First-principles study on effects of local

Coulomb repulsion and Hund's coupling in

ferromagnetic semiconductor CrGeTe3
Cite as: J. Appl. Phys. 128, 123901 (2020); https://doi.org/10.1063/5.0015566
Submitted: 29 May 2020 . Accepted: 04 September 2020 . Published Online: 22 September 2020

Eunjung Ko , and Young-Woo Son

COLLECTIONS

Paper published as part of the special topic on 2D Quantum Materials: Magnetism and Superconductivity

ARTICLES YOU MAY BE INTERESTED IN

Cleavable magnetic materials from van der Waals layered transition metal halides and
chalcogenides
Journal of Applied Physics 128, 110901 (2020); https://doi.org/10.1063/5.0023729

Electronic and optical properties of low-dimensional group-IV monochalcogenides

Journal of Applied Physics 128, 121101 (2020); https://doi.org/10.1063/5.0016003

Thermodynamic analysis of GaInN-based light-emitting diodes operated by quasi-resonant

optical excitation
Journal of Applied Physics 128, 123103 (2020); https://doi.org/10.1063/5.0008041

J. Appl. Phys. 128, 123901 (2020); https://doi.org/10.1063/5.0015566 128, 123901

© 2020 Author(s).
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

First-principles study on effects of local Coulomb

repulsion and Hund’s coupling in ferromagnetic
semiconductor CrGeTe3
Cite as: J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566
Submitted: 29 May 2020 · Accepted: 4 September 2020 · View Online Export Citation CrossMark
Published Online: 22 September 2020

Eunjung Ko and Young-Woo Sona)

AFFILIATIONS
Korea Institute for Advanced Study, Seoul 02455, South Korea

Note: This paper is part of the Special Topic on 2D Quantum Materials: Magnetism and Superconductivity
a)
Author to whom correspondence should be addressed: [email protected]

ABSTRACT
Based on a first-principles computational method treating the local Coulomb repulsion of U and Hund’s coupling of J separately, we investi-
gate the effects of these important interactions on the electronic and magnetic structures of a layered ferromagnetic semiconductor
CrGeTe3. We show that explicit inclusion of a density functional for the coupling J is critical to obtain proper bandgaps of a single-layer
and bulk systems with a reasonable U, unlike a usual approximation treating those together. After obtaining reliable U and J, we also study
the effects of strain on its energy gap. Within 2% of tensile and compressive strains, respectively, there is no strain-induced semiconductor-
to-metal transition, contrary to the results without J, thus highlighting important roles of exchange interactions.

Published under license by AIP Publishing. https://doi.org/10.1063/5.0015566

I. INTRODUCTION the roles of local Coulomb repulsion as well as exchanges and

their interplay is still lacking.
Recently, the novel magnetic properties of van der Waals
(vdW) ferromagnetic (FM) layered semiconducting materials have The bulk CGT is an intrinsic FM semiconductor with the crit-
ical transition temperature (Tc) of 61 K27 and the experimental
attracted a lot of interest owing to intriguing electronic and mag-
bandgap (Eg) of 0.2–0.7 eV.23,28–30 There have been several theoret-
netic properties shown in their mono- or few-layers samples.1–14
ical studies on changes in Tc, magnetic ordering, and bandgap
Since many vdW layered FM semiconducting materials in two-
from bulk to ML.20–26 Broadly speaking, for CGT compounds, the
dimensions (2D) have correlated transition metal ions and neigh-
electron–electron interactions have been considered based on two
boring chalcogens,15–17 a proper description of their electronic and different computational approaches: (1) computational methods
magnetic properties poses an interesting challenge in treating a based on density functional theory (DFT) with additional on-site
local correlation and exchange together with covalent bondings Coulomb interaction of U and Hund exchange parameter of J,31
between atoms. Among semiconducting 2D FM materials, a in which an isotropic screened on-site Coulomb parameter
monolayer (ML) and bulk CrGeTe3 (CGT) compounds attract par- Ueff ¼ U  J (called as DFT + U method) is added.20–25 (2) DFT
ticular attention regarding that aspect. Recent experiments demon- with hybrid functional computation in which the local or semilocal
strated that the bulk semiconducting FM ordering along the density functional exchange is mixed with exact non-local Hartree–
out-of-plane direction is maintained in its semiconducting bilayer Fock exchange.20,21 The former methods showed unexpectedly
samples.18,19 On the other hand, its computed ground state prop- decreasing Eg behavior as Ueff increases20–22 while typical DFT + U
erties of ML and bulk sensitively depend on a way of computing computations on correlated materials show the opposite trend.32
Coulomb repulsions and exchange interactions.20–26 We note that Even with a reasonable strength of Ueff, the DFT + U method
several theoretical and computational studies20–26 have been per- points to a metallic bulk CGT.21 This uncommon behavior of the
formed to understand their electronic and magnetic properties CGT bandgap as a function of the Hubbard U correction has been
with various computational methods. However, a critical study on attributed to the disparate wave function characteristics of

