Available via license: CC BY 4.0
Content may be subject to copyright.
crystals
Article
Hydrogenation of HOPG-supported Gold
Nanoparticles: Features of Initial Stages
Andrey K. Gatin, Maxim V. Grishin *, Nadezhda V. Dokhlikova, Sergey Yu. Sarvadii and
Boris R. Shub
N. N. Semenov Federal Research Center for Chemical Physics, Russian Academy of Sciences (FRCCP RAS),
Moscow 119991, Russia
*Correspondence: mvgrishin68@yandex.ru
Received: 10 June 2019; Accepted: 5 July 2019; Published: 9 July 2019
Abstract:
The features of deuterium adsorption on the surface of gold nanoparticles deposited on
highly oriented pyrolytic graphite (HOPG) were determined. The results showed that deuterium
adsorption on gold nanoparticles takes place at room temperature. The results also showed that the
filling of the nanoparticles’ surfaces with the adsorbate occurs from the graphite–gold interface until
the entire surface is covered by deuterium. The results of quantum chemical simulations are used to
explain the experimental data. A simple model of the observed effects is proposed.
Keywords:
gold nanoparticles; highly oriented pyrolytic graphite; interface; deuterium adsorption;
STM; DFT
1. Introduction
Many modern catalysts are composed of nanomaterials. In particular, gold nanoparticles are used
to catalyze the low-temperature oxidation of CO [
1
] and selective oxidative coupling of methanol to
methyl formate [
2
]. Gold nanoparticles are also used as catalysts for selective isomerization of epoxides
to allylic alcohols [
3
], benzylation of aromatics [
4
], and production of vinyl chloride and vinyl acetate
monomers [
5
], among other substances. Gold nanoparticles are used in catalytic reactions of alkene
hydrogenation [
6
], in the chemoselective hydrogenation of crotonaldehyde to crotyl alcohol [
7
], in the
chemoselective hydrogenation of nitro compounds [
8
], and in the heterogeneous hydroformylation
of olefins [
9
]. In addition, catalysts based on gold nanoparticles are used in hydrogenation reactions
assisted by hydrogen transfer, including the reduction of carbonyl compounds [
10
,
11
] and the
hydrochlorination of amines [
12
] and alkynes [
13
]. The dissociative adsorption of molecular hydrogen
and the interaction of the generated atomic hydrogen with nanoparticles’ surfaces play an important
role in these processes. Using calorimetry, TEM and XRD, direct synthesis of water from O
2
and
H
2
was found to occur over silica-supported gold nanoparticles at 383–433 K and under several
kilopascals of pressure [
14
]. The results of corresponding DFT calculations are also presented in this
paper. The results obtained for Al
2
O
3
-supported gold nanoparticles [
15
,
16
] supported a conclusion [
14
]
regarding the interaction of gold and hydrogen. Coordinatively unsaturated Au atoms and/or such
defects as vacancies and impurity atoms could be centers of hydrogen adsorption and dissociation.
However, in [
17
], using XPS and mass spectrometry methods, it was found that for the gold-based
nanocatalysts supported on TiO
2
, hydrogen dissociation takes place on the perimeter of interfaces
between the gold and TiO
2
. At the same time, calculations [
18
] differ from the conclusion reached
above. It is specifically stated that centers of H
2
dissociation must be low-coordinated gold atoms
not directly bound to the stoichiometric and reduced TiO
2
support. In [
19
] it was suggested that
hydrogen dissociation occurs on low-coordinated gold atoms, then hydrogen atoms migrate to the
substrate. In our previous works [
20
,
21
], using methods of scanning tunneling microscopy (STM)
Crystals 2019,9, 350; doi:10.3390/cryst9070350 www.mdpi.com/journal/crystals
Crystals 2019,9, 350 2 of 14
and spectroscopy (STS), we observed the dissociative adsorption of hydrogen on gold nanoparticles
deposited on graphite. Thus, direct participation of gold nanoparticles in the processes of molecular
hydrogen dissociation is not in doubt.
At the same time, despite the long-term studies of nanocatalytic systems, including those based
on gold nanoparticles, the mechanisms of elementary acts of even model reactions remain unclear.
One of the reasons for the difficulties is that most of the methods used for the study of physicochemical
(including catalytic) properties of nanoparticles—photoelectron microscopy, mass spectrometry, and
calorimetry—provide information averaged over a large ensemble of nanoparticles in different states.
As a result, it is impossible to link the local structural features of nanoparticles with their chemical
properties. This problem can be solved using probe methods—in particular, scanning tunneling
microscopy and spectroscopy, the spatial resolution of which allows us to investigate chemical
interactions between gaseous reagents and solid surfaces at the level of single molecules and single
defects. Previously [
21
], we found that adsorption of hydrogen leads to a change in the electronic
structure of gold nanoparticles. Furthermore, the results of [
22
] have revealed that electron density
distortions introduced by the hydrogen adatom are local. These facts have allowed us to observe the
evolution of the distribution of adsorbed deuterium atoms on the surface of gold nanoparticles while
the sample exposure to molecular deuterium is increasing.
The main purpose of this work is to determine the locations of adsorbed deuterium on the
surface of gold nanoparticles, as well as to establish filling dynamics of nanoparticles’ surfaces with an
adsorbate as the exposure increases. The paper presents the results of an experimental and numerical
study of the initial stages of deuterium adsorption on gold nanoparticles deposited on the surface of
highly oriented pyrolytic graphite (HOPG).
2. Experiment
The experiments were carried out at a facility consisting of a scanning tunneling microscope
(UHV VT STM, Omicron NanoTechnology, Taunusstein, Germany), Auger spectrometer (CMA-100,
Omicron NanoTechnology, Taunusstein, Germany), quadrupole mass spectrometer (HAL 301 PIC,
Hiden Analytical Limited, Warrington, UK) and auxiliary equipment, at T =300 K and a residual gas
pressure of P =2
×
10
−10
Torr. We used tips made of platinum-iridium and tungsten wires prepared by
standard methods as STM probes.
