Physics Letters A 342 (2005) 459–467

www.elsevier.com/locate/pla

First-principles study of the stability and Jahn–Teller distortion

of nickel clusters
Zun Xie a,b , Qing-Min Ma b,∗ , Ying Liu b,c,∗ , You-Cheng Li b,a
a School of Material Science and Engineering, Hebei University of Technology, 300130 Tianjin, China
b Department of Physics, Hebei Normal University, 050016 Shijiazhuang, Hebei, China
c National Key Laboratory for Materials Simulation and Design, 100080 Beijing, China

Received 18 March 2005; received in revised form 26 May 2005; accepted 26 May 2005
Available online 2 June 2005
Communicated by R. Wu

Abstract
The density functional method for molecules has been employed to study the structural and magnetic properties of NiN
(N = 2–13) clusters. The calculated results show that the Jahn–Teller effect plays an important role in determining the ground
state of certain geometric structures. Ground states have been assigned for these clusters. For Ni5 and Ni7 , new ground states
have been found. The second derivative of the binding energy with cluster size shows that besides N = 13, N = 10 is also a
magic number. Most importantly, it is found that from N = 6 to 13, the spin moment is 8µB in all cases, which verifies the
quantum confinement of 4s electrons in small nickel clusters.
 2005 Elsevier B.V. All rights reserved.

PACS: 36.40.Mr; 36.40.Cg; 61.46.+w

Keywords: First-principles calculations; Density functional for molecules; Jahn–Teller distortion

1. Introduction technological fields. Since the properties evolve with

size and composition, suitably designed clusters with
In the past three decades, transition-metal clusters tailored properties are promising in the synthesis of
(TMCs), due to their unique geometrical structures, nanostructures. Among the 30 kinds of pure TMCs,
physical, chemical, electronic and magnetic proper- nickel clusters are the primary target of many research
ties, have attracted much attention in scientific and groups [1–9] because of their extensive catalytic and
important magnetic properties. In comparison with the
alkali metal clusters, the localized behavior of unfilled
* Corresponding authors. 3d electrons of nickel clusters results in enormous
E-mail address: yliu@mail.hebtu.edu.cn (Y. Liu). complexity and dominates most of properties. Nickel
0375-9601/$ – see front matter  2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.physleta.2005.05.067
460 Z. Xie et al. / Physics Letters A 342 (2005) 459–467

