Polyhedron 134 (2017) 376384

Contents lists available at ScienceDirect

Polyhedron
journal homepage: www.elsevier.com/locate/poly

Electronic structures of endohedral fullerenes with scandium, titanium

and iron atoms and metal-carbon clusters
M.V. Ryzhkov a,, N.I. Medvedeva a, B. Delley b
a
Institute of Solid State Chemistry, Ural Branch of the Russian Academy of Sciences, 620990 Ekaterinburg, Russian Federation
b
Paul Scherrer Institut WHGA 123, CH-5232, Villigen PSI, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history: The electronic structures of the endohedral forms of the C28, C40, C60 and C80 fullerenes encapsulating the
Received 14 February 2017 3d-atoms Sc, Ti and Fe, and in the case of C80 with M2C2 clusters (M = Sc, Ti, Fe) were investigated using
Accepted 3 June 2017 the ab initio Dmol3 method. For the C60 and C80 cages, we found preferable positions of the metal atoms
Available online 30 June 2017
and their carbon clusters near the internal surface of the fullerene shell. The analysis of the formation
energy variation and the role of the magnetic state of 3d-atoms is presented.
Keywords: 2017 Elsevier Ltd. All rights reserved.
Fullerenes
Metallocarbohedrenes
Ab initio methods
Stability
Magnetic ordering

1. Introduction
The discovery of fullerenes in 1985 [1] initiated the synthesis of
many useful materials based on carbon cages of various sizes.
Nowadays, there is an opinion that the most prospective modifica-
tions are based on the so-called endohedral fullerenes (EFs),
formed by the encapsulation of various micro-objects inside the
fullerene cages [2]. According to the results of experimental studies
[27], the EFs interior can contain individual atoms, molecules,
metal clusters, binary clusters of MxCy or MxNy types and more
complex objects as well. Such species are interesting because of
the possibility of their applications in biology and medicine; on
the other hand, electron transfer from the internal species to the
carbon cage makes these systems an ideal object for investigations
of the properties and behavior of entrapped atoms or clusters [8,9].
Analysis of the experimental works showed that the probabili-
ties for the encapsulation of various atoms inside the fullerene
cages are considerably different using both the arc discharge and
laser ablation methods for their generation. Additionally, the
majority of the elements in the periodic table were never detected
inside the fullerene shells, and the production yield of EFs with
metals is higher than that with non-metals. However, the probabil-
ities of EF formation with various metals are considerably different,
e.g. the elements Sc and Y (as well as their clusters with carbon and
nitrogen) give rise to endohedral particles in higher yields than Ti
Corresponding author. Fax: +7 343 3744495.
E-mail address: ryz@ihim.uran.ru (M.V. Ryzhkov).
and Zr [7]. It is interesting that most transition metals could not
form EFs, but the presence of Cu, Ni and Fe can increase the yield
of other metal endohedral species [7,10]. The nature of these cat-
alytic processes is not yet clear.
The aim of the present work was quantum-chemical modelling
of EFs containing atoms and metalcarbon clusters of the 3d-met-
als Sc, Ti and Fe; a comparison of the peculiarities of the electronic
structures of these objects, which can be responsible for the differ-
ent probabilities of their formation.
2. Objects of the calculations
In the present paper, we considered EFs of four sizes: (1) the
smallest stable endofullerenes M@C28 (Td); (2) M@C40 (Td), the size
of C40 being close to optimal for one-metal-atom encapsulation; (3)
M@C60 with non-central positions of one metal atom; (4)
M2C2@C80 with the metalcarbon clusters inside the fullerene cage
(M = Sc, Ti, Fe). The shapes of these particles are shown in Fig. 1.
The results of the syntheses of endohedral M@C28 (Td) clusters
with Ti, Zr and actinides are reported in Refs [1113]. The elec-
tronic structure calculations of these objects [1416] focused on
the properties of the heavy elements inside this quite small cage.
In our earlier DFT calculations on neutral An@C28 particles [17],
performed for the almost all actinides (An) from Th to Md, we
found that the variation of the dissociation energy (Ed) for the
An@C28 ? An + C28 reaction was not monotonous across the acti-
nide row and that Pa@C28 was the most stable cluster in this series
(Ed  12 eV). On the other hand, the charge transfer from the metal

