Confinement-Induced Growth of Au Nanoparticles Entrapped in

Article pubs.acs.org/JPCC

Confinement-Induced Growth of Au Nanoparticles Entrapped in Mesoporous TiO2 Thin Films Evidenced by in Situ ThermoEllipsometry Eduardo D. Martínez,† Cédric Boissière,‡ David Grosso,‡ Clément Sanchez,‡ Horacio Troiani,§ and Galo J. A. A. Soler-Illia*,† †

Gerencia Química, Comisión Nacional de Energía Atómica, Centro Atómico Constituyentes, Av. Gral Paz 1499, San Martin, B1650KNA, Buenos Aires, Argentina ‡ Laboratoire de Chimie de la Matière Condensée de Paris, UMR CNRS 7574 Collège de France, Bat. C. 11, place Marcelin Berthelot, 75231 Paris Cedex 05, France § Instituto Balseiro, Comisión Nacional de Energía Atómica, Centro Atómico Bariloche, Av. Bustillo Km 9.5, San Carlos de Bariloche, 8400, Argentina S Supporting Information *

ABSTRACT: Metal-porous oxide nanocomposites present great interest in optical devices and heterogeneous catalysis. For these applications, particle shape and size control, as well as accessibility, are critical aspects. In this work, gold nanoparticles (NPs) were infiltrated into mesoporous TiO2 thin films (MTTF) by an impregnation-reduction method. In situ ellipsometry measurements were performed during thermal treatment to follow in real time the changes in the optical constants and thickness of the composites systems while being submitted to continuous heating at different rates, from room temperature up to 600 °C. Complementary characterization by UV−visible spectrophotometry, grazing incident wide angle scattering (GIWAXS), and Xray reflectometry (XRR) were performed. TEM microscopy was used to analyze the morphological changes in the composite films after the thermal treatment. Our experiments demonstrate that particle coarsening starts at temperatures below 200 °C through the processes of ripening and particle migration, leading to changes in the particle size distribution (PSD) until a mechanical restriction, due to the porous geometry, induces a change of the particle shape from spherical to ellipsoidal. This results in an internal stress that swells the mesoporous film. The effect of the gold filling fraction and the heating ramp was analyzed. A mechanism based on the kinetics of particle migration and coalescence, through the modeling of the localized surface plasmon resonance under the dipolar approximation, is proposed to explain the changes in the optical properties of these composites materials and unravel the thermal activated processes occurring in metal crystallites supported on porous oxide frameworks.

■

INTRODUCTION

the material. Mesoporous materials are indeed interesting hosts for metal nanoparticles, as they control NP size by confinement, increase NP stabilization, avoid their aggregation, provide a network of accessible channels, and add highly controlled NP−surface interactions as well as confinement and curvature effects.7−12 NPs included in mesoporous thin films present fast pore accessibility and highly controllable and reproducible NP

Noble metal nanoparticles with highly controlled size, shape, and environment constitute a very active field of research in materials chemistry. Their morphology-dependent optical or catalytic features are well understood nowadays, leading to a wide range of applications mostly in plasmon-based sensors or heterogeneous catalysts.1−4 Noble metal NPs can be produced by a variety of routes and used as the main building blocks of more complex nanostructures.5,6 This integration process frequently introduces an inorganic interface that also has to be controlled and plays a central role in the final properties of © 2014 American Chemical Society

Received: January 14, 2014 Revised: May 20, 2014 Published: May 20, 2014 13137

dx.doi.org/10.1021/jp500429b | J. Phys. Chem. C 2014, 118, 13137−13151

The Journal of Physical Chemistry C

Article

(LSPR) is extensively comprehended in the frame of the coupling between light and the conduction electrons in the confined space of NPs.35,36 Since the pioneering work of Mie regarding the absorption and scattering of small spherical particles, several modifications to the problem, like the one introduced by Gans for nonspherical shapes, the contributions of Persson37 and Kreibig38−40 on the interface effects in supported metal clusters, and the temperature effects on plasmon resonance by Doremus,41 Kreibig,42 and El-Sayed,43 provide a broad and robust theoretical framework for studying the thermal induced changes in noble metal nanoparticle supported in porous matrices. For this purpose, several concepts should be recalled. (i). Under the assumption of noninteracting particles, the extinction coefficient can be decomposed into its multipolar contributions; however, for particles smaller than ∼10 nm, the dipolar component is sufficient to describe their optical properties. The absorption spectrum related to the extinction coefficient is characterized by its amplitude, broadness, and peak position. The extinction cross-section for spherical NPs in the quasi-static approximation reads as35,36

