Supporting Information
Spin-orbit torque in single-molecule junctions from ab initio
María Camarasa-Gómez,1, 2 Daniel Hernangómez-Pérez,1, 3 and Ferdinand Evers1
1
Institute of Theoretical Physics, University of Regensburg, 93040 Regensburg, Germany
2
Department of Molecular Chemistry and Materials Science,
Weizmann Institute of Science, Rehovot 7610001, Israel
3
CIC nanoGUNE BRTA, Tolosa Hiribidea 76, 20018 San Sebastián, Spain
Keywords: spin-orbit torque; density functional theory; out-of-equilibrium; single-molecule junctions
Email: [email protected]
[email protected]
CONTENTS
I. Computational details 2
II. Out-of-equilibrium density matrix with spin-orbit coupling 3
III. Relation with the equilibrium density matrix 6
IV. Self-consistent DFT-NEGF cycle with spin-orbit coupling 7
V. Observables 9
VI. Additional results 11
A. Additional figures 11
B. Additional tables 13
C. Junction geometry 13
References 15
1
I. COMPUTATIONAL DETAILS
Geometry. The geometry for the copper–vanadocene molecular junction was obtained
following a well-established two-step process. In the first step, we optimize the coordinates
of the molecular atoms, and the molecular atoms with the diamine anchor groups, employing
the trust-radius enhanced version of the Broyden-Fletcher-Shanno-Goldfarb algorithm1 . In
a second step, the copper atoms are added and the geometry optimized to obtain the correct
contact angle between the molecule and the pyramidal copper tip. Each copper pyramidal
cluster used in this work has 11 atoms and was cut in the (111) direction. The geometries
are considered to be structurally relaxed once the force components per atoms are smaller
than the threshold value 10−2 eV/Å.
DFT calculations. For our spin-orbit torque calculations, we employ the FHI-aims
package1 . As exchange-correlation approximation we employ the non-empirical Perdew-
Burke-Ernzerhof (PBE) exchange-correlation functional2 . For simplicity, we avoid the
use of hybrid functionals due to inherent problems in describing metallic properties of
nanostructures3 . We include as well scalar relativistic corrections to the kinetic energy
employing the atomic zeroth-order regular approximation4 . The Kohn-Sham states5 are
represented in an all-electron numeric atom-center basis set (see also Sec. II in the Support-
ing Information). For computational efficiency, we consider the optimized ‘light’ pre-defined
settings. Spin-orbit (SO) coupling is considered as implemented in the FHI-aims code. This
code provides the so-called non-self-consistent second-variational method6 , in which, the
SO coupling is reintroduced (at first order in (E − v̂)/c2 , with v̂ the scalar potential oper-
ator) after a regular SCF calculation with scalar relativistic corrections. The Kohn-Sham
Hamiltonian with SO coupling is then rebuilt explicitly and further diagonalized to obtain
SO-corrected eigenenergies and eigenstates.
For the density functional theory7,8 (DFT) calculations, we obtain the ground state of
the system using as convergence criteria for the self-consistent field cycle 10−6 electrons/Å3
for the particle density, 10−4 eV for the Kohn-Sham eigenvalue sum, and 10−7 eV for the
total energy.
Transport code. In the self-consistent field cycle shown in the main text (fur-
ther details in Sec. II), we have interfaced the density functional theory package with
2
the quantum transport module AITRANSS9,10 , where non-equilibrium Green’s function
formalism11,12 (NEGF) is implemented. This package has been extended to incorporate
spin-orbit interaction13 .
II. OUT-OF-EQUILIBRIUM DENSITY MATRIX WITH SPIN-ORBIT COUPLING
General definitions. We provide here technical details on the generalization of the
non-equilibrium density matrix for finite systems with absorbing boundary conditions9,10,14 .