Nanoparticles were deposited on the surface of HOPG by the impregnation method. To achieve
this deposition, an aqueous solution of HAuCl
4
with a metal concentration of 5
×
10
−6
g/mL was
applied to the substrate surface. Then, the sample was dried, placed in a vacuum chamber and
annealed in ultrahigh vacuum at T =500–750 K for several hours. The required duration of annealing
was established by the results of STM study of the sample surface morphology. The fraction of substrate
surface covered by nanoparticles did not exceed 0.05.
Morphology and electronic structure of the sample surfaces at the level of single nanoparticles,
as well as the results of their modification due to the interaction with adsorbate, were determined
using topographic and spectroscopic STM measurements. As the adsorbate we used a molecular
isotope of hydrogen—deuterium (D
2
). The use of D
2
made it possible to more accurately control
the exposure of the samples in the injected gas. It is known that nanocontact formed by the metal
sample and the conducting tip demonstrates the S-shaped dependence of the STM tunnel current on
voltage (volt-ampere characteristic (VAC)) [
23
]. The change in elemental composition of the sample
(for instance, due to chemical reactions), can lead to the transition of its electronic structure from metal
to semiconductor type or to a significant decrease in the density of states in the vicinity of the Fermi
level. This results in the appearance of a zero-current section within the S-shaped VAC curve attributed
to the band gap of the material [
24
–
27
]. Thus, changes in VAC curve shape can indicate chemical
composition change over the surface of nanoparticles. Hereafter, the term “VAC of nanoparticles”
refers to the VAC of STM tunnel contact formed by STM tip and the surface of gold nanoparticles
Crystals 2019,9, 350 3 of 14
placed on the graphite support, and the term “VAC of graphite” refers to the VAC of STM tunnel
contact formed by STM tip and the pure graphite surface.
Analysis of the elemental composition of the samples’ surfaces was carried out by Auger
spectroscopy. The obtained data were compared with the results from spectroscopic STM measurements.
The gas composition at all stages of the experiment was controlled by the results of mass
spectrometric measurements. In the experiments described below, the molecular deuterium pressure
did not exceed P =1×10−6Torr. Exposure value was measured in Langmuir, 1 L =1×10−6Torr ×s.
Numerical Experiment
Quantum chemical simulation of the interaction between atomic deuterium and the interface
formed by gold clusters and the carbon substrate—supercluster Au
13
C
138
—was carried out within
the framework of density functional theory (DFT) using software packages OpenMX-3.7 (OMX)
and QuantumEspresso-5.1.1 (QE). All the calculations were carried out in a generalized gradient
approximation using the PBE functional without regard for spin polarization. The GGA-PBE
approximation met the requirements of the study, since the main focus was on calculating the
projected density of states (PDOS) for the atoms of the model during deuterium adsorption. The
atomic bases in OMX are specified using localized functions—pseudoatomic orbitals. For Au atoms
we used the s1p1d1-set with 17 valence electrons (N
val
) and a cut-offradius (R
cut
) of 7 Å. For C atoms,
s1p1-set,
Nval =4
, and R
cut
=5 Å were used, while for D atoms, the more accurate set of s3p2-set,
Nval =1
and R
cut
=7 Å was used. Base sets of flat waves were used in QE. In this study, the cut-off
energy of 24 Ry and ultrasoft pseudopotentials was chosen, taking into account relativistic effects for
gold. The spatial volume of the cell was 25
×
25
×
25 Å
3
, and the characteristic size of the model system
was 16
×
13
×
6 Å
3
, which allowed us to minimize the influence of neighboring superclusters. Note
that no calculations of PDOS were made for the atoms located on the edge of the supercluster during
the simulation; therefore, the effect of dangling bonds could be neglected. The use of two software
packages allowed us to cross-test the data of the numerical experiment.
As a model of a gold nanoparticle located at the edge of the HOPG terrace we used a supercluster
Au
13
C
138
. The atomic structure of C
138
is an extracted supercell of bulk graphite, the lower layer of
which consists of 90 carbon atoms, and the upper one of 48. During the calculations, the positions of
all C
138
atoms were fixed in order to avoid the influence of the transformation of the atomic structure
on the density of states of C atoms. In addition, control calculations showed that the change in the
chemical potential of C
138
was less than 1% during relaxation of the atomic structure. As a model of a
gold nanoparticle, an icosahedral isomer of the 13-atomic gold cluster was taken. Interacting with
C
138
—mostly with carbon atoms of the upper layer—the atomic structure of Au
13
changed, forming
two regions: the interface, where Au atoms were bound to C atoms, and the vertex, where Au atoms
were bound to Au atoms. The binding energy between Au13 and C138 was estimated by the formula:
Ebond =E(Au13C138)−E(C138)−E(Au∗
13), (1)
where E(Au
13
*) is the total energy of the Au
13
cluster with modified atomic structure. The obtained
value of –1.2 eV corresponds to the lower limit of the binding energy and is consistent with the results
of the STM/STS experiment. Note that since the purpose of the study is to search for general qualitative
trends in PDOS changes, the Au
13
atomic structure obtained is a suitable model within the framework
of our research, although it is unlikely to correspond to the global minimum of potential energy. During
interaction with deuterium atoms, the atomic structure of obtained gold–carbon system was fixed,
except in a few cases asterisked in Figure 6. It was necessary to use such an approach as we needed
to separate the electron and the geometric effects of deuterium adsorption; the first one was due to
the formation of the bond, and the second one was associated with changes in the atomic structure of
Au
13
C
138
caused by deuterium adsorption. The method of numerical experimentation was essentially
Crystals 2019,9, 350 4 of 14
the calculation and summation of PDOS for atoms near the site of D adsorption before and after the
interaction. The obtained graphs were compared with VAC data acquired in the STM experiments.
3. Results and Discussion
This work includes STM experiments on molecular deuterium interaction with the surface of gold
nanoparticles, and quantum chemical simulation of atomic deuterium adsorption onto the gold cluster
in contact with two graphene nanocells modeling the edge of the HOPG terrace. In the theoretical
portion, we consider the causes of hydrogen dissociative adsorption on the surface of gold nanoparticles
and the influence of hydrogen chemisorption on the STM tunneling spectra.
