Computational and Theoretical Chemistry 1021 (2013) 135–143

Contents lists available at SciVerse ScienceDirect

Computational and Theoretical Chemistry

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

Structures, relative stabilities, and electronic properties of potassium

clusters Kn (13 6 n 6 80)
Andrés Aguado
Departamento de Física Teórica, Atómica y Óptica, Universidad de Valladolid, Valladolid 47071, Spain

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

Article history: We report the global minimum (GM) structures, energetic and electronic properties of Kn clusters with up
Received 24 April 2013 to 80 atoms, obtained by combining basin hopping unbiased optimizations (based on a many-body
Received in revised form 26 June 2013 empirical potential) with subsequent reoptimization of several candidate structures employing a density
Accepted 29 June 2013
functional theory method which accounts for van der Waals dispersion interactions (vdW-DFT). A com-
Available online 5 July 2013
parison with other alkali cluster systems shows that the GM structures of Kn coincide with those of Nan
more often than with those of Csn clusters. Nevertheless, the inclusion of dispersion interactions produces
Keywords:
for several clusters changes in the GM structures, which become either identical to those of Csn clusters,
Potassium clusters
Density functional theory
or completely novel (i.e. not reported for other alkali systems). Many GM structures are based on poly-
van der Waals interactions icosahedral packing; some others contain Kasper polyhedral units with disclination lines. Atoms with a
Magic clusters low coordination number (or adatoms) are avoided in the GM structures following a geometric shell clos-
Global optimization ing: such additional atom is instead inserted into the outermost atomic shell of the cluster. The calculated
Basin hopping ionization potentials, electron affinities and HOMO–LUMO gaps are found to reproduce perfectly the elec-
tron shell closings observed in photoionization measurements. The electric dipole moments of most clus-
ters are close to zero, a distinguishing feature of metallicity previously observed for Nan clusters.
However, some clusters like K26, K47 or K49 have sizable electric dipole moments. The cluster stabilities
are found to explain the experimental abundances inferred from mass spectra. The enhanced stabilities
(magic numbers) can be interpreted in terms of electron shell closing effects for the smaller clusters and
of geometrical packing effects for the larger clusters. Although there is not a sharply defined critical size
separating both regimes, it is estimated to be around 55 atoms.
Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction
There has been an ongoing interest during last years in under-
standing the geometric packing preferences of small atomic clus-
ters. Together with the electronic structure, cluster geometry
determines several physical and chemical properties such as reac-
tivity and optical absorption, which are of general interest in Nano-
science. Typically, the structure and other cluster properties can
differ substantially from those of isolated molecules and macro-
scopic systems, due to quantum confinement effects on the elec-
tron cloud and also to the large proportion of surface atoms.
Alkali metal clusters were amongst the first nanoscopic systems
investigated with mass spectroscopic techniques [1]. Those semi-
nal studies revealed that clusters with 8, 20, 34, 40, 58, . . . electrons
have an enhanced stability as compared to other cluster sizes. The
magic sizes were understood in terms of a well defined electron
shell structure with the help of the structureless jellium model
[2,3]. Since then, alkali clusters have remained an interesting ob-
E-mail address: aguado@metodos.fam.cie.uva.es
ject of study because they are representative of simple metal clus-
ters with delocalized electrons. In particular, first principles
methods such as quantum chemistry methods and density func-
tional theory have been applied to study many structural, thermal,
electronic and optical properties of sodium clusters [4–31], con-
tributing much to our present understanding of the physico-chem-
ical properties of alkali metal systems.
The number of studies devoted to potassium clusters is compar-
atively much smaller. This is somehow surprising, because on the
experimental front, results for the abundance mass spectra, ioniza-
tion potentials (IPs), polarizabilities, optical absorption properties,
and relevant fragmentation channels are available since the early
days of cluster physics [32–37]. In particular, the measured IP val-
ues show sharp drops (an indication of an electronic shell closing)
right after the sizes N = 8, 18, 20, 34, 40, 58, 92 and weaker but sig-
nificant features for N = 14, 26, 30 in good agreement with the pre-
dictions from a spheroidal Nilsson model [38]. The electron shell
closings expected at N = 68, 70 on the base of jellium model predic-
tions are not observed in the experiments, however. A liquid drop
model study incorporating shell-correction terms and a consider-
ation of electronic entropy [39] was also able to reproduce all the