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-1

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

conduction and valence band edges21 or a strong covalency orbitals. We found that the effect of additional U on Ge 4p and
between Cr d-orbital and Te p-orbital.22 Considering the failure of Te 5p orbitals is almost negligible. The atomic structures were
the conventional DFT + U approach and an appropriate bandgap relaxed by the spin-polarized PBE (spPBE) method until all the
from a hybrid functional calculation, Menichetti et al. emphasized atomic forces are smaller than 0.01 eV Å−1. A rigorous semicore
an important role of non-local interactions between all relevant pseudopotential of Cr was generated using the 3s2, 3p6, 3d5, and
orbitals.21 Despite many theoretical computations for CGT, each 4s0 ionic configuration. In addition, another rigorous semicore
effect of Hubbard parameter U and exchange parameter J on band pseudopotential of Cr generated using the 3p6, 3d5, and 4s1 elec-
structures has not been examined so far. So, it would be still pre- tron configurations was also tested, confirming the same results.
mature to conclude that the DFT + U approach may be inappropri- The valence configurations for Ge and Te pseudopotentials are
ate to understand the material properties of CGT compounds. It is 3d10, 4s2, 4p2 and 4d10, 5s2, 5p4, respectively. Real-space grids
necessary to investigate band structures altered by U and J parame- were generated with 800 Ry cutoff energy and pseudo-atomic
ters using a more rigorous approach by Lichtenstein et al.33 in orbitals (PAOs) of double-zeta polarization basis set are used to
which U and J are treated separately. expand the electronic wave functions. For the Brillouin-zone inte-
The semiconducting FM ordering from bulk to ML indicates gration, a 16  16  1 k-grid is used for the structure relaxation
that it originates mainly from the intra-layer superexchange cou- and electronic structure calculations.
pling through a Cr–Te–Cr bond forming an angle of 91.4°.27 It Bulk CGT has a rhombohedral symmetry of the space group
implies that the octahedral structure of Te atoms surrounding a Cr R3 with the hexagonal cell parameter a = b = 6.8196 Å and
atom can be very crucial for the Eg and spin ordering. The effect of c = 20.3710 Å at 5 K.27 The monolayer CGT consists of edge-sharing
the octahedron distortion on the Eg and magnetic ordering was CrTe6 octahedrons forming a honeycomb lattice and Ge-dimmers
considered by several experimental and theoretical studies.34,35 The residing at the central hexagon as shown in Fig. 1. A vacuum spacing
pressure-induced spin reorientation transition from the c axis to of 16 Å was used to avoid the interactions between the layers. The
the ab plane was shown to originate from a variation of the Cr–Te ML lattice constant is not yet experimentally known while our
bonding distance owing to pressure.34 The enhancement of 2D fer- relaxed ML CGT parameters by the spPBE method are
romagnetism and the increase of Eg due to tensile strain were also a = b = 6.992 Å with 2.821 Å for Cr–Te distance, 2.660 Å for Ge–Te,
notified by a couple of first-principles calculations studies with a and 85.21° for Te–Cr–Te angle. These values are different from the
simple isotropic Hubbard U correction without J.34,35 However, the corresponding bulk experimental ones: 2.737 (2.765) Å for Cr–Tetop
effect of any specific change in the octahedron, such as the (Cr–Tebottom) distance, 2.558 Å for Ge–Te, and 85.73° (85.95°) for
bonding distance or angular change, on the electronic structure Tetop–Cr–Tetop (Tebottom–Cr–Tebottom) angle, respectively. For spe-
with proper consideration of on-site Coulomb as well as Hund’s cific positions of Cr, Tetop, and Tebottom, see Fig. 1. This indicates
coupling remains unclear. Considering that the exchange interac- that the CrTe6 octahedron has an inversion symmetry in ML CGT,
tion is crucial in determining characters of conduction band states while in the layered CGT compounds rotation angles of the upper
when the Hund’s coupling is strong, it is important to explore a and lower triangles of Te atoms within a single layer differ due to
possible variation of electronic structure with more systematic com- interlayer interactions.
putational methods. We estimated U and J of Cr 3d orbitals by using a PAO-based
In this paper, we present a first-principles study on the elec- constrained DFT (cDFT) method which was implemented into the
tronic and magnetic properties of CGT compounds as functions of SIESTA code and tested for a layered iridate.38 In our method, with
U, J, and thickness. We show that the decreasing energy gap with a constraint that a given number of spin-up electrons are equally
increasing U within the DFT + U method originates from the
enhanced bandwidth associated with the spin-majority empty eg
orbital states as increasing U. With introducing a separate func-
tional for J, we show that the exchange interaction reduces the
bandwidth and simultaneously alters the characteristics of states
associated with its conduction band minimum, resulting in a
proper energy gap value with reasonably estimated U and J. We
also report that the exchange of J is a critical factor in obtaining the
Eg variation under biaxial strains.