loading. In addition, the plasmonic resonance of the Au NPs is affected by the intrinsic dielectric constant of the host mesoporous matrix, which can be finely tuned by changing the inorganic framework or the porosity. This leads to the possibility of producing systems with finely tuned plasmons due to a combination of NPs sizes and distributions as well as the matrix refractive index,13 which have recently been applied to optical sensing14 and catalysis.15 Understanding the evolution of NP size and shape under mesopore confinement along thermal treatment of such nanocomposite film systems is a central issue in modern materials chemistry. Several processes develop along thermal treatment that result in the gradual change of the nanocomposite structure, since the high surface energy of the dispersed metal phase corresponds to a local thermodynamic minimum rather than a global one. These processes are key in the final morphology of the confined NP, and can be used to attain high control of shape and size. As in colloidal systems, NP arrays submitted to high temperature evolve through coarsening processes, changing significantly the particle size distribution and the properties derived therefrom. Understanding the changes that take place upon nanocomposite heating is essential for high temperature applications, such as in catalysis or high-T sensing. In addition, a detailed understanding of the phenomena taking place upon thermal treatment in NP-mesoporous material permits control of the confined NP growth and therefore tune the plasmonic properties. In the case of supported metal clusters, because of its interesting properties in catalysis, these phenomena have been studied both theoretically and experimentally for a long time.16,17 They involve interparticle mass transfer via surface or vapor diffusion, in a process resembling Ostwald ripening, and also by means of particle migration, i.e., a surface intraparticle atom diffusion that results in a Brownian-like motion of the NPs, followed by the coalescence of colliding particles. Both of these mechanisms invoke thermally activated processes. A complete treatment of this issue was reviewed by Wynblatt and Gjostein18 explaining the different stages in nanoparticle coarsening, both thermodynamically and kinetically. The ripening models were improved considerably in a more recent work by Parker and Campbell who applied them successfully for the Au/TiO2(110) system.19 Direct observations of coarsening processes have been reported by either STM or TEM techniques on flat surfaces that mimic model catalyst systems. Goodman and coworkers20−22 have studied the coarsening of Pt and Au NPs on (110) TiO2 surfaces by STM while Parker and Campbell have reviewed the models for the sintering processes.19,23 Other interesting examples of STM studies in this system can be found,24 and in similar systems as well, like Pd/TiO2.25,26 TEM analysis performed on different systems, such as Au/CeO2,27 Au/Si3N4,28 Pt/SiO2,29 Ni/SiO2,30 and Au/SiO2,31 have also provided some insights into the coarsening mechanism. There is clear evidence and agreement that both temperature and atmosphere affect the coarsening kinetics, retarding or accelerating the rates of the ripening and migration mechanisms.22,32,33 In the case of mesoporous materials, only very recently nucleation, growth, and diffusive sintering of SnO2 within SiO2 mesofilms were analyzed using spectroscopic ellipsometry.34 In this regard, optical characterization represents an additional tool for composite materials containing noble metal nanoparticles since their localized surface plasmon resonance

Cext(λ , r , ϵm) =

24π 2 3/2 3 ϵm r λ ×

ϵ″(λ , r ) [ϵ′(λ , r ) + 2ϵm]2 + ϵ″(λ , r )2

(1)

where λ is the wavelength, ε = ε′ + iε″ is the complex dielectric function of the NPs, εm is the dielectric function of the surrounding environment, and r is the particle radius. (ii). Particle shape correlates with the absorption spectrum through the excitation conditions of LSPR; for the case of spherical particles, a well-defined band with one maximum is observed, while anisotropic particles, like rods or ellipsoids, may develop two different modes of resonance depending on the polarization of the incident light relative to the main axis of symmetry of the particles.35,44 One of the bands is located at lower plasmon energy (longer wavelength) and is associated with the longitudinal polarization mode (i.e., incident electric field in the long axis symmetry direction), while the other plasmon band, located at higher energies, is attributed to the transversal mode (i.e., perpendicular to the long axis). As the particles become more anisotropic, the spectral position of the longitudinal plasmon mode shift to longer wavelength (lower plasmon energy) and the modification of the spectrum depends on the kind of anisotropy characteristic of the NPs, e.g., oblate or prolate spheroids.35,36,44 In systems with particles in random orientations an average is detected by any large-area characterization technique. (iii). In real systems, particle size distribution (PSD) must be considered in any calculation of the absorption spectrum, especially if dynamical processes like coarsening are changing the distribution of particle size or the state of aggregation. In this regard, changes in the absorption spectrum provide relevant information about the PSD evolution, in particular, as ripening processes tend to decrease the population of smaller particles increasing the larger ones, and having in mind the major broadness of the LSPR associated with smaller particles, a ripening process should reduce the broadness of the absorption spectrum and increase its amplitude.35,36 (iv). Another way in which the metallic phase of the nanocomposites studied here may evolve is by aggregation through particle migration, leading to a final stage of 13138



