DSpace at VNU: Naphthalene adsorptions on graphene using Cr Cr-2 Fe Fe-2 linkages: Stability and spin perspectives from first-principles calculations

Chemical Physics Letters 614 (2014) 238–242

Contents lists available at ScienceDirect

Chemical Physics Letters
journal homepage: www.elsevier.com/locate/cplett

Naphthalene adsorptions on graphene using Cr/Cr2 /Fe/Fe2 linkages:
Stability and spin perspectives from first-principles calculations
Viet Q. Bui, Hung M. Le ∗
Department of Materials Science, University of Science, Vietnam National University, Ho Chi Minh City, Vietnam

a r t i c l e

i n f o

Article history:
Received 1 August 2014
In final form 18 September 2014
Available online 28 September 2014

a b s t r a c t
We present a first-principles study of naphthalene adsorption on graphene via coordination bonds with
Cr/Cr2 /Fe/Fe2 . The obtained structures possess great binding stability, and the geometry alignment of
C10 H8 is distorted. Especially, the use of Cr/Fe dimer further enhances the binding stability of C10 H8 on
graphene. From binding energy analysis, the adsorption of C10 H8 on metal–graphene is observed to be
more favorable than the adsorption of metal–C10 H8 on graphene. When empirical dispersion corrections
are introduced, the binding energy is improved by 0.78–1.40 eV. Interestingly, various degrees of magnetism are observed with respect to the metal identity, atom/dimer utilization, and bonding interactions.
© 2014 Elsevier B.V. All rights reserved.

Graphene has been a hot trend in material research for years

[1,2] because of its excellent electronic properties and great thermostability as proven through a variety of experimental studies
[3–5]. Especially, the spintronic–electronic aspects with the aim
to control the electronic properties tend to get more particular
interests [6,7]. During the past few years, studies of the electronic
structures and magnetic properties of those two-dimensional
structures have gained remarkable achievements. Among them,
the studies of metal atom adsorptions on graphene [8–16] using
density functional theory (DFT) calculations [17,18] have shown
different perspectives in the insights of electronic structures.
Besides, various aspects such as system configurations, binding
energies (stability), atomic diffusion barriers, magnetization, and
work function are also of concern. Those studies can be regarded as
a premise for developing applications in advanced semiconductors,
nanomagnetic devices, as well as gas sensors.
The adsorptions of transition metal dimers on graphene were
shown to have relatively low binding energies (from 0.16 to
0.27 eV), which might result in high mobility of the adatoms
and dimers on the surface [12]. Therefore, it is of importance to
seal the adsorbed metal atoms with functional groups (ligands).
By adopting first-principles calculations, Avdoshenko et al. [19]
explored the electronic properties of graphene–metal–benzene
complex, and the ligand was capable of controlling the properties

∗ Corresponding author.
E-mail addresses: , (H.M. Le).
/>0009-2614/© 2014 Elsevier B.V. All rights reserved.

of absorbed metal atoms, thereby imposed bosonic/fermionic characters. The theoretical interactions between two graphene layers
and Cr were also investigated in the graphene–Cr–graphene intercalating nanostructures [20]. From the analysis of binding energy
and electronic structures, it was shown that high Cr-occupation

rate gave an unstable structure, while the loosened occupations
(lower concentration) of Cr resulted in more stable structures. In
another study reported by Le et al. [21], C60 was utilized as a ligand
to decorate the graphene surface via coordination bonds with Cr
(G–Cr–C60 ); meanwhile, interesting magnetic properties exhibited by G–Cr–C60 itself and its combination with a metal cluster
(Ni4 /Pd4 /Pt4 ) were found.
Acknowledging the significance of graphene–metal–ligand
structures in spintronic and electronic applications, in the present
study, we demonstrate a theoretical investigation of threecomponent graphene-based nanostructures, in which naphthalene
(C10 H8 , abbreviated as Np) is attached to graphene via coordination bonds with a Cr/Fe atom or Cr2 /Fe2 dimer (for convenience,
the structures are denoted as G–M–Np or G–M2 –Np). Structural
stability and magnetic moments of those structures will be deliberately discussed in the light of electronic structure data given by
DFT calculations.
First-principles calculations based on DFT are performed within
local spin density approximation in order to evaluate the degree
of spin polarization in the bonding schemes between Cr/Fe
and graphene/naphthalene. All DFT calculations are executed
using Quantum Espresso, an open-source computational package
[22]. The Perdew–Burke–Ernzerhof (PBE) exchange-correlation


