied. Open-shell environments give rise to a scalar embedding potential and an additional vector potential that corresponds to spin magnetization. A relativistic embedding framework that retains the full magnetization vector form of the spin degrees of freedom can be particularly advantageous for studies of late transition metal, lanthanide, and actinide molecular complexes where relativistic effects play an essential role in their chemical properties and reactivities.

In this work, we seek to introduce a relativistic projection-based embedding scheme within the Dirac formalism, where the full vector form of magnetization density and non-collinear spin are retained throughout the embedding treatment. We considered three tests cases: the spin–orbit splitting of the valence t2g orbitals of W(CO)6, the exchange coupling constant of [(H3N)4Cr(OH)2Cr(NH3)4]4+, and the potential curve of WF6 dissociation. We first consider the splitting of the t2g manifold of W(CO)6 as a case where only scalar-potential embedding is needed due to the closed-shell nature, but inclusion of spin–orbit coupling is required. The exchange coupling constant of [(H3N)4Cr(OH)2Cr(NH3)4]4+ tests how well a local treatment of a complex two-spin-center interaction is modeled. We keep the spin centers in the embedded subsystem and use scalar-potential embedding. Finally, we consider the dissociation of WF6, where the spin centers are placed in two different subsystems. Proper handling of relativistic effects and embedding magnetic field are needed for this test. The method introduced here is capable of treating open-shell systems with strong relativistic effects.

In the spinor basis, the matrix representation of the kinetically balanced modified Dirac equation (in a.u.) is7,8

VTT14c2WTCL+CLCS+CS=S020212c2TCL+CLCS+CSϵ+00ϵ,
(1)

where c is the speed of light. V, T, and S are the two-component nonrelativistic potential energy, kinetic energy, and overlap matrices, respectively. W is the relativistic potential matrix, defined as (σp)V(σp), where p is the linear momentum operator and σ contains Pauli spin matrices. {ϵ+} and {ϵ} are the sets of positive/negative eigenvalues with corresponding molecular orbital coefficients (CL+CS+)T for the positive and (CLCS)T for the negative energy solutions.

In this work, we use the exact-two-component (X2C)16–31 technique to “fold” the pseudo-large component coefficients CS into the large component by a one-step unitary transformation, resulting in an electron-only Hamiltonian (see Ref. 24 for numerical implementations). To lower the computational cost, here we only use the bare Coulomb operator instead of transforming from the four-component two-electron operator, leading to the one-electron X2C approach. In the one-electron X2C framework, the transformation (or “picture change”) is independent of the two-electron operator. This simplification leads to a one-step procedure to construct the transformation matrix through the diagonalization of the one-electron four-component core Hamiltonian. This approximation neglects the “picture-change” of the two-electron and embedding operators. To empirically correct the neglect of two-electron spin–orbit coupling, we make use of the Boettger scaling factor32 

hijSNSO=hijX2C-SOQliQljZiZjhijX2C-SO,
(2)

where hSNSO is the scaling-factor-corrected spin–orbit Hamiltonian (screened-nuclear-spin–orbit Hamiltonian), i and j are basis functions, hX2C-SO is the spin–orbit part of the one-electron X2C Hamiltonian, Q is the number of electrons in the filled shell, l is the angular momentum, and Z is the nuclear charge where the basis function is centered. We work in a general Kramers-unrestricted formalism for representation of the one-electron wave functions.

The X2C transformation yields a set of spinor molecular orbitals, which are complex linear combinations of AO spin-orbitals. Spinor molecular orbitals in two-component relativistic electronic structure method can be written as

ψi(x)=ϕiα(r)ϕiβ(r),
(3)

where the spatial functions {ϕkα(r)}, {ϕkβ(r)} are expanded in terms of a common set of real atomic orbital basis functions with complex-valued coefficients, {cμiα} and {cμiβ}. One-body density matrix γ can be defined as

γμνστ=iocccμiσcνiτ*,σ,τ{α,β}.
(4)

In one-component nonrelativistic electronic structure methods, due to spin collinearity, only one of the magnetization densities can be non-zero,30,33 which is often conveniently chosen to be the z-component of the magnetization density vector. In the relativistic regime, the complete vector form of the magnetization density or the current density must be retained in any embedding framework in order to properly simulate open-shell systems and spin-driven reactions, such as intersystem crossing and homogeneous bond dissociation. We will work in the Pauli matrix basis where the spin-blocked Kohn–Sham matrix F and density matrix γ are cast as30,34–36

F=n=03Fnσn,
(5)
γ=n=03γnσn,
(6)

where

σ0=1001,σ1=0110,σ2=0ii0,σ3=1001.
(7)

The resulting scalar (F0) and spin/magnetic parts (F1, F2, F3) of the Kohn–Sham matrix are

F0=h0+J[γ0]ζK[γ0]+Vxc[γ0],
(8)
Fn=hnζK[γn]+Vxc[γn],n=1,2,3,
(9)

where h is the core Hamiltonian, J and K are the Coulomb and exchange matrices, respectively, ζ is the coefficient of the exact exchange in hybrid density functional theory (DFT), and Vxc is the torque-free non-collinear density functional kernel.29,30,37

In the Pauli matrix basis, γ now have a well-defined physical meaning: γ0 is the scalar charge density and γ1–3 are magnetization densities that give rise to the magnetic moments. A major advantage of developing embedding schemes in this framework is that each magnetization component can be independently aligned between the high-level and low-level subsystems, and components of vector embedding potentials are also decoupled. In addition, the Pauli matrix basis makes it easier to include external electromagnetic fields related to spectroscopies.34–36 