clusters constitute one of the most studied sets of clus- density approximation (LDA) based on computational
ters and, therefore, are an excellent system for explor- schemes [9–11] such as the linearized-augmented-
ing new theoretical or experimental approaches. plane-wave, the linearized-muffin-tin-orbital etc. Be-
Structural stability, electronic structure, and mag- cause the size of the clusters and the complexity of
netic properties are three important linked problems the nickel atom do not allow a configuration inter-
in cluster research, since the magnetic moment de- action approach [6,7], the existing ab initio results
pends on the electronic structure, which in turn de- are still somewhat inconsistent with respect to each
pends on the geometry. Knowledge concerning these other, and the true ground states are uncertain. Reuse
questions can give us additional insight into the phys- et al. [12] carried out the first systematic study of the
ical and chemical properties. Unfortunately, it is very geometries of NiN (N = 2–6, 8, 13) clusters using an
difficult to determine the geometry of nickel clusters approach based on a linear combination of atomic
consisting of less than 100 atoms, even though it is in molecular-orbitals within the density functional for-
this size range that nickel clusters exhibit the most in- malism. They obtained calculated geometries, and the
teresting electronic, magnetic, and optical properties. total spin magnetic moments for these nickel clusters.
Experimental determination of the cluster geometry is It is well known that there are abundant low-lying
difficult, since most clusters are too large for spec- structural isomers whose energies are very close to
troscopic probes but too small for diffraction probes. the ground state energy and that the Jahn–Teller ef-
Theory is usually employed to determine the cluster fect [13] is important. Subsequently, Reuse et al. pub-
geometry, but there are often several low-lying struc- lished studies for Ni13 [14], Ni7 , and Ni8 [1] clus-
tural isomers whose energies are very close to that of ters including the Jahn–Teller effect. A more complete
the ground state [1,2]. Nevertheless attempts are being study of the smallest NiN clusters is, however, still cru-
made, experimentally and theoretically, to obtain the cial for a comprehensive understanding of the property
geometries of nickel clusters. of these clusters.
On the experimental side, chemical-probe exper- In general, the density functional method with its
iments have given very valuable, although indirect, useful local density approximation and its relatively
information concerning the structure of these clus- simple computational scheme provides an alternate
ters. Parks and coworkers [3] have used the reactions theoretical tool for studying transition-metal complex-
of nickel clusters with ammonia and with water to ity and the results obtained so far are comparable or
predict the equilibrium geometric structure. Parks et better than those obtained at the HF level. The first-
al. [4] have probed the equilibrium geometric struc- principles cluster method referred to as the density
ture of small nickel clusters via molecular adsorption functional for molecules (DMol) [21] is an accurate
of nitrogen onto their surfaces. However, it is ar- and efficient local density functional calculation for
gued that the current experimental studies have not molecules and has been applied to many problems
confirmed the coexistence of isomers [5]. On the the- in molecular clusters, such as chemisorption, surface
oretical side, the research techniques include mole- reconstruction, and the ground state of highly corre-
cular dynamics (MD) simulations, the embedded- lated TMCs. Recently, Nayak et al. [5] have studied
atom method (EAM), first-principles calculation, and the structure of Ni7 using the DMol method together
effective-medium-theory (EMT). The MD simulations with MD simulations. They have used the Hedin–
to date are either based on the tight-binding frame- Lundqvist/Janak–Moruzzi–Williams local correlation
work or based on ab initio methods, which provide functions and double numerical basis sets with 4p
a good compromise between accuracy and computa- polarization functions in their calculations. More re-
tional speed. The EMT permits a fair estimation of cently, using the DMol cluster method, Zhang et al.
binding energy. In general, for the smallest clusters, [15,16] have studied the structural stability of LaN
the best credible method is still first-principles cal- clusters and Wang et al. [17,18] have successfully
culations. Most of the ab initio calculations on NiN studied the electronic structures of icosahedral TiN
clusters have been performed within the framework clusters.
of the self-consistent field at the nonlocal Hartree– In order to clarify the stable structures for NiN
Fock (HF) level of approximation [6–8], or the local (N = 2–13) clusters, we have performed a global min-
 Z. Xie et al. / Physics Letters A 342 (2005) 459–467 461

