PHYSICAL REVIEW B 83, 165417 (2011)

Effect of edge reconstruction and passivation on zero-energy states and magnetism in triangular
graphene quantum dots with zigzag edges
O. Voznyy,1 A. D. Guclu,1 P. Potasz,1,2 and P. Hawrylak1
1

Institute for Microstructural Sciences, National Research Council of Canada, Ottawa, Canada
2
Institute of Physics, Wroclaw University of Technology, Wroclaw, Poland
(Received 13 October 2010; revised manuscript received 2 March 2011; published 14 April 2011)
We present the results of ab initio calculations of the effect of reconstruction and passivation of zigzag
edges on the electronic and magnetic properties of triangular graphene quantum dots. We find that, similarly
to nanoribbons, hydrogen-passivated ideal zigzag edges are energetically favored over the pentagon-heptagon
zigzag. However, the reconstructed edge is more stable in the absence of hydrogen, thus, delayed passivation
with H may lock the dot in such an unfavorable configuration. Both hydrogen-passivated edge morphologies
lead to a band of states at the Fermi level. Unlike in nanoribbons, this quasidegenerate band results in net spin
polarization for structures with zigzag edge of all sizes studied here. For triangular dots with pentagon-heptagon
zigzag edge, a larger width of the zero-energy band is predicted, leading to the loss of net magnetization.
DOI: 10.1103/PhysRevB.83.165417

PACS number(s): 73.22.Pr, 68.65.Pq, 75.75.c

Graphene, an atomically thick honeycomb lattice of carbon

atoms, exhibits fascinating properties due to the relativisticlike nature of quasiparticle dispersion close to the Fermi
level.15 Graphenes potential for nanoelectronics applications, particularly magnetization of edges, and its use for
spintronics motivated considerable amount of research in
graphene nanoribbons610 and, more recently, graphene quantum dots.1125 Large optical absorptivity, tunable electronic
levels, high charge mobility, and nontoxicity make graphene
nanostructures attractive also for photovoltaic applications.26
In low-dimensional graphene structures, the overall shape
and the character of the edges drastically affect the electronic
properties near the Fermi level.25,2729 In particular, theoretical
models predict that, in triangular graphene quantum dots
(TGQDs) with exclusively zigzag edges, the energy spectrum
near the Fermi level collapses to a band of degenerate states,
isolated from the rest of the spectrum by a well-defined gap,
with states predominantly localized on the edges.1621,2325 It
was shown that, unlike in graphene nanoribbons that have
no net magnetic moment, in this band of degenerate states,
strong electron-electron interactions lead to ferromagnetism
and peculiar magnetic19,21,24,30 and optical22,25 properties,
e.g., magnetic moment proportional to the dot size and
controlled by an external gate. TGQDs might also offer
additional advantages for third-generation solar cells utilizing
the presence of the intermediate band in the gap31 and multiexciton generation (MEG).32 MEG was already demonstrated
in carbon nanotubes,33,34 while a recent theoretical study35
suggests that it is also possible in graphene QDs, and that
localization of states on the edges increases MEG efficiency.
Despite the fact that TGQDs with zigzag edges have
not been demonstrated yet, recent experimental works suggest that they are conceivable in the future. Typically, the
solution-derived organic chemistry methods produce graphene
nanostructures with H-passivated edges with a predominantly
armchair structure.10,36,37 Other techniques, such as etching
lithography or mechanical exfoliation, result in nanostructures with a mixture of zigzag and armchair edges, and
their reconstructed counterparts. The edges morphology is
found to be highly dynamic under nonequilibrium preparation
1098-0121/2011/83(16)/165417(5)