Article

chamber for stabilization of the mesostructure and then submitted to a thermal treatment at 60 °C for 30 min, 130 °C for 30 min, and calcination of the organic surfactant at 350 °C for 2 h.13 After this treatment, the films were washed with acetone and immersed in an ethanol:water 1:1 v:v mixture under strong agitation in order to remove any residue from the surfactant decomposition. The films were finally dried at 130 °C. Gold Infiltration. The infiltration procedure of TiO2-F127 (TF) films entails a deposition−precipitation method consisting of the infiltration with a gold precursor and its further reduction with NaBH4. This was done by immersing the film for 1 min in a 1 mM solution of HAuCl4·3H2O (SigmaAldrich) at pH 4; after removal, the film is dried with dry air and then immersed for another minute in a 10 mM NaBH4 freshly prepared solution for the reduction of the adsorbed AuCl4− ion. This procedure is referred to as a reduction step (RS) and can be repeated as many times as desired. A pink color due to LSPR of gold nanoparticles is developed after three steps of reduction.13 Characterization. UV−vis Spectrophotometry. For the mesoporous films deposited on glass substrates the absorption band was followed by conventional UV−vis spectrophotometry in an HP Agilent 8453 spectrophotometer. Care was taken to guarantee normal incidence of the scanning beam. Ellipsometry. The mesoporous thin films without infiltration were first characterized by ellipso-porosimetry (EP). This technique allows the determination of the thickness and the optical constants n(λ) and k(λ) in the 300−950 nm range. Film porosity was determined by spectrometric ellipsometry while varying the water vapor pressure in a controlled humidity chamber. By the use of a Lorentz−Lorenz effective medium theory, a water adsorption isotherm is obtained from which the total accessible volume and, through the Kelvin equation, the pore size distribution can be extracted.56 The measurements were performed in a commercial SOPRA GES5 spectrometric ellipsometer in microspot configuration and the data was analyzed by the use of Winelli II software. Before the EP measurements, samples were washed with acetone, ethanol, and water and dried at 130 °C, in order to avoid pore blocking. The thermo-ellipsometry measurements were performed using a spectroscopic ellipsometer (Woollam M200U, software VASE) at a fixed incident angle of 70° and a polarization angle fixed at 45°. Thermal ellipsometry measurements were performed by placing the film samples on a hot plate, controlled by a home modified LINKAM heating cell. By setting the dynamical scan mode in the WVASE software, measurements were periodically taken to follow the changes in the optical parameters of the samples, as described by Bass et al.53 The data obtained was then analyzed by proposing a model for the film optical properties and fitting the model parameters to match the experimental data. For the proper modeling of the samples studied in this work, the substrate material, i.e., silicon, was first measured at different temperatures and described by a model consisting on a CauchyUrbach law for the optical range of the spectra, two TaucLorentz functions for the transitions in the UV region, and two Gaussian absorptions, provided that Kramers-Krönig relations were properly satisfied. In addition, a thin layer of SiO2, which is normally present in silicon wafers, was added to the model with the optical constant of silica taken from Palik.57 For the description of mesoporous thin films loaded with gold nanoparticles, several effective medium models were tested

coalescence, which is very fast compared to the migration kinetics;18 in the transition stage prior to coalescence, a partial clustering of particles with strong plasmon coupling is expected to produce a broadening and a red-shift of the LSPR as was experimentally observed in recent reports.45 Coalescence of particles results in a dramatic change of the PSD since the total number of particles is significantly reduced, with the final state composed of a smaller number of bigger particles. In this work, a study is presented in order to propose a mechanism of evolution of the PSD, particle shape, and the corresponding optical properties of gold NP embedded in mesoporous thin films. For that goal, a model system formed by gold infiltration within highly organized mesoporous TiO2 thin films (MTTF) has been selected. This particular system, Au/TiO2, is perhaps one of the most studied due to its well documented catalytic activity;4,46−49 especially efficient catalysts are made with gold NPs of 20−25 Å in size.47 The main features of these composites include the controlled dispersion of the metal phase into the tuned porosity of mesoporous materials. Nowadays there are well established methods for the synthesis of mesoporous oxide materials of different compositions based on organic surfactants serving as templating agents and the control of hydrolysis and condensation reactions in the sol−gel process.50−52 The particular features of mesoporous thin films, allowing multilayered structures, provide an excellent matrix for their inclusion with metal nanoparticles since both the porous volume and the film thickness can be accurately controlled. A complete treatment on the modeling of the optical properties in gold loaded mesoporous TiO2 thin films with Im3m cubic mesostructure produced through a highly reproducible method was recently presented.13 The optical characterization by UV−visible spectrophotometry and spectroscopic ellipsometry provided complementary information from which structural features, like mean particle size and filling fraction, could be extracted. In the present work, Au NPmesoporous TiO2 nanocomposites were synthesized and submitted to different thermal treatments. The change in the optical properties was characterized by UV−visible spectrophotometry and ellipsometry, and modeled in order to understand the plasmonic features. A recently developed in situ thermal ellipsometry method53 was applied to follow the kinetics of shape transformation of the included Au NP, allowing us to develop a growth model that proceeds by coarsening mechanisms.