imization of the total energy using the first-principles Table 1

cluster method. We have found that high symmetry Calculated atomic bond lengths a0 (in Å) and binding energies Eb
clusters have heavy degeneracy and that the Jahn– (in eV) of the Ni2 dimer in the ground state
Teller effect is prominent. Above all, for clusters of 5 PWC BP BLYP PBE Expt.
and 7 atoms, we have found new ground states. To our a0 2.064 2.096 2.112 2.096 2.1545 ± 0.0004
knowledge, the existing first-principles calculations Eb −6.62893 −5.99681 −5.63756 −5.97857 −2.042 ± 0.002
of NiN (N = 9–12) clusters are not sufficient, and
there are only results based on tight-binding molecu- (GGA) corrections. The popular B88 functional [23]
lar dynamics (TBMD) simulations [19] and the EAM is often used in conjunction with the PW91 functional
study [20]. In this Letter, we present our research re- or the LYP [24] correlation functional. The PBE [25]
sults for the smallest NiN (2
N
13) clusters. functional is another GGA functional in which all
the parameters are fundamental constants. The results
of test calculations are summarized in Table 1. The
2. Computational details binding energy of the cluster system is defined as the
difference in the total energy between the interacting
The DMol cluster method can yield accurate and atom system and the free-atom system.
efficient self-consistent calculations using a rapidly From Table 1, it is obvious that the results of GGA
convergent three-dimensional numerical integration are much closer to the experimental data than those
scheme [21]. As is known, the basis sets and the of LDA. The results of BLYP are closest to the ex-
exchange-correlation function used in the localized- perimental data in binding energy and bond length.
orbital method can affect the calculated results to some It should also be noted that the BLYP functional has
extent. The inner orbitals such as 3s and 3p sometimes been applied successfully to several other systems, in-
are important for the ground-state properties of nickel cluding carbon nanotubes, TiN clusters [17,18], LaN
clusters. We chose a double-numerical basis with po- clusters [15,16]. In consideration of above calculated
larized functions (DNP), which froze only the 1s, 2s, results and these applications, we adopted the BLYP
2p orbitals in the present study. The size of the DNP exchange-correlation functional in the present study.
basis sets is comparable to that of the Gaussian 6– Both in experiment and in theory, the determina-
31 G** sets, but the numerical basis set is much more tion of the ground state of the Ni2 dimer has been the
accurate than a Gaussian basis set of the same size. subject of much controversy. The latest experiments
Experimental evidence on the high reactivity of nickel [26] on Ni2 clusters, using resonant two-photon ion-
clusters suggests that their ground states have open ization spectroscopy, show a ground-state bond length
shells. In the density functional approach, this means of 2.1545 ± 0.0004 Å for 58 Ni2 , and a revised value for
that spin polarized calculations are required to find the the binding energy of −2.042 ± 0.002 eV. We find the
correct electronic configuration of the ground state. ground-state binding energy to be −5.63756 eV, which
First of all, using several kinds of exchange- is much lower than the experimental data, and a bond
correlation functionals, we have performed all-electron length of 2.112 Å, which is close to the experimen-
test calculations on the Ni2 dimer. The functionals tal data. It is well known that the density functional
used in the test include the following: the Perdew– method sometimes estimates too low for the binding
Wang (1992) (PWC) functional [22]; the Perdew– energy but gives good bond distances. Our calcula-
Wang (1991) (PW91) with the Becke (B88) exchange tions show the total spin to be S = 1, which is consis-
functional (denoted as BP together); the Lee–Yang– tent with earlier matrix isolation measurements [27]
Parr (LYP) correlation functional with B88 (denoted and many theoretical predictions [12]. However, the
as BLYP together); and the Perdew, Burke, Enzerhof most recent jet-cooled study of the Ni2 dimer indicated
(PBE) functional. The first of these functionals, be- that the ground state was a mixture of triplet and sin-
longs to the LDA, and is a recent parameterization glet states. These differences are extremely small but
of the Ceperley and Alder data, which corrects some still need further work.
Vosko–Wilk–Nusair functional problems through fit- For accurate calculations, we chose an octupole
ting. The others are all general gradient approximation scheme for the multipolar expansion of the charge
462 Z. Xie et al. / Physics Letters A 342 (2005) 459–467

density and Coulomb potential. In the optimizations, of the cluster belongs to a point group with a single
the energy gradient and atomic displacement converge main axis, this operation does not change the symme-
to 1 × 10−4 Hartree/Bohr and 1 × 10−4 Å, respec- try. It is important that all of these structural distortions
tively. The charge density in the self-consistent itera- are small, just enough to lift the energy level degen-
3
tions converges to 1 × 10−6 e/Å , which corresponds eracy. It should be noted, however, that for a cubic
to a total energy convergence of 1 × 10−5 Hartree. point group with several main axes, the above oper-
As the number of possible configurations increases ation would lower the symmetry of the cluster.
very rapidly with the cluster size, it is obvious that one
cannot explore all possible configurations. In general,
clusters with higher symmetry and fewer low coordi-
3. Results and discussions
nation surface atoms are expected to be more stable.
The initial geometry was therefore constructed accord-
ing to a certain symmetry. We then examined almost The calculated results are summarized in Table 2
all candidate configurations for up to 13 atoms appear- and some sketches of the geometric structures are
ing in the literature. In order to determine the ground shown in Fig. 1. As mentioned by Alonso [28], for
state of each cluster, by relaxation, starting from those the TMCs, the equilibrium geometry is determined by
configurations, we have employed Jahn–Teller the- competition between two factors. One is that a com-
ory [13], which states that an electronic system occu- pact structure has the maximum number of bonds. The
pying an energy level with degeneracy is unstable and other factor is that the directional bonding is com-
a structural distortion will occur to remove that de- patible with the orientation and the filling of the d
generacy, especially in those systems whose highest- orbital. The two factors both enhance the binding en-
occupied molecular orbital (HOMO) is not fully occu- ergy. The first factor above means that, normally, the
pied. For almost every cluster size, we performed var- larger the number of bonds, the more stable the clus-
ious distortion involving moving one or several atoms ter. In Table 2, besides a few special clusters, we give
along the main axis of the cluster. If the configuration only those structures with a high number of bonds. The