conditions, and interconversion between different types of

edge reconstructions is often observed.29,3842 Techniques
for preparation of graphene nanostructures with controllededge morphology are constantly emerging, e.g., nanotube
unzipping43 and Joule heating,40,44 and some of them, such
as anisotropic etching using Ni (Ref. 46) or Co (Ref. 47)
nanoparticles and carbothermal decomposition of SiO2 ,45
can already produce exclusively zigzag edges and triangular
shapes. Other alternatives based on patterned hydrogenation
rather than etching were also proposed.48
Theoretical predictions for infinite edges suggest that, with
hydrogen passivation, armchair configuration is the most
favorable one, with only a slight energy advantage over
zigzag (ZZ). Without H, a reconstructed edge terminated
by pentagon-heptagon pairs (ZZ57 ) is predicted to be the
lowest in energy, followed by armchair.8,39,49,50 For finite-size
structures, however, confinement effects and the presence of
corners would affect the stability of the overall structure and
the preferred edge morphology. For example, ZZ edges are
suggested to be dominating for small carbon clusters, both
with and without H passivation, while injection of pentagon
and heptagon defects may lower the total energy for some
structures.51
Determining the degree of edge passivation (one, two,
or no H atoms attached to edge carbon8 ) in experimental
structures is still complicated.42,45 Under the damaging high
flux of electrons in transmission electron microscopy (TEM)
there is, likely, no H. A consequent hydrogenation of such
unpassivated edges is possible.8,27 Interconversion between
reconstructions requires overcoming high-energy barriers,
even for nonpassivated carbon edges.44 Thus, locking in a
nonoptimal configuration remains quite possible. So far, only
TGQDs with ideal H-passivated ZZ and armchair edges were
studied. Effects of pentagons and heptagons injection, leading
to mixing of the sublattices, has not been addressed yet.
In this paper, using ab initio methods, we investigate the
robustness of TGQD properties of interest (zero-energy states
and magnetism) versus edge reconstructions and passivation as
a function of size. We use the most feasible with current manufacturing technique reconstructions (ZZ and ZZ57 ),38,39,41

165417-1

2011 American Physical Society

 LU,
POTASZ, AND HAWRYLAK
VOZNYY, GUC

PHYSICAL REVIEW B 83, 165417 (2011)

while our conclusions are expected to be general for any

configuration of pentagon-heptagon defects. We discover the
loss of magnetization in ZZ57 due to the increased width of the
zero-energy band. A more detailed study reveals why, in larger
ZZ TGQDs, with similar band dispersion, magnetization is not
affected.
Calculations have been performed within the density functional theory approach as implemented in the SIESTA code.52
We have used the generalized gradient approximation (GGA)
with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation
functional,53 double- plus polarization (DZP) orbital bases
for all atoms (i.e., 2s, 2p, and 2d orbitals for carbon, thus,
both and bonds are included on equal footing), TroullierMartins norm-conserving pseudopotentials to represent the
cores, 300-Ry real-space mesh cutoff for charge density (with
symmetrization sampling to further improve the convergence),
of vacuum between the
and a supercell with at least 20 A
periodic images of the TGQDs. Geometries were optimized
were reached, and
until the forces on atoms below 40 meV/A
exactly the same geometries were used for the comparison of
total energies of the ferromagnetic (FM) and antiferromagnetic
(AFM) configurations. Our optimized C-C bond length for
overestimates the experimental value
bulk graphene of 1.424 A
by 3%, typical for GGA.
Figure 1 shows the different TGQD structures considered in
this paper. Depending on the parity of the amount of atoms in
the edge of TGQD, the requirement of the ZZ57 reconstruction
of the edge leads to several possible structures. The three rings
at the corner can have 5-7-5, 7-6-5, or 6-6-5 arrangements,
presented in Figs. 1(b)1(d) (for the sake of comparison of total
energies, we investigate only those reconstructions conserving
the amount of atoms). Among reconstructed corners, only the
structure in Fig. 1(b) conserves the mirror symmetry of the
TGQD; however, according to our calculations, it is the least
stable due to strong distortion of the corner cells. Thus, in the
remainder of this paper, we will be presenting results utilizing
the configuration shown in Fig. 1(c) for an even, and that in
Fig. 1(d) for an odd, number of atoms n on a side of the
triangle.
Passivation by hydrogen is an important requirement
for the observation of the band of nonbonding states. Our
calculations show that, without hydrogen passivation, the
bonds hybridize with the bonds on the edge, destroying the
condition of the well-defined equivalent bonds on a bipartite
lattice and thus the zero-energy band itself. In Fig. 2(a),
we address the stability of hydrogen passivation for ZZ and
ZZ57 edges on the example of a triangle with n = 8 atoms
on a side (97 carbon atoms total and number of passivating hydrogens NH = 3n + 3 = 27). For hydrogen-passivated
structures, ZZH is 0.3 eV per hydrogen atom more stable
than ZZ57 H since, in the latter structure, the angles between
the bonds significantly deviate from the ideal 120 and the
total energy is affected by strain. In the absence of hydrogen,
however, the structure has to passivate the dangling bonds
by itself, e.g., by reconstructing the edge. Indeed, the ZZ57
reconstruction becomes 0.4 eV more stable. It is important to
note that hydrogen passivation is a favorable process for both
structures, even relative to the formation of H2 molecules,
and not only atomic hydrogen (i.e., formation of the H-H
bond can not compensate the energy loss due to breaking the

FIG. 1. (Color online) Triangular graphene quantum-dot edge

configurations considered in this paper: (a) Ideal zigzag edges ZZ,
(b) ZZ57 reconstruction with pentagon-heptagon-pentagon corner
configuration, (c) ZZ57 reconstruction with heptagon-hexagonpentagon corner, and (d) ZZ57 reconstruction with hexagon-hexagonpentagon corner.