We partition the quantum system into two quantum subsystems, the cluster and the environment, where it is assumed that the cluster is the region of interest. The charge density and magnetization density vector are partitioned as

γ0tot=γ0clu+γ0env,
(10)
mtot=mclu+menv,m=(γ1,γ2,γ3),
(11)

leading to a partition of the total energy as

Etotγ0tot,mtot=Ecluγ0clu,mclu+Eenvγ0env,menv+Eintγ0clu,mclu,γ0env,menv,
(12)

where Eint is the interaction energy. The interaction energy accounts for the non-additive contribution to the energy upon explicit cluster–environment interactions. The embedding perturbation matrices arising from the interaction energy term in Eq. (12) are

V=Vextenv+Jγ0totJγ0cluζKγ0totKγ0clu+Vxcγ0totVxcγ0clu+VTγ0totVTγ0clu,
(13)
B=ζKmtotKmclu+VxcmtotVxcmclu+VTmtotVTmclu,
(14)

where Vextenv is the external potential due to nuclei in the environment.

In Eq. (13), V′ is the matrix representation of the scalar embedding potential

V=δEintγ0clu,mclu,γ0env,menvδγ0clu,
(15)

and (VTγ0totVTγ0clu) is the matrix representation of the scalar potential due to the non-interacting electronic kinetic energy (TS),

(VTγ0totVTγ0clu)  =δTSγ0clu,mclu,γ0env,menvTSγ0clu,mcluTSγ0env,menvδγ0clu.
(16)

In Eq. (14), B is the matrix representation of the embedding magnetic field vector,

B=δEintγ0clu,mclu,γ0env,menvδmclu,
(17)

and VTmtotVTmclu is the matrix representation of the magnetic field due to the non-interacting electronic kinetic energy

VTmtotVTmclu  =δTSγ0clu,mclu,γ0env,menvTSγ0clu,mcluTSγ0env,menvδmclu.
(18)

The non-additive kinetic energy potential VT is difficult to compute analytically since it is challenging to obtain a good analytic model for the kinetic energy as a function of density. See Ref. 4 for more details regarding approximations of VT.

The partitioning above is, in principle, exact for a given level of theory. One can use the embedding perturbations in Eqs. (13) and (14), computed at the low level of theory, in a calculation on the cluster at the high level of theory

F=F0+Vσ0+n(Fn+μBBn)σn,
(19)

where Bn is the nth component of the magnetic field vector, μB is the Bohr magneton, and the embedding magnetic field is coupled to the high-level spin degrees of freedom via a spin-Zeeman interaction.7,36,38

We will focus on density functional theory (DFT) based embedding methods in this work.39–51 An embedding approach that avoids the non-additive kinetic energy contribution to Eq. (13) is the projection-based scheme.6,52 In projection-based embedding, a projection operator is introduced to enforce orthogonality between the orbitals of the cluster and environment. Thus, the non-additive kinetic energy contribution to Eqs. (13) and (14) is zero by construction,52 which gives rise to the following embedding perturbations:

VProj=Vextenv+Jγ0totJγ0cluζKγ0totKγ0clu+Vxcγ0totVxcγ0clu,
(20)
BnProj=ζKmtotKmclu+VxcmtotVxcmclu.
(21)

The resulting embedding two-component Kohn–Sham matrix in Pauli matrix representation for the high-level calculation is

FProj=F0+μPP0env+VProjσ0+nFn+μPPnenv+μBBnProjσn,
(22)

where

Pnenv=SγnenvS,n=0,1,2,3,
(23)

is the renormalized density for the environment subsystem. μP (usually about 106 a.u.) is a parameter for the projection term that elevates the energy of orbitals in the environment to maintain the wave function orthogonality between different subsystems.6,52

The separation of the total system into subsystems in projection-based quantum embedding is done by assigning the molecular orbitals of the total system to different subsystems. Note that schemes involving the choice of nuclei for projection-based embedding have been used to facilitate an automated choice of molecular orbitals.52 Projection-based embedding has been shown to be able to handle strongly bonded systems since the relevant bonding orbitals can be embedded as opposed to cutting across a bond.53–55 

As pointed out in our previous work,38 there is ambiguity in Eq. (19) in the absence of a preferred spin axis (e.g., in an external magnetic field) since the high- and low-level calculations may differ in spin orientation. In addition to the spin-dependent embedding magnetic field, the projection operator also depends upon the spin frame of the low-level calculation [see Eqs. (22) and (23)]. We propose using the low-level cluster density matrix as an initial guess for the embedded high-level cluster calculation. This strategy biases the high-level embedded calculation toward a solution with the same spin frame.

For projection-based embedding, the low- and high-level calculations were performed in a locally modified version of the Chronus Quantum software package.56 The triple-ζ Sapporo basis set with diffuse functions57 was used for W, the 6-311G(d) basis set58 was used for C, O, and F in the W-containing complexes, and 6-31G59–61 was used for [(H3N)4Cr(OH)2Cr(NH3)4]4+ in both high- and low-level calculations. A value of μP = 106 a.u. was used for the projection operator.

In the experimental photoelectron spectrum of W(CO)6, two peaks were observed with a 0.29 eV separation and a 2:1 ratio in intensity.62,63 These features can be understood at the orbital level,64 where the degenerate b2g and eg orbitals (of the t2g manifold) split into e′ and e upon inclusion of spin–orbit coupling. To correctly model the W(CO)6 spectrum, electronic structure methods that include spin–orbit effects should be used even though ground-state W(CO)6 is a closed-shell system.

When the fine-structure splitting of the high-level system is the subject of interest, the low-level method can be chosen to be nonrelativistic because the splitting is mainly due to the spin–orbit coupling localized at the high-level site. The photoelectron spectrum of W(CO)6 is used to benchmark the relativistic-in-nonrelativistic embedding scheme.