3.1. Experimental Portion
Gold nanoparticles were distributed on the surface of the HOPG in such a way that most of them
were arranged on graphite surface defects (terrace edges). Nanoparticles had an oblate spheroidal
shape. The maximum of the lateral size distribution was in the range of 4–8 nm, with an average height
of approximately 1.5–2 nm. An example image of the HOPG surface area with gold nanoparticles
deposited on it is shown in Figure 1a. Figure 1b,c demonstrate the averaged VAC of nanoparticles
(dots B) and graphite (dots A) and the Auger electrons spectrum of the sample, respectively. The results
of spectroscopic measurements revealed that VACs of nanoparticles and graphite were similar—they
possessed an S-shape VAC similar to cases of tunnel contact formed by conductors. It could also be
noted that the absolute value of the tunneling current measured on the nanoparticle in the entire
voltage range was slightly higher than that of the tunneling current measured on graphite. That is, the
local conductivity of nanoparticles was slightly higher than the conductivity of graphite. Such behavior
occurred because the density of states in the vicinity of the Fermi level of metal nanoparticles was
higher than that of the HOPG (semimetal). The absence of residual impurities on gold was confirmed
by the elemental analysis of the sample surface: Auger spectra show maxima at energies of 272 eV
and 69, 141, and 160 eV, corresponding to carbon and gold, respectively. Noticeable signals of other
elements are not detected.
The exposure of the sample to deuterium resulted in a significant perturbation of the local electron
density of nanoparticles. The change in the electronic structure of nanoparticles as a result of increased
exposure to D2is shown in Figure 2a–f. Each part in this figure consists of three elements:
(1) The first element is a topography image of the HOPG surface area with nanoparticles deposited
on it. Blue crosses indicate points where VAC was measured;
(2) The second element shows the averaged VAC curves corresponding to the points indicated on
the topography image (curve A: area of graphite without nanoparticles; curve B: peripheral area of
nanoparticles; curve D: central area of nanoparticles);
(3) The last element is a “current” image of the same surface area, which is a map of the surface
distribution of tunneling current values at a certain voltage value (darker points correspond to higher
values of tunneling current). This type of image informs the local electron density of the sample surface
in an explicit form.
Crystals 2019,9, 350 5 of 14
Crystals 2019, 9, x FOR PEER 5 of 14
Figure 1. Gold nanoparticles deposited on HOPG: (a) topography image, (b) VAC of nanoparticles
and HOPG before exposure to deuterium, (c) Auger spectrum of the sample.
As seen from the above data, after the exposure of the sample to deuterium equal to 200 L, an
area appeared on the nanoparticles’ periphery in which the measured value of the tunneling current
was lower than one measured on the graphite (hereafter this area is called the “area of reduced
conductivity”). However, the value of the tunneling current measured in the center of nanoparticle
was still bigger than the one measured on the graphite (Figure 2a). That is, after exposure to
deuterium, we could observe two different areas on the surface of the gold nanoparticle—the central
area with high conductivity, and the peripheral one with low conductivity. In addition, the VAC
measured at several individual points on the periphery of the nanoparticle had a zero-current
section. Before the exposition of the sample to deuterium, the existence of such curves was not
observed.
As we can see, further increasing the sample’s exposure to deuterium up to 800 L resulted in
significant changes: the width of the reduced conductivity region at the periphery of the particle
increased, as did the number of points where such VACs containing zero-current sections were
detected (Figure 2b). The central part of the nanoparticles remained an area of high conductivity.
With an increase in exposure to deuterium up to 1200 L, some points in the central region of
nanoparticles started to demonstrate VAC with a zero-current section. However, the average
conductivity of the nanoparticle central region was still slightly higher than the conductivity of
graphite (Figure 2c). With further exposure to deuterium up to 1600 L, the width of the high
conductivity area decreased, and the number of points with zero-current section within the VAC
increased even further (Figure 2d). Finally, when the exposure reached approximately 2000 L, it
became impossible to identify areas with high and low conductivity on the surface of the
Figure 1.
Gold nanoparticles deposited on HOPG: (
a
) topography image, (
b
) VAC of nanoparticles and
HOPG before exposure to deuterium, (c) Auger spectrum of the sample.
As seen from the above data, after the exposure of the sample to deuterium equal to 200 L, an
area appeared on the nanoparticles’ periphery in which the measured value of the tunneling current
was lower than one measured on the graphite (hereafter this area is called the “area of reduced
conductivity”). However, the value of the tunneling current measured in the center of nanoparticle
was still bigger than the one measured on the graphite (Figure 2a). That is, after exposure to deuterium,
we could observe two different areas on the surface of the gold nanoparticle—the central area with
high conductivity, and the peripheral one with low conductivity. In addition, the VAC measured at
several individual points on the periphery of the nanoparticle had a zero-current section. Before the
exposition of the sample to deuterium, the existence of such curves was not observed.
As we can see, further increasing the sample’s exposure to deuterium up to 800 L resulted in
significant changes: the width of the reduced conductivity region at the periphery of the particle
increased, as did the number of points where such VACs containing zero-current sections were detected
(Figure 2b). The central part of the nanoparticles remained an area of high conductivity. With an
increase in exposure to deuterium up to 1200 L, some points in the central region of nanoparticles
started to demonstrate VAC with a zero-current section. However, the average conductivity of the
nanoparticle central region was still slightly higher than the conductivity of graphite (Figure 2c). With
further exposure to deuterium up to 1600 L, the width of the high conductivity area decreased, and the
number of points with zero-current section within the VAC increased even further (Figure 2d). Finally,
when the exposure reached approximately 2000 L, it became impossible to identify areas with high
and low conductivity on the surface of the nanoparticle. The fraction of points with VAC containing
Crystals 2019,9, 350 6 of 14
a zero-current section was about one-third of the total number of points where VAC measurements
were carried out. So, after the exposure to 2000 L, the surface of nanoparticles became homogeneous,
and VACs with and without zero-current sections became almost randomly distributed over the entire
surface of the nanoparticle. Formal averaging of VACs measured at the center of the particle and at its
periphery demonstrates that conductivity of both the central and peripheral regions of the particle
was lower than that of the graphite (Figure 2e). Note that this effect was observed both for isolated
nanoparticles and for nanoparticles agglomerated in clusters. A further increase in exposure up to
3000 L did not lead to noticeable changes in the form of VAC curves measured on the surfaces of
nanoparticles (Figure 2f).
