Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93

Contents lists available at ScienceDirect

Journal of Molecular Structure: THEOCHEM

journal homepage: www.elsevier.com/locate/theochem

Structural and electronic properties of stable Aun (n = 2–13) clusters:

A density functional study
Ajanta Deka, Ramesh C. Deka *
Department of Chemical Sciences, Tezpur University, Napaam, Tezpur, Assam 784 028, India

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

Article history: All-electron scalar relativistic calculations have been performed to investigate the electronic structures of
Received 12 May 2008 neutral gold clusters (Aun, n = 2–13) in the gas phase using density functional theory with the generalized
Received in revised form 5 September 2008 gradient approximation. Full geometry optimizations of topologically different clusters and clusters
Accepted 8 September 2008
belonging to different symmetry groups have been carried out. Binding energies, ionization potentials,
Available online 22 September 2008
electron affinities, and chemical hardness values are calculated and they are found to be comparable with
the available experimental and theoretical results. The most stable structure of each of the cluster has a
Keywords:
two-dimensional planar configuration. A three dimensional distorted Y shaped structure (4b) for Au4, a
Gold clusters
Density functional theory
tri-capped triangle (6b), a chair (6f), and a see-saw structure (6j) for Au6, an eclipsed sandwich structure
Binding energies (7g) for Au7, a condensed trigonal bipyramid (9e) and a boat shaped structure (9f) for Au9, a staggered
Ionization potentials sandwich (11c) and an eclipsed sandwich structure (11d) for Au11, a ladderane structure (12d) for
Hardness Au12, and a staggered (13d) and a distorted sandwich structure (13e) for Au13 are characterized for the
first time in this work.
Ó 2008 Elsevier B.V. All rights reserved.

1. Introduction rized according to the level of difficulties encountered in their

computational treatment. In this sense, alkali-metal clusters
Recently, nanometer-sized systems and structures have been a proved particularly useful because they could be easily character-
topic of great interest for experimental and theoretical investiga- ized experimentally and investigated with less computational ef-
tions [1–3]. While many researchers in this field use a variety of fort compared to the other metal clusters. Hence, they have been
techniques and materials to create and study such systems, most widely studied both theoretically and experimentally [6]. Compar-
share a common element: some kind of cluster or quantum dot. atively much less is known about transition metal clusters where
A cluster is a group of atoms that come together by self-assembly. electronic structure and chemical properties depend upon the
They can be comprised of anywhere between 2 and 10000 parti- interplay between s and d electrons. The coinage metals with their
cles, and represent a sort of bridge between the atomic and bulk filled d shell and singly occupied s shell (electronic structure
size regimes. nd10(n+1)s1, with n = 3, 4 and 5 for Cu, Ag and Au, respectively)
Much of the interest in clusters is fuelled by the many possible can be regarded as a bridge between the ‘‘simple” s-only alkali-
applications of clusters and the related nanoscale technologies. metals and the more complicated transition metals. As a result,
Nano clusters not only exhibit interesting electronic and optical they have been subjected to many experimental [7–11] and theo-
properties intrinsically associated with their low dimensionality retical investigations [12–16].
and the quantum confinement effect, but also represent the critical In recent years, gold nanoclusters have received considerable
components in the potential nanoscale electronic and photonic de- attention due to their important role as building blocks in nanoscale
vice applications [4,5]. The role of clusters in chemical systems is electronic [17], optical and medical diagnostic devices [18]. Besides,
also intriguing. As particle size decreases, the ratio of surface atoms the optical and electronic properties of nanosized clusters are also
to interior atoms increases. For example, a 38 atom cuboctahedral representative of how molecular scale behavior evolves into bulk so-
cluster has 32 surface atoms and 6 interior atoms, while a 13 atom lid-state phenomena [19]. In particular, small gold clusters have at-
icosahedral cluster has 12 surface atoms and only one interior tracted interest as tips and contacts in molecular electronic circuits
atom. Clearly, cluster materials are chemically more reactive than [20] and also as potential chemical catalysts [21]. Recent experi-
bulk materials. Theoretically, these metal clusters may be catego- ments have shown that, when dispersed as fine particles on metal
oxide surfaces, gold facilitates a number of oxidation reactions, in
* Corresponding author. Tel.: +91 3712 267008; fax: +91 3712 267005. particular the CO?CO2 oxidation [22,23]. Interestingly enough, the
E-mail address: ramesh@tezu.ernet.in (R.C. Deka). activity of gold clusters has been found to depend critically on the

0166-1280/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.theochem.2008.09.018
84 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93