II. METHODS
We performed first-principles electronic structure calculations
based on the noncollinear DFT with the additional Hubbard repulsion
of U and exchange interaction of J including the spin–orbit coupling
(SOC). For ionic potentials, the norm-conserving pseudopotentials36 FIG. 1. Atomic structure of the (a) top and (b) side view of monolayer (ML)
were used with the SIESTA code.37 The generalized-gradient approxi- CrGeTe3. The blue circle is for Cr, the cyan for Ge, the orange for the top Te,
mation (GGA), parameterized by the Perdew-Burke-Ernzerhof (PBE) and the olive for the bottom Te. In (a) and (b), the red solid line represents the
hexagonal unit cell. In (a), the local axis for a Cr octahedron is indicated by
formula, was used as an exchange-correlation functional. The rota- orange arrows.
tionally invariant approach33 was used for U calculation for Cr 3d

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-2

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

distributed to all d orbitals and that another given number of spin- structures. In the trigonally distorted CrTe6 octahedron, the t2g
down electrons to the others, the d-electron dependence of the orbitals can be decomposed into je0 1 i, je0 2 i, and ja1 i and eg orbit-
DFT total energy can be regarded as the form of the collinear-spin als into je1 i and je2 i in the local axes, respectively,39 as indicated
double-counting energy functional, i.e., by orange arrows in Fig. 1(a). The kets are defined as
je0 1 i ¼ cos αjxyi  sin αjyzi, je0 2 i ¼ cos αjx2  y2 i þ sin αjzxi,
U XJ
EDFT (n"d , n#d ) ¼ E0 þ nd (nd  1)  nσ (nσ  1), (1) ja1 i ¼ jz 2 i, je1 i ¼ cos αjyzi þ sin αjxyi, and je2 i ¼ cos αjzxi 
2 σ
2 d d sin αjx2  y2 i where α represents the other distortion deviated
from the exact trigonal distortion.39 In the present geometry, α is
where nd ¼ n"d þ n#d is the number of electrons in a d shell of an about 26.57°. In Fig. 2, the t2g and eg bands for Cr atoms are high-
atom, nσd is the number of electrons of spin σ in the d shell, and E0 lighted, simultaneously distinguishing the majority and minority spin
is an energy independent of d electron distribution. Calculation of channel. It turns out that ML CGT has an indirect gap. The valence
the exchange parameter J is straightforward from Eq. (1). In the band maximum (VBM) locates at the Γ point regardless of U and J
case that an even number of electrons, say, 2n, are in a d shell of a values and has 5p orbital characters of Te atoms dominantly. Unlike
given atom, the parameter J is given by the VBM, the conduction band minimum (CBM) having mainly 3d
orbitals of Cr atoms locates on the line connecting the Γ and K
J ¼ EDFT (n"d ¼ n, n#d ¼ n)  EDFT (n"d ¼ n þ 1, n#d ¼ n  1), (2) points and its precise position depends on U and J values.
The t2g and eg orbital-projected DOS by the spPBE, noncollin-
which requires two cDFT calculations. In the case that an odd ear PBE + SOC, and noncollinear PBE + SOC + U method are
number of electrons, say, 2n+1, are in the d shell, the J is given by shown in Figs. 3(a)–3(d), respectively. With the spPBE method, the
t2g and eg bands are roughly separated in energy by the crystal field
2J ¼ EDFT (n"d ¼ n, n#d ¼ n þ 1)  EDFT (n"d ¼ n  1, n#d ¼ n þ 2), (CF) splitting of 10Dq of ∼2.72 eV [the energy between the filled
(3) t2g and empty eg bands in the spin-majority band as drawn in
Fig. 3(a)]. The spin-majority t2g bands and a part of bands associ-
which also requires two cDFT calculations. Then, the on-site ated with eg states are filled due to the pd hybridization. We note
Coulomb parameter U is obtained by using three spin-unpolarized that the CBM has mainly the spin-minority empty t2g state. The
cDFT calculations, estimated Ueff (the energy between the spin-majority filled and
spin-minority empty t2g bands) is 3.20 eV, and the effective JH (the
U  J/2 ¼ EDFT (n"d ¼ n#d ¼ n, ns ¼ 0) energy between the empty spin-majority and empty spin-minority
eg bands) is 1.41 eV. The approximate values for CF, Ueff, and JH
 2EDFT (n"d ¼ n#d ¼ n  1/2, ns ¼ 1)
are estimated using the maximum peak of projected DOS for each
þ EDFT (n"d ¼ n#d ¼ n  1, ns ¼ 2), (4) state.23,40 The SOC shifts the levels of whole bands of the Cr 3d
band toward the lower energy slightly, keeping the CF splitting,
where ns is the number of electrons in a s orbital of the same atom. Ueff, and JH the same. But the Te 5p band in the VBM is almost
For ML CGT, U of 5.97 eV and J of 1.00 eV were obtained from the unaffected by the SOC so that the variation of CBM by the SOC
cDFT calculations using Cr 3d orbitals whose cutoff radius is 1.58 reduces the bandgap from 0.6 eV to 0.36 eV.
Å. Our Ueff (=U – J = 4.97 eV) is larger than a self-consistently cal- The additional U on Cr d orbitals shifts bands in energy
culated one (3.9 eV) in a previous theoretical study.21 This differ- further as shown in Fig. 3(c). For the case of U of 1 eV without J,
ence is reasonable considering the projection-dependent value of the spin-majority filled t2g band moves toward the lower energy,
Ueff. Although U and J values were estimated by cDFT, we investi- while the empty t2g and eg spin-minority bands move toward the
gated electronic structures of ML CGT as U increases from 1 to higher energy. This leads the spin-majority empty eg bands to be at
6 eV when J is 0 or 1 eV to examine the changes of gaps as func- the CBM and induces the larger CF splitting than one by the SOC.
tions of U and J and to compare our results with others. For larger U values more than 1 eV, the spin-majority empty eg
band is still at the CBM as shown in Fig. 3(c). We note that the
III. RESULTS AND DISCUSSION effective U approach20–25 in previous studies is essentially the same
as the current picture without J.
A. Band structures with Hubbard repulsion and With introducing J separately, the band structure alters quali-
exchange interaction tatively. As shown in Fig. 3(d), with including J, unlike the case
We first explore the changes in electronic structures induced by without J shown in Fig. 3(c), the spin-minority empty t2g state
the SOC and then by the DFT + U method. Figure 2 shows the band becomes the CBM state with U of 2 and 3 eV, respectively. In
structures obtained by (a) the spPBE method, (b) the noncollinear general, the spin-majority and minority empty eg bands show J
PBE + SOC, and (c) and (d) the noncollinear PBE + SOC + U in ML dependence, but the rest bands remain almost unaltered. From the
CGT. To compare with previous DFT results,20 the ML lattice projected DOS in Fig. 3, we estimate the approximate values for
parameters taken from the experimental bulk lattice parameters27 CF, Ueff, and JH in the ML CGT as functions of U and J parameters
were used. To investigate the spin-resolved band structures of Cr 3d in Table I. All values show a general increasing trend as U
orbitals, we further consider the projected density of states (DOS) of increases. If we compare the values in the case of U = 1 eV and
Cr 3d orbitals to t2g and eg orbital eigenstates of octahedral atomic J = 0 to those of U = 2 eV and J = 1 eV, the CF and Ueff of the latter