V.Q. Bui, H.M. Le / Chemical Physics Letters 614 (2014) 238–242

functional [23,24] is employed with the ultrasoft pseudopotentials
for all atoms involved [25,26]. The k-points are generated using the
Monkhorst–Pack method with a chosen mesh of (6 × 6 × 1), which
is sufficient to ensure the reliability of total energy calculations. The
kinetic energy cut-off is selected as 45 Rydberg (612 eV) for planewave expansions. The two-dimensional lattices (atomic positions
and unit cell parameters) are simultaneously optimized using the
Broyden–Fletcher–Goldfarb–Shanno algorithm [27] with an energy

convergence criterion of 10−5 eV/cell and a gradient convergence
˚
criterion of 10−3 eV/A/atom.
The two-dimensional unit cells are
constructed by employing large c lattice parameters (30 Bohr or
˚ The optimized structures and electronic structure data
15.88 A).
herein are reported without considering the empirical dispersion
correction terms. It was shown in an earlier study that dispersion
interactions contributed a significant role in the binding between
graphene and aromatic molecules [28]. Therefore, we also perform
additional optimizations with the D2 empirical dispersions corrections [29,30] and update the binding energies.
A (4 × 4) supercell of graphene containing 32 carbon atoms
is employed to host the M–Np and M2 –Np complexes. The chosen graphene cell is wide enough to avoid possible interactions
between the attached complexes due to periodic boundary conditions. In the first part, the periodic graphene sheet is decorated
with the C10 H8 ligand using only one transition metal atom (either
Cr or Fe) as the bridging atom. In the G–Cr–Np structure (Figure 1a),
the Cr atom clings to the hollow site of two honeycomb units
in graphene and C10 H8 . This binding behavior is similar to that
observed in two previous studies [19,21]. In the case of G–Fe–Np,
two different configurations are found. As shown in Figure 1b and
c, the Fe atom can settle either on the hollow site or on top of a C
atom in graphene, while it only binds to the hollow site of Np.
In the second part, we attempt to use a transition-metal dimer
to fix both aromatic rings of naphthalene instead of only one like in
the previous cases. In the case of Cr dimer, only one stable configuration can be found, in which two Cr atoms occupy the hollow sites
in two adjacent hexagonal honeycomb units in both graphene and
naphthalene, as shown in Figure 1d. In the G–Fe2 –Np case, there are
two stable configurations given by structural optimizations; one
structure (Figure 1f) is similar to the Cr dimer case, while the other

is much distinctive because one Fe atom accommodates at the hollow site and another locates on top of a carbon atom (Figure 1e).
Interestingly enough, the different arrangements of Fe atoms in
G–Fe–Np and G–Fe2 –Np result in different magnetic behaviors as
will be discussed in a later part of this Letter. In total, we report
six configurations in this study: G–Cr–Np, G–Fe–Np(1) , G–Fe–Np(2) ,
G–Cr2 –Np, G–Fe2 –Np(1) , and G–Fe2 –Np(2) (see Figure 1).
In the first structure (G–Cr–Np), it can be seen clearly that
Cr connects to six C atoms in both naphthalene and graphene;
however, the plane containing naphthalene is not parallel to
graphene. Indeed, the ligand is highly distorted as we observe vari˚ while the Cr–C(graphene)
ous Cr–C(Np) bond lengths (2.14–2.23 A),
˚ In terms of bonding, the interbonds are almost identical (2.20 A).
acting behavior of naphthalene as a ligand is distinguished from
that of benzene as seen in a previous study, where benzene was
symmetrically aligned [19]. As a ligand, C60 even behaves more
differently because the geometry of C60 allows itself to rotate and
achieve most stable states by assembling high geometry distortions
[31].
The structural stability of a structure can be evaluated by calculating two different binding energies:
(a)

(1)

(b)

(2)

Ebinding = Egraphene + EM –Np − Estructure
Ebinding = Egraphene–M + ENp − Estructure

239

Table 1
Binding energies and magnetizations of the investigated nanostructures (the values
given by PBE calculations with D2 dispersion corrections are shown in parentheses).
Binding energy
(eV)
(a)

G–Cr–Np
G–Fe–Np(1)
G–Fe–Np(2)
G–Cr2 –Np
G–Fe2 –Np(1)
G–Fe2 –Np(2)

MT
( B /cell)

MA
( B /cell)

0.11
(0.11)
1.89
(1.06)
2.00
(2.00)
2.47
(2.44)

1.03
(0.94)
1.34
(1.32)