http://dx.doi.org/10.1016/j.poly.2017.06.032
0277-5387/ 2017 Elsevier Ltd. All rights reserved.
 M.V. Ryzhkov et al. / Polyhedron 134 (2017) 376384
M@C28 M@C40 M@C60 M2C2@C80
Fig. 1. Geometrical structures of M@C28, M@C40, M@C60 endohedral metallofullere-
nes and M2C2@C80 cluster endofullerenes. The dark spheres correspond to atoms
inside the fullerene cage.
atom to the fullerene cage monotonously decreased in this row
from 2.1e (Th) to 0.9e (Md). The corresponding properties of
3d-metals inside the C28 cage were not evaluated.
The strong similarity between the properties of empty C28 and
C40 fullerenes was first reported in Ref. [18]; moreover, according
to Ref. [19], the C40 cluster of Td symmetry is highly reactive and
could behave as tetravalent superatom. However, theoretical sim-
ulations of empty C40 cages of various symmetries [20,21] showed
that the pristine D2 isomer was slightly more stable than that of Td
symmetry. On the other hand, our calculations of various endohe-
dral An@C40 clusters with actinide atoms from Th to Md [22]
showed that EFs of Td symmetry were noticeably more stable than
the corresponding endohedral particles of the D2 type. We also
found that the variation of Ed for the An@C40 ? An + C40 reaction
was not monotonous and that Pa@C40 (Td) was the most stable
cluster in this row (Ed  6.4 eV). In contrast to An@C28, the charge
transfer from the metal atom to the C40 cage was not monotonous:
in the first half of the row, the effective charge on the actinide atom
(QAn) decreased from 2.2e (Th) to 1.3e (Am), after that QAn
increased up to 1.6e for Cm and Bk and then again decreased to
1.0e for Md. On the base of these results, the present modelling
of Sc, Ti and Fe atoms behavior inside the C40 fullerene was per-
formed for particles of Td symmetry (Fig. 1).
The radius of icosahedral C60 fullerene (nearly 3.6 ) is too large
for one metal atom, even for uranium, which was experimentally
detected inside this cage [23]. This means that a non-central posi-
tion of the encapsulated atom also could be realized in M@C60 EFs.
For this reason, in our earlier work [24], we investigated the stabil-
ity of An@C60 complexes (An = Th-Md) with central and non-cen-
tral positions of the metal atom. The results of our calculations
showed that in the first half of the actinide row, the values of Ed
for the clusters with non-central An positions were higher than
those for the most symmetrical particles by 14 eV. For the heavier
actinides, the corresponding energy gain was in the 0.11.0 eV
range. We found that in the case of a non-central starting position,
the metal atoms closely approached a hexagonal carbon ring
(Fig. 1), and the final symmetry for the majority of the An@C60 clus-
ters was Cs. These results are in agreement with the Li@C60 exper-
imental structure, which was established precisely by XRD
crystallography [25]. Thus, in the present paper, we study the elec-
tronic structure of M@C60 EFs (M = Sc, Ti, Fe) corresponding to Cs
symmetry, with the metal atoms located near the C6 hexagon, as
shown in Fig. 1.
The possibility to generate the C80 fullerenes encapsulating var-
ious objects (M, M2, M2C2, M3N, etc.) was experimentally con-
firmed in many works (see the reviews of Lu et al. [7,8]). In the
species with two and more entrapped atoms, the interplay of the
interaction inside the encapsulated object and the interaction
between the encapsulated atoms and the carbon cage is of essen-
tial interest in connection with the stability and production yields
of such systems. DFT calculations of Ti2@C80 [26] showed that the
two Ti atoms are located as non-interacting ions near opposite
377
Cs D2h C2v
Fig. 2. Geometrical structures and point symmetries of the three most stable
isomers of the M2C2 clusters. The large spheres correspond to metal atoms.
sides of the C80 wall. On the other hand, in the cases of encapsu-
lated clusters containing four, five and more atoms, the interac-
tions between these MxCy, MxNy and other objects are certainly
important [79].
Earlier in our investigations on the electronic structure of small
iron-carbon clusters [27], the binomial scheme for the theoreti-
cal generation of all stable isomers was suggested. In the case of
the Fe2C2 particle, twelve isomers were found. In Fig. 2, the three
most stable structures of Cs, D2h and C2v symmetry are shown.
An analogous study performed for the Sc2C2 and Ti2C2 clusters
showed that the three lowest-energy isomers were also of Cs, D2h
and C2v types; however, the order of stability of these structures
for Sc, Ti and Fe was different. In Table 1, we have listed the
atomization energies (Ea) for the three most stable M2C2 isomers
(M = Sc, Ti, Fe). The Ea (which is sometimes called the binding
energy or the total bond energy) of a neutral system is defined
as Ea = Etot Esum, where Etot is the total energy of the system and
Esum is the sum of the total energies of all isolated neutral atoms in
the system. As seen in Table 1, the distorted parallelogram of Cs
symmetry is the ground structure for the Ti2C2 and Fe2C2 clusters,
while for the Sc2C2 particle this structure is the second isomer. On
the other hand, rhombuses of the D2h type are the second-most-
stable structure for the titanium and iron clusters and the
ground-state isomer for the Sc2C2 system. Only the three-dimen-
sional structure of the C2v type is the third-most-stable isomer
for all three clusters (Table 1). The results of these calculations also
showed that the electronic structure of the titanium clusters con-
verged to a spin-restricted configuration, whereas in the case of
the iron species and the two most stable scandium particles,
high-spin states of the metal atoms and ferromagnetic (FM) order-
ing of their spin moments were obtained (Table 1). Note, the mag-
netic moments are localized primarily on the metal atoms.
Since the aim of the present paper was the search for the most
stable M2C2@C80 clusters, hereby we considered all three lowest-
energy M2C2 isomers (M = Sc, Ti, Fe) as the core of C80 fullerene.
In the initial configurations, the nine M2C2 clusters (with optimized
geometries for isolated objects) were located near the center of C80
cage (Fig. 1), then these initial structures were subjected to geom-
etry optimization, leading to new equilibrium positions of all the
atoms in the systems.
3. Computational methods
Geometry optimization of all the clusters was performed using
the DMol3 method [28] with a double numerical atomic basis set
with polarization functions (dnp). The generalized gradient
approximation (GGA) in the PBE form [29] was used in all calcu-
lations. The Coulomb potential was computed using the model
density obtained as the decomposition of electron density into
multipolar components, including those with l = 3 (octupole).
378 M.V. Ryzhkov et al. / Polyhedron 134 (2017) 376384
Table 1
Atomization energies (eV) and total magnetic moments (in parenthesis) obtained for
the three most stable isomers of the M2C2 clusters (M = Sc, Ti, Fe).
Symmetry/Cluster Sc2C2 Ti2C2 Fe2C2
17.26 (2 lB) 18.82 (0 lB) 16.65 (6 lB)
D2h
17.13 (2 lB) 18.93 (0 lB) 17.06 (6 lB)
Cs
16.59 (0 lB) 18.57 (0 lB) 16.59 (6 lB)
C2v
Optimization of the cluster geometry was done until the change in
the value of the maximum energy gradients was less than 0.001
atomic units (Hartree/Bohr), the atomic displacements converged
to within 0.001 and the global orbital cutoff was 8.0 . To con-
firm that the stationary points correspond to minima, the vibra-
tional frequencies were computed for some of the investigated
clusters.
The scheme of Hirshfeld [30] was used for the calculation of
atomic charges (QAt) and spin-moments (SAt). In this approach,
the partition of electron density between atoms at each point r
of the spatial integration is made by the use of a partition func-
tion, which is large for small r RA (A is an atom index) and small
for large r RA. A shortcoming of this scheme is the dependence of
QAt and SAt on the shape of the partition function. On the other
hand, the most natural boundary for charge density partitioning
is the locus of points of the q(r) minimum [31]. A method for the
realization of this Baders idea using numerical spatial integration
incorporated in the Discrete Variational method (DVM) [32,33]
was reported in Ref. [34]. In our previous works, this scheme was
applied to FenCm nanoparticles [27,35], as well as to endohedral
fullerenes [17,22,24] and clusters of actinide dioxides [36]. A com-
parison of various methods for atomic effective charge calculations
