Structure and reactivity of Ru nanoparticles supported on modified graphite surfaces: A study of the model catalysts for ammonia synthesis

Article abstract of DOI:10.1021/ja031718s

Supported ruthenium metal catalysts have higher activity than traditional iron-based catalysts used industrially for ammonia synthesis. A study of a model Ru/C catalyst was carried out to advance the understanding of structure and reactivity correlations in this structure-sensitive reaction where dinitrogen dissociation is the rate-limiting step. Ru particles were grown by chemical vapor deposition (CVD) via a Ru₃(CO)₁₂ precursor on a highly oriented pyrolytic graphite (HOPG) surface modified with one-atomic-layer-deep holes mimicking activated carbon support. Scanning tunneling microscopy (STM) has been used to investigate the growth, structure, and morphology of the Ru particles. Thermal desorption of dissociatively adsorbed nitrogen has been used to evaluate the reactivity of the Ru/HOPG model catalysts. Two different Ru particle structures with different reactivities to N₂ dissociation have been identified. The initial sticking coefficient for N₂ dissociative adsorption on the Ru/HOPG model catalysts is ～10^-6, 4 orders larger compared to that of Ru single-crystal surfaces.

Full text of DOI:10.1021/ja031718s

Published on Web 06/18/2004
Structure and Reactivity of Ru Nanoparticles Supported on
Modified Graphite Surfaces: A Study of the Model Catalysts
for Ammonia Synthesis
Zhen Song, Tanhong Cai, Jonathan C. Hanson, Jose A. Rodriguez, and Jan Hrbek*
Contribution from the Chemistry Department, BrookhaVen National Laboratory,
Upton, New York 11973
Received December 15, 2003; E-mail: hrbek@bnl.gov
Abstract: Supported ruthenium metal catalysts have higher activity than traditional iron-based catalysts
used industrially for ammonia synthesis. A study of a model Ru/C catalyst was carried out to advance the
understanding of structure and reactivity correlations in this structure-sensitive reaction where dinitrogen
dissociation is the rate-limiting step. Ru particles were grown by chemical vapor deposition (CVD) via a
Ru3(CO)12 precursor on a highly oriented pyrolytic graphite (HOPG) surface modified with one-atomic-
layer-deep holes mimicking activated carbon support. Scanning tunneling microscopy (STM) has been
used to investigate the growth, structure, and morphology of the Ru particles. Thermal desorption of
dissociatively adsorbed nitrogen has been used to evaluate the reactivity of the Ru/HOPG model catalysts.
Two different Ru particle structures with different reactivities to N2 dissociation have been identified. The
initial sticking coefficient for N2 dissociative adsorption on the Ru/HOPG model catalysts is ∼10-6, 4 orders
larger compared to that of Ru single-crystal surfaces.
Introduction
catalytic activity by modifying the electronic properties of the
active particles and by affecting the growth and ultimately the
structure and morphology of the active particles that change
the abundance of the active sites of the catalyst. Specifically,
in the case of the Ru/C system, it has been found that the
graphitization and texture of a carbon support are two crucial
factors influencing the activity of the catalyst.8-10 It is known
that the texture of a substrate plays an important role in
controlling the dispersion of the supported particles. Liang et
al.8 studied different carbon-supported Ru catalysts toward the
ammonia synthesis reaction and found that catalytic activities
in terms of turnover frequencies have the same trend as the
electrical conductivities of the carbon support, a parameter
indicating the extent of graphitization of the C materials. It has
also been found that very high temperature (up to 1900 °C)
treatment of a carbon support is needed before loading Ru to
achieve a good activity of the catalysts.8-10 Such a high
temperature treatment basically induces a significant graphitiza-
tion of the carbon material. However, no explanation of the
correlations between graphitization and reactivity has yet been
reported.
Ru catalysts have shown much higher activity at lower
temperatures and pressures in ammonia synthesis than the
conventional iron catalyst, and they are becoming the next
generation catalysts for ammonia synthesis.1-4 Studies of this
class of catalysts include: (1) the determination of the active
site on Ru for N2 dissociation, known to be the step-limiting
process in ammonia synthesis, (2) the optimization of the Ru
structure for the best activity, and (3) the search of a good
support for Ru. It has been found that the nature of the supports
influences greatly the Ru catalytic activity and that carbon is
the most active among many of supports investigated thus far,
including oxides, nitrides, zeolites, etc.5-7 Although carbon-
supported Ru catalyst has been used industrially in the ammonia
synthesis process,2 the stability of the carbon support is still a
problem for the limited lifetime of the catalyst.3 To find a more
efficient and stable Ru catalyst, it is necessary to understand
why the activity of the Ru/C system is superior to others and
what factors control the catalyst stability.
The function of a support in a catalyst performance is one of
the important issues in catalysis. The support can influence the
(1) Kowalczyk, Z.; Jodzis, S.; Raro´g, W.; Zielin´ski, J.; Pielaszek, J.; Presz, A.
Appl. Catal., A 1999, 184, 95.