2210-271X/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.comptc.2013.06.040
136 A. Aguado / Computational and Theoretical Chemistry 1021 (2013) 135–143
experimental maxima in the IP, and additionally predicting new
weaker maxima at sizes N = 46, 50 and 54, which would be ob-
served only at very low temperatures. The abundance mass spectra
show that all clusters with an electron shell closing are magic,
which demonstrates that the enhanced stabilities are correlated
with the electronic shell structure. No enhanced abundance is ob-
served at N = 55, where an icosahedral geometric shell closing is
expected, probably because the clusters in the experimental beam
are in a hot liquid phase. Secondary magic numbers (associated
with less prominent maxima in the abundance mass spectra) are
observed for N = 14, 26, 30 [36]. These features at N = 14, 26, 30
have been recently confirmed in an experimental study of the
2þ
abundances of Kþ N and KN cluster ions extracted from helium drop-
lets [40]. This study identifies also a substantial odd–even alterna-
tion in the abundances of clusters containing from 34 to 40
electrons, and a local stability maximum for clusters with about
46 electrons.
The first theoretical calculations [38,39,41,42] on potassium
clusters were based on the jellium model and its variants and, as
mentioned in the previous paragraph, were quite successful in
reproducing and explaining the results of photoionization experi-
ments. Nevertheless, those models do not take into account the de-
tailed structure of the ionic skeleton, which is essential to
understand the properties of solid-like clusters and the possible
impact of geometric packing on their stabilities. First-principles
studies based on either density functional theory (DFT) or quantum
chemical methods were reported for very small clusters containing
up to about 10 atoms only [43–51]. Banerjee et al. [52] reported ab
initio results for KN clusters containing an even number of atoms
ranging from N = 2 to N = 20. In this study, no global optimization
of cluster structures was attempted, however, and the structures of
KN were simply assumed to be the same as those of NaN clusters.
Very recently, Abdalla and coworkers [53] have reported a global
optimization study of potassium clusters with up to 20 atoms,
employing a genetic algorithm coupled to a density functional
tight binding (DFTB) method. In this work, the sampling of the po-
tential energy surface is extensive, so the quality of the results de-
pends mainly on the accuracy of the approximations employed to
build the DFTB model. Specifically, the model is parametrized by
first assuming that all potential energy matrix elements involving
non-nearest-neighbor atoms are negligible, and then fitting to DFT
results on the K2 molecule obtained with the B3LYP functional. We
have not found any other first-principles study dealing with potas-
sium clusters of larger size.
The aim of the present work is to provide a more definitive
assessment of the structures of Kn clusters in a broad size range
n = 13–80, employing global optimization methods and ab initio
DFT calculations that incorporate a description of van der Waals
dispersion interactions (vdW-DFT) [54]. The main motivation for
this study comes from the scarcity of theoretical results for
N = 13–20 and the total absence of them for bigger clusters. We
would like to reproduce and interpret the experimental results
about ionization potentials and cluster abundances in a broad size
range, and also assess the possible impact of geometric shell struc-
ture onto the stability of cold, solid-like, clusters. Another motiva-
tion is to offer a comparison with our previous results on sodium
[27] and cesium [55] clusters. In those works, we demonstrated
that NaN and CsN clusters adopt very different global minimum
(GM) structures for most sizes, a result which was a priori quite
unexpected. Also, while most NaN clusters adopt the lowest possi-
ble spin multiplicity, many CsN clusters have high magnetic mo-
ments and behave indeed as magnetic superatoms. The
additional effects of vdW interactions and sd-hybridization, which
are much more important for cesium than for sodium, explain the
differences. Although potassium clusters have been many times
assumed to be very similar to sodium clusters [52], the previous
results on Na and Cs clusters cast some doubts on this assumption.
It is therefore interesting to explicitly check its validity.
The paper is organized as follows: The theoretical approach for
locating the GM structures is briefly summarized in Section 2,
where we also include some benchmark tests of the accuracy of
our DFT calculations. Section 3 describes the structures, stabilities,
and also some electronic properties of Kn clusters. Finally, Section 4
summarizes the main conclusions of our work.
2. Theoretical methods
The methods and theoretical level employed to locate the puta-
tive GM structures are the same as in our previous works on so-
dium [27] and cesium [55] clusters. Briefly stated, the first-
principles calculations are performed at the Kohn–Sham [56] DFT
[57] level. We employ the SIESTA code [58] with different imple-
mentations of the exchange–correlation functional: the spin-polar-
ized local density approximation (LDA functional), as
parameterized by Perdew–Zunger [59] and based on Ceperley–Al-
der simulations of the electron gas [60]; the spin-polarized gener-
alized gradient approximation (GGA), as implemented by Perdew,
Burke and Ernzerhof (PBE functional) [61]; and the fully non-local
van der Waals functional (vdW-DFT), as proposed by Klimes, Bow-
ler, and Michaelides (KBM functional) [62]. This functional has
been implemented by Román-Pérez and Soler [63] in the SIESTA
code. The main results are obtained with the KBM functional.
LDA and PBE are used only for benchmark purposes and also to
perform an initial, lower-cost optimization of cluster structures,
which are then reoptimized with the KBM functional.
Each potassium atom contributes one electron to the valence
electron density. The effect of core electrons is represented by
norm-conserving pseudopotentials [65] which contain scalar rela-
tivistic and non-linear core corrections [66]. The cluster wave func-
tion is expanded into a basis set of localized atom-centered
orbitals, of DZP2 quality. A 100 Ryd cutoff is employed for the fast
Fourier transform grid. Table 1 shows benchmark results obtained
for the dimer and bulk limits with these computational settings.
The KBM functional is found to perform much better than either
LDA or GGA approximations in describing the several properties,
suggesting that vdW corrections are not negligible for potassium.
This observation is in line with the results of previous works
[67,68].
We would like to emphasize that the improvement obtained
when using the dispersion corrected functional is a direct conse-
quence of a more accurate representation of the interactions, and
does not happen by chance. In a metal, the dispersion interaction
can be understood as the response of the density of the electron
Table 1
Equilibrium bond distance d, cohesive energy Ecoh, and vertical ionization potential
(VIP) of the K2 dimer, and lattice constant a, bulk modulus B and cohesive energy Ecoh
of bulk potassium, as predicted by several exchange–correlation functionals, are
compared to experimental results. Experimental values are taken from Ref. [64] for
bulk potassium, and from Ref. [48] for the dimer.
K2 d (Å) Ecoh (eV) VIP (eV)
LDA 3.799 0.331 4.498
PBE 3.959 0.238 4.180
KBM 3.930 0.251 4.092
Exp. 3.924 0.260 4.061
K (bulk) a (Å) Ecoh (eV) B (GPa)
LDA 5.157 1.03 4.41
PBE 5.264 0.87 3.39
KBM 5.219 0.92 3.83
Exp. 5.212 0.94 3.70
 A. Aguado / Computational and Theoretical Chemistry 1021 (2013) 135–143
gas (formed by the valence electrons) to its own fluctuations,
mediated by the correlations that exist between charge density
fluctuations localized at distant (non-overlapped) regions. This
interaction, which is carried out by plasmons, is a non-local
many-body correlation effect which cannot be described by any
semilocal correlation functional. In the work of Dion et al. [54], dis-
persion is defined through a many-body treatment of the electron
gas involving the response function of the metal. The functional of
Dion et al. [54] is an explicit non-local functional of the electron
density, of the following form:
Z Z
1 on each atom was smaller than 0.005 eV/Å.
EvdW
nl ¼ dr1 dr2 nðr1 Þ/ðr1 ; r2 Þnðr2 Þ;
2
where n(r) is the valence electron density. The kernel / satisfies two
important properties: (1) it correctly tends to zero for homogeneous
densities, so the interaction occurs between density fluctuations
and (2) at long distances, it reproduces the correct asymptotic
dependence of the dispersion energy, proportional to 1=r 612 . Notice
that the functional depends exclusively on electronic variables,
and in particular r12 does not correspond to an interatomic distance
but to the distance between two localized charge density fluctua-
tions. The accuracy of this method for metallic solids has been re-
cently demonstrated [68].
Just to complete our description, we notice that the functional