showed that the most realistic values were obtained using the spa-
tial integration method [34]. For this reason we also used the DV
method in the present work. For each final cluster configuration
obtained by DMol3, the DV calculation was carried out, where
QAt and SAt were computed as the integrals of the charge and spin
density inside the domain bounded by the points of their
minimum.
4. Results and discussion
4.1. Endohedral metallofullerenes M@C28 (M = Sc, Ti, Fe)
According to previous investigations, the optimized structure of
pristine C28(Td) is a polyhedron containing the three groups of non-
equivalent carbon atoms, with the cage radius in the range 2.40
2.48 for the various types of C atoms. To obtain the properties
of M@C28 clusters, we performed geometry optimization of three
EFs, where Sc, Ti and Fe were located in the center of the C28 cage.
For an estimation of the stabilizing effect of the metal atom inside
the fullerene we used the formation energy values Eform, which is
the difference between the atomization energies of the optimized
M@Cn cluster (n = 28, 40, 60) and the empty Cn cage. The latter
Table 2
Atomization (Ea) and formation (Eform) energies (eV), effective charges (e) and spin moments (lB) on the metal atoms in the M@C28 clusters (M = Sc, Ti, Fe) obtained in DMol3
calculations using the Hirshfeld scheme and in DV calculations using spatial integration.
Cluster Ea Eform
Sc@C28 204.33 6.85
Ti@C28 204.06 6.58
Fe@C28 198.48 1.00
value for the C28 particle, obtained in the previous calculations, is
197.48 eV. The results obtained for the three M@C28 clusters
are summarized in Table 2. We had found that all the M@C28 par-
ticles are more stable than the pristine C28 fullerene. The value of
the atomization energy of the most stable Sc@C28 cluster is less
than that of the C28 cage by 6.85 eV; in the least stable case,
Fe@C28, the decrease of Ea is near to 1 eV (Table 2). Analysis of
the geometrical parameters showed that the deformation of the
cage due to addition of metal atoms was small: the radial expan-
sion of the MC distances was near to 0.04 , which was noticeably
less than 0.9 in the case of actinide atom encapsulation [17].
As seen in Table 2, the non-magnetic (or spin-restricted) config-
uration was obtained for the Sc@C28 and Ti@C28 species, whereas in
the optimized Fe@C28, the spin moment of the metal atom was
close to 3.5 lB. Note that at the beginning of our investigations,
the calculations of all clusters were spin-unrestricted, i.e. in the
starting point, the formal (non-zero) spin moments of the isolated
atoms were used as the initial values. The obtained results mean
that the electronic structure of the scandium and titanium clusters
converged to the non-magnetic (or spin-restricted) configuration
of all atoms in these species. Surprisingly, similar values of SFe near
3.5 lB were obtained using both methods of electronic structure
calculation and the different schemes of spin density partition.
On the other hand, the effective charges on the metal atoms com-
puted by DMol and DV methods were considerably different
(Table 2). The values of QM obtained using the integral procedure
[34] are more than two times larger than those computed using
the Hirshfeld scheme. It is worth mentioning that the values of
the charge transfer from the metal atom to the carbon cage
obtained using the Hirshfeld method do not correlate with the
variation of Eform, whereas the decrease of the corresponding val-
ues computed by spatial integration [34] agrees with the Eform vari-
ation on going from the Sc to Fe clusters.
According to both methods, the highest occupied molecular
orbital (HOMO) was of the C2p type with small admixtures of
C2s states in all the M@C28 complexes. However, the nature of
the lowest unoccupied molecular orbital (LUMO) was different in
these objects. In the case of Sc encapsulation, the Fermi level (used
as a zero of the energy scale) was located between two nearly
degenerate molecular orbitals (MOs) of C2p character and the
HOMOLUMO gap was less than 0.1 eV. Note that both MOs were
not localized on some specific carbon sites, but were rather dis-
tributed over the cage. The energy of the lowest Sc3d MO was
higher than EF by 3.5 eV. In the Ti@C28 complex, the LUMO contain-
ing the main contribution from the Ti3d atomic orbital (AO) was
shifted from the HOMO to a higher energy by 1.48 eV. It is worth
mentioning that the vacant Ti3d MOs were localized inside the
cage, and the lowest vacant C2p orbital distributed over the cage
was higher than EF by 2.18 eV. In the spin-polarized case of iron
atom encapsulation, the occupied levels of the spin up Fe3d MOs
were located near the center of the occupied part of the C2p band
in the energy region from 4.9 to 3 eV below EF, while the vacant
spin down MOs of Fe3d character were located 0.5 eV higher than
EF. As in the case of scandium, both the HOMO and LUMO were of
C2p type and the corresponding gap was near 0.1 eV. Thus, the
levels of the vacant M3d as well as the M4s states became closer
DMol3 DVM
QM SM QM SM
0.34 0.00 1.36 0.00
0.41 0.00 1.17 0.00
0.36 3.45 0.95 3.48
 M.V. Ryzhkov et al. / Polyhedron 134 (2017) 376384
to the occupied part of the C2p band on going from Sc to Fe, fol-
lowed by the increase of M3d and 4s AOs admixtures in the occu-
pied C2p MOs, as a result the effective charges on metal atoms
decreased in this row. As seen in Table 2, the absolute values of
atomization and formation energies obtained for Ti@C28 were less
than those of the most energetically stable Sc@C28 by 0.27 eV,
while the HOMOLUMO gap of the former was considerably
greater than that of the latter and of the empty C28 cage (0.6 eV).
Thus, we can expect that the titanium complex is relatively more
stable than other M@C28 systems.
Finally, we tested the role of the magnetic state of the iron atom
in the Fe@C28 complex. To simulate a hypothetical non-magnetic
configuration of this system, we performed additional spin-
restricted calculations on Fe@C28, where the initial and final spin
moments were equal to zero for all atoms. The results we obtained
showed that the atomization energy increased up to 196.17 eV,
which was higher than that of the empty cage ( 197.48 eV) by
1.31 eV. Thus, the presence of the Fe atom in the non-magnetic
state inside the C28 fullerene made this cluster unstable with
respect to the Fe@C28 ? Fe + C28 reaction. Analysis of effective
charges showed that the charge transfer from the non-magnetic
iron atom to the carbon cage decreased in both DMol and DV
calculations.
4.2. Endohedral M@C40 fullerenes (M = Sc, Ti, Fe)
For the investigation of M@C40 EFs, we performed geometry
optimizations and calculated the atomization energies for the
three clusters, where the Sc, Ti and Fe atoms were located in the
center of the C40 (Td) cage (Fig. 1). According to previous investiga-
tions, the optimized structure of the C40(Td) fullerene is a polyhe-
dron with the three groups of non-equivalent carbon atoms, and
the cage radius calculated for various types of C atoms are in the
range 2.853.11 . For an estimation of the formation energy for
the optimized M@C40 species, we used the value of 291.07 eV
for the atomization energy of the empty C40 (Td) cage, as obtained
in the previous calculations. The results obtained for the optimized
M@C40 clusters are listed in Table 3. As in the previous case, we
found that the stability of all M@C40 particles is higher than that
of the pristine C40 fullerene. The absolute value of Eform for the
most stable Sc@C40 cluster is near 4.23 eV, which is noticeably less
than that of the Sc@C28 system (6.85 eV). A similar trend was
obtained for the titanium clusters, where the absolute value of
Eform decreased from 6.58 to 3.41 eV on going from Ti@C28 to
Ti@C40. On the other hand, in the least stable case of Fe@C40, a
small increase of |Eform| from 1.00 to 1.07 eV was obtained (Tables
2 and 3). However, it should be mentioned that the variation of
Eform on going from Sc to Fe is similar for both types of clusters.
According to our previous calculations on An@C40 [22], the
deformation of the cage radius due to addition of the actinide atom
was less than 0.09 ; in the present case of 3d atom encapsulation,
the changes of the geometrical parameters were less than 0.01 .
As seen in Table 3, the non-magnetic configuration was only
obtained for the Sc@C40 cluster, whereas in the optimized Ti@C40
and Fe@C40 particles, the spin moments of the metal atoms were
Table 3