■

EXPERIMENTAL SECTION Synthesis Procedures. TiO2 Mesoporous Thin Films. Mesoporous TiO2 thin films (MTTF) were synthesized by dipcoating according to a previously optimized procedure.54,55 The block copolymer Pluronics F127 (Aldrich, Mw ∼ 12600) used as a template was dissolved in an ethanol solution of titanium tetrachloride (Merck,TiCl4); water was added to promote hydrolysis of the Ti precursor, the condensation being controlled by the HCl formed in situ. The final molar ratio TiCl4:F127:H2O:EtOH was 1:0.005:10:40. The solution was heated up to 32 °C and was dip-coated at a withdrawal speed of 1.0 mm s−1 in a humidity controlled environment at 35%RH. Typical film thicknesses from 50 to 300 nm were obtained by varying the dip-coating withdrawal speed from 1 to 4 mm s−1. Silicon was the substrate for the ellipsometry samples and glass was the substrate for the spectrophotometer measurements. In all cases the films were preserved for 24 h in a 50% RH 13139



Article

XRD2 beamline of LNLS synchrotron in Campinas, Brazil. For GI-XRD, a fixed angle was used in order to attain an X-ray penetration length of the incident beam equal to half of the film thickness; an APD0008 detector (FMB Oxford) was programmed to scan the diffracted beams in a wide angle range (10−45°), in order to resolve the crystalline nature of the matrix and the metal particles. For the XRR measurements, the samples were placed in a homemade controlled humidity chamber in a θ−2θ configuration. The critical angle of reflection θc was obtained under dry nitrogen flux; θc2 is proportional to the electronic density of the composite film.65,66 XRR measurements of the empty mesoporous film and infiltrated films with different number of reduction steps were determined at 0% relative humidity. This is necessary in order to avoid water condensation that leads to errors in the θc determination. The filling fraction of the porosity with Au nanoparticles (FF%) was obtained from the electronic densities, as described in detail by Fuertes et al.67

including Bruggeman and Maxwell-Garnett. Although the fitting of the experimental data was good, the gold filling fraction and the intensity of the plasmon absorption could not be obtained simultaneously, as they are strongly related; therefore, a simpler model was proposed in which a CauchyUrbach dispersion law (eq 2a) is considered for TiO2 and a Lorentz oscillator (eqs 2a and 2b) accounts for the LSPR absorption of gold nanoparticles. In all the cases studied in this work the fitting of this proposed model, in the spectral range between 400 and 1000 nm, was excellent (Mean Square Error (MSE) < 10) and allowed us to determine the changes in the TiO2 matrix and those due to the infiltrated gold. 1.24 ⎛ ⎞2 B εC − U(λ) = ⎜A n + 2n + iAk e Bk ( λ − Ck)⎟ ⎝ ⎠ λ

εL(λ) =

(2a)

Amp ·Br ·En En 2 −

2

( 1.2398 λ )

− iBr ·En

■

(2b)

RESULTS AND DISCUSSION Mesoporous TiO2 Thin Films (MTTF). Under the preparation procedures used in this work, MTTF are highly organized on both types of substrates, silicon and glass. Twodimensional SAXS diagrams of MTTF performed at low angle (3°) and normal incidence (90°) shown in the Supporting Information (Figure S1) correspond to Im3m cubic mesostructures with a d110 distance of 11.9 ± 0.3 nm, in excellent agreement with previously reported MTTF synthesis through the chloride route.13,54 Figure 1 shows EP measurements of

The parameters An, Bn, Ak, and Bk of the Cauchy-Urbach model and the parameters Amp, En, and Br of the Lorentz functions describing the plasmon band were the fitting parameters adjusted to match the experimental data. Ck was fixed in 3.1 eV as an onset for TiO2 absorption. In eqs 2a and 2b, Amp corresponds to the area below the Lorentz curve, Ek(eV) to the position of the LSPR band maximum and Br(eV) to its broadness. In particular, the changes of the parameters corresponding to the plasmon band are intimately related to the gold NPs size, shape, their dielectric environment, and their coupling interactions. To compensate for the substrate temperature effect, the final results for the parameters arise from the linear average of the obtained values after fitting the data using the substrate information measured at room and final temperatures (see SI). The effective dielectric function obtained after the fitting can be translated into the absorption spectrum by the use of eq 3. σ (λ ) =

4πh 2.303λ

1 ( ϵ1, ef (λ)2 + ϵ2, ef (λ)2 − ϵ1, ef (λ)) 2 (3)

where ε1,ef and ε2,ef account for the real and imaginary part of the dielectric function, respectively, and h is the film thickness. Ellipsometry has proven to be a versatile and useful technique for the characterization of thin films containing metal NPs.58−62 In situ studies have been made by Oates and co-workers to analyze Ag NP growth in a polymer matrix63 and on SiO2 substrate,64 applying a similar model for the description of the dielectric function. Electron Microscopy (TEM-FESEM). The samples for TEM measurements were prepared by depositing a small amount of the scratched pieces of the films onto an ethanol drop placed on carbon coated copper grids. The random dispersion of the fragments occasionally leaves some of these pieces in a cross sectioned position allowing inspection of the film thickness. Images were taken in a JEOL JEM1011 operating at 120 kV and a Philips CM 200 electron microscope operating at 200 kV equipped with a LaB6 emission filament and an ultrathin objective lens for high resolution images. Field emission− scanning electron microscopy (FE-SEM) images were obtained with a ZEISS LEO 982 GEMINI field emission electron microscope using an in-lens detector to improve resolution. Grazing Incidence X-ray Diffraction (GI-XRD) and X-ray Reflectometry (XRR). Both techniques were performed at the

