Band Unfolding Made Simple

Sara G. Mayo, Felix Yndurain and Jose M. Soler

Dpto. Fı́sica de la Materia Condensada
Universidad Autónoma de Madrid
Cantoblanco, 28049 Madrid, Spain
(Dated: June 27, 2019)
We present a simple view on band unfolding of the energy bands obtained from supercell calcula-
tions. It relies on the relationship between the local density of states in reciprocal space (qLDOS)
and the fully unfolded band structure. This provides an intuitive and valid approach not only for
periodic, but also for systems with no translational symmetry. By refolding into the primitive Bril-
arXiv:1812.03925v2 [cond-mat.mtrl-sci] 26 Jun 2019

louin zone of the pristine crystal we recover the conventional unfolded bands. We implement our
algorithm in the Siesta package and apply it to defects on Si and graphene.

INTRODUCTION METHOD

Plots of so called energy bands are the most basic and The basic steps of our unfolding method are schemat-
used tool in interpreting the calculated electronic struc- ically summarized in Fig. 1. We start by considering the
ture of simple crystals. Such plots represent the energy of energy bands not just as dispersion relations, but as the
the Bloch orbitals as a function of their crystal momen- density of states in the (first) simulation Brillouin zone
tum in the primitive Brillouin zone (PBZ), that is, E (k). (SBZ), the reciprocal of the real-space simulation (super)
These theoretical band structures can be obtained within cell (SC):
the tight-binding approximation or the density functional X
nSBZ (K, ) = δ( − K,i ) (1)
theory (DFT) and they have a direct connection with
i
the results of angle resolved photoemission spectroscopy
(ARPES) experiments. However, the simplicity of this where K is a wave vector in the SBZ, and K,i is the
approach disappears when the calculations involve large eigenvalue of its ith Bloch state, that is, band energy
supercells with many atoms. As the size of the cell in [15].
real space increases, the first Brillouin zone in reciprocal The normalization of nSBZ (K, ), as well as those of
space shrinks and more lines populate the band struc- other densities defined below, is such that they are den-
ture, hindering the extraction of useful information and sities of states per unit of macroscopic volume (as well as
the comparison with experiments. per unit of their respective independent variables), what
facilitates the comparison between simulation cells of dif-
Several authors have already developed techniques to
ferent volumes.
unfold the supercell Brillouin zone into the primitive Bril-
Next, we split the normalized weight δ( − K,i ) ac-
louin zone. Some of the existing works focus on algo-
cording to the squared Fourier coefficients of the corre-
rithms within the tight-binding approximation [1–4] or
sponding wave function. Summing over Bloch states, we
first principles calculations, employing as basis sets linear
construct the fully unfolded bands as the spectral density,
combinations of atomic orbitals (LCAO) [5], plane waves
which can also be considered as the local density of states
[6] or Wannier functions [7]. Another methodology stud-
in reciprocal space (qLDOS):
ies the electronic structures of alloy systems [8, 9]. Other
authors have delved into the theory of the problem and XZ
they have developed general formulations of it [10–14]. n(q, ) = dK |ψ̃K,i (q)|2 δ( − K,i )
i SBZ
Most of these methods search a direct relationship be- X
tween Brillouin zones of the primitive crystal cell and = |ũKq ,i,Gq |2 δ( − Kq ,i ), (2)
i
of the simulation supercell, by expanding the supercell
states in a basis set with the periodicity of the crystal. where ψK,i (r) is a Bloch wave function (normalized in
In contrast, we aim to simplify the procedure and provide the SC), uK,i (r) ≡ ψK,i (r)e−iKr is its periodic part, and
a link between existing methods by dividing the unfold- ψ̃K,i (q), ũK,i,G are their respective Fourier transforms:
ing problem in two steps. First, we consider the fully
unfolded bands, extended to the whole reciprocal space, Z
1
through a Fourier decomposition of the Bloch wave func- ψK,i (r) = dq ψ̃K,i (q) eiqr , (3)
tions of the system [14]. This results in a non-periodic (2π)3/2 ∞
description, interesting by itself. Second, by what we call
a refolding into the crystal primitive Brillouin zone, we
X
ψ̃K,i (q) = (2π)3/2 δ(K + G − q) ũK,i,G , (4)
recover the conventional unfolded bands of other authors. G
 2

these “bands” have a different weight at each Brillouin

10
5
(a) zone (Fig. 1(c)).
As can be seen by comparing with Eq. (2), the qLDOS
V(x)