0.20
(0.19)
2.39
(1.52)
3.27
(3.23)
3.36
(3.24)
2.36
(2.01)
1.88
(1.77)

(b)

Ebinding

Ebinding

1.78
(2.90)
1.27
(2.24)
1.62
(2.08)

2.71
(4.04)
1.62
(2.99)
1.42
(2.82)

1.80
(2.82)
1.85
(2.69)
2.25
(3.03)
3.34
(4.48)
2.10
(3.42)
2.43
(3.51)

where Egraphene , Egraphene–M , ENp , and ENp–M represent the total energies of pure graphene, graphene decorated with metal atom/dimer,
C10 H8 , and metal–C10 H8 , respectively, while Estructure denotes the
total energy of the optimized complex. The positive binding energy
value is indicative of a stable structure. The two binding energy
quantities above critically demonstrate the relative stability with
respect to experimental synthesizing methods. In experiments,
the synthesis of G–M–Np/G–M2 –Np can be achieved by either
attaching the C10 H8 –metal complex on a pure graphene surface
(a)
(expressed by Ebinding ) or attaching C10 H8 on a metal–graphene

(b)

surface (expressed by Ebinding ). The favorability of a synthesizing
method in experimental reality can be evaluated by making direct
comparisons of the two binding energies.
(a)
By applying Eqs. (1) and (2) for the G–Cr–Np case, Ebinding
(b)

and Ebinding are 1.78 and 1.80 eV, respectively. The binding energy
results indicate that the resulted structure is highly stable;
also, it is demonstrated that the direct adsorption of C10 H8 on
metal–graphene surface is somewhat more favorable. In other
words, the bond between Cr adatom and C10 H8 seems to be quite
stronger than the bond between Cr and the graphene. For con(a)
(b)
venience, we summarize Ebinding and Ebinding of the investigated
nanostructures in Table 1. The updated binding energies with
empirical dispersion corrections are raised by 0.78–1.40 eV, as
listed in Table 1 (shown in parentheses). It can be seen in Table 1
(a)
that solely for G–Cr–Np, Ebinding (2.90 eV) is quite higher than the
(b)

corresponding Ebinding (2.82 eV) when van der Waals corrections are
introduced.
Besides studying structural stability, we also evaluate spin
polarization, which have much influence on the magnetic property.
By interpreting density of states (DOS) from electronic structure
data, the total (MT ) and absolute (MA ) magnetizations (also listed

in Table 1) are derived and reported for each structure. The estimation of total magnetization in G–Cr–Np shows that the structure
exhibits a weak magnetic moment of 0.11 B /cell, while the absolute magnetic moment is 0.20 B /cell. In fact, the absolute magnetic
moment indicates a significant anti-ferromagnetic amount within
the structure. For the electronic structure analysis, we only examine data from the PBE calculations without dispersion corrections.
By examining the total DOS of G–Cr–Np (Figure 2a), we observe that
the weak spin polarization mainly occurs prior to the Fermi level
(highest-occupied bands). While Cr appears to contribute a positive magnetic moment (0.15 B ), C atoms (from both graphene and
naphthalene) tend to contribute a resisting amount, which causes
anti-ferromagnetic effects within the structure. Interestingly, the
spin polarization of Cr 3d orbitals contributes the ferromagnetism


240

V.Q. Bui, H.M. Le / Chemical Physics Letters 614 (2014) 238–242

Figure 1. Side and top views of six optimized nanostructures using PBE calculations without dispersion corrections: (a) G–Cr–Np, (b) G–Fe–Np(1) , (c) G–Fe–Np(2) , (d)
G–Cr2 –Np, (e) G–Fe2 –Np(1) , and (f) G–Fe2 –Np(2) . Teal is used for graphene.

Figure 2. Total DOS of (a) G–Cr–Np, (c) G–Fe–Np(1) , (e) G–Fe–Np(2) and PDOS of 3d
orbitals of (b) G–Cr–Np, (d) G–Fe–Np(1) , (f) G–Fe–Np(2) . The Fermi level is positioned
at 0.

(as much as 98%). Subsequently, we calculate the degree of spin
polarization in each 3d subshell by analyzing the partial DOS (PDOS)
of 3d orbitals. Compared to the other 3d orbitals, the 3dz2 subshell
is most polarized as shown in Figure 2b. The spin polarization terms
of metal 3d orbitals in all investigated structures are summarized
in Table 2.
The most stable form of G–Fe–Np(1) (shown in Figure 1b) is

0.11 eV lower in energy compared to G–Fe–Np(2) (Figure 1c). The
Table 2
Spin polarization terms (

B)

of G, Np, and 3d orbitals of each metal atom.