Table 2
Calculated results for different structures of NiN (N = 2–13) clusters. The table lists the symmetry, binding energy Eb (in eV), the HOMO
state, the energy difference
E between the LUMO and HOMO, the electron occupation number n in the HOMO, and the total spin moment
µs (in µB )
N Sym Eb HOMO
E n µs N Sym Eb HOMO
E n µs
3 D3h −9.16767 E 0.368 1 2 C2v −35.38311 A1 0.161 1 8
C2v −9.27671 A1 0.131 1 2 C4v −35.84272 E 0.371 2 8
4 Td −13.39243 T2 0.458 1 4 C2 −35.85273 B 0.255 1 8
C3v −13.37616 E 0.019 1 4 10 Td −40.49849 T1 0.507 3 8
Cs −13.46724 A 0.228 1 4 D2 (A) −40.49849 B3 0.506 1 8
D2d −13.47127 B2 0.222 1 4 C3v −39.97942 A1 0.228 1 8
5 D3h −17.62936 E 0.227 2 4 D4d −39.99117 E1 0.059 2 8
C2v (A) −17.77462 A2 0.302 1 4 D2 (B) −39.99273 B3 0.030 1 8
C4v −17.96244 E 0.663 2 6 11 D4d (A) −43.34163 A2 0.224 1 8
C2v (B) −17.98921 B2 0.685 1 6 D4d (B) −43.26211 B2 0.477 1 8
6 Oh −22.11106 T2U 0.252 2 8 12 C5v −48.40911 E2 0.124 2 8
D4h −22.19598 EU 0.125 2 8 13 Ih −52.05962 HU 0.195 3 8
D2h −22.21052 B1U 0.119 1 8 D5d (i) −52.71926 A2U 0.112 1 8
7 D5h −26.68841 E1 0.170 2 6 D3d (i) −52.71943 A2U 0.124 1 8
C3v −26.87943 A1 0.371 1 8 D2h (i) −52.44182 B1U 0.229 1 8
C2 −26.63191 A 0.654 1 8 D5h (d) −52.27069 A2 0.134 1 8
8 D2d −31.38768 E 0.487 2 8 Oh (c) −51.66657 T1G 0.129 2 8
C2v −31.27790 B2 0.245 1 8 D4h (c) −51.74011 B2U 0.233 1 6
D6h −30.21057 E1U 0.154 1 6 D2h (c) −51.87265 B3U 0.082 1 6
9 D3h −35.82140 E 0.139 1 8 D3d (c) −51.88782 EU 0.16 2 6
 Z. Xie et al. / Physics Letters A 342 (2005) 459–467 463

5: D3h and C2v (A) 5: C4v and C2v (B) 6: Oh , D4h and D2h

7: D5h 7: C3v 7: C2

8: D2d 8: C2v 8: D6h

9: C4v and C2 9: C2v 9: D3h

10: Td and D2 (A) 10: D4d and D2 (B) 10: C3v

11: D4d (A) 11: D4d (B) 12: C5v

13: Ih , D5d , D3d , and D2h 13: D5h 13: Oh , D4h , D2h , and D3d

Fig. 1. Some structural sketches and their symmetries of NiN (N = 5–13) clusters.

second factor above promotes the occurrence of the We have found that the structure with C2v symme-
Jahn–Teller effect and results in the known complex- try (isosceles triangular) and average bond length of
ity of TMCs. 2.2798 Å has the lowest energy. The case of Ni3 is
For the Ni3 cluster we have tried two structures of special significance because of the existing contro-
(C2v and D3h ). The C2v is a distorted form of D3h . versy between theory and experiment. Moskovits and
464 Z. Xie et al. / Physics Letters A 342 (2005) 459–467