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-3

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

FIG. 2. Band structures of ML CrGeTe3 obtained from (a) spPBE, (b) noncollinear PBE + SOC, (c) noncollinear PBE + SOC + U (U = 1 eV and J = 0 eV), and (d) noncollin-
ear PBE + SOC + U (U = 3 eV and J = 1 eV) methods. In band structures, the t2g (eg) bands for a Cr atom are colored red (blue) and the left (right) panel shows the major-
ity (minority) spin channel. The ML lattice parameters were taken from the experimental bulk lattice parameters.27

are larger than ones of the former while JH shows an opposite than without J. On the other hand, the spin-minority empty t2g
order. This tendency holds for all the other similar comparisons state is lower in energy than the spin-majority empty eg states with
because only the empty eg bands are affected by J for a given U. U of 2–3 eV and J of 1 eV. Thus, within these U and J values, the
We summarize variations of energy band edges associated spin-minority empty t2g state is at the CBM determining the size of
with relevant orbital states as functions of U and J in Fig. 4(a). We the energy gap.
note first that the lower edge of the spin-majority empty eg bands Based on computed values in Table I and Fig. 4(a), we draw
with J is higher in energy than one without J for the same U in schematic band diagrams describing the effects of U with and without
Fig. 4(a). It is shown that the bandwidth of the spin-majority J in Figs. 4(b) and 4(c), respectively. One of the notable observations
empty eg band decreases with J, resulting in the larger Eg with J is the gradually increasing bandwidth of the spin-majority empty eg

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-4

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

FIG. 3. Projected density of states

(DOS) of t2g (red) and eg (blue) bands
for a Cr atom in ML CrGeTe3 by (a)
spPBE, (b) noncollinear PBE + SOC,
(c) noncollinear PBE + SOC + U
(Hund’s parameter J = 0 eV), and (d)
noncollinear PBE + SOC + U (J = 1 eV)
methods. The minus (plus) side in a
panel shows the majority (minority)
spin DOS. In (c) and (d), the
U-dependent DOS profile is presented.
The black and blue dashed lines indi-
cate the upper edge and lower edge of
the majority empty eg bands, respec-
tively. The red dashed lines represent
the upper edge of the majority filled t2g
bands. The black dashed lines are flat
with a zero slope while the blue and
red dashed lines have a negative
slope, implying the increasing band-
width of eg bands as the Ueff (U − J)
increases. In (a)–(d), the ML structure
taken from the experimental bulk lattice
parameter27 was used. The valence
band maximum EVBM associated with
the p orbital of Te atom is set to zero.

