Journal of Superconductivity and Novel Magnetism (2019) 32:805–819

https://doi.org/10.1007/s10948-018-4742-4

ORIGINAL PAPER

Magnetic Behavior of Superconductor 2H-NbSe2 Intercalated

with Iron: First Principle Study
Hemza Kouarta1 · Kamel Zanat2 · Hafid Belkhir1

Received: 9 May 2018 / Accepted: 17 May 2018 / Published online: 15 June 2018
© Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract
We investigated the structural, electronic and magnetic properties of iron-intercalated niobium selinium, F ex NbSe2 for
x = 0 and 0.25, from first principle calculations. The tendency towards the localization in superconductors criterion which
has been extracted from our results placed the NbSe2 compound in vicinity of Fe-based superconductors. The DFT plus on-
site coulomb repulsion U (DFT+U) and the exact exchange for correlated electrons (DFT+EECE)-based hybrid functional
were used including the spin-orbit coupling (SOC) to understand the role of correlation and SOC effects in this type
of systems. The calculation predicts the correct antiferromagnetic (AF) ground state, from both the generalized gradient
approximation (GGA) and GGA+U. A shift up in the Fermi energy was observed after the intercalation of 2H-NbSe2 with
iron, indicating a charge transfer from Fe to the host compound. The obtained magnetic moment of Fe is enhanced by a
correlation effect in both DFT+U and DFT+EECE over the small value from GGA-only. Moreover, a large unquenched
orbital magnetic moment is saturated to morb ∼ 0.6μB under a moderate correlation effect of Ueff ∼ 2.0eV . We also
show that the correlation effect is important along side with SOC in order to get a true picture for the band filling consistent
with that of the crystal field splitting and a large unquenched morb which has been found experimentally in such systems.
Therefore, the combination of correlation and SOC effects is a decisive choice for further studies of such systems.

Keywords Transition metal dishalcogenides · Intercalation · Electronic band structure · Electron-electron interactions ·
Spin-orbit effects

1 Introduction faster resonance (in THz) comparably to that of ferromagnet

(in GHz), in order to generate new spintronic devices [2].
The growing interest of the scientific community in Materials with layered structure are of the most subjected
spintronics, topological insulators, the recent discovery of topics to intensive research studies in the past decade.
controlling the antiferromagnet [1], and the enhancement Transition metal dichalcogenides (TMDCs) are among
of ferromagnet using the antiferromagnetic coupling where this attractive family that exhibit fascinating and unusual
spin-orbit coupling plays a crucial role opens the way to physical properties [3], such as charge density wave (CDW),
a new research area of spin-orbitronic. By focusing on superconductivity, and the competition between them at
different aspects of the antiferromagnet, we can pursue the low temperature [4]. Because of the fast development
new progress in magnetic recording, mainly because of the regarding the new physical properties on graphene, much
experimental and theoretical research has been made
recently. This renewed the interest in these types of
 Hemza Kouarta materials. Transition metal dichalgogenides having the
hemzakouarta.phy@gmail.com
formula 2H-MX2 (M: Transition metal, X: S, Se, and
1
Te) with hexagonal structure, also known as a quasi-two
Faculty of Science, Department of Physics, Laboratory
Studies of Surfaces and Interfaces of Solid Materials
dimensional layered structure, are formed by two slabs,
(LESIMS), University of Badji Mokhtar, P.O. Box 12, Annaba each slab is a sandwich of (X-M-X) atomic layers (see
23000, Algeria Fig. 1). The atomic layers in the sandwich are strongly
2 Laboratoire de Physique, Université 8 Mai 1945, BP 401, related by a covalent bonding while the slabs are holding
24000 Guelma, Algeria by weak van der Waals (vdW) bondings [3, 4]. The intrinsic
806 J Supercond Nov Magn (2019) 32:805–819

Fig. 1 a Unit cell of the hexagonal crystal structure of 2H-NbSe2 . b The constructed 2×2×1 supercell. c The unit cell of F e1/4 NbSe2 intercalated
with F e

property of vdW gap, in 2H-MX2 structure, makes these Recent experimental work on F ex T aS2 has brought out a
compounds a proper choice to form a new 2D system [5] new magnetic phenomenon like the sharp switching in the
or as a parent material for intercalation compounds [6] magnetization curve, a very large magnetoresistance driven
with different types of molecules and atoms [7], in order from both anisotropic magnetoresistance and magnon
to get new devices exhibiting highly suitable properties for suppression [14–17].
particular technological applications. 2H-NbSe2 is one of the most famous superconductors
Niobium and tantalum dichalcogenides M(Nb, Ta)X2 are in the family of transition metal dichalcogenides certainly
the most interesting members, since they are metals showing because of its higher Tc among others [4]. Indeed, the
superconductivity at low temperature [4, 7, 8]. Intercalating intensive investigation to understand its critical point
them with the first row of transition metals (T = Cr, Mn, between CDW and the origin of superconductivity [18–23]
Fe, Co, and Ni) to form Tx MX2 , they present interesting reveals always a new physical phenomenon that exhibits in
electrical and magnetic properties associated to the change this compound [24]. The insertion of a few concentration
of structural and chemical crystal, depending strongly on the of Fe in this particular system, 0.01 ≤ x ≤ 0.1, shows a
amount of the intercalation (x%) [6, 9–11]. The presence of dramatic change in its properties. First of all, it shows the
3D transition metal atoms in these compounds makes them collapse of superconductivity [25, 26]. Scanning tunneling
correlated materials [12] with the appearance of a different microscope (STM) measurement shows a decrease in the
magnetic ordering at high concentration with the formation energy gap of the CDW, a formation of a 3a0 × 3a0
of superlattices [6, 9, 11]. The magnetic configuration superlattice below a temperature of T = 35K, and the
depends generally on the 3D atoms intercalated in the vdW appearance of Kondo-like behavior [26–28]. At a higher
gap. It gives the same magnetic ordering, ferromagnetic concentration, x ≥ 0.20, the structure of the superlattices
(FM), or antiferromagnetic (AF), for both Nb and Ta becomes a 2a0 × 2a0 order, with the observation of charge
members. Nevertheless, the only particularity seems to be transfer [27, 28]. The antiferromagnetic order was identified
in the case of iron (Fe), where the choice of the host leads at x = 25% in F ex NbSe2 , where the Néel temperature
to different phases, FM in F ex T aX2 and AF in F ex NbX2 reached a maximum, TN = 175K [29] and TN =
(x = 1/4, 1/3). Furthermore, the increase of concentration 155K [9], then decreases to 122K in F e1/3 NbSe2 [27].
(x%) leads to the transition between the two magnetic The reported effective and spin magnetic moments are very
phases [13]. The preferential direction of the magnetic large in F e1/4 NbSe2 , meff = 5.22μB [29]; 5.17μB [6, 9];
moment changes from perpendicular alignment to the c- and mspin = 4.9μB [6]. Moreover, significant spin-orbit
axis for “V, Cr, and Mn” to the parallel alignment for the coupling was revealed from the anisotropic behavior of the
case of “F e” and reverts again to perpendicular for “Co” susceptibility [9], similar to what is observed in the case of
and “Ni” intercalates [9, 10]. Again, the Fe intercalation F e1/3 NbS2 and F e1/3 T aS2 [9–11].
into MX2 is a particularity case. This difference in the easy Particularly, these results were confirmed recently in
axis of magnetization leads to an anomalous behavior in the F e1/4 T aS2 by K. Ko et al. [30], using X-ray absorption
Hall effect, electrical resistivity, and susceptibility [6, 9–11]. spectroscopy (XAS), X-ray magnetic circular dichroism
J Supercond Nov Magn (2019) 32:805–819 807