The experimental structure of W(CO)6 is used in this work.65 In projection-based embedding, the partitioning formally occurs with respect to molecular orbital. Since the molecular orbitals for W(CO)6 have well-defined symmetries, we did not perform any localization for the selection of projected orbital space. In Table I, we report several different ways that we partition the orbital space of W(CO)6. The embedding schemes were motivated by choosing the environment orbitals with little spatial overlap with the W atom. In embedding scheme A, we chose the core orbitals of C and O, scheme B adds CO σ orbitals, and C includes the CO π orbitals. See supplementary material for images of orbitals referenced in Table I.

TABLE I.

Various embedding schemes defined in terms of the nonrelativistic orbitals.

Environment orbitals
 A 4a1g,2eg,3t1u,3eg,5t1u,6a1g  
 B A+8t1u,5eg,8a1g,9t1u,11t1u  
 C B+10t1u,2t2u,1t1g  
Environment orbitals
 A 4a1g,2eg,3t1u,3eg,5t1u,6a1g  
 B A+8t1u,5eg,8a1g,9t1u,11t1u  
 C B+10t1u,2t2u,1t1g  

Table II presents the splitting of the highest-occupied t2g-derived orbitals of W(CO)6 as a function of electronic structure method and embedding scheme. In this benchmark study, the nonrelativistic one-component PBE0 exchange-correlation functional66 is used in the low level of W(CO)6. Because the splitting is due to spin–orbit coupling, when nonrelativistic PBE0 without spin–orbit is used for the high-level subsystem, it failed to predict the correct fine-structure splitting. In contrast, when the high-level subsystem is modeled with the relativistic X2C-PBE0 approach,30 the predicted splitting (0.22 eV) agrees well with the experimental value of 0.29 eV.63 We find that the splitting is relatively independent of the partition choice; removing 42% of the electrons from the relativistic X2C-PBE0 calculation yields an error less than 0.01 eV with respect to a calculation on the total system. Because the schemes primarily remove orbitals originating from the CO molecules, this result is reasonable since the spin–orbit is localized to the W atom.

TABLE II.

Splitting of the valence t2g-derived orbitals of W(CO)6 (in eV) for various embedding schemes (see Table I for definitions), compared to full cluster calculations using PBE0 and X2C-PBE0. Neclu is the number of electrons explicitly modeled with the high-level theory in the cluster.

MethodX2C-PBE0X2C-PBE0X2C-PBE0
schemePBE0X2C-PBE0ABCExpt.63 
Δt2g 0.00 0.22 0.22 0.22 0.22 0.29 
Neclu 158 158 134 110 92  
MethodX2C-PBE0X2C-PBE0X2C-PBE0
schemePBE0X2C-PBE0ABCExpt.63 
Δt2g 0.00 0.22 0.22 0.22 0.22 0.29 
Neclu 158 158 134 110 92  

The [(H3N)4Cr(OH)2Cr(NH3)4]4+ complex, denoted di-Cr(III), is an interesting test case due to the antiferromagnetic superexchange-coupled Cr(III) centers.36,67–69 In this second benchmark study, we compute the exchange-coupling constant of di-Cr(III) as

J=EAFMEFM12,
(24)

where EAFM is the energy of the antiferromagnetic state and EFM is the energy of the ferromagnetic state. Note that Eq. (24) assumes an isotropic, nonrelativistic spin Hamiltonian. In this benchmark system, we focus on testing if we can remove a significant amount of the electrons while maintaining the correct spin physics.

We used the computed geometry from Ref. 36 for di-Cr(III) and chose nonrelativistic BH&HLYP70 for the low-level calculation and X2C–BH&HLYP as the high-level calculation. For the partitioning, we removed 64 electrons primarily from NH3 orbitals (see the supplementary material for more details), leaving the unpaired electrons on each Cr(III) in the high-level calculation. Since the electrons in the environment are paired, the embedding magnetic field is zero, and scalar-potential embedding is used.

Table III lists the exchange-coupling constant of di-Cr(III) from an embedded X2C–BH&HLYP calculation, compared to those from the full X2C calculation. In comparing the X2C–BH&HLYP calculation on the total system with the embedded calculation, we find that we can remove 45% of the electrons from the system and introduce an error of only 0.2 cm−1 in J. This result suggests that the methodology presented can locally model couplings between multiple spin centers.

TABLE III.

Exchange coupling constant J of [(H3N)4Cr(OH)2Cr(NH3)4]4+ (in cm−1). Neclu is the number of electrons explicitly modeled with the high-level theory in the cluster.

MethodFull X2C–BH&HLYPEmbedded X2C–BH&HLYP
J −4.1 −3.9 
Neclu 142 78 
MethodFull X2C–BH&HLYPEmbedded X2C–BH&HLYP
J −4.1 −3.9 
Neclu 142 78 

In the third benchmark study, we investigate the dissociation energy (ΔE) and ground-state potential curve for the WF6 → WF5 + F reaction. Specifically, we considered singlet WF6 in an octahedral geometry dissociating to triplet WF5 + F, where the WF5 has a trigonal bipyrimidal geometry (see supplementary material for structures). In order to correctly predict the energetics, the low-level system is treated with the spin-free X2C-PBE0 approach in which only scalar-relativistic effects are included, resulting in a relativistic-in-scalar-relativistic embedding scheme.