The spinless case was originally presented in Ref. 10 and it is extended here to non-collinear
systems with spin-orbit coupling. The non-equilibrium density matrix is defined within the
NEGF from the energy integration of the lesser Green’s function, Ĝ< := Ĝ< (E),
Z +∞
1
ρ̂ = dE Ĝ< , (S1)
2πi −∞
which in turn can be computed from the Keldysh equation, Ĝ< = ĜΣ̂< Ĝ† . Here, Ĝ := Ĝ(E)
is the retarded Green’s function of the extended molecule
Ĝ = (E11 − ĤKS − Σ̂)−1 , (S2)
with 11 the identity operator, ĤKS the Kohn-Sham Hamiltonian and Σ̂ := Σ̂(E) = Σ̂L (E) +
Σ̂R (E) the self-energy operator of the left (L) and right (R) leads. The operator Σ̂< :=
Σ̂< (E) represents the lesser self-energy and is given by
h i
Σ̂< = i f (E − µL )Γ̂L (E) + f (E − µR )Γ̂R (E) , (S3)
with f (E − µα ) the Fermi-Dirac distribution function of each reservoir, characterized by
chemical potential µα , and Γ̂ := Γ̂(E) is the anti-Hermitian part of the self-energy
h i
†
Γ̂α = i Σ̂α − Σ̂α , (S4)
with α ∈ {L, R}.
3
Basis set and reconstructed Kohn-Sham Hamiltonian. The Kohn-Sham equa-
tions are solved within the FHI-aims package using a numeric atom-center basis set, φj (r) =
⟨r|φj ⟩ := ⟨r|j⟩. Each Kohn-Sham spinor ψµ (r) = ⟨r|ψµ ⟩ := ⟨r|µ⟩, where µ = 1, · · · Nstates
with Nstates = 2N and N being the number of spin-resolved orbitals, can be expanded as
follows
↑
ψµ (r) XN
c↑
ψµ (r) = = µj φj (r), (S5)
ψµ↓ (r) j=1 c↓µj
where cσµj are the spin-projected Kohn-Sham coefficients. In general, the Kohn-Sham coef-
ficients can be recast into a complex quadratic matrix of dimension 2N × 2N , so that Eq.
(S5) reads
2N
X
ψµ (r) = Bνµ φν (r). (S6)
ν=1
Whenever the set of atom-center orbitals is non-orthogonal, it can be orthogonalized by
a Löwdin orthogonalization procedure15 using the real-valued symmetric overlap matrix
⟨φν |φν ′ ⟩ if ν, ν ′ ≤ N and σ = σ ′ =↑,
Sνν ′ = ⟨ν|ν ′ ⟩ = (S7)
⟨φν−N |φν ′ −N ⟩ if ν, ν ′ ≥ N and σ = σ ′ =↓,
so that the orthogonal basis relates to the non-orthogonal one by
2N
−1/2
X
|ν̃⟩ = Sν ′ ,ν |ν ′ ⟩. (S8)
ν ′ =1
The Kohn-Sham Hamiltonian can therefore be reconstructed from the set of Kohn-Sham
spinors and energies, ϵµ in a straightforward way
2N
X 2N X
X 2N
H KS
= |ψµ ⟩ϵµ ⟨ψµ | = |ν⟩Bνµ ϵµ [B ∗ ]µν ′ ⟨ν ′ |,
µ=1 µ=1 ν,ν ′ =1
2N
X 2N
X 2N
X 1/2 1/2
= |ν̃⟩Sν̃ν Bνµ ϵµ [B ∗ ]µν ′ Sν ′ ν̃ ′ ⟨ν̃ ′ |. (S9)
µ=1 ν,ν ′ =1 ν̃,ν̃ ′ =1
which can be recast in a matrix multiplication form, ĤKS = S 1/2 BϵB † S 1/2 , where ϵ is the
Kohn-Sham energy (diagonal) matrix.