(XMCD), angle-resolved photoemission spectroscopy coupling using the DFT+U and EECE approximations with
(ARPES), many-body model calculation, and first princi- and without spin orbit coupling. Finally, the conclusion is
ples band calculation using WIEN2K code. The reported reported in Section 4.
orbital magnetic moments are very large and unquenched,
μorb = 1.0μB , coming from a strong spin-orbit cou-
pling (SOC). This results in a large magnetocrystalline 2 Computational Details
anisotropy and a huge anisotropic field (HA = 60T ) that
depasess the one of a rare earth [30, 31]. This is associated We have performed first-principles density functional the-
with an Ising-type [30] and Ruderman-Kitte-Kasuya-Yosida ory calculations [35, 36] using the all electron full poten-
(RKKY) indirect exchange interaction between the far sep- tial linear-augmented plane wave method plus the local
arated Fe-spin, which was suggested earlier as a smaller orbitals technique (FP-LAPW+lo) [37, 38] as implemented
super-exchange interaction [9, 32]. in WIEN2k code [34]. The exchange-correlation potential
Recently, Y. Koh et al. [33] found a smaller TN = 106K for the host, 2H-NbSe2 , was chosen to be the revised van
in F e1/4 NbSe2 , from both curves of magnetization and der Waals density functional (rev-vdW-DF2), as proposed
resistivity vs temperature. The obtained band structure and by I. Hamada [39], over the ordinary local density and
Fermi surface from the ARPES data show folded bands generalized gradient approximations (LDA and GGA). For
due to the superlattice of 2a0 × 2a0 and a shift in the the ternary, F e1/4 NbSe2 , the exchange-correlation poten-
Fermi energy (EF ) indicating a charge transfer from F e tial was taken in the framework of the GGA with the
atom to the host material NbSe2 . Unfortunately, neither parameterization of Perdew-Burke-Ernzerhof (PBE) [40].
spin nor orbital magnetic moments were reported in their The optimized structural parameters, used for host mate-
work. On the other side, knowing the electronic structure rial 2H-NbSe2 , are obtained from the minimization of total
is a fundamental key to clarify the change in the physical energy versus volume and c/a with the relaxation of the
properties made by the intercalation in such systems. We atomic positions, starting from the experimental parameters,
believe that there is no detailed theoretical investigation a = b = 3.44Å, c = 12.55Å and zSe = 0.117 [41]. In
made yet on the Fe1/4 NbSe2 . the case of the intercalated compound F ex NbSe2 , a super-
In this work, we present a detailed ab initio calculation of cell of 2 × 2 × 1 was employed, Fig. 1b, in order to deal
the 2H-NbSe2 and F e1/4NbSe2 for structural, electronic, with the amount of intercalation (x = 25%) where F e
and magnetic properties using a full potential linear - atoms are placed in the vdW gaps, between Se layers. The
augmented plane wave (FP-LAPW) method within the new unit cell remains hexagonal, containing 8 atoms of
WIEN2k code [34]. We found that the degree of localization Nb, 16 atoms of Se, and 2 atoms of Fe. In the AF phase,
in superconductors in NbSe2 compounds puts it in the the magnetic interactions is assumed to be FM intra- and
vicinity of iron-based superconductors. The AF magnetic AF inter-Fe layers [9, 10]. The structural parameters of
order was found to be the magnetic ground state of the supercell calculations are obtained by a full geometri-
F e1/4 NbSe2 after the geometrical optimization using GGA cal optimization with total energy minimization for volume,
and GGA+U, which is consistent with the experiments [9, c/a ratio, and force relaxations for the internal atomic posi-
10, 33], with a remarkable dependency of the c-parameter tion parameters. Bulk modulus was determined by fitting
on the considered magnetic order. Beside that, the calculated the energy-volume curve to the Birch-Murnaghan equation
Bulk modulus B0 shows a large difference between of state (BM-eos) [42]. The muffin-tin radii (Rmt ) were
magnetic phases suggesting the possibility of changing the chosen to be 2.3, 2.2, and 2.1 a.u. for the Nb, Se, and Fe
magnetic order by applying a pressure. The correlation atoms, where the valence states used are 4d4 5s1 for Nb,
effect and spin-orbit coupling were found to be important 3d10 4s2 4p4 for Se, and 3d6 4s2 for Fe. Moreover, the semi-
factors for further studies of such systems, especially the core states 4s2 4p6 and 3p6 for Nb and Fe, respectively, are
magnetocrystalline anisotropy properties, where we have treated as a valence using WIEN2K local orbital technique.
obtained a large unquenched orbital magnetic moment The plane wave expansion cutoff (Kmax ) of the wave func-
after GGA+U+SOC. Moreover, the bonding nature of the tion in the interstitial region is controlled by the parameter
pure and the intercalated system and the effect of the Rmt × Kmax . Values of 7 and 9 were employed for the
hybridization between orbitals are illustrated and discussed. host and the ternary compounds, respectively (where Rmt is
The remainder of the paper is organized as follows. the smallest muffin-tin radius). The cutoff for the Fourier
Section 2 describes the simulation technique used in this expansion of the potential and the charge density is equal
paper. In Section 3, we present our results beginning with to Gmax = 16Ry. The integration of the Brillouin zone
the structural properties according the magnetic order, then was performed using 32 × 32 × 14 k-points mesh for the
the effect of the electron correlation and the spin-orbit host structure and 16 × 16 × 7 for all the supercell cal-
808 J Supercond Nov Magn (2019) 32:805–819

culations. The relativistic effect was taken into account, The stacking along the c-axis goes from Se-Nb-Se, in this
including SOC as a perturbation for the valence electrons sequence, forms a single slab with a strong covalent bonding
and the full one for the core electrons [43]. The SOC is between the three atomic layers and a very week bonding
switched off when we need to see its effect. The LDA+U between the slabs of vdWs type, where two single slabs
method is used here for taking into account the electron cor- define the primitive cell of this structure.
relation effects present in this type of materials [12, 30, 44] In Table 1, we present the calculated lattice parameters
in the context given by Anisimov et al. [45, 46], with the for the host compound. The failure of LDA and GGA is very
approximation correction for the self-interaction correction obvious due to the absence of a term describing correctly
(SIC). In this implementation, only the effective coulomb the important vdW gap existing in this layered material.
parameter Ueff = U − J is meaningful, where U is the The underestimation in LDA and the overestimation in
in-site coulomb interaction parameter and J is the Hunds GGA of crystal structural parameters are very clear from
coupling parameter. The correlation effect is applied only to Table 1, especially over the z-axis where the vdW forces
the strongly correlated electrons at 3d states of Fe atoms. We play an important role between the slabs. Compared to the
have also used the exact exchange for correlated electrons experimental value, the c-parameter is smaller by about
(EECE) functional as introduced by Novak et al. [47]. This 0.50Å in LDA and has a large difference of ∼ 1.15Å in
hybrid method has not been used before in such systems, GGA. This weak vdW interaction is one of the significant
but its efficiency is already demonstrated in other corre- shortcomings of DFT in the treatment of the structural
lated materials. The exact exchange, or the Hatree-Fock properties. Several methods were proposed, with affordable
exchange, is applied only to specific orbitals and inside the computational effort, to take into account vdW forces [50,
atomic sphere. Here, the 3d orbital of Fe, by a fraction (α) of 51] (for a review, see ref [52]). Among the many functionals
typically 25% replaces the local density exchange obtained bench marked, rev-vdW-DF2 [39] is one of the methods
from LDA or GGA. recently proposed with the best performance in description
of solids [51, 52]. Indeed, lattice parameters obtained by this
method applied to NbSe2 , see Table 1, are in a very good
3 Results and Discussion accordance with the experimental values [3, 41, 48, 49] and
other DFT calculations including vdW correction [53, 54]
3.1 Structural Properties for both in-plane “a” and out-of-plane “c” parameters. For
the z-coordination of Se atom, we have found a value of
Figure 1a shows the bulk structure of NbSe2 , which has LDA = 0.107c with LDA and zGGA = 0.132c with GGA,
zSe Se
exp
an hexagonal structure with space group P 63 /mmc [3, which are out of range of all experimental values zSe =
4, 48] containing 2 f.u./cell, two atoms of Nb, each one 0.116c − 0.125c [41, 48, 49], and the latest theoretical ones
theo = 0.116c − 0.118c [20, 53, 54]. However, the GGA
surrounding by six atoms of chalcogenides in trigonal zSe
prismatic symmetry [3, 41, 48, 49] at the Wychoff’s position value is approximately the same reported in ref [18], zSe =
2b(0, 0, ±1/4), and four atoms of Se at 4f (1/3, 1/3, z). 0.134c, but we did not find any experiments confirming

Table 1 Structural parameters for NbSe2 and F e1/4 NbSe2 after full optimization of lattice parameters (a and c or Veq and coa) and internal
parameter for atomic position (zSe ). The bulk modulus B0 is obtained by fitting the results to BM-eos. mspin represents the spin magnetic moment
of one Fe atom. E is the total energies of F e1/4 NbSe2 in the NM, FM, and AF phases, taking the latter as the reference of energy. The
experimental data are taken from ref [29, 41]

a (Å) c (Å) c/a zSe Veq (Å3 ) B0 (GP a) mspin (μB ) E(meV )