0 is the Fourier-space equivalent of the real-space local den-

-5 sity of states (rLDOS),
-3 -2 -1 0 1 2 3 Z
x/a
VSC X
n(r, ) = dK |ψK,i (r)|2 δ( − K,i ). (6)
1 (2π)3 i SBZ
0 (b) The total density of states (DOS) can be obtained by
(K)

-1 integration of either n(r, ) or n(q, ):

-2
Z Z
-5 0 5
1 1
n() = dr n(r, ) = dq n(q, ). (7)
2Ka/ VSC SC (2π)3 ∞
1
Since |ψ̃K,i (q)|2 is the probability of measuring mo-
0 (c)
(q)

mentum q of a given electron, n(q, ) is the probability

-1 of finding an electron (or an empty state) in the sys-
-2 tem with energy  and momentum q, and it can thus be
-5 0 5 directly related with ARPES results if matrix element
2qa/
effects are taken into account [14, 16, 17].
1
The qLDOS, that we call fully unfolded bands is the
0 (d) same as the spectral weight of other references [10, 12, 13]
(k)

-1 and the plane-wave unfolded spectra introduced by Ko-

sugi et al. [14]. Therefore, our approach is a different
-2
-5 0 5 description, rather than a new method that yields differ-
2ka/ ent results. Our emphasis is to provide a clear and simple
link with previous methods through the (L)DOS, as well
FIG. 1: Scheme of the unfolding method. (a) 1D model of as to generalise band structure analysis to non periodic
a chain of pseudo-atoms (dots) with an attractive gaussian systems.
potential (line). The atoms are paired, with a Peierls distor- The last step in our method is to refold n(q, ) into a
tion of 2% relative to their undistorted distance a. The en- refolding Brillouin zone (RBZ) as
ergy origin is the average potential and its units are ~2 a2 /me . X
The unit cell of the distorted chain is shaded. (b) First two nRBZ (k, ) = n(k + g, ), (8)
bands of the distorted chain, with its PBZ shaded. A ran- g
dom K point is singled out (dashed line). The (normalized)
weights of the Bloch states at this K (black dots), are split where k is within the RBZ and g = πn/a
according to the squared Fourier coefficients of their respec- P are its recipro-
2
cal lattice vectors. Notice that, since G |ũK,i,G | = 1,
tive wave functions (red dots), appearing at q = K + G. Dot P
then G n(K + G, ) = n SBZ (K, ), that is, refolding
areas are proportional to weights. (c) Fully unfolded bands,
obtained through the procedure described in (b) for all Ks n(q, ) back into the SBZ recovers the original bands.
in the PBZ. The non-periodic unfolded weights are the local Frequently, the simulation cell will be a supercell of the
DOS in Fourier space (qLDOS) n(q, ). (d) Bands refolded refolding cell. In these cases, the RBZ will be a supercell
into the PBZ of the undistorted chain (shaded region). The of the SBZ, and vectors g will belong to the set of Gs
weight at each k (blue dots) is the sum over g of the unfolded (Fig. 1(d)). Nevertheless, this condition is not necessary
weights in (c), at points q = k + g. The continuous line is the in our method, and in fact it will not be true in many
first band of the undistorted chain.
cases, as for simulation cells of liquids or amorphous sys-
Z tems, or of defects that induce strong deformations [18].
1 The above full unfolding/refolding method can be im-
ũK,i,G = dr uK,i (r) e−iGr , (5)
VSC SC mediately applied in a plane wave DFT code [19–21],
since the Fourier coefficients of the Bloch wave functions
with VSC the volume of the SC and G its reciprocal wave
are then directly available. The slow decay with mo-
vectors. In Eq. (2), Kq and Gq are the unique vectors
mentum of all-electron wavefunctions can be addressed
such that: Kq is within the SBZ; Gq is a reciprocal wave
by using pseudopotentials or by introducing a momen-
vector; and Kq + Gq = q. A state ψK,i will contribute
tum cutoff. For a basis of atomic orbitals, we expand the
to n(q, ) at points q = K + G for all G = πN/a vectors,
Bloch states as
due to Bloch’s theorem. We emphasize that q extends XX
to infinity and n(q, ) is not periodic in q: although the ψK,i (r) = cK,i,µ φµ (r − R − rµ ) eiK(R+rµ ) , (9)
energies at which n(q, ) can be nonzero are periodic, R µ
 3