form of Dion et al. [54] is appropriate for the electron gas of a me-
tal. The group of Perdew has proposed an alternative point of view
to describe the physics of dispersion interactions in metals [67],
which is more appropriate to the limit of a Wigner crystal, where
electrons localize at lattice sites and are separated by regions of
very low density. Within this framework, the dispersion interac-
tion in solid potassium would be considered to arise between
K+–K+ ion pairs, and would be screened (as compared to its value
in free space) by the valence electron density. This screening effect
results in an effective interaction between ions which is calculated
via second-order time-dependent perturbation theory and involves
once more plasmonic excitations, somehow connecting this point
of view to that of Dion et al. The difference in philosophy between
both methods is clear. In the method of Perdew, the dispersion en-
ergy correction is written in a more ‘‘classical’’ way as
EvdW
nl ¼ C 6 =R6ij , where Rij is an interatomic distance, and the
many-body screening effect of the valence electron gas is included
only in the calculation of the C6 coefficient. Both methodologies,
however, provide results of similar accuracy for the alkali metals.
In this work, we have followed the functional proposed by Dion
et al. [54].
For each cluster size, we need to generate first a number of rea-
sonable trial structures, which are then introduced as initial geom-
etries for the first-principles unconstrained relaxations. To this
end, we have employed unbiased basin hopping optimizations
[69] based on a many-body Gupta empirical potential [70–72].
The appropriate parameters for potassium are given by Li et al.
[73]. From the pool of structures thus generated, we have selected
13(C2 ) 14(C2 ) 16(D3h )
20(C3) 21(C2v ) 24(C2v )
Fig. 1. Putative GM structures for KN clusters with N = 13–34 atoms. The size and approximate point group symmetry are provided below each cluster. Several isomers are
shown (for example, for N = 16) when they are nearly degenerate (total energy difference smaller than 0.01 eV).
137
up to 100 (depending on cluster size) varied structural motifs for
subsequent optimization at the DFT level. The selection is based
on several structural indicators as detailed in our previous work
[27], which aims at maintaining a high structural diversity within
the considered sample. DFT optimizations were first performed
with the LDA functional because of its lower computational ex-
pense. Finally, we re-optimized at the vdW-DFT level all the iso-
mers which, according to the LDA calculations, have total
energies up to 0.4 eV above the putative GM. All the equilibrium
geometries shown in the next section were obtained through
unconstrained conjugate gradients optimization, until the force
ð1Þ
3. Results and discussion
3.1. Global minimum structures
Fig. 1 shows a representative sample of the GM structures lo-
cated in the size range N = 13–34. For N = 13–15 we obtain oblate
deltahedral structures based on the packing of pentagonal bipyra-
mid units. In the size range N = 16–18, the clusters have a single
endohedral atom enclosed inside a high-symmetry polyhedral
shell. The shell of K16 is the so-called Z15 Kasper polyhedron
[74], containing three disclination lines, although an isomer based
on a defective icosahedron is almost equally stable. Similarly, K17
and the C2v isomer of K18 contain four and five disclination lines,
respectively. The Cs isomer of K18 is instead built by adding one
atom to the structure of K17. Another structure with a single central
atom is the C3 isomer of K20, which was also obtained in a recent
study of sodium clusters [29]. From N = 19 to N = 34 the structures
are dominated by poly-icosahedral (pIh) packing. For example, the
structures of K19–K23 are based on the double icosahedron, a triple
icosahedron is obtained for K24, while one of the degenerate struc-
tures of K27 contains four K13 icosahedral units. Clusters with
N = 28–33 atoms are obtained by simply removing atoms from
the perfect core–shell pancake structure of K34. This structure is
formed by interpenetrating seven K13 icosahedral units. The seven
core atoms are in a pentagonal bipyramid arrangement, which re-
sults in a high D5h point group symmetry.
Comparing our results for N = 13–20 with those of Abdalla et al.
[53] we find a qualitative agreement for N = 13–16 and N = 20.
These authors also locate deltahedral structures for N = 13–16,
although not exactly coinciding with our predictions. For K20, the
GM of Abdalla and coworkers coincides with our C3 isomer. How-
ever, for N = 16–19 their structures are quite different from ours,
and of much lower symmetry. For all clusters with N = 13–20, we
have checked that Abdalla’s structures are at least local minima
at the level of DFT-KBM calculations. The differences may originate
from the different levels of theory employed and from the approx-
imations inherent to the parametrization of their tight-binding
model. A better agreement is found between our results and those
reported by Banerjee et al. [52], who considered just clusters with
16(C1) 17(D2) 18(Cs) 18(C2v ) 20(Cs)
26(Cs) 27(C2v ) 27(C2v ) 30(Cs) 34(D5h )
138 A. Aguado / Computational and Theoretical Chemistry 1021 (2013) 135–143
35(C3v ) 37(Cs) 38(C3v )
44(C2v ) 46(Cs) 49(C1)
Fig. 2. A representative selection of the putative GM structures for clusters with N = 35–55 atoms. The rest of the caption is the same as in Fig. 1.
an even number of atoms but a fully ab initio technique as in the
present work. We locate the same GM structure for N = 14 and
18. For K20, our structure coincides with one of the nearly degener-
ate structures of Banerjee. For K16, however, Banerjee et al. failed to
consider the Kasper polyhedron, so a complete comparison of the
two calculation levels is not possible.
Fig. 2 shows relevant structures in the size range N = 35–55.
Similarly to K34, K35 also consists of seven K13 icosahedral units,
resulting in another perfect core–shell arrangement, but the core
atoms adopt a different C3v symmetry. Poly-icosahedral packing
continues to dominate the structures up to N = 41, and most struc-
tures in this size range can indeed be obtained by removing atoms
from a 45-atom perfect anti-Mackay icosahedron, a structure ob-
tained by covering a 13-atom icosahedron with a complete anti-
Mackay overlayer. The only exception to this trend occurs for
K39, which has the same 7-atom core as K35 and is obtained from
it by incorporating the four additional atoms into a crowded shell,
which lowers the global symmetry from C3v to C3.
We observe a structural transition at N = 42, a critical size be-
yond which the clusters cannot anymore accomodate the strain
inherent to pIh structures. Nevertheless, the transition does not in-
volve just a change from anti-Mackay to Mackay overlayers. K42,
for example, has eight core atoms in a hexagonal bipyramid
arrangement, and although the shell is quite amorphous, an
approximate disclination line can be easily distinguished in
Fig. 2. The local environment of some of the core atoms continues
56(C2v ) 57(Cs) 57(C2v )
61(C2v ) 61(C1) 64(C1 )
74(D3d ) 75(C2) 76(C1 )
Fig. 3. A representative selection of the putative GM structures for clusters with N = 56–80 atoms. The rest of the caption is the same as in Fig. 1.
38(Cs) 39(C3 ) 40(C5v ) 42(C1)
50(C1 ) 52(Cs) 52(Cs) 55(Ih )
to be icosahedral, so pIh packing still prevails in some parts of the
cluster. K44 has a 9-atom core and a respectable C2v symmetry. An
analysis of the local environment of those inner atoms reveals that
the cluster contains a non-trivial combination of Ih13 and Z14 and
Z15 Kasper polyhedral units [74]. This type of structures persists
up to N = 52, where a transition towards Mackay icosahedral struc-
tures occurs which is maintained until the Mackay icosahedron is
completed at N = 55. In the size range N = 44–52 the number of
core atoms steadily increases from nine to eleven, and the struc-
tures become more and more prolate. The transition at N = 52 is
thus accompanied by a significant change in the cluster shape to-
wards nearly spherical structures.
The putative GM structures in the size range N = 56–80 are
shown in Fig. 3. Immediately after the icosahedral shell closing at
N = 55, potassium clusters avoid to place the first few additional
atoms on a new, more external, atomic shell. For K56–K58, those
atoms are admitted into the surface layer of K55 through the gen-
eration of a number of rosette-like defects [75], which consist of
6-fold vertices that substitute the 5-fold vertices of the original ico-
sahedron. For K57 only, there is a nearly degenerate structure with
two adatoms occupying neighboring AM sites. K59 has two degen-
erate structures: theC6v isomer is a perfect core–shell structure
with 15 internal atoms. Its core is the Z14 Kasper polyhedron with
D6d symmetry, but the growth of the shell follows a different pat-
tern (Mackay-like and anti-Mackay-like, respectively) on both
sides of the disclination line, lowering the global symmetry to
58(C2 ) 59(C3v ) 59(C6v ) 60(C1 )
66(C1 ) 67(C1) 71(C5) 73(Cs)
77(C2 ) 78(C1 ) 80(D3 )
 A. Aguado / Computational and Theoretical Chemistry 1021 (2013) 135–143 139