Atomization and formation energies, effective charges and spin moments on the metal atoms obtained in the DMol3 and DV calculations for the M@C40 clusters (M = Sc, Ti, Fe).
Cluster Ea Eform
Sc@C40 295.30 4.23
Ti@C40 294.48 3.41
Fe@C40 292.14 1.07
379
larger than 1.5 and 3 lB, respectively. In the present case, the val-
ues of STi and SFe obtained in the DMol and DV calculations differ
by 0.11 and 0.33 lB, respectively (Table 3). However, the difference
in the spin-density integrals computed for metal atoms using the
two methods is considerably less than that in the values of QM
obtained using the Hirshfeld scheme and integral procedure [34].
The latter values of QM are three times larger than those obtained
in the DMol calculations (Table 3). It is worth mentioning that in
the case of M@C40 species, the charge transfer from the metal atom
to the carbon cage obtained in both approaches correlates with the
variation of Eform on going from Sc to Fe.
Similar to the M@C28 EFs, in all the M@C40 complexes, the
HOMOs were of the C2p type, as well as the LUMOs in the scan-
dium and iron systems. As a result, in the former and latter objects,
the HOMOLUMO gaps were less than 0.1 eV. Analysis of these
HOMO and LUMO structures showed that they were almost uni-
formly distributed over the C40 cage. In Sc@C40, the levels of the
vacant Sc3d MOs were located 0.85 eV higher than EF. In the
Ti@C40 complex, the gap between the HOMO of C2p character
and the lowest vacant Ti3d (spin up) LUMO was only 0.24 eV,
which was considerably less than the corresponding gap in
Ti@C28 (1.48 eV). However, these vacant Ti3d states were localized
inside the cage, whereas the energy of the first vacant MO of C2p
character (also uniformly distributed over the cage) was higher
than EF by 1.57 eV. The splitting between the Ti3d spin up and spin
down levels was near to 1 eV, whereas the corresponding value for
the Ti4s states was close to 0.5 eV. In the Fe@C40 molecule, the
levels of the spin up Fe3d orbitals were located near the center
of the occupied part of the C2p band in the energy region from
4.6 to 2.6 eV below EF, while the energies of the spin down
MOs of Fe3d character were higher than EF by 0.23 eV. As in the
case of smaller systems, the average energies of the vacant M3d
and M4s states became closer to the occupied part of the C2p band
on going from Sc to Fe. As a result, the increase of the M3d and 4s
orbitals admixtures in the occupied C2p MOs induced a decrease of
the effective charges on the metal atoms in this row (Table 3). The
absolute values of atomization and formation energies obtained for
Ti@C40 were less than those of Sc@C40 by 0.82 eV, however, the
above-mentioned structure of the vacant MOs allowed us to expect
the relative reactivity or solubility of Ti@C40 could be larger than
that of the scandium complex.
As in the previous case, we tested the role of the magnetic state
of the titanium and iron atoms in the M@C40 complexes. According
to the spin-restricted calculations performed for the Ti@C40 and
Fe@C40 clusters, the atomization energies increased up to
294.04 and 290.13 eV, respectively. Note, the latter value was
higher than that of the empty C40 cage by 0.94 eV. Thus, the pres-
ence of the non-magnetic Ti atom inside the C40 fullerene
decreases its stability by 0.44 eV, and the encapsulation of the Fe
atom in the non-magnetic state makes the Fe@C40 cluster unstable
with respect to the Fe@C40 ? Fe + C40 reaction. Analysis of the
effective charges obtained in the DMol and DV spin-restricted cal-
culations also showed that the charge transfer from the non-mag-
netic metal atoms to the carbon cage was less than that obtained in
both spin-polarized calculations.
DMol3 DVM
QM SM QM SM
0.49 0.00 1.53 0.00
0.37 1.71 1.33 1.60
0.35 3.14 1.01 3.47
380 M.V. Ryzhkov et al. / Polyhedron 134 (2017) 376384
4.3. Endohedral M@C60 fullerenes (M = Sc, Ti, Fe)
To evaluate the influence of the Sc, Ti and Fe atoms on the char-
acteristics of the M@C60 endohedral metallofullerenes, we per-
formed geometry optimizations and calculated the atomization
energies for the three clusters, where the Sc, Ti and Fe atoms were
located near the C6 hexagon of the carbon cage (Fig. 1). In the start-
ing configuration of these systems, we used the geometrical
parameters of a pristine C60 cage with Ih symmetry and an
atomization energy of 453.14 eV. As mentioned above, the pres-
ence of a metal atom inside the cage reduced the symmetry of the
final M@C60 species to Cs. The results obtained for the optimized
M@C60 clusters are summarized in Table 4. As in the previous
cases, we found that the stability of all M@C60 particles is higher
than that of the empty C60 fullerene; however, the most stable
M@C60 cluster was obtained for titanium atom encapsulation.
The absolute value of Eform for Ti@C60 is close to 3.88 eV, which is
slightly higher than the value of 3.62 eV obtained for the corre-
sponding scandium cluster and also higher than that of the
Ti@C40 system (3.41 eV). On the other hand, in the cases of the
Sc@C60 and Fe@C60 species, the expected decrease of |Eform| from
4.23 to 3.62 eV and from 1.07 to 0.75 eV, respectively, was
obtained (Tables 3 and 4). In the optimized geometry of the
Sc@C60 cluster, the ScC distances for the six carbon atoms of the
nearest C6 ring were in the range 2.232.28 , and the average
ScC bond length was close to 2.25 . The corresponding average
TiC and FeC distances were near to 2.15 and 2.62 , respectively.
As seen in Table 4, these geometrical parameters are in agreement
with the non-monotonous variation of Eform on going from Sc to Fe
clusters.
As in the case of the M@C40 species, a non-magnetic configura-
tion was obtained for the Sc@C60 cluster only. The spin moment of
Ti was close to 0.38 and 0.58 lB, according to DMol and DV calcu-
lations, respectively. Note, these values were noticeably less than
those obtained for the Ti@C40 particle. On the other hand, the spin
moment of the Fe atom inside the C60 cage (3.63 and 4.09 lB) was
higher than that in the optimized Fe@C40 cluster. As in the previous
cases, the values of QM computed using the Hirshfeld scheme were
nearly three times less than those obtained using the integral pro-
cedure [34]; however, the results of both approaches demon-
strated the same monotonous decrease of QM on going from the
Sc to the Fe clusters, which did not correlate with the non-mono-
tonous variation of Eform (Table 4).
In contrast to the M@C28 and M@C40 EFs, in all the M@C60 com-
plexes, the HOMO and LUMO were of the same character: in
Sc@C60 and Ti@C60, these orbitals were of the C2p type, and in
the Fe@C60 system, both orbitals were of Fe3d character. Although
only six carbon atoms were the nearest neighbors to the metal, the
total contributions from the C2p AOs of these sites to the HOMO
and LUMO were less than 18 and 16% in the Sc@C60 and Ti@C60
EFs, respectively; i.e., these orbitals were not localized on specific
atoms. In the scandium complex, the levels of the vacant Sc3d
MOs were distributed between 1.7 and 2.6 eV above EF. In the
spin-polarized case of Ti@C60, the considerable overlap of the
Ti3d (spin up) and C2p states was found near the Fermi level. As
a result of strong hybridization between the C2p and spin up
Table 4
Atomization and formation energies, effective charges and spin moments on the metal atoms obtained in the DMol3 and DV calculations for the M@C60 clusters (M = Sc, Ti, Fe).
Cluster Ea Eform
Sc@C60 456.76 3.62
Ti@C60 457.02 3.88
Fe@C60 453.89 0.75
Ti3d AOs, the nearly degenerated HOMO and LUMO also contained
noticeable admixtures (17%) of the Ti3d orbitals. The vacant Ti3d
spin down band was located within a 0.352.55 eV interval above
EF. In Fe@C60, the levels of the MOs with the main contributions
from the spin up Fe3d states were distributed among the occupied
part of the C2p band from 4.17 to 3.07 eV below EF. In contrast
to the smaller iron complexes, in Fe@C60, the Fermi level was
located just in the spin down Fe3d band, i.e. both the HOMO and
LUMO were of the Fe3d type, and the corresponding gap was close
to 0.1 eV. As in the cases of smaller systems, the energies of the
vacant M3d and M4s states became closer to the occupied part
of the C2p band on going from Sc to Fe. Consequently, the increase
of the M3d and 4s AOs admixtures in the occupied C2p MOs
induced a decrease of the effective charges on the metal atoms in
this row (Table 4). Thus, the HOMOLUMO gap was small in all
of the investigated M@C60 systems; however, in all these objects,