G

Np

3dz2

3dzx

3dzy

3dx2 −y2

3dxy

−0.03
0.17
0.05
−0.04

−0.01
0.02
−0.20

−0.19

G–Fe2 –Np(1)

0.01

−0.14

G–Fe2 –Np(2)

0.07

−0.01

0.12
0.05
0.28
0.42
0.42
−0.01
0.09
0.02
0.02

0.00
0.64
0.57
0.10
0.10
−0.10

0.43
0.33
0.34

0.00
0.55
0.56
0.18
0.18
−0.11
0.60
0.18
0.18

0.01
0.20
0.33
0.42
0.42
−0.03
0.16
0.05
0.05

0.01
0.25
0.31
0.23
0.23
−0.05

0.13
0.07
0.07

G–Cr–Np
G–Fe–Np(1)
G–Fe–Np(2)
G–Cr2 –Np

geometric configuration of G–Fe–Np(1) is somewhat similar to that
of G–Cr–Np, i.e. Fe is bound to two honeycomb units in naphthalene and graphene. Binding energy calculations (Table 1) suggest
that the Fe–naphthalene interaction is stronger than the interaction between Fe and graphene in both G–Fe–Np(1) and G–Fe–Np(2) .
The metastable configuration, G–Fe–Np(2) , has an odd bonding
behavior, in which Fe only establishes a bond to one C atom from
graphene, while it interacts with six C atoms from naphthalene
with different bond distances. Lowdin charge analysis [32] shows
that the charge of Fe (+0.26) in G–Fe–Np(1) is actually less positive than that (+0.31) in the metastable state, G–Fe–Np(2) . Indeed,
this is explainable because the Fe atom in G–Fe–Np(2) locates
on top of C in graphene, which allows the 2pz orbital of that
C atom to receive more partial charge from the metal. In terms
of magnetic alignments, G–Fe–Np(1) and G–Fe–Np(2) exhibit total
magnetizations of 1.89 B /cell and 2.00 B /cell, respectively. Interestingly, the absolute magnetization of G–Fe–Np(2) (3.27 B /cell)
is 37% higher than the absolute magnetization of G–Fe–Np(1)
(2.39 B /cell), which indicates higher anti-ferromagnetic effect in
G–Fe–Np(2) (see Figure 2b and c).
An interesting phenomenon can be observed as we look at
the binding energies of G–Fe–Np(1) and G–Fe–Np(2) . Even though
(a)
(b)
G–Fe–Np(1) is the most stable configuration, its Ebinding and Ebinding

are lower than those of G–Fe–Np(2) . Therefore, the adsorption of
naphthalene on graphene–Fe (with Fe sitting on top of a C atom)
would stabilize the valence electrons of Fe better; as a result, the
binding energy for this process becomes higher. When graphene is
directly decorated with the Fe–naphthalene complex, the binding
energy results suggest that Fe favor to bind on top of C. Overall, the
binding energy results suggest the favorability of C10 H8 adsorption
on graphene–Fe in actual experimental synthesis.
In general, the use of metal dimer enhances the bonding strength
between C10 H8 and graphene. When a Cr dimer is utilized to bridge
C10 H8 and graphene, only one stable configuration can be observed.
Two Cr atoms are likely to establish bonds with the honeycomb
units in both naphthalene and graphene. Moreover, there is an
˚
interaction between the two Cr atoms with a distance of 2.65 A.
In this case, two aromatic rings in C10 H8 are held tightly by the
two metal atoms. Thus, the resulted binding energies are larger
than the previously reported cases. Again, we conceive that covering graphene–Cr2 with C10 H8 might be more favorable, because
(b)
(a)
Ebinding is larger (3.34 eV) than Ebinding . While the use of a single
bridging Cr atom results in a small magnetic moment, we observe
that the use of Cr dimer raises the total magnetic moment to


V.Q. Bui, H.M. Le / Chemical Physics Letters 614 (2014) 238–242

Figure 3. Total DOS of (a) G–Cr2 –Np, (c) G–Fe2 –Np(1) , (e) G–Fe2 –Np(2) and PDOS
of 3d orbitals of (b) G–Cr2 –Np, (d) G–Fe2 –Np(1) , (f) G–Fe2 –Np(2) . The Fermi level is
positioned at 0. The PDOS of the second metal atom are given in dashed lines.