C6v. The C3v structure contains 14 internal atoms and is obtained by
adding three adatoms on top of a C3v isomer of K56. Similarly, most
clusters with N = 60–62 show adatoms on top of K56, which is thus
a more favorable seed for cluster growth than the Mackay
icosahedron. The C2v isomer of K61 is, however, an exception to this
rule.
A structural transition, similar to the one that occurs at N = 42,
is observed at N = 63 and extends up to N = 66. The structures in
this size range contain 17 or 18 internal atoms and a Z15 Kasper
polyhedral unit at their cores. The 6-fold disclination lines can be
clearly appreciated in Fig. 3. A new transition occurs at N = 67
and persists up to N = 72. In fact, all the structures in this size range
can be obtained by removing atoms from (or adding one atom to)
the perfect core–shell arrangement of K71. The packing in this clus-
ter is of an interesting mixed type: the core, with D5h symmetry,
contains 19 atoms in a double icosahedral arrangement, so it
shows pIh packing; the shell tries to provide a Mackay-like over-
layer on top of the pIh core, which is not exactly possible without
generating a high strain in the bonds along the equatorial region.
This strain and the resulting geometrical frustration are relaxed
by twisting the two sides of the Mackay shell by opposite angles
about the 5-fold axis, which lowers the global symmetry to C5.
We would like to emphasize that the structure of K71 is different
from the one that would be obtained by interpenetrating two
Ih55 units (a higher order version of pIh packing), as in that case
the 19-atom core would not show pIh packing.
Our calculations predict that this type of mixed packing domi-
nates the structures of the remaining sizes up to N = 80. All these
clusters attempt to build a Mackay-like overlayer on top of a more
compactly packed core. Perfect core–shell arrangements, which
can be considered geometrical shell closings, are obtained for
K74, K77 and K80. The core in K74 contains 20 atoms and is based
on two fused and distorted Ih13 units. The two K13 icosahedra share
only six atoms, as opposed to seven in a double icosahedron,
resulting in a slightly lower D3d symmetry. For this size in particu-
lar, the core symmetry is fully respected by the shell, which means
that the structure of K74 is just a bigger size version of the structure
of the K20 core, or in other words, that it consists of two Ih55 units,
interpenetrated in exactly the same way as the two Ih13 units of
detailed investigation is warranted for each individual material if
accurate results are to be obtained.
The effect of van der Waals interactions can be assessed by
comparing the results obtained at the vdW-DFT level of theory
(shown in Figs. 1–3) with those obtained under the LDA and GGA
approximations (not shown explicitly). Qualitatively, the main ef-
fect is the same as that reported for cesium clusters [55], namely
a higher stabilization of more compact structures. For example,
the GM structure of Cs15 is a Z14 Kasper polyhedron with one cen-
tral atom [55], while Na15 adopts a less compact deltahedral struc-
ture without internal atoms. For K15, the deltahedral structure is
0.1 eV more stable than the Kasper polyhedron at the LDA or
GGA levels, but only 0.06 eV more stable at the vdW-DFT level. This
is a general observation for all sizes. In some cases, this extra sta-
bilization is enough to invert the energetic order of isomers ob-
tained at the LDA or GGA levels. One example is K16, whose
structure coincides with Na16 if vdW effects are neglected but with
Cs16 when vdW effects are included. Considering that the bench-
mark results in Table 1 demonstrate that the KBM functional pro-
vides a more accurate description of potassium properties, we
believe that the KBM predictions on cluster structures are also
more reliable. vdW effects are therefore not negligible for potas-
sium clusters, and should be included in calculations aimed at
obtaining high quality results.
3.2. Electronic properties
In order to identify the main electronic shell closings, we show
in Fig. 4 the energy gap between the highest occupied and lowest
unoccupied molecular orbitals (HOMO–LUMO gap), the vertical
ionization potential (VIP) and the vertical electron affinity (VEA)
of potassium clusters. We observe a substantial odd–even alterna-
tion, mostly in the HOMO–LUMO gaps, which gradually dissap-
pears as the cluster size increases. In fact, the HOMO–LUMO gaps
of clusters with more than 70 atoms are already very small and
0.5 14 20
18
εHOMO-LUMO (eV)

