Plasmonic monitoring of catalytic hydrogen generation by a single nanoparticle probe.

Abstract

Plasmonic nanostructures such as gold nanoparticles are very useful for monitoring chemical reactions because their optical properties are highly dependent upon the environment surrounding the particle surface. Here, we designed the catalytic structure composed of platinized cadmium sulfide with gold domains as a sensitive probe, and we monitored the photocatalytic decomposition of lactic acid to generate hydrogen gas in situ by single-particle dark-field spectroscopy. The plasmon band shift of the gold probe throughout the reaction exhibits significant particle-to-particle variation, and by simulating the reaction kinetics, the rate constant and structural information (including the diffusion coefficient through the shell and the relative arrangement of the active sites) can be estimated for individual catalyst particles. This approach is versatile for the monitoring of various heterogeneous reactions with distinct components at a single-particle level.

Full text

Page 1

Plasmonic Monitoring of Catalytic Hydrogen Generation by a Single
Nanoparticle Probe
Daeha Seo, Garam Park, and Hyunjoon Song*
Department of Chemistry, Korea Advanced Institute of Science and Technology, Daejeon 305-701, Korea
*
S Supporting Information
ABSTRACT: Plasmonic nanostructures such as gold nano-
particles are very useful for monitoring chemical reactions
because their optical properties are highly dependent upon the
environment surrounding the particle surface. Here, we
designed the catalytic structure composed of platinized
cadmium sulfide with gold domains as a sensitive probe, and
we monitored the photocatalytic decomposition of lactic acid
to generate hydrogen gas in situ by single-particle dark-field
spectroscopy. The plasmon band shift of the gold probe
throughout the reaction exhibits significant particle-to-particle
variation, and by simulating the reaction kinetics, the rate constant and structural information (including the diffusion coefficient
through the shell and the relative arrangement of the active sites) can be estimated for individual catalyst particles. This approach
is versatile for the monitoring of various heterogeneous reactions with distinct components at a single-particle level.
■INTRODUCTION
Heterogeneous catalysts have been widely used in various
chemical transformations from the synthesis of chemical
resources and pollutant removal to electrochemical cells for
energy conversion. Heterogeneous catalytic systems are rather
sophisticated because the associated reactions mainly occur at
the interface between multiple phases, where the solid surface
provides active reaction sites. Numerous catalytic factors
influence the reaction rates, including the surface facets, the
active surface atoms at edges and kinks, and the atomic
arrangement, as well as the average surface area and
composition.1−6To simplify the reaction system, solid
substrates with single-crystalline faces have been employed as
model catalysts to investigate surface-dependent reaction
properties.7−10Recently, the model catalysts have been
extended to three-dimensional structures bearing metal nano-
particles and have provided valuable information regarding
actual catalytic reactions.11However, these approaches have
fundamental limitations due to ensemble average measure-
ments because the individual catalyst particles have broad
distributions of particle size and reaction environment. Another
serious challenge is in situ reaction monitoring during the
reaction process. For the detection of intermediate stages, the
reactions have been investigated using special surface
techniques, which usually require high-vacuum conditions
with powerful irradiation of light or subatomic particles,
producing situations that are far from the real reaction
conditions.8−10
Surface plasmon resonance (SPR) is a phenomenon
involving the coupling of irradiated light and conducting
electrons. It leads to intense light scattering, which is highly
dependent upon the dielectric constants of the surrounding
media.12Accordingly, it is useful in monitoring changes in the
surfaces of metallic nanostructures. Anker et al. employed SPR
for highly sensitive chemical and biomolecular detection,13and
Larsson et al. used it to continuously monitor gas-phase
reactions.14These plasmon-based techniques are advantageous
in terms of real-time, label-free, and high-resolution sensing
without requiring special treatment of the specimen.
Recent developments in surface plasmon measurement have
made it possible to study chemical properties at a single-particle
level. Novo et al. and Xu et al. reported the direct observation
of chemical reactions on individual gold nanocrystal surfa-
ces.15,16The optical property of single gold nanorods was
modulated by electrochemical charge injection, which would
guide the redox reaction mechanisms at the level of single
electrons.17Hydrogen adsorption and desorption on individual
Pd nanoparticles were also detected by introducing plasmonic
gold nanostructures.18,19
In the present study, we introduced gold probes into a
heterogeneous catalytic system and monitored the reaction
progress in situ by single-particle dark-field spectroscopy.
Specially designed catalytic structures bearing the gold probes
carried out the hydrogen generation reaction, and the SPR peak
change provided critical information regarding individual
particle properties, such as reaction kinetics and the geometry
of individual catalytic nanoparticles.
■EXPERIMENTAL SECTION
Synthesis of AuPt/CdS Hollow Cubes. Silver cubes were
synthesized via the polyol process developed by Zhang et al.20First,
Received:
Published: December 13, 2011
October 5, 2011
Article
pubs.acs.org/JACS
© 2011 American Chemical Society
1221
dx.doi.org/10.1021/ja2093663 | J. Am. Chem.Soc. 2012, 134, 1221−1227