size of the particle and the nature of the substrate. For example, the crossover from planar to non planar gold clusters occurs at
when supported on the Mg(OH)2 surface, the best catalytic activity the cluster consisting of 15 gold atoms. Sankaran and Viswana-
for oxidation of CO is observed for a system size of 13 atoms [24] than [42] performed DFT calculation on Aun clusters (n = 2–12)
while the optimal activity of gold clusters supported on TiO2 surfaces using B3LYP functional and LANL2DZ effective core potential ba-
occurs for sizes of 2–3 nm (corresponding to a few hundred atoms, sis set without imposing any symmetry and observed a 2D–3D
approximately). Since the chemical and physical properties of gold conversion at Au5. Li et al. [44] have located the 2D–3D turnover
clusters depend greatly on their physical structures, significant for neutral gold clusters at Au12. Thus, inspite of such large num-
efforts have been made to determine the most stable configurations ber of studies dedicated to the structure of both neutral and io-
of gold clusters in this size range. Before developing any practical nic gold clusters, there appears to be little consensus regarding
applications, it is important to understand the chemical, thermody- the ‘‘turnover point” from clusters in which planar isomers are
namic, electronic and optical properties of the nanoclusters. Apart lowest in energy to those in which non planar isomers dominate.
from the most stable cluster isomers, other possible low-energy Therefore, we were motivated to investigate the 2D-to-3D con-
structures must also be investigated. Unfortunately, direct and com- version of neutral gold cluster using density functional theory with
prehensive experimental information of smaller gold clusters is dif- all electron scalar relativistic effects. Most studies on gold clusters
ficult to acquire. Therefore, smaller gold clusters have been make use of effective core potential (ECP), in which the core elec-
theoretically investigated using various computational methods trons are replaced by an effective potential, thereby eliminating
over the past few years [25–43]. Most of these studies have shown the need for core basis functions. The use of ECP reduces the
that neutral and charged gold clusters favor two-dimensional (2D) computational expense drastically. In contrast to these earlier
structures to unusually large sizes while few of them have found calculations, the work described in this paper uses a first principles
the lowest energy structure to be three-dimensional (3D). The occur- all-electron method based on DFT incorporating scalar relativistic
rence of planar gas-phase clusters with large cluster sizes is attrib- effects for the treatment of core electrons. We have attempted to
uted to the strong relativistic bonding effects in gold that reduce predict a number of possible low-energy structural isomers for
the s–d energy gap and thus inducing hybridization of the atomic each Aun cluster (n = 2–13) with unconstrained geometry optimi-
5d–6s orbitals causing an overlap of the 5d shells of neighboring zations starting from different initial candidate geometries.
atoms in the cluster [26].
Häkkinen and Landman [25] have investigated the low-energy 2. Computational method
structures of gold clusters and their anions for n = 2–10 atoms
using density functional theory (DFT) employing scalar relativis- The isomers of gold clusters Aun (n = 2–13) have been fully opti-
tic pseudopotentials for the 5d106s1 valence electrons of gold mized using Kohn–Sham Theory [45,46]. We used the generalized
and a generalized gradient approximation (GGA). They located gradient approximation (GGA) in our calculations. At the GGA le-
the planar to 3D transition for neutral clusters at n = 7. They also vel, we have chosen the BLYP functional [47,48] which incorpo-
compared their DFT based electronic structure calculations of rates Becke’s exchange and Lee–Yang–Parr correlation. The
small gold cluster anions with experimental photoelectron spec- efficiency of the BLYP functional in studying gold clusters have
troscopy (PES) data [26]. Walker [28] performed DFT calculations been demonstrated by Bonačić-Koutecký et al. [49] in their work
to determine the optimized geometries of neutral Aun (n = 2–11) on silver–gold bimetallic clusters. The basis set chosen is DNP
and cationic Aun+ (n = 2–9) clusters and found that the 2D-to-3D [50]. The DNP basis functions are the double-numerical atomic
transition for neutral clusters occurs at Au11. Furche et al. [30] orbitals augmented by polarization functions, i.e., functions with
performed a combined experimental and theoretical study of angular momentum one higher than that of the highest occupied
gold cluster anions for n < 13. The experimental study consisted orbital in the free atom. The size is comparable with the Gaussian
of ion mobility experiments and the theoretical study was done 6-31G** basis set, but the DNP basis set is supposed to be more
by using ab-initio molecular dynamics. By comparing theoretical accurate than a Gaussian basis set of similar size. This basis set is
and experimental results, they assigned structures for the gold suitable even for treating completely dissociated atoms. Gold being
cluster anions and located the transition from two-dimensional a heavy atom, relativistic effects become important. So, all-electron
to three-dimensional cluster to occur at n = 12. While several calculations with scalar relativistic corrections are used. While cal-
ab-initio calculations on gold nanoclusters in the past have been culations involving all-electron scalar relativistic (AER) method are
limited by symmetry constraints [31,32], an unconstrained glo- more difficult to perform due to the huge computational expense,
bal minima search [33] is necessary for locating all possible they are supposed to provide better accuracy than those involving
low-energy isomers. Considerable progress has been made to- effective core potentials as a result of the lesser number of approx-
wards this goal by Garzón et al. [34] and Soler et al. [35] who imations in AER. The advantages of the all-electron scalar relativis-
predicted amorphous structures for the most stable isomers of tic method over the effective core potential methods have been
intermediate size Aun (n = 38,55,75) nanoclusters based on an demonstrated by Orita et al. [51] in their work on CO adsorption
exhaustive minima search using generic optimization methods on Pt(111). They found that the AER method can correctly predict
that employed classical Gupta n-body potentials. Similarly, the atop adsorption site and gives adsorption energies which are
empirical potentials have been employed in the past to deter- closer to the experimental values than those predicted by ECP
mine the lowest energy structures of medium sized gold clusters methods. Fine grid mesh points are employed for the matrix inte-
[36–39]. However, recently Zhao and coworkers [40] have used a grations. Self-consistent field procedures are done with a conver-
combined scheme of DFT with empirical generic algorithms to gence criterion of 106 a.u. on the energy and electron density.
generate a number of possible structural isomers for gold clus- All the calculations are carried out using the DMol3 program
ters of up to 20 atoms, which were further optimized at the [52]. The nature of the stationary points has been characterized
DFT level to determine the lowest energy structures. They have by running vibrational frequency calculations at the same level
found that small Aun clusters adopt planar structures up to of theory. In order to determine the ground electronic state for a
n = 6, flat cage structures are preferred in the range of n = 10– given cluster size, test calculations were initially performed in
14 and a structural transition from flat-cage-like structure to two spin states, singlet, and doublet, for neutral clusters having
compact near-spherical structure occurs at around n = 15. Xiao even and odd number of gold atoms, respectively. Then the same
and Wang [41] used a plane wave DFT basis set and the PW91 calculations were repeated in higher spin states, viz, triplet and
functional for Au14 and Au20 to predict by interpolation that quartet. An analysis of the results revealed that though the
 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93 85

structures remained the same, calculations in higher spin states Table 1

gave higher energies (2.1–3.6 eV) for all the clusters studied. So Calculated properties of Au2

we decided to perform the calculations in the minimum spin Method Re (Å) xe (cm1) De (eV)
states, singlet and doublet. For charged clusters the choice of sin- BLYP/DNP 2.53 171 2.203
glet or doublet state was made depending upon the number of BOP/DNP 2.61 137 2.025
electrons being even or odd, respectively. For a given cluster size, BP/DNP 2.57 148 2.324
the most stable structure is determined from the relative energy HCTH/DNP 2.60 135 2.007
PBE/DNP 2.57 147 2.400
of the isomers. This is further confirmed by comparing their bind- PW91/DNP 2.57 148 2.426
ing energies. rPBE/DNP 2.60 140 2.140
In order to determine the size evolution of the stability of the VWN_BP/DNP 2.57 148 2.325
clusters and their electronic properties, we have determined the VWN/DNP 2.44 205.8 2.982
Experimental 2.47 191 2.302
binding energy per atom, the second difference of binding energy,
the fragmentation energy, the first vertical ionization potential, Experimental values are taken from Refs. [15,54].
electron affinity and chemical hardness of the stable structures.
The average binding energy per atom is computed from:
experimental value of 2.302 eV, but the vibrational frequency is
Eb ðnÞ ¼ ½nEðAu1 Þ  EðAun Þ=n ð1Þ far too low. Among all the available functionals, it is observed that
where E(Aun) is the energy of a cluster with n atoms. the BLYP functional gives the best estimate of all the three param-
eters considered. Hence we choose this functional for the present
The fragmentation energy for a cluster of size n has been calcu-
study.
lated from:
For all the Aun clusters (n = 2–13) a number of low-energy
Dðn; n  1Þ ¼ EðAun  1Þ þ EðAu1 Þ  EðAun Þ ð2Þ structures are obtained along with the most stable structure
(Fig. 1). The total energy of all the clusters and their relative ener-
For the most stable structure for each cluster size n, the second
gies along with the symmetry point group are given in Table 2.
difference of binding energy D2Eb(n) has been obtained as:
Each structure is identified by its shape and symmetry. The energy
D2 Eb ðnÞ ¼ 2Eb ðnÞ  Eb ðn þ 1Þ  Eb ðn  1Þ ð3Þ differences are given relative to the most stable structure. How-
ever, it should be noted that DFT can not distinguish the structures
where Eb(n) is the binding energy per atom of the cluster with n whose energy difference falls within the ca. 0.2 eV range [55]. We
atoms. take this energy as the ‘‘isomer-coexistence interval”. The binding
The vertical ionization potential (IP) and electron affinity (EA) energy per atom, the second difference of binding energy, the frag-
are calculated as: mentation energy, the average bond length, the HOMO–LUMO en-
IP ¼ EðAuþn Þ  EðAun Þ ð4Þ ergy gap, the first vertical ionization potential and electron affinity
and the value of chemical hardness for each structure are given in
EA ¼ EðAun Þ  EðAun Þ ð5Þ
Table 3. In what follows, we will first discuss and compare struc-
where E(Aun+) and E(Aun) are the total energy of the cationic and ture and energetics of Aun clusters and then their electronic
anionic clusters, respectively, at the optimized geometry of the neu- properties.
tral cluster.
Chemical hardness is used as a parameter to determine the 3.1. Structure and energetics
reactivity of a system. In density functional theory, hardness (g)
of an electronic system is defined as the second derivative of en- Au2: We first present our calculated values of Au2 and compare
ergy (E) with respect to the number of electrons (N) at constant them with the reported experimental and theoretical values. The
external potential, vð~
rÞ [53] calculated bond length, binding energy and vibrational frequency
!   for Au2 are 2.53 Å, 2.214 eV and 171 cm1, respectively. These
1 o2 E 1 ol GGA calculated values are in satisfactory agreement with experi-
g¼ ¼ ð6Þ
2 oN2 2 oN mðrÞ mental results (Re = 2.47 Å, Eb = 2.30 eV and t = 191 cm1)
mðrÞ
[56,57]. The BLYP/DNP level of theory gives 2.4% error in bond
where l is the chemical potential of the system. length, 8.8% error in binding energy and 10.4% error in vibrational
Using the finite difference approximation and the Koopmans’ frequency. However, our values are still better than some of the
theorem, global hardness, g can be approximated as theoretical values reported earlier [28,40]. The LDA results of Wang
IP  EA et al. [40] gave a value of 2.55 Å for Au–Au bond length. Using the
g¼ ð7Þ PW91PW91 exchange-correlation functional, Walker [28] found
2
that the calculated bond length is 2.56 Å with Stuttgart 1997 basis
where IP and EA are the first vertical ionization potential and elec- set, 2.55 Å with LANL2DZ and 2.54 Å with CRENBL. The previous
tron affinity, respectively, of the chemical system. theoretical calculations show that the error in the calculated vibra-
tional frequencies is >10% using various functionals and basis sets
3. Results and discussion [28]. Using the PW91PW91 exchange-correlation functional, the
calculated frequency is 126 cm1 with Stuttgart 1997 basis set,
In order to determine the best exchange-correlation functional 170 cm1 with LANL2DZ and 172 cm1 with CRENBL [28]. Our
to be used for the calculations, we made a comparative study of binding energy values are in very good agreement with
the bond length, vibrational frequency and dissociation energy of PW91PW91/LANL2DZ (2.20 eV) and PW91PW91/CRENBL
the gold dimer using various functionals. The results are presented (2.23 eV) levels of theory. The value of Au–Au bond length deter-
in Table 1. The experimental value of the bond length for Au2 is mined by us agrees well with the CCSD(T) value of 2.512 Å [14].
2.47 Å [54]. Among the functionals used, the VWN functional gives The calculated ionization potential (IP) for Au2 falls within 2.2%
the best value of bond length, but the vibrational frequency and of the experimental value.
dissociation energy are overestimated. The dissociation energy cal- Au3: Three isomers (3a–3c) are obtained with the linear one
culated by the VWN_BP and the BP functionals are close to the being lowest in energy. This is in agreement with DFT–LDA results
86 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93

