Interface structure of Ni nanoparticles on MgO (1 0 0): A combined HRTEM and molecular dynamic study

Article abstract of DOI:10.1016/j.susc.2006.01.045

Ni clusters with an average size of 4 nm, supported on MgO micro-cubes were studied by high resolution electron microscopy (HRTEM) and image simulations by the multislice technique. Regular defects were evidenced in the metal clusters at the interface. Molecular dynamic calculations of a 4 nm cluster indicates the same type of defects.

Full text of DOI:10.1016/j.susc.2006.01.045

Surface Science 600 (2006) L86–L90
www.elsevier.com/locate/susc
Surface Science Letters
Interface structure of Ni nanoparticles on MgO (100):
A combined HRTEM and molecular dynamic study
a
a,
a
b
a
S. Sao-Joao , S. Giorgio *, C. Mottet , J. Goniakowski , C.R. Henry
a
CRMCN-CNRS 1, Campus de Luminy, Case 913, 13288 Marseille Cedex 9, France
b
INSP, Campus de Boucicaut, 140 rue de Lourmel, 75015 Paris, France
Received 4 November 2005; accepted for publication 30 January 2006
Available online 20 February 2006
Abstract
Ni clusters with an average size of 4 nm, supported on MgO micro-cubes were studied by high resolution electron microscopy
(HRTEM) and image simulations by the multislice technique. Regular defects were evidenced in the metal clusters at the interface.
Molecular dynamic calculations of a 4 nm cluster indicates the same type of defects.
ꢀ 2006 Elsevier B.V. All rights reserved.
Keywords: Interface; HRTEM; Molecular dynamic
1. Introduction
the particular case of Ni layers grown on MgO single crys-
tals, a complex structure was evidenced by using surface
diffraction [12].
For reactions of hydrogenation, the catalytic activity
and selectivity of Ni nanoparticles supported on MgO were
found higher than in the commercial Ni catalysts [13].
The comparative studies of the Ni catalytic activities for
CH4 reforming, on different substrates as MgO, TiO2,
Al2O3, show that the Ni clusters properties strongly depend
on the metal support interactions [14].
The recent interest for supported Ni clusters is mainly
due to their catalytic properties for the carbon nanotubes
production [15].
Here, the interface Ni/MgO is studied at the atomic
scale by HRTEM, associated to image analysis and image
simulations by the multislice technique and molecular dy-
namic simulations.
The physical and chemical properties of supported metal
clusters can be modified by the interactions with the sub-
strate and by the structural variations due to their limited
size [1,2].
Many high resolution transmission electron microscopy
(HRTEM) studies were performed for the determination of
the atomic structure and the formation of intermediate
compounds in the metal–ceramic interfaces [3]. The inter-
face structure of bulk Cu on Al2O3 was determined for a
thick Cu film (1 lm) on a sapphire (0001) single crystal
[4] and for an Ag bulk layer on MgO [5]. The same tech-
nique was used in the case of Cu clusters larger than
10 nm, embedded in a Al2O3 matrix [6], and for nanoparti-
cles supported on various oxides, as Pd/MgO [7], Pd/ZnO
[8], Au/MgO [9], PdCu/MgO [10] and Au/TiO2 [11]. Most
of these interfaces are non reactive, and some of them (Ag/
MgO (100) and Pd/MgO have been observed in situ by
grazing incidence X-ray scattering. By this technique, in
2. Preparation technique
Pure Ni particles were prepared by the decomposition of
Ni(acac)2 on MgO oxide powders. The Ni(acac)2 (Fluka),
from 40 to 90 mg, was dissolved in acetonitrile (200 ml).
MgO (Fluka) clean powders (200 mg) with a surface
*
Corresponding author.
E-mail address: giorgio@crmcn.univ-mrs.fr (S. Giorgio).
Associated to Aix Marseille Universities.
1
0039-6028/$ - see front matter ꢀ 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2006.01.045