Page 2

a silver cube dispersion in ethanol (2.6 mL, 0.19 mmol with respect to
the original silver precursor concentration) was added to a boiling
aqueous solution of PVP (Mw= 55000, 50 mL, 0.60 mmol). Next,
HAuCl4(10 mL, 0.016 mmol) and H2PtCl6(10 mL, 0.011 mmol)
aqueous solutions were added dropwise to the boiling mixture for 7
min, followed by reflux for another 3 min. After centrifugation, the
resultant AuPt−Ag hollow cubes were dispersed in water (30 mL).
Second, the aqueous dispersion of the AuPt−Ag hollow cubes (26
mL) and PVP (0.049 g, 0.44 mmol) was added to water (200 mL).
The Na2S aqueous solution (8.9 mL, 10 mM) was added, and the
reaction mixture was allowed to stir for 1 d. After centrifugation, the
resultant AuPt/Ag2S hollow cubes were dispersed in methanol (10
mL).
Third, the methanol dispersion of the AuPt/Ag2S hollow cubes (9.0
mL), Cd(NO3)2solution in methanol (8.0 mL, 8.0 mmol), and
tributylphosphine (TBP, 2.2 mL, 9.9 mmol) was added to methanol
(40 mL), and the reaction mixture was stirred for 3 d. After
centrifugation, the resultant AuPt/CdS hollow cubes were dispersed in
methanol (5.0 mL).
Synthesis of Au@Pt/CdS Cubes. Au@Ag core−shell cubes were
synthesized via the seed-mediated polyol process developed by Park et
al.21First, the Au@Ag core−shell cube dispersion in ethanol (5.0 mL,
0.076 and 0.53 mmol with respect to the original gold and silver
precursor concentrations, respectively) was added to the boiling
aqueous solution of PVP (100 mL, 0.31 mmol). Next, the H2PtCl6(12
mL, 0.0056 mmol) aqueous solution was added dropwise to the
boiling mixture for 3 min, followed by reflux for another 3 min. After
centrifugation, the resultant Au@Pt−Ag cubes were dispersed in water
(10 mL).
Second, PVP (0.017 g, 0.15 mmol) was added to the aqueous
dispersion of the Au@Pt−Ag cubes (5.0 mL). The Na2S aqueous
solution (2.2 mL, 10 mM) was added, and the reaction mixture was
allowed to stir for 1 d. After centrifugation, the resultant Au@Pt/Ag2S
cubes were dispersed in methanol (10 mL).
Third, the Cd(NO3)2solution in methanol (1.3 mL, 1.3 mmol) and
TBP (0.37 mL, 1.7 mmol) was added to the methanol dispersion of
the Au@Pt/Ag2S cubes (3.0 mL), and the reaction mixture was
allowed to stir for 3 d. After centrifugation, the resultant Au@Pt/CdS
cubes were dispersed in methanol (5.0 mL).
Monitoring Photocatalytic Decomposition of Lactic Acid.
Glass slides and coverslips were cleaned using aqua regia and piranha
solutions, and ultimately a large amount of distilled water. The
reaction chamber was formed using double-sided tape on the glass
slide. The highly diluted catalyst dispersion (∼1/1000 dilution of the
original dispersion with methanol) was cast onto the glass slide, and
the sample was allowed to dry for 1 min. For the reaction, 10% lactic
acid was added onto the reaction chamber. A coverslip covered the
chamber.
Dark-field scattering measurements were performed with an