Crystals 2019, 9, x FOR PEER 6 of 14
nanoparticle. The fraction of points with VAC containing a zero-current section was about one-third
of the total number of points where VAC measurements were carried out. So, after the exposure to
2000 L, the surface of nanoparticles became homogeneous, and VACs with and without zero-current
sections became almost randomly distributed over the entire surface of the nanoparticle. Formal
averaging of VACs measured at the center of the particle and at its periphery demonstrates that
conductivity of both the central and peripheral regions of the particle was lower than that of the
graphite (Figure 2e). Note that this effect was observed both for isolated nanoparticles and for
nanoparticles agglomerated in clusters. A further increase in exposure up to 3000 L did not lead to
noticeable changes in the form of VAC curves measured on the surfaces of nanoparticles (Figure 2f).
Figure 2. Topography image, VAC, and current image of gold nanoparticles and HOPG at different
experimental stages. The letter designation of the curves corresponds to the points on the
topography. After exposure to deuterium: (a) 200 L; (b) 800 L; (c) 1200 L.
Figure 2. Cont.
Crystals 2019,9, 350 7 of 14
Crystals 2019, 9, x FOR PEER 7 of 14
Figure 2 (cont.). Topography image, VAC and current image of gold nanoparticles and HOPG at the
different stages of experiment. The letter designation of the curves corresponds to the points on the
topography. After exposure to deuterium: (d) 1600 L; (e) 2000 L; (f) 3000 L.
In our previous work [22], it was shown that adsorption of atomic hydrogen on the surface of
free gold nanoclusters results in a decrease in the density of states in the vicinity of the Fermi level,
leading to a decrease in the conductivity of the gold nanocluster. Taking into account the results of
this work and the experimental data described above, one can conclude that dissociative adsorption
of deuterium on the surface of gold nanoparticles takes place and that the filling of nanoparticles
with adsorbate begins mainly with the gold–graphite interface, followed by the filling of the rest of
the nanoparticle surface. It should be noted that the time of transition of the system to the final
(stationary) state is very long (up to 25–30 h). This long transition time indicates the presence of
significant activation barriers both for the acts of surface migration of deuterium atoms and for the
processes of structural rearrangement of nanoparticles.
3.2. Theoretical Portion and Quantum Chemical Simulation
In this section we consider why gold nanoparticles, unlike monocrystals, are able to adsorb
hydrogen, and why hydrogen adsorption significantly affects the tunnel spectra of STM. Our
considerations are built on the basis of model concepts that allow an analytical solution.
Figure 2.
Topography image, VAC, and current image of gold nanoparticles and HOPG at different
experimental stages. The letter designation of the curves corresponds to the points on the topography.
After exposure to deuterium: (a) 200 L; (b) 800 L; (c) 1200 L; (d) 1600 L; (e) 2000 L; (f) 3000 L.
In our previous work [
22
], it was shown that adsorption of atomic hydrogen on the surface of free
gold nanoclusters results in a decrease in the density of states in the vicinity of the Fermi level, leading
to a decrease in the conductivity of the gold nanocluster. Taking into account the results of this work
and the experimental data described above, one can conclude that dissociative adsorption of deuterium
on the surface of gold nanoparticles takes place and that the filling of nanoparticles with adsorbate
begins mainly with the gold–graphite interface, followed by the filling of the rest of the nanoparticle
surface. It should be noted that the time of transition of the system to the final (stationary) state is
very long (up to 25–30 h). This long transition time indicates the presence of significant activation
barriers both for the acts of surface migration of deuterium atoms and for the processes of structural
rearrangement of nanoparticles.
Crystals 2019,9, 350 8 of 14
3.2. Theoretical Portion and Quantum Chemical Simulation
In this section we consider why gold nanoparticles, unlike monocrystals, are able to adsorb
hydrogen, and why hydrogen adsorption significantly affects the tunnel spectra of STM. Our
considerations are built on the basis of model concepts that allow an analytical solution.
Hydrogen adsorption on the surface of metals, and gold in particular, has been studied many
times. According to the Newns-Anderson model, one can describe the surface as a set of single-electron
states with energy
εk
, and the adsorbate interacting with it as a single valence state with energy
εa
.
Therefore, when the adsorbate approaches the surface, the two sets of states will relate through the
interaction matrix with elements Vak, and the adsorbate density of states will be
na=1
π
∆(ε)
(ε−εa−Λ(ε))2+∆(ε)2, (2)
where
∆(ε)=πX
k
|Vak|2δ(ε−εk), (3)
Λ(ε)=P
πZ∆(ε0)
ε−ε0dε0. (4)
As can be seen,
∆
(
ε
) is a local projection of the density of states of the metal near the adsorbate. If
we consider the valence shells formed by the sp- and d-states in the case of noble and transition metals
— the most interesting case for catalysis — then we can simplify the dependence
∆
(
ε
) with the next
assumption—
∆
(
ε
) does not depend on energy for the sp-band, and the width of
∆
(
ε
) is small for the
d-band.
Thus, two stages can be distinguished in the process of atom/molecule adsorption on the metal
surface: the interaction of the adsorbate with the sp-states of the metal in the weak-chemisorption
approximation, and the interaction of the adsorbate with the d-states of the metal in the
strong-chemisorption approximation. Weak chemisorption leads to broadening of the density of
adsorbate states. Bonding and antibonding states are formed below and above the d-band and the
level of adsorbate as a result of strong chemisorption (Figure 3).
Crystals 2019, 9, x FOR PEER 8 of 14
Hydrogen adsorption on the surface of metals, and gold in particular, has been studied many
times. According to the Newns-Anderson model, one can describe the surface as a set of
single-electron states with energy ɛk, and the adsorbate interacting with it as a single valence state
with energy ɛa. Therefore, when the adsorbate approaches the surface, the two sets of states will
relate through the interaction matrix with elements Vak, and the adsorbate density of states will be
𝑛=1
𝜋∆(𝜀)
(𝜀−𝜀−Λ(𝜀))+∆(𝜀) , (2)
where
∆(𝜀)=𝜋
|𝑉|𝛿(𝜀−𝜀
)
, (3)
Λ(𝜀)=𝑃
𝜋∆(𝜀′)
𝜀−𝜀′𝑑𝜀′. (4)
As can be seen, Δ(
ɛ
) is a local projection of the density of states of the metal near the adsorbate. If we
consider the valence shells formed by the sp- and d-states in the case of noble and transition metals — the
most interesting case for catalysis — then we can simplify the dependence Δ(ɛ) with the next
assumption—Δ(ɛ) does not depend on energy for the sp-band, and the width of Δ(ɛ) is small for the
d-band.