C-H bond) [Fig. 2(a)]. The same conclusions hold for larger
TGQDs as well. Thus, the H-passivated edge, required for
magnetism, is easily achievable and we will present further
only the results for hydrogen-passivated structures omitting
the index H (i.e., use ZZ instead of ZZH). These results are
also consistent with the ones for infinite edges in graphene
nanoribbons.8,49,50 In Fig. 2(b), we investigate the relative
stability of hydrogen-passivated ZZ and ZZ57 structures as
a function of the linear size of the triangles. The largest
TGQD that we have studied has n = 23 atoms on a side of
the triangle (622 carbon atoms total). The fact that the energy
per edge atom increases with size signifies that the bulk limit
has not yet been reached. Clearly, the ZZ structure remains the
ground state for the range of sizes studied here. Nevertheless,
since most of the current experimental techniques involve
the unpassivated edge, for which ZZ57 is more stable, its
consequent hydrogenation may propagate the reconstruction
into the final structure, where it may be locked due to high
interconversion barrier.44
It was shown previously that the number of states in the
zero-energy band equals the difference between the number

FIG. 2. (a) Relative total energies of hydrogen-passivated and

nonpassivated TGQDs with reconstructed and nonreconstructed
edges for the case of n = 8 (NH = 27). (b) Total energy difference
between hydrogen-passivated ZZ57 and ZZ configurations as a
function of the number of atoms on a side of the triangle. Presented
values are energy per hydrogen atom.

165417-2

EFFECT OF EDGE RECONSTRUCTION AND PASSIVATION . . .

FIG. 3. (Color online) Energy spectra of the ground states for (a)
ZZ and (b) ZZ57 configurations for a hydrogen-passivated triangular
dot with n = 12. Spin-up states are shown in black squares and spindown states are shown in red circles. On the right-hand side, charge
densities of the filled part of zero-energy bands are shown. Circular
outlines show the population of only one sublattice in the ZZ structure
and both sublattices in ZZ57 .

of atoms in A and B graphene sublattices.19,54 Zero-energy

states are localized exclusively on the sublattice to which
the ZZ edges belong and are exactly degenerate within
the nearest-neighbor tight-binding model.18,19,23 Figure 3
compares the DFT electronic spectra near the Fermi level for
the ground states of hydrogen-passivated unreconstructed (ZZ)
and reconstructed (ZZ57 ) TGQDs with n = 12. Introduction of
the ZZ57 edge reconstruction smears the distinction between
sublattices. Nevertheless, the zero-energy band survives in
a reconstructed (ZZ57 ) TGQD. As can be seen from the
electronic density of occupied states of the band, they are
still predominantly localized on the edges. Moreover, the
number of zero-energy states remains the same. However,
the dispersion of this band increases almost threefold due
to reduction of the structure symmetry. Lifting of the band
degeneracy becomes observed even in the nearest-neighbor
tight-binding model with equal hoppings (not shown), and is
more pronounced for the structures in Figs. 1(c) and 1(d),
which additionally lift the reflection symmetry present in
Fig. 1(b).
Magnetization of the ZZ configuration was investigated
in detail through mean-field18,19 and exact diagonalization21
calculations. It was shown that the electrons in the zero-energy
band are spin polarized. The up- and down-spin edge states
are split around the Fermi level such that only up-spin states
are filled. Our calculated dispersion of the up-spin states is
0.03 eV/state [Fig. 3(a)]. On the other hand, the ground
state of the ZZ57 configuration is antiferromagnetic, i.e.,
there is no splitting between the up- and down-spin states
[Fig. 3(b)]. Nevertheless, calculations for the ferromagnetic
ZZ57 can still be performed by adjusting the Fermi level