Fig. 1. The calculated structures of gold clusters (Aun, n = 2–13) at BLYP/DNP level of calculations.

of Wang et al. [40] who predicted a linear structure and also with and an energy difference of 0.010 eV as compared to the linear
earlier CASSCF [58] studies but disagrees with the lowest energy molecule. The highest energy structure is a triangle 3c (D3h) with
triangular structure reported by Walker [28]. The next higher in an energy difference of 0.052 eV over the lowest energy linear
energy is a V shaped structure (3b) with a bond angle of 154.2° structure. Another study [44] has the V shaped cluster as the
 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93 87

Table 2
Shape, symmetry, total energy and relative energy values for Aun clusters, n = 2–13 optimized at the BLYP/DNP level of theory

n Structure no. Shape Symmetry Total energy including ZPVE Relative energy
2 2 Linear D1h 1001298.46578
3 3a Linear D1h 1501948.02002 0.000
3 3b V C2v 1501948.00969 0.010
3 3c Triangle D3h 1501947.96816 0.052
4 4a Rhombus Cs 2002598.48497 0.000
4 4b Distorted Y C1 2002598.42509 0.060
4 4c Y C2v 2002598.34697 0.138
4 4d Zigzag Cs 2002598.10427 0.381
4 4e Tetrahedron Td 2002597.47693 1.008
5 5a W C2v 2503248.77241 0.000
5 5b X D2h 2503248.38505 0.387
5 5c Trigonal bipyramid D3h 2503247.96743 0.805
6 6a Triangle C2v 3003899.84412 0.000
6 6b Tricapped triangle C1 3003899.84324 0.001
6 6c Pentagonal pyramid C5v 3003898.92761 0.917
6 6d Bicapped butterfly C2h 3003898.84505 0.999
6 6e Parallelogram C1 3003898.84369 1.000
6 6f Chair C2h 3003898.83910 1.005
6 6g Boat C1 3003898.51200 1.332
6 6h Capped trigonal bipyramid Cs 3003898.31692 1.527
6 6i Octahedron Oh 3003898.17314 1.671
6 6j See-saw C2h 3003898.03159 1.813
6 6k Zigzag C1 3003897.81153 2.003
7 7a Capped triangle Cs 3504549.50540 0.000
7 7b Hexagon D6h 3504549.44987 0.056
7 7c Bi-edge-capped trigonal bipyramid Cs 3504549.11205 0.393
7 7d Extended W C2v 3504549.05293 0.453
7 7e Face capped octahedron C3v 3504548.88791 0.618
7 7f Pentagonal bipyramid D5h 3504548.83410 0.671
7 7g Eclipsed sandwich D3h 3504547.67017 1.835
8 8a Tetra-edge-capped rhombus D2h 4005200.43709 0.000
8 8b Capped hexagon Cs 4005200.14378 0.293
8 8c Three dimensional structure C2v 4005199.73131 0.706
8 8d Bi-edge-capped octahedron D2h 4005198.83354 1.604
9 9a Bi-edge-capped hexagon I C2v 4505850.54997 0.000
9 9b Bi-edge-capped hexagon II Cs 4505850.53316 0.017
9 9c Bi-edge-capped hexagon III D2h 4505850.47574 0.074
9 9d Tetra-edge-capped square pyramid C4v 4505850.27459 0.275
9 9e Condensed trigonal bipyramid D2h 4505848.62416 1.926
9 9f Boat D3 4505848.52930 2.021
10 10a Tri-capped hexagon D2h 5006501.24075 0.000
10 10b Square antiprism C1 5006500.07229 1.168
10 10c Three dimensional structure I C1 5006499.00000 1.922
10 10d Three dimensional structure II D4d 5006498.84051 2.400
11 11a Tetra-capped hexagon Cs 5507151.46912 0.000
11 11b Tetra-edge-capped hexagon D2h 5507151.30703 0.162
11 11c Staggered sandwich C2h 5507149.27896 2.190
11 11d Eclipsed sandwich D5h 5507149.14500 2.324
12 12a Pentacapped hexagon D3h 6007802.52572 0.000
12 12b Three dimensional structure I D2d 6007800.84008 1.686
12 12c Three dimensional structure II C1 6007800.77947 1.746
12 12d Ladderane D3d 6007800.52935 1.996
12 12e Distorted condensed triangle C2v 6007800.42735 1.938
13 13a Hexa-capped hexagon Cs 6508452.31362 0.000
13 13b Condensed hexagon D2h 6508452.24670 0.067
13 13c Tricapped boat C2v 6508451.61101 0.703
13 13d Staggered sandwich D6d 6508450.93911 1.375
13 13e Distorted sandwich C2h 6508450.71491 1.599

All energies are in eV.

lowest in energy, followed by the linear structure, and has the tri- symmetry, a ‘‘distorted Y” shaped one (4b) and a ‘‘planar Y” shaped
angular geometry as the highest energy structure. The accuracy of structure (4c). The rhombus is slightly distorted with 0 opposite
the DFT method for calculating the energy difference between gold edges revealing Au–Au bond lengths of 2.64 and 2.69 Å A, respec-
clusters has been estimated to be 0.1 eV [30] (or 0.2 eV [55]) and so tively. The 4a structure was reported to be the most stable struc-
the three structures are probably not distinguishable at this level of ture by Walker [28] and Zhao et al. [29]. The other structures are
theory. The values of binding energy per atom for the structures 3a, a zigzag structure (4d) and a tetrahedron geometry (4e). Structures
3b and 3c are 1.209, 1.205 and 1.191 eV, respectively. 4d and 4e are higher in energy than 4a by 0.381 and 1.008 eV,
Au4: For Au4 we have obtained five stable structures as shown respectively. The ‘‘distorted Y” structure (C1) is three dimensional,
in Fig. 1 (4a–4e). The calculations reveal three structures (4a, 4b and to the best of our knowledge, it has not been reported in earlier
and 4c) within the ‘‘isomer-coexistence interval” of 0.2 eV (Table theoretical
0
studies. The average bond length for this structure is
2). Thus, the lowest energy structures are a rhombus (4a) with Cs A. Another structure, a square shaped one was also obtained.
2.68 Å
88 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93