Figure 3. Schematic representation of the change in the local electronic structure of an atom in the
case of adsorption on the surfaces of simple (alkali) and transition/noble metals.
Thus, two stages can be distinguished in the process of atom/molecule adsorption on the metal
surface: the interaction of the adsorbate with the sp-states of the metal in the weak-chemisorption
approximation, and the interaction of the adsorbate with the d-states of the metal in the
strong-chemisorption approximation. Weak chemisorption leads to broadening of the density of
adsorbate states. Bonding and antibonding states are formed below and above the d-band and the
level of adsorbate as a result of strong chemisorption (Figure 3).
The interactions of the adsorbate with the sp-band (with the surface states) differ little for
various noble and transition metals. Therefore, the metal d-band—mainly the location of its
center—plays the key role in determining the adsorption properties. For example, for Pt-group
transition metals, the d-band is not completely occupied and is located in the vicinity of the Fermi
level; therefore, in the case of adsorption, the anti-bonding state will be above the Fermi level,
Figure 3.
Schematic representation of the change in the local electronic structure of an atom in the case
of adsorption on the surfaces of simple (alkali) and transition/noble metals.
Crystals 2019,9, 350 9 of 14
The interactions of the adsorbate with the sp-band (with the surface states) differ little for various
noble and transition metals. Therefore, the metal d-band—mainly the location of its center—plays
the key role in determining the adsorption properties. For example, for Pt-group transition metals,
the d-band is not completely occupied and is located in the vicinity of the Fermi level; therefore, in
the case of adsorption, the anti-bonding state will be above the Fermi level, providing remarkable
chemisorption properties of the surface. However, the d-band of noble metals (copper, silver and
gold) is below the Fermi level and is completely occupied. Therefore, in the case of adsorption, the
antibonding state will be below the Fermi level too, and the surface will be more inert (Figure 4a). In
addition, within the subgroup of noble metals, Pauli’s repulsion is maximal for gold, which is the most
inert and the “noblest of metals” due to this fact [
28
]. The binding energy of the hydrogen atom on the
surface of a gold monocrystal is less than half of the dissociation energy of a hydrogen molecule. As a
result, the layer of adsorbed H atoms is unstable.
However, experimentation on hydrogen interaction with gold nanoparticles given in this paper
and other works [
20
,
29
–
31
] demonstrates the existence of a fairly stable form of hydrogen adsorption.
How can these results be explained? PDOS calculations for clusters of noble Au
13
and transitional Ni
13
have shown that the conclusions drawn from a simple model of surface adsorption are also valid in
cases of limited systems. The d-band of the Au
13
cluster is located below the Fermi level, and after
adsorption of hydrogen its center is shifted down, and the antibonding state
σ
*is formed (Figure 4c).
In the case of Ni
13
cluster — cluster of transition metal — the d-band is in the vicinity of the Fermi
level, and the d-band shift is much smaller; therefore, no formation of the occupied antibonding state
σ*takes place (Figure 4d).
Crystals 2019, 9, x FOR PEER 9 of 14
providing remarkable chemisorption properties of the surface. However, the d-band of noble metals
(copper, silver and gold) is below the Fermi level and is completely occupied. Therefore, in the case
of adsorption, the antibonding state will be below the Fermi level too, and the surface will be more
inert (Figure 4a). In addition, within the subgroup of noble metals, Pauli's repulsion is maximal for
gold, which is the most inert and the “noblest of metals” due to this fact [28]. The binding energy of
the hydrogen atom on the surface of a gold monocrystal is less than half of the dissociation energy of
a hydrogen molecule. As a result, the layer of adsorbed H atoms is unstable.
However, experimentation on hydrogen interaction with gold nanoparticles given in this paper
and other works [20,29–31] demonstrates the existence of a fairly stable form of hydrogen
adsorption. How can these results be explained? PDOS calculations for clusters of noble Au13 and
transitional Ni13 have shown that the conclusions drawn from a simple model of surface adsorption
are also valid in cases of limited systems. The d-band of the Au13 cluster is located below the Fermi
level, and after adsorption of hydrogen its center is shifted down, and the antibonding state σ* is
formed (Figure 4c). In the case of Ni13 cluster — cluster of transition metal — the d-band is in the
vicinity of the Fermi level, and the d-band shift is much smaller; therefore, no formation of the
occupied antibonding state σ* takes place (Figure 4d).
Figure 4. Diagrams of energy levels (a,b) and PDOS representations (c,d) for metal clusters
interacting with a hydrogen atom: (a,c) Au13 cluster; (b,d) for Ni13 cluster.
However, in the case of gold nanoparticles in comparison with the bulk, the contribution of
surface states to the whole binding energy Eads increases with size reduction of nanoparticle—the
well-known “size effect”:
Figure 4.
Diagrams of energy levels (
a,b
) and PDOS representations (
c,d
) for metal clusters interacting
with a hydrogen atom: (a,c) Au13 cluster; (b,d) for Ni13 cluster.
Crystals 2019,9, 350 10 of 14
However, in the case of gold nanoparticles in comparison with the bulk, the contribution of surface
states to the whole binding energy E
ads
increases with size reduction of nanoparticle—the well-known
“size effect”:
Eads =ρ(εd)Ed+ρ(εss)Ess
ρ(εd)+ρ(εss), (5)
where
ρ
(
εd
)and
ρ
(
εss
)are the peak amplitudes of the density of d-states and surface states, and E
d
and
E
ss
are corresponding interaction energies. The
ρ
(
εd
)/
ρ
(
εss
)ratio is negligible for the surface of bulk
gold, but it increases rapidly with nanoparticle size reduction. Also note that E
ss
>E
d
. These changes
are due to the location of the “band” of the surface-states. For noble metals, the peak
ρ
(
εss
)is ~0.5 eV
lower than the Fermi level. In terms of two-state formalism, this means that the antibonding state of an
Au-H “surface molecule” is unoccupied and does not compensate for the bonding-state contribution
to the binding energy, in contrast with interaction just with the d-band states. This concept is shown
schematically in Figure 5. Thus, it is due to the interaction with surface states of gold nanoparticles
that the binding energy of the H adatom increases and provides existence of stable adsorption forms
observed in experiments.