Figure 1. Ellipso-porosimetry measurements performed on TF film supported on silicon in the adsorption and desorption cycles. (a) Change in the refractive index at 630 nm at different values of relative humidity, (b) thickness dependence, (c) reconstruction of the adsorption isotherm, and (d) pore size distribution extracted from the isotherm data.

MTTF deposited by dip-coating on a silicon substrate at a withdrawal speed of 1.0 mm·s−1 and calcined at 350 °C. The adsorption−desorption curves correspond to a type IV isotherm presenting H2 hysteresis loops. This feature is attributable to a three-dimensional pore array with restrictions, which is a typical feature of distorted cubic mesopore systems.68 A pore volume fraction of 46% was obtained for this system with a sample to sample variation of less than 4%. The pore size distribution (PSD) calculated by applying the Kelvin equation to the isotherm data leads to a pore average radius of 5 ± 1 nm 13140



Article

TEM images of the unloaded and loaded films after ten reduction steps (Au10RS) are shown in Figure 3. As-produced

and an average interpore neck radius of 3.0 ± 0.3 nm, obtained from the adsorption and desorption curves, respectively. Gold Infiltration. The procedure of gold reduction on TF films deposited on glass substrate was studied by UV−visible spectrophotometry on films deposited on transparent substrates. Details of the infiltration procedure and the critical aspects involved in the reproducibility of the NP synthesis, together with the detailed modeling of the optical properties, can be found in our previous work.13 Figure 2a shows the

Figure 3. Electron microscopy images of the mesoporous TF films on glass before (a) and after (b) 10 RS gold infiltration. Cross section views of the loaded films were obtained by TEM (c) and FE-SEM (d).

nanoparticles are small and present isotropic morphology; coincidently, simulation of the UV−vis spectra correlates well with small, spherical nanoparticles (see below). A cross-section view (Figure 3c) demonstrates a homogeneous NP filling throughout film thickness. Similar information can be observed in the FE-SEM image in Figure 3d. Thermal Treatment at Fixed Temperatures. Nanocomposite films loaded with gold after ten reduction steps (Au10RS) were submitted to thermal treatment at a fixed temperature. A large-size, homogeneous sample was cut into four pieces; each of them was annealed in air atmosphere at different final temperatures (60, 130, 200, and 350 °C) for 2 h. After cooling, the samples were measured by UV−vis spectroscopy; absorption spectra are shown in Figure 4a. A clear red-shift was observed in all cases and the maximum absorbance value was also altered. This suggests that thermal treatment triggers important structural changes in the nanocomposite that are reflected in a change of the optical properties. GI-XRD measurements performed on the thermally treated samples display patterns compatible with metallic Au. A decrease in the full width at half-maximum of the (111) gold peak (B111) is observed for higher temperatures, that can be related to an increase in the mean particle size (D111) through the Scherrer formula69,70 (eq 4). Figure 2. Infiltration of TF/Glass system by reduction of AuCl4− with NaBH4. (a) Absorption spectra after different number of reduction steps. (b) Changes in the maximum intensity and peak position and (c) filling fraction of the porous volume calculated from XRR measurements.

Dhkl =

Kλ Bhkl cos(θhkl)

(4)

The use of this formula is a rough approximation for the real values of the particle size, since the smaller particles in a polydisperse system determine the breadth of the diffraction peak; in addition, the constant value K spans from 0.89 to 1.4 depending on the mechanical stress acting on the crystallites. For this reason, the results in particle size obtained this way tend to be smaller than the values extracted from the TEM images. In any case, the narrowing in the (111) diffraction peak with increasing annealing temperature indicates an increase in the mean particle volume. An interesting consideration is whether the particle growth mechanism is through a migration and coalescence process, where two or more particles come

obtained spectra; the intensity and the position of the LSPR band are progressively increased and red-shifted with further reduction steps (Figure 2b). Figure 2c shows an increase in the filling fraction of the porous volumeFF%, calculated from XRR measurementsalong reduction. An excellent correlation is observed between the nanoparticle contents and the increase in absorbance, which is expected from the increasing NP size, filling fraction of the films, and the plasmon coupling interaction of particles. 13141



Article

Figure 5. Electron microscopy images of the mesoporous TF-Au10RS films on glass after 2 h of treatment at 200 °C (a, b) and 350 °C (c− g).