Dilella [29] found the geometry of Ni3 in solid argon has fewer bonds. In contrast, the trigonal bipyramid
to be C2v triplet with an apex angle estimated to be C2v (A) configuration optimizes the number of bonds
between 90◦ and 100◦ . Studies using both TBMD sim- but forces d orbitals into a symmetry which is not op-
ulations [19] and EAM [20] for Ni3 indicated that the timal for their bonding (the d orbitals in an atom have
D3h (equilateral triangle) structure was the most sta- square symmetry). Therefore, the C2v (B) with fewer
ble. Reuse et al. [12] suggested a C2v ground state bonds is more stable than C2v (A). The C2v (B) stud-
with triplet state and an apex angle of 61◦ . In good ied in the present work is thus a new ground state of
agreement with Reuse et al., our results indicate that the Ni5 cluster obtained here for the first time by a
the ground state is a C2v structure with S = 1 and first-principles calculation.
an apex angle of 57.198◦ . At the same time, we have The initial configuration of Ni6 is a tetragonal
found that the binding energy of the equilateral trian- bipyramid. Both D4h and D2h are distortional tetrago-
gle configuration is 0.11 eV higher than that of our C2v nal bipyramids derived from the octahedron (Oh sym-
structure. metry). The rectangular octahedron (Oh ) has a three-
For Ni4 , we have investigated the tetrahedron struc- fold HOMO state partially occupied, and is unstable
ture. The C3v , Cs , and D2d structures are all distortions according to Jahn–Teller theory. The distortion from
from the tetrahedron Td form. The Td structure, a reg- Oh to D4h has changed the threefold to twofold degen-
ular tetrahedron, has a triplet HOMO state partially eracy, fully occupied. The further distortion from D4h
filled, and its binding energy is little higher than that to D2h has completely lifted the twofold degeneracy.
of the Cs structure with a singlet HOMO state. The The D2h structure is energetically the lowest isomer,
D2d structure (a rhombus bent into a butterfly struc- with a bond length ranging from 2.372 to 2.400 Å.
ture), another distortion tetrahedron [12] has the low- This is in qualitative agreement with the predictions
est binding energy and S = 2. We conclude that the of Menon et al. [19] and Almlof et al. [30] who con-
D2d and Cs structures are the energetically lowest two sidered Jahn–Teller distortions in octahedral Ni6 .
isomers, and that they have the same spin. In the case of Ni7 , we have considered a capped
Most theoretical calculations of Ni5 [12,20] have octahedron, a pentagonal bipyramid, and a tri-capped
found that the ground state to be a trigonal bipyramid tetrahedron structure. They all have the same num-
(D3h symmetry). In our work, we tried the two struc- ber of bonds. The capped octahedron (C3v ) and the
ture D3h and C4v . The C2v (A) is a distortion from D3h , pentagonal bipyramid (D5h ) have similar binding en-
and C2v (B) from the square pyramid C4v . The D3h ergy. The latter has a twofold HOMO state fully oc-
and C4v structures have both twofold HOMO states cupied and is only 0.0273 eV/atom higher in total-
fully filled. The Jahn–Teller distortion from D3h to C2v energy than the C3v structure. It is well known that
(A) has lifted the twofold degeneracy and lowered the the structure, which has a fully occupied HOMO state,
binding energy. Similarly, from C4v to C2v (B), the is relatively stable, although there is degeneracy in
twofold degeneracy of C4v is also lifted. The lowest the HOMO state. Consequently we suggest that the
energy structure is just the C2v (B) distortion of the capped octahedron (C3v ) and the pentagonal bipyra-
square pyramid structure, the base being a rhombus mid (D5h ) are two nearly degenerate isomers. Some
rather than a square. The binding energy is 0.027 eV theoretical calculations have predicted the ground state
lower than that of the C4v structure. This conclu- to be a pentagonal bipyramid, but other density func-
sion agrees with that of the TBMD study by Menon tional calculations [1,5,31] and EAM results [20] have
et al. [19], who called this structure C2v (B) as a distor- suggested that these two types of structures are nearly
tional Td . In our case, the distortion structure of D3h is degenerate isomeric forms. The corresponding aver-
0.215 eV less favorable than that of C4v . It is interest- age binding energies per atom were found to differ by
ing to note that the number of bonds in C2v (B) is fewer only 0.03–0.07 eV/atom, i.e., very similar to our find-
than in C2v (A). We have further calculated the average ings. Compared with C3v , the tri-capped tetrahedron
bond lengths (2.351 Å for (B) and 2.373 Å for (A)), structure (C2 symmetry) has the same spin and singlet
which shows that C2v (B) is more compact than C2v HOMO state, but its binding energy is 0.25 eV higher
(A). In addition, the tetragonal pyramid of Ni5 permits than that of the C3v structure, which may be regarded
undistorted d-orbital combinations to form though it as a third isomer [20] for Ni7 .
 Z. Xie et al. / Physics Letters A 342 (2005) 459–467 465