where cK,i,µ are expansion coefficients and φµ are atomic

orbitals centered at position R + rµ (R being SC lattice
vectors). Substituting into Eq. (5) we find

(2π)3/2 X
ũK,i,G = cK,i,µ φ̃µ (K + G) e−iGrµ , (10)
VSC µ

where φ̃µ (q) is the Fourier transform of the numerical

atomic orbital φµ (r), with well defined angular momen-
tum quantum numbers (lµ , mµ ), that can be decomposed
into radial and angular parts:

φµ (r) = φµ (r) Ylµ ,mµ (r̂) (11)

φ̃µ (q) = φ̃µ (q) Ylµ ,mµ (q̂) (12)

∞
r Z
2
φ̃µ (q) = (−1)lµ r2 dr jlµ (qr) φµ (r), (13)
π 0

with Yl,m (r̂) spherical harmonics and jl (x) spherical

Bessel functions.
This algorithm has been implemented in the Siesta
package [22]. After a converged DFT Siesta calculation,
the hamiltonian and overlap matrices, in the atomic basis
set, are calculated and written in a file. This file, as well
as those specifying the radial numerical atomic orbitals,
are read by an external utility program that calculates
the fully unfolded and refolded spectra at the desired
q and k band lines. Some Siesta subroutines are also
used by the unfolding/refolding program to obtain the
wave function coefficients at each required K point of
the SBZ, as well as to perform the Fourier transforms in
Eq. (13).
FIG. 2: (a) Refolded bands, into the crystal PBZ, of a single
vacancy in bulk Si. Energies are relative to the Fermi level,
APPLICATIONS set at zero (horizontal dotted line). Vertical dotted lines mark
high symmetry points of the PBZ. The colorbar is the scale
for the LDOS.; (b) (conventional) bands of Si FCC crystal;
We apply our previously described Siesta implemen- (c) difference between the refolded LDOS of the defective Si
tation to a Si FCC crystal with a single vacancy, a model and the pristine crystal, zoomed in the gap region, near the
of amorphous Si (a-Si), a monolayer of graphene with a Γ point. States arising (vanishing) due to the vacancy appear
(585) divacancy, and a rotated graphene bilayer under in hot (cold) colours. The black (red) dashed line is the Fermi
pressure. We employed the GGA-PBE [23] functional level of the perfect (defective) system.
for exchange and correlation and double-ζ + polariza-
tion (DZP) basis sets (double-ζ (DZ) for a-Si).

of the supercell approach and they become smoother in

Vacancy in Si crystal
a larger SC.