inverted microscope (Carl Zeizz, Axiovert 40). The sample was
illuminated by a halogen lamp (35 W) using a dark-field condenser
(N.A. = 1.3) with immersion oil. Light was collected through a 43×
microscope objective lens (N.A. = 0.98) and captured by a 640 × 480
pixel color video camera (SONY, SSC-DC80) for images and a CCD
camera (ANDOR, NEWTON DU971N) with monochromator
(Dongwoo, 500i) for spectra. The extinction spectrum was integrated
for 45 s and was extracted by background subtraction and lamp
spectrum correction.
■RESULTS AND DISCUSSION
To demonstrate, we chose the photocatalytic decomposition of
lactic acid on platinized cadmium sulfide (Pt/CdS), which
yields pyruvic acid and hydrogen gas.22The Pt/CdS system is
known to efficiently decompose organic molecules due to the
large bandgap of CdS (2.410 eV). The reaction was expected to
cause a large change in the refractive index (n) between the
reactants (1.425 for lactic acid) and the products (1.416 for
pyruvic acid and 1.000 for H2gas). If the catalyst structure is a
hollow shell with the active sites located inside, the reaction
mainly occurs on the internal surface, and hydrogen gas should
accumulate in the vacancy.
To effectively probe the reaction, we introduced Au(0)
domains into the catalyst structure in two ways. The first was to
locate the Au domains in close proximity to the active Pt sites
to detect hydrogen evolution directly (Figure 1a). In this
arrangement, the reaction progress can be monitored from very
early moments, but the Au domains may interact electronically
with the active Pt sites and affect catalytic behaviors. The
second way to detect hydrogen evolution is to separate the Au
domains by some distance from the Pt sites to observe the
actual reaction progress on the Pt/CdS catalyst surface without
any interference (Figure 1b). These two distinct catalyst
designs are referred to as AuPt/CdS hollow cubes and Au@Pt/
CdS cubes, respectively.
The catalysts were synthesized via three steps. For the former
structure, partial galvanic replacement of the silver cubes20with
platinum and gold precursors yielded AuPt−Ag alloy hollow
cubes. Subsequent chalcogenization with S2−and cationic
exchange with Cd2+yielded AuPt/CdS hollow cubes. Scanning
electron microscopy (SEM), high-angle annular dark-field-
scanning electron microscopy (HAADF-STEM), high-resolu-
tion transmission electron microscopy (HRTEM), and X-ray
diffraction (XRD) data show that the structures are completely
converted from silver cubes to AuPt−Ag alloy hollow cubes,
AuPt/Ag2S hollow cubes, and eventually AuPt/CdS hollow
cubes, respectively, with the preservation of the cubic
frameworks (Supporting Information, Figures S1−S3, and
Figure 1. Structural design and synthetic scheme of catalytic particles bearing gold probes: (a) AuPt/CdS hollow cubes; (b) Au@Pt/CdS cubes.
Journal of the American Chemical Society
Article
dx.doi.org/10.1021/ja2093663 | J. Am. Chem.Soc. 2012, 134, 1221−1227
1222