For Ni8 , we have selected three configurations and is the second isomers. The C3v structure has no de-
(D2d , D6h and one bi-capped octahedron). The D2d generacy in the HOMO state and is the third isomers
structure, called [20] ‘bisdisphenoid’, has the lowest of Ni10 . One additional comment should be made. The
binding energy and a twofold HOMO state fully filled. second derivative of cluster binding energy with re-
This conclusion is in best agreement with both exper- spect to cluster size for Ni10 is 1.80262 eV, greater
iment [6] and an EAM study by Grigoryan et al. [20]. than zero, which suggests that N = 10 is a so-called
The bi-capped octahedron C2v has a singlet HOMO magic number, and that Ni10 is very stable, like Ni13 .
state and a slightly higher binding energy compared This is the first time that the magic number N = 10
to D2d . The D6h structure is unstable because of the has been found using a first-principles calculation and
higher binding energy and twofold HOMO state par- is consistent with the results of Grigoryan et al. [20].
tially occupied. We have not found more stable dis- For Ni11 , we have tried two D4d structures: D4d
tortion structures for either D2d or D6h . It is worth (A) and D4d (B). The D4d (A) structure has the low-
noting that the number of bonds in the D6h structure est energy and a singlet HOMO state. The D4d (B)
is the highest for all the structures discussed here, but structure is found to be quite similar in energy to that
it is unstable. From the optimized structure of D6h in of the D4d (A). It therefore appears that they are two
Fig. 1, we can see that the coordination of the two nearly degenerate isomeric forms. The Ni12 cluster
apex atoms is up to 7, greater than half the coordi- with C5v symmetry has a fully occupied HOMO state
nation number of a face-center cube. Its 6-atom base with twofold degeneracy. However, it is difficult to dis-
must therefore be expanded, which could weaken the tort C5v into a more stable state whose HOMO state is
binding energy. not degenerate.
For Ni9 , three configurations have been studied For Ni13 , there are three potential modes compet-
(C2v , D3h , and C4v ). The C2 structure is a distortion ing in cluster growth: icosahedral (i), decahedral (d)
of C4v . The lowest energy structure obtained is the C2 , and cuboctahedral (c) isomers [32], and it is therefore
which has a singlet HOMO state. The D3h and C2 have essential to compare the results of structures. The four
similar energy, but the D3h has a twofold HOMO state, structures D5d (i), D3d (i), D2h (i), and Ih , all belong
incompletely occupied. A Jahn–Teller distortion could to an icosahedral structure. Only Ih is a regular icosa-
possibly lift the degeneracy or make the HOMO state hedron; all the others are relaxed icosahedrons. The
occupied fully, but further distortion of D3h is rather structures, excluding Ih , have similar binding energies
difficult. The C2v structure with a similar binding en- and no degeneracy in the HOMO state. The D3d (i) has
ergy is also a low-lying isomer. From Fig. 1, we can the lowest binding energy, as was predicted by Reuse
see that the number of bonds in the D3h configuration et al. [2]. The Ih structure has the highest binding en-
is greater by one than in C4v or C2v but D3h is less sta- ergy and a degenerate HOMO of 5 with only three
ble than C4v and C2v . We hypothesize that one reason electrons. The D5d (i), D2h (i), and D3d (i) structures
for this is that the D3h structure makes the distortion can be considered as degenerate isomers of i-Ni13
of d orbitals have trigonal symmetry, and that the bond while Ih is not stable. In the same way, for d-Ni13 , D5h
emphasized in Fig. 1 is too short being 0.1–0.138 Å (d) is stable. For c-Ni13 , D2h (c), D3d (c) and D4h (c)
shorter than most of the other bonds. are degenerate isomers while Oh is not stable. Further-
For Ni10 , we consider the following structures: more, it is clear that i-Ni13 clusters are of lower energy
tetra-capped octahedron (Td symmetry), C3v and D4d . compared to the other two kinds of isomers, which in-
The D2 (A) is a distortion of Td and D2 (B) of the D4d dicates that Ni13 prefers to be i-isomer. Most of earlier
symmetry structures. Of these geometries, the distor- theoretical studies [33] were based only on an icosahe-
tional D2 (A) structure of Td symmetry is found to dral geometry. Although our studies have been carried
be the most stable, for it has the lowest binding en- out three geometries, nevertheles, our results find the
ergy and a singlet HOMO state. It is necessary to note icosahedral geometry to be the most stable, in agree-
that this distortion has completely lifted the threefold ment with experiment [4].
degeneracy of the HOMO state, though the binding en- From the above analysis, it is clear that there are
ergy does not show much reduction. The D2 (B) has many low-lying states with very close-lying energies.
also lifted the twofold degeneracy of the D4d structure The Jahn–Teller distortions, which lower the structural
466 Z. Xie et al. / Physics Letters A 342 (2005) 459–467