2H-NbSe2 LDA 3.400 12.090 3.557 0.107 121.10 53 / /

GGA 3.477 13.690 3.937 0.132 143.38 8 / /
rev-vdW-DF2 3.456 12.531 3.625 0.116 129.66 37 / /
Expta 3.445 12.55 3.642 0.116 128.77 –
F e1/4 NbSe2 NM 6.926 12.362 1.78 0.108 513.62 79 0.00 139
(GGA) FM 6.925 12.841 1.85 0.115 541.77 61 2.86 97
AF 6.934 12.819 1.84 0.114 538.35 57 2.83 0.0
Exptb 6.933 12.706 1.83 0.121 528.91 / 5.25

a From [41]
√
b From [29], the magnetic moment represent the meff = 8C1 as extracted from Curie-Wiess magnetic susceptibility, χ = (C1 /T − θ) − C2
J Supercond Nov Magn (2019) 32:805–819 809

such a high theoretical value. The value of zSe obtained from intercalated compound in the NM order is less than of that
rev-vdW-DF2 is in good accordance with the experimental in the host material, either the calculated 12.53Å or the
ones. experimental 12.55Å ones, giving the minimal parameters
The bulk modulus B0 extracted by fitting the energy- for : Se height 0.108Å, Fe−Fe out of plane distance 6.181Å,
volume curves to BM-eos, obtained from total energy and Fe−Se bond length, 2.437Å (see Table 2). For the
calculations of B0 = 8GP a with GGA and B0 = FM configuration, an elongation of the c-lattice with a
53GP a with LDA approximations, is strongly fluctuated larger value 12.84Å, also Se height of 0.115Å, the Fe−Fe
that occurred for the structural parameters. However, the bond distance between slabs 6.420Å, but the in-plane
B0 = 37GP a obtained with rev-vdW-DF2 is comparable lattice parameters are a little small, therefore the Fe−Fe
to the latest theoretical work [53, 54], where we did not in-plane bond distance is smaller. The magnetic order,
find any stated experimental data in the literature. However, therefore, considerably influences the crystalline structure.
the Nb-Se bond length, 2.573Å, the height of the sandwich, This finding shows the high strength of the magnetic
3.31Å, and the vdW gaps, 3.12Å, are very closer to a series interactions in the c-direction. We will remember this later
of experimental results [20, 48, 55]. The accuracy of our when we discuss magnetic interaction couplings and their
previous results gives a good starting point to proceed with direction.
intercalated compound. The stable state, AF order, is a mediate phase between
The supercell of 2 × 2 × 1 employed here gives two good those two orders, NM and FM, where the AF interaction
choices of vacancies, within the vdW gaps, having an octa- of Fe atoms between the layers tends to reduce the c-
hedral symmetries [4, 41] where Fe can occupy interstices, lattice (and the Se height) and therefore the out-of-plane
at Wyckoff positions 2a[(0, 0, 0) and (0, 0, 1/2)]. This is Fe−Fe bond distance, more than the FM interaction. The
illustrated in Fig. 1c. Simulation results for the optimized in-plane lattice parameter in the AF phase is expanded by
structural parameters and the calculated total energies of a small but noticeable value (a larger Fe−Fe in-plane bond
F e1/4 NbSe2 , using the GGA approximation, in the NM, distance) with respect to that in the FM phase. Thus, the
FM, and AF magnetic configurations, are summarized in lattice parameters of AF phase are in good agreement with
Table 1 for a good comparison. The obtained results are in a those of experiments [29], as seen in Table 1. It is clear that
good agreement with the experimental data [9, 29, 33]. the equilibrium volume closer to the experimental finding
The AF phase was found to be the ground state, as stated is that of AF phase (see Table 1). Moreover, the calculated
before in the old experiments [4, 6] and confirmed recently bulk modulus B0 shows a very clear difference related to the
by Y. Koh et al. [33]. Indeed, FM and NM solutions are considered magnetic order. A drop from 79GP a for NM to
higher magnetic states than the AF by EF M = 97meV 61GP a in FM and 57GP a in AF which reflects the collapse
and ENM = 197meV , respectively, and require more of the rigidity upon setting up of magnetic interactions, a
energy to stabilize them (see Table 1). The local spin decrease of about 28% with respect to AF magnetic ground
magnetic moment, in the AF order, mspin = ±2.83μB /F e state. According to these values, it can be possible to tune
atom is very small compared to the experimental value the magnetic behavior by applying a certain pressure to
exp
mspin = 4.9μB /F e atom [6]. We will return to it later the system. It is worth to note that F. Hulliger et al. [29]
in Sections 3.2 and 3.3. The optimized lattice parameters noted that the soft pristine compound 2H-NbSe2 becomes
have a considerable difference between the NM case and hard upon intercalation by 3d-transition metal element. This
the magnetic ones, especially the c-lattice parameter, while trend is well reproduced by our calculated results of bulk
the in-plane lattice parameters have comparable values. modulus where B0 = 37GP a for NbSe2 was increased to
If we ignore the Fe magnetism (no spin-polarization in B0 = 79GP a for intercalated compound F e1/4 NbSe2 .
the NM order), the c-axis is reduced to a smaller value The bond lengths after intercalation, see Table 2, show
12.33Å, hence the c/a ratio, comparing to the experimental an analog to the previous statements; beside the change
value 12.71Å [29] or to the AF and FM orders, 12.84Å from the host, the shorter (larger) the lattices, the smaller
and 12.82Å, respectively. Moreover, the c-axis of the (larger) the bonds, hence a small (big) volume. Thus,

Table 2 Bond lengths between atoms in F e1/4 NbSe2 obtained from full optimization of the geometrical structure in each magnetic phase in Å.
The Nb-Se is equal to 2.573Å in the bulk system, 2H-NbSe2

Fe-Fe(out-plane) Fe-Se(1) (Se(2) ) Fe-Nb(1) (Nb(2) ) Nb(1) -Se(1) (Se(2) ) Nb(2) -Se(2) (Se(1) )

NM 6.181 2.437 (4.214) 3.092 (4.641) 2.603 (4.351) 2.595 (2.650)

FM 6.420 2.535 (4.263) 3.211 (4.724) 2.622 (4.346) 2.581 (2.639)
AF 6.409 2.532 (4.275) 3.205 (4.722) 2.619 (4.377) 2.579 (2.642)
810 J Supercond Nov Magn (2019) 32:805–819

also related to whether taking into account the magnetic various experimental data of angle-resolved photoemission
interactions or not, as suggested before. The bond lengths spectroscopy (ARPES) [23, 24, 33] and other band structure
of Fe with the first closest atoms (Se(1) and Nb(1) ) and the calculations [18–20, 24, 56, 57]. Below the Fermi level, Se-
second ones (Se(2) and Nb(2) ) in NM phase are smaller 4p bands are dominated and mixed with some character
than to those upon putting interactions (FM and AF), but of Nb-4d states which is implying a hybridization between
this leads to an increase of the Nb(1) -Se(1) bonds as the them. In the vicinity of EF , three bands are present, see
F e-Se(1) gets smaller. The Nb(2) -Se(2) bonds are also Fig. 2b, where the SOC is added. The lowest band is Se-
affected. Hence, the surrounding six Se(1,2) atoms around pz state with small empty states near the -point. The two
Nb(2) in the octahedral symmetry are slightly distorted other bands are derived from the doublet dx 2 −y 2 /dxy and the
in the three states, and due to the effective modification singlet dz2 of Nb-4d states. Moreover, those bands have a
between atoms, the Fe with the surrounding atoms has very narrow bandwidth where their minimum positioned in
some modification indicating a distortion of the F eSe6 . M direction and it takes two thirds of this distance from 
We should also mention that the Fe−Se bond lengths are at energy 0.54eV under EF .
smaller than Nb(1) -Se(1) bonds probably due to a formation The Nb-dz2 and Se-pz states intersect many times
of strong bonding between the iron and selenium, as can be without any hybridization due to the different parities of
seen later in the electron density plot (see Fig. 9). This is those bands as stated by Mazin et al. [20].
in good agreement with recent theoretical work of Chan et In Fig. 2b which includes spin-orbit effect, three bands
al. [56] and Li et al. [57, 58] on 2H-NbSe2 intercalated with cross EF through M--K. The upper two bands consist
V, Ti, and Ta, respectively. of bonding and anti-bonding dz2 orbitals mainly in M
direction, as measured from the ARPES data [21, 23] and
3.2 GGA Electronic Structure Properties DFT calculation [22]. We have also a contribution from Nb-
dx 2 −y 2 /dxy around the K-point [20] with a lifting of the
The electronic band structure is essential for understanding degeneracy at this point due to the spin-orbit coupling. The
the modification of electronic and magnetic properties by lowest band is the anti-bonding Se-pz orbital. We should
intercalating it with magnetic atoms. For 2H-NbSe2 , the mention also that Se-pz band is not detected most of the
band structure along high symmetry directions is shown in time in the ARPES measurement due to the broadening
Fig. 2a using GGA-PBE approximation. As can be seen in coming along with the strong dispersion in kz [21, 23, 33].
this figure, the compound exhibits a special band structure The DOS of NbSe2 is plotted in Fig. 3, showing the
which characterizes systems with a layered structure where projected density of states (PDOS) for Nb-4d and Se-4p
intermediate bands located at EF are separated from higher bands. Due to the crystal symmetry of Nb atoms in the
and lower bands by energy gaps. Furthermore, the figure trigonal prismatic, 4d states split into three well-defined
shows bands crossing EF which is an indication of a states, two states having a sharp peaks of DOS at EF ,
metallic behavior of this compound, in accordance with the consisted of dz2 state and a doublet state of dx 2 −y 2 and

Fig. 2 Band structure plot for Nb-dx2-y2/dxy Nb-dz2 Se-pz

NbSe2 along the Brillouin zone Nb-d Se-s Se-p
6 4
high-symmetry directions. a We
neglect spin-orbit coupling, 4
showing the orbital characters of
Nb-4d (circle), Se−4p (square), 2
and Se−4s (diamond). b, A 2
0
closeup of the band structure
around the Fermi energy,
Energy(eV)