Crystals 2019, 9, x FOR PEER 10 of 14
𝐸
=𝜌(𝜀
)𝐸
+𝜌(𝜀
)𝐸
𝜌(𝜀
)+𝜌(𝜀
) , (5)
where ρ(ε
d
) and ρ(ε
ss
) are the peak amplitudes of the density of d-states and surface states, and E
d
and
E
ss
are corresponding interaction energies. The ρ(ε
d
)/ρ(ε
ss
) ratio is negligible for the surface of bulk
gold, but it increases rapidly with nanoparticle size reduction. Also note that E
ss
> E
d
. These changes
are due to the location of the “band” of the surface-states. For noble metals, the peak ρ(ε
ss
) is ~0.5 eV
lower than the Fermi level. In terms of two-state formalism, this means that the antibonding state of
an Au-H “surface molecule” is unoccupied and does not compensate for the bonding-state
contribution to the binding energy, in contrast with interaction just with the d-band states.
This concept is shown schematicas.
Figure 5. Correlation diagram illustrating the difference in position of antibonding orbitals formed
due to the interaction of adatom with surface states and d-band states.
We now consider how the conductivity of STM tunneling contact is affected by hydrogen
adsorption on the surface of gold nanoparticles. At kT >> eU, where U is bias voltage, the Fermi
distribution
1/
)1()(
−
+=
kT
ef
ε
ε
can be given as a Heaviside function and the tunneling current can
be written as
𝐼(𝑈)=𝑐𝑜𝑛𝑠𝑡 𝑀
𝜌
(𝜀+ 𝑒𝑈)𝜌
(𝜀)𝑑𝜀
, (6)
where ρ
tip
(
ε
) is density of states of the probe, ρ
k
(
ε
) is local (projected on the k
th
atom) density of states
of the sample, and M
k
is the matrix element of wave function overlaps for the probe and the k
th
atom
of the sample. To more clearly establish the dependence we are interested in, we assume that ρ
tip
(
ε
)
has no singularities in the vicinity of the Fermi level as it is usually observed in experiments. ρ
k
(
ε
)
and M
k
can be approximated by values averaged over the region of tunneling contact:
𝑀
=𝑀=𝑐𝑜𝑛𝑠𝑡 , (7a)
𝜌
(𝜀)=𝜌
(𝜀)=1 𝐾𝜌
(𝜀)
. (7b)
The tunneling current is now determined only by the density of states of the sample:
Figure 5.
Correlation diagram illustrating the difference in position of antibonding orbitals formed due
to the interaction of adatom with surface states and d-band states.
We now consider how the conductivity of STM tunneling contact is affected by hydrogen
adsorption on the surface of gold nanoparticles. At kT >> eU, where Uis bias voltage, the Fermi
distribution
f(ε) = (1+eε/kT )−1
can be given as a Heaviside function and the tunneling current can
be written as
I(U)=const X
k
0
Z
−eU
Mkρtip(ε+eU)ρk(ε)dε, (6)
where
ρtip
(
ε
)is density of states of the probe,
ρk
(
ε
)is local (projected on the k
th
atom) density of states
of the sample, and M
k
is the matrix element of wave function overlaps for the probe and the k
th
atom
of the sample. To more clearly establish the dependence we are interested in, we assume that
ρtip
(
ε
)
has no singularities in the vicinity of the Fermi level as it is usually observed in experiments.
ρk
(
ε
)and
Mkcan be approximated by values averaged over the region of tunneling contact:
Mk=M=const , (7a)
ρk(ε)=ρsample(ε)=1/KX
k
ρk(ε). (7b)
Crystals 2019,9, 350 11 of 14
The tunneling current is now determined only by the density of states of the sample:
I(U)=const ×M×ρtip(0)
0
Z
−eU
ρsample(ε)dε. (8)
At low-bias voltages, the decisive contribution to binding energy is made by the interaction
of the H-adatom state with surface states of the gold nanoparticle. Note that the peak of surface
states decreases as hydrogen atoms fill the nanoparticle surface—adsorption is known to influence the
distribution of surface state densities [
32
]. Therefore, it is due to the reduction of density of states in the
vicinity of the Fermi level that we can detect the acts of hydrogen (deuterium) interaction (dissociative
adsorption) with gold nanoparticles in STM-STS experiments.
The density of states of individual atoms can also be calculated using DFT modeling and compared
with the VAC measured in STM experiments, and we have performed this for the abovementioned
supercluster Au
13
C
138
which is a model of a gold nanoparticle at the edge of the graphite terrace.
Summation over a set of projectors simulates a situation typical for STM experiments on metal clusters
when the curvature radius of the probe is much less than the curvature radius of the cluster and when
the tunneling current is formed by several atoms. In these cases, matrix elements of the wave function
of the probe and the cluster atoms are assumed to be the same for simplicity. The results are shown in
Figure 6.
As can be seen, in the case of D atom adsorption over the Au–C interface, the density of surface
states decreases (Figure 6a). In the case of adsorption under the interface, a decrease in the density
of surface states is also observed, and the binding energy of the adatom with the supercluster is
higher than in the previous case (Figure 6b). For adsorption under the interface, the effect of the
Au atoms’ mobility on the value of the binding energy with the D atom was also investigated. As
shown by calculations, if we allow the two nearest Au atoms to move, the binding energy will increase
by
0.46 eV (+2Au*)
, and if we remove the “freezing” from degrees of freedom for all Au atoms, the
binding energy will increase by only 0.05 eV (+13Au*). Thus, it is possible to describe the interaction
of D with a supercluster as local, since the perturbation of the atomic and electronic structure occurs on
the atoms closest to D. In Figure 6c, one can see that the adsorption of the D atom at the top of the Au
13
cluster does not lead to any noticeable decrease in the density of surface states. The binding energy
of the D atom with the supercluster is calculated for three similar cases. Note that the difference in
the adsorption energy at the interface and at the top is small (i.e., the surface migration of D adatoms
is constrained because of a significant activation barrier). Existence of this barrier was confirmed
experimentally in our previous work [
20
], where we described strong bonding between hydrogen and
gold nanoparticles on the surface of graphite and the long transition time (up to 25–30 h) of the system
to the final (stationary) state. Of course, the real values of binding energy are likely to be significantly
lower than the calculated ones, which is at least partly due to the absence of the correction of basis set
superposition error, but such comparison of calculated values is valid.