Table 3
Binding energy/atom, second differential of binding energy, fragmentation energy, average Au–Au distance, HOMO–LUMO gap, first vertical ionisation potential, electron affinity
and hardness for Aun (n = 2–13) clusters
0
n Structure no. Eb D2Eb D Average bond length (Å
A) HOMO–LUMO gap IP EA g
2 2 1.102 2.203 2.53 1.637 9.294 1.820 3.737
3 3a 1.209 -0.175 1.423 2.55 2.231 8.754 3.361 2.696
3 3b 1.205 2.55 2.005 8.679 3.347 2.666
3 3c 1.191 2.68 0.007 7.416 2.208 2.604
4 4a 1.490 0.148 2.334 2.68 0.927 7.996 2.387 2.804
4 4b 1.475 2.68 0.966 8.130 2.585 2.773
4 4c 1.455 2.62 0.976 7.972 2.484 2.744
4 4d 1.395 2.68 1.140 8.337 2.867 2.735
4 4e 1.238 2.75 1.513 7.309 2.430 2.440
5 5a 1.623 0.086 2.156 2.68 1.142 7.790 2.962 2.414
5 5b 1.546 2.65 0.505 7.115 2.720 2.198
5 5c 1.462 2.78 0.190 7.267 2.643 2.312
6 6a 1.843 0.272 2.940 2.68 1.727 7.840 2.243 2.798
6 6b 1.843 2.68 1.721 7.839 2.244 2.797
6 6c 1.690 2.73 1.641 7.633 2.291 2.671
6 6d 1.676 2.78 0.189 7.472 3.051 2.211
6 6e 1.676 2.69 0.176 7.413 3.050 2.181
6 6f 1.675 2.68 0.180 7.473 3.054 2.210
6 6g 1.621 2.73 0.475 7.420 2.667 2.377
6 6h 1.588 2.71 0.630 7.576 2.883 2.346
6 6i 1.564 2.55 1.806 7.541 2.983 2.279
6 6j 1.541 2.68 0.684 7.338 3.010 2.164
6 6k 1.504 2.55 0.832 7.860 3.477 2.192
7 7a 1.798 0.170 1.530 2.69 1.077 7.225 2.870 2.178
7 7b 1.790 2.70 1.396 6.900 2.921 1.989
7 7c 1.742 2.76 0.690 6.927 2.798 2.065
7 7d 1.733 2.69 0.467 7.367 3.374 1.997
7 7e 1.710 2.78 1.306 6.710 3.094 1.808
7 7f 1.702 2.78 0.784 7.069 2.778 2.146
7 7g 1.536 2.74 0.591 6.798 2.866 1.966
8 8a 1.923 0.127 2.856 2.67 1.420 7.512 2.834 2.339
8 8b 1.887 2.70 0.729 7.342 2.945 2.198
8 8c 1.835 2.76 1.332 7.389 2.407 2.491
8 8d 1.723 2.76 0.144 7.447 3.326 2.061
9 9a 1.930 0.056 1.982 2.70 0.584 7.146 3.239 1.954
9 9b 1.928 2.70 0.786 7.373 3.467 1.953
9 9c 1.922 2.71 0.051 7.148 3.064 2.042
9 9d 1.899 2.72 1.113 7.319 3.195 2.062
9 9e 1.716 2.78 0.073 6.678 2.962 1.858
9 9f 1.705 2.78 0.530 7.163 3.371 1.896
10 10a 1.993 0.053 2.560 2.71 1.172 7.349 2.967 2.191
10 10b 1.876 2.79 0.822 7.047 2.736 2.156
10 10c 1.801 2.76 0.255 6.831 2.924 1.954
10 10d 1.753 2.70 0.845 6.709 2.841 1.934
11 11a 2.002 0.054 2.097 2.71 0.913 7.025 3.553 1.736
11 11b 1.988 2.70 0.727 7.004 3.606 1.699
11 11c 1.803 2.87 0.346 6.653 3.071 1.791
11 11d 1.791 2.80 0.011 6.565 3.018 1.774
12 12a 2.066 0.088 2.765 2.71 0.938 7.236 3.209 2.014
12 12b 1.939 2.80 0.229 7.360 3.035 2.163
12 12c 1.934 2.78 0.468 6.934 3.093 1.921
12 12d 1.913 2.80 1.423 7.051 3.344 1.853
12 12e 1.904 2.74 0.356 6.863 3.207 1.828
13 13a 2.047 1.657 2.71 0.033 6.770 3.343 1.714
13 13b 2.042 2.71 0.612 6.921 3.760 1.580
13 13c 1.993 2.80 0.539 6.487 3.051 1.718
13 13d 1.941 2.76 0.070 6.842 3.424 1.709
13 13e 1.924 2.79 0.058 7.415 3.250 2.082

All energies are in eV.

However, it was found to be a transition state. Other authors have one with C2v symmetry (5a). This structure has a binding energy
also predicted that the most stable Au4 cluster has a rhombic struc- per atom value of 1.623 eV which is higher than that for the struc-
ture with the ‘‘planar Y” shaped structure lying higher in energy tures 5b and 5c. There is an X shaped structure 5b (D2h) lying
[26,44,49]. Our results match this trend in energy. The rhombic 0.387 eV higher in energy. A higher energy three dimensional tri-
and tetrahedron structures have binding energies per atom of gonal bipyramidal structure with D3h symmetry is also found in
1.49 and 1.24 eV, respectively. Second-order many-body perturba- our BLYP/DNP level calculation. This structure is 0.805 eV higher
tion theory calculations [59] give binding energy per atom values in energy than the ‘‘W” shaped structure. Our results are in agree-
of 1.73 and 1.37 eV for these structures. ment with the results of Walker [28]. The binding energy per
Au5: In case of Au5, three stable structures were obtained as atom for the structures 5b and 5c are 1.546 and 1.462 eV,
shown in Fig. 1 (5a–5c). The lowest energy cluster is a ‘‘W” shaped respectively.
 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93 89