the HOMO and LUMO were not localized on specific atoms of the
cage surface. Consequently, we can expect a correlation between
the relative reactivity or solubility of these systems and the values
of their formation energies (Table 4).
Evaluation of the role of the Ti atoms magnetic state in the elec-
tronic structure of Ti@C60 showed that the atomization energy
obtained in the additional spin-restricted calculation of this cluster
was nearly the same ( 457.0 eV) as in the spin-polarized
approach. The average distance between the non-magnetic tita-
nium atom and the carbon atoms of the nearest hexagon (2.14 )
was also close to the corresponding value (2.15 ) obtained in
the spin-polarized calculation. In contrast to previous modelling
of EFs with a non-magnetic iron atom, the spin-restricted calcula-
tions of the Fe@C60 cluster showed a decrease of the atomization
energy from 453.89 to 454.23 eV and an increase of |Eform| from
0.75 to 1.09 eV. Nevertheless, the encapsulation of the Fe atom was
considerably less favorable than that of the scandium and titanium
atoms. Surprisingly, the non-magnetic state of the iron atom
induced an essential decrease of the average Fe-C distance from
2.62 to 2.01 . In contrast to the central positions of the metal
atoms inside the fullerene cages, the effective charges obtained
in the DMol and DV spin-restricted calculations of the Ti@C60
and Fe@C60 species showed different trends: the QTi (0.45) and
QFe (24) values computed using the Hirshfeld scheme were slightly
larger or nearly the same as in the spin-polarized calculations
(Table 4), while the integral DV charges of the non-magnetic met-
als decreased from 1.29 to 1.09 and from 0.86 to 0.77 for Ti and Fe,
respectively. In the latter case, this decrease of charge transfer did
not correlate with the above mentioned decrease of the FeC bond
lengths in spin-restricted calculation.
4.4. Endohedral M2C2@C80 cluster-fullerenes (M = Sc, Ti, Fe)
In the present work we considered the two types of C80 fuller-
ene: the most symmetrical Ih and the experimentally detected
C2v(5) [37,38] isomers. In the case of the more complex
M2C2@C80 EFs, the formation energies were computed as the differ-
ence between the atomization energy of the optimized M2C2@C80
particle and the sum of the atomization energies of the isolated
M2C2 cluster and the empty C80 cage. The latter values for the Ih
DMol3 DVM
QM SM QM SM
0.51 0.00 1.58 0.00
0.43 0.38 1.29 0.58
0.24 3.63 0.86 4.09
 M.V. Ryzhkov et al. / Polyhedron 134 (2017) 376384
and C2v(5) isomers obtained in the DMol3 PBE calculation were
close to 608.08 and 608.47 eV, respectively, and the corre-
sponding atomization energies for the M2C2 clusters are listed in
Table 1. Note that in the earlier and less accurate calculations of
empty C80 fullerenes, the obtained energies of both isomers were
either equal [39] or the Ih cage was defined as energetically more
preferable than the C2v(5) cage [40]. The results obtained for the
most stable M2C2@C80(Ih) and M2C2@C80(C2v) EFs are summarized
in Tables 5 and 6, respectively.
In the case of the symmetrical C80(Ih) cage, we found that the
maximum value of |Ea| for the scandium systems was achieved
for the least stable Sc2C2 (C2v) isomer, and as expected, the largest
value of |Eform| (8.08 eV) among the all investigated complexes was
obtained. For the three Ti2C2@C80(Ih) EFs, the maximum |Ea| was
obtained for the second Ti2C2 (D2h) isomer and only in the case
of iron, the encapsulation of most stable Fe2C2 (Cs) cluster gave
the maximum value of |Ea|. However, since the difference in energy
of the three isolated M2C2 isomers (Table 1) was higher than the
stabilizing effect of their encapsulation in C80(Ih) cage (Table 5),
the maximum values of |Eform| were obtained for the third C2v iso-
mer for all the metal atoms.
In the modelling of complexes with the less symmetrical
C80(C2v) fullerene, the interaction between the encapsulated clus-
ter and the cage induced considerable deformation of the M2C2
geometry, depending on the initial orientation of the cluster with
respect to the cage wall (we tested the three initial orientations
of the symmetry plane for each cluster type). For instance, the
geometry optimization of the encapsulated Sc2C2 butterfly (C2v)
particle induced a considerable increase of the Sc-C2-Sc dihedral
angle (from 74 to 124) in one calculation and converged to a pla-
nar structure in the other calculation. Since the energies of these
optimized complexes were nearly the same, both are shown in
Table 6. It is interesting to compare the features of these most
stable Sc2C2@C80(C2v) complexes with the corresponding results
of experimental works [37,38]. According to experiment [38], the
Sc-C2-Sc dihedral angle of the butterfly type Sc2C2 cluster was
127 or 131, which is larger than our value of 124. The experimen-
tal value of the CC bond length in the Sc2C2 particle was close to
1.2 , whereas in our calculation, a value of 1.28 was obtained for
the C2 dimer. Kurihara et al. [38] found that one Sc atom was near a
hexagon and the second was near a [6,6] bond, while the both the-
oretical positions of metal sites were near [6,6] bonds. On the other
hand, the shortest Sccage distances obtained in our calculations
(2.21 ) were in close agreement with the experimental values
(2.192.23 ) [38]. It is worth mentioning that in our complex with
the planar Sc2C2 cluster (Table 6), one scandium atom interacts
with a hexagon and the second atom is near a [6,6] bond. The pos-
sibility of planar Sc2C2 cluster formation inside a C80(C2v) cage was
reported in Ref. [37].
As in the case of the Ti2C2@C80(Ih) EFs, the maximum |Ea| was
obtained for the second Ti2C2 (D2 h) isomer encapsulated in the
C2v(5) cage; however, in contrast to the Ih fullerene, the difference
in energy of the three isolated Ti2C2 isomers (Table 1) was less than
Table 5
Atomization and formation energies for the EFs encapsulating the three most stable isomers of the M2C2 clusters (M = Sc, Ti, Fe). The sum of the spin moments on the iron atoms
(lB) obtained using Hirshfeld (H) and DV integral procedures (I) in the calculations of Fe2C2@C80(Ih) complexes.
Cluster Sc2C2@C80(Ih)
Ea Eform
632.46 7.12
632.29 7.08
632.75 8.08
381
the stabilizing effect of their encapsulation in the C80(C2v) cage
(Table 6), and the maximum value of |Eform| was also obtained
for the D2h isomer (5.93 eV). The energies obtained for the iron
complexes were noticeably less than the corresponding values
for the scandium and titanium systems, and as in the case of the
titanium clusters, the encapsulation of the Fe2C2 (D2h) isomer gave
the maximum values of |Ea| and |Eform| (Table 6).
In contrast to the smaller fullerenes, in all the M2C2@C80(Ih)
complexes, the HOMO and LUMO were of the same C2p character,
even in the case of the iron clusters, though the admixtures of the
Fe3d AOs in these orbitals increased up to 22%. The HOMOLUMO
gap in all the Ih EFs were equal or less than 0.1 eV. On the other
hand, in the majority of the M2C2@C80(C2v) systems, the gap was
quite large. A maximum value of 0.8 eV was obtained for the com-
plex with the Sc2C2 cluster of the butterfly type, with almost a
maximum value of |Ea| (632.29 eV) where the LUMO was of Sc3d
character. The second gap of 0.55 eV was also obtained for the EF
with a Sc2C2 cluster of the trapezium shape (|Ea| = 631.93 eV),
where both HOMO and LUMO were of the C2p type with noticeable
contributions from Sc3d AOs (3 and 13%, respectively). Surpris-
ingly, in the complex with the Sc2C2(D2h) cluster, the gap was only
0.05 eV, although the LUMO was of the Sc3d nature, as in the most
stable scandium EF. In the titanium systems, the gap varied from
0.2 eV for the butterfly cluster to 0.4 eV for the rhombuses clus-
ter (|Ea| = 633.22 eV) and to a maximum value of 0.5 eV for the
trapezium Ti2C2 cluster. Note, only in the latter case was the
LUMO of the Ti3d type, though in the two other titanium com-
plexes the main contributions to both HOMO and LUMO were from
C2p states. In the Fe2C2@C80(C2v) EFs, the levels of the MOs with
the main contributions from Fe3d states of both spin directions
were distributed among the occupied part of the C2p band. The
interchange of the C2p and Fe3d levels in the region below and
above EF resulted in a variation of the HOMO and LUMO character:
in the complex with the C2v cluster, the former was of C2p and the
latter of Fe3d type, while in the two other iron systems, both orbi-
tals were of the Fe3d nature. Note, in all the investigated M2C2@C80