4
Markovian out-of-equilibrium density matrix. For simplicity, we assumed in this
paper that the leads are closed shell and non-polarized with vanishing spin-orbit inter-
action. Therefore, the self-energy is block-diagonal in spin space and spin independent,
Σ̂σα = Σ̂σ̄α ≡ Σ̂α , with σ = {↑, ↓} and σ̄ corresponding to flip σ. The effect of the spin-orbit
interaction in the Green’s function will thus come from the off-diagonal blocks of the Kohn-
Sham Hamiltonian labeled by the spin indices σ, σ̄. Taking advantage of the Markovian ap-
proximation considered in the absorbing boundary condition scheme for the self-energy9,10,14 ,
we perform a rotation of the Green’s function to the basis that diagonalizes the complex
matrix ĤΣ = ĤKS + 112 ⊗ Σ̂. Because this matrix is non-Hermitian, we choose to work with
the right set of eigenvalues and eigenvectors ĤΣ B = BZ where Z = diag(Z1 , . . . , Z2N ), with
Zµ = ReZµ + iImZµ , is the eigenvalue matrix and B the matrix that contains the right
eigenvectors in each column. In this basis, the Green’s function and the matrix blocks in
spin-space are given by
′
Ĝσσ = B σ Ĝ([B −T ]σ )T , (S10)
with B −T = (B −1 )T . Here, Ĝ := Ĝ(E) = [E11 − Z]−1 and B σ corresponds to a N × 2N
matrix extracted from B by projecting the eigenvectors into the spin subspace labeled by
σ. Using Eq. (S10), the lesser Green’s function reads
′ ′
X
G<,σ,σ = B σ Ĝ([B −T ]σ1 )T Σ̂<,σ1 [(B −1 )† ]σ1 Ĝ ∗ (B σ )† . (S11)
σ1
We substitute Eq. (S11) into Eq. (S1) and use Eq. (S3) in the zero-temperature limit, in
which the Fermi-Dirac distribution reduces to the Heaviside step function, f (E − µα ) →
θ(E − µα ) to find Z µα
σσ ′
X 1 ˆ ′
ρ = B σ
dE Ĝ Γ̃α Ĝ (B σ )† ,
∗
(S12)
α
2π −∞
where we have defined Γ̃ˆ := P ([B −T ]σ )T Γ̂σ ([B −1 ]† )σ . We can recast this expression into
α σ α
′ ′
ˆ σ )† where the matrix Jˆ := P Jˆα is given by
a familiar form by writing ρσσ = B σ J(B α
Z µα
1 ˆ Ĝ ∗ (E).
Jˆα = dE Ĝ(E)Γ̃α (S13)
2π −∞
5
This integral has been evaluated analytically using contour integration10 , where its matrix
elements are given by (Jα )µν = (Γ̃α )µν Fµν (µα ) with µ, ν = 1, . . . , 2N ,
ϵ2µ + ηµ2
1 1 h 1
Fµν (µ) = − 2πi + ln
2π (ϵµ − ϵν ) + i(ηµ − ην ) 2 ϵ2ν + ην2
i
ηµ ην
− iarctan − iarctan , (S14)
ϵµ ϵν
and ϵµ = µ − ReZµ and ηµ = ImZµ . The matrix Jˆ is simply the non-equilibrium generaliza-
tion of the occupation numbers at equilibrium, as shown in Sec. III.
III. RELATION WITH THE EQUILIBRIUM DENSITY MATRIX
The general form of the density operator of a quantum system defined in a finite-
dimensional Hilbert space is given by16,17
X
ρ̂ = fµ |ψµ ⟩ ⟨ψµ | , (S15)
µ
where the (fractional) occupations are given by 0 ≤ fµ ≤ 1 and the sum runs over all the
available states, µ = 1, . . . , Nstates . At equilibrium, the quantum state |ψµ ⟩ is a Kohn-Sham
state, which in the presence of spin-orbit interaction is represented as a two-component
spinor
↑
X |ψµ ⟩
|ψµ ⟩ = |σ⟩ ⊗ |ψµσ ⟩ ≡ , (S16)
σ |ψµ↓ ⟩
PN
where we remind that |ψµσ ⟩ = σ
j=1 cµj |φj ⟩ is expanded in the basis of spin-resolved Kohn-
Sham orbitals (Nstates = 2N ), similar to Eq. (S5). Using Eqs. (S15) and (S16), the
spin-resolved blocks of the equilibrium density matrix in the presence of SO interaction can
be expressed in the following way
′ ′
ρσσ = cσ f(cσ )† . (S17)
Here, cσ are N × 2N matrices with the spin-resolved Kohn-Sham coefficients at equilibrium
(analogous to the B σ matrices for the open quantum system) and f is a diagonal matrix of
6
dimension 2N × 2N in which the occupation numbers are in the diagonal, i.e. fµν = fµ δµν .