PHYSICAL REVIEW B 83, 165417 (2011)

for up and down spins independently. The energy spectrum

obtained in such a way is similar to the one for ZZ, but with
negative min . The interplay of the max and the spin-up
band dispersion in such FM calculations can be monitored
to predict whether the ground state will be ferromagnetic
(min > 0) or antiferromagnetic (min < 0). Apart from the
significant increase of the band dispersion, we note that the
spin-up and -down splitting max reduces by a factor of 2 in
the ZZ57 structure. One can see from the charge-density plot
in Fig. 3(b) that zero-energy states can now populate both
A and B sublattices even close to the center of the dot (see
outlined regions). We speculate that the resulting reduction
in the peak charge density on each site is responsible for
the reduced on-site repulsion between spin-up and spin-down
electrons. Stronger dispersion and reduced up-down spin
splitting favor kinetic energy minimization versus exchange
energy and destroy the ferromagnetism in ZZ57 . It should be
noted that partial polarization can still be possible in ZZ57 .
Particularly, we observed it for structures with symmetric
corners [Fig. 1(b)], which exhibit smaller dispersion.
Our conclusions based on the analysis of the energy spectra
are supported by the total energy calculations depicted in
Fig. 4. For the ZZ structure, the gap min is always positive
and the total energy of the FM configuration is lower than that
of AFM (blue squares). For the ZZ57 configuration, on the
contrary, the ground state clearly remains AFM for all sizes
with the exception of the case with n = 4. Here the band consists of only three states and their dispersion can not overcome
the splitting between spin-up and spin-down states, resulting
in FM configuration being more stable. The total energy difference between the FM and AFM configurations for ZZ remains
almost constant (in the range 0.30.5 eV) for the triangle sizes
studied here, and reduces with size if divided by the number of
edge atoms. Such a small value, comparable to the numerical
accuracy of the method, makes it difficult to make reliable
predictions regarding magnetization of larger dots.
To investigate whether the magnetization of the edges
would be preserved on a mesoscale, we plot in Fig. 5 the

FIG. 4. (Color online) Total energy difference between ferromagnetic and antiferromagnetic states as a function of the size of
the triangle for hydrogen-passivated ZZ (blue squares) and ZZ57 (red
circles). For ZZ, the ground state is ferromagnetic for all sizes studied,
while for ZZ57 , it is antiferromagnetic for n > 4.

165417-3

 LU,
POTASZ, AND HAWRYLAK
VOZNYY, GUC

2.0

13 12

40 23

PHYSICAL REVIEW B 83, 165417 (2011)

4

1.5

CB

ZZ FM

1.0

E (eV)

0.5
0.0

min

max

-0.5
-1.0
-1.5
-2.0
0.00

VB

0.05

0.10

0.15

0.20

0.25

0.30

1/n

FIG. 5. (Color online) Scaling of the energy gaps with the inverse
linear size of ZZ TGQDs. Full energy spectra of the structures
calculated in this work are shown. Open symbols correspond to
spin-down and filled symbols to spin-up states.

evolution of the energy spectra with the TGQD size. For

this plot, we performed an additional calculation for the case
of n = 40 (1761 carbon atoms total). We did not perform
the geometry optimization for this case due to the high
computational cost, however, based on the results for smaller
structures, we expect that this would have a minor effect
on the spectrum. This allows us to notice the reduction
of the splitting max between the spin-up and spin-down
states with the growing size, which was not appreciated
previously.19 Our GGA gap between zero-energy bands (min )
and that between the valence and conduction bands are larger
than LDA gaps reported previously,19 as also observed for
graphene nanoribbons.55 Both gaps show sublinear behavior,
complicating the extrapolation to triangles of infinite size. This
behavior, however, should change to linear for larger structures

K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang,

S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, Science 306, 666
(2004).
2
K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I.
Katsnelson, I. V. Grigorieva, S. V. Dubonos, and A. A. Firsov,
Nature (London) 438, 197 (2005).
3
Y. B. Zhang, Y. W. Tan, H. L. Stormer, and P. Kim, Nature (London)
438, 201 (2005).
4
S. Y. Zhou, G. H. Gweon, J. Graf, A. V. Fedorov, C. D. Spataru,
R. D. Diehl, Y. Kopelevich, D. H. Lee, S. G. Louie, and A. Lanzara,
Nat. Phys. 2, 595 (2006).
5
A. H. C. Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and
A. K. Geim, Rev. Mod. Phys. 81, 109 (2009).
6
K. Nakada, M. Fujita, G. Dresselhaus, and M. S. Dresselhaus, Phys.
Rev. B 54, 17954 (1996).
7
K. Wakabayashi, M. Fujita, H. Ajiki, and M. Sigrist, Phys. Rev. B
59, 8271 (1999).