systems, the HOMO and LUMO were not localized on specific
atoms of the cage surface. It is worth mentioning that only in the
Sc2C2@C80(C2v) complexes were the largest absolute values of for-
mation energy and the HOMOLUMO gap achieved simultane-
ously. The obtained results allowed us to expect that the
maximum relative reactivity or solubility of investigated EFs had
to be realized in the Sc2C2(butterfly)@C80(C2v) complex.
As mentioned above, the interplay of the interaction inside the
encapsulated cluster and the interaction between this cluster and
the carbon cage is of essential interest. To compare these two types
of interaction, we considered the bond order for various pairs of
MC atoms: (1) the nearest neighbors in the M2C2 clusters; (2)
the MC pairs connecting the cluster with the fullerene wall. In
contrast to the quite simple structure of the M@C60 EFs, where
the metal atom was located above the center of a hexagon, in the
M2C2@C80 complexes, the positions of the metal atoms near the
C6 and C5 rings of the fullerene were more complicated. In M2C2@-
Ti2C2@C80(Ih) Fe2C2@C80(Ih)
Ea Eform Ea Eform SRM(H)/SRM(I)
633.64 6.74 627.22 2.49 FM
6.18/6.52
632.91 5.90 627.43 2.29 AFM
1.43/1.84
633.50 6.85 627.31 2.64 AFM
0.06/0.11
382 M.V. Ryzhkov et al. / Polyhedron 134 (2017) 376384
C80(Ih) EFs with the most symmetrical Sc2C2 and Fe2C2 rhombuses
(D2h), both metal atoms were located above the center of C5 rings
with the corresponding MC distances near to 2.28 (Sc) and 2.37
(Fe). In the complex with the Ti2C2 (D2h) cluster, both titanium
atoms were located above the C6C5 border with TiC distances
in the 2.212.65 range. In all EFs containing the M2C2 trapezium
(Cs), both metal atoms were located close to hexagons, but not
above their centers. As a result, the number of carbon neighbors
were 6 for Ti and Fe and 7 for Sc with the MC distances in the
intervals 2.112.58 (Ti), 2.102.43 (Fe) and 2.182.59 (Sc). In
all the EFs with the third M2C2 (C2v) isomer, the optimized posi-
tions of both metal atoms were shifted from those above hexagon
centers, and the number of carbon neighbors of each metal atom
varied from 5 in the iron complex to 7 and 8 in the systems with
titanium and scandium atoms, respectively. The corresponding
MC bond lengths were in the intervals 2.242.52 (Sc), 2.172.60
(Ti) and 2.222.49 (Fe). In the M2C2@C80(C2v) EFs, the typical
positions of the metal atoms were near the hexagons and above
the C6C6 and C6C5 borders; the corresponding MC bond lengths
were in the intervals 2.212. 60 (Sc), 2.182.63 (Ti) and 2.28
2.59 (Fe).
Though the coordination of the metal atoms and the MC dis-
tances were noticeably different for the various types of
M2C2@C80 complexes, the structure of the MC bonds in these
objects had some common features. To illustrate the bond order
for various pairs of MC atoms, we present the results obtained
for the three Ih EFs containing M2C2 (C2v) clusters with the maxi-
mum values of |Eform| (Table 5). The roles of various atomic orbitals
(AOs) in the inter-atomic bonding could be compared using the
values of the overlap populations (nij) for the corresponding pairs
of i and j AOs of the neighboring atoms [41]. Note, the values of
nij are not directly connected with the atomization or formation
energies; however, the overlap populations can be useful for eval-
uation of the transformation of bond order of the various pairs of
atomic states. For simplicity, the carbon atoms of the C80 cage con-
nected with the metal atoms M1 and M2 were divided to the three
groups: 1) C11 and C21 groups, for which the M1C and M2C dis-
tances were between 2.1 and 2.3 ; 2) C12 and C22 groups, where
M1C and M2C bond lengths were in the range 2.32.4 ; 3) the
most distant atoms were ascribed to C13 and C23 groups, for which
the M1C and M2C distances were longer than 2.4 . For each
group, the average values of overlap populations are listed in
Table 7.
Analysis of the nij values for all the complexes showed that the
Sc3d and Ti3d orbitals played an overwhelming role in the ScC
and TiC bonding, while the contributions from Sc, Ti 4s and 4p
AOs were negligibly small. On the other hand, the overlap popula-
tions for the Fe4s, 4pC2p states are comparable with those for the
Fe3dC2p interaction. As seen in Table 7, the ScC bonding inside
the Sc2C2 cluster is slightly stronger than the interaction of both
metal atoms with the neighbors of the cage belonging to the C12
and C21 groups, for which the ScC distances are close to those
inside the cluster. The contributions from the Sc3d C2p orbitals
for the more distant carbon atoms of other groups are considerably
Table 6
Atomization and formation energies and the sum of the spin moments on the iron atoms for the M2C2@C80(C2v) EFs with the three most stable isomers of M2C2 clusters.
Cluster Sc2C2@C80(C2v)
Ea Eform
631.54 5.81
631.93 6.33
632.29 7.23
632.30 7.24
less. In contrast to the scandium system, the nij values for the Ti3d
C2p AOs of atoms inside the cluster were more than two times lar-
ger than those for the interactions between metal atoms and the
fullerene cage (Table 7). In the case of the iron complex, the FeC
bond lengths inside the Fe2C2 cluster (1.95 ) are shorter than
those in the titanium system (2.06 ): however, the FeC interac-
tions inside this cluster are considerably weaker than the corre-
sponding TiC bonding. As seen in Table 7, the main
contributions from the Fe3d states to bonding inside the Fe2C2
cluster are similar to those in the scandium complex, though the
corresponding ScC distances (2.26 ) are considerably longer.
It is worth mentioning that the nij values for the Fe4s, 4pC2p
states are relatively small for the short FeC bond lengths inside
the Fe2C2 cluster. However, these values become of the same order
of magnitude or even larger than nij for the Fe3dC2p AOs corre-
sponding to the interactions between the cluster and the fullerene
cage, where the inter-atomic distances are longer than 2.2 . Thus,
we can conclude that only in the Ti2C2@C80 complexes, is the
metalcarbon bonding inside the Ti2C2 cluster considerably stron-
ger than that between the cluster and the fullerene cage. In the
case of the iron systems, the nature of FeC interactions are more
complicated; however, one can conclude that the Fe2C2C80 inter-
action is still weaker than that inside the encapsulated cluster. In
the scandium complexes, both types of bonding are comparable,
and as a result, maximum values of formation energy are achieved.
It is important to note that the overlap populations obtained for
the most stable Sc2C2(butterfly)@C80(C2v) complex were similar
to those of the Sc2C2(butterfly)@C80(Ih) system shown in Table 7.
As shown in Tables 5 and 6, non-zero magnetic moments were
obtained only for the iron complexes, whereas the electronic struc-
tures of all the scandium and titanium systems converged to the
spin-restricted configuration. Note, the two isolated Sc2C2 isomers
(D2h and Cs), which had total spin moments of 2 lB (Table 1), trans-
formed to non-magnetic clusters inside the fullerene cage. It is
important to mention that although all the iron complexes con-
verged to the spin-polarized state, only in the cases of the Fe2C2(-
D2h)@C80(Ih) and Fe2C2(C2v)@C80(C2v) complexes, were the
directions of both metal atom spin moments the same; in other
words, ferromagnetic (FM) ordering was obtained, as in all isolated
Fe2C2 particles (Table 1). When the two other types of Fe2C2 clus-
ters were encapsulated in the C80(Ih) and C80(C2v) cages, the orien-
tation of the spin moment of one Fe atom became the opposite to
that of the other metal atom (AFM ordering). As a result, the sum of
the magnetic moments decreased considerably, especially in the
case of the butterfly@C80(Ih) and trapezium@C80(C2v) complexes,
where the compensation of the spin moments of the metal atoms
was nearly complete (Tables 5 and 6).
To evaluate the role of the magnetic state of the iron atoms, we
performed the spin-restricted calculations for the more stable Fe2-
C2@C80 (Ih) complexes. The results of these calculations showed
that only the Cs cluster retained its trapezium shape, whereas
the D2h rhombuses with a FM ordering of the magnetic moments
of the Fe atoms transformed to a non-magnetic zig-zag chain,
as shown in Fig. 3, and the third butterfly C2v isomer converged
Ti2C2@C80(C2v) Fe2C2@C80(C2v)
Ea Eform Ea Eform SRM(H)/SRM(I)
633.22 5.93 627.70 2.58 AFM
2.55/2.57
632.48 5.08 627.37 1.84 AFM
0.38/0.44
632.57 5.53 626.62 1.56 FM
6.00/6.05
 M.V. Ryzhkov et al. / Polyhedron 134 (2017) 376384 383