′
ˆ σ′ )† , yields
Comparison with the non-equilibrium expression derived in Sec. II, ρσσ = B σ J(B
the interpretation of J as the non-equilibrium generalization of the occupation numbers at
equilibrium.
IV. SELF-CONSISTENT DFT-NEGF CYCLE WITH SPIN-ORBIT COUPLING
In this section we provide additional details on the implementation and physics of the self-
consistent DFT-NEGF cycle10,13 presented in the main text. This self-consistent cycle ex-
plicitly includes the effect of SO coupling, compared to available tools in the literature10,18–22 ,
which do not incorporate the effect of SO interaction. In this regard, also our DFT-NEGF
self-consistent cycle extends previous works in which the non-equilibrium density matrix
with SO coupling is calculated at first order in the linear voltage and in a non-self-consistent
way23–25 .
Density-matrix update. The self-consistent DFT-NEGF cycle requires to update the
Kohn-Sham Hamiltonian at every step, which itself is a functional of the density matrix. We
choose for the update to employ the blocks Re [ρ̂σσ ]. In other words, it is enough to update
the densities, as DFT physical properties are indeed a functional of the particle density26,27 .
Self-energy and charge neutrality condition. Although in principle one could com-
pute the exact self-energy of the semi-infinite lead, see e.g. Refs. 18 and 28, here we employ
instead the computationally advantageous absorbing boundary condition scheme9,10,14 , as
mentioned in Sec. II. Within the this scheme, the self-energy can be written as
X
Σ̂α = |µ̃⟩ [δϵ − iη]δµ̃ν̃ ⟨µ̃| , (S18)
µ̃,ν̃∈Sα
where the real part, δϵσ , corresponds to a compensating energy shift and the imaginary part,
η, to a local absorption rate, both only active in the subspace of atoms that belong to the
outermost external layers of the metallic cluster representing the electrodes, Sα .
The absorption rate, η, is a parameter which is chosen so that physical quantities for a
given system (density of states, transmission function . . . ) are stable and almost invariant
7
under moderate changes in the value for η (here, we consider η = 0.05 Ha). In addition, from
Eq. (S12) it is clear that the non-equilibrium density matrix has a parametric dependence
on δϵ and EF , the latter understood as µL = µR := EF at zero bias. First, using an initial
guess for EF and assuming that screening occurs within our extended molecule boundaries,
we fix the value for δϵ given a Fermi energy, also ensuring the charge neutrality within some
tolerance
|Nelec − ⟨N ⟩| ≤ ζNelec , (S19)
and
N X
X
⟨N ⟩ = Tr [ρ̂(δϵ)] = ρσσ
ii (δϵn ). (S20)
i=1 σ
A reasonable value for this tolerance parameters is ζ = 10−4 . The convergence of this
self-consistent cycle produces ρ̂(δϵ∞ ).
Next, we correct for nonphysical charge accumulation at the boundaries of the metal
cluster given the optimal δϵ∞ , by minimizing the external charge at the interface
XX
δQouter = ρσσ ref
ii − Q , (S21)
i∈Sα σ
where Qref = NSα ZSα , being NSα the number of atoms of the contact region and ZSα is the
total charge per atom. We compute these charges by using a Löwdin population analysis. In
precedent studies10 , it has been proven that δQouter is a linear function of EF (and therefore
δϵ∞ ). The optimized self-energy is obtained for δQouter (δϵ∗ ) = 0, which fixes δϵ∗ and the
corresponding Fermi energy EF∗ (δϵ∗ ) for any ensuing non-equilibrium calculation.