band (width between black and blue dashed lines) as U gradually minority empty t2g band should move toward the higher energy
increases regardless of J. Due to the effect of U-dependent band- with increasing U, the gap should increase. Especially, near U of
width, once the spin-majority empty eg state locates at the CBM, 3 eV, the lower edges of the spin-majority empty eg and the spin-
the gap decreases as increasing U, as already discussed in other minority empty t2g bands compete for the CBM.
studies.20–22 On the other hand, as mentioned in the previous par- Having established the critical role of J in the ML, now, we
agraph, with J, the lower edge of the spin-minority empty t2g band focus on energy gaps as a function of thickness. Figure 5(a) sum-
becomes the CBM in the case of U of 2 and 3 eV. Since the spin- marizes the energy bandgaps of ML CGT as a function of U with J
and without J, respectively. As mentioned before, without J, we
TABLE I. Crystal field splitting (CF), effective U parameter (Ueff ), Hund’s coupling obtain the decreasing gap as increasing U (>1 eV), consistent with
(JH) determined from Fig. 3 with respect to U and J values. U, J, CF, Ueff, and JH the previous DFT calculations.20–22 With J of 1 eV, the gap
are in eV units and magnetocrystalline anisotropy energy (MAE) is in meV/f.u., increases as U increases up to 3 eV and then decreases. We note
where f.u. is a formula unit referring to Cr2Ge2Te6. that ML bandgaps with J of 1 eV are larger than those without J for
all U’s up to 6 eV and that the obtained gap is the largest when
U (J = 0) 1 2 3 4 5 6 U = 3 eV and J = 1 eV. This general gap behavior holds for the mul-
CF 2.99 3.29 3.71 4.02 4.41 4.63 tilayered structures as shown in Figs. 5(b)–5(d). Ultimately, for
Ueff 3.87 4.6 5.42 6.1 7.29 8.34 bulk CGT, the system becomes to be metallic if U is larger than
JH 1.81 2.06 2.50 2.95 3.42 4.07 3 eV without J, agreeing with previous studies.20–22 Since the size of
MAE −0.07 0.27 0.65 1.05 1.43 1.83 energy gap systematically decreases as the number of layers
U (J = 1) 2 3 4 5 6 increases for a given set of U, the erroneous metallic behavior of
CF 3.23 3.89 4.35 4.61 4.89 bulk CGT with increasing U originates from the inaccurate descrip-
Ueff 3.96 5.53 6.45 7.16 7.93 tion of electronic structure of the single layer by the simple effective
JH 0.75 1.08 1.42 2.03 2.50 DFT + U approach. The largest gap in our bulk CGT is 0.37 eV
MAE 0.20 0.32 0.54 0.85 1.19 when U is 2 eV and J is 1 eV and the second-largest Eg is 0.34 eV
when U is 3 eV and J is 1 eV. Both the values are within the

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-5

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

FIG. 4. (a) U and J-dependent variation of band peak (circle), majority/minority lower band edge (triangle), majority spin upper band edge (square) of t2g (red) and eg
(blue) bands of a Cr atom in ML CGT in the noncollinear PBE + SOC + U method. The open (closed) symbols represent J of 0 eV (1 eV). The ML lattice parameters were
taken from the experimental bulk lattice parameters.27 Schematic band structure figures of t2g (red) and eg (blue) bands for Cr atoms as U increases when (b) J is 0 (eV)
and (c) J is 1 (eV). In (b) and (c), the left (right) side of vertical lines represents the majority (minority) spin.

experimental bulk gap of 0.24,28 0.38,29 or 0.7 eV.23,30 Our com- decreases. The increasing magnetic moment per Cr atom with U is
puted energy gaps of ML and bulk CGT are also consistent with caused by the enhanced DOS of spin-majority filled eg bands and
our cDFT estimation described above and with those of 0.33 eV by the diminishing DOS of spin-minority filled eg bands at each J
and 0.56 eV in the previous HSE06+SOC calculations, respec- as shown in Fig. 3(d).
tively.20 This indicates that the proper U and J values for the pro- We compute magnetocrystalline anisotropy energy (MAE)
jected orbitals are important in simulating electronic and magnetic of ML CGT for all U and J values. The MAE can be obtained by
properties of these compounds. On the other hand, as shown in subtracting the total energy with magnetization along the
Fig. 2(d), at these U and J values, both the spin-minority empty t2g out-of-plane direction from the one along the in-plane direction
state and spin-majority empty eg state are at the CBM while in the (MAE = Ein-plane − Eout-of-plane). All results are summarized in
HSE06+SOC calculations20,21 the latter is at the CBM. Table I. For almost all cases, our MAEs are positive, implying
Regarding on magnetic moments, our calculated values of ML that the magnetization is along the out-of-plane direction. The mag-
and bulk CGT with U = 3 eV and J = 1 eV are 3.51 and 3.47 μB/Cr, nitude and direction of our MAEs are consistent with a previous
respectively, agreeing with previous results.20–22 The calculated study using HSE calculation21,41 and those20 using DFT + U approach
moments are larger than 3 μB/Cr while the experimental bulk mag- including spin–orbit coupling while other HSE or U + SOC calcula-
netic moment per Cr atom is of 2.827 or 3 μB/Cr.30 As U increases, tions reported the negative MAEs.21,35,42
the magnetic moment per Cr atom linearly increases and one of Our calculated exchange coupling parameters J1, J2, and J3 are
the Ge atom does not change while the moment of Te atom slightly −4.773 meV, −0.324 meV, and −0.646 meV for ML CGT when