where the effect of edges reduces,22 converging both gaps to

zero, as expected for Dirac fermions. An important difference
from the nearest-neighbor tight-binding calculation23 is the
growing dispersion of the zero-energy bands. Combined with
the reduction of the valence-conduction gap, this leads to the
overlap of the zero-energy band with the valence band, even for
finite sizes, as indeed observed for the n = 40 case (see Fig. 5),
while in ZZ57 structures, it becomes visible already at n = 23
(not shown). Nevertheless, it does not affect the magnetization
of the edges, as indeed confirmed by our calculation for
n = 40, and can be compared to a magnetization of the
infinitely long hydrogen-passivated nanoribbons, where the
edge state overlaps in energy with the valence band but, in
k space, those bands do not actually cross.8 Our results thus
suggest that magnetization of the edges for infinitely large
triangles survives in the limit of zero temperature.
In conclusion, we have investigated the relative stability of
ZZ and ZZ57 edge reconstructions, both with and without hydrogen passivation, in triangular graphene quantum dots. Our
results suggest that pentagon-heptagon defects are possible in
such dots. The effect of edge reconstruction on electronic and
magnetic properties of the dots was investigated as a function
of size. The band of zero-energy states is found to survive
despite the fact that reconstruction smears out the distinction
between the two sublattices. However, the reduction of the dot
symmetry due to edge reconstruction is responsible for the
increased dispersion (width) of the zero-energy band, while
the mixing of sublattices reduces the splitting between spin-up
and spin-down bands. These two effects combined destroy the
magnetism in dots with reconstructed edges. Nevertheless, for
the dots with ideal ZZ edge, spin up-down splitting remains
always larger despite the increase of the zero-energy band
dispersion with the dot size, and magnetism survives even in
the limit of infinitely large dots.
The authors thank M. Korkusinski for fruitful discussions,
NRC-CNRS CRP, NRC-NSERC-BDC nanotechnology program, Canadian Institute for Advanced Research, Institute for
Microstructural Sciences, and QuantumWorks for support.

T. Wassmann, A. P. Seitsonen, A. M. Saitta, M. Lazzeri, and

F. Mauri, Phys. Rev. Lett. 101, 096402 (2008).
9
L. Yang, M. L. Cohen, and S. G. Louie, Phys. Rev. Lett. 101, 186401
(2008).
10
J. Cai et al., Nature (London) 466, 470 (2010).
11
B. Wunsch, T. Stauber, and F. Guinea, Phys. Rev. B 77, 035316
(2008).
12
J. Wurm, A. Rycerz, I. Adagideli, M. Wimmer, K. Richter, and
H. U. Baranger, Phys. Rev. Lett. 102, 056806 (2009).
13
A. R. Akhmerov and C. W. J. Beenakker, Phys. Rev. B 77, 085423
(2008).
14
F. Libisch, C. Stampfer, and J. Burgdorfer, Phys. Rev. B 79, 115423
(2009).
15
Z. Z. Zhang, K. Chang, and F. M. Peeters, Phys. Rev. B 77, 235411
(2008).
16
M. Ezawa, Phys. Rev. B 76, 245415 (2007).
17
M. Ezawa, Phys. Rev. B 77, 155411 (2008).

165417-4

EFFECT OF EDGE RECONSTRUCTION AND PASSIVATION . . .