Among the reactions involved in ammonia synthesis, N2
dissociative adsorption on the catalyst surface is the rate-
determining step.11 It has been shown that the active site for
N2 dissociation is the so-called B5 site, made of five Ru atoms:
(2) Bielawa, H.; Hinrichsen, O.; Birkner, A.; Muhler, M. Angew. Chem., Int.
Ed. 2001, 40, 1061.
(3) Kowalczyk, Z.; Jodzis, S.; Raro´g, W.; Zielin´ski, J.; Pielaszek, J. Appl. Catal.,
A 1998, 173, 153.
(4) Schlo¨gl, R. Angew. Chem., Int. Ed. 2003, 42, 2004.
(5) Szmigiel, D.; Bielawa, H.; Kurtz, M.; Hinrichsen, O.; Muhler, M.; Raro´g,
W.; Jodzis, S.; Kowalczyk, Z.; Znak, L.; Zielin´ski, J. J. Catal. 2002, 205,
205.
(6) Hansen, T. W.; Hansen, P. L.; Dahl, S.; Jacobsen, C. J. H. Catal. Lett.
2002, 84, 7.
(7) McClaine, B. C.; Siporin, S. E.; Davis, R. J. J. Phys. Chem. B 2001, 105,
7525.
(8) Liang, C.; Wei, Z.; Xin, Q.; Li, C. Appl. Catal., A 2001, 208, 193.
(9) Raro´g-Pilecka, W.; Szmigiel, D.; Komornicki, A.; Zielin´ski, J.; Kowalczyk,
Z. Carbon 2003, 41, 579.
(10) Raro´g, W.; Kowalczyk, Z.; Sentek, J.; Skladanowski, D.; Zielin´ski, J. Catal.
Lett. 2000, 68, 163.
(11) Hinrichsen, O. Catal. Today 1999, 53, 177.
9
8576
J. AM. CHEM. SOC. 2004, 126, 8576-8584
10.1021/ja031718s CCC: $27.50 © 2004 American Chemical Society

Structure and Reactivity of Ru Nanoparticles
A R T I C L E S
two at step edges and three at the lower terraces.12,13 Theoretical
calculations and experiments have shown that the rate of N2
dissociation is completely dominated by the steps.12 Therefore,
the ammonia synthesis reaction is extremely structure-sensitive
on Ru. A marble model calculation showed that for round-
shaped Ru particles, the abundance of B5 sites had a sharp
maximum at the particle size around 2.0 nm in diameter.13
However, it was found experimentally that the NH3 synthesis
activity of the Ru catalysts increased with the size of Ru particles
up to 5 nm in diameter and then became constant with the
additional increase.8,9 A better understanding of the particle size-
dependent reactivity of the Ru catalysts requires the knowledge
of the structure and morphology of the Ru particles with atomic
resolution, especially that of the steps on the particle surfaces.
Transmission electron microscopy (TEM) studies of Ru
particles grown on various supports, such as C, BN, Si3N4, MgO,
MgAl2O4, and TiO2, showed that in most cases Ru appears as
round particles.13-15 Some of the TEM images showed clearly
orientational growths of Ru crystals relative to the support, e.g.,
Ru/SiN4, Ru/BN, and Ru/TiO2.13-15 However, no evidence has
been found for the orientational/epitaxial growth of Ru on
graphite samples, probably because of the use of poorly
crystallized graphitic supports.13 Since a TEM image is a
projection image, no detail information about the steps at the
particle surfaces has been provided.
In this article, we present a scanning tunneling microscopy
(STM) study of the growth of Ru particles on a highly oriented
pyrolytic graphite surface by chemical vapor deposition via a
Ru3(CO)12 precursor. To simulate the structure of activated
carbon, we etched the surface with oxygen to form randomly
distributed, one-atomic-layer-deep nanoholes on the surface.
Unlike a TEM image, an STM image provides a three-
dimensional structure of the Ru particles grown on the highly
oriented pyrolytic graphite (HOPG) surface, especially the step
structure on the surfaces of Ru particles, which is related closely
to the reactivity of the catalyst. Our results have shown that a
well-ordered, stepped graphite surface is crucial for the epitaxial
growth of Ru particles. Ru grows laterally at the steps or hole
edges of the HOPG surface and forms layered nanocrystals with
the (0001) surfaces parallel to the graphite surface. The texture
of the hole-modified HOPG surface plays a role in controlling
the size and shape of the Ru particles. Reactivity study of the
Ru/HOPG toward N2 dissociation by using temperature-
programmed desorption (TPD) shows a high efficiency of the
dissociation of N2 on the layered/stepped Ru nanocrystals. These
results help to understand the importance of graphitization and
texture of the carbon support for the preparation of Ru/C
catalysts and point to a favorable Ru structure with a high
activity for the ammonia synthesis process.
distributed one-atomic-layer-deep holes on the HOPG surface. A freshly
cleaved HOPG sample was ion-sputtered in a UHV chamber with the
100 eV Ne+ for ∼15-60 s. Then, the sample was oxidized in air at a
temperature of 530 °C for ∼10-20min. Figure 1 shows two modified
HOPG surfaces with one-layer-deep holes of 7 and 32 nm in diameter,
respectively. The holes on the HOPG surfaces were generated by
etching of the top graphite layer around the defect sites created by the
ion sputtering. The density and the size of the holes could be adjusted
by the time for sputtering and oxidation, respectively. In this way, we
could control the texture of the HOPG surface.