S. Sao-Joao et al. / Surface Science 600 (2006) L86–L90
L87
can be seen with a periodicity of 1.24 nm. Below the clus-
´
ters, some moire fringes in the substrate were attributed
to local deformations in MgO. In the power spectrum in
inset, the (200) reflexions are splitted, and correspond to
the d200 distance of bulk MgO and bulk Ni.
By numerical analysis of the image [18], the intensity
profiles were recorded along the [100] direction parallel
to the interface. The average distance between the (200)
fringes is plotted in Fig. 2b as a function of the Ni layer
number from the interface. The two first layers at the inter-
face, which contain the defects, are disturbed, then the lat-
tice corresponds to bulk Ni.
The structure and defects at the interface were enhanced
by Fourier filtering [19,20]. In Fig. 2c, three dislocations
are seen with an average distance of 1.25 nm. This distance
Fig. 1. TEM image of Ni clusters on a MgO cube: an overview.
specific area of 30 m2 gꢀ1 were mixed in the solution and
stirred at 80 ꢁC until complete evaporation of the solvent.
After drying at 80 ꢁC in air, the powders were slowly an-
nealed in vacuum (10ꢀ4 mbar), until 450 ꢁC, then reduced
under H2 at 400 mbar during 3 h at the same temperature
[16,17].
The powders were dispersed on a copper grid covered
with a carbon film and observed in standard electron
microscopy, with a Jeol 2000 FX microscope. The metal
particles seen in Fig. 1 have nanometer size distribution
(3.5 nm, 1 nm).
3. HRTEM observation
The preparations were observed with a Jeol 3010 micro-
scope for high resolution imaging (300 KV, Cs = 0.6 mm).
In Fig. 2a, a Ni cluster is seen in profile view along a
[100] direction. The shape is limited by two [011] edges.
At the interface, in the two first metal layers, three defects
Fig. 2b. Average distance between the (200) lattice planes normal to the
interface in Ni, from the interface to the top of the cluster.
Fig. 2a. HRTEM image of a Ni cluster in profile view, supported on
MgO. In inset is the Power Spectrum showing the splitting of the 200 spots
from Ni and from MgO.
Fig. 2c. Fourier filtered image of Fig. 2a with the 020 reflexions. Three
dislocations are indicated by arrows.

L88
S. Sao-Joao et al. / Surface Science 600 (2006) L86–L90
corresponds to the superimposition of seven unit cells of Ni
on six unit cells of MgO.
a 2.1 nm thick Ni layer (12 ML), which is about the height
of the cluster in Fig. 2, in epitaxy on a 63 nm thick MgO
layer (30 ML).
In Fig. 3, a Ni cluster is seen in top view on the sub-
´
strate. Two sets of moire fringes are parallel to the [100]
and [010] directions, with a periodicity of about 1 nm. In
the corresponding power spectrum, the satellites due to
4. Molecular dynamics calculations
´
the moire fringes are seen around the (200) spots. The
Prediction of the atomic structure can be performed by
quenched molecular dynamics simulations using semi-
empirical potentials as done recently on the Pd/MgO
(100) system [22]. The energetic model is based on a tight
binding many-body potential for the metal–metal bonds
and a surface energy potential approach fitted to ab initio
calculations for the metal–oxide ones. The details of the
energetic model description can be found in [22]. The
parameters for the Ni–Ni interactions are given in [23]
and the ones for the Ni–MgO (100) ones are listed in
[24]. The Ni atomic relaxations are obtained by the
quenched molecular dynamics procedure at 0 K, whereas
the MgO (100) substrate is kept rigid. This approximation
proved satisfactory for palladium clusters on MgO (100).
In the case of Ni, the metal–oxide interaction is of a similar
strength and the elastic properties of nickel differ only little
from these of palladium. Based on the elasticity-derived
arguments developed in [25] we estimate that nickel should
accommodate the major part of the interface-induced
distortion.
superimposition of the (200) lattice fringes of bulk Ni
(d200 = 0.176 nm) and bulk MgO (d200 = 0.21 nm) gives
moire fringes with a calculated distance D = 1.08 nm,
´
which is the distance measured in the experimental images.
Furthermore, the superimposition of Ni layers on vary-
ing thickness of MgO were simulated by the multislice tech-
nique with the EMS program [21]. The supercells with Ni,
and MgO were built with the dimensions 2.1 · 2.1 nm2
which corresponds to the superimposition of 6 · 6 unit cells
of Ni on 5 · 5 unit cells of MgO. In the images (Fig. 4a, b),
´
the moire fringes loose contrast as the thickness of the sub-
´
strate increases. The moire fringes are no longer visible for
The coordinates of atoms in a Ni cluster supported
on MgO obtained from molecular dynamic calculations
were used for HRTEM simulations with the mutislice tech-
nique (the considered cluster contains about 4000 Ni
atoms). For the cluster observed in profile view along a
[100] direction (Fig. 5), the super-cell lattice parameters
were a = 6.3150 nm and b = 4.7681 nm. For the cluster
observed in profile view along a [110] direction (Fig. 7),
the super-cell lattice parameters were a = 7.1570 nm and
b = 4.9786 nm.
The calculations were performed with a resolution of
512 · 512 pixels. The defects in the simulated images
(Fig. 6a) were evidenced in the image (Fig. 6b), four dislo-
cations parallel to [100] are seen with a spacing of five
Fig. 3. HRTEM image of a Ni cluster (001) oriented on MgO, in top
view. The power spectrum in inset shows the satellite reflections around
´
the central spot and the 200 spots, due to the moire fringes.
Fig. 4. (a), (b) Simulated images of the superimposition of 6 · 6 · 6 unit cells of Ni layers on 5 · 5 unit cells of MgO with increasing thickness (respectively
42 nm and 63 nm).