-2
Energy(eV)

including spin-orbit coupling,

showing the orbital character of -4 0
the bands crossing the Fermi
level, Nb-dz2 (diamond), -6
Nb-dx 2 y 2 /dxy (circle), and
Se-pz (square). The Fermi level -8
is set at zero energy -2
-10

-12

-14 -4
Γ M K Γ A Γ M K Γ A
(a) (b)
J Supercond Nov Magn (2019) 32:805–819 811

dz2 that gave the covalent bonding in this system [3, 4, 7]. The
dx2-y2 / dxy resulting DOS at the Fermi level is higher which indicates a
1.2
dxz / dyz good conductivity of the material.
The narrowing of the band width with high density of
0.8 states around EF invokes the problem of the localized states
Nb over the itinerant ones. The ratio of the experimental kinetic
PDOS (States / eV)

0.4 energy (extracted from optical measurement or conductiv-

ity) and the band structure Kexp /Kband is commonly used
0 to quantify the tendency towards the localization in super-
pz conductors [59]. It is very interesting that this ratio which
0.6
p x / py is calculated by the optical conductivity, 0.52–0.40 [60], is
also determined directly, although it has less accuracy, by
0.4 the inverse of the renormalization of the calculated LDA
band structure to the measured ARPE-spectra, by about
0.2
Se 2.2 at -point [61]. Thus, a direct comparison between the
ARPES measurement and band structure calculation can be
made by a direct renormalization of the minimum of the
0
-5 -4 -3 -2 -1 0 1 2 3 4 5 d bands exp = 0.19eV to our calculated one band =
Energy (eV) 0.53eV , with exp /band = 0.36 which is consistent with
Fig. 3 Projected density of states of Nb-4d and Se-4p in 2H-NbSe2 that calculated by Rahn et al., 0.45 ± 0.12 [21]. The cal-
using GGA-PBE approximation culated ratio placed our compound in vicinity of iron-based
superconductors.
Upon the intercalation with F e, the band structure
dxy and the last state is a doublet state formed by dyz and became more complicated having a lot of bands because of
dxz which shows a well-separated bonding anti-bonding the larger number of atoms in super-cell. Also, one would
bands with no contribution at EF . The Se−4p states have a expect the folding of the bands due to the supercell. Hence,
doublet state of px and py and a singlet of pz orbital; the we have plotted the band structure for the supercell of
contribution at EF is rather small. These states are strongly 2 × 2 × 1NbSe2 without Fe intercalation (sc-NbSe2 ) for a
hybridized with Nb-4d states in a spread range of energy good comparison. The band structure of sc-NbSe2 is plotted
which is related to the bond length between the two atoms in Fig. 4a. The increase of the number of bands along the

Nb-d Se-p Fe-d Se-p Nb-d

4 4

3 3

2 2

1 1
Energy(eV)

Energy(eV)

0 0

-1 -1

-2 -2

-3 -3

-4 -4
Γ M K Γ A Γ M K Γ A
(a) (b) (c)
Fig. 4 a Band structure plot for the 2 × 2 × 1NbSe2 supercell, show- the primitive Bloch representation. The symbol size represents the
ing the band folding of NbSe2 . b Band structure of F e1/4 NbSe2 in Bloch spectral weight with the corresponding color map. The Fermi
the AF state with the orbital characters of Fe−3d (circle), Nb-4 (dia- level is set at zero energy
mond), and Se-4p(square). c Band structure of supercell unfolded to
812 J Supercond Nov Magn (2019) 32:805–819

high symmetry directions is clearly apparent compared to However, Y. Koh et al. [33] did not observe any intercalant-
that of pristine unit cell. The folding of bands is very clear. derived bands from the ARPES data, from both the Fermi
For example, the bands that were along the -M direction surface and the binding energy which is not consistent with
in NbSe2 at M-point are folded to -point in the supercell, our finding of the GGA band structure calculation.
especially Nb-dz2 bands which cross EF at -M direction For F e1/4 NbSe2 , the PDOS of the AF phase, which
in the middle for the pristine compound are now situated is the ground state found from the GGA calculations, is
at M-point for the supercell. Also, the band that was near shown in Fig. 5. The Fe atoms are surrounded by six Se
the M-point just under the Fermi level, along -M-K, that atoms forming an octahedron slightly distorted between the
belongs to the Nb-dz2 is now at the -point in a reverse layers, due to the elongation in the c-axis with a trigonal site
direction. symmetry of D3 d. The Fe−3d split first to a triplet lower
Figure 4b shows the band structure of the host lying state t2g and a doublet eg state due to strong octahedral
material after intercalating it with F e using GGA- crystal field. Due to the elongated trigonal distortion, the t2g
PBE approximation. By introducing F e, the bands of state splits further to a higher single state of a1g (dz2 ) and
F e (circles) in the upper region, with respect to EF , can a doublet state egπ , with no change to remaining eg doublet
be distinguishable and enter directly below the Nb-4d states [8, 9]. Nb and Se atoms have the same symmetry as in
bands (diamonds), at a forbidden energy interval from 1.0 the pristine compound with the same splitting for 4d and 4p
to 2.0eV . In the lower region, which is located between states, respectively. However, each atom has two site types
−2.0 and −3.0eV , we can see a broad mixing between the depending on the distance from Fe atoms that we called:
Se−4p (square), originated from pristine unit cell, and the Nb(1) , Se(1) (Nb(2) , Se(2) ) near (far from) Fe atom, while
Fe−3d. There is also some contribution from the Nb−4d. Nb and Se without indices represent atoms in the pristine
Near the Fermi level, when going from the upper region compound. Figure 5 shows a clear difference between the
to lower one, we can recognize clear localized bands, flat PDOS of each type of atoms. In this figure, the spin majority
bands with no clear dispersion character, that belongs to channel of F e is almost fully occupied but with more
the F e atoms where the intersection with bands of Nb-
4d is very clear, indicating a hybridization between F e-3d
and Nb-4d bands. We will return to it in more details in 1.2
DOS discussions. These Nb-4d bands shift downwards in 0.8 Fe
0.4
energy remarkably at the -point. Indeed, Nb-dz2 lowered 0
from 1.25eV in NbSe2 to 0.82eV in the intercalated one; -0.4
-0.8 d z2
also, Se-pz bands are dropped totally below EF . The main -1.2 dxz / dyz
contribution of bands to EF crossing is Fe atoms with a dx2-y2 / dxy
0.6 Nb1
lifting of the degeneracy at  and K points just at EF .
0.3
For a qualitative interpretation of our results, we have
PDOS (States / eV)

0
performed the unfolding of the supercell bands for the -0.3
ternary compound by computing the Bloch spectral weight -0.6 Nb2
to obtain the band structure in the BZ of the primitive cell 0.2

from which the supercell is constructed. Nevertheless, the Se1

0.1
band structure in the supercell representation is also needed 0
in the comprehension of the band character of the folded
-0.1
bands as long as we cannot determine it with the unfolding p x / py
technique, while the latter is an essential tool to compare the -0.2 pz
0.3
calculated supercell band structure with host (undoped or no
0.15
intercalated) and experimental ARPES counterpart. 0
Figure 4c shows the Bloch spectrums with their weights -0.15
of unfolded bands obtained from GGA calculation. The -0.3 Se2
bands with a small symbol size (dark-blue) have a negligible -6 -5 -4 -3 -2 -1 0 1 2 3 4 5
Energy (eV)
weight, while the bands with large symbol size (dark-red)
represent NbSe2 original bands. The shift down of the Fig. 5 Projected density of states of each atom in F e1/4 NbSe2 in AF
bands of the host atoms is very clear especially at the - phase using GGA-PBE approximation. From top to bottom, we plot
PDOS of Fe−3d, Nb(1) -4d and Nb(2) -4d, Se(1) -4p and Se(2) -4p. We
point which is an indication of an electron filling coming
plot the PDOS only for one Fe atom. Since the Nb atoms are not
from Fe. Moreover, we can distinguish the bands crossing polarized, we gathered the PDOS of Nb(1) and Nb(2) sites in the same
EF with extra Fe-like bands with non-negligible weights. plot; only spin-up channel is shown
J Supercond Nov Magn (2019) 32:805–819 813