Page 3

Figure 2a−c). The resultant AuPt/CdS hollow cubes have an
average edge size of 66 ± 10 nm with an average wall thickness
of 10 ± 3 nm. Individual hollow cubes have a rough surface
with coarse domains (Figure 2b). The HRTEM image shows
that the particles possess polycrystalline nature with single-
crystal domains of CdS (Figure 2c). The elemental mapping
image indicates that Au, Pt, S, and Cd components are
distributed evenly over the hollow structure (Supporting
Information, Figure S4). The energy dispersive X-ray spectros-
copy (EDS) data show that some compositional variations of Pt
and Au exist in individual particles, although the total atomic %
of Pt and Au is invariable (Supporting Information, Figure S5).
By contrast, the Au@Pt/CdS cubes are relatively uniform
both in shape and in composition. The partial galvanic
replacement reaction of Au@Ag core−shell cubes21with the
Pt precursor yielded Au@Pt−Ag alloy hollow cubes, and
chalcogenization with S2−and cationic exchange with Cd2+
afforded Au@Pt/CdS cubes. A series of characterization results
prove the complete conversion from the Au@Ag core−shell
cubes to the final Au@Pt/CdS cubes, respectively (Supporting
Information Figures S6−S8, and Figure 2d−f). The resulting
Au@Pt/CdS cubes have an average edge size of 140 ± 20 nm
with the average wall thickness of 20 ± 4 nm. The average
diameter of the Au cores is estimated to be 60 ± 11 nm. On the
basis of the SEM image and EDS analysis, each nanocube is
observed to comprise double layers of inner Pt and outer CdS
shells with a spherical Au core at the center. The double layers
of Pt and CdS are also distinct in the elemental mapping data
(Supporting Information Figures S9, S10). The outermost CdS
layers have large single-crystalline domains with a distance of
0.31 nm between the neighboring lattice fringe images,
corresponding to that of monoclinic CdS(101) (Figure 2f).
Such clear separation of the Pt and CdS layers is mainly
attributed to a combination of galvanic replacement and the
Kirkendall effect during the sulfidation reaction. The partial
galvanic replacement reaction yielded Pt layers located at the
faces of the original silver cubes. The remaining Ag(0)
components of the Pt−Ag alloy hollows reacted with S2−to
form Ag2S shells outside the Pt layers. The double-shell
formation is elucidated by the nanoscale Kirkendall effect,23in
which the difference in the diffusion coefficients of Ag (DAg=
9.981 × 10−2cm2s−1at 298 K) and S (DS= 1.901 × 10−2cm2
s−1at 298 K)24across the Pt layers leads to a net Ag flow
outward and an opposite flow of vacancies inward, forming a
hollow shell structure. The ensuing cation exchange converted
the Ag2S into the CdS shells outside the Pt layers distinctively.
In the AuPt/CdS hollow cubes, such distinctive layer formation
was not observed, presumably because of the coarse deposition
of Pt and Au together on the original silver cube surface during
the partial galvanic replacement.
The optical properties of the catalyst particles were measured
both in bulk aqueous dispersions (n = 1.333) by UV−vis
spectroscopy and in individual particles immersed in an
emersion oil (n = 1.515) by dark-field spectroscopy. The
bulk spectrum of the AuPt/CdS hollow cubes has a broad peak
at a maximum of 550 nm, while the single-particle signal
appears at 575 nm, which are consistent with each other
considering the surrounding media (Supporting Information,
Figure S11). The extinction peaks of the Au@Pt/CdS cubes
appear at 485 and 590 nm (Figure 3a, black). The former can
be assigned to that of the Pt/CdS hollow shells, by confirming
Figure 2. Electron microscopy images of catalytic particles. SEM, HAADF-STEM, and HRTEM images of (a−c) AuPt/CdS hollow cubes, and (d−f)
Au@Pt/CdS cubes.
Figure 3. Optical properties of Au@Pt/CdS cubes. (a) UV−vis and
(b) single-particle dark-field spectra of free-standing Au spheres (red),
Pt/CdS hollow cubes (blue), and Au@Pt/CdS cubes (black).
Journal of the American Chemical Society
Article
dx.doi.org/10.1021/ja2093663 | J. Am. Chem.Soc. 2012, 134, 1221−1227
1223