We model a vacancy in a Si FCC crystal using a The most relevant changes occur around the Fermi
63-atom supercell. Its refolded bands into the crystal level. A new state arises within the gap, with higher
PBZ vectors are depicted in Fig. 2(a), and can be com- weight around the Γ point. The top of the valence band,
pared with the bands of the periodic crystal, in Fig. 2(b). around Γ, decreases in energy and in weight. We show
Changes are appreciated at a careful sight. The refolded the difference between the crystal and defective refolded
bands become blurred and widened due to the appear- spectra in Fig. 2(d) around Γ, at the gap, to remark these
ance of small splittings. Some of these are a consequence changes.
 4

Amorphous Si regions of c-Si band degeneracies. This similarity is inde-

pendent of the size of the a-Si simulation cell employed.
A clear example of disorder is an amorphous solid. In We attribute it to the similarity of the local structure in
this case, we cannot talk about a proper band structure, a-Si and c-Si in bond distances and angles [24].
but yet the energy dispersion of the electron states pro-
vides interesting results. We studied the fully unfolded
Divacancy in graphene
and refolded bands of a-Si using supercells of 216, 512
and 1024 atoms, modelled by Igram et al. [24], obtaining
equivalent results for all cases. We present the results for Graphene is a material with unique electronic proper-
the 512-atom supercell. ties, but highly sensitive to structural disorder. The pres-
ence of defects leads to significant changes on its bands,
specially around the Fermi level. Many types of defects
have been studied by theorists, such as adatoms, vacan-
cies or Stone-Wales defects, with the aim of predicting
their properties and, potentially, using them to tailor the
functionalities of graphene. Hence, unfolding their band
stuctures will shed light on how they modify the original
graphene bands.

FIG. 4: (a) Relaxed structure of the (585) divacancy in

graphene. The dashed lines are the mirror planes of the de-
fect. (b) PBZ of graphene, its symmetry modified by the
FIG. 3: (a) Fully unfolded bands of a 512-atom cubic cell (585) defect.
of amorphous silicon, along the (111) and (100) directions.
Vertical lines indicate high symmetry points of the first and We consider a (585) divacancy in a graphene mono-
second Brillouin zones of the crystal. They are shown for ref- layer, a defect that has been synthesised and charac-
erence but they are not special in the a-Si simulation cell. The
terised by Ugeda et al. [26]. The (585) defect is formed
red line is a fit to a free-electron dispersion relation with an
effective mass m∗ = 1.1; (b) refolding of the a-Si bands into by two adjacent C vacancies rearranged into two pen-
the PBZ of crystalline silicon. Dashed lines are the crystal tagons and one octagon, as shown in Fig. 4(a), with no
bands. dangling σ bonds. We modelled the divacancy in simula-
tion cells of different sizes, from 7 × 7 to 14 × 14. Here we
Fig. 3(a) show the fully unfolded bands along two direc- report the 11 × 11, 12 × 12 and 13 × 13 as representative.
tions of reciprocal space, corresponding to the L − Γ and The conventional bands of the supercell calculations
Γ−X directions of the crystal, up to the second Brillouin within their SBZs, Fig. 5, reveal that the former six-fold
zone. As expected from the isotropic character of the symmetry of the Brillouin zone is broken, leading to in-
amorphous solid, its unfolded bands are essentially iden- equivalent K and M points and two emerging mirror
tical in both directions, yielding a widened free-electron planes, Fig. 4(b), as happens in the atomic structure.
dispersion with an effective mass m∗ = 1.1 and a gap of The K and K 0 points no longer present a Dirac cone,
1.5 eV, consistent with existing values [24, 25]. although all of them present a band crossing of EF at
The same bands, refolded to the PBZ of the crystal, are different points in their Brillouin zones: near K 0 for the
shown in Fig.3(b). At close magnification and inspection, 11 × 11 SC, around Γ for the 12 × 12 SC and between M 0
a few localised states appear in the gap due to defects and K for the 13 × 13 SC. Besides this, it is hard to find
in the a-Si model [24]. Interestingly enough, despite its similarities between them. To relate these overcrowded
isotropic and non-periodic structure, and the incommen- spectra with the band structure of graphene, we compute
surability of its simulation cell with that of the crystal, the refolded bands into the PBZ in the surroundings of
the refolded bands of a-Si appear as a blurred version of K and K 0 points of graphene.
the crystalline silicon (c-Si) bands, specially at low en- Fig. 6(a) and (c) show the 11 × 11 and 13 × 13 SCs
ergies. Also, higher LDOS of a-Si can be appreciated in bands refolded into the PBZ. The Fermi level crossing is
 5

FIG. 5: Conventional bands of a (585) divacancy in graphene modelled in (a) 11 × 11, (b) 12 × 12 and (c) 13 × 13 simulation
cells, in their respective SBZs.