In this study, equilibrium geometries of WF6 and WF5 were optimized using two-component PBE0 in the development version of Gaussian.71 CRENBL effective core potential72 was used for the core electrons and spin–orbit of W; double-ζ Sapporo basis set57 was used for the 5s, 5p, 5d, and 6p electrons of W, and def2-SVP basis set73 for F. In order to obtain the dissociation curve, we carried out relaxed potential energy scan in which all molecular degrees of freedom except the W–F bond are optimized.

Canonical orbitals will not be as well localized for the distorted WF6 geometries along the potential curve. Thus, we partitioned the system based on localized orbitals. We used Pipek–Mezey localization74 due to previous good performance for projection-based embedding.6 We consider two embedding schemes for the WF6 → WF5 + F reaction. For WF6, we projected out six π orbitals, and for WF5 + F, we projected out the dissociated F p orbitals and three π orbitals from the high-level subsystem (see the supplementary material images of orbitals). Note that the two schemes do not map to each other since the dissociated F pz orbital forms a σ bond that has a significant contribution from W basis functions. At the equilibrium bond length, the partitioning scheme is designed to keep the bonding electrons in the same subspace; at the long distance, one of the bonding electrons is projected out to create two product radicals. Since the unpaired electrons are in separate subsystems, both the embedding scalar potential and embedding magnetic field are included.

Table IV compares dissociation energies computed with spin-free PBE0 (SF-PBE0), X2C-PBE0, and embedded X2C-PBE0 in SF-PBE0. The dissociation energies were computed as the difference between the energy of WF5 20 Å away from a fluorine atom and the energy of WF6 at equilibrium. We include an experimental value for reference in Table IV, but a direct comparison between computed and experimental values is not recommended due to our neglect of thermal corrections and zero-point energy in the calculations. The addition of spin–orbit coupling lowers the dissociation energy from 5.23 eV to 5.11 eV in going from SF-PBE0 to X2C-PBE0. The embedded X2C-in-SF-PBE0 dissociation energy is in quantitative agreement (0.03 eV difference) with X2C-PBE0, indicating that projection-based embedding is an effective tool for including localized spin–orbit effects.

TABLE IV.

Dissociation energy (ΔE) of WF6 (in eV) for scalar-relativistic PBE0 (SF-PBE0), X2C-PBE0, and embedded X2C-PBE0 in SF-PBE0 (X2C-in-SF-PBE0).

SF-PBE0X2C-PBE0X2C-in-SF-PBE0Expt.75 
ΔE 5.23 5.11 5.14 5.25 
SF-PBE0X2C-PBE0X2C-in-SF-PBE0Expt.75 
ΔE 5.23 5.11 5.14 5.25 

Figure 1 shows the potential energy curve corresponding to the dissociation of WF6 computed using a relaxed potential energy scan. During the singlet region of the potential curve, the bonding electrons are kept in the same subspace. Starting from 3.00 Å, one of the bonding electrons is projected out in order to simulate the homogeneous dissociation path. Inevitably, this creates an inconsistent treatment throughout the potential energy curve scan, but it offers an opportunity to analyze the importance of embedding magnetic field in open-shell systems. We find a good agreement between X2C-in-SF-PBE0 and X2C-PBE0 except in the 2.75 Å–3.25 Å region. This discrepancy is due to the presence of a singlet/triplet crossing and a choice of partitioned MOs (see the supplementary material) that does not continuously map between the singlet and triplet. The former problem can be addressed by moving to an electronic structure method capable of a balanced treatment of ground and excited states, and the latter problem is an active area of recent research.76–79 

FIG. 1.

Relaxed potential curve of WF6 with (blue) X2C-PBE0 and (red) X2C-in-SF-PBE0 embedding, compared to (black) X2C-in-SF-PBE0 embedding with scalar-potential embedding only.

FIG. 1.

Relaxed potential curve of WF6 with (blue) X2C-PBE0 and (red) X2C-in-SF-PBE0 embedding, compared to (black) X2C-in-SF-PBE0 embedding with scalar-potential embedding only.

Close modal

In order to analyze the importance of the magnetic embedding component, we also report in Fig. 1 the WF6 potential curve computed using X2C-in-SF-PBE0 without the embedding magnetic field. Owing to the closed-shell nature in the bonding region, the embedding magnetic field is zero in the range of 1.75 Å–2.75 Å. In the intermediate interaction distance (2.75 Å–3.25 Å), the embedding magnetic field plays an important role of introducing as much as 0.2 eV interaction energy to the system compared to those computed with scalar-potential embedding only. As the interaction distance increases, the magnetic embedding effects decrease quickly and the potential energy curves computed with and without embedding magnetic fields are indistinguishable.

We introduced a relativistic quantum embedding formalism capable of variational treatments of relativistic effects, including scalar relativity and spin–orbit coupling. Projection-based embedding was extended to a relativistic two-component formalism for the first time, where the full spin magnetization vector is explicitly modeled. Two projection-based relativistic embedding protocols, relativistic-in-nonrelativistic and relativistic-in-scalar-relativistic schemes, were benchmarked and applied to studies of molecular complexes. The spin–orbit splitting for the valence t2g-derived orbitals of W(CO)6 was computed using ground-state relativistic Kohn–Sham density functional theory and the result agrees quantitatively with experimental photoelectron spectroscopy. Using embedding, we can remove almost half of the electrons of the W(CO)6 complex from the high-level treatment and introduce an error less than 0.01 eV with respect to a calculation on the total system. We also applied the relativistic-in-nonrelativistic embedding scheme to the exchange coupling constant of [(H3N)4Cr(OH)2Cr(NH3)4]4+ and found that almost half of the electrons can be removed and only introduce an error of 0.2 cm−1, indicating that local treatments of complex interactions between spin centers can be achieved. The dissociation energy and dissociation potential curve of WF6 were studied with the relativistic-in-scalar-relativistic scheme. We find a quantitative agreement of 0.03 eV in the dissociation energy with a relativistic calculation on the total system and a potential curve that is well behaved in all regions except the singlet-triplet crossing. The embedding presented in this work shows promise for treatments of heavy-element derived spin phenomena since it can accurately and efficiently include spin–orbit coupling in systems with multiple spin centers.