Page 4

the peak at 490 nm of the Pt/CdS hollow particles (blue) that
are directly synthesized from the silver cubes. The 590 nm peak
is the extinction of the Au cores, which is significantly red-
shifted as compared to that of the free-standing Au spheres with
identical diameters (542 nm, red). A similar peak shift is
observed in the single-particle dark-field spectra (Figure 3b).
The spectrum of a single particle exhibits a broad peak at 474
nm arising from the Pt/CdS hollow shell (blue), and a sharp
peak at 629 nm assignable to that of the Au core. A red-shift of
52 nm from the peak of a free-standing Au sphere (577 nm,
red) is ascribed to the plasmon coupling of the Au core and the
Pt/CdS hollow shell.25,26
To check the relationship between the refractive index of
surrounding medium (n) and the maximum plasmon peak shift
(Δλ),12the catalyst particles were immersed in four different
media, air, water, ethylene glycol (n = 1.438), and immersion
oil, respectively, and the maximum extinction peaks were
measured (Supporting Information, Figure S12). From the
slopes of the linear fits, the sensitivity factors, Δλ/Δn, were
estimated to be 34.5 nm RIU−1for the AuPt/CdS hollow cubes
and 32.9 nm RIU−1for the Au@Pt/CdS cubes, respectively.
Such low sensitivity factors as compared to those (∼60 nm
RIU−1) of Au nanoparticles27are due to the existence of other
components such as Pt and CdS around the Au probes.
On the basis of these optical properties, the two different
catalysts were employed for hydrogen gas evolution by the
photocatalytic decomposition of lactic acid, and the reactions
were monitored by single-particle dark-field spectroscopy.
Under white-light irradiation with a halogen lamp (35 W)
through a dark-field condenser (N.A. = 1.3), photons
coincident with the band gap are absorbed in the CdS shells
to generate photoelectrons, which transfer to the Pt sites, and
react with protons in solution to generate hydrogen gas. The
operational model of dark-field spectroscopy and the detection
scheme on a single particle are depicted in Figure 4a.
At t = 0, 10% lactic acid solution in water was introduced as a
proton source, and the plasmon scattering signal of an
individual particle was measured every minute. Figure 4b
shows the extinction peak change of a single AuPt/CdS hollow
cube during the reaction. The plasmon band at 569 nm
Figure 4. Monitoring of hydrogen evolution by single-particle dark-field spectroscopy. (a) Schematic diagram of single-particle measurement and
detection on a single Au@Pt/CdS catalyst particle. (b) Extinction spectra and (c) plasmon band shifts of a single AuPt/CdS hollow cube along the
reaction progress. (d) Extinction spectra and (e) plasmon band shifts of a single Au@Pt/CdS cube along the reaction progress. (c and e insets)
Kinetic plots along the reaction progress based on eq 1.
Journal of the American Chemical Society
Article
dx.doi.org/10.1021/ja2093663 | J. Am. Chem.Soc. 2012, 134, 1221−1227
1224