FIG. 6: Bands around the Fermi level of a (585) divacancy in graphene, refolded into the PBZ, obtained from (a) 11 × 11,
(b) 12 × 12 and (c) 13 × 13 simulation cells. K 0g denotes the K 0 point of the PBZ of graphene. K 0n indicates a K 0 point of
the (n × n) SBZ.

FIG. 7: 2D bands of a (585) divacancy refolded into the PBZ, in the surroundings of the K0 point, obtained from (a) 11 × 11,
(b) 12 × 12 and (c) 13 × 13 simulation cells. The darkness times size of the dots is proportional to the LDOS. The units for k
are 1/a0 , where a0 is the lattice parameter. (Interactive figures available as supplementary material).

located around K 0 of graphene, which corresponds to a coincident with the tip of the lower cone.
K 0 point in both SBZs. One of the bands conforming
the former Dirac cone stays almost unaltered, while the In the case of the 12 × 12 SC, Fig. 6(b), we see a sim-
other one is split, forming two cone tips shifted in the ilar behaviour, with slight differences. One band of the
kx direction and connected by a state. In the 2D bands cone remains almost unaltered, with a small gap of 0.04
plot of Fig. 7(a) and (c) this is seen with more clarity. eV opening at K 0 . The other one is split, with one of
The Fermi level decreases from its original value, being its fragments conforming a flat state at the Fermi level,
leading to a single band crossing. In Fig. 7(b) this flat
 6

state and the cone tips can be appreciated. Here, the K 0 We study a bilayer rotated 5.08◦ , at equilibrium under
of graphene coincides with a Γ point of the 12 × 12 SBZ. increasing pressures up to 1.63 GPa. We employ a GGA
Despite the differences between the three sizes, a gen- exchange-correlation functional including Van der Waals
eral trend is clearly identified after refolding into the PZB interactions [33, 34]. As none of the monolayers lattice
of graphene, which is not the case in the traditional bands orientations has a prevalence over the other, refolding
description of Fig. 5. All cases present a splitting and a into the PBZ of one of them is not particularly infor-
shift in kx of the Dirac cone, as well as a single band mative. Therefore, we consider the fully unfolded bands
between the two cone tips as the lowest unoccupied elec- as the adequate tool to analyze this system. We first
tronic state. The Fermi level is located at the tip of the show the fully unfolded bands of a graphene monolayer,
lower cone. We remark that, in 3n × 3n SCs, the K 0 of Fig. 8, up its second Brillouin zone, for comparison pur-
the PBZ corresponds to a Γ point in the SBZ, unlike in poses. Two paraboloids with gaps opening along them
the (3n + 1) × (3n + 1) and (3n + 2) × (3n + 2) cases. This and a six-fold symmetry and are clearly distinguished,
is consistent with the formal differences existing between and conform the dispersion relations of the σ and π or-
both kinds of supercells [27]. Finally, the refolded bands bitals. In the case of the bilayer, we will restrict the
obtained around K are inverted in ky compared to the
refolded bands around K 0 shown above.

Pressure in rotated graphene bilayers

Another path to modify the electronic behaviour of

graphene is given by rotated graphene bilayers. These
are defect-free systems with a Moiré pattern, involving
large periodicities. At small angles, the interaction be-
tween the two layers induces two saddle points in the
band structure, along with two logarithmic van Hove sin-
gularities in the DOS. As the angle decreases, the singu-
larities approach and, eventually, at the so called magic
angles (θ = 1.1◦ ), they collapse into a single peak at the
Fermi level [28]. The same effect has been recently re-
ported to occur for larger angles, when external pressure
is applied [29–32].