Au6: The stable structures obtained in our calculations for Au6 triangular structure. The next higher energy structure is a planar
are shown in Fig. 1 (6a–6k). We have obtained two nearly degen- extended W (7d, C2v) with an energy difference of 0.453 eV over
erate lowest-lying Au6 structures, a triangle (6a, C2v) and a tri- the capped triangle. Higher energy structures include a face capped
capped triangle (6b, C1). The structure 6b lies only 0.001 eV higher octahedron (7e, C3v), a pentagonal bipyramid (7f, D5h) and an
in energy compared to the structure 6a. They have the same value eclipsed sandwich structure 7g with D3h symmetry. This eclipsed
of binding energy per atom of 1.843 eV. The tri-capped triangle is sandwich structure consists of two triangular rings, one on top of
three dimensional and consists of four triangles, one lying in the the other connected 0 by a gold atom at the middle. The average
plane and three above the plane – one on each side of the first tri- bond length is 2.74 Å A. This structure was not reported in earlier
angle. All the four triangles are isosceles. The one lying on the studies. It has the highest energy among all the structures with
plane has Au–Au bond lengths of 2.71 and 2.76 Å and for the other the energy difference between this and the capped trianglar struc-
triangles the bond lengths lie between 2.63 and 2.67 Å. A lowest ture being 1.835 eV. A comparison of the structures 6i and 7e
energy triangular structure for Au6 with D3h symmetry was pre- shows0 that the Au–Au bond distances which have a value of
dicted by other workers [26,28,44]. Hence, within the DFT error, 2.78 ÅA in the octahedron structure (6i) changes as a result of face
there is a 2D–3D coexistence for neutral gold clusters at Au6. The capping. In the structure 7e,
0
the Au–Au distances in the octahedron
next higher energy structure is a pentagonal pyramid (6c) with varies from
0
A. The capped gold atom lies at a distance
2.74 to 2.91 Å
C5v symmetry which is also reported by Bonačić-Koutecky et al. of 2.71 ÅA from an atom in the capped face of the octahedron. We
[2]. Among the other stable structures for Au6 are a bicapped but- did not find a 2D–3D coexistence for Au7 as observed for Au6.
terfly structure (6d) with C2h symmetry, a planar parallelogram Au8: The lowest energy structure for Au8 cluster is clearly a pla-
structure (6e) which is a combination of two rhombuses, a chair nar tetra-edge-capped rhombus (8a) with D2h symmetry. Our cal-
structure (6f) with C2h symmetry. In structure
0
6d, the Au–Au bond culated minimum energy structure is in agreement with those
length varies between 2.64 and 2.80 Å A for the basic butterfly struc-
0
reported in the literature [26,28,44]. Wang et al. [40] reported that
ture while for the capping the bond lengths are 2.63 and 2.64 Å A. a distorted bicapped octahedron is the lowest energy structure. In
The chair form of Au6 consists of a planar central Au4 ring and our case we have obtained a bi-edge-capped octahedron as the
two Au atoms lying above and below the ring connected to two highest energy structure. In an earlier paper, Häkkinen and Land-
opposite edges. The 0
average Au–Au distance for the Au04 ring of this man [25] reported a capped tetrahedron as the lowest energy
A while for the other bonds it is 2.67 Å
structure is 2.63 Å A. The values structure for neutral Au8 cluster. The other structures obtained in
of binding energy per atom, the first vertical ionisation potential our calculation are a capped hexagon (8b, Cs) and a three dimen-
and electron affinity for this structure are 1.675, 7.473 and sional bi-edge-capped octahedron (8d, D2h). Another three dimen-
3.054 eV, respectively. Other higher energy structures include a sional structure (8c) with C2v symmetry is obtained which is higher
boat (6g, C1), capped trigonal bipyramid (6h, Cs), an octahedron in energy than the minimum energy planar structure by 0.706 eV.
(6i, Oh) and a see-saw structure (6j) consisting of two equilateral Structure 8d has the highest energy with the energy difference be-
triangles connected through their vertices. The two triangles are tween this and the minimum energy planar structure being
non overlapping. The highest energy structure (6k) is a zigzag 1.604 eV. The binding energy per atom value for these isomers of
one (C1). The pentagonal pyramid, capped bipyramid, rhombus Au8 lie in the range 1.923–1.723 eV. The minimum energy planar
and boat structures have also been reported by Walker [28]. In case structure 8a has the highest value of binding energy per atom as
of the trigonal bipyramid structure0
5c, the Au–Au bond distances expected.
for the
0
triangular ring is 2.78 A
Å and the other bond distances are Au9: Au9 has three structures within an energy difference of
2.75 ÅA. In the capped trigonal bipyramid the Au–Au 0
distance for 0.1 eV. The calculated lowest energy structure of Au9 is a ‘‘bi-
the triangular ring varies between 2.740
A. The other bond
and 2.77 Å edge-capped hexagon I” (9a) with C2v symmetry. This structure is
lengths lie between
0
2.71 and 2.85 ÅA. The capped Au atom lies at a in agreement with the results of Walker [28]. Recent calculations
distance of 2.53 ÅA from the nearest atom in the triangular ring. The by Xiao and Wang [41] and that of Bonačić-Koutecký and cowork-
groups of Michaelian [36] and Wilson [37] found that the bipyra- ers [2] predict the ‘‘bi-edge-capped hexagon II” structure (9b) of
mid is the lowest energy structure of Au6 using n-body Gupta po- Fig. 1 as the lowest energy structure for Au9. We found this struc-
tential and empirical Murrell-Mottram many body potential, ture to be only 0.017 eV higher in energy than the capped hexagon
respectively. Both n-body Gupta and the Murrell-Mottram many- I structure. We also note that Wang et al [40] and Häkkinen and
body potentials are designed to reproduce properties of bulk crys- Landman [25] predicted that the lowest energy structure for Au9
talline gold which may not reproduce the properties of clusters. To has a three dimensional structure. Here, the lowest energy three
the best of our knowledge, the tri-capped triangle (6b), the chair dimensional structure is found to be 0.275 eV higher in energy
(6f), and the see-saw structure (6j) have not been reported in compared to the minimum energy structure. Clearly, there is no
any previous experimental and theoretical work. 2D–3D coexistence for the stable isomers of Au9. The two dimen-
Au7: Au7 has seven stable structures (7a–7g). Within the iso- sional ‘‘bi-edge-capped hexagon III” structure (9c, D2h) also lies
mer-coexistence interval, the lowest energy structures for Au7 within the isomer-coexistence interval of the lowest energy struc-
are a planar capped triangle with Cs symmetry (7a, Cs) and a hexa- ture 9a as does the structure 9b. Thus the three two dimensional
gon with one gold atom occupying the centre of the ring (7b, D6h). structures 9a, 9b and 9c may be said to coexist as the minimum en-
A planar capped triangle with Cs symmetry was also obtained as ergy structure. Higher energy stable structures include the three
the minimum energy structure by other workers by adding a gold dimensional tetra-edge-capped square pyramid (9d, C4v), a con-
atom to the planar triangular structure of Au6 [2,26,28,44]. Then densed trigonal bipyramid structure (9e, D2h) and a boat structure
we have a bi-edge-capped trigonal bipyramid structure (7c) which with D3 symmetry (9f). The condensed trigonal bipyramid struc-
is about 0.393 eV higher in energy than the hexagonal structure. ture (9e, D2h) consists of two trigonal bipyramids condensed
Wang et al. [40] reported a three dimensional pentagonal bipyra- through a vertex of the0 triangular ring. The average bond length
mid structure as the lowest energy structure for Au7. In our case A. To the best of our knowledge, this struc-
for this cluster is 2.78 Å
this pentagonal bipyramid structure was found to be 0.671 eV ture has not been reported previously. The boat shaped one with
higher in energy than the minimum energy structure. Gilb et al. D3 symmetry is another new structure obtained in this study.
[60] found a hexagon with one central atom as the lowest energy Among all the stable structures obtained for Au9, this boat struc-
structure for Au7 cation. In our case this structure is found to be ture has the highest energy with the energy difference between
next higher in energy as compared to the minimum energy capped this and the most stable bi-edge-capped hexagon I being 2.021 eV.
90 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93