Crystals 2019,9, 350 12 of 14
Crystals 2019, 9, x FOR PEER 12 of 14
Figure 6. The projected density of electron states for various active centers in the supercluster
Au13C138: (a) adsorption over the Au–C interface; (b) adsorption under the Au–C interface; (c)
adsorption at the top of the Au13 cluster. The cases of various numbers of mobile gold atoms are
marked as “+2 Au*” and “+13 Au*”.
Thus, the results of quantum chemical calculations correlate with results of the above
experiment, which together demonstrate that at small exposures of HOPG-supported gold
nanoparticles to deuterium, adsorption occurs predominantly on the periphery of gold
nanoparticles close to the gold–carbon interface. In addition, our calculations prove the conclusions
of [20,21], where hydrogen adsorption on HOPG-supported gold nanoparticles was shown to
decrease local conductivity of tunnel nanocontact.
4. Conclusions
Experimental study and quantum chemical simulation of deuterium adsorption on
HOPG-supported gold nanoparticles were performed. The periphery of gold nanoparticles has been
found experimentally to be the preferred region for deuterium adsorption. Quantum chemical
simulation has shown deuterium adsorption in this region to be more beneficial in energy in
comparison with other sites of the gold–carbon system. With increasing exposure and saturation of
the interface adsorption sites, deuterium interacts with the rest of the gold nanoparticle surface,
gradually covering it from periphery to the center.
Figure 6.
The projected density of electron states for various active centers in the supercluster Au
13
C
138
:
(
a
) adsorption over the Au–C interface; (
b
) adsorption under the Au–C interface; (
c
) adsorption at the
top of the Au
13
cluster. The cases of various numbers of mobile gold atoms are marked as “+2 Au*”
and “+13 Au*”.
Thus, the results of quantum chemical calculations correlate with results of the above experiment,
which together demonstrate that at small exposures of HOPG-supported gold nanoparticles to
deuterium, adsorption occurs predominantly on the periphery of gold nanoparticles close to the
gold–carbon interface. In addition, our calculations prove the conclusions of [
20
,
21
], where hydrogen
adsorption on HOPG-supported gold nanoparticles was shown to decrease local conductivity of
tunnel nanocontact.
4. Conclusions
Experimental study and quantum chemical simulation of deuterium adsorption on
HOPG-supported gold nanoparticles were performed. The periphery of gold nanoparticles has
been found experimentally to be the preferred region for deuterium adsorption. Quantum chemical
simulation has shown deuterium adsorption in this region to be more beneficial in energy in comparison
with other sites of the gold–carbon system. With increasing exposure and saturation of the interface
adsorption sites, deuterium interacts with the rest of the gold nanoparticle surface, gradually covering
it from periphery to the center.
Crystals 2019,9, 350 13 of 14
Author Contributions:
Conceptualization, A.K.G. and M.V.G.; methodology, M.V.G.; formal analysis, N.V.D.;
investigation, A.K.G.; writing - original draft preparation, A.K.G. and N.V.D.; writing - review and editing, M.V.G.
and S.Y.S.; visualization, S.Y.S. and N.V.D.; supervision, B.R.S.; project administration, A.K.G. and M.V.G.; funding
acquisition, A.K.G.
Funding: This research was funded by the Russian Science Foundation, project No. 18-73-00195.
Acknowledgments:
Quantum chemical calculations were carried out using the computing power of the
Interdepartmental supercomputer center of the Russian Academy of Sciences.
Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the
study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to
publish the results.
References
1.
Haruta, M.; Yamada, N.; Kobayashi, T.; Iijima, S. Gold Catalysts Prepared by Coprecipitation for
Low-Temperature Oxidation Of Hydrogen and of Carbon Monoxide. J. Catal.
1989
,115, 301–309. [CrossRef]
2.
Wittstock, A.; Zielasek, V.; Biener, J.; Friend, C.M.; Baumer, M. Nanoporous Gold Catalysts for Selective
Gas-Phase Oxidative Coupling of Methanol at Low Temperature. Science
2010
,327, 319–322. [CrossRef]
[PubMed]
3.
Raptis, C.; Garcia, H.; Stratakis, M. Selective Isomerization of Epoxides to Allylic Alcohols Catalyzed by
TiO2-Supported Gold Nanoparticles. Angew. Chem. Int. Ed. 2009,48, 3133–3136. [CrossRef]
4.
Taylor, S.F.R.; Sa, J.; Hardacre, C. Friedel-Crafts Alkylation of Aromatics with Benzyl Alcohol over
Gold-Modified Silica. ChemCatChem 2011,3, 119–121. [CrossRef]
5.
Meyer, R.; Shaikhutdinov, S.K.; Freund, H.-J. Surface Chemistry of Catalysis by Gold. Gold Bull.
2004
,37,
72–124. [CrossRef]
6. Hashmi, A.S.K.; Hutchings, G.J. Gold Catalysis. Angew. Chem. Int. Ed. 2006,45, 7896–7936. [CrossRef]
7.
Zhu, Y.; Tian, L.; Jiang, Z.; Pei, S.; Xie, S.; Qiao, M.; Fan, K.J. Heteroepitaxial Growth of Gold on Flowerlike
Magnetite: An Efficacious and Magnetically Recyclable Catalyst for Chemoselective Hydrogenation of
Crotonaldehyde to Crotyl Alcohol. J. Catal. 2011,281, 106–118. [CrossRef]
8.
Corma, A.; Serna, P. Chemoselective Hydrogenation of Nitro Compounds with Supported Gold Catalysts.
Science 2006,313, 332–334. [CrossRef]
9.
Liu, X.; Hu, B.; Fujimoto, K.; Haruta, M.; Tokunaga, M. Hydroformylation of Olefins by Au/Co
3
O
4
Catalysts.
Appl. Catal. B 2009,92, 411. [CrossRef]
10.