FIG. 9: Conventional bands in the SBZ and DOS of a

FIG. 8: Fully unfolded bands of a graphene monolayer, up graphene bilayer rotated an angle of 5.08◦ , (a) at equilib-
to the second Brillouin zone. The squared area marks a Dirac rium geometry; (b) under a pressure of 0.70 GPa and (c)
cone, region considered for further study in the bilayer case. under 1.63 GPa.
 7

FIG. 10: 2D fully unfolded bands of a graphene bilayer rotated 5.08◦ , in the surroundings of the first K points of the
monolayers, at different pressures: (a) equilibrium distance. The arrows indicate the saddle points that originate the van Hove
singularities; (b) 0.04 GPa; (c) 0.23 GPa; (d) 0.70 GPa; (e) 1.06 GPa and (f ) 1.63 GPa. The darkness and size of the dots
is the scale for LDOS, relative to the maximum value of each data set. Weights smaller than 100 times the maximum are
neglected (interactive figures available as supplementary material).

◦
unfolding region to the surroundings of a K (and K 5.08 ) flat bands. We remark how their weight in this area of
point (red square). reciprocal space is small compared to that of the merging
In Fig. 9 we show the conventional band structures of bands, despite the sharp peak on the DOS of Fig. 9(c).
the bilayer at the equilibrium distance, at a middle stage This is an indicator of delocalization in q, and is not un-
and under a pressure of 1.63 GPa, next to their corre- expected, as these states are known to be well localized
sponding DOS. The saddle points and van Hove singu- in the AA stacking region in real space [26, 29].
larities can be appreciated.
Fig. 10 depicts the evolution of the fully unfolded spec-
tra under increasing pressures. At equilibrium configu- CONCLUSIONS
ration, Fig. 10(a) shows a neat picture of the interaction
between the cones of both monolayers, as well as the sad- We have presented a simple formulation of the band
dle point emerging in between. The relative maximum unfolding problem, a tool necessary to extract useful in-
intensities of the LDOS are homogeneous in energies, and formation from the band structure of large supercell cal-
the states present high dispersion in energies. As pres- culations. The idea of a full unfolding that expands the
sure is applied, Fig. 10(b),(c),(d) and (e), the cones flat- bands not only to the primitive cell, but to the full recip-
ten and the saddle points move towards the Fermi level. rocal space, allows to treat this problem as a decompo-
We appreciate as well that the two pairs of bands im- sition of the wave functions into its Fourier coefficients.
mediately over and below the cones lower their energies A refolding recovers the conventional unfolded bands in
and start to merge. Also, in Fig. 10(d) and (e), higher the PBZ of the crystal. It is feasible for any eigenstate,
weights correspond to these merging bands, whereas the regardless of the basis used. In the case of plane wave
cone states around the Fermi level tend to be delocalized codes this implementation shall be almost immediate.
in many qs. In Fig. 10(f) the cones have collapsed into We have successfully applied our algorithm to obtain
 8