itinerant character, while the spin minority channel is almost this and as mentioned before, the effect of correlation is
empty and has a more localized character. The two doubly expected to be important in this type of materials [30]
degenerate states (dx 2 −y 2 /dxy and dyz /dxz ) have a sharp and the need to use more advantaged methods beyond
peak at Fermi level with a very localized character, while the normal LDA and GGA approximations to describe the
dz2 states have no contribution to EF . This large DOS at exchange and correlation effects in this systems is not
EF with a narrow width emerges the problem of electron questionable. The DFT+U is one of the most efficient
correlation in this system, which is not properly treated approaches that is proved to be a good approximation for
by approximations of exchange-correlation potentials used this type of 3d-transition metals [45, 46, 62]. The exact
in DFT calculations. The three degenerate states in the exchange correlation effect (EECE) is also approved to
minority spin channel show small equally occupied states be used for the treatment of electron correlation problems
and very large empty DOS which reflect the oxidation without dealing with the double counting terms of LDA+U
of F e+2 with intermediate spin configurations [9], with methods.
a magnetic moment of 2.80μB and an orbital moment
of 0.17μB per Fe atom. The hybridization between the 3.3 Strong Electron Correlation Effect
Nb(1) -4d states and Fe−3d states is weak according to
the changes in the Nb(1) -4d PDOS but still strong with The GGA+U method was used to remedy the correlation
the Se(1) -4p. In addition to that, there is an absence of effect which is attributed generally to the 3d-Fe states.
polarization on Nb sites, no magnetic moment induced, It was used the self-interaction correction (SIC). This
due to dual hybridization with two neighbor iron atoms approximation uses only one variable as a parameter for the
in the c-direction coupled antiferromagnetically. The main on-site electron interaction. This parameter represents the
hybridization is between the dxz /dyz of both F e − Nb(1) effective coulomb interaction Ueff = U − J , where U is
in the range of −3.6 to −1.4 eV , and a small one of both the Hubbard parameter of the static coulomb interaction and
dz2 states in the range of −0.8 to −0.2eV and from 1.5 to J is Hund’s rule coupling of the exchange interaction. We
2.0eV . For Se(1) -4p states, the hybridization to F e-3d is in note that the electron correlation correction is applied on
wide range at the occupied states and around EF between Fe−3d states only. Moreover, because of the strong spin-
px /py and dx 2 −y 2 /dxy . This induces a magnetic moment of orbit coupling revealed from the experiments [9] and the
0.02μB at Se(1) site. The Nb(1) -d and the Se(1) -4p states important, but not so large, value of the orbital magnetic
hybridize in the unoccupied states in a range of 1.6 to moment (morb ) obtained from GGA+SOC calculation, one
4.5eV and weaken in the occupied states compared to the cannot exclude the spin-orbit coupling together with the
pure compound due to the effect of Fe atom. The strong electron correlation effect [12, 30, 44]. To see their effect,
hybridization between Fe and Se(1) atoms can be interpreted “SOC” should be switched off wherever it is needed. In
as a formation of a strong covalent bonding between these this part, we have used the crystal structure parameters as
two atoms as the bond length between them is now smaller obtained in the previous section.
than Nb(1) -Se(1) and Nb-Se bond length (see Table 2). To shed light on the electron correlation effect for the
In Nb(2) -4d, we can recognize the compensation between magnetic phase transition in F e0.25NbSe2 compound, we
the dz2 and the dx 2 −y 2 /dxy around EF , where the DOS of have calculated the difference of the total energy between
dz2 becomes mostly occupied and the dx 2 −y 2 /dxy mostly the FM and AF phases E = EF M − EAF for several
empty, with more hybridization with the Se(2) -pz and Se(2) - values of Ueff , in the range of 0.0 to 4.8eV . Hence, this
px /py . The difference between Nb(2) -Se(2) and Nb-Se definition means that the AF phase is lower in energy for
bond lengths is small, which reflects the small change of positive values while the negative energies put the FM
PDOS of farthest atoms from dopant. The metallic behavior phase as the magnetic ground state. The calculations were
of 2H-NbSe2 is enhanced by iron intercalation, where the done for the preceding two magnetic phases, FM and AF,
DOS at the Fermi level is much more dominated by the including both correlation and SOC effects. Figure 6a shows
3d − F e states, and the contribution from other atoms is the evolution of E with Ueff . The GGA+U including SOC
rather small. The magnetic moment is ±2.97μB /F e atom, (GGA+U+SOC) predicts also an AF ground state which is
which is smaller than the experimental one 5.17μB [41]. always larger than the FM state. As it is seen in this figure,
The picture of the band filling with intercalation, shifting the AF phase is enhanced by further increasing Ueff where
up of EF in the band structure, is consistent with the picture the AF phase is lower in energy over the whole range of
of the rigid band model proposed before [9] and which has Ueff values.
been seen recently in the ARPES experiment [33]. The Fe Hence, hereafter, we will consider only the stable
intercalant-derived bands at EF disagree with our finding magnetic phase, the AF order. The evolution of mspin , morb ,
from the GGA band structure calculations. In addition to and the ratio morb /mspin per Fe atom, with increasing
814 J Supercond Nov Magn (2019) 32:805–819

500 of F e intercalated NbSe2 (in the AF phase) and possible

(a)

EFM - EAFM (meV)

400 interplay between them, GGA+U total energy calculations
300 are converged firstly without SOC. Then, we bring all
200 ΔΕ
mspin together including SOC, performing GGA+U+SOC, to
100 morb
assess all effects. Thus, the electron correlation effect was
0 morb/mspin
(b) 0.6
treated according to two schemes, the DFT+U with SIC
double counting term correction and EECE method with no
mspin ( μΒ)

morb ( μΒ)
3.2
0.4
double counting problem. Moreover, DFT+U calculations
3
0.2 need as input the effective coulomb parameter Ueff applied
2.8
0
on the d-correlated orbitals of F e atoms. In the remaining
of this study, we will set Ueff = 3.8eV (which corresponds
0 1 2 3 4 5
Ueff (eV)

to U = 4.5eV and J = 0.7eV ), while the EECE method

Fig. 6 a Variation of the energy difference between FM and AF phases
E = EF M −EAF . b Spin (mspin ), orbital (morb ) magnetic moments,
is a parameter-free method. The difference between the two
and the ratio between them (morb /mspin ) with increasing values of schemes gives us a good opportunity to confirm our results
Ueff away from the ambiguous approximations adopted for these
methods. To this end, we have shown also the PDOS of
EECE+SOC band structure calculations for F e1/4 NbSe2
the values of Ueff , is plotted in Fig. 6b. A monotonic with a reasonable amount of Hatree-Fock exact exchange
increasing of mspin from 2.80 to 3.30μB can be seen equal to (α) = 25%.
when the Ueff increases from more localized. Concerning To allow a better comparison, we have plotted the band
the correlation effect, the orbital moment increased from structure in Fig. 7 and the PDOS in Fig. 8 for GGA+U,
0.15eV from GGA-only exchange-correlation to 0.54μB at GGA+U+SOC, and EECE+SOC. Concerning GGA+U,
Ueff = 1.7eV and gets saturated at about Ueff = 2.0eV Fig. 7a, changes in the band structure of F e1/4 NbSe2 by
to morb ∼ 0.60μB . Thus, a moderate correlation effect the effect of Ueff can be well seen on F e bands where they
is sufficient to unquenched the orbital moment. The ratio shifted away from EF with a flat band dispersion (bands
between morb and mspin has a reasonable value that can be with an atomic-like character). Indeed, sharp peaks in the
checked experimentally. PDOS of Fe−d bands, as plotted in Fig. 8a, reflect this
To get more insight on the strong electron correlation and strong electron localization, especially the bands around
the SOC effects on the electronic and magnetic properties “1eV ” that correspond mainly to the dx 2 −y 2 /dxy state of

Fig. 7 Band structure plot for F e1/4 NbSe2 along the Brillouin zone as (a) with including SOC effect, GGA+U+SOC. c The band struc-
high-symmetry directions. a The band characters of Fe−3d (circle), ture plotted in (b) unfolded to the primitive Bloch representation. The
Nb−4d (diamond), and Se-4p (square) obtained from GGA+U. In (b), symbols size and color represent the Bloch spectral weights
J Supercond Nov Magn (2019) 32:805–819 815

6 2 ×2 4

3
×4 (a) 1
(b) 2
(c)
Fe Fe Fe
0 0 0

-3 -1 -2 d z2
d z2 d z2 dxz/ dyz
-6 -2 -4 ×4
dxz / dyz dxz/ dyz 1
0.6 Nb1 0.4 dx2-y2 / dxy
dx2-y2 / dxy dx2-y2/ dxy Nb1 Nb1
0.3 0.2 0.5

PDOS (States / eV)

PDOS (States / eV)

PDOS (States/eV)

0 0 0

-0.3 -0.2
-0.5
Nb2 Nb2 Nb2
-0.6 -0.4
-1
0.2
Se1 Se1
0.1
0.1
Se1 0.1

0 0 0

-0.1
-0.1 -0.1
px / py -0.2 px/ py px/ py
pz pz pz
0.3 0.3
0.2

0 0 0

Se2 Se2 -0.2 Se2

-0.3 -0.3
-6 -5 -4 -3 -2 -1 0 1 2 3 4 -6 -5 -4 -3 -2 -1 0 1 2 3 4 -6 -5 -4 -3 -2 -1 0 1 2 3 4
Energy (eV) Energy (eV) Energy (eV)