Images of the projected orbitals for W(CO)6, [(H3N)4Cr(OH)2Cr(NH3)4]4+, and WF6 can be found in the supplementary material along with the geometries of WF6 and WF5.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

This work was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award No. LAB 17-1775, as part of the Computational Chemical Sciences Program. The development of the two-component method is funded by the U.S. Department of Energy (Grant No. DE-SC0006863). The development of the Chronus Quantum open source software package was supported by the National Science Foundation (Grant No. OAC-1663636). Computations were facilitated through the use of advanced computational, storage, and networking infrastructure provided by the Hyak supercomputer system at the University of Washington, funded by the Student Technology Fee.

1.
A. S. P.
Gomes
and
C. R.
Jacob
, “
Quantum-chemical embedding methods for treating local electronic excitations in complex chemical systems
,”
Annu. Rep. Prog. Chem., Sect. C: Phys. Chem.
108
,
222
277
(
2012
).
2.
F.
Libisch
,
C.
Huang
, and
E. A.
Carter
, “
Embedded correlated wavefunction schemes: Theory and applications
,”
Acc. Chem. Res.
47
,
2768
2775
(
2014
).
3.
C. R.
Jacob
and
J.
Neugebauer
, “
Subsystem density-functional theory
,”
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
4
,
325
362
(
2014
).
4.
T. A.
Wesolowski
,
S.
Shedge
, and
X.
Zhou
, “
Frozen-density embedding strategy for multilevel simulations of electronic structure
,”
Chem. Rev.
115
,
5891
5928
(
2015
).
5.
Q.
Sun
and
G. K.-L.
Chan
, “
Quantum embedding theories
,”
Acc. Chem. Res.
49
,
2705
2712
(
2016
).
6.
S. J. R.
Lee
,
M.
Welborn
,
F. R.
Manby
, and
T. F.
Miller
 III
, “
Projection-based wavefunction-in-DFT embedding
,”
Acc. Chem. Res.
52
,
1359
1368
(
2019
).
7.
K. G.
Dyall
and
K.
Fægri
,
Introduction to Relativistic Quantum Chemistry
(
Oxford University Press on Demand
,
2007
).
8.
M.
Reiher
and
A.
Wolf
,
Relativistic Quantum Chemistry: The Fundamental Theory of Molecular Science
(
John Wiley & Sons
,
2014
).
9.
W.
Liu
,
Handbook of Relativistic Quantum Chemistry
(
Springer-Verlag Berlin Heidelberg
,
2017
).
10.
S.
Crampin
, “
An embedding scheme for the Dirac equation
,”
J. Phys.: Condens. Matter
16
,
8875
(
2004
).
11.
M.
James
and
S.
Crampin
, “
Relativistic embedding method: The transfer matrix, complex band structures, transport, and surface calculations
,”
Phys. Rev. B
81
,
155439
(
2010
).
12.
A. S. P.
Gomes
,
C. R.
Jacob
, and
L.
Visscher
, “
Calculation of local excitations in large systems by embedding wave-function theory in density-functional theory
,”
Phys. Chem. Chem. Phys.
10
,
5353
(
2008
).
13.
A. S. P.
Gomes
,
C. R.
Jacob
,
F.
Réal
,
L.
Visscher
, and
V.
Vallet
, “
Towards systematically improvable models for actinides in condensed phase: The electronic spectrum of uranyl in Cs2UO2Cl4 as a test case
,”
Phys. Chem. Chem. Phys.
15
,
15153
(
2013
).
14.
A. W.
Götz
,
J.
Autschbach
, and
L.
Visscher
, “
Calculation of nuclear spin-spin coupling constants using frozen density embedding
,”
J. Chem. Phys.
140
,
104107
(
2014
).
15.
M.
Olejniczak
,
R.
Bast
, and
A. S. P.
Gomes
, “
On the calculation of second-order magnetic properties using subsystem approaches in a relativistic framework
,”
Phys. Chem. Chem. Phys.
19
,
8400
8415
(
2017
).
16.
W.
Kutzlenigg
and
W.
Liu
, “
Quasirelativistic theory equivalent to fully relativistic theory
,”
J. Chem. Phys.
123
,
241102
(
2005
).
17.
W.
Liu
and
D.
Peng
, “
Infinite-order quasirelativistic density functional method based on the exact matrix quasirelativistic theory
,”
J. Chem. Phys.
125
,
044102
(
2006
).
18.
D.
Peng
,
W.
Liu
,
Y.
Xiao
, and
L.
Cheng
, “
Making four- and two-component relativistic density functional methods fully equivalent based on the idea of “from atoms to molecule”
,”
J. Chem. Phys.
127
,
104106
(
2007
).
19.
M.
Ilias
and
T.
Saue
, “
An infinite-order relativistic Hamiltonian by a simple one-step transformation
,”
J. Chem. Phys.
126
,
064102
(
2007
).
20.
W.
Liu
and
D.
Peng
, “
Exact two-component Hamiltonians revisited
,”
J. Chem. Phys.