Page 5

gradually shifts to a shorter wavelength, and the maximum peak
shift (Δλmax) is −14 nm during the reaction. If the environment
around the Au probe is completely converted from the reactant
solution (10% lactic acid in water, n = 1.342) to hydrogen gas
(n = 1.000), the plasmon band shift is estimated to be −12 nm
by Mie theory, indicating that our measurement is reasonable
within the error range. In a control experiment, the present
reaction conditions in the absence of lactic acid did not exhibit
any plasmon shifts during the reaction period (Supporting
Information, Figure S13), which excludes the possibility of
plasmon frequency change by photoelectron injection from
CdS (ωp2= de2/meffε0, where ωpis plasmon frequency, d is
electron density, e is electron charge, meffis effective mass of the
electron, and ε0is vacuum permittivity).28After the irradiation,
the catalyst particle showed a peak shift pattern similar to that
of the fresh catalyst, indicating high stability of the catalyst
against photocorrosion.22
Figure 4c shows the relationship between the plasmon band
shift and reaction time. The effective refractive index (neff) of
the solution is given by neff= fnreact+ (1 − f)nprod, where nreact
and nprodare the refractive indexes of the reactant and products,
respectively, and f is a fraction of the reactant with respect to
the total reagent concentration.29Therefore, the plasmon shift
can be converted to f or the concentration of lactic acid, [L], by
using the sensitivity factor (34.5 nm RIU−1) of the AuPt/CdS
hollow cubes. From the point of view of the reaction kinetics,
lactic acid decomposition is a first-order reaction. If we suggest
that the hydrogen generated from the reaction is in a gaseous
state,30the hydrogen gas hinders the diffusion of the reactants
on the active Pt sites and subsequently slows the reaction rate.
This effect is reflected by the kinetic equation assuming that the
hindrance of the active sites is commensurate with the
hydrogen gas concentration. The observed reaction rate, kobs,
is then corrected to be {(A − ax)/A}·k, where A is the pre-
exponential Arrhenius factor, a is the site hindrance factor by
hydrogen gas, x is the hydrogen gas concentration, and k is the
actual reaction rate on the Pt sites. It is noted that A is a
measure of the rate at which collisions occur irrespective of the
particles’ energy. The solution to the first-order kinetic
equation with the site hindrance factor is (Supporting
Information, illustration S1):
−
⎝
xx
[L][L]
sat0sat
−−
=
⎛
⎜
⎞
⎠
⎟
xxx
x
kt
ln
[L]
satsat0
0
(1)
where [L]0is the original concentration of lactic acid, x is the
hydrogen concentration, and xsatis the hydrogen concentration
at the saturation state. The experimental data are plotted on the
basis of this equation (Figure 4c, inset), and k is estimated to be
5.3 × 10−6M−1s−1using a least-squares regression.
Figure 4d is a graph of the change in the extinction peak
during the reaction time for a single Au@Pt/CdS cube. The
Δλmaxis −4.6 nm, which is nearly one-third of the value of the
AuPt/CdS hollow cube. This is because the Au core is, on
average, 20 nm removed from the active Pt sites on the hollow
shell; thus, the surroundings around the Au core are not
completely filled with hydrogen gas even in the saturation state.
The kinetic analysis estimates k to be 2.6 × 10−6M−1s−1
(Figure 4e, inset). It is interesting that the peak shift is delayed
for 3 min after the reaction starts (Figure 3e), which is also due
to the distance between the Au probe and the actual reaction
site (vide infra).
Ten individual particles for both catalyst structures were
investigated by dark-field spectroscopy (Table 1). Figure 5a
Table 1. Estimation of Rate Constant (k), Diffusion Coefficient (D), and Pt−Au Distance (d) of 10 Individual (a) AuPt/CdS
Hollow Cubes and (b) Au@Pt/CdS Cubes
(a) Individual AuPt/CdS Hollow Cubes
rate constant k (10−6M−1min−1)
particle no.peak shift (nm)
−16.7
−14.6
−10.4
−18.5
−9.4
−6.3
−14.0
−8.1
−6.8
−11.8
−10.7 ± 4.1
reverse peak shift (nm)diffusion coefficient D (10−19m2s−1)
1
2
3
4
5
6
7
8
9
10
average
5.8
14.2
9.9
9.0
9.3
7.2
5.3
17.3
7.6
4.1
9.5 ± 4.5
(b) Individual Au@Pt/CdS Cubes
5.2
7.4
3.7
1.2
2.9 2.8
particle no.
1′
2′
3′
4′
5′
6′
7′
8′
9′
10′
average
peak shift (nm)
−4.5
−5.5
−2.9
−3.9
−4.6
−2.8
−8.1
−3.3
−4.8
−3.9
−4.4 ± 1.5
rate constant k
(10−6M−1min−1) reverse peak shift (nm)
diffusion coefficient D
(10−19m2s−1)delay time (s) Pt−Au distance d (nm)
17.8
25.3
17.2
16.9
16.6
19.2
17.1
16.3
38.6
17.8
20.3 ± 7.0
6.1
8.4
4.8
4.1
2.6
2.7
6.8
2.2
12.8
4.6
6.5 ± 3.2
3
8
3
3
4
10
2
4
12
4
2.4
2.2
1.8
2.1
1.8 1.2
2.7 2.9
5 ±3
Journal of the American Chemical Society
Article
dx.doi.org/10.1021/ja2093663 | J. Am. Chem.Soc. 2012, 134, 1221−1227
1225