Au10: The lowest energy structure for Au10 is a tricapped hexa- Au13: Five stable structures were obtained with two planar
gon (10a) having D2h symmetry which is in agreement with the structures, a hexa capped hexagon (13a, Cs) and a planar con-
result obtained by Walker [28] and Li et al. [44]. The other stable densed hexagon structure (13b, D2h) as lowest structures within
structures are three dimensional. The energy difference between the ‘‘isomer-coexistence interval”. Hou and coworkers [29] have
the minimum energy planar structure and the next higher energy also obtained the condensed hexagon structure consisting of two
three dimensional square antiprism structure (10b) is 1.168 eV. hexagonal rings condensed through a vertex as the most stable
Structure 10b consists of two square pyramids placed back to back structure. Taking the hexagonal ring as the base, Häkkinen et al.
in staggered conformation. This structure has been obtained by [26] have also found a planar structure as the minima in their
Furche et al. [30] and by Häkkinen et al. [26] in their study on gold study on gold cluster anions. However, their structure contains
cluster anions. The other two higher energy structures (10c and capping at different positions whereas there is no capping in our
10d) have symmetry 0C1 and D4d symmetry with average bond structure. Kappes and coworkers [60] have found a three dimen-
lengths 2.79 and 2.70 Å A, respectively. The highest energy structure sional C2v structure as the lowest energy one in their study on gold
10d is 2.4 eV higher in energy than the minimum energy planar cluster cations. The other stable structures for Au13 are three
structure. dimensional and higher in energy than the planar ones. Among
Au11: In agreement with the results of Häkkinen et al. [26] for them is a tricapped boat structure (13c, C2v) and two hitherto unre-
gold cluster anions, we find a planar tetra capped hexagon (11a, ported structures: one is a staggered sandwich structure (13d, D6d)
Cs) as the minimum energy structure for Au11. The next higher en- consisting of two six member planar rings placed one on top of the
ergy structure (11b) is also planar and is separated by an energy other with one gold atom in between the planes. The six member0
difference of 0.162 eV from the structure 11a. Structure 11a has rings are regular hexagons with the Au–Au distance being 2.70 Å A
been reported by several groups [26,30,44]. We have also obtained while the distance of the central
0
gold atom from the atoms in
two previously unreported structures – one of them is a staggered the hexagonal ring is 2.96 Å A. A comparison of structures 11c and
sandwich structure (11c, C2h). In this case, the atoms in the two 13d, both of which are staggered sandwiches, shows that in the
pentagonal rings are placed in staggered position with respect to case of the pentagonal ring (11c), the Au–Au distance
0
is not uni-
each 0other. The average bond length of the pentagonal ring is form with the average bond length being 2.90 Å A, while the hexag-
2.90 ÅA while the central gold atom is separated
0
from an atom in onal ring shows equal Au–Au bonds. On going from Au11 to Au13
the pentagonal ring by a distance of 2.74 Å A. The other is an eclipsed there is an increase in the distance between the 0 central gold atom
sandwich structure (11d, D5h). This structure consists of two over-0 and atoms in the planar ring from 2.74 to 2.96 Å A. The other struc-
lapping regular pentagonal rings with Au–Au distance of 2.79 Å A ture is a distorted sandwich structure (13e, C2h) with two distorted
with a gold atom in between the two planes. All the distances be- non planar
0
Au6 rings. The average bond length of this structure is
tween the constituent atoms 0
of the pentagonal ring and the central A.
2.79 Å
gold atom are equal (2.79 Å A). The distance between an atom0
in the
pentagonal ring and the gold atom at the centre is 2.79 Å A. The pen-
0
3.2. Variation of average Au–Au bond length
tagonal rings are separated from each other by a distance of 2.82 Å A.
The binding energy per atom value for the two structures are 1.803 In order to compare the variation of average Au–Au bond length
and 1.791 eV, respectively. of Aun clusters with the increase of cluster size we choose the most
Au12: Again, we find a planar penta-capped hexagon structure stable structure for each of the clusters. The variation of average
to be the lowest in energy for Au12 (12a, D3h). This structure con- Au–Au bond length of all the lowest energy structures as a function
sists of a hexagonal ring with capping at five positions, three on of cluster size is shown in Fig. 2. The average 0bond length of Au 0 n
the edges and two at the vertices. This is in agreement with the re- clusters increases as we go from n = 2 (2.53 Å A) to n = 4 (2.68 Å
A).
sults of Furche et al. [30] and Häkkinen et al. [26] for gold cluster For Au4 to Au6, the average Au–Au bond length of the minimum
anions. This structure has also been obtained by Hou and cowork- energy structures are 0
exactly same followed by an increase of the
ers [29] in their study of two dimensional gold clusters. Li et al. same at Au7 (2.70
0
A). At Au8, the average Au–Au bond length de-
Å
[44] also obtained this structure. However, this structure was creases (2.67 ÅA) from that of Au7. This decrease of bond length at
found to be higher in energy by 0.022 eV than the minimum energy Au8 reflects the change of packing of Au atoms in Au8 which has
structure, which they found to be three dimensional. Then, we a capped rhombus structure as the most stable one. The most sta-
have two- and three-dimensional structures 12b (D2d) and 12c ble rhombus structure of Au4 has the similar Au–Au bond lengths. 0
(C1) lying 1.686 and 1.746 eV, respectively higher in energy than There is an increase
0
of average Au–Au bond length for Au9 (2.70 ÅA)
the minimum energy planar structure. Structure 12b consists of and Au10 (2.71 Å A) clusters from that of Au8. For Au11, Au12 and Au13
six four membered rings, two of which are planar and the rest clusters the average Au–Au bond lengths are same as that of Au10.
are butterfly
0
shaped. The average bond length for the planar0
ring
A while for the butterfly shaped rings it is 2.81 Å
is 2.77 Å A. The next
higher in energy is a ladderane structure (12d) having D3d symme-
try with four stacked triangles, one on top of the other in staggered
positions. All the triangles are equilateral.
0
The top and bottom tri-
angles have Au–Au bond length 0 of 2.81 Å A while for the middle two
the Au–Au bond length is 2.98 Å A. The binding energy per atom for
this structure is 1.913 eV. This exotic structure was not reported in
earlier studies. The highest in energy is a distorted condensed tri-
angle (12e) with C2v symmetry. This structure consists of two
hexanuclear gold triangles (similar to structure 6a) joined in a par-
allel conformation. The two triangles deviate slightly from planar-
ity, the Au–Au–Au bond angle of the deviated edge being 151.7°. Li
et al. [44] have found this structure to be the minimum ?energy
structure for Au12. In our study, we find that the 12e structure
is 1.938 eV higher in energy than the minimum energy planar
structure 12a. Fig. 2. Variation of average bond length with respect to cluster size.
 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93 91

The variation of average Au–Au bond lengths in our calculations is 0.3

slightly different from that reported by Wang et al. [40]. With the
0.2
increase of cluster size they obtained a structural change from pla-
0.1

Δ2Eb(eV)
nar to 3D, the onset of three dimensional structure occurring at
n = 7. On the other hand, all the minimum energy structures in 0.0
our calculations are planar.
-0.1