2.47 B /cell (MA = 3.36 B /cell). From Lowdin charge analysis, it is
found that the spin of one Cr atom do not oppose the other; in
fact, both contribute similar positive amounts of magnetization
(1.34 B ), while graphene and naphthalene give negative (antiferromagnetic) moments (Figure 3a and Table 2). Among the 3d
orbitals, 3dz2 and 3dx2 −y2 are the most polarized subshells with
spin polarization terms of 0.42 B as shown in Figure 3b and Table 2.
The other 3d subshells give less significant magnetic contributions
ranging from 0.10 B to 0.23 B .
In the last case, we investigate the possibility of attaching C10 H8
on graphene using two Fe atoms. We observe two distinctive
equilibrium structures (shown in Figure 1e and f) as introduced
previously. The former structure, which has one Fe atom locating
on top of C in graphene, is more stable by 0.20 eV in total energy.
This result is somewhat contradicting to the previous observations in the G–Fe–Np cases. Recall that when bridging of C10 H8 and
graphene using one Fe atom, the structure with Fe located on the
hollow sites of two honeycomb units is more energetically stable.
(a)
(b)
The binding energies (Ebinding and Ebinding ) of G–Fe2 –Np(1) are
1.62 eV and 2.10 eV, respectively, and those of G–Fe2 –Np(2) are 1.42
and 2.43 eV, respectively. At this point, the calculated binding energies allow us to arrive at the first conclusion regarding experimental
synthesis, i.e. attaching the C10 H8 group on a metal–graphene sur(b)
face should be more favorable due to the fact that Ebinding is higher
(a)

than the corresponding Ebinding in most cases. Both structures
exhibit smaller magnetic moments (1.03 B /cell from G–Fe2 –Np(1)
and 1.34 B /cell from G–Fe2 –Np(2) ) compared to the single Fe
case (G–Fe–Np). Especially, G–Fe2 –Np(1) is the sole case in which

we observe a small anti-ferromagnetism from one Fe atom (see
Table 2). For G–Fe2 –Np(2) , in terms of bonding and magnetic alignments, the roles of two Fe atoms are almost similar as illustrated
by the PDOS distributions (Figure 3f).
It is of particular interest to examine how G–Fe–Np(1) could be
transformed to G–Fe–Np(2) and vice versa. This can be achieved by
performing a nudged-elastic-band (NEB) scan [33]. By optimizing
10 intermediate structures, we have found activation energies of
0.84 and 0.73 eV for the forward and backward directions, respectively. At the transition state, the structure is found to exhibit a
total magnetic moment of 2.11 B , slightly higher than those found
in the two equilibrium structures. Moreover, a local minimum is
also found near the transition state as shown in Figure 4a. The

241

Figure 4. Relative energy profiles of (a) G–Fe–Np(1) ↔ G–Fe–Np(2) and
G–Fe2 –Np(1) ↔ G–Fe2 –Np(2) transformations given by NEB scans using PBE
calculations without D2 corrections. The reaction barriers of both transformations
are found to be above 0.8 eV.

same procedure is executed to find 10 intermediate structures of
the G–Fe2 –Np(1) ↔ G–Fe2 –Np(2) transformation; however, due to
energy convergence difficulty, a rough result is reported herein,
which indicates a barrier height of 0.88 and 0.67 eV for the forward
and backward directions, respectively (Figure 4b). The magnetic
moment found at the transition state is 0.71 B , which is lower
than those of the two equilibrium structures. From the NEB results,
we can make a second conclusion, i.e. the mobility of Fe atoms on
the graphene surface is more prohibited due to strong interactions
between Fe and naphthalene.
In summary, we have shown in this study that naphthalene

can be attached to the graphene surface via coordination bonds
with Cr/Fe atom or dimer with great stability. In a previous study,
the binding of transition metal dimers on graphene was found
to be weak and have high surface mobility [12]. With the use of
naphthalene as a binding ligand on metal, the metal–graphene
binding is shown to be more stabilized. As a reference for experimental synthesis, we adopt different binding energy evaluation
schemes, and the adsorption of C10 H8 molecule on a metal-attached
graphene surface is more thermodynamically favorable in most
cases; in other words, the metal–naphthalene bond is stronger
than the interaction between metal and graphene. The inclusion
of dispersion corrections for long-range interactions shows that
the attachment of Np on graphene–metal is further enhanced by
0.78–1.40 eV. When Fe is utilized as bridging atoms, we observe
different configurations of G–Fe–Np/G–Fe2 –Np, which possess different structure stability and magnetic moments. Overall, good
structural stability and interesting magnetism of the investigated
nanostructures may elicit potential spintronic and electronic applications.
Supplementary material
˚ of the optiThe unit cell parameters and atomic positions (in A)
mized structures (with and without D2 dispersion corrections) are
given in one supplementary file.
Acknowledgement
We are grateful to the computing support from the University
of Science, Vietnam National University. This work is funded by
Vietnam National University under grant B2014-18-03.