Fig. 8 Projected density of states of F e1/4 NbSe2 in the AF phase atoms are not polarized, we gathered the PDOS of Nb(1) and Nb(2)
using GGA+U in (a), GGA+U+SOC in (b), and EECE+SOC in (c) sites in the same plot; only spin-up channel is shown
for all atoms. We plot the PDOS only for one F e atom. Since the Nb

Fe atom which have a very narrow energy width with very reflect d↓1 occupation. It is very clear that the F e-dz2 band
high DOS. Around EF , DOS is more reduced but still is shifted away from EF and becomes fully empty with
retains its metallic character of the compound. The main no clear contribution of F e-d states at EF comparing to
contribution is by Nb-d, and a almost filled minority spin GGA+U only calculation.
of F e-dz2 states with a broad density. In addition, the DOS Figure 8c represents the unfolded band structure after the
of Se(1) shows broad changes around and below EF due GGA+U+SOC calculation. From this figure, it is clear that
to the strong hybridization with Fe−3d states which induce the Fe bands with non-negligible weights are far from EF
a magnetic moment at Se sites of 0.03μB . The magnetic level. Bands with small weights near -point are present
moment of Fe is raised to an integer value, ±3.12μB /F e which reflect the folding bands as detected experimentally,
atom which is expected from DFT+U calculations with see figure 5 in [33].
more localized character, but it is still smaller than that The PDOS of Fe shows that the states dx 2 −y 2 /dxy and
obtained experimentally of 4.90μB [6]. dxz /dyz are mixed. This is more apparently in minority spin
Henceforth, we switch on the SOC effect; the total energy states. Moreover, these states are splitted into two sharp
calculation of GGA+U+SOC is plotted in Fig. 7b for the peaks located at −1 and 1eV with an atomic orbital-like
band structure, and Fig. 8b for the PDOS of each atom. character. It seems that the original states dx 2 −y 2 /dxy and
From these figures, we see dramatic changes due to the dxz /dyz are mixed, in order to fulfill the Hamiltonian of
inclusion of the SOC effect where they are more obvious the system. The obtained states are a linear combination of
in Fe−d PDOS. In the Fe−d states, Fig. 8b, all bands are atomic orbitals (αd2 + βd−1 and αd−2 − βd1 ), which are
occupied for the majority spin channel with a localized splitted under SOC effect resulting in the large unquenched
character reserving the picture of high spin state of F e+2 . orbital magnetic moment. We used the notation LML (ML
While minority spin channel is partially occupied, which being the z-projection of the orbital angular momentum
816 J Supercond Nov Magn (2019) 32:805–819

L). Indeed, a large morb = 0.61μB is obtained by The major differences between the calculation with and
GGA+U+SOC calculation. It is worth to note that the without SOC are mainly the electron occupations in the
expectation value of L in the non-mixed d-states (basis) minority spin of Fe bands and the presence of the latter at
used here, which are the real spherical harmonics, is zero EF level. All calculations give almost the same occupation
which means that morb is quenched. The obtained spin for majority spin electrons, while minority spin occupation
magnetic moment is mspin = 3.24 μB per F e atom, where strongly depends on the SOC effect. For GGA+U, dz2 is
morb /mspin = 0.19 , which is a little smaller than the partially occupied, and presents at EF , while dx 2 −y 2 /dxy
XMCD value of 0.33 in F e1/4 T aS2 [30]. The DOS at EF and dxz /dyz occupations are negligible to give d↑5 (a1g )1↓
is almost coming from the host atoms. For the PDOS of for GGA+U, whereas in GGA+U+SOC the dz2 occupation
Nb and Se atoms, Fig. 8b, only Se(1) -p orbitals have a is negligible and the two states dx 2 −y 2 /dxy and dxz /dyz
significant change validating the strong hybridization with are partially occupied with the same amounts as they are
the Fe−3d found before in GGA and GGA+U, yield an mixed and splitted under SOC effect to give d↑5 (egπ )1↓ for
induced magnetic moment of 0.017μB at Se(1) sites. Koh GGA+U+SOC.
et al. interpret the change of the resistivity behavior at TN After GGA+SOC calculations, where the correlation
as a strong interaction between the electron of the host effect is absent, the orbital magnetic moment of Fe atoms
and the magnetic moment on Fe [33], which is in good is unquenched mo = 0.17 μB (see Table 3). In GGA+U
accordance with our results. On the other hand, we can not calculation, without SOC, the intercalant-derived dz2 bands
see any spin polarization on the Nb(1) site, except the usual at EF seems to be a wrong picture comparing to the ARPES
hybridization with Se-4p. This means that the Nb atom is experiment. Moreover, the crystal field splitting theory and
somehow screened from the Fe atoms by the conduction the band filling obtained by calculation are not consistent.
electrons of the Se atoms, and hence no formation of bond After GGA+U+SOC, the only bands crossing EF are those
(or hybridization) between the Fe and Nb atoms can be seen coming from Nb−d bands of the host material. These results
from the PDOS. are in a good agreement with the ARPES experiment [33],
Figure 8c represents the EECE+SOC calculations which where no bands crossing EF come from the intercalated
yield a PDOS with a broad agreement with GGA+U+SOC atoms. Furthermore, the band occupation obtained by this
ones and coincide very well with the picture describing method, d↑5 (egπ )1↓ , is very consistent with the crystal field
the strong effect of correlation and SOC. The mixing splitting picture and the F e L2,3 -edge X-ray absorption
of the dx 2 −y 2 /dxy and dyz /dxz are identical with more spectroscopy (XAS) experimental results for Fe intercalated
localization, from that of GGA+U+SOC, observed from the in T aS2 [30].
sharp PDOS picks. The same trends are retained for the In order to confirm our findings concerning the magnetic
remaining Nb and Se atoms where the dominant DOS at properties, we have summarized our results obtained for
EF are still coming from Nb1(2) atoms. Furthermore, the spin and orbital moments using different approximations in
calculation predicts an unusual orbital magnetic moment Table 3. As can be seen in this table, the spin magnetic
at F e atoms, a value of morb = 0.72μB , and mspin = moment of Fe is enhanced by adding some correlation to the
3.16μB for the spin magnetic moment, giving the ratio of system for the two methods used and also by considering the
morb /mspin = 0.23 greater than that obtained with Hubbard SOC effect. For the orbital magnetic moment, we can notice
correction GGA+U+SOC. an important enhancement of morb that came from the

Table 3 Spin (mspin ) and orbital (morb ) magnetic moments in (μB ) of F e1/4 NbSe2 for Fe and Se(1) atoms obtained from various approximations.
Ueff = 3.8eV is used in GGA+U/+SOC calculations. For experimental values mT = meff , see footnote in Table 1

LDA/GGA EECE EECE+SOC Expt

- SOC U U+SOC

Fe mspin 2.64/2.83 2.60/2.80 3.09/3.12 3.18/3.24 3.01 3.16 4.90a

morb – 0.15/0.17 – 0.53/0.61 – 0.72
mT 2.64/2.83 2.75/2.97 3.09/3.12 3.71/3.85 – 3.88 5.17[a] /5.25b

Se mspin 0.019/0.021 0.01/0.020 0.028/0.030 0.018/0.017 0.027 0.025 –

a From ref [6]