3.3. Binding energy and relative stability -0.2

-0.3
The binding energy values for all the Aun (n = 2–13) clusters are 2 3 4 5 6 7 8 9 10 11 12 13
given in Table 3. Our calculated binding energies are lower than Number of Au atoms in cluster
those reported by Wang et al. [40] using DFT–LDA methods. How-
ever, it is known that GGA underestimates the binding energy by Fig. 4. Variation of second energy differential of binding energy per atom with
0.3–0.4 eV/atom with respect to LDA based values [32]. As a com- cluster size.
parison, our GGA calculation yields a value of 1.102 eV/atom for
Au2 as against a value of 1.22 eV/atom [40] by DFT–LDA methods.
Our calculated binding energy is in very good agreement with the even-numbered Aun clusters are relatively more stable than the
experimental value of 1.15 eV/atom [54]. The variation of calcu- neighbouring odd-numbered ones. Among all the minimum energy
lated binding energy per atom (Eb) of the most stable structure structures, triangular Au6 has the largest D2Eb value of 0.272 eV/
for each cluster as function of cluster size is shown in Fig. 3. As atom, indicating its high stability.
can be seen from the plot, the per atom binding energy increases
with cluster size and has a maximum value at Au6 in agreement 3.4. HOMO–LUMO gap
with the results of Häkkinen and Landman [25], Xiao and Wang
[41] and Bonačić-Koutecky and coworkers [2]. This indicates that The energy gap between the highest occupied molecular orbital
triangular Au6 is more stable than its neighbours. The increase of (HOMO) and the lowest unoccupied molecular orbital (LUMO) for
Eb(n) with n indicates the greater stability as the cluster grows in each stable structure has been calculated. A large HOMO–LUMO
size. This may be due to the fact that with increasing cluster size, gap has been considered as an important requisite for the chemical
the average number of nearest neighbours increases, which pro- stability of gold clusters [61]. The HOMO–LUMO gap for all the 58
motes greater average number of interactions per atom. For the stable structures are given in Table 3.
same cluster size, binding energy can be used to determine the rel- The variation in the HOMO–LUMO gap for the most stable
ative stability of the isomers [29]. structures for each cluster size is shown in Fig. 5. As seen from
In order to analyze the stability and size-dependent properties the plot, the HOMO–LUMO energy gap shows an odd–even oscilla-
of gold clusters, we also calculated fragmentation energies, tion for n > 4. The even-numbered clusters have larger HOMO–
D(n,n-1), using Eq. (2) and second differences in binding energy, LUMO gap and are relatively more chemically stable than the
D2Eb using equation 3 as defined in the methods section. The frag- odd-numbered neighbours. Odd–even alterations in the HOMO–
mentation energies are sensitive to the relative stabilities that can LUMO gap values have often been predicted by earlier calculations
be observed in mass abundance spectra. Fig. 3 shows D(n,n-1) of for other small metal clusters [62,63] as also for small Aun clusters
gold clusters which oscillate as the cluster size increases and indi- [23,26,35]. Most importantly, HOMO–LUMO gaps evaluated from
cate that the clusters with even number of atoms are more stable photoelectron spectroscopy (PES) data of gold cluster anions have
than the clusters with odd number of atoms. The fragmentation also shown oscillations [8,64]. Odd–even oscillation behaviors can
energy value for the most stable Au6 triangular cluster has the be understood by the effect of electron pairing in orbitals. Even-
highest value of 2.940 eV indicating high stability of triangular sized clusters have an even number of s valence electrons and a
Au6 cluster. Fig. 4 shows the variation of the second differential doubly occupied HOMO, while it is singly occupied for odd-sized
of binding energy per atom with cluster size n. In cluster physics, clusters. The electron in a doubly occupied HOMO feels a stronger
the quantity D2Eb is commonly known to represent the relative effective core potential due to the fact that the electron screening
stability of a cluster of size n with respect to its neighbours is weaker for electrons in the same orbital than for inner-shell elec-
[25,29]. The plot shows an odd–even oscillation with the even- trons. It is crucial to note the unusually high values of the HOMO–
numbered clusters having larger D2Eb values. This indicates that LUMO gap for Au3 unlike the HOMO–LUMO gaps obtained from

Fig. 3. Variation of binding energy per atom and fragmentation energy with respect
to cluster size. Fig. 5. Variation of HOMO–LUMO gap with cluster size.
92 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93

earlier DFT calculations [23,35]. The different ground state struc- 4. Conclusions
tures may cause a large discrepancy for Au3. However, our pattern
of HOMO–LUMO gap is in good agreement with that obtained by We have investigated low-energy geometries, binding energy,
Wang et al. [62]. Other than Au3, the HOMO–LUMO gap is particu- relative stability, HOMO–LUMO gap and chemical hardness of
larly large for Au2, Au6 and Au8 in agreement with other theoretical gold clusters Aun, n varying from 2 to 13 using BLYP/DNP level
[25,28,44] studies. of theory. Different initial guess geometries were considered
for all the clusters which lead to a number of structural isomers
3.5. Ionization potential, electron affinity and chemical hardness for each cluster size. Of these isomers, 12 structures were not re-
ported previously. The minimum energy structures for all the
In cluster science, ionization potential is used as an important clusters are found to be planar with coexistence of 3D structures
property to study the change in electronic structure of the clus- for Au4 and Au6 within the DFT error. It is observed that some of
ter with size. We have calculated the vertical ionization potential our minimum energy structures obtained by using AER method
of all the stable structures. This has been obtained from consid- are different from those reported with ECP calculations. Odd–
eration of the total energies of the neutral cluster and the cat- even alterations are evident in the HOMO–LUMO gaps, chemical
ionic cluster at the geometry of the neutral cluster. We hardness, second energy difference of binding energy and frag-
compare our values of ionization potential with experimental mentation energy. Even-numbered clusters are found to be rela-
results. For linear chain trimer, our value (8.75 eV) is larger com- tively more stable compared to the odd-numbered ones
pared to the experimental value (7.50 eV) [65]. Wang et al. [40] indicating that electron spin-pairing stabilizes these clusters.
have determined an even higher value for the first vertical ioni- The calculated results agree well with available experimental
zation potential of the linear trimer (9.05 eV). However, for the data and some of the earlier theoretical investigations of gold
Au3 triangle, our value of ionization potential (7.42 eV) compares nanoclusters. The systematic investigations of structural and
well with the experimental value (7.50 eV) [65]. This may indi- electronic properties of small gold clusters present in this study
cate that for Au3, the linear chain, which is only 0.052 eV lower will provide an insight into understanding more complicated
in energy than the triangle is not much stable than the triangu- gold clusters and clusters of other metals.
lar structure. For W shaped Au5 cluster, our value of ionization
potential (7.79 eV) is very close to the experimental value Acknowledgements
(8.0 eV) [65]. For triangular Au6, the difference between our va-
lue (7.84 eV) and the experimental value (8.8 eV) is large. In The authors gratefully acknowledge Dr. Ashwini Kr. Phukan,
Fig. 6, the calculated ionization potentials of the minimum en- Tezpur University, India and Prof. Notker Rösch, Technische Uni-
ergy structures as a function of cluster size have been plotted. versität, München, Germany for helpful suggestions and continu-
The plot shows even–odd alternative behaviour. The even-sized ous encouragement. This work was partly supported Department
clusters with an even number of s valence electrons have larger of Science and Technology (DST), New Delhi and Council of Scien-
values of ionization potential compared to their immediate tific and Industrial Research (CSIR), New Delhi.
neighbours. This is in agreement with previous studies [25,26].
Similar to ionization potential, we calculated vertical electron References
affinity values for all the minimum energy structures and they
are plotted in Fig. 6 as a function of cluster size. The vertical [1] C.R.C. Wang, S. Pollack, M.M. Kappes, J. Chem. Phys. 94 (1991) 2496.
electron affinity values also show an odd–even alterative behav- [2] V. Bonačić-Koutecký, P. Fantucci, J. Koutecký, Chem. Rev. 91 (1991) 1035.
[3] W. Ekardt (Ed.), Metal Clusters, Wiley, Chichester, 1999.
iour; but in this case, the nature is opposite. The odd-sized clus- [4] R.P. Andres, T. Bein, M.M. Dorogi, S. Feng, J.I. Henderson, C.P. Kubiak, R.
ters have higher value of electron affinity compared to the even- Osifichin, R. Reifenberger, Science 272 (1996) 1323.
sized ones. [5] M. Dorogi, J. Gomez, R. Osifichin, R.P. Andres, R. Reifenberger, Phys. Rev. B 52
(1995) 9071.
We have also calculated the chemical hardness of all the sta- [6] M. Schmidt, C. Ellert, W. Kronmüller, H. Haberland, Phys. Rev. B 59 (1999)
ble structures based on finite difference approximation. The var- 10970.
iation of chemical hardness with cluster size is shown in Fig. 6. [7] D. Schooss, S. Gilb, J. Kaller, M. Kappes, F. Furche, A. Kohn, K. May, R. Ahlrichs, J.
Chem. Phys. 113 (2000) 5361.
As can be seen in Fig. 6, the chemical hardness values also show
[8] J. Ho, K.M. Erwin, W.C. Lineberger, J. Chem. Phys. 93 (1990) 6987.
an oscillating behaviour as the cluster size increases. The odd– [9] M.B. Knickelbein, G.M. Koretsky, J. Phys. Chem. A 102 (1998) 580.
even oscillation of the chemical hardness indicates that the [10] D. Leopold, J. Ho, W.C. Lineberger, J. Chem. Phys. 86 (1987) 1715.
even-numbered gold clusters with higher hardness are more sta- [11] G. Dietrich, S. Krückeberg, K. Lützenkirchen, L. Schweikhard, C. Walther, J.
Chem. Phys. 112 (2000) 752.
ble than their neighbouring odd-numbered clusters. [12] K. Balasubramanian, M.Z. Liao, Chem. Phys. 127 (1988) 313.
[13] R. Rousseau, D. Marx, J. Chem. Phys. 112 (2000) 761.
[14] R. Wessendrup, T. Hunt, P. Schwedtfeger, J. Chem. Phys. 112 (2000) 9356.
[15] P. Pyykkö, Angew. Chem. Int. Ed. 43 (2004) 4412.
[16] E.A. Ivanova Shor, V.A. Nasluzov, A.M. Shor, G.N. Vayssilov, N. Rösch, J. Phys.
Chem. C 111 (2007) 12340.
[17] A.C. Templeton, W.P. Wuelfing, R.W. Murray, Acc. Chem. Res. 33 (2000) 27.
[18] R. Elghanian, J.J. Storhoff, R.C. Mucic, R.L. Letsinger, C.A. Mirkin, Science 277
(1997) 1078.
[19] S. Chen, R.S. Ingram, M.J. Hostetler, J.J. Pietron, R.W. Murray, T. Schaaff, J.T.
Khoury, M. Alvarez, R.L. Whetten, Science 280 (1998) 2098.
[20] F.F. Fan, A.J. Bard, Science 277 (1997) 1791.
[21] M. Valden, X. Lai, Science 281 (1998) 1647.
[22] M. Haruta, Catal. Today 36 (1997) 153.
[23] J.D. Grunwaldt, A. Baiker, J. Phys. Chem. B 103 (1999) 1002.
[24] D.A.H. Cunningham, W. Vogel, H. Kageyama, S. Tsubota, M. Haruta, J. Catal. 177
(1998) 1.
[25] H. Häkkinen, U. Landman, Phys. Rev. B 62 (2000) R2287.
[26] H. Häkkinen, B. Yoon, U. Landman, X. Li, H.-J. Zhai, L.-S. Wang, J. Phys. Chem. A
107 (2003) 6168.
[27] B. Yoon, H. Häkkinen, U. Landman, J. Phys. Chem. A 107 (2003) 4066.
[28] A.V. Walker, J. Chem. Phys. 122 (2005) 094310.
Fig. 6. Variation of IP, EA and chemical hardness with cluster size. [29] J. Zhao, J. Yang, J.G. Hou, Phys. Rev. B 67 (2003) 085404.
 A. Deka, R.C. Deka / Journal of Molecular Structure: THEOCHEM 870 (2008) 83–93 93