FIG. 5. U and J-dependent bandgap

profile for (a) ML (b) bilayer (BL), (c)
trilayer (TL), and (d) bulk CrGeTe3
in the noncollinear PBE + SOC + U
method. The open squares represent
the case of J = 0 (eV) and the closed
squares for J = 1 (eV). The experimen-
tal bulk lattice parameters were used
for the ML, BL, TL, and bulk.27

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-6

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

FIG. 6. (a) Strain-dependent distances

between Cr and Te atoms (closed
black squares) and Ge and Te (open
red circles) in ML CrGeTe3. The angle
between neighboring bonds of Te–Cr
atoms indicated by a red arrow in
Fig. 1(b) is drawn by closed blue
circles for ML CrGeTe3. (b) Variation of
Mulliken valence charge density ρ (red
symbol lines) and magnetic moment
(blue symbol lines) as a function of
strain for Cr (square), Ge (circle), and
Te (triangle) atoms. In (b), the open
(closed) symbol indicates the case of
the Coulomb parameter U = 3.0
(4.0) eV and the Hund’s parameter
J = 0.0 (1.0) eV. Bandgap Eg as func-
tions of strain and U when (c) J is 0 eV
and (d) J is 1 eV.

U = 3 eV and J = 1 eV, respectively. We calculated the exchange cou- Under the applied biaxial strains, the Eg increases monotoni-
pling parameters using the four-state mapping analysis approach43–45 cally for all U’s with and without J as shown in Figs. 6(c) and 6(d).
using 3 × 3 × 1 supercells. Our J1 of −4.773 meV is consistent with a In cases of U = 2 and 3 eV with J = 1 eV in Fig. 6(d), the slopes of
value of −6.24 meV21 from a previous HSE computation. gap variation with respect to U are steeper than the others. This is
because the t2g orbital states of CBMs for these U and J values are
more sensitive to strain than the eg states for other U’s. In addi-
B. Effect of the CrTe6 octahedron distortion on the tion, we tested the energy gap variation with constrained relaxa-
bandgap tions such that under a given strain the angle and distances
Now, we investigate the strain effect. Our results and previ- in Fig. 6(a) are relaxed separately while fixing one of them
ous studies34,35 demonstrate that the energy gap of ML CGT with (not shown here). From this, we confirm that variation of the
ferromagnetic order can be controlled by the biaxial strain. Here, Te–Cr–Te angle, not the bonding distances, is a main cause of the
the compressive and tensile strains within 2% are considered. strain-induced gap changes, thus ensuring again that the local
With increasing strains, the Cr–Te distance, Ge–Te distance, and interactions within the octahedron are critical in determining the
the Te–Cr–Te angle are almost linearly increased as shown in energy gap size.
Fig. 6(a). The Mulliken charge (ρ)46 and the magnetic moment In particular, the Eg variation with J in Fig. 6(d) does not
per Cr, Ge, and Te atom are almost independent of the strain as show any possibility of the strain-induced semiconductor-to-metal
shown in Fig. 6(b). transition while the gap variation without J in Fig. 6(c) indicates