Finite bias. Once the self-energy is properly parametrized, finite bias in the system
can be considered by setting a difference in the chemical potentials of the two reservoirs,
µL − µR = eVbias with µL > µR by construction. We note that, within this scheme, the
reference chemical potential, µ̄ = (µL + µR )/2, which corresponds to the Fermi energy at
zero bias, is not fixed. Therefore, we can still adjust the value to make sure that the charge
neutrality condition is satisfied at finite bias within a given precision, Nζ . In practical
terms, we consider the finite bias calculation to be converged if |Tr[ρ̂(µ̄)] − Nelec | < Nζ with
Nζ = 10−7 electrons.
8
Magnetic anisotropy term. As explained in the main text, if the magnetic anisotropy
energy is too small compared to the voltage window, the spin density will oscillate in the self-
consistent cycle at finite bias. A possible solution to this scenario is to perform a “magnetic
stabilization” of the spin density towards the easy magnetization axis of the system. This is
done by adding a local magnetic anisotropy term in the atom where most of the spin density
(or spin magnetic moment) is located. In practical terms, we add a term to the Kohn-Sham
Hamiltonian in the step where we build the Kohn-Sham Hamiltonian to incorporate it into
the non-equilibrium density matrix, see Eq. (S2). In the molecular orbital basis, the term
reads
′
XX
Σσσ
MAE = σ∆δσ,σ ′ |p, j⟩ ⟨p, j| . (S22)
p∈A j∈OA
Here, ∆ is a parameter and A is the set of atoms where the term is non-zero (with corre-
sponding orbitals OA .
Exchange-correlation field. The components of the exchange-correlation magnetic
field, formally obtained as the functional derivative of the XC energy with respect to the
magnetization, can be computed from the spin-blocks of the Kohn-Sham Hamiltonian as
x
Bxc = (ĤKS )↑↓ + (ĤKS )↓↑ , (S23)
y
Bxc = i[(ĤKS )↑↓ − (ĤKS )↓↑ ], (S24)
z
Bxc = (ĤKS )↑↑ − (ĤKS )↓↓ . (S25)
V. OBSERVABLES
After convergence of the self-consistent DFT-NEGF cycle (see main text), we can proceed
with the calculation of physical observables or other magnitudes of interest. For example,
from the knowledge of the self-consistent Green’s function at finite bias, we can evaluate the
finite bias transmission function
h i
T (E, Vbias ) = Tr Γ̂L Ĝ(E, Vbias )ΓR Ĝ† (E, Vbias ) . (S26)
9
It can be shown straightforwardly that the transmission function can be written as a sum
over the spin-resolved (spin-conserving and spin-flip) components
X
T (E, Vbias ) = Tσ,σ′ (E, Vbias ), (S27)
σ,σ ′
the latter adopting a particularly simple form in the basis in which the matrix ĤΣ is diagonal
h ′
i
Tσ,σ′ (E, Vbias ) = Tr Γ̃σL Ĝ Γ̃σR Ĝ ∗ . (S28)
Vbias
Here, we have defined the rotated anti-Hermitian part of the left/right self-energy as
Γ̃σL = (B σ )† ΓL B σ , (S29a)
Γ̃σR = ([B −T ]σ )T ΓR [(B −1 )† ]σ , (S29b)
ˆ is the Green’s function in the right eigenbasis of ĤΣ . Finally,
and we recall that Ĝ := G(E)
the charge current can be expressed as11,29
Z +∞
2e
I(E, Vbias ) = dE [f (E − µL ) − f (E − µR )] T (E, Vbias ). (S30)
h ∞
Other observables are obtained by a trace of the non-equilibrium density matrix at fixed
bias with the corresponding operator, see also main text.