K20. The core of K77 has 21 atoms but otherwise is quite similar 0.4
to the core of K74. K78 and K79 have 22 atoms in a low-symmetry 30 34 40 58
0.3
core. Finally, the core of K80 contains 23 atoms and shows perfect
46 50
pIh packing of three Ih13 units. The symmetry of the core is there- 0.2 60 64
fore D3h. Similarly to the situation discussed for K71, the growth of 0.1
a Mackay-like overlayer involves some geometrical frustration that
0
lowers the global symmetry down to D3. Also, for the same reasons 3.5 18
stated for K71, this cluster is not obtained by interpenetrating three 3.4 14
Ih55 units (moreover, such a structure would contain 24 core atoms
3.3 20
VIP (eV)

instead of 23). 34 40 58
It is interesting to compare the structures of KN clusters with 3.2 26 30 46 50
those previously reported by us for clusters of other alkali metals 3.1 60 64
[27,29,55]. In those studies, sodium and cesium clusters were 3
found to adopt very different structures for most sizes, despite
both being alkali metals. Traditionally, the structures of potassium
and sodium clusters have been assumed to be akin to each other 1.6
VEA (eV)

(see, for example, [52]). This assumption is based on the similari- 1.4 60 64
46 50 58
ties observed in the abundance mass spectra, ionization potentials
and other properties of these two alkali clusters. A detailed com- 1.2 34 40
30
parison of our ab initio results confirms that KN structures coincide 24
1
with NaN structures much more frequently than with CsN struc- 14 20
0.8
tures. Exceptions occur only for N = 16, 39, 45, 48, 49, 52, 76, 78, 10 20 30 40 50 60 70 80
where the GM structure of KN coincides with that of CsN; also, for N
N = 37, 42–44, 58, 61, 63, 66, 67–69, 79, the GM structures here re-
Fig. 4. HOMO–LUMO gaps (upper graph), vertical ionization energies (middle
ported are specific to potassium, as they do not coincide with those graph) and vertical electron affinities (lower graph) of Kn clusters. The main
of sodium or cesium clusters. All in all, we conclude that, due to the electron shell closings (sizes preceeding a large drop in the HOMO–LUMO and/or
huge complexity of the potential energy landscape of clusters, a VIP values, or marked minima in the VEA values) are explicitly indicated.
140 A. Aguado / Computational and Theoretical Chemistry 1021 (2013) 135–143
the clusters seem to be approaching the gap closure characteristic
of a metallic state. The odd–even alternation of electronic proper-
ties occurs on top of an average trend which coincides with the
predictions of classical charged sphere models.
Large drops in the VIP curve are associated with the opening of
an electron shell. Electronically magic clusters will also have a
large HOMO–LUMO gap and will display a local minimum in the
VEA curve. According to these criteria, the cluster sizes with a par-
ticularly stable electron shell structure are N = 14, 18, 20, 26, 30,
34, 40, 46, 50 and 58. Some of these numbers (N = 18 and 26, for
example) do not show up as marked minima in the VEA curve,
which instead shows a local minimum for K24. After N = 58 the
curves are smoother although the stabilities of K60 and K64 are also
somewhat enhanced as compared to neighboring sizes. No feature
is detected at N = 68 or N = 70 even though these are electron shell
closings according to some jellium models.
All the maxima observed in the experimental photoabsorption
spectra [33,36,38], as well as the absence of specific features for
N = 68, 70, are reproduced by our calculations. We predict addi-
tional maxima in the VIP (N = 24, 46, 50, 64, 74) which are not ob-
served in the experiments but are predicted by the liquid drop
model with shell corrections devised by Yannouleas et al. [39].
As explained by these authors, the weaker nature of these local
maxima is expected to be detected only in experiments performed
at very low temperatures. In this connection, we point out that, gi-
ven the estimated melting points of potassium clusters [76], the re-
ported experimental VIP values most probably refer to liquid-like
clusters.
Fig. 5 shows the size dependence of the cluster shape, as quan-
tified by the Hill–Wheeler parameters [2,77]. b measures the de-
gree of quadrupolar deviation from sphericity, with a value b = 0
indicating a perfectly spherical cluster; c determines the ‘‘sense’’
of that deviation, i.e. whether the shape is prolate or oblate. c = 0
corresponds to an axially symmetric prolate aggregate (cigar-like),
0.4
13
25-26
0.3
47-48
0.2
β
16 45
0.1 80
66
20 39-40
0 55-56
Oblate
1
0.5
γ
0 5
Prolate
10 20 30 40 50 60 70 80
N
Fig. 5. Hill–Wheeler deformation parameters b (top panel) and c (bottom panel) of
potassium clusters. The dashed red lines indicate those c values corresponding to
perfectly oblate, triaxial or oblate shapes. (For interpretation of the references to
color in this figure legend, the reader is referred to the web version of this article.)
while for c = p/3 we have an oblate axially symmetric shape (pan-
cake). For intermediate c values, cluster deformation is triaxial.
The overall trend for b is to decrease, indicating a tendency for
bigger clusters to adopt more spherical shapes. On top of that aver-
age trend, a series of oscillations between spherical and non-spher-
ical shapes is observed in Fig. 5, which is in fair general agreement
with the predictions of jellium models although geometric shell
closings introduce some additional features. For example, clusters
with about 20, 40 and 58 electrons are more spherical than the
rest, while clusters in between two spherical shell closings (for
example, K13, K26 or K48) are the least spherical ones. However,
K39 is slightly more spherical than K40 due to a geometric shell
closing. Similarly, the perfect icosahedron at N = 55 is more spher-
ical than K58. We do not observe any enhanced sphericity at N = 68,
70 even if these sizes correspond to electron shell closings in some
jellium models [2,3]. These observations suggest that geometric
packing arguments dominate the structures of the biggest clusters
here considered.
The prolate/oblate character of the quadrupolar deformation is
also in general agreement with the structureless electron shell
models. In general terms, the structures tend to have prolate shape
deformations immediately after an electron shell closing, and ob-
late shapes when approaching an electron shell closing from below
(see also [26]). Nevertheless, for clusters with more than 55 atoms
the presence of geometric shell closings causes several exceptions
to such general rule. We conclude that, although the jellium model
continues to be a valuable guide for the cluster shapes even at zero
Kelvin, conclusions based on this model should not be overempha-
sized. An ab initio description, including the ionic structures explic-
itly, is required in order to obtain an accurate description of alkali
cluster structures.
A resurgence of interest in the screening properties of small al-
kali particles comes from recent beam deflection experiments per-
formed by Bowlan et al. [78]. In that work, the electric dipole
moments of small cold Nan clusters were measured and found to
be essentially zero (60.002 Debye/atom for n 6 20). This is a distin-
guishing feature of metallicity and apparently implies that the per-
fect screening properties of alkali metals are maintained down to
the nanoscale region. We have recently shown that an exhaustive
search for the GM structures employing ab initio methods can
approximately reproduce the experimental observations about
the electric dipole moments of sodium clusters [29], and predict
similar results for cesium clusters [55]. Nevertheless, a detailed
comparison of the results obtained at different levels of theory
[30], including several exchange–correlation functionals within
DFT and a variety of quantum-chemical methods, suggests that
the theoretical dipoles obtained from the most accurate methods
are significant up to an error of about 10 mDebye/atom. The results
of Fig. 6 show that Bowlan’s expectations can be also transferred to
15
μ (mDebye/atom)
10
0
10 20 30 40 50 60 70 80
N
Fig. 6. Electron dipole moments of Kn clusters. In cases of near-degeneracy, the
different isomers typically have very similar l values.
 A. Aguado / Computational and Theoretical Chemistry 1021 (2013) 135–143 141

potassium clusters, confirming the generality of the result, at least
for the family of alkali clusters. Nearly all the dipole moments of KN
clusters are below 10 mDebye/atom, and our recent work [29] sug-
gests that these values will be even smaller when zero-point vibra-
tional effects are taken into account. There are only three clusters
(K26, K47 and K49) with a more sizable dipole moment. If the find-
ings of Bowlan et al. were completely general, it might imply that
we have not located the correct GM structures for these sizes.
Otherwise, they may just represent exceptions to the ‘‘zero dipole’’
rule.
Finally, we mention here that most potassium clusters have the
lowest spin multiplicity, i.e. a singlet for clusters with an even
number of atoms and a doublet for clusters with an odd number
of atoms. In this sense KN clusters are also more similar to sodium
than to cesium clusters [55]. Exceptions to this rule occur for some
sizes where geometric packing effects dominate the structure: K54
is a spin quintet, K55 and K71 are spin quartets, and K16 and K56 are
spin triplets.
3.3. Cluster stability
Fig. 7 shows three different stability measures: the cohesive en-
ergy (or binding energy per atom) Ecoh ðNÞ ¼ E1  ENN , the evapora-
tion energy Eevap(N) = (E1 + EN1)  EN, and the second energy
difference D2(N) = EN1 + EN+1  2EN = Eevap(N)  Eevap(N + 1). While
the cohesive energy quantifies the total internal energy content of
a cluster and is therefore a measure of its global (or absolute) sta-
bility, Eevap and D2 provide ‘‘local’’ stability measures by comparing
the energy of a cluster of size N to that of clusters with neighboring
sizes, and so they are more suitable quantities to interpret the re-
sults of abundance mass spectra.
Cohesive energies are referenced to the results of a four param-
eter, liquid drop like, smooth fit. This particular representation is
convenient to highlight the oscillatory pattern of the cohesive
energies about the average trend. Clusters with 18–20, 34–40
20
0.01 18
Ecoh-Efit (eV)
and 55–60 atoms clearly stand on top of the average behavior as
regards absolute stability. These ranges are in qualitative agree-
ment with the main spherical electron shell closings.
The three indicators essentially predict the same set of magic
numbers, which occur for N = 14, 18, 20, 26, 30, 34, 40, 46, 50,
55, 58, 71, 74, 77 and 80, with some secondary magic numbers also
noticed in the figure. For clusters with up to 58 atoms, nearly all
the magic numbers have an electronic origin, as they coincide with
the predictions of spheroidal Nilsson model [38] while their struc-
tures (see Section 3.1) do not correspond in general to geometric
shell closings. An exception for small sizes is K34, a geometric shell
closing with high D5h symmetry, whose enhanced stability may
thus have an important contribution coming from geometric
packing.
In the size range N P 55, we notice that K55 is a more prominent
magic number than K58 based on the three different indicators
shown in Fig. 7. The enhanced stability of K55 is clearly associated
to the icosahedral shell closing, as no specific feature is observed in
the electronic properties for this size (see Section 3.2). The same is
true of other prominent magic numbers observed in this size range,
i.e. N = 71, 74, 77, 80. In all these cases, the structures present a
perfect core–shell arrangement, with a complete Mackay overlayer
covering a compact core, so they can be identified as geometric
shell closings. The experimental abundance spectra do not identify
significant features at these sizes, suggesting that the cluster beam
is made up of liquid-like clusters and that the experiments are
blind to the geometric magic numbers typical of cold, solid clus-
ters. These features are expected to be recovered in experiments
that determine the dissociation energies of solid clusters by
accounting for the latent heat of melting, as exemplified by exper-
iments on aluminum clusters [79–81]. In summary, we conclude
that electron shells become of secondary importance as compared
to geometric shells in determining the relative stabilities of solid
potassium clusters with more than N  55–58 atoms.
It may be interesting to explicitly comment about the accuracy
of the structures derived with the Gupta potential. It is well known
that this type of semiempirical potentials cannot be made fully
transferable to different atomic environments, and most of the
available parametrizations have been fitted to bulk properties. This