Table 7
Overlap populations (10 3 e per one pair of interacting atoms) of the C2p and valence orbitals of metal atoms and the corresponding MC distances () (in parenthesis) obtained
for the M2C2 (C2v) clusters inside C80 fullerene.

Orbitals C atoms of M2C2 M1 neighbors of C80 cage M2 neighbors of C80 cage

C1 C2 C11 C12 C13 C21 C22 C23

Sc1 3d 119 (2.27) 120 (2.26) 102 65 64

Sc2 3d 125 (2.25) 116 (2.28) 99 66 64
Ti1 3d 245 (2.05) 232 (2.07) 96 64 55
Ti2 3d 244 (2.05) 235 (2.07) 94 65 55
Fe1 3d 127 134 35 23
4s 13 (1.96) 14 (1.94) 14 21
4p 13 9 41 27
Fe2 3d 127 131 32 13
4s 16 (1.95) 13 (1.95) 25 7
4p 11 10 38 17

to the same planar trapezium as that of the Cs cluster. As expected,
the rhombuses ? zig-zag chain transformation of the Fe2C2
(D2h) isomer was accompanied by a noticeable variation of all
the FeC bond lengths: 1) the four metalcarbon distances in the
rhombuses (1.97 ) decreased to 1.83 for each FeC pair in the
chain; 2) the average FeC bond length between the metal atoms
and the nearest neighbors of the C5 rings in the fullerene cage also
decreased from 2.37 to 2.03 . Although this structural deforma-
tion was quite considerable, the corresponding variation of
atomization energy was quite small: the value of |Ea| increased
from 627.22 eV (Table 5) to 627.35 eV with a corresponding
increase of |Eform| for the non-magnetic state. On the other hand,
the deformation of the Fe2C2@C80 complex with the trapezium-
shape (Cs) cluster was noticeably less: the structure of the Fe2C2
cluster was nearly the same, and the average distances between
the two metal atoms and the carbon atoms of the nearest hexagons
in the fullerene wall decreased from 2.13 and 2.37 to 2.09 and
2.21 . However, in this case, the values of |Ea| and |Eform|
decreased by 0.26 eV on going from the AFM to the non-magnetic
state. Thus, the non-magnetic state of the Fe2C2@C80 complex is
more stable than that with a FM ordering of the Fe atoms spin
moments and is less stable than the system with AFM ordering.
It is worth mentioning that the considerable increase of overlap
populations induced by the decrease of the FeC bond lengths
did not change the relation between the interactions inside and
outside the Fe2C2 cluster. In the non-magnetic state, the FeC
bonding inside the cluster was also stronger than that between
the Fe2C2 cluster and the cage.
Analysis of the effective charges on the atoms of the M2C2 clus-
ters encapsulated in the C80 fullerene, presented in Table 8, showed
that the differences in the values obtained using the DMol-Hirsh-
feld and DV-Integral procedures were even greater than in the pre-
vious cases. For the M2C2@C80 systems, the latter values are
substantially larger than QM2C2 obtained using the Hirshfeld
scheme. It is also worth mentioning that the trends of charge trans-
fer variation computed using both methods do not correlate, nei-
ther with the variation of Eform nor with the variation of Ea
obtained for each metal (Tables 5, 6 and 8). However, the decrease
of QM2C2 on going from scandium to the titanium and iron systems,
obtained using both procedures, was in agreement with the varia-
tion of the average formation energy in this row (Tables 5, 6 and 8).
Note, in almost all the C80(Ih) complexes, the charges on the M2C2
clusters were larger than in the C80(C2v) systems. The results of the
spin-restricted calculations showed that the transformation of the
Fe2C2 (D2h) isomer from rhombuses to zig-zag chain accompa-
nied by the increase of |Eform| also induced the increase of the DV
charge on cluster from 2.05 to 2.60. On the other hand, the charge
transfer from the trapezium-shape cluster decreased from 2.13
to 2.05 in the non-magnetic state. The results of the calculations
using the Hirshfeld procedure were quite surprising: in all the
non-magnetic complexes, negative charges on Fe2C2 clusters were
obtained.