10
VI. ADDITIONAL RESULTS
A. Additional figures
Fig. S1. Magnetization, mα , at the vanadium atom as a function of the voltage bias, Vbias , applied
across the molecular junction. We display the traces obtained for several representative values of
the magnetic anisotropy term in the two relevant spatial directions (a) x̂ and (b) ẑ. The dashed
horizontal line represents the value expected for the local magnetization in the isolated vanadocene
molecule, oriented in the ẑ direction. The inset in each panel corresponds to a zoom of the data
inside the dashed light beige rectangle.
11
Fig. S2. Magnetization, my at the vanadium atom as a function of the voltage bias, Vbias , applied
across the molecular junction. The dashed horizontal line represents the value expected for the
local magnetization in the isolated vanadocene molecule, directed in the ẑ direction, shown here as
comparison. The inset corresponds to a zoom of the data inside the dashed light brown rectangle.
0.50
0.25
0.00
δty [µeV]
−0.25
∆ [eV]
−0.50 -1.0
-0.5
−0.75 -0.25
1.0
−1.00
−0.3 −0.2 −0.1 0.0 0.1 0.2 0.3
Vbias [V]
Fig. S3. Local SO torque response, δty , exerted at the spin located in the vanadium atom as a
function of the voltage bias, Vbias , applied across the molecular junction.
12
B. Additional tables
∆ = 0 eV ∆ = −1.0 eV
No SO coupling SO coupling No SO coupling SO coupling
HOMO-1 -3.7168 -3.7187 -3.4979 -3.4997
HOMO -3.5860 -3.5737 -3.3572 -3.3552
LUMO -1.8668 -1.8676 -1.6289 -1.6294
LUMO+1 -1.6314 -1.6311 -1.3698 -1.3696
TABLE I. Kohn-Sham energies of the most representative molecular orbitals of the isolated
vanadocene molecule. All energies are given in eV. We show the values for two representative
cases ∆ = 0 and ∆ = −1.0 eV. The level splitting due to SO coupling is of the order of meV.
Concerning the magnetic anisotropy terms under consideration here, the shifts in the Kohn-Sham
energies are an order of magnitude smaller. These shifts do not influence the relative ordering of
the energy levels.
C. Junction geometry
atom -3.31867693 -0.06187045 4.35673893 Cu
atom -5.20228950 0.72186008 5.79267608 Cu
atom -2.88558216 0.72993814 6.64372828 Cu
atom -4.09535943 -1.41077248 6.24771781 Cu
atom -2.27267900 -2.89247000 8.96326200 Cu
atom -7.08737900 1.53856800 7.16908800 Cu
atom -4.67552700 1.54376600 8.04046800 Cu
atom -2.26367600 1.54896400 8.91184800 Cu
atom -5.88595400 -0.67955000 7.63048500 Cu
atom -3.47410400 -0.67435200 8.50186500 Cu
atom -4.68453000 -2.89766800 8.09188200 Cu
atom -2.65423855 -0.33020629 2.36545816 N
atom -3.12384327 -1.18682727 2.05423888 H
atom -3.08850991 0.43767169 1.84247646 H
atom -2.58445402 -0.84769963 -1.78526729 H
13
atom -0.65214719 1.75915198 1.95730800 H
atom -1.51811663 -0.64597132 -1.84818991 C
atom -1.24285835 -0.40516733 2.05183247 C
atom -0.36476566 0.71081090 1.92292851 C
atom -0.63615589 -2.70993339 -2.04454787 H
atom -1.40634938 1.59613252 -1.76495526 H
atom -0.49063183 -1.63397809 -1.98050888 C
atom -0.90029606 0.63736550 -1.84383017 C
atom -0.19691969 -0.49224728 -0.01553840 V
atom -0.47001932 -1.60175256 1.96908726 C