55 58 observation alone casts some reasonable doubts about the ex-

0.005 34 60
40 64 pected accuracy of the Gupta potential for clusters. But even more
14
7174
26 30 77 important, at least for small alkali clusters, is the observation that
0 46 50
electronic shell effects are so important that the cluster shapes fol-
-0.005 low very well the jellium model predictions (see Fig. 5). Obviously,
1 the electron shell effects typical of superatoms are not incorpo-
55 rated into the Gupta potential, which is an explicit function of
0.9 atomic coordinates only. Therefore, it is no surprise that the DFT-
18 34 38
Eevap (eV)

20 26 30 58 7780
0.8 14 44 46 50 64 7174 derived GM structures do not coincide with the Gupta-derived
40
GM structures, except for a few sizes such as N = 55, where the ico-
0.7
sahedral geometric shell closing is well described by the Gupta
0.6 model. Nevertheless, as we emphasized in Section 2, the Gupta po-
tential is a very useful external tool to generate a diversity of plau-
0.3
20 55 sible structural motifs, many of which would be overlooked if
14 34
0.2 using just some sort of intuition or educated guess. This is why
18 40
26 30 46
50 58 71 74 we employ structural indicators rather than just Gupta energies
Δ2 (eV)

0.1 64 77
in order to choose the structures that are re-optimized at the
0 DFT level. A more complete discussion about these issues is to be
-0.1 found in our previous work [27].
-0.2
10 20 30 40 50 60 70 80
N 4. Conclusions