242

V.Q. Bui, H.M. Le / Chemical Physics Letters 614 (2014) 238–242

Appendix A. Supplementary data
Supplementary material related to this article can be found, in
the online version, at doi:10.1016/j.cplett.2014.09.047.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]

K.S. Novoselov, et al., Science 306 (2004) 666.
K.S. Novoselov, et al., Nature 438 (2005) 197.
R. Prasher, Science 328 (2010) 185.
C. Berger, et al., J. Phys. Chem. B 108 (2004) 19912.
H.C. Schniepp, et al., J. Phys. Chem. B 110 (2006) 8535.
E. Rudberg, P. Salek, Y. Luo, Nano Lett. 7 (2007) 2211.
Y.W. Son, M.L. Cohen, S.G. Louie, Nature 444 (2006) 347.
M. Wu, E.-Z. Liu, M.Y. Ge, J.Z. Jiang, Appl. Phys. Lett. 94 (2009) 102505.
C. Cao, M. Wu, J. Jiang, H.-P. Cheng, Phys. Rev. B 81 (2010) 205424.
V. Zólyomi, A. Rusznyák, J. Kürti, C.J. Lambert, J. Phys. Chem. C 114 (2010) 18548.
S. Malola, H. Häkkinen, P. Koskinen, Appl. Phys. Lett. 94 (2009) 043106.
H. Johll, H.C. Kang, E.S. Tok, Phys. Rev. B 79 (2009) 245416.

X. Liu, C.Z. Wang, Y.X. Yao, W.C. Lu, M. Hupalo, M.C. Tringides, K.M. Ho, Phys.
Rev. B 83 (2011) 235411.
[14] H. Sevinc¸li, M. Topsakal, E. Durgun, S. Ciraci, Phys. Rev. B 77 (2008) 195434.
[15] G. Cantele, Y.S. Lee, D. Ninno, N. Marzari, Nano Lett. 9 (2009) 3425.

[16] A. Krasheninnikov, P. Lehtinen, A. Foster, P. Pyykkö, R. Nieminen, Phys. Rev.
Lett. 102 (2009) 126807.
[17] P. Hohenberg, Phys. Rev. 136 (1964) B864.
[18] W. Kohn, L.J. Sham, Phys. Rev. 140 (1965) A1133.
[19] S.M. Avdoshenko, I.N. Ioffe, G. Cuniberti, L. Dunsch, A.A. Popov, ACS Nano 5
(2011) 9939.
[20] V.Q. Bui, H.M. Le, Y. Kawazoe, D. Nguyen-Manh, J. Phys. Chem. C 117 (2013)
3605.
[21] H.M. Le, H. Hirao, Y. Kawazoe, D. Nguyen-Manh, Phys. Chem. Chem. Phys. 15
(2013) 19395.
[22] P. Giannozzi, et al., J. Phys. Condens. Matter 21 (2009) 395502.
[23] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.
[24] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 78 (1997) 1396.
[25] D. Vanderbilt, Phys. Rev. B 41 (1990) 7892.
[26] A. Dal Corso, Phys. Rev. B 64 (2001) 235118.
[27] D.F. Shanno, J. Optim. Theory Appl. 46 (1985) 87.
[28] E.G. Gordeev, M.V. Polynski, V.P. Ananikov, Phys. Chem. Chem. Phys. 15 (2013)
18815.
[29] S. Grimme, J. Comput. Chem. 27 (2006) 1787.
[30] V. Barone, M. Casarin, D. Forrer, M. Pavone, M. Sambi, A. Vittadini, J. Comput.
Chem. 30 (2009) 934.
[31] H.M. Le, H. Hirao, Y. Kawazoe, D. Nguyen-Manh, J. Phys. Chem. C 118 (2014)
21057.
[32] D. Sanchez-Portal, E. Artacho, J.M. Soler, Solid State Commun. 95 (1995) 685.
[33] D. Sheppard, R. Terrell, G. Henkelman, J. Chem. Phys. 128 (2008) 134106.