5. Conclusions

Our ab initio study of four types of EFs containing Sc, Ti and Fe

Fig. 3. Final geometrical structure of the Fe2C2@C80 complex obtained in a spin-
restricted calculation of the initial system with a rhombuses-type (D2h) Fe2C2
cluster inside the C80 cage.
atoms, as well as their metalcarbon clusters, showed that the
properties of these 3d-metals strongly depend on their coordina-
tion in the nano-sized species. In the case of small M2C2 particles,
the stabilities of the iron and scandium clusters are similar,
whereas the corresponding titanium species are energetically more
favorable. On the other hand, in M@C28 and M@C40 EFs, where the
MC bond lengths are practically not variable and the number of
carbon neighbors is considerably higher than in the other known
molecules and solids, the stabilizing effect of Sc encapsulation is
noticeably stronger than that of titanium, and is considerably
stronger than that of iron. However, when the number of nearest
carbon atoms in the M@C60 EFs decreases, the absolute value of
the formation energy for the titanium complex again becomes
higher than that for the scandium system. In the most complicated
M2C2@C80 EFs, the values of |Eform| for the scandium complexes
become higher than those for the titanium systems due to the
interplay between MC interactions inside and outside the M2C2
cluster. This result, as well as the considerably smaller energies
384 M.V. Ryzhkov et al. / Polyhedron 134 (2017) 376384
Table 8
Total effective charges of the M2C2 clusters (M = Sc, Ti, Fe) encapsulated in C80(Ih)/C80(C2v) fullerenes obtained using Hirshfeld (H) and integral (I) procedures.
Cluster Sc2C2
Q (H) Q (I)
0.42/0.16 2.87/2.61
0.43/0.34 3.07/2.97
0.40/0.34 2.99/2.74
for EFs containing iron atoms, can be responsible for the different
probabilities of EF formation with various metals. According to
our study, the stabilizing or destabilizing role of the magnetic
states of the metal atoms strongly depends on the number of car-
bon atoms in the fullerene shell which are the nearest neighbors to
the encapsulated object. In the investigated cases of large coordi-
nation numbers and stable geometries of M@C28 and M@C40 com-
plexes, the obtained values of |Eform| considerably decrease for the
non-magnetic state of these EFs. On the other hand, in the Fe@C60
complex, the stability of the non-magnetic state increases due to
interaction with six carbon atoms and the FeC bond length trans-
formation. In our modelling of Fe2C2@C80 EFs, the non-magnetic
state is more stable than that with FM ordering and is less stable
than the corresponding complex with AFM ordering of the iron
atoms spin moments.
As mentioned above, the vibrational frequencies were com-
puted for some of the investigated objects and could be used to
estimate the influence of the vibrations and standard statistical
thermodynamics on the stability of the EFs. The calculated enthal-
pies for the M@C28 particles showed that the enthalpy additions to
the zero-point energy (Table 2) at T = 298.15 K were close to 112
(Sc@C28), 109 (Ti@C28) and 108 kcal/mol (Fe@C28); i.e. the variation
of these values (1 eV = 23.06 kcal/mol) is considerably less than the
formation energy difference on going from the Sc to Fe systems. A
similar analysis for the M@C60 complexes showed that the varia-
tion of DH was equal or less than 0.02 eV (0.4 kcal/mol), which
was considerably less than the difference in Ea and Eform on going
from the Sc to Fe systems (Table 4). Thus, the impact of vibrations
on the stabilizing role of the metal atoms in EFs is not important.
Similar conclusions were made for the An@C28 clusters [22] and
for the binary, ternary and quaternary metalcarbon particles [42].
Acknowledgment
This work was supported by the Russian Foundation for Basic
Research, grant 14-03-00324.
References
[1] H.W. Kroto, J.R. Heath, S.C. OBrien, R.F. Curl, R.E. Smalley, Nature 318 (1985)
162.
[2] M.S. Dresselhaus, G. Dresselhaus, P. Eklund, Science of Fullerenes and Carbon
Nanotubes, Academic Press, San-Diego, Ca, 1996.
Ti2C2 Fe2C2
Q (H) Q (I) Q (H) Q (I)
0.30/0.04 2.55/2.51 0.14/0.13 2.05/2.04
0.23/0.16 2.85/2.80 0.00/0.09 2.13/2.10
0.21/0.13 2.40/2.54 0.15/013 1.86/1.67
[3] M. Takada, B. Umeda, E. Nishibori, M. Sakata, Y. Saito, M. Ohno, H. Shinohara,
Nature 377 (1995) 46.
[4] R. Jaffiol, A. Debarre, C. Julien, D. Nutarelli, P. Tchenio, A. Taninaka, B. Cao, T.
Okazaki, H. Shinohara, Phys. Rev. B 68 (2003) 014105.
[5] Y. Lian, Z. Shi, X. Zhou, X. He, Z. Gu, Carbon 38 (2000) 2117.
[6] L. Dunsch, S. Yang, Phys. Chem. Chem. Phys. 9 (2007) 3067.
[7] X. Lu, L. Feng, T.J. Akasaka, S. Nagase, Chem. Soc. Rev. 41 (2012) 7723.
[8] X. Lu, T.J. Akasaka, S. Nagase, Acc. Chem. Res. 46 (2013) 1627.
[9] P. Jin, C. Tang, Z. Chen, Coord. Chem. Rev. 270271 (2014) 89.
[10] S. Stevenson, M.A. Mackey, M.C. Thompson, H.L. Coumbe, P.K. Madasu, C.E.
Coumbe, J.P. Phillips, Chem. Commun. (2007) 4263.
[11] T. Guo, M.D. Diener, Y. Chai, M.J. Alford, R.E. Haufler, S.M. McClure, T. Ohno, J.H.
Weaver, G.E. Scuseria, R.E. Smalley, Science 257 (1992) 1661.
[12] H. Funasaka, K. Sugiyama, K. Yamamoto, T. Takahashi, J. Appl. Phys. 78 (1995)
5320.
[13] P.W. Dunk, N.K. Kaiser, M. Mulet-Gas, A. Rodriguez-Fortea, J.M. Poblet, H.
Shonohara, C.L. Hendrickson, A.G. Marshall, H. Kroto, J. Am. Chem. Soc. 134
(2012) 9380.
[14] K. Jackson, E. Kaxiras, M.R. Pederson, J. Phys. Chem. 98 (1994) 7805.
[15] K. Zhao, R.M. Pitzer, J. Phys. Chem. 100 (1996) 4798.
[16] J.-P. Dognon, C. Clavaguera, P. Pyykko, J. Am. Chem. Soc. 131 (2009) 238.
[17] M.V. Ryzhkov, A.L. Ivanovskii, B. Delley, Comp. Theor. Chem. 985 (2012) 46.
[18] M. Cui, H. Zhang, M. Ge, J. Feng, W. Tian, C. Sun, Chem. Phys. Lett. 309 (1999)
344.
[19] X. Lu, Z. Chen, Chem. Rev. 105 (2005) 3643.
[20] E. Albertazzi, C. Domene, P.W. Fowler, T. Heine, G. Seifert, C. Van Alsenoy, F.
Zerbetto, Phys. Chem. Chem. Phys. 1 (1999) 2913.
[21] A.A. El-Barbary, H.I. Lebda, M.A. Kamel, Comput. Mater. Sci. 46 (2009) 128.
[22] M.V. Ryzhkov, B. Delley, Comp. Theor. Chem. 1013 (2013) 70.
[23] M. Diener, C.A. Smith, D.K. Veirs, Chem. Mat. 9 (1997) 1773.
[24] M.V. Ryzhkov, A.L. Ivanovskii, B. Delley, Nanosyst.: Phys., Chem., Math. 5
(2014) 1.
[25] S. Aoyagi, E. Nishibori, H. Sawa, K. Sugimoto, M. Takata, Y. Miyata, R. Kitaura, H.
Shinohara, H. Okada, T. Sakai, Y. Ono, K. Kawachi, K. Yokoo, S. Ono, K. Omote, Y.
Kasama, S. Ishikawa, T. Komuro, H. Tobita, Nat. Chem. 2 (2010) 678.
[26] A.N. Enyashin, Yu.N. Makurin, A.L. Ivanovskii, Comp. Mater. Sci. 36 (2006) 26.
[27] M.V. Ryzhkov, A.L. Ivanovskii, B. Delley, Chem. Phys. Lett. 404 (2005) 400.
[28] B. Delley, J. Chem. Phys. 92 (1990) 508.
[29] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.
[30] F.L. Hirshfeld, Theor. Chim. Acta 44 (1977) 129.
[31] R.F.W. Bader, Atoms in Molecules, Clarendon Press, Oxford, 1990.
[32] E.J. Baerends, D.E. Ellis, Chem. Phys. 2 (1973) 41.
[33] M.R. Press, D.E. Ellis, Phys. Rev. B 35 (1987) 4438.
[34] M.V. Ryzhkov, J. Struct. Chem. 39 (1998) 933.
[35] M.V. Ryzhkov, B. Delley, Theor. Chem. Acc. 131 (2012) 1144.
[36] M.V. Ryzhkov, A.Ya. Kupryazhkin, J. Nucl. Mater. 384 (2009) 226.
[37] H. Kurihara, X. Lu, Y. Liduka, N. Mizorogi, Z. Slamina, T. Tsuchiya, T. Akasaka, S.
Nagase, J. Am. Chem. Soc. 133 (2011) 2382.
[38] H. Kurihara, X. Lu, Y. Liduka, H. Nikawa, M. Hachiya, N. Mizorogi, Z. Slamina, T.
Tsuchiya, S. Nagase, T. Akasaka, Inorg. Chem. 51 (2012) 746.
[39] A.N. Enyashin, A.L. Ivanovskii, Russ. J. Inorg. Chem. 51 (2006) 1483.
[40] K. Nakao, N. Kurita, M. Fujita, Phys. Rev. B 49 (1994) 11415.
[41] R.S. Mulliken, Annu. Rev. Phys. Chem. 29 (1978) 1.
[42] M.V. Ryzhkov, N.I. Medvedeva, B. Delley, Polyhedron 109 (2016) 182.