Plane defects like stacking faults and multiple twins are clearly visible in the HRTEM measurements (Figure 5d−f). This is a strong indication of particle migration and coalescence at higher temperatures.71 Due to polydispersity in particle size, a significant change in the population sizes is difficult to assess by TEM analysis. Thermo-ellipsometry measurements, detailed in the next section, will contribute with additional in situ information about the changes at low temperatures. However, the TEM information gives valuable information about a plausible mechanism of thermal evolution: particle shape does not appreciably change below 130 °C, although coarsening processes take place at higher temperatures. Indeed, while the spherical particle volume before heating is on the order of 200 nm3, the ellipsoidal particles observed at 200 °C have a typical volume between 300 and 500 nm3 (see Table S1, SI). This major difference could only arise as a result of the coarsening of the original particles through mass transport mechanisms that must have started at temperatures between 130 and 200 °C. It should be mentioned that the orientation of the anisotropic particles, although aligned in the direction of the organized porosity, is still a short-range phenomena, micrometer scale, compared to the millimeter size of the light spot in UV−visible or ellipsometry characterization; therefore, the measured optical properties do not show anisotropy, but an average of the different orientations in the measured area. Modeling the Optical Response. The absorption spectrum shown in Figure 4a before and after the thermal treatment could be accurately modeled following the work of Sánchez et al.13 extending the Mie formulation to a Mie-Gans theory for randomly oriented oblate ellipsoidal particles.35 It should be mentioned that it is not possible to explain the observed red-shift by modeling our system as composed by a mere polydispersion of spherical particles, as particle sizes are confined in the 2−10 nm range, where the dipolar approximation holds, and the plasmon peak position is independent of size. Figure 6a,b shows simulated UV−visible spectra for prolate and oblate Au NP respectively, with different aspect ratios. The extinction cross section obtained by simulation is far more consistent with the experimental absorption spectra shown in Figure 4 for oblate ellipsoids than for prolate ones. An oblate particle shape is expected given the pore geometry after the

Figure 4. (a) UV−vis absorption spectra of TF-Au10RS submitted to thermal treatment for 2 h at the indicated final temperatures. (b) GIWAXS measurements of Au (111) peak of the samples after heating. (c) Scherrer calculation of crystallite size.

together, or if an evaporation-coarsening mechanism similar to Ostwald ripening, is taking place. In the first case, the particles, considered as monocrystalline, retain their characteristic crystal domain size and no narrowing is expected for the diffractogram peaks; on the contrary, evaporation of smaller particles and precipitation of the material on the larger ones should extend the crystal size of the latter, resulting in the narrowing of the diffraction peaks, as observed. TEM Analysis of the Thermal Treated Films. TEM of samples submitted to 80 and 130 °C show that the particle shape remains essentially spherical; no clear difference could be observed in the morphology or size of the gold particles in the loaded films compared to the unheated samples (histograms available in the SI). On the contrary, samples treated at temperatures above 200 °C show marked changes in the NP morphology. A typical change from the original spherical NP shape at low temperature to an ellipsoidal one is shown in Figure 5. Some alignment in the NP orientation coincident with the mesopore structure can be appreciated, although particles still remain disperse as isolated entities. At 350 °C, aggregated particles forming rods or wire fragments can be easily distinguished inside the porous framework (Figure 5c). 13142



Article

Figure 6. Comparison between simulation of the extinction spectra of ellipsoidal Au particles and the experimental absorption spectra from Figure 4a. Dependence of the calculated extinction cross section for prolate (a) and oblate (b) particles of varying aspect ratio h = a/b. (c) Extinction spectra of TF/Glass Au-10RS heated at the indicated temperature for 2 h and the simulated spectra considering a mean aspect ratio for the particles. The initial spherical particle radius was 3.5 nm, the dielectric constant of the surrounding medium was εm = 2.65 (n = 1.62). (d) Comparison between the spectral position of the absorption maximum as a function of temperature treatment and as a function of the aspect ratio.

uniaxial contraction of the MTTF during calcination.56 The excellent coincidence of the experimental and model spectra considering oblate ellipsoids with no preferential orientation is presented in Figure 6c. An estimation of the mean aspect ratio was obtained that is consistent with the observed particles in Figure 5b; in particular, the modeling corresponding to spherical nanoparticles (i.e., h = 1) agrees very well with the experimental spectrum of the nanocomposites at 25 °C. For the sample treated at 350 °C, the plasmon band is slightly narrower for the simulated spectrum, probably due to the enlarged particle size and dispersion in the shape of the particles as observed by TEM. Thermo-Ellipsometry Experiments. Stationary Thermal Treatments. In a first series of experiments, MTTF deposited on silicon and submitted to ten steps of reduction (TF/Si Au10RS) were placed on the hot plate of the ellipsometer. A heating program was set to rapidly heat up to the final temperature that was held stable for 2 h. Simultaneously, ellipsometry measurements were taken periodically to follow the changes in the optical properties of the composite film. The final temperatures studied were 80, 130, 200, and 350 °C. As shown in Figure 7, the plasmon peak position (parameter En) shifts during the heating process. A slight blue-shift is observed at first for temperatures above 100 °C, followed by a strong red-shift. Once the final temperature was reached, which