b From ref [29]
J Supercond Nov Magn (2019) 32:805–819 817

Fig. 9 Electron charge density

for a NbSe2 and b F e1/4 NbSe2

strong correlation effect regardless the approximations used. obtained by a direct comparison between the experimental
For the magnetic moment of Se, mSe , its value increases and our calculated band structure for NbSe2 compound.
when the localization becomes important and declines by Our results placed NbSe2 in the vicinity of iron-based
adding the SOC. For the Nb(1) site, we cannot find any superconductors localized materials. We have found a
induced magnetic moment in all calculations. strong dependency of the structural parameters on the
To examine the nature of bonds between the different magnetic configurations considered in this work. By
atoms, Fig. 9 shows the electron charge density for examining the band structure and the DOS of the pristine
the intercalated F e1/4 NbSe2 and the pristine 2H-NbSe2 and intercalated system, the effect of intercalation with Fe
compounds, especially between Fe and its surrounding is confirmed to be a charge transfer from the Fe atoms to the
atoms Se(1) and Nb(1) . We chose the plane that passes host NbSe2 material.
through these three atoms and containing also the other The theoretical scheme of GGA and the on-site repulsion
two types Se(2) and Nb(2) . The bond between the Nb−Se U including SOC predicts the correct antiferromagnetic
in the pristine system is more covalent than ionic. If we ground state found in the experiments. The obtained
look to the F e-Se(1) bonds in the F e1/4 NbSe2 , they are magnetic moment of F e is enhanced by introducing
partially ionic and partially covalent compared to Nb(2) - correlation effects. The spin-orbit coupling leads to an
Se(2) bonds, which are more covalency. One can also see unquenched orbital magnetic moment in the Fe atom
that no direct bonding is formed between the Fe and Nb from GGA and a larger value is obtained by introducing
atoms. The partial ionization of Fe is in agreement with the correlation effect from both GGA+U and EECE methods.
experimental result [33]. Accordingly, RKKY interaction is Moreover, the large unquenched orbital magnetic moment
already proposed as an exchange-coupling mechanism in is saturated to morb ∼ 0.6μB under a moderate correlation
such systems [9, 30, 32] of the far separated Fe atoms [9, effect of Ueff ∼ 2eV .
30, 32]. The other mechanism proposed is the indirect The correlation effect was found to be an important
superexchange coupling. The interaction between Fe and factor in this type of system alongside with the spin-orbit
Nb(1) in F e1/4 NbSe2 is barely detected in our work; no coupling, where the result of the absence of one or both
spin polarization or magnetic moment was found on the of them seems to be not consistent with the experiment.
Nb(1) site, while the interaction of Se(1) -4p orbital with After DFT+U+SOC, there is no intercalant-derived bands
both F e-3d and Nb(1) -4d orbitals is very strong. Hence, crossing the EF level and the Fe−3d electron occupations
the superexchange is proposed as an exchange coupling are d↑5 (egπ )1↓ which is very consistent with the crystal field
mechanism along the c-axis. The interaction is high through splitting picture. The good match between the results of
Se atoms in a sequence of F e-Se(1) -Nb(1) -Se(1) -F e. DFT+U+SOC and DFT+EECE+SOC methods confirms
our findings. The superexchange coupling between the Fe
atom is found to be more likely through the chain Fe-
4 Conclusion Se(1) -Nb( 1)-Se(1) -Fe. Excellent agreements between our
theoretical calculations and experiments were obtained
We have presented structural, electronic, and magnetic using DFT+U+SOC.
properties of Fe-intercalated F ex NbSe2 , for x = 0% and On the basis of our findings, from the values of the bulk
25%, within DFT using the all-electron FP-LAPW method. modulus and the obtained bond lengths of F e1/4 NbSe2
The tendency towards the localization in superconductors is in the different magnetic order considered in this paper,
818 J Supercond Nov Magn (2019) 32:805–819

the possible change of the stable magnetic ground state 16. Hardy, W.J., Chen, C.W., Marcinkova, A., Ji, H., Sinova,
by applying a pressure would be of interest. Moreover, J., Natelson, D., Morosan, E.: Phys. Rev. B 91, 054426
(2015). https://doi.org/10.1103/PhysRevB.91.054426. https://link.
the obtained large orbital magnetic moment can lead to aps.org/doi/10.1103/PhysRevB.91.054426
very anisotropic properties like a large magnetocrystalline 17. Choi, Y.J., Kim, S.B., Asada, T., Park, S., Wu, W., Horibe, Y.,
anisotropy that is already reported in similar compounds. Cheong, S.-W.: EPL 86(3), 37012 (2009). https://doi.org/10.
Further studies on this issue are also still required, taking 1209/0295-5075/86/37012
18. Corcoran, R., Meeson, P., Onuki, Y., Probst, P.A., Springford,
into account both correlation and SOC effects.
M., Takita, K., Harima, H., Guo, G.Y., Gyorffy, B.L.: J.
Phys. Condens. Matter 6(24), 4479 (1994). http://stacks.iop.org/
0953-8984/6/i=24/a=010
Acknowledgements One of the authors (H.K.) would like to thank 19. Rossnagel, K., Seifarth, O., Kipp, L., Skibowski, M., Voß, D.,
Dr. Ali Hamidani, University of Guelma-Algeria, for many useful Krüger, P., Mazur, A., Pollmann, J.: Phys. Rev. B 64(23), 235119
comments and also Professor Samir Lounis from the Peter Grunberg (2001)
institute for their valuable insights and discussions. 20. Johannes, M.D., Mazin, I.I., Howells, C.A.: Phys. Rev. B
73, 205102 (2006). https://doi.org/10.1103/PhysRevB.73.205102.
https://link.aps.org/doi/10.1103/PhysRevB.73.205102
21. Rahn, D., Hellmann, S., Kalläne, M., Sohrt, C., Kim, T., Kipp, L.,
References Rossnagel, K.: Phys. Rev. B 85(22), 224532 (2012)
22. Flicker, F., Van Wezel, J.: arXiv:1502.06816 (2015)
1. Wadley, P., Howells, B., Železný, J., Andrews, C., Hills, V., Cam- 23. Arguello, C., Rosenthal, E., Andrade, E., Jin, W., Yeh, P., Zaki, N.,
pion, R.P., Novák, V., Olejı́nk, K., Maccherozzi, F., Dhesi, S.S., Jia, S., Cava, R., Fernandes, R., Millis, A., et al.: Phys. Rev. Lett.
Martin, S.Y., Wagner, T., Wunderlich, J., Freimuth, F., Mokrousov, 114(3), 037001 (2015)
Y., Kuneš, J., Chauhan, J.S., Grzybowski, M.J., Rushforth, 24. Bawden, L., Cooil, S., Mazzola, F., Riley, J., Collins-McIntyre,
A.W., Edmonds, K.W., Gallagher, B.L., Jungwirth, T.: Science L., Sunko, V., Hunvik, K., Leandersson, M., Polley, C.,
351(6273), 587 (2016). https://doi.org/10.1126/science.aab1031. Balasubramanian, T., et al.: Nat. Commun. 7 (2016)
http://science.sciencemag.org/content/351/6273/587 25. Hauser, J.J., Robbins, M., DiSalvo, F.J.: Phys. Rev. B 8,
2. Gomonay, O., Jungwirth, T., Sinova, J.: Phys. Stat. Solidi (RRL)– 1038 (1973). https://doi.org/10.1103/PhysRevB.8.1038. https://
Rapid Res. Lett. 11(4), 1700022 (2017). https://doi.org/10.1002/ link.aps.org/doi/10.1103/PhysRevB.8.1038
pssr.201700022 26. Whitney, D.A., Fleming, R.M., Coleman, R.V.: Phys. Rev. B 15,
3. Wilson, J.A., Yoffe, A.D.: Adv. Phys. 18, 193 (1969). https://doi. 3405 (1977). https://doi.org/10.1103/PhysRevB.15.3405. https://
org/10.1080/00018736900101307 link.aps.org/doi/10.1103/PhysRevB.15.3405
4. Wilson, J., Salvo, F.D., Mahajan, S.: Adv. Phys. 24(2), 117 27. Dai, Z., Xue, Q., Gong, Y., Slough, C.G., Coleman, R.V.: Phys.
(1975). https://doi.org/10.1080/00018737500101391 Rev. B 48, 14543 (1993). https://doi.org/10.1103/PhysRevB.48.
5. Chhowalla, M., Shin, H.S., Eda, G., Li, L.J., Loh, K.P., Zhang, H.: 14543. https://link.aps.org/doi/10.1103/PhysRevB.48.14543
Nat. Chem. 5(4), 263 (2013) 28. Coleman, R.V., Dai, Z., Gong, Y., Slough, C.G., Xue, Q.:
6. van Den Berg, J.V., Sherwood, R.: J. Phys. Chem. Solids 32(1), J. Vac. Sci. Technol., B: Microelectron. Nanometer Struct.–
167 (1971). https://doi.org/10.1016/S0022-3697(71)80019-7. Process., Meas., Phenom. 12(3), 1801 (1994). https://doi.org/10.
http://www.sciencedirect.com/science/article/pii/S0022369771800 1116/1.587603 http://avs.scitation.org/doi/abs/10.1116/1.587603
197 29. Hulliger, F., Pobitschka, E.: J. Solid State Chem. 1(2), 117
7. Friend, R., Yoffe, A.: Adv. Phys. 36(1), 1 (1987). https://doi.org/ (1970). https://doi.org/10.1016/0022-4596(70)90001-0. http://
10.1080/00018738700101951 www.sciencedirect.com/science/article/pii/0022459670900010
8. Vandenberg-Voorhoeve, J.M.: Structural and magnetic properties 30. Ko, K.T., Kim, K., Kim, S.B., Kim, H.D., Kim, J.Y., Min, B.I.,
of layered chalcogenides of the transition elements, pp. 423–457. Park, J.H., Chang, F.H., Lin, H.J., Tanaka, A., Cheong, S.W.:
Springer Netherlands, Dordrecht (1976). https://doi.org/10.1007/ Phys. Rev. Lett. 107, 247201 (2011). https://doi.org/10.1103/Phys
978-94-010-1478-6 RevLett.107.247201. https://link.aps.org/doi/10.1103/PhysRevLett.
9. Parkin, S.S.P., Friend, R.H.: Philos. Mag. B 41(1), 65 (1980). 107.247201
https://doi.org/10.1080/13642818008245370 31. Kirchmayr, H.R.: J. Phys. D. Appl. Phys. 29(11), 2763 (1996).
10. Parkin, S.S.P., Friend, R.H.: Philos. Mag. B 41(1), 95 (1980). http://stacks.iop.org/0022-3727/29/i=11/a=007
https://doi.org/10.1080/13642818008245371 32. Suzuki, N., Yamazaki, Y., Teshima, T., Motizuki, K.: Physica, B:
11. Friend, R.H., Beal, A.R., Yoffe, A.D.: Philos. Mag. 35(5), 1269 Condens. Matter 156, 286 (1989). https://doi.org/10.1016/0921-45
(1977). https://doi.org/10.1080/14786437708232952 26(89)90655-8. http://www.sciencedirect.com/science/article/pii/
12. Morosan, E., Natelson, D., Nevidomsky, A.H., Si, Q.: Adv. Mater. 0921452689906558
24(36), 4896 (2012). https://doi.org/10.1002/adma.201202018 33. Koh, Y., Cho, S., Lee, J., Yang, L.X., Zhang, Y., He, C., Chen,
13. Narita, H., Ikuta, H., Hinode, H., Uchida, T., Ohtani, T., Wakihara, F., Feng, D.L., Arita, M., Shimada, K., Namatame, H., Taniguchi,
M.: J. Solid State Chem. 108(1), 148 (1994). https://doi.org/10.1006/ M., Kim, C.: Jpn. J. Appl. Phys. 52(10S), 10MC15 (2013). http://
jssc.1994.1022. http://www.sciencedirect.com/science/article/pii/ stacks.iop.org/1347-4065/52/i=10S/a=10MC15
S002245968471022X 34. Blaha, P., Schwarz, K., Madsen, G., Kvasnicka, D., Luitz, J.: An
14. Morosan, E., Zandbergen, H.W., Li, L., Lee, M., Checkelsky, J.G., augmented plane wave+ local orbitals program for calculating
Heinrich, M., Siegrist, T., Ong, N.P., Cava, R.J.: Phys. Rev. B crystal properties (2001)
75, 104401 (2007). https://doi.org/10.1103/PhysRevB.75.104401. 35. Hohenberg, P., Kohn, W.: Phys. Rev. 136, B864 (1964).
https://link.aps.org/doi/10.1103/PhysRevB.75.104401 https://doi.org/10.1103/PhysRev.136.B864. https://link.aps.org/
15. Checkelsky, J.G., Lee, M., Morosan, E., Cava, R.J., Ong, N.P.: doi/10.1103/PhysRev.136.B864
Phys. Rev. B 77, 014433 (2008). https://doi.org/10.1103/PhysRevB. 36. Kohn, W., Sham, L.J.: Phys. Rev. 140, A1133 (1965).
77.014433. https://link.aps.org/doi/10.1103/PhysRevB.77.0144 https://doi.org/10.1103/PhysRev.140.A1133. https://link.aps.org/
33 doi/10.1103/PhysRev.140.A1133
J Supercond Nov Magn (2019) 32:805–819 819