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-7

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

the transition. Wang et al.35 showed that the strain-induced 10

J. U. Lee, S. Lee, J. H. Ryoo, S. Kang, T. Y. Kim, P. Kim, C. H. Park, J. G. Park,
semiconductor-to-metal transition was observed near 2% com- and H. Cheong, Nano Lett. 16, 7433 (2016).
11
pressive strains by the spPBE + SOC method. Since the gap D. Zhong, K. L. Seyler, X. Linpeng, R. Cheng, N. Sivadas, B. Huang,
E. Schmidgall, T. Taniguchi, K. Watanabe, M. A. McGuire, W. Yao, D. Xiao,
decreases systematically as the thickness increases as shown in
K. M. C. Fu, and X. Xu, Sci. Adv. 3, e1603113 (2017).
Fig. 5, we may deduce wrongly that few layers CGT could be 12
M. Bonilla, S. Kolekar, Y. Ma, H. C. Diaz, V. Kalappattil, R. Das, T. Eggers,
metallic under weak biaxial strains. Our calculations with sepa- H. R. Gutierrez, M. H. Phan, and M. Batzill, Nat. Nanotechnol. 13, 289
rate exchange interactions clearly rule out such a possibility for (2018).
few layers and bulk CGTs, partly agreeing with a recent study 13
D. J. O’Hara, T. Zhu, A. H. Trout, A. S. Ahmed, Y. K. Luo, C. H. Lee,
employing hybrid functionals.35 M. R. Brenner, S. Rajan, J. A. Gupta, D. W. McComb, and R. K. Kawakami,
Nano Lett. 18, 3125 (2018).
14
IV. CONCLUSION T. J. Williams, A. A. Aczel, M. D. Lumsden, S. E. Nagler, M. B. Stone,
J. Q. Yan, and D. Mandrus, Phys. Rev. B 92, 144404 (2015).
We investigated each effect of the local Coulomb repulsion 15
S. Manzeli, D. Ovchinnikov, D. Pasquier, O. V. Yazyev, and A. Kis, Nat. Rev.
of U and Hund’s coupling of J on the electronic and magnetic Mater. 2, 17033 (2017).
16
structures of a layered ferromagnetic semiconductor CrGeTe3. M. R. Molas, C. Faugeras, A. O. Slobodeniuk, K. Nogajewski, M. Bartos,
Our results show the importance of the coupling J with reason- D. M. Basko, and M. Potemski, 2D Mater. 4, 021003 (2017).
able U values for proper bandgaps of CGT compounds with and
17
W. S. Yun, S. W. Han, S. C. Hong, I. G. Kim, and J. D. Lee, Phys. Rev. B 85,
without external strain. From a series of ab initio calculations, we 033305 (2012).
18
estimate that the on-site Hubbard repulsion of about 3–4 eV and C. Gong, L. Li, Z. Li, H. Ji, A. Stern, Y. Xia, T. Cao, W. Bao, C. Wang,
Y. Wang, Z. Q. Qiu, R. J. Cava, S. G. Louie, J. Xia, and X. Zhang, Nature 546,
the exchange interaction of 1 eV are quite suitable values for the
265 (2017).
two important parameters for further theoretical studies on CGT 19
X. Zhang, Y. Zhao, Q. Song, S. Jia, J. Shi, and W. Han, Jpn. J. Appl. Phys. 55,
compounds. 033001 (2016).
20
Y. Fang, S. Wu, Z. Z. Zhu, and G. Y. Guo, Phys. Rev. B 98, 125416 (2018).
ACKNOWLEDGMENTS 21
G. Menichetti, M. Calandra, and M. Polini, 2D Mater. 6, 045042 (2019).
22
S. Kang, S. Kang, and J. Yu, J. Electron. Mater. 48, 1441 (2019).
Y.-W. Son was supported by the NRF of Korea (No. 23
Z. Hao, H. Li, S. Zhang, X. Li, G. Lin, X. Luo, Y. Sun, Z. Liu, and Y. Wang,
2017R1A5A1014862, SRC program: vdWMRC Center) and KIAS Sci. Bull. 63, 825 (2018).
individual Grant (No. CG031509). E. Ko was supported by a 24
N. Sivadas, M. W. Daniels, R. H. Swendsen, S. Okamoto, and D. Xiao, Phys.
KIAS individual Grant (No. CG075001). E. Ko thanks Beom Rev. B 91, 235425 (2015).
Hyun Kim, Se Young Park, and Hyungjun Lee for useful discus- 25
X. Li and J. Yang, J. Mater. Chem. C 2, 7071 (2014).
26
sions. This work was supported by the Center for Advanced B. Siberchicot, S. Jobic, V. Carteaux, P. Gressier, and G. Ouvrard, J. Phys.
Computation at KIAS and by the National Supercomputing Chem. 100, 5863 (1996).
27
Center with supercomputing resources including technical support V. Carteaux, D. Brunet, G. Ouvrard, and G. Andre, J. Phys. Condens. Matter
(No. KSC-2019-CRE-0208). 7, 69 (1995).
28