Before the deposition of Ru, the HOPG sample was heated to 900
K in UHV for 1 h to remove the impurities (such as O2, H2O, CO,
etc.) from the surface. We deposited Ru on the modified HOPG surface
at 550 K by CVD via a Ru3(CO)12 precursor in a preparation chamber,
which was separated from the STM chamber by a gate valve. The
Ru3(CO)12 was kept in a glass vial at room temperature and connected
with the preparation chamber with a gate valve.16 The exposure of Ru
carbonyl was controlled by dosing time, and the surface concentration
of Ru was determined by Auger electron spectroscopy (AES) using
the published Auger data.17 This approach can lead to errors of factor
of 2 in surface concentration determination.18 Since the strongest AES
peaks of Ru and C overlap at 272-273 eV, the Ru concentration was
calculated using the Ru 231 eV peak together with the mixed peak at
272 to 273 eV. The fraction of surface covered by Ru particles was
also measured independently from STM images.
After the preparation, the samples were transferred under vacuum
to the STM chamber (Omicron, base pressure ∼5 × 10-11 Torr). The
STM measurements were performed at room temperature with a
tungsten tip.
N2 dissociation on the Ru/HOPG was studied by TPD in a separate
UHV chamber (base pressure ∼4 × 10-10 Torr) equipped with a high-
pressure cell. N2 adsorption could be carried out in either the main
chamber or the high-pressure cell. In the main chamber we used a tube
doser with the enhancement factor19 of ∼3 for dosing nitrogen. The
effect of the hot filament of the ion gauge on the probability for
dinitrogen dissociative chemisorption in our chamber was much less
than previously reported.20 With the sample mounted 0.6 m away and
out of the direct line-of-sight from the ion gauge, we measured a 2-fold
increase in nitrogen surface concentration with the filament switched
on during N2 exposure. The high-pressure cell, isolated from the main
chamber by a gate valve, allowed the adsorption of N2 at pressures up
to 1 atm and the direct transfer of the sample to the main chamber for
the TPD measurement after evacuation. The modified HOPG substrates
were characterized by STM before installation into the TPD chamber.
The thermocouple was cemented to the back of the sample. TPD spectra
were recorded with a heating rate of 2 K/s. High purity 14N2 (99.999%)
and isotope 15N2 gases were used as received. Since both N2 and CO
have mass spectrum of parent species at m/e ) 28, we monitored
nitrogen signal at fragment with m/e ) 14 where 14N+ represented 14%
of the N2+ intensity, while CO2+ fragment was only 0.8% of the parent
CO+ signal (for a more detailed discussion, see ref 20).
Some of the Ru/HOPG samples used in the TPD study were first
measured by STM and then transferred through air to the TPD chamber;
the others were directly prepared in the TPD chamber. The samples
transferred through air were reduced by H2 in the TPD chamber until
no O2 was detected by AES.
Experimental Section
Our preliminary experiments showed that Ru3(CO)12 sticks only on
step edges of the HOPG surface. To obtain an evenly Ru-decorated
surface, we first prepared a modified HOPG surface with randomly
Results and Discussions
Structure of the Ru/HOPG Model Catalyst. Figures 2 and
3 show the STM images and analysis of Ru deposited on the
(12) Dahl, S.; Logadottir, A.; Egeberg, R. C.; Larsen, J. H.; Chorkendorff, I.;
To¨rnqvist, E.; Nørskov, J. K. Phys. ReV. Lett. 1999, 83, 1814.
(13) Jacobsen, C. J. H.; Dahl, S.; Hansen, P. L.; To¨rnqvist, E.; Jensen, L.; Topsøe,
H.; Prip, D. V.; Møenshaug, P. B.; Chorkendorff, I. J. Mol. Catal. A 2000,
163, 19.
(16) Cai, T.; Song, Z.; Chang, Z.; Liu, G.; Rodriguez, J. A.; Hrbek, J. Surf. Sci.
2003, 538, 76.
(17) Davis L. E.; MacDonald, M. C.; Palmberg, P. W.; Riach, G. E.; Weber, R.
E. Handbook of Auger Electron Spectroscopy, 2nd ed.; Physical Electronics
Inc.: Eden Prairie, MN, 1976.
(18) Woodruff, D. P.; Delchar, T. A. Modern Techniques of Surface Science;
Cambridge University Press: Cambridge, 1988; p 146.
(19) Campbell, C. T.; Valone, S. M. J. Vac. Sci. Technol., A 1985, 3, 408.