Fig. 7. Stability measures for Kn clusters: the upper graph shows the cohesive We have reported the GM structures, electronic properties and
energies, referenced to Efit, which is a four-parameter fit of the form Efit = A0 + A1N1/
3
+ A2N2/3 + A3N; the middle graph shows the evaporation energies and the lower
stabilities of potassium clusters containing from 13 to 80 atoms,
graph shows the second-energy differences. Clusters with enhanced stability obtained employing a first-principles DFT method that incorpo-
(magic sizes) are shown for each of the stability indicators. rates a description of van der Waals interactions. KN clusters adopt
142 A. Aguado / Computational and Theoretical Chemistry 1021 (2013) 135–143
a variety of structural motifs in different size ranges: deltahedral
and poly-icosahedral packing dominates for N 6 41; Mackay icosa-
hedra are the GM for N = 52–55; structures containing Kasper poly-
hedral units and disclination lines have the lowest energy for
N = 16–18, 42–51, 63–66; for N = 56–62 the structures contain ro-
sette-like defects on the outermost atomic shell; finally, for N P 67
we observe a mixed packing pattern in which a distorted Mackay-
like overlayer is grown around a poly-icosahedral core. Our results
improve over previous theoretical determinations of global min-
ima for several sizes with N 6 20. A comparison of the structures
of potassium clusters with those previously obtained for sodium
and cesium clusters demonstrates that KN clusters share more sim-
ilarities with NaN than with CsN clusters on average, but at the
same time there are numerous exceptions to this simple rule
which justify a separate and dedicated study of potassium clusters.
The superior accuracy of the vdW-DFT functional as compared to
usual LDA and GGA approximations has been demonstrated
through benchmark calculations on the dimer and bulk. Inclusion
of vdW interactions has been found to modify the GM structure
of several KN clusters. Therefore, we have concluded that vdW
interactions should be included in accurate studies of potassium
clusters.
The identified electronic shell closings are in very good agree-
ment with the experimentally determined photoionization spectra.
For clusters with more than 58 atoms, the size dependence of the
electronic properties becomes substantially smoother and feature-
less. The cluster shapes are also in good agreement with those pre-
dicted by spheroidal jellium models, although the detailed atomic
structure (which is neglected in jellium models) is found to intro-
duce non-negligible modifications on top of the average jellium
trends. With few exceptions, KN clusters are found to adopt the
lowest possible spin multiplicity. They also possess negligible elec-
tric dipole moments, as was also found for sodium and cesium
clusters. Metallic-like screening seems thus to be a rather general
property of small alkali clusters.
We have also reported the relative stabilities and magic num-
bers of KN clusters. The theoretical results reproduce all the magic
numbers observed in experimental abundance spectra and which
coincide with electronic shell closings, but additionally predict
new magic numbers which are associated with geometric shell
closings. The electron shell structure is found to dominate the sta-
bilities of clusters with less than about 55–58 atoms, while geo-
metric packing effects explain the magic numbers for bigger
clusters. Theoretical calculations thus become an ideal comple-
ment to the experimental mass spectra which typically miss the
geometric shell closings if the cluster beam contains hot, liquid-
like, clusters.
Acknowledgements
I gratefully acknowledge the support of the Spanish ‘‘Ministerio
de Ciencia e Innovación’’, the European Regional Development
Fund and ‘‘Junta de Castilla y León’’ (Project Nos. FIS2011-22957
and VA104A11-2).
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.comptc.2013.
06.040.
References
[1] W.D. Knight, K. Clemenger, W.A. de Heer, W.A. Saunders, M.Y. Chou, M.L.
Cohen, Electronic shell structure and abundances of sodium clusters, Phys. Rev.
Lett. 52 (1984) 2141.
[2] W.A. de Heer, The physics of simple metal clusters: experimental aspects and
simple models, Rev. Mod. Phys. 65 (1993) 611.
[3] M. Brack, The physics of simple metal clusters: self-consistent jellium model
and semiclassical approaches, Rev. Mod. Phys. 65 (1993) 677.
[4] V. Bonacić-Koutecký, P. Fantucci, J. Koutecký, Quantum chemistry of small
clusters of elements of groups Ia, Ib, and IIa: fundamental concepts,
predictions, and interpretation of experiments, Chem. Rev. 91 (1991) 1035.
[5] U. Röthlisberger, W. Andreoni, Structural and electronic properties of sodium
microclusters (n = 2–20) at low and high temperatures: new insights from ab
initio molecular dynamics studies, J. Chem. Phys. 94 (1991) 8129.
[6] U. Röthlisberger, W. Andreoni, P. Giannozzi, Thirteen-atom clusters:
equilibrium geometries, structural transformations, and trends in Na, Mg, Al,
and Si, J. Chem. Phys. 96 (1992) 1248.
[7] F. Spiegelman, R. Poteau, B. Montag, P.-G. Reinhard, Global structure of small
Na clusters in different approaches, Phys. Lett. A 242 (1998) 163.
[8] L. Kronik, I. Vasiliev, M. Jain, J.R. Chelikowsky, Ab initio structures and
polarizabilities of sodium clusters, J. Chem. Phys. 115 (2001) 4322.
[9] M. Schmidt, H. Haberland, Phase transitions in clusters, C.R. Phys. 3 (2002) 327.
[10] I.A. Solov’yov, A.V. Solov’yov, W. Greiner, Structure and properties of small
sodium clusters, Phys. Rev. A 65 (2002) 053203.
[11] M. Moseler, B. Huber, H. Häkkinen, U. Landman, G. Wrigge, M.A. Hoffmann, B.
von Issendorff, Thermal effects in the photoelectron spectra of Na N clusters
(N = 4–19), Phys. Rev. B 68 (2003) 165413.
[12] K.R.S. Chandrakumar, T.K. Ghanty, S.K. Ghosh, Static dipole polarizability and
binding energy of sodium clusters Nan (n = 1–10): a critical assessment of all
electron based post Hartree–Fock and density functional methods, J. Chem.
Phys. 120 (2004) 6487.
[13] A. Aguado, Competing thermal activation mechanisms in the melting-like
transition of NaN (N = 135–147) clusters, J. Phys. Chem. B 109 (2005) 13043.
[14] M.S. Lee, S. Chacko, D.G. Kanhere, First-principles investigation of finite-
temperature behavior in small sodium clusters, J. Chem. Phys. 123 (2005)
164310.
[15] F. Baletto, R. Ferrando, Structural properties of nanoclusters: energetic,
thermodynamic and kinetic effects, Rev. Mod. Phys. 77 (2005) 371.
[16] A. Aguado, J.M. López, Small sodium clusters that melt gradually: melting
mechanisms in Na30, Phys. Rev. B 74 (2006) 115403.
[17] M.S. Lee, D.G. Kanhere, Effects of geometric and electronic structure on the
finite temperature behavior of Na58, Na57, and Na55 clusters, Phys. Rev. B 75
(2007) 125427.
[18] S. Zorriasatein, M.S. Lee, D.G. Kanhere, Electronic structures, equilibrium
geometries, and finite-temperature properties of Nan (n = 39–55) from first
principles, Phys. Rev. B 76 (2007) 165414.
[19] E.G. Noya, J.P.K. Doye, D.J. Wales, A. Aguado, Geometric magic numbers of
sodium clusters: interpretation of the melting behaviour, Eur. Phys. J.D 43
(2007) 57.
[20] O. Kostko, B. Huber, M. Moseler, B. von Issendorff, Structure determination of
medium-sized sodium clusters, Phys. Rev. Lett. 98 (2007) 043401.
[21] S.M. Ghazi, M.S. Lee, D.G. Kanhere, The effects of electronic structure and
charged state on thermodynamic properties: An ab initio molecular dynamics
investigation on neutral and charged clusters of Na39, Na40 and Na41, J. Chem.
Phys. 128 (2008) 104701.
[22] S. Núñez, J.M. López, A. Aguado, Atomic layering and related postmelting
effects in small liquid metal clusters, Phys. Rev. B 79 (2009) 165429.
[23] C. Hock, C. Bartels, S. Strassburg, M. Schmidt, H. Haberland, B. von Issendorff, A.
Aguado, Premelting and postmelting in clusters, Phys. Rev. Lett. 102 (2009)
043401.
[24] M. Itoh, V. Kumar, T. Adschiri, Y. Kawazoe, Comprehensive study of sodium,
copper, and silver clusters over a wide range of sizes 2 6 N 6 75, J. Chem. Phys.
131 (2009) 174510.
[25] S.M. Ghazi, S. Zorriasatein, D.G. Kanhere, Building clusters atom-by-atom:
from local order to global order, J. Phys. Chem. A 113 (2009) 2659.
[26] B. Huber, M. Moseler, O. Kostko, B. von Issendorff, Structural evolution of the
sodium cluster anions Na 
20  Na57 , Phys. Rev. B 80 (2009) 235425.