new characterizations of non-trivial physical systems. [11] H. Huang et al., New J. Phys. 16, 033034 (2014)
The fully unfolded bands provide a distribution of the [12] P. V. C. Medeiros, S. Stafström, and J. Björk, Phys. Rev.
states as a function of their energy and momenta, allow- B 89, 041407 (2014)
ing a direct comparison with experimental photoemission [13] O. Rubel, A. Bokhanchuk, S. J. Ahmed, and E. Assmann,
Phys. Rev. B 90, 115202 (2014)
spectra, as well as a way to determine a value of the ef- [14] T. Kosugi, H. Nishi, Y. Kato, and Y. Matsushita, J.
fective mass of the system under study in a chosen direc- Phys. Soc. Jpn. 86, 124717 (2017)
tion of reciprocal space. Refolding into the primitive cell [15] Following conventional practice, we write the Bloch wave
yields clear band spectra that allow comparison with the vector K as a subscript, even though it is a continuous
crystal bands, even identifying crystal-like patterns in an variable.
amorphous solid. [16] F. J. Himpsel and D. Eastman, J. Vac. Sci. Technol. 16,
The outcomes of this work prove that the underlying 1297 (1979)
[17] C. Q. Han et al., Phys. Rev. B 90, 085101 (2014)
state distribution in reciprocal space is much richer than
[18] In these cases, however, the vector K such that K + G =
what conventional band structures can evince, transcend- k + g, will depend not only on k but also on g. This will
ing the existence of any real or imposed periodicity. make the calculation of nRBZ (k, ), at given k points,
considerably more expensive.
[19] https://www.abinit.org/
AKCNOWLEDGEMENTS [20] http://www.quantum-espresso.org/
[21] http://www.vasp.at/
We are grateful to B. Battarai and D. Drabold for shar- [22] J. M. Soler et al., J. Phys.: Condens. Matter 14, 2745
(2002)
ing the coordinates of their a-Si model with us. JMS
[23] J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev.
thanks Prof. P. B. Allen for introducing him into the Lett.77, 3865 (1996)
unfolding problem. Funded by Spain’s MINECO grant [24] D. Igram, B. Bhattarai, P. Biswas and D.A. Drabold, J.
FIS2015-64886-C5-5-P. Non-Cryst. Solids 492, 27-32 (2018)
[25] H. D. Barber, Solid-State Electron. 10, 1039 (1967)
[26] M. M. Ugeda, I. Brihuega, F. Hiebel, P. Mallet, J.
Y. Veuillen, J. M. Gomez-Rodriguez, and F. Yndurain,
Phys. Rev. B 85, 121402(R) (2012)
[1] T. B. Boykin and G. Klimeck, Phys. Rev. B 71, 115215 [27] J. M. Garcı́a-Lastra, Phys. Rev. B 82, 235418 (2010)
(2005) [28] Y. Cao, V. Fatemi, S. Fang, K. Watanabe, T. Taniguchi,
[2] T. B. Boykin, N. Kharche, G. Klimeck and M. Korkusin- E. Kaxiras, and P. Jarillo-Herrero, Nature (London) 556,
ski, J. Phys.: Condens. Matter 19, 036203 (2007) 43 (2018)
[3] T. B. Boykin, A. Ajoy, Physica B 531, 130-138 (2018) [29] F. Yndurain, Phys. Rev. B 99 045423 (2019)
[4] T. G. Dargam, R. B. Capaz, and B. Koiller, Phys. Rev. [30] A, Lopez-Bezanilla, Phys. Rev. Materials 3, 054003
B 56, 9625 (1997) (2019)
[5] C. C Lee, Y. Yamada-Takamura and T. Ozaki, J. Phys.: [31] S. Carr, S. Fang, P. Jarillo-Herrero and E. Kaxiras, Phys.
Condens. Matter 25 345501 (2013) Rev. B 98, 085144 (2018)
[6] M. X. Chen and M. Weinert, Phys. Rev. B 98, 245421 [32] M. Yankowitz, S. Chen, H. Polshyn, K.Watanabe, T.
(2018) Taniguchi, D. Graf, A. F. Young and C. R. Dean, Sci-
[7] W. Ku, T. Berlijn and C. C. Lee, Phys. Rev. Lett.104, ence 363 1059 (2019)
216401 (2010). [33] K. Berland and P. Hyldgaard, Phys. Rev. B 89, 035412
[8] V. Popescu and A. Zunger, Phys. Rev. Lett. 104, 236403 (2014)
(2010) [34] G. Romn-Prez and J. M. Soler, Phys. Rev. Lett. 103,
[9] V. Popescu and A. Zunger, Phys. Rev. B 85, 085201 096102 (2009)
(2012) [35] G. Trambly de Laissardire, D. Mayou, and L. Magaud,
[10] P. B. Allen, T. Berlijn, D. A. Casavant, and J. M. Soler, Phys. Rev. B 86, 125413 (2012)
Phys. Rev. B 87, 085322 (2013)