(14) Hansen, T. W.; Wagner, J. B.; Hansen, P. L.; Dahl, S.; Topsøe, H.; Jacobsen,
C. J. H. Science 2001, 294, 1508.
(15) Komaya, T.; Bell, A. T.; Weng-Sieh, Z.; Gronsky, R.; Engelke, F.; King,
T. S.; Pruski, M. J. Catal. 1994, 149, 142.
9
J. AM. CHEM. SOC. VOL. 126, NO. 27, 2004 8577

A R T I C L E S
Song et al.
Figure 1. STM images of two HOPG surfaces modified by one-atomic-layer-deep holes with diameters of (a) 7 and (b) 32 nm. Image size: (a) 400 × 400
nm2 and (b) 800 × 800 nm2.
min for the second sample. No oxygen is detected by AES at
either of the Ru coverages, indicating that the Ru3(CO)12 has
completely decomposed and that the deposited Ru is in metallic
state. XANES and XRD measurements (not shown) corroborated
the formation of metallic Ru on HOPG.
Figure 2a shows that the growth of Ru starts from the step
edges and the edges of holes of the modified HOPG substrate.
The Ru particles appear as round islands. The apparent height
distribution is shown in Figure 2b, displaying two maxima at
∼0.5 and ∼1.5 nm, respectively. Similarly, the apparent
diameter distribution of the particles shows also two peaks at
∼3.5 and ∼6.5 nm, respectively (Figure 2c), indicating a
bimodal growth of Ru on the HOPG surface. Our previous
studies on the CVD growths of Mo and Ru on a Au(111) surface
have shown that the nucleation of metal clusters occurred before
their anchoring on the surface.16,21 The bimodal CVD growth
Figure 2. (a) STM image of a small hole-modified Ru/HOPG surface (2
atom % Ru). Image size: 200 × 200 nm2. (b) Height and (c) diameter
distributions of the Ru particles in (a).
of Ru on HOPG may proceed by the nucleation of Ru during
small hole-modified HOPG surface displayed in Figure 1a at
two Ru coverages. The corresponding AES spectra are shown
in Figure 4, together with a clean HOPG AES spectrum. The
Ru dosing time is 20 min for the first sample and another 25
the Ru3(CO)12 decomposition to form small particles and by
the sintering of the small particles to big ones. The Ru surface
concentration of this sample measured by AES is ∼2 atom %,
(21) Song, Z.; Cai, T.; Rodriguez, J. A.; Hrbek, J.; Chan, A.; Friend, C. J. Phys.
Chem. B 2003, 107, 1036.
(20) Dietrich, H.; Geng, P.; Jacobi, K.; Ertl, G. J. Chem. Phys. 1996, 104, 375.
9
8578 J. AM. CHEM. SOC. VOL. 126, NO. 27, 2004

Structure and Reactivity of Ru Nanoparticles
A R T I C L E S
Figure 3. (a) STM image of a small hole-modified Ru/HOPG surface
(30 atom % Ru). Ru forms flat, layered particles that completely cover the
holes. Image size: 200 × 200 nm2. (b) The orientations of the steps of the
Ru particles (black and dotted arrows) shown in (a), and the HOPG lattice
orientation (gray arrows) obtained from an atomic resolution STM image
shown. Image size: 1.1 × 1.1 nm2. The gray arrows are rotated by 30°
from the hexagonal atomic structure of the HOPG surface because the STM
displays every other C atom of the graphite surface.29 (c) A close-up
differential image of a Ru particle. Image size: 70 × 55 nm2. (d) A line
profile across the Ru particle shown in (c).
Figure 5. Two possible epitaxial growths of Ru on a HOPG surface,
denoted by the large hexagons with the sides equal to the Ru lattice constant
(0.271 nm). The two Ru lattices are rotated relatively by 30°. The small
circles represent the carbon atoms at the top (dark gray) and the second
(light gray) layers of the graphite surface.
Figure 3a is an STM image of a higher Ru surface concentra-
tion, ∼30 atom % (AES measurement), on the same HOPG
surface. The apparent Ru area fraction measured by STM is
36%. At this Ru coverage, the etched holes on HOPG have
been completely covered by the Ru particles. STM images show
that the Ru particles have layered structure with the steps on
the particles running along six directions (Figure 3b). Figure
3c is a close-up image of one Ru particle, displaying the layered
crystalline structure. A line profile across the crystal is shown
in Figure 3d. The highest part of the particle reaches ∼3 nm,
and the width is ∼ 50 nm. On average, the height of the Ru
particles in Figure 3a have increased by a factor of ∼3, and the
apparent coverage increased by a factor of ∼18, compared to
those in Figure 2a. This indicates that Ru grows much faster
laterally than vertically.
Many of the Ru islands show regular triangular or truncated
triangular shapes derived from hexagonal structures, implying
a (0001) surface orientation. The single step height on the Ru
particles is 0.22 ( 0.03 nm, consistent with the step height
(0.214 nm) on a Ru (0001) surface. The six side orientations of
the Ru crystals, shown in Figure 3b, can be treated as two sets
of threefold symmetry axes, relatively rotated by 30° around
the surface normal. One set is coincident with the orientation
of the atomic resolution image of HOPG surface (Figure 3b).