Figure 7. Plasmon peak position (En) evolution of TF-Au10RS heated up to the mentioned final temperatures. The arrows indicate the time at which the final temperature was reached.

is marked in the figure with vertical arrows, only small changes in the peak position were further detected. There is a positive correlation between the total red-shift and the final temperature, in coincidence with what was observed on glassdeposited samples (Figure 4a). Only a small variation in the plasmon wavelength is observed after cooling, indicating that the changes observed are irreversible and they are not the result of a temperature effect on the matrix index or the plasmon resonance. Continuous Heating Treatments. In these experiments, the Au loaded titania films (TF/Si Au10RS) were mounted in the thermal ellipsometry device described in the Experimental Section, at room temperature. A closed cup with quartz 13143



Article

Figure 8. Results of the thermo-ellipsometry performed on an empty TF film (triangular, red symbols) and a Au-loaded film after 10 reduction steps (circle, black symbols): (a) the fitting result of the thickness and the Cauchy parameters during heating; (b) the Lorentz parameters describing the LSPR in the Au loaded film (10 reduction steps).

must also be considered as suggested by El-Sayed43 and Doremus.41 According to the mentioned authors, the effect of temperature on the plasmon band, without considering coarsening processes, is to decrease slightly the amplitude of the absorption spectrum and increase its broadness; this is because of the reduced relaxation time due to higher electron− phonon scattering at higher temperatures. With this in mind, the detected decrease in the plasmon amplitude is also consistent with their observation of LSPR in gold particles for this same temperature range. In the second region (80 °C< T < 170 °C) the peak position of the LSPR is shifted progressively to lower energies, from 2.117 eV (585.6 nm) to 2.07 eV (599 nm), while the broadness decreases. Ripening and other solid state reactions such as particle migration can be the cause of this observation, as will be discussed in detail in the next section. The third region (170−260 °C) is the more interesting one; this is the temperature region in which particle shape was observed to change from spherical to ellipsoidal and the film thickness was observed to markedly increase. Several phenomena seem to be taking place in parallel. A critical temperature at 210 °C (T2) is defined when the broadness reaches a local minimum. The energy peak position changes alternatively around a value of 2.064 eV (600 nm). After this process is complete, a fourth temperature region above 270 °C is characterized by a reduction in broadness of the plasmon band without significant changes in the energy peak position. The same experiments were performed on TF samples loaded after 5 RS or 20 RS. The evolution of the optical parameters profiles was similar. The energy of the plasmon band, due to the higher plasmon interactions, is lower for the more loaded samples, consistent with the observations in the absorption spectra of Figure 2a. Some slight fluctuation in the critical temperature T2 is observed while changing the filling fraction (see SI). If the observed changes are due to clustering, coalescence, and shape distortion, the independence of the

windows was placed over the plate to enclose the system. An air flux of 0.2 L/min was connected to purge the chamber and provide a dry atmosphere. A controlled heating ramp of 5 °C/ min and a final temperature of 350 °C were set before the dynamical mode of acquisition was started. For comparison, pristine mesoporous TiO2 films (MTTF) were measured under the same conditions. Figure 8a shows the thickness evolution with temperature for pure or Au-loaded MTTF. In the case of pure MTTF, film thickness remains constant up to 250 °C, when a slight decrease is observed to start, probably due to condensation of the inorganic matrix. For the Au-loaded film, however, a much more important change takes place at approximately 200 °C increasing the film thickness by nearly 10%. The parameters concerning the Lorentz oscillator for the LSPR of the Au-loaded MTTF are displayed in Figure 8b showing the changes in amplitude, peak position, and broadness during the heating process. From Figure 8 it is clear that the amplitude of the LSPR band (Amp) and the optical parameter An present essentially the same behavior in their thermal evolution profiles. This suggests that these fitted parameters are strongly correlated and cannot be fitted independently; in fact, since the An parameter measures the constant component of the refractive index, it is deeply affected by the presence of the metal phase, which increases the refractive index of the nanocomposite, especially in the short wavelength region, where interband absorptions are present. Four different regions can be distinguished, as well as several critical temperatures where the profiles in the parameters evolution change their trends. In the first region (T < 80 °C) an increase in the LSPR energy from 2.108 eV (588 nm) to 2.117 eV (585.6 nm) can be assigned to desorption of water molecules from the pores or inorganic walls. This process decreases the dielectric constant of the environment surrounding the particles. Although water desorption by itself could explain the decrease in the amplitude of the plasmon band, a thermal effect on electron scattering 13144