37. Sjöstedt, E., Nordström, L., Singh, D.: Solid State Commun. 48. Marezio, M., Dernier, P., Menth, A., Hull, G.: J. Solid State Chem.
114(1), 15 (2000). https://doi.org/10.1016/S0038-1098(99)00577- 4(3), 425 (1972). https://doi.org/10.1016/0022-4596(72)90158-2.
3. http://www.sciencedirect.com/science/article/pii/S00381098990 http://www.sciencedirect.com/science/article/pii/0022459672901
05773 582
38. Madsen, G.K.H., Blaha, P., Schwarz, K., Sjöstedt, E., Nordström, 49. Selte, K., Kjekshus, A.: Acta Chem. Scand. 18(3) (1964)
L.: Phys. Rev. B 64, 195134 (2001). https://doi.org/10.1103/Phys 50. Dion, M., Rydberg, H., Schröder, E., Langreth, D.C., Lundqvist,
RevB.64.195134. https://link.aps.org/doi/10.1103/PhysRevB.64.19 B.I.: Phys. Rev. Lett. 92, 246401 (2004). https://doi.org/10.1103/
5134 PhysRevLett.92.246401
39. Hamada, I.: Phys. Rev. B 89, 121103 (2014). https://doi.org/10. 51. Peng, H., Yang, Z.H., Perdew, J.P., Sun, J.: Phys. Rev. X
1103/PhysRevB.89.121103. https://link.aps.org/doi/10.1103/Phys 6, 041005 (2016). https://doi.org/10.1103/PhysRevX.6.041005.
RevB.89.121103 https://link.aps.org/doi/10.1103/PhysRevX.6.041005
40. Perdew, J.P., Burke, K., Ernzerhof, M.: Phys. Rev. Lett. 77, 3865 52. Berland, K., Cooper, V.R., Lee, K., Schröder, E., Thonhauser, T.,
(1996). https://doi.org/10.1103/PhysRevLett.77.3865. https://link. Hyldgaard, P., Lundqvist, B.I.: Rep. Prog. Phys. 78(6), 066501
aps.org/doi/10.1103/PhysRevLett.77.3865 (2015). http://stacks.iop.org/0034-4885/78/i=6/a=066501
41. Meerschaut, A., Deudon, C.: Mater. Res. Bull. 36(9), 1721 53. Bucko, T., Hafner, J., Lebegue, S., Angyán, J.G.: J. Phys. Chem.
(2001). https://doi.org/10.1016/S0025-5408(01)00646-8. http:// A 114(43), 11814 (2010)
www.sciencedirect.com/science/article/pii/S0025540801006468 54. Bučko, T.C.V., Lebègue, S., Hafner, J., Ángyán, J.G.: Phys. Rev.
42. Birch, F.: Phys. Rev. 71, 809 (1947). https://doi.org/10.1103/Phys B 87, 064110 (2013). https://doi.org/10.1103/PhysRevB.87.0641
Rev.71.809. https://link.aps.org/doi/10.1103/PhysRev.71.809 10. https://link.aps.org/doi/10.1103/PhysRevB.87.064110
43. Schwarz, K., Blaha, P., Madsen, G.: Comput. Phys. Commun. 55. Mattheiss, L.: Phys. Rev. B 8(8), 3719 (1973)
147(1), 71 (2002). https://doi.org/10.1016/S0010-4655(02)00206- 56. Chen, L., Li, C., Tang, H., Li, H., Liu, X., Meng, J.: RSC Adv. 4,
0. http://www.sciencedirect.com/science/article/pii/S0010465502 9573 (2014). https://doi.org/10.1039/C3RA47237J
002060. Proceedings of the Europhysics Conference on Com- 57. Li, H., Chen, L., Zhang, K., Liang, J., Tang, H., Li, C., Liu,
putational Physics Computational Modeling and Simulation of X., Meng, J., Wang, Z.: J. Appl. Phys. 116(10), 103709 (2014).
Complex Systems https://doi.org/10.1063/1.4895489
44. Mankovsky, S., Chadova, K., Ködderitzsch, D., Minár, J., Ebert, 58. Li, H., Liu, S., Chen, L., Wu, J., Zhang, P., Tang, H., Li, C., Liu,
H., Bensch, W.: Phys. Rev. B 92, 144413 (2015). https://doi.org/10. X., Wang, Z., Meng, J.: RSC Adv. 4(101), 57541 (2014)
1103/PhysRevB.92.144413. https://link.aps.org/doi/10.1103/Phys 59. Basov, D.N., Chubukov, A.V.: Nat. Phys. 7(4), 272 (2011).
RevB.92.144413 https://doi.org/10.1038/nphys1975
45. Anisimov, V.I., Solovyev, I.V., Korotin, M.A., Czyżżyk, M.T., 60. Qazilbash, M.M., Hamlin, J.J., Baumbach, R.E., Zhang, L., Singh,
Sawatzky, G.A.: Phys. Rev. B 48, 16929 (1993). https://doi.org/10. D.J., Maple, M.B., Basov, D.N.: Nat. Phys. 5(9), 647 (2009).
1103/PhysRevB.48.16929. https://link.aps.org/doi/10.1103/Phys https://doi.org/10.1038/nphys1343
RevB.48.16929 61. Lu, D.H., Yi, M., Mo, S.K., Erickson, A.S., Analytis, J., Chu, J.H.,
46. Liechtenstein, A.I., Anisimov, V.I., Zaanen, J.: Phys. Rev. B Singh, D.J., Hussain, Z., Geballe, T.H., Fisher, I.R., Shen, Z.X.:
52, R5467 (1995). https://doi.org/10.1103/PhysRevB.52.R5467. Nature 455(7209), 81 (2008). https://doi.org/10.1038/nature07263
https://link.aps.org/doi/10.1103/PhysRevB.52.R5467 62. Messiad, M., Zanat, K., Hamidani, A.: J. Magn. Magn. Mater.
47. Novák, P., Kuneš, J., Chaput, L., Pickett, W.E.: Phys. Status Solidi 441(Supplement C), 424 (2017). https://doi.org/10.1016/j.jmmm.
(b) 243(3), 563 (2006). https://doi.org/10.1002/pssb.200541371 2017.06.018. http://www.sciencedirect.com/science/article/pii/
S030488531731106X