The small angle deviations arise from the drift of the microscope.
On the basis of these results, we propose that the growth of
Ru on a graphite surface is epitaxial and has two orientations,
both with the (0001) surfaces parallel to the graphite (0001)
surface. One of them has the same orientation as the HOPG
hexagonal lattice (indicated by k1, k2, and k3 vectors); the other
is 30° rotated around the surface normal, indicated by kr1, kr2,
and kr3 (see Figure 5). Considering that the nearest-neighbor
atom distance of the Ru(0001) surface is 0.271 nm, and that of
the HOPG surface is 0.142 nm, the mismatches of the Ru (0001)
Figure 4. AES spectra of the samples shown in Figure 2 (spectrum a) and
Figure 3 (spectrum b). Spectrum c is from a clean HOPG sample.
while the Ru covered area fraction obtained from STM images
is ∼7%. The disagreement can be caused by the STM tip
convolution that overestimates the Ru particle diameters; on the
other hand, a contribution of the carbon AES signal under the
Ru particles leads to a lower estimate of the Ru surface
concentration by AES.
9
J. AM. CHEM. SOC. VOL. 126, NO. 27, 2004 8579

A R T I C L E S
Song et al.
holes are fully covered by Ru and the growing of particles are
more symmetrical.
Stability of the Ru Particles upon Oxidation and Reduc-
tion. Ru catalysts can be oxidized by exposure to air, and they
can be reduced to metallic state in the H2. To test the stability
of the Ru on HOPG during these processes, we oxidized and
then reduced the Ru/HOPG sample that is shown in Figure 3a.
The oxidation was carried out in O2 of ∼2 × 10-7 Torr at 800
K for 1 h, and the reduction was carried out in H2 of ∼2 ×
10-5 Torr at 550 K for 2 h. The sample was then annealed to
800 K for 1 h. After the oxidation, oxygen was detected on the
sample by AES (not shown), indicating the formation of
ruthenium oxide. Meanwhile, the surface Ru concentration
decreased to 15 atom %, a 50% loss from the original sample.
This loss is most probably caused by the sublimation of
ruthenium oxide during the oxidation at high temperature.23 AES
measurements also showed that the oxide was readily reduced
by the H2 treatment without further loss of Ru from the sample.
We tried to image the oxidized Ru sample by STM and found
that it was impossible to obtain a stable image. The Ru particles
were easily pushed away by the STM tip. This could be caused
by the bond weakening between nanoparticles and the graphite
substrate after the oxidation. After the reduction, the STM
measurement again became possible. Figure 8a is an STM image
of Ru/HOPG after the oxidation and reduction. The Ru particles
still keep the flat shape, and the additional coalescence between
particles has also occurred. The height distribution of the Ru
particles on this sample is slightly broader than that of the as-
deposited sample, as shown in Figure 8b. The average heights
are 3.3 and 4.8 nm for the as-deposited Ru and that after the
oxidation and reduction treatments, respectively.
Figure 6. Atomic resolution images (10 × 10 nm2) of two Ru particles
with different orientations. (a), (c) Top view and (b), (d) three-dimensional
images.
lattice (shown as the big hexagons in Figure 5) to the two
corresponding hexagonal structures derived from the HOPG
lattice are -4.7% and +9.8%, respectively.
The atomic resolution STM images of the Ru crystals, shown
in Figure 6a and c, directly prove this interpretation. Both images
show that the surface atoms of Ru are arranged in hexagonal
structure, indicating the (0001) surface. The selected two crystals
are rotated 30° to each other. The crystal shown in Figure 6c
has the same orientation as that of the HOPG substrate.
Overall, a fraction of the Ru in the catalysts is lost as an
oxide at higher temperatures. However, the morphology of the
Ru particles on HOPG is stable, keeping the flat, layered form.
N2 Dissociation on the Ru/HOPG Surfaces. Figure 9A
shows that there is a large background desorption of nitrogen
from HOPG. The N2 TPD spectra from a clean modified HOPG
Parts a and b of Figure 7 are STM images of an as-deposited
Ru (14 atom %)/HOPG surface that was modified with large
holes (Figure 1b). The Ru particles start to grow at the step or
hole edges of the HOPG surface. Some of the large Ru particles
have formed by aggregation of small particles. Figure 7c,d
shows images of the sample after annealing to 800 K, which
has resulted in further aggregation and better crystallization of
the Ru particles. Unlike the growth on small holes, Ru particles
growing on the large hole-modified HOPG surface have an
elongated shape. Two different Ru orientations rotated by 30°
are also found on this sample, shown by Figure 7e. The edge
angles of the particles imply a (0001) surface orientation. Figure
7c,d also shows that the Ru crystallites grown on the area with
higher density of the holes are smaller and that the big Ru
crystals are usually found at large extended graphite surface
areas.
surface after exposure to 2 × 1011 L (curve a) and 1 × 1011
L
(curve c) of N2 at room temperature were monitored with m/e
) 14 to distinguish between CO and N2 contributions. The
featureless curve b in Figure 9A is a blank TPD of m/e ) 14
spectrum recorded right after acquiring curve a without exposure
to N2. The broad desorption peak probably arises from the
molecular N2 adsorbed at the defect sites of HOPG and/or the
subsurface N2 formed by penetration of the N2 molecules into
the sublayer of the HOPG through the defect sites. We treat
this broad desorption feature as a background signal and subtract
it from the TPD curves as shown in Figure 9B, curves c and d.