[27] A. Aguado, O. Kostko, First-principles determination of the structure of NaN
and NaN clusters with up to 80 atoms, J. Chem. Phys. 134 (2011) 164304.
[28] A. Aguado, Stepwise melting in Naþ 41 : a first-principles critical analysis of
available experimental results, J. Phys. Chem. C 115 (2011) 13180.
[29] A. Aguado, A. Vega, L.C. Balbás, Structural and zero-point vibrational effects on
the electric dipole moments and static dipole polarizabilities of sodium
clusters, Phys. Rev. B 84 (2011) 165450.
[30] A. Aguado, A. Largo, A. Vega, L.C. Balbás, On the electric dipole moments of
small sodium clusters from different theoretical approaches, Chem. Phys. 399
(2012) 252.
[31] A. Aguado, M.F. Jarrold, Melting and freezing of metal clusters, Annu. Rev.
Phys. Chem. 62 (2011) 151.
[32] W.D. Knight, W.A. de Heer, K. Clemenger, W.A. Saunders, Electronic shell
structure in potassium clusters, Solid State. Commun. 53 (1985) 445.
[33] W.A. Saunders, K. Clemenger, W.A. de Heer, W.D. Knight, Photoionization and
shell structure of potassium clusters, Phys. Rev. B 32 (1985) 1366.
[34] M.M. Kappes, M. Schär, P. Radi, E. Schumacher, On the manifestation of
electronic structure effects in metal clusters, J. Chem. Phys. 84 (1986) 1863.
[35] W.D. Knight, W.A. de Heer, W.A. Saunders, K. Clemenger, M.Y. Chou, M.L.
Cohen, Alkali metal clusters and the jellium model, Chem. Phys. Lett. 134
(1987) 1.
[36] M.L. Cohen, M.Y. Chou, W.D. Knight, W.A. de Heer, Physics of metal clusters, J.
Phys. Chem. 91 (1987) 3141.
 A. Aguado / Computational and Theoretical Chemistry 1021 (2013) 135–143
[37] C. Bréchignac, P. Cahuzac, J.P. Roux, Photoionization of potassium clusters:
neutral and ionic cluster stabilities, J. Chem. Phys. 87 (1987) 229.
[38] W.D. Knight, W.A. de Heer, W.A. Saunders, Shell structure and response
properties of metal clusters, Z. Phys. D 3 (1986) 109.
[39] C. Yannouleas, U. Landman, Electronic entropy, shell structure, and size-
evolutionary patterns of metal clusters, Phys. Rev. Lett. 78 (1997) 1424.
[40] L.A. der Lan, P. Bartl, C. Leidlmair, H. Schöblel, S. Denifl, T.D. Märk, A.M. Ellis, P.
Scheier, Submersion of potassium clusters in helium nanodroplets, Phys. Rev.
B 85 (2012) 115414.
[41] M.Y. Chou, A. Cleland, M.L. Cohen, Total energies, abundances, and electronic
shell structure of lithium, sodium, and potassium clusters, Solid State.
Commun. 52 (1984) 645.
[42] S. Saito, M.L. Cohen, Jellium-model calculation for dimer decays of potassium
clusters, Phys. Rev. B 38 (1988) 1123.
[43] M. Manninen, Structures of small alkali-metal clusters, Phys. Rev. B 34 (1986)
6886.
[44] C. Bréchignac, P. Cahuzac, J.P. Roux, D. Pavolini, F. Spiegelmann, Adiabatic
decomposition of mass-selected alkali clusters, J. Chem. Phys. 87 (1987) 5694.
[45] F. Spiegelmann, D. Pavolini, Ab initio calculations of the electronic structure of
small Nan, Naþ þ
n , Kn, and Kn clusters (n 6 6) including core-valence interaction,
J. Chem. Phys. 89 (1988) 4954.
[46] A.K. Ray, S.D. Altekar, Ab initio many-body perturbation-theoretic study of
small potassium clusters, Phys. Rev. B 42 (1990) 1444.
[47] A. Kornath, R. Ludwig, A. Zoermer, Small potassium clusters, Angew. Chem. Int.
Ed. 37 (1998) 1575.
[48] I.S. Lim, P. Schwerdtfeger, T. Söhnel, H. Stoll, Ground state properties and static
dipole polarizabilities of the alkali dimers from Kn2 to Frn2 ðn ¼ 0; þ1Þ from scalar
relativistic pseudopotential coupled cluster and density functional studies, J.
Chem. Phys. 122 (2005) 134307.
[49] E. Flrez, P. Fuentealba, A theoretical study of alkali metal atomic clusters: from
Lin to Csn (n = 2–8), Int. J. Quantum Chem. 109 (2009) 1080.
[50] L. Padilla-Campos, E. Chávez, Electronic properties of small Kn (n 6 8) and
bimetallic KnCum (n, m 6 4) clusters, J. Mol. Struct. THEOCHEM 958 (2010) 92.
[51] J. Centeno, P. Fuentealba, Big bang methodology applied to atomic clusters, Int.
J. Quantum Chem. 111 (2011) 1419.
[52] A. Banerjee, T.K. Ghanty, A. Chakrabarti, Ab initio studies and properties of
small potassium clusters, J. Phys. Chem. A 112 (2008) 12303.
[53] S. Abdalla, M. Sprinborg, Y. Dong, Isolated and deposited potassium clusters:
energetic and structural properties, Surf. Sci. 608 (2013) 255–264.
[54] M. Dion, H. Rydberg, E. Schröder, D.C. Langreth, B.I. Lundqvist, Van der Waals
density functional for general geometries, Phys. Rev. Lett. 92 (2004) 246401.
[55] A. Aguado, Discovery of magnetic superatoms and assessment of van der
Waals dispersion effects in Csn clusters, J. Phys. Chem. C 116 (2012) 6841.
[56] W. Kohn, L.J. Sham, Self-consistent equations including exchange and
correlation effects, Phys. Rev. 140 (1965) A1133.
[57] P. Hohenberg, W. Kohn, Inhomogeneous electron gas, Phys. Rev. 136 (1964)
B864.
[58] J.M. Soler, E. Artacho, J.D. Gale, A. García, J. Junquera, P. Ordejón, D. Sánchez-
Portal, The SIESTA method for ab initio order-N materials simulation, J. Phys.:
Condens. Matter 14 (2002) 2745.
[59] J.P. Perdew, A. Zunger, Self-interaction correction to density-functional
approximations for many-electron systems, Phys. Rev. B 23 (1981) 5048.
143
[60] D.M. Ceperley, B.J. Alder, Ground state of the electron gas by a stochastic
method, Phys. Rev. Lett. 45 (1980) 566.
[61] J.P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made
simple, Phys. Rev. Lett. 77 (1996) 3865.
[62] J. Klimes, D.R. Bowler, A. Michaelides, Chemical accuracy for the van der Waals
density functional, J. Phys.: Condens. Matter 22 (2010) 022201.
[63] G. Román-Pérez, J.M. Soler, Efficient Implementation of a van der Waals
density functional: application to double-wall carbon nanotubes, Phys. Rev.
Lett. 103 (2009) 096102.
[64] G.I. Csonka, J.P. Perdew, A. Ruzsinszky, P.H.T. Philipsen, S. Lebègue, J. Paier, O.A.
Vydrov, J.G. Ángyán, Assessing the performance of recent density functionals
for bulk solids, Phys. Rev. B 79 (2009) 155107.
[65] R. Hamann, M. Schlüter, C. Chiang, Norm-conserving pseudopotentials, Phys.
Rev. Lett. 43 (1979) 1494.
[66] S.G. Louie, S. Froyen, M.L. Cohen, Nonlinear ionic pseudopotentials in spin-
density-functional calculations, Phys. Rev. B 26 (1982) 1738.
[67] J. Tao, J.P. Perdew, A. Ruzsinszky, Long-range van der Waals attraction and
alkali-metal lattice constants, Phys. Rev. B 81 (2010) 233102.
[68] J. Klimes, D.R. Bowler, A. Michaelides, Van der Waals density functionals
applied to solids, Phys. Rev. B 83 (2011) 195131.
[69] D.J. Wales, J.P.K. Doye, Global optimization by basin-hopping and the lowest
energy structures of Lennard-Jones Clusters containing up to 110 atoms, J.
Phys. Chem. A 101 (1997) 5111.
[70] R.P. Gupta, Lattice relaxation at a metal surface, Phys. Rev. B 23 (1981) 6265.
[71] V. Rosato, M. Guillopé, B. Legrand, Thermodynamical and structural properties
of f.c.c transition metals using a simple tight-binding model, Philos. Mag. A 59
(1988) 321.
[72] F. Cleri, V. Rosato, Tight-binding potentials for transition metals and alloys,
Phys. Rev. B 48 (1993) 22.
[73] Y. Li, E. Blaisten-Barojas, D.A. Papaconstantopoulos, Structure and dynamics of
alkali-metal clusters and fission of highly charged clusters, Phys. Rev. B 57
(1998) 15519.
[74] D.J. Wales, Energy Landscapes, Cambridge University Press, Cambridge, 2003.
[75] E. Aprà, F. Baletto, R. Ferrando, A. Fortunelli, Amorphization mechanism of
icosahedral metal nanoclusters, Phys. Rev. Lett. 93 (2004) 065502.
[76] A. Aguado, Orbital-free molecular dynamics study of melting in K20, K55, K92,
K142, Rb55, and Cs55 clusters, Phys. Rev. B 63 (2001) 115404.
[77] D.L. Hill, J.A. Wheeler, Nuclear constitution and the interpretation of fission
phenomena, Phys. Rev. 89 (1953) 1102.
[78] J. Bowlan, A. Liang, W.A. de Heer, How metallic are small sodium clusters?,
Phys Rev. Lett. 106 (2011) 043401.
[79] A.K. Starace, C.M. Neal, B.P. Cao, M.F. Jarrold, A. Aguado, J.M. López, Correlation
between the latent heats and cohesive energies of metal clusters, J. Chem.
Phys. 129 (2008) 144702.
[80] A.K. Starace, C.M. Neal, B.P. Cao, M.F. Jarrold, A. Aguado, J.M. López, Electronic
effects on melting: comparison of aluminum cluster anions and cations, J.
Chem. Phys. 131 (2009) 044307.
[81] B.P. Cao, A.K. Starace, O.H. Judd, I. Bhattacharyya, M.F. Jarrold, J.M. López, A.
Aguado, Activation of dinitrogen by solid and liquid aluminum nanoclusters: a
combined experimental and theoretical study, J. Am. Chem. Soc. 132 (2010)
12906.