18

J. Fernandez-Rossier and J. J. Palacios, Phys. Rev. Lett. 99, 177204

(2007).
19
W. L. Wang, S. Meng, and E. Kaxiras, Nano Lett. 8, 241 (2008).
20
J. Akola, H. P. Heiskanen, and M. Manninen, Phys. Rev. B 77,
193410 (2008).
21
A. D. Guclu, P. Potasz, O. Voznyy, M. Korkusinski, and
P. Hawrylak, Phys. Rev. Lett. 103, 246805 (2009).
22
A. D. Guclu, P. Potasz, and P. Hawrylak, Phys. Rev. B 82, 155445
(2010).
23
P. Potasz, A. D. Guclu, and P. Hawrylak, Phys. Rev. B 81, 033403
(2010).
24
H. Sahin, R. T. Senger, and S. Ciraci, J. Appl. Phys. 108, 074301
(2010).
25
T. Yamamoto, T. Noguchi, and K. Watanabe, Phys. Rev. B 74,
121409 (2006).
26
X. Yan, X. Cui, B. Li, and Liang-Shi Li, Nano Lett. 10, 1869
(2010).
27
Y. Kobayashi, K. Fukui, T. Enoki, and K. Kusakabe, Phys. Rev. B
73, 125415 (2006).
28
O. V. Yazyev and M. I. Katsnelson, Phys. Rev. Lett. 100, 047209
(2008).
29
K. A. Ritter and J. W. Lyding, Nat. Mater. 8, 235 (2009).
30
E. H. Lieb, Phys. Rev. Lett. 62, 1201 (1989).
31
A. Luque and A. Marti, Adv. Mater. 22, 160 (2010).
32
A. J. Nozik, Nano Lett. 10, 2735 (2010).
33
J. P. Gabor, K. Zhong, M. Bosnick, M. Park, and M. McEuen,
Science 325, 1367 (2009).
34
R. Baer and E. Rabani, Nano Lett. 10, 3277 (2010).
35
J. McClain and J. Schrier, J. Phys. Chem. C 114, 14332 (2010).
36
M. D. Watson, A. Fechtenkotter, and K. Mullen, Chem. Rev.
(Washington, DC, U.S.) 101, 1267 (2001).
37
X. Li, X. Wang, L. Zhang, S. Lee, and H. Dai, Science 319, 1229
(2008).
38
C. O. Girit, J. C. Meyer, R. Erni, M. D. Rossell, C. Kisielowski,
L. Yang, C. H. Park, M. F. Crommie, M. L. Cohen, S. G. Louie, and
A. Zettl, Science 323, 1705 (2009).

PHYSICAL REVIEW B 83, 165417 (2011)

39

P. Koskinen, S. Malola, and H. Hakkinen, Phys. Rev. B 80, 073401

(2009).
40
X. T. Jia, M. Hofmann, V. Meunier, B. G. Sumpter, J. CamposDelgado, J. M. Romo-Herrera, H. B. Son, Y. P. Hsieh, A. Reina,
J. Kong, M. Terrones, and M. S. Dresselhaus, Science 323, 1701
(2009).
41
A. Chuvilin, J. C. Meyer, G. Algara-Siller, and U. Kaiser, New J.
Phys. 11, 083019 (2009).
42
K. Suenaga and M. Koshino, Nature (London) 468, 1088 (2010).
43
D. V. Kosynkin, A. L. Higginbotham, A. Sinitskii, J. R. Lomeda,
A. Dimiev, B. K. Price, and J. M. Tour, Nature (London) 458, 872
(2009).
44
M. Engelund, J. A. Furst, A. P. Jauho, and M. Brandbyge, Phys.
Rev. Lett. 104, 036807 (2010).
45
B. Krauss, P. Nemes Incze, V. Skakalova, L. P. Biro, K. V. Klitzing,
and J. H. Smet, Nano Lett. 10, 4544 (2010).
46
L. C. Campos, V. R. Manfrinato, J. D. Sanchez-Yamagishi, J. Kong,
and P. Jarillo-Herrero, Nano Lett. 9, 2600 (2009).
47
F. Schaffel, J. H. Warner, A. Bachmatiuk, B. Rellinghaus,
B. Buchner, L. Schultz, and M. H. Rummeli, Nano Res. 2, 695
(2009).
48
M. Wu, X. Wu, Y. Gao, and X. C. Zeng, J. Phys. Chem. C 114, 139
(2010).
49
P. Koskinen, S. Malola, and H. Hakkinen, Phys. Rev. Lett. 101,
115502 (2008).
50
C. K. Gan and D. J. Srolovitz, Phys. Rev. B 81, 125445 (2010).
51
D. P. Kosimov, A. A. Dzhurakhalov, and F. M. Peeters, Phys. Rev.
B 81, 195414 (2010).
52
J. M. Soler, E. Artacho, J. D. Gale, A. Garcia, J. Junquera,
P. Ordejon, and D. Sanchez-Portal, J. Phys. Condens. Matter 14,
2745 (2002).
53
J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865
(1996).
54
S. Fajtlowicz, P. E. John, and H. Sachs, Croat. Chem. Acta 78, 195
(2005).
55
E. Rudberg, P. Salek, and Y. Luo, Nano Lett. 7, 2211 (2007).

165417-5