This background signal, found in all high-pressure adsorption
experiments, is absent in the high-vacuum ones. Additionally,
a small contribution (0.8%) from doubly ionized CO to the
signal of m/e ) 14 (Figure 9B, curve b) is also subtracted from
the total m/e ) 14 signal (Figure 9B, curve a). The final curve
d is that of N2 desorption signal from Ru particles. This data
processing procedure is validated by an isotope experiment (see
below).
The reason for the formation of elongated shape particles is
unknown. One possibility is that an asymmetric electric field
is present around a particle at an edge of a large hole due to the
broken symmetry at the edge. This electric field could influence
the state and diffusion of the Ru3(CO)12 molecules and frag-
ments.22 In the case of the small hole decorated surface, the
Figure 10 shows representative TPD spectra from samples
of different morphologies after the data processing described
(22) Pascual, J. I.; Jackiw, J. J.; Song, Z.; Weiss, P. S.; Conrad, H.; Rust, H.-P.
(23) CRC Handbook of Chemistry and Physics, 46th ed.; Chemical Rubber
Co.: Cleveland, Ohio, 1965-1966.
Phys. ReV. Lett. 2001, 86, 1050.
9
8580 J. AM. CHEM. SOC. VOL. 126, NO. 27, 2004

Structure and Reactivity of Ru Nanoparticles
A R T I C L E S
Figure 7. STM images of large hole-modified Ru/HOPG surface (∼14 atom % Ru). (a), (b) As deposited. (c), (d) After annealing to 800 K. The image
sizes: (a), (c), and (d) 400 × 400 nm2, (b) 100 × 100 nm2. (e) The Ru crystallites step orientations.
Figure 8. (a) STM image of the Ru/HOPG sample (15 atom % Ru) after the oxidation and reduction treatment. Image size: 200 × 200 nm2. (b) The height
distributions of the Ru/HOPG sample before and after the oxidation and reduction.
above. Curve a is a TPD spectrum of 15N2 (m/e ) 30) from a
sample (Ru: ∼15 atom %) with the large flat particles similar
to that shown in Figure 3a after exposure to 6 × 104 L of 15N2
at 400 K under high vacuum. The isotope experiment confirms
that the peak at 520 K is due to the N2 desorption. This
experiment has been reproduced for several samples that have
been measured by STM either before or after the TPD
measurements. The morphologies of these samples are shown
in Figures 3a, 7c and d, or 8a, in which the Ru particles expose
flat, layered structure of the (0001) surface. The N2 adsorption
has been carried with both 14N2 and 15N2 at both low (∼10-5
Torr) and high (∼10 Torr, 14N2 only) pressures. All the TPD
spectra show a peak at ∼500 K. At larger exposure, the peak
position shifts to lower temperature as expected for the second-
order desorption of N2 formed by recombination of the atomic
nitrogen on the surface.
Curve c is the N2 TPD spectrum (m/e ) 14) from a sample
with ∼2 atom % Ru of a small, round particles (similar to those
shown in Figure 2a), after an exposure of 2 × 1011
L
14N2 at
400 K. Small exposures of N2 comparable to those for curves
a and b did not result in detectable TPD peak on this sample.
To allow direct comparison with curve a, curves b and c have
been multiplied by 7, the parent/fragment ratio of N2+/N+. The
TPD spectrum c shows the N2 desorption from the round particle
peaks at 650 K.
Curve b is from a sample prepared exactly the same way as
for the sample shown in Figure 7a,b. STM images have shown
the modification holes on both samples are the same; AES
spectra show the Ru surface concentrations are also identical.
We, therefore, assume that the TPD curve b is from a sample
with morphology imaged in Figure 7a, which shows both small
round and large flat Ru particles. The N2 adsorption was carried
9
J. AM. CHEM. SOC. VOL. 126, NO. 27, 2004 8581

A R T I C L E S
Song et al.
Figure 9. TPD spectra of N2 (m/e ) 14) from (A) clean HOPG surface and (B) Ru/HOPG surface. Panel (A): (a) 2 × 1011 L N2 at 300 K, (b) no N2
exposure, and (c) 1 × 1011 L N2 at 300 K. Panel (B): simultaneous record of (a) m/e ) 14 and (b) m/e ) 28 (multiplied by 0.008) after 2 × 1010 L N2
exposure at 300 K, (c) subtraction of (b) from (a), the dotted line represents a scaled contribution from a clean HOPG, and (d) subtraction of the dotted curve
from (c).