symmetry, help determine the ground states for NiN

clusters for N
13. These distortions allow for the
lowering of the energy through the removal of some
of the degeneracy present in more symmetric geome-
tries. In searching for the ground state, we find that the
binding energy does not exhibit a smooth parabolic
form when the cluster is expanded without changing
the symmetry. This indicates that there are some false
stable states very near the ground states for certain
symmetries, which increases the difficulty of the cal-
culation.
In the following, we discuss the magnetic moments
of NiN clusters. Very recently, Zhou et al. [34,35]
have studied the role of orbital polarization in low-
dimensional transition metal systems, and Wan et al.
[36] have demonstrated the effect of orbital polariza-
tion, surface enhancement, and quantum confinement Fig. 2. Spin moment as a function of cluster size.
in nanocluster magnetism. For the spin moment (µs )
of Ni13 , Wan et al. obtained 0.58 µB /atom within a leads to a quantum terrace in the s electron occupation
rather general tight-binding framework and Reuse et as a function of the cluster size, from N = 9 to 13,
al. [12] obtained 0.62 µB /atom by the linear combi- and the total number of s electrons remains at eight
nation of atomic molecular-orbital approach. All these independent of the cluster size. Our results therefore
results are in good agreement with that of the present support Wan et al.’s findings. Furthermore our results
study; namely, µs = 0.615 µB /atom for Ni13 . show that the quantum terrace from N = 13 to 9 may
It is worth mentioning that our calculated results in fact extend to the 6-atom cluster.
have shown that from N = 6 to 13, the spin moment
remains constant at 8µB (see Fig. 2), that is to say, the
mean spin moments of the clusters decrease monoton- 4. Conclusions
ically with size. Apsel et al. [37] found that the mean
magnetic moment exhibited characteristic oscillations In summary, we have obtained optimized geome-
with respect to the nickel cluster size: in the range tries of a number of low-energy geometrical isomers
of N
13, there was one pronounced minimum at of NiN (N = 2–13) clusters using the DMol method
N = 13 and another less pronounced minimum at based on density functional theory. The Jahn–Teller
N = 6. At first sight our calculated results seems to be effect is shown to play an important role in optimizing
inconsistent with Apsel et al. Wan et al. [36], however, the ground states. Reduced symmetry can easily lift
have already shown that it is the orbital moment that the degeneracy of the HOMO state and there are many
is responsible for this disagreement between the mo- isomers near the ground state. For Ni5 and Ni7 , new
ment value and the oscillation of total moments with ground states are found. As in the case of the 13-atom
respect to the cluster size. Our results are crucial in cluster, Ni10 is also a magic cluster. Above all, we ob-
reaching an understanding of the work of Wan et al. tained that from N = 6 to 13, the spin moments are all
on the spin moment for N = 9–13. Within a simple 8µB , which exactly verifies the quantum confinement
electronic shell model [38], the spin moment (in µB ) of 4s electrons in very smaller nickel clusters.
is equal to the number of unpaired electrons, i.e., the
number of holes, which is equal in turn to the num-
ber of 4s electrons in the nickel atom. Fig. 2 therefore Acknowledgements
also indicates the total number of s electrons as a func-
tion of cluster size (2
N
13). Our results exactly The authors would like to thank D.B. Zhang for
verify the findings of Wan et al., volume confinement his assistance and Dr. N.E. Davison for his help
 Z. Xie et al. / Physics Letters A 342 (2005) 459–467 467