[30] F. Furche, R. Ahlrichs, P. Weis, C. Jacob, S. Gilb, T. Bierweiler, M.M. Kappes, J. [47] A.D. Becke, Phys. Rev. A 38 (1988) 3098.
Chem. Phys. 117 (2002) 6982. [48] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785.
[31] C.L. Cleveland, U. Landman, T.G. Schaaff, M.N. Shafigullin, P.W. Stephens, R.L. [49] V. Bonačić-Koutecký, J. Burda, R. Mitric, M. Ge, G. Zampella, P. Fantucci, J.
Whetten, Phys. Rev. Lett. 79 (1997) 1873. Chem. Phys. 117 (2002) 3120.
[32] O.D. Häberlen, S.-C. Chung, M. Stener, N. Rösch, J. Chem. Phys. 106 (1997) [50] B. Delly, D.E. Ellis, J. Chem. Phys. 76 (1982) 1949.
5189. [51] H. Orita, N. Itoh, Y. Inada, Chem. Phys. Lett. 384 (2004) 271.
[33] I.L. Garzón, A. Posada-Amarillas, Phys. Rev. B 54 (1996) 11796. [52] B. Delly, J. Chem. Phys. 92 (1990) 508.
[34] I.L. Garzón, K. Michaelian, M.R. Beltrán, A. Posada-Amarillas, P. Ordeján, E. [53] K. Fukui, Science 218 (1982) 747.
Artacho, D. Sánchez-Portal, J.M. Soler, Phys. Rev. Lett. 81 (1998) 1600. [54] G.A. Bishea, M.D. Morse, J. Chem. Phys. 95 (1991) 5646.
[35] J.M. Soler, M. Beltrán, K. Michaelian, I.L. Garzón, P. Ordeján, D. Sánchez-Portal, [55] K. Morokuma, Phil. Trans. Roy. Soc. London A 360 (2002) 1098.
E. Artacho, Phys. Rev. B 61 (2000) 5771. [56] M.D. Morse, Chem. Rev. 86 (1986) 1049.
[36] K. Michaelian, N. Rendon, I.L. Garzón, Phys. Rev. B 60 (2000) 2000. [57] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, CRC, Boca Raton,
[37] N.T. Wilson, R.L. Johnston, Eur. Phys. J. D 12 (2000) 161. 2003.
[38] M. Soler, I.L. Garzón, J.D. Joannopoulos, Solid State Commun. 117 (2001) 621. [58] K.G. Balasubramanian, M.Z. Liao, J. Chem. Phys. 86 (1987) 5587.
[39] I.L. Garzón, M.R. Beltrá, G. González, I. Gutierrez-González, K. Michaelian, J.A. [59] A.K. Ray, Comp. Mat. Sci. 25 (2002) 279.
Reyes-Nava, J.I. Rodriguez-Hernádez, Eur. Phys. J. D 24 (2003) 105. [60] S. Gilb, P. Weis, F. Furche, R. Ahlrichs, M.M. Kappes, J. Chem. Phys. 116 (2002)
[40] J. Wang, G. Wang, J. Zhao, Phys. Rev. B 66 (2002) 035418. 4094.
[41] L. Xiao, L. Wang, Chem. Phys. Lett. 392 (2004) 452. [61] J. Li, X. Li, H.J. Zhai, L.S. Wang, Science 299 (2003) 864.
[42] M. Sankaran, B. Viswanathan, Bull. Catal. Soc. India 5 (2006) 26. [62] J. Wang, G. Wang, J. Zhou, Phys. Rev. B 64 (2003) 205411.
[43] F. Remacle, E.S. Kryachko, J. Chem. Phys. 122 (2005) 044304. Adv. Quantum [63] J. Zhao, Eur. Phys. J. D 14 (2000) 309.
Chem. 47 (2004) 423.. [64] K.J. Taylor, C.L. Pettiette-Hall, O. Cheshnovsky, R.E. Smalley, J. Chem. Phys. 96
[44] X. Li, H. Wang, X. Yang, Z. Zhu, J. Chem. Phys. 126 (2007) 84505. (1992) 3319.
[45] P. Hohenberg, W. Kohn, Phys. Rev. 136 (1964) B864. [65] C. Jackschath, I. Rabin, W. Schulze, Ber. Bunsenges Phys. Chem. 86 (1992)
[46] W. Kohn, L. Sham, Phys. Rev. 140 (1965) A1133. 1200.