Figure 11. 15N2 (a) and 15N 14N (b) TPD spectra from a Ru/HOPG sample
(∼10 atom % Ru) first exposed to 15N2 (2 × 105 L) and then to 14N2
(4 × 104 L) at 400 K.
Figure 10. TPD spectra from (a) large flat Ru crystallites (∼15 atom %
Ru), 6 × 104 L 15N2 at 400 K, (b) small, round and large, flat Ru particles
recombination of 14N and 15N atoms on the Ru surfaces, this
experiment verified dissociative adsorption of N2 on the Ru
particle surface.
(∼14 atom % Ru), 3 × 104
L
14N2 at 400 K, and (c) small, round Ru
particles (∼2 atom % Ru), 2 × 1011 L 14N2 at 300 K. The STM images are
the representative surface morphologies of the corresponding TPD curves.
Figure 12 shows two TPD spectra from the same sample after
two different exposures to N2. The lower exposure curve (the
same one shown in Figure 10b) is from an as-deposited sample,
and we used it to calculate the N2 dissociative adsorption
probability, which is assumed to be close to the initial sticking
coefficient s0. The amount of adsorbed N is determined by
comparing the peak area of the N2 TPD curve with that of CO
saturated adsorbed on a Ru (0001) surface at room temperature,
where θCO ) 0.6.24 Interestingly, the sticking coefficient of N2
on the Ru/HOPG system is ∼1 × 10-6, 4 orders larger than
that on a Ru (0001) single-crystal surface (9 × 10-11) measured
by Dahl et al. at 400 K.25 Although at low coverage the N
concentrations on the two different structures of the Ru particles
are nearly the same, N coverage on the flat particles increased
much more than that on the round particles. Additionally, the
desorption peak from the round Ru particles shifts significantly
(from 710 to 620 K), more than expected for the coverage
dependence of N2 desorption. A structural change of these
out in the main chamber of 4 × 10-6 Torr at the sample
temperature of 400 K. Two TPD peaks have been observed,
one at ∼500 K, and the other at ∼700 K. We suggest that the
∼500 K peak in the curve b originates from the large flat Ru
particles (see curve a) and that the 700 K peak (b) and the 650
K peak (c) have the same origin, namely the round Ru particles.
The shift in the peak positions may arise partially from the N
coverage difference due to the second-order desorption of N2
and partially from the size difference of the particles on the
surfaces. The small round particles on the as-deposited sample
(b) have an apparent height of ∼1 nm (Figure 7a,b), while the
round Ru particles on the annealed sample (c) are taller (an
apparent height of ∼3 nm), as measured by STM after the TPD
experiment.
Figure 11 shows the isotope exchange experiment. The Ru/
HOPG sample (400 K) was first exposed to 15N2 (2 × 105 L),
and then the dosing gas was switched to 14N2 (2 × 104 L).
Signals from m/e ) 30 and 29 were recorded for the TPD
spectra. Since 14N15N (m/e ) 29) can only be obtained by
(24) Pfnu¨r, H.; Feulner, P.; Menzel, D. J. Chem. Phys. 1983, 79, 4613.
(25) Dahl, S.; To¨rnqvist, E.; Chorkendorff, I. J. Catal. 2000, 192, 381.
9
8582 J. AM. CHEM. SOC. VOL. 126, NO. 27, 2004

Structure and Reactivity of Ru Nanoparticles
A R T I C L E S
The high dissociation rate of N2 on Ru/HOPG can be ascribed
to the high density of steps on the particle surfaces. Besides
this, our STM images (Figure 6b,d) show that the surfaces of
the Ru particles are not perfectly flat. This implies the presence
of dislocations within the Ru particles and on their surfaces and,
consequently, induction of surface strain. Wintterlin et al.27
observed directly by STM that a stretched Ru(0001) surface
was much more active to NO dissociation; the N coverage at
the stretched part of the strained surface was much higher, while
it was lower at the compressed part. Therefore, in our system,
it is possible that the atomic N has higher concentration at certain
strained areas on the Ru particle surfaces, which induces higher
fraction of the low-temperature desorption peak of N2.
Dietrich et al.20,28 studied nitrogen TPD from opened surfaces
of single-crystal Ru surfaces, (101h0) and (112h1), after exposure
to N2 gas. On these two surfaces, two and four layers of Ru
atoms are exposed, respectively. The N2 desorption from both
surfaces are found at ∼650 K. We can safely assume that a Ru
particle with a round shape has larger fraction of opened surfaces
than the one with a flat shape. The N2 desorption peak at ∼650
K found from small round particles could be due to the
desorption from those opened surfaces.
Figure 12. TPD spectra of N2 at m/e ) 14 from a Ru/HOPG sample
(∼14 atom % Ru) after exposure to N2 of (a) 3 × 104 L and (b) 2 × 1010
L at 400 K.
particles must have occurred during the temperature ramping
of the TPD experiments. STM images (Figure 7) have shown
the sintering of the Ru particles upon annealing. Since the Ru
particles with the flat, layered structure are more stable and have
larger capacity for atomic N, they may have an ideal structure
and morphology required for the ammonia synthesis.