131
,
031104
(
2009
).
21.
W.
Liu
, “
Ideas of relativistic quantum chemistry
,”
Mol. Phys.
108
,
1679
1706
(
2010
).
22.
T.
Saue
, “
Relativistic Hamiltonians for chemistry: A primer
,”
ChemPhysChem
12
,
3077
3094
(
2011
).
23.
Z.
Li
,
Y.
Xiao
, and
W.
Liu
, “
On the spin separation of algebraic two-component relativistic Hamiltonians
,”
J. Chem. Phys.
137
,
154114
(
2012
).
24.
D.
Peng
,
N.
Middendorf
,
F.
Weigend
, and
M.
Reiher
, “
An efficient implementation of two-component relativistic exact-decoupling methods for large molecules
,”
J. Chem. Phys.
138
,
184105
(
2013
).
25.
P. K.
Tamukong
,
M. R.
Hoffmann
,
Z.
Li
, and
W.
Liu
, “
Relativistic GVVPT2 multireference perturbation theory description of the electronic states of Y2 and Tc2
,”
J. Phys. Chem. A
118
,
1489
1501
(
2014
).
26.
F.
Egidi
,
J. J.
Goings
,
M. J.
Frisch
, and
X.
Li
, “
Direct atomic-orbital-based relativistic two-component linear response method for calculating excited-state fine structures
,”
J. Chem. Theory Comput.
12
,
3711
3718
(
2016
).
27.
J. J.
Goings
,
J. M.
Kasper
,
F.
Egidi
,
S.
Sun
, and
X.
Li
, “
Real time propagation of the exact two component time-dependent density functional theory
,”
J. Chem. Phys.
145
,
104107
(
2016
).
28.
L.
Konecny
,
M.
Kadek
,
S.
Komorovsky
,
O. L.
Malkina
,
K.
Ruud
, and
M.
Repisky
, “
Acceleration of relativistic electron dynamics by means of X2C transformation: Application to the calculation of nonlinear optical properties
,”
J. Chem. Theory Comput.
12
,
5823
5833
(
2016
).
29.
F.
Egidi
,
S.
Sun
,
J. J.
Goings
,
G.
Scalmani
,
M. J.
Frisch
, and
X.
Li
, “
Two-component non-collinear time-dependent spin density functional theory for excited state calculations
,”
J. Chem. Theory Comput.
13
,
2591
2603
(
2017
).
30.
A.
Petrone
,
D. B.
Williams-Young
,
S.
Sun
,
T. F.
Stetina
, and
X.
Li
, “
An efficient implementation of two-component relativistic density functional theory with torque-free auxiliary variables
,”
Eur. Phys. J. B
91
,
169
(
2018
).
31.
W.
Liu
and
Y.
Xiao
, “
Relativistic time-dependent density functional theories
,”
Chem. Soc. Rev.
47
,
4481
4509
(
2018
).
32.
J. C.
Boettger
, “
Approximate two-electron spin-orbit coupling term for density-functional-theory DFT calculations using the Douglas-Kroll-Hess transformation
,”
Phys. Rev. B
62
,
7809
7815
(
2000
).
33.
F.
Ding
,
J. J.
Goings
,
M. J.
Frisch
, and
X.
Li
, “
Ab initio non-relativistic spin dynamics
,”
J. Chem. Phys.
141
,
214111
(
2014
).
34.
T. F.
Stetina
,
S.
Sun
,
D. B.
Williams-Young
, and
X.
Li
, “
Modeling magneto-photoabsorption using time-dependent complex generalized Hartree-Fock
,”
ChemPhotoChem
3
,
739
746
(
2019
).
35.
S.
Sun
,
D.
Williams-Young
, and
X.
Li
, “
An ab initio linear response method for computing magnetic circular dichroism spectra with nonperturbative treatment of magnetic field
,”
J. Chem. Theory Comput.
15
,
3162
3169
(
2019
).
36.
S.
Sun
,
D. B.
Williams-Young
,
T. F.
Stetina
, and
X.
Li
, “
Generalized Hartree-Fock with non-perturbative treatment of strong magnetic field: Application to molecular spin phase transition
,”
J. Chem. Theory Comput.
15
,
348
356
(
2019
).
37.
J. J.
Goings
,
F.
Egidi
, and
X.
Li
, “
Current development of non-collinear electronic structure theory
,”
Int. J. Quantum Chem.
118
,
e25398
(
2018
).
38.
C. E.
Hoyer
,
D. B.
Williams-Young
,
C.
Huang
, and
X.
Li
, “
Embedding non-collinear two-component electronic structure in a collinear quantum environment
,”
J. Chem. Phys.
150
,
174114
(
2019
).
39.
G.
Senatore
and
K. R.
Subbaswamy
, “
Density dependence of the dielectric constant of rare-gas crystals
,”
Phys. Rev. B
34
,
5754
(
1986
).
40.
P.
Cortona
, “
Self-consistently determined properties of solids without band-structure calculations
,”
Phys. Rev. B
44
,
8454
(
1991
).
41.
T. A.
Wesolowski
and
A.
Warshel
, “
Frozen density functional approach for ab initio calculations of solvated molecules
,”
J. Phys. Chem.
97
,
8050
8053
(
1993
).
42.
N.
Govind
,
Y. A.
Wang
,
A. J. R.
Da Silva
, and
E. A.
Carter
, “
Accurate ab initio energetics of extended systems via explicit correlation embedded in a density functional environment