S. Sao-Joao et al. / Surface Science 600 (2006) L86–L90
L89
Fig. 5. Equilibrium shape obtained by molecular dynamic calculation of a
Ni cluster seen in profile view along [100].
Fig. 8a. Simulated HRTEM image by the multislice technique of the
calculated Ni cluster from Fig. 7.
Fig. 6a. Simulated HRTEM image by the multislice technique of the
calculated Ni cluster (Fig. 5).
Fig. 8b. Filtered image in the Fourier space of the simulated HRTEM
image of the cluster on Fig. 8a, showing the dislocations present in the
original image.
MgO layers. In the [110] direction, two defects in the cal-
culated image (Fig. 8a) are better seen in the phase image
(Fig. 8b).
In both cases, the distance between two dislocations is
slightly lower than in the experimental images. Indeed,
the HRTEM images correspond to the growth of seven
Ni layers on six MgO layers, which represent a distance
of 1.26 nm instead of 1.05 nm, between neighboring de-
fects. This difference is due to the shortening of Ni–Ni dis-
tances inside the calculated cluster, where the experimental
Ni–Ni distances preserve practically their bulk values. This
shortening, induced by a large proportion of surface
(under-coordinated) atoms, is somewhat overestimated in
our model of metal–metal interaction. In the present case
the effect may be additionally enhanced by the considerably
different equilibrium shapes of the experimental and the
calculated clusters, the latter being characterized by a lar-
ger aspect ratio. This difference can be hardly explained
by deficiencies of the computational model. We remind
that Pd clusters treated within the same framework showed
a tendency to an underestimation of the aspect ratio [22].
Since in the present case we find an opposite effect, we tend
to question the hypothesis of non-reactivity which is at the
origin of the theoretical model, and suggest that nickel may
oxidize to some extend at the experimental interface.
Fig. 6b. Filtered image in the Fourier space of the simulated HRTEM
image of the cluster on Fig. 6a. The dislocations are indicated by arrows.
Fig. 7. Equilibrium shape of the Ni cluster on MgO (100) calculated by
molecular dynamic simulation, seen in profile view along the [110]
direction.