Ikariya, T.; Blacker, A.J. Asymmetric Transfer Hydrogenation of Ketones with Bifunctional Transition
Metal-Based Molecular Catalysts. Acc. Chem. Res. 2007,40, 1300–1308. [CrossRef]
11.
Su, F.-Z.; He, L.; Ni, J.; Cao, Y.; He, H.-Y.; Fan, K.-N. Efficient and Chemoselective Reduction of Carbonyl
Compounds with Supported Gold Catalysts under Transfer Hydrogenation Conditions. Chem. Commun.
2008,30, 3531–3533. [CrossRef] [PubMed]
12.
He, L.; Qian, Y.; Ding, R.-S.; Liu, Y.-M.; He, H.-Y.; Fan, K.-N.; Cao, Y. Highly Efficient Heterogeneous
Gold-Catalyzed Direct Synthesis of Tertiary and Secondary Amines From Alcohols and Urea. ChemSusChem
2012,5, 621–624. [CrossRef] [PubMed]
13.
Uozumi, Y.; Hamasaka, G. Hydrochlorination of Alkynes with Titaniasupported Gold Nanoparticles. Synfacts
2017,13, 0890.
14.
Barton, D.G.; Podkolzin, S.G. Kinetic Study of a Direct Water Synthesis over Silica-Supported Gold
Nanoparticles. J. Phys. Chem. B 2005,109, 2262–2274. [CrossRef] [PubMed]
15.
Bus, E.; van Bokhoven, J.A. Hydrogen Chemisorption on Supported Platinum, Gold, and
Platinum–Gold-Alloy Catalysts. Phys. Chem. Chem. Phys. 2007,9, 2894–2902. [CrossRef] [PubMed]
16.
Shimizu, K.; Miyamoto, Y.; Kawasaki, T.; Tanji, T.; Tai, Y.; Satsuma, A.J. Chemoselective Hydrogenation of
Nitroaromatics by Supported Gold Catalysts: Mechanistic Reasons of Size-and Support-Dependent Activity
and Selectivity. Phys. Chem. C 2009,113, 17803–17810. [CrossRef]
17.
Fujitani, T.; Nakamura, I.; Akita, T.; Okumura, M.; Haruta, M. Hydrogen Dissociation by Gold Clusters.
Angew. Chem. Int. Ed. 2009,48, 9515–9518. [CrossRef] [PubMed]
18.
Boronat, M.; Illas, F.; Corma, A. Active Sites for H
2
Adsorption and Activation in Au/TiO
2
and the Role of
the Support. J. Phys. Chem. A 2009,113, 3750–3757. [CrossRef] [PubMed]
Crystals 2019,9, 350 14 of 14
19.
Panayotov, D.A.; Burrows, S.P.; Yates, J.T.; Morris, J.R. Mechanistic Studies of Hydrogen Dissociation and
Spillover on Au/TiO2: IR Spectroscopy of Coadsorbed CO and H-Donated Electrons. J. Phys. Chem. C
2011
,
115, 22400–22408. [CrossRef]
20.
Gatin, A.K.; Grishin, M.V.; Kirsankin, A.A.; Kharitonov, V.A.; Shub, B.R. Individual Nanoparticles of
Aluminum, Gold, Nickel, and Platinum Deposited on a Pyrolytic Graphite Surface. Nanotechnol. Russ.
2013
,
8, 36–45. [CrossRef]
21.
Grishin, M.V.; Gatin, A.K.; Dokhlikova, N.V.; Kirsankin, A.A.; Kulak, A.I.; Nikolaev, S.A.; Shub, B.R.
Adsorption and Interaction of Hydrogen and Oxygen on the Surface of Separate Crystalline Gold
Nanoparticles. Kinet. Catal. 2015,56, 532–539. [CrossRef]
22.
Dokhlikova, N.V.; Kolchenko, N.N.; Grishin, M.V.; Shub, B.R. Electron delocalization in heterogeneous
systems AunHm.Nanotechnol. Russ. 2016,11, 7–11. [CrossRef]
23.
Scanning Tunneling Microscopy, I. General Principles and Applications to Clean and Absorbate-Covered Surfaces;
Guntherodt, H.J., Wiesendanger, R., Eds.; Springer: Berlin, Germany, 1994; p. 280.
24.
Binnig, G.; Rohrer, H.; Gerber, C.; Weibel, E. Tunneling Through a Controllable Vacuum Gap. Appl. Phys.
Lett. 1982,40, 178–180. [CrossRef]
25. Meyer, E.; Hug, H.J.; Bennewitz, R. Scanning Probe Microscopy; Springer: Berlin, Germany, 2004; p. 210.
26.
Hamers, R.J.; Wang, Y.J. Atomically-Resolved Studies of the Chemistry and Bonding at Silicon Surfaces.
Chem. Rev. 1996,96, 1261–1290. [CrossRef] [PubMed]
27.
Hamers, R.J.; Tromp, R.M.; Demuth, J.E. Surface Electronic Structure of Si (111)-(7
×
7) Resolved in Real
Space. Phys. Rev. Lett. 1986,56, 1972–1975. [CrossRef] [PubMed]
28. Hammer, B.; Norskov, J.K. Why gold is the noblest of all the metals. Nature 1995,376, 238–240. [CrossRef]
29.
Manzoli, M.; Chiorino, A.; Vindigni, V.; Boccuzzi, F. Hydrogen Interaction with Gold Nanoparticles and
Clusters Supported on Different Oxides: A FTIR Study. Catal. Today 2012,181, 62–67. [CrossRef]
30.
Silverwood, I.P.; Rogers, S.M.; Callear, S.K.; Parker, S.F.; Catlow, C.R.A. Evidence for a Surface Gold Hydride
on a Nanostructured Gold Catalyst. Chem. Commun. 2016,52, 533–536. [CrossRef]
31.
Watkins, W.L.; Borensztein, Y. Mechanism of Hydrogen Adsorption on Gold Nanoparticles and Charge
Transfer Probed by Anisotropic Surface Plasmon Resonance. Phys. Chem. Chem. Phys.
2017
,19, 27397–27405.
[CrossRef]
32. Belen’kii, A.Y. Electronic surface states of crystals. Sov. Phys. Uspekhi 1981,24, 412–425. [CrossRef]
©
2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