with the language. This work was supported by the [15] D.B. Zhang, J. Shen, J. Chem. Phys. 120 (2004) 5104.
science fund of Hebei Province (A2005000143 and [16] D.B. Zhang, J. Shen, J. Chem. Phys. 120 (2004) 5081.
A2004000141), the Hebei Science and Technology [17] S.Y. Wang, W.H. Duan, D.L. Zhao, et al., Phys. Rev. B 65
(2002) 165424.
Committee Doctoral Foundation (03547008D), the [18] S.Y. Wang, J.Z. Yu, H. Mizuseki, et al., J. Chem. Phys. 120
Hebei Education Committee Doctoral Foundation (2004) 8463.
(B2004115), and the Hebei Normal University Doc- [19] M. Menon, J. Connolly, N. Lathiotakis, et al., Phys. Rev. B 50
toral Foundation (103260). (1994) 8903.
[20] V.G. Grigoryan, M. Springborg, Phys. Rev. B 70 (2004)
205415.
[21] B. Delley, J. Chem. Phys. 92 (1990) 508.
References [22] J.P. Perdew, Y. Wang, Phys. Rev. B 45 (1992) 13244.
[23] A.D. Becke, J. Chem. Phys. 88 (1988) 2547.
[24] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 786.
[1] N. Desmarais, C. Jamorski, F.A. Reuse, et al., Chem. Phys.
[25] T. Tsuneda, T. Suzumura, K. Hirao, J. Chem. Phys. 110 (1999)
Lett. 294 (1998) 480.
10664.
[2] F.A. Reuse, S.N. Khanna, S. Bernel, Phys. Rev. B 52 (1995)
[26] J.C. Pinegar, J.D. Langenberg, C.A. Arrington, et al., J. Chem.
R11650.
Phys. 102 (1995) 666.
[3] E.K. Parks, B.J. Winter, T.D. Klots, et al., J. Chem. Phys. 94
[27] M. Moskovits, J.E. Hulse, J. Chem. Phys. 66 (1977) 3988.
(1991) 1882.
[28] J.A. Alonso, Chem. Rev. 100 (2000) 637.
[4] E.K. Parks, L. Zhu, S.J. Riley, et al., J. Chem. Phys. 100 (1994)
[29] M. Moskovits, D.P. DiLella, J. Chem. Phys. 72 (1980) 2267.
7206.
[30] Z. Yu, J. Almlof, J. Phys. Chem. 95 (1991) 9167.
[5] S.K. Nayak, B. Reddy, B.K. Rao, et al., Chem. Phys. Lett. 253
[31] S.N. Khanna, M. Beltran, P. Jena, Phys. Rev. B 64 (2001)
(1996) 390.
235419.
[6] H. Basch, M.D. Newton, J.W. Moskowitz, J. Chem. Phys. 73
[32] J. Akola, H. Häkkinen, M. Manninen, Phys. Rev. B 58 (1998)
(1980) 4492.
3601.
[7] M. Tomonari, H. Tatewaki, T. Nakamura, J. Chem. Phys. 85
[33] M.S. Stave, A.E. DePristo, J. Chem. Phys. 97 (1992) 3386.
(1986) 2875.
[34] L. Zhou, J.T. Wang, D.S. Wang, et al., Mater. Trans., JIM 40
[8] M.A. Nygren, P.E.M. Siegbahn, U. Wahlgren, et al., J. Phys.
(1999) 1237.
Chem. 96 (1992) 3633.
[35] L. Zhou, D.S. Wang, Y. Kawazoe, Phys. Rev. B 60 (1999)
[9] J.M. Soler, A.R. Williams, Phys. Rev. B 42 (1990) 9728.
9545.
[10] R. Yu, D. Singh, H. Krakauer, Phys. Rev. B 43 (1991) 6411.
[36] X.G. Wan, L. Zhou, J.M. Dong, et al., Phys. Rev. B 69 (2004)
[11] P. Mlynarski, D.R. Salahub, J. Chem. Phys. 95 (1991) 6050.
174414.
[12] F.A. Reuse, S.N. Khanna, Chem. Phys. Lett. 234 (1995) 77.
[37] S.E. Apsel, J.W. Emmert, J. Deng, et al., Phys. Rev. Lett. 76
[13] F.S. Ham, J. Lumin. 85 (2000) 193.
(1996) 1441.
[14] F.A. Reuse, S.N. Khanna, S. Bernel, Phys. Rev. B 52 (1995) [38] N. Fujima, T. Yamaguchi, Phys. Rev. B 54 (1996) 26.
R11650.