,”
Chem. Phys. Lett.
295
,
129
134
(
1998
).
43.
J.
Neugebauer
,
M. J.
Louwerse
,
E. J.
Baerends
, and
T. A.
Wesolowski
, “
The merits of the frozen-density embedding scheme to model solvatochromic shifts
,”
J. Chem. Phys.
122
,
094115
(
2005
).
44.
C. R.
Jacob
and
L.
Visscher
, “
Calculation of nuclear magnetic resonance shieldings using frozen-density embedding
,”
J. Chem. Phys.
125
,
194104
(
2006
).
45.
M. H.
Cohen
and
A.
Wasserman
, “
On hardness and electronegativity equalization in chemical reactivity theory
,”
J. Chem. Phys.
125
,
001121
1139
(
2006
).
46.
O.
Roncero
,
M. P.
de Lara-Castells
,
P.
Villarreal
,
F.
Flores
,
J.
Ortega
,
M.
Paniagua
, and
A.
Aguado
, “
An inversion technique for the calculation of embedding potentials
,”
J. Chem. Phys.
129
,
184104
(
2008
).
47.
J. D.
Goodpaster
,
N.
Ananth
,
F. R.
Manby
, and
T. F.
Miller
, “
Exact nonadditive kinetic potentials for embedded density functional theory
,”
J. Chem. Phys.
133
,
084103
(
2010
).
48.
S.
Laricchia
,
E.
Fabiano
, and
F.
Della Sala
, “
Frozen density embedding with hybrid functionals
,”
J. Chem. Phys.
133
,
164111
(
2010
).
49.
S.
Fux
,
C. R.
Jacob
,
J.
Neugebauer
,
L.
Visscher
, and
M.
Reiher
, “
Accurate frozen-density embedding potentials as a first step towards a subsystem description of covalent bonds
,”
J. Chem. Phys.
132
,
164101
(
2010
).
50.
M.
Pavanello
and
J.
Neugebauer
, “
Modelling charge transfer reactions with the frozen density embedding formalism
,”
J. Chem. Phys.
135
,
234103
(
2011
).
51.
B.
Senjean
,
M.
Tsuchiizu
,
V.
Robert
, and
E.
Fromager
, “
Local density approximation in site-occupation embedding theory
,”
Mol. Phys.
115
,
48
62
(
2016
).
52.
F. R.
Manby
,
M.
Stella
,
J. D.
Goodpaster
, and
T. F.
Miller
 III
, “
A simple, exact density-functional-theory embedding scheme
,”
J. Chem. Theory Comput.
8
,
2564
2568
(
2012
).
53.
T. A.
Barnes
,
J. D.
Goodpaster
,
F. R.
Manby
, and
T. F.
Miller
 III
, “
Accurate basis set truncation for wavefunction embedding
,”
J. Chem. Phys.
139
,
024103
(
2013
).
54.
D. V.
Chulhai
and
J. D.
Goodpaster
, “
Improved accuracy and efficiency in quantum embedding through absolute localization
,”
J. Chem. Theory Comput.
13
,
1503
1508
(
2017
).
55.
B.
Hégely
,
P. R.
Nagy
, and
M.
Kállay
, “
Dual basis set approach for density functional and wave function embedding schemes
,”
J. Chem. Theory Comput.
14
,
4600
4615
(
2018
).
56.
D. B.
Williams-Young
,
A.
Petrone
,
S.
Sun
,
T. F.
Stetina
,
P.
Lestrange
,
C. E.
Hoyer
,
D. R.
Nascimento
,
L.
Koulias
,
A.
Wildman
,
J.
Kasper
,
J. J.
Goings
,
F.
Ding
,
A. E.
DePrince
 III
,
E. F.
Valeev
, and
X.
Li
, “
The Chronus quantum (ChronusQ) software package
,”
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
10
,
e1436
(
2019
).
57.
T.
Noro
,
M.
Sekiya
, and
T.
Koga
, “
Sapporo-(DKH3)-nZP (n = D, T, Q) sets for the sixth period s-, d-, and p-block atoms
,”
Theor. Chem. Acc.
132
,
1363
(
2013
).
58.
R.
Krishnan
,
J. S.
Binkley
,
R.
Seeger
, and
J. A.
Pople
, “
Self-consistent molecular orbital methods. XX. A basis set for correlated wave functions
,”
J. Chem. Phys.
72
,
650
654
(
1980
).
59.
R.
Ditchfield
,
W. J.
Hehre
, and
J. A.
Pople
, “
Self-consistent molecular-orbital methods. IX. An extended Gaussian-type basis for molecular-orbital studies of organic molecules
,”
J. Chem. Phys.
54
,
724
728
(
1971
).
60.
W. J.
Hehre
,
R.
Ditchfield
, and
J. A.
Pople
, “
Self-consistent molecular orbital methods. XII. Further extensions of Gaussian-type basis sets for use in molecular orbital studies of organic molecules
,”
J. Chem. Phys.
56
,
2257
2261
(
1972
).
61.
V. A.
Rassolov
,
J. A.
Pople
,
M. A.
Ratner
, and
T. L.
Windus
, “
6-31G* basis set for atoms K through Zn
,”
J. Chem. Phys.
109
,
1223
1229
(
1998
).
62.
G. M.
Bancroft
,
L.
Dignard-Bailey
, and
R. J.
Puddephatt
, “
Photoelectron study of additivity and ligand field effects on the tungsten 5d orbitals in WCO6nPR3n compounds
,”
Inorg. Chem.
23
,
2369
2370
(
1984
).
63.
J.
Wu
,
G. M.
Bancroft
,
R. J.
Puddephatt
,
Y. F.
Hu
,
X.
Li
, and
K. H.
Tan
, “