M. Suzuki, B. Gao, K. Koshiishi, S. Nakata, K. Hagiwara, C. Lin, Y. X. Wan,
H. Kumigashira, K. Ono, S. Kang, S. Kang, J. Yu, M. Kobayashi, S. W. Cheong,
DATA AVAILABILITY and A. Fujimori, Phys. Rev. B 99, 161401 (2019).
29
The data that support the findings of this study are available Y. F. Li, W. Wang, W. Guo, C. Y. Gu, H. Y. Sun, L. He, J. Zhou, Z. B. Gu,
from the corresponding author upon reasonable request. Y. F. Nie, and X. Q. Pan, Phys. Rev. B 98, 125127 (2018).
30
H. Ji, R. A. Stokes, L. D. Alegria, E. C. Blomberg, M. A. Tanatar, A. Reijnders,
L. M. Schoop, T. Liang, R. Prozorov, K. S. Burch, N. P. Ong, J. R. Petta, and
REFERENCES
R. J. Cava, J. Appl. Phys. 114, 114907 (2013).
31
1
K. S. Burch, D. Mandrus, and J.-G. Park, Nature 563, 47 (2018). S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, and A. P. Sutton,
2
J. G. Park, J. Phys. Condens. Matter 28, 301001 (2016). Phys. Rev. B 57, 1505 (1998).
3 32
Y. Deng, Y. Yu, Y. Song, J. Zhang, N. Z. Wang, Z. Sun, Y. Yi, Y. Z. Wu, S. Wu, A. G. Petukhov, I. I. Mazin, L. Chioncel, and A. I. Lichtenstein, Phys. Rev. B
J. Zhu, J. Wang, X. H. Chen, and Y. Zhang, Nature 563, 94 (2018). 67, 1531061 (2003).
4
A. K. Geim and I. V. Grigorieva, Nature 499, 419 (2013). 33
A. I. Lichtenstein, V. I. Anisimov, and J. Zaanen, Phys. Rev. B 52, R5467 (1995).
34
5
M. Xu, T. Liang, M. Shi, and H. Chen, Chem. Rev. 113, 3766 (2013). Z. Lin, M. Lohmann, Z. A. Ali, C. Tang, J. Li, W. Xing, J. Zhong, S. Jia,
6
G. R. Bhimanapati, Z. Lin, V. Meunier, Y. Jung, J. Cha, S. Das, D. Xiao, Y. Son, W. Han, S. Coh, W. Beyermann, and J. Shi, Phys. Rev. Mater. 2, 051004 (2018).
35
M. S. Strano, V. R. Cooper, L. Liang, S. G. Louie, E. Ringe, W. Zhou, S. S. Kim, K. Wang, T. Hu, F. Jia, G. Zhao, Y. Liu, I. V. Solovyev, A. P. Pyatakov,
R. R. Naik, B. G. Sumpter, H. Terrones, F. Xia, Y. Wang, J. Zhu, D. Akinwande, A. K. Zvezdin, and W. Ren, Appl. Phys. Lett. 114, 092405 (2019).
N. Alem, J. A. Schuller, R. E. Schaak, M. Terrones, and J. A. Robinson, ACS 36
N. Troullier and J. L. Martins, Phys. Rev. B 43, 1993 (1991).
37
Nano 9, 11509 (2015). D. Sánchez-Portal, P. Ordejón, E. Artacho, and J. M. Soler, Int. J. Quantum
7
M. Gibertini, M. Koperski, A. F. Morpurgo, and K. S. Novoselov, Nat. Chem. 65, 453 (1997).
38
Nanotechnol. 14, 408 (2019). N. Lee, E. Ko, H. Y. Choi, Y. J. Hong, M. Nauman, W. Kang, H. J. Choi,
8
B. Huang, G. Clark, D. R. Klein, D. MacNeill, E. Navarro-Moratalla, Y. J. Choi, and Y. Jo, Adv. Mater. 30, 1805564 (2018).
39
K. L. Seyler, N. Wilson, M. A. McGuire, D. H. Cobden, D. Xiao, W. Yao, D. I. Khomskii, Transition Metal Compounds (Cambridge University Press,
P. Jarillo-Herrero, and X. Xu, Nat. Nanotechnol. 13, 544 (2018). 2014).
9 40
B. Huang, G. Clark, E. Navarro-Moratalla, D. R. Klein, R. Cheng, K. L. Seyler, E. Pavarini, E. Koch, D. Vollhardt, and A. L. Eds, The LDA+DMFT Approach
D. Zhong, E. Schmidgall, M. A. McGuire, D. H. Cobden, W. Yao, D. Xiao, to Strongly Correlated Materials Modeling and Simulation (Forschungszentrum
P. Jarillo-Herrero, and X. Xu, Nature 546, 270 (2017). Jülich GmbH Institute for Advanced Simulations Grafische, 2011).

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-8

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

41
H. L. Zhuang, Y. Xie, P. R. C. Kent, and P. Ganesh, Phys. Rev. B 92, 035407 (2015). 44
H. J. Xiang, E. J. Kan, S. H. Wei, M. H. Whangbo, and X. G. Gong, Phys. Rev. B
42
C. Xu, J. Feng, H. Xiang, and L. Bellaiche, Npj Comput. Mater. 4, 57 (2018). 84, 224429 (2011).
43
H. Xiang, C. Lee, H. J. Koo, X. Gong, and M. H. Whangbo, Dalton Trans. 42, 45
M. H. Whangbo, H. J. Koo, and D. Dai, J. Solid State Chem. 176, 417 (2003).
823 (2013). 46
R. S. Mulliken, J. Chem. Phys. 23, 1833 (1955).

J. Appl. Phys. 128, 123901 (2020); doi: 10.1063/5.0015566 128, 123901-9

Published under license by AIP Publishing.