L90
S. Sao-Joao et al. / Surface Science 600 (2006) L86–L90
5. Discussion
References
[1] C.R. Henry, Surf. Sci. Rep. 31 (1998) 231.
[2] H.J. Freund, Surf. Sci. 500 (2002) 271.
[3] T. Epicier, C. Esnouf, J. Phys. III France 4 (1994) 1811.
Experimental images of the Ni/MgO interface in cross-
section were obtained. They correspond to epitaxial rela-
tions: (001) Ni k (001) MgO with [100] Ni k [100]
MgO. They show regular defects identified as dislocations,
with a periodicity of 1.24 nm in the metal, in the three first
layers from the interface with the oxide. This spacing
between the defects corresponds to the superimposition
of seven unit cells of Ni over six unit cells of MgO.
[4] G. Dhem, M. Ruhle, G. Ding, R. Raj, Proc. ICEM 13 (1994) 277
¨
(Paris).
[5] A. Trampert, F. Ernst, C.P. Flynn, H.F. Fischmeister, M. Ruhle,
¨
Acta Metall. 40 (1992) S227.
`
[6] A. Peyrot, M.B. Ricoult, S. Hagege, Proc. ICEM 13 (1994) 265
(Paris).
[7] S. Giorgio, C.R. Henry, C. Chapon, J.M. Penisson, J. Cryst. Growth
100 (1990) 254.
In the experimental image in top view of a Ni cluster on
´
MgO, the moire fringes are separated by 1 nm. Simulated
[8] S. Giorgio, C.R. Henry, C. Chapon, Microsc. Microanal. Micro-
struct. 6 (1995) 237.
[9] S. Giorgio, C. Chapon, C.R. Henry, G. Nihoul, J.M. Penisson,
Philos. Mag. A 64 (1991) 87.
[10] S. Giorgio, C.R. Henry, Microsc. Microanal. Microstruct. 8 (1997)
379.
images by the multislice technique of the superimposition
of 6 · 6 unit cells of Ni on 5 · 5 unit cells of MgO, show
moire fringes with the same spacing, 1.08 nm.
´
The atomic structure of a Ni cluster on MgO was pre-
dicted by molecular dynamics calculations and the atoms
coordinates were used to simulate the HRTEM image by
the multislice technique. They clearly indicate defects with
a spacing of five MgO unit cells and six Ni unit cells.
The misfit between bulk Ni (a = 0.352 nm) and MgO
(a = 0.42 nm), is 17%. In order to accommodate Ni (001)
on MgO (001), with Ni atoms located on oxygen ions,
two coincidence meshes are possible. As observed in the
experimental images, seven Ni atoms are superimposed
on six oxygen ions in the [100] directions of MgO, which
corresponds to a low misfit of 2.2%. The images obtained
for the calculated cluster agree well with the experimental
results, the slightly higher misfit (3.4%) being due to a
stronger bond contraction within the calculated cluster.
Whereas the experimental and calculated images in top
view, normal to the Ni/MgO interface are in a good agree-
ment, in profile view, the experimental images give a lower
cluster’s aspect ratio and reveal a better accommodation
between the Ni and MgO lattices. In our opinion this
may be due to a more complex structure of the experimen-
tal interface, with a possibility of partial oxidation of
deposited nickel. This latter goes beyond the scope of our
computational model.
[11] S. Giorgio, C.R. Henry, B. Pauwels, G. Van Tendeloo, Mater. Sci.
Eng. A 297 (2001) 197.
[12] A. Barbier, G. Renaud, O. Robach, J. Appl. Phys. 84 (1998) 4259.
[13] M. Serra, P. Salagre, Y. Cesteros, F. Medina, J.E. Sueiras, J. Catal.
209 (2002) 202.
[14] Z. Xu, Y.M. Li, J.Y. Zhang, L. Chang, R.Q. Zhou, Z.T. Duan, Appl.
Catal., A 210 (2001) 45.
[15] C. Park, M.A. Keane, J. Catal. 221 (2004) 386.
[16] S. Giorgio, C.R. Henry, Eur. Phys. J. AP 20 (2002) 23.
[17] S. Sao-Joao, S. Giorgio, J.M. Penisson, C. Chapon, S. Bourgeois, C.
Henry, J. Phys. Chem. B 109 (2005) 342.
[18] S. Giorgio, C. Chapon, C.R. Henry, G. Nihoul, Philos. Mag. B 67
(1993) 773.
[19] S. Giorgio, G. Nihoul, C. Chapon, C.R. Henry, Microsc. Microanal.
Microstruct. 3 (1992) 61.
[20] M.J. Hytch, E. Snoeck, P. Baules, J.C. Ousset, M.J. Casanove, S.
Dubourg, J.F. Bobo, J. Appl. Phys. 89 (2001) 5414.
[21] P.A. Stadelman, Ultramicroscopy 21 (1987) 131.
[22] W. Vervisch, C. Mottet, J. Goniakowski, Phys. Rev. B 65 (2002) 1649.
´
[23] C. Mottet, J. Goniakowski, F. Baletto, R. Ferrando, G. Treglia,
Phase Transitions 77 (2004) 101.
[24] http://www.crmcn.univ-mrs.fr/gModel/webparam.pdf.
[25] W. Vervisch, C. Mottet, J. Goniakowski, Eur. Phys. J. D 24 (2003)
311.