After several rounds of TPD experiments, the activity of the
catalyst decreases. One possible explanation is that the active
sites on Ru are poisoned by carbon coming from the CO
impurity in the N2 gas and from the graphite substrate. We have
found that oxidation and reduction treatment can reactivate the
Ru/HOPG catalyst. The reactivated sample shows the same TPD
features as those presented in Figure 10a,c.
Unlike the close-packed Ru(0001) surface, subsurface N can
form on both of the opened surfaces, with the concentrations
possibly 10 times more than that on the surface.20,28 This
difference shows an additional advantage of the (0001) surface,
which can hold the N atoms at the surface for the further NH3
synthesis reactions.
In summary, we assume that two major structural features
can be related to the high activity in ammonia synthesis of the
Ru/C system: (1) the well-ordered graphite substrate, corre-
sponding to high extent of graphitization and (2) the layered
and strained (0001) surfaces of the Ru particles. The former
allows the lateral growth of Ru on graphite, and in this way the
Ru surface area could be larger than that of ball-like particles.
The latter results in a relatively high density of B5 sites for N2
dissociation, and the (0001) terraces serves as storage for the
surface N atoms that migrated from the step edges after the
dissociation. The strain on the surfaces of the flat Ru particles
lowers the adsorption energy of N atoms, which could also
contribute to the high activity of the catalyst.
In conclusion, we suggest that the TPD peak at ∼650 K is
from small, round Ru particles, and the peak at ∼500 K is from
large flat Ru particles. The broad desorption feature starting at
∼380 K is from adsorption of N2 at defect sites of the HOPG
surface during high-pressure exposures.
Reactivity and Structures. Diekho¨ner et al.26 reported TPD
of N2 from a Ru(0001) single-crystal surface after exposing
Ru(0001) to an atomic beam of N. They found that at low
coverage (ΘN ≈ 0.3), desorption of N2 occurs at ∼800 K,
consistent with that obtained by adsorbing N2 on Ru(0001).25
With the increase of the N coverage, weakly bonded species
formed on the surface and desorbed at temperature between
∼400-600 K. Specifically, when the N coverage was larger
than 0.5, the low-temperature desorption (centered at 500 K)
dominated the TD spectrum. However, by exposing the Ru(0001)
single-crystal surface to N2, the N coverage is always less than
0.3;25 there has been no report of the low-temperature desorption.
It is interesting that the N2 desorption peak from the flat, layered
Ru particles in our experiment appears at ∼500 K. This could
imply a fact that the N2 dissociation rate on the Ru/HOPG is
high and allows the formation of high N coverage on the flat
Ru particles. The observed N coverage of ∼0.5 on Ru/HOPG
confirms this deduction.
Conclusions
We have manipulated the growth of Ru nanoparticles on a
hole-modified HOPG surface and prepared two kinds of Ru
clusters, round ones and flat, layered ones, by controlling the
texture of the HOPG surface and the surface concentration of
Ru. The growth of Ru on graphite is epitaxial. Two growth
orientations have been found. Both have the (0001) surfaces
parallel to the graphite surface. One has the same orientation
as the graphite lattice; the other is 30° rotated around the surface
normal. The texture of the graphite surface influences the size
and shape of the Ru particles.
(27) Wintterlin, J.; Zambelli, T.; Trost, J.; Greeley, J.; Mavrikakis, M. Angew.
Chem., Int. Ed 2003, 42, 2850.
(28) Dietrich, H.; Jacobi, K.; Ertl, G. J. Chem. Phys. 1997, 106, 9313.
(29) Hembacher, S.; Giessibl, F. J.; Mannhart, J.; Quate, C. F. Proc. Natl. Acad.
Sci. U.S.A. 2003, 100, 12539.
(26) Diekho¨ner, L.; Mortensen, H.; Baurichter, A.; Huntz, A. C. J. Vac. Sci.
Technol., A 2000, 18, 1509.
9
J. AM. CHEM. SOC. VOL. 126, NO. 27, 2004 8583

A R T I C L E S
Song et al.
The N2 adsoption/desorption behaviors on the round and flat
Ru particles are different. N2 that dissociates on round Ru
particles partially diffuses to the bulk of Ru particles, and the
part staying at surface desorbs at ∼650 K. N2 that dissociates
on flat Ru particles forms a high density of atomic N on the
(0001) terraces because of the large number of step sites and
strained (0001) surface, and it desorbs at ∼500 K. The
possibility to form such a high density of atomic N on the
surface may be the reason for the high activity of Ru/C catalysts
in the NH3 synthesis.
The extent of graphitization is important for the activity of
the Ru/C catalyst. A well-ordered graphite surface directs the
epitaxial growth of Ru and allows the formation of the flat,
layered Ru crystals.
Acknowledgment. This research was supported by the U.S.
Department of Energy, Division of Chemical Sciences, under
Contract DE-AC02-98CH10886.
JA031718S
9
8584 J. AM. CHEM. SOC. VOL. 126, NO. 27, 2004