atom 0.96639067 0.20188917 1.79667540 C
atom -0.85066276 -2.61744856 2.05066946 H
atom 1.86585139 0.80528680 1.70418569 H
atom 0.75577010 -0.94761530 -2.08095580 C
atom 0.51233728 0.45304829 -1.97646388 C
atom 0.90284977 -1.22043125 1.82595422 C
atom 1.26155242 1.23968547 -2.02757459 H
atom 2.03096840 -1.55757029 -2.39009416 N
atom 2.78415331 -1.05519807 -1.90829725 H
atom 2.05185165 -2.51665718 -2.02791050 H
atom 1.74653248 -1.90400741 1.77309523 H
atom 2.69567186 -1.73940944 -4.38823462 Cu
atom 2.19661522 -1.46467877 -6.75320968 Cu
atom 4.53432536 -1.42684065 -6.02779276 Cu
atom 3.29165075 -3.55646920 -5.97407617 Cu
atom 0.18267600 -3.41980600 -9.00216500 Cu
atom 1.47362300 -1.20480200 -9.06190000 Cu
atom 3.93188400 -1.13342100 -8.33511500 Cu
atom 6.39014400 -1.06204100 -7.60832900 Cu
atom 2.64093700 -3.34842600 -8.27538000 Cu
atom 5.09919600 -3.27704500 -7.54859500 Cu
atom 3.80824900 -5.49204900 -7.48886000 Cu
14
1 V. Blum, R. Gehrke, F. Hanke, P. Havu, V. Havu, X. Ren, K. Reuter, and M. Scheffler, “Ab initio
molecular simulations with numeric atom-centered orbitals,” Computer Physics Communications
180, 2175 – 2196 (2009).
2 J. P. Perdew, K. Burke, and M. Ernzerhof, “Generalized gradient approximation made simple,”
Phys. Rev. Lett. 77, 3865–3868 (1996).
3 W. Gao, T. A. Abtew, T. Cai, Y.-Y. Sun, S. Zhang, and P. Zhang, “On the applicability of
hybrid functionals for predicting fundamental properties of metals,” Solid State Communications
234-235, 10–13 (2016).
4 E. van Lenthe, J. G. Snijders, and E. J. Baerends, “The zero-order regular approximation for
relativistic effects: The effect of spin–orbit coupling in closed shell molecules,” The Journal of
Chemical Physics 105, 6505–6516 (1996).
5 W. Kohn and L. J. Sham, “Self-consistent equations including exchange and correlation effects,”
Phys. Rev. 140, A1133–A1138 (1965).
6 W. P. Huhn and V. Blum, “One-hundred-three compound band-structure benchmark of post-
self-consistent spin-orbit coupling treatments in density functional theory,” Phys. Rev. Mater.
1, 033803 (2017).
7 M. Dreizler and E. K. U Gross, Density Functional Theory: An Approach to the Quantum
Many-Body Problem, 1st ed. (Springer, Berlin, 1990).
8 R. G. Parr and W. Yang, Density Functional Theory of Atoms and Molecules, 1st ed. (Oxford
University Press, Oxford, 1989).
9 A. Arnold, F. Weigend, and F. Evers, “Quantum chemistry calculations for molecules coupled
to reservoirs: Formalism, implementation, and application to benzenedithiol,” The Journal of
Chemical Physics 126, 174101 (2007).
10 A. Bagrets, “Spin-polarized electron transport across metal–organic molecules: A density func-
tional theory approach,” J. Chem. Theory Comput. 9, 2801–2815 (2013).
11 M. Di Ventra, Electrical Transport in Nanoscale Systems, 1st ed. (Cambridge University Press,
Cambridge, 2008).
12 J. C. Cuevas and E. Scheer, Molecular Electronics, 2nd ed. (World Scientific, 2010).
15
13 M. Camarasa-Gómez, Ab Initio Electronic Transport in Single-Molecule Junctions: Quantum
Interference Effects and Spin-Orbit Torque, Ph.D. thesis, Universität Regensburg (2021).