Photoelectron spectra of trimethylphosphine-substituted tungsten carbonyls: Ligand field effects, ligand additivity effects, and core-valence ionization correlations
,”
Inorg. Chem.
38
,
4688
4695
(
1999
).
64.
M. B.
Hall
, “
Use of spin-orbit coupling in the interpretation of photoelectron spectra. I. Application to substituted rhenium pentacarbonyls
,”
J. Am. Chem. Soc.
97
,
2057
2065
(
1975
).
65.
S.
Arnesen
and
H.
Seip
, “
Studies on failure of first born approximation in electron diffraction. V. Molybdenum and tungsten hexacarbonyl
,”
Acta Chem. Scand.
20
,
2711
(
1966
).
66.
C.
Adamo
and
V.
Barone
, “
Toward reliable density functional methods without adjustable parameters: The PBE0 model
,”
J. Chem. Phys.
110
,
6158
(
1999
).
67.
J. B.
Goodenough
,
Magnetism and the Chemical Bond
(
Interscience Publishers
,
1963
), Vol. 1.
68.
S.
Decurtins
and
H. U.
Guedel
, “
Superexchange in bis-(μ-hydroxo)-bridged chromium (III) dimers. Optical spectroscopic study of singly excited 2Eg4A2g pair states
,”
Inorg. Chem.
21
,
3598
3606
(
1982
).
69.
S.
Decurtins
,
H. U.
Guedel
, and
A.
Pfeuti
, “
Optical spectroscopy of [(NH3)4Cr(OH)2Cr(NH3)4]4+ and [(en)2Cr(OH)2Cr(en)2]4+
,”
Inorg. Chem.
21
,
1101
1106
(
1982
).
70.
M.
Caricato
,
G. W.
Trucks
,
M. J.
Frisch
, and
K. B.
Wiberg
, “
Electronic transition energies: A study of the performance of a large range of single reference density functional and wave function methods on valence and Rydberg states compared to experiment
,”
J. Chem. Theory Comput.
6
,
370
383
(
2010
).
71.
M. J.
Frisch
,
G. W.
Trucks
,
H. B.
Schlegel
,
G. E.
Scuseria
,
M. A.
Robb
,
J. R.
Cheeseman
,
G.
Scalmani
,
V.
Barone
,
G. A.
Petersson
,
H.
Nakatsuji
,
X.
Li
,
M.
Caricato
,
A. V.
Marenich
,
J.
Bloino
,
B. G.
Janesko
,
R.
Gomperts
,
B.
Mennucci
,
H. P.
Hratchian
,
J. V.
Ortiz
,
A. F.
Izmaylov
,
J. L.
Sonnenberg
,
D.
Williams-Young
,
F.
Ding
,
F.
Lipparini
,
F.
Egidi
,
J.
Goings
,
B.
Peng
,
A.
Petrone
,
T.
Henderson
,
D.
Ranasinghe
,
V. G.
Zakrzewski
,
J.
Gao
,
N.
Rega
,
G.
Zheng
,
W.
Liang
,
M.
Hada
,
M.
Ehara
,
K.
Toyota
,
R.
Fukuda
,
J.
Hasegawa
,
M.
Ishida
,
T.
Nakajima
,
Y.
Honda
,
O.
Kitao
,
H.
Nakai
,
T.
Vreven
,
K.
Throssell
,
J. A.
Montgomery
, Jr.
,
J. E.
Peralta
,
F.
Ogliaro
,
M. J.
Bearpark
,
J. J.
Heyd
,
E. N.
Brothers
,
K. N.
Kudin
,
V. N.
Staroverov
,
T. A.
Keith
,
R.
Kobayashi
,
J.
Normand
,
K.
Raghavachari
,
A. P.
Rendell
,
J. C.
Burant
,
S. S.
Iyengar
,
J.
Tomasi
,
M.
Cossi
,
J. M.
Millam
,
M.
Klene
,
C.
Adamo
,
R.
Cammi
,
J. W.
Ochterski
,
R. L.
Martin
,
K.
Morokuma
,
O.
Farkas
,
J. B.
Foresman
, and
D. J.
Fox
, Gaussian Development Version Revision I.11+,
2016
.
72.
R. B.
Ross
,
J. M.
Powers
,
T.
Atashroo
,
W. C.
Ermler
,
L. A.
LaJohn
, and
P. A.
Christiansen
, “
Ab initio relativistic effective potentials with spin–orbit operators. IV. Cs through Rn
,”
J. Chem. Phys.
93
,
6654
6670
(
1990
).
73.
F.
Weigend
and
R.
Ahlrichs
, “
Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy
,”
Phys. Chem. Chem. Phys.
7
,
3297
3305
(
2005
).
74.
J.
Pipek
and
P. G.
Mezey
, “
A fast intrinsic localization procedure applicable for ab initio and semiempirical linear combination of atomic orbital wave functions
,”
J. Chem. Phys.
90
,
4916
4926
(
1989
).
75.
K. G.
Dyall
, “
Bond dissociation energies of the tungsten fluorides and their singly charged ions: A density functional survey
,”
J. Phys. Chem. A
104
,
4077
4083
(
2000
).
76.
J. D.
Goodpaster
,
T. A.
Barnes
,
F. R.
Manby
, and
T. F.
Miller
 III
, “
Accurate and systematically improvable density functional theory embedding for correlated wavefunctions
,”
J. Chem. Phys.
140
,
18A507
(
2014
).
77.
M.
Welborn
,
F. R.
Manby
, and
T. F.
Miller
 III
, “
Even-handed subsystem selection in projection-based embedding
,”
J. Chem. Phys.
149
,
144101
(
2018
).
78.
M.
Bensberg
and
J.
Neugebauer
, “
Direct orbital selection for projection-based embedding
,”
J. Chem. Phys.
150
,
214106
(
2019
).
79.
D.
Claudino
and
N. J.
Mayhall
, “
Automatic partition of orbital spaces based on singular value decomposition in the context of embedding theories
,”
J. Chem. Theory Comput.
15
,
1053
1064
(
2019
).

Supplementary Material