14 F. Evers and A. Arnold, “Molecular conductance from ab initio calculations: Self energies and
absorbing boundary conditions,” in CFN Lectures on Functional Nanostructures - Vol. 2, Lecture
Notes in Physics, Vol. 820, edited by C. Röthig, G. Schön, and M. Vojta (Springer, 2011) pp.
27–53.
15 P.-O. Löwdin, “On the non-orthogonality problem connected with the use of atomic wave func-
tions in the theory of molecules and crystals,” The Journal of Chemical Physics 18, 365–375
(1950).
16 J. J. Sakurai and J. Napolitano, Modern quantum mechanics, 3rd ed. (Cambridge University
Press, Cambridge, 2021).
17 H. P. Breuer and F. Petruccione, The theory of open quantum systems, 1st ed. (Oxford University
Press, New York, 2002).
18 A. R. Rocha, V. M. García-Suárez, S. Bailey, C. Lambert, J. Ferrer, and S. Sanvito, “Spin
and molecular electronics in atomically generated orbital landscapes,” Phys. Rev. B 73, 085414
(2006).
19 N. Papior, N. Lorente, T. Frederiksen, A. García, and M. Brandbyge, “Improvements on non-
equilibrium and transport Green function techniques: The next-generation transiesta,” Com-
puter Physics Communications 212, 8–24 (2017).
20 W. Dednam, C. Sabater, O. Tal, J. J. Palacios, A. E. Botha, and M. J. Caturla, “Refined
electron-spin transport model for single-element ferromagnetic systems: Application to nickel
nanocontacts,” Phys. Rev. B 102, 245415 (2020).
21 B.-H. Huang, Y.-F. Lai, and Y.-H. Tang, “Validity of dft-based spin-orbit torque calculation
for perpendicular magnetic anisotropy in iron thin films,” AIP Advances 13, 015034 (2023).
22 B.-H. Huang, Y.-H. Fu, C.-C. Kaun, and Y.-H. Tang, “Determining perpendicular magnetic
anisotropy in Fe/MgO/Fe magnetic tunnel junction: A DFT-based spin–orbit torque method,”
Journal of Magnetism and Magnetic Materials 585, 171098 (2023).
23 B. K. Nikolić, K. Branislav, Kapildeb Dolui, Marko D. Petrović, Petr Plecháč, Troels Markussen,
and Kurt Stokbro, “First-principles quantum transport modeling of spin-transfer and spin-orbit
torques in magnetic multilayers,” in Handbook of Materials Modeling: Applications: Current
and Emerging Materials, edited by Wanda Andreoni and Sidney Yip (Springer International
16
Publishing, 2018) pp. 1–35.
24 K. D. Belashchenko, Alexey A. Kovalev, and M. van Schilfgaarde, “First-principles calculation
of spin-orbit torque in a Co/Pt bilayer,” Phys. Rev. Mater. 3, 011401 (2019).
25 K. Dolui, M. D. Petrović, K. Zollner, P. Plecháč, J. Fabian, and B. K. Nikolić, “Prox-
imity spin–orbit torque on a two-dimensional magnet within van der Waals heterostructure:
Current-driven antiferromagnet-to-ferromagnet reversible nonequilibrium phase transition in bi-
layer CrI3 ,” Nano Letters 20, 2288–2295 (2020).
26 R. O. Jones, “Density functional theory: Its origins, rise to prominence, and future,” Rev. Mod.
Phys. 87, 897–923 (2015).
27 F. Evers, R. Korytár, S. Tewari, and J. M. van Ruitenbeek, “Advances and challenges in single-
molecule electron transport,” Rev. Mod. Phys. 92, 035001 (2020).
28 M. Brandbyge, J.-L. Mozos, P. Ordejón, J. Taylor, and K. Stokbro, “Density-functional method
for nonequilibrium electron transport,” Phys. Rev. B 65, 165401 (2002).
29 M. Büttiker, Y. Imry, R. Landauer, and S. Pinhas, “Generalized many-channel conductance
formula with application to small rings,” Phys. Rev. B 31, 6207–6215 (1985).
17