Article

evolution profiles on the filling fraction suggests that the critical temperatures might be determined by other features of the MTTF matrix, such as the porous structure and pore size. Further research needs to be performed on this issue. In another set of experiments, the heating ramp during the thermal treatment was analyzed by performing thermoellipsometry measurements on different pieces of a TFAu10RS film. In each case a heating ramp of 5, 10, 20, and 50 °C/min was set in the temperature controller until a final temperature of 400 °C. The thermal evolution profiles are similar to those observed for samples heated at 5 °C/min (see SI), although some minor variation in the critical temperatures and the features of the optical changes are observed. It is interesting to note that the energy peak position of the plasmon band seems to depend entirely on the temperature but not on the heating rate, as can be extracted from the similar slopes in the En profiles; this indicates that the changes in the distribution of the metal particles occurs fast enough to reach an equilibrium condition at each temperature, as observed in the isothermal experiments. Even for the case of 50 °C/min, the slope of the change in the energy peak position is similar to the one observed for 5 °C/min; however, there appears to be a delay in the onset of the energy drop (T1) starting at 92.5, 93.7, 110.7, and 126.2 °C for the samples heated at 5, 10, 20, and 50 °C/min, respectively. Also, the energy shift produced at the critical temperature T2 is different for each case, diminishing for 20 °C/min and completely absent for the case of 50 °C/ min. Coarsening Mechanism. Up to temperature T2 (Region 2 and part of Region 3), a trend is observed: a continuous redshift of the maximum is accompanied by an increase in the peak amplitude and decrease in the LSPR broadness. This NP coarsening can be explained in principle either by mass transfer through diffusive sintering, or by means of particle migration and coalescence. Any of these two mechanisms will result in a significant change of the particle size distribution,72 increasing the mean particle size and decreasing the population of the smaller NPs. Below we discuss the evidence obtained by thermoellipsometry in light of both possible mechanisms. Regarding the LSPR broadness (Br) evolution in this region, its persistent decrease indicates a continuous NP size increase. This is consistent with both Ostwald or NP aggregation mechanisms, which cannot be distinguished from these measurements. The observed red-shift of the LSPR in region 2 (decreasing En) could be the result of different effects: on one side, aggregation of the particles, prior to their coalescence, will change the system from a homogeneous NP distribution to clusters of strongly interacting particles45,73,74 which resonate at lower energies. This effect is commonly used in colorimetric sensors.2 The presence of this effect supports a particle migration mechanism. Although the temperatures invoked here are lower than the ones usually reported in coarsening studies,20−30 an additional effect of the high curvature framework in which the particles are located must be considered, which is different from the usually flat surface analyzed in common studies. This high curvature could actually act as an additional capillary driving force to confine the metal phase into the pores.18 Even though the Tammann temperature, which sets the starting temperature at which bulk mobility of the metal particles becomes measurable, is low for the case of gold (395 °C),75 the thermal stability of supported gold NPs is known to be strongly dependent on the support

material, being more stable when Au is directly bonded to Si atoms in SiO2 than to Ti atoms in TiO2.76 Another way in which the particle evolution could result in the observed red-shift is by increasing the interface damping parameter (A parameter) present in the description of the damping frequency, γ(r, A) = γ∞ + AvF/r, in the dielectric function of NPs.37,39,13 As the surface contact between the particles and the TiO2 matrix may change during particle growth, the interface damping will change proportionally: for the case of Ag NPs on TiO2, the A parameter was reported to change from 0.6 for simply deposited spherical particles to 1.8 for fully embedded particles.40 Since both NP growth mechanisms (ripening and migration) could in principle be present, it is not possible to distinguish which is the main process governing each region only from the thermoellipsometry results. With the optical information presented in Figure 4a and the nanostructure revealed by TEM at specific temperatures (Figure 5) taken as a reference, this information leads to the suggestion that the system evolves through a complex pathway, in which a combination of several mechanisms with different activation energies can take place depending on the temperature. However, thermoellipsometry is a useful tool to determine the activation energies of a dominant process in a given region, as will be discussed below. Kinetic Analysis of the Second Region. In order to ascertain the activation energies of the intermediate temperatures, in which growth processes begin to take place, the thermal behavior in region 2 for samples heated at a rate of 5 °C/min can be analyzed on the basis of kinetic models for solid state reactions. Here, we follow the procedure proposed by PérezMaqueda77 by defining the degree of conversion α from an initial to a final state (0 < α < 1). The rate of the reaction follows the Arrhenius temperature dependence as dα = Af (α) e−Ea / RT dt

(5)

where dα/dt is the rate of conversion that can be expressed as dα/dt = βdα/dT, with β = dT/dt the heating rate. Ea is the activation energy and R the gas constant. Equivalently, eq 5 can be written as ⎛ dα ⎞ ln⎜ /f (α)⎟ = cA − Ea /RT ⎝ dt ⎠

(6)

The kinetic model associated with the reaction under study can be written in the Sestak-Berggren parametrization77 f(α) = c(1 − α)nαm. As observed above, the amplitude of the absorption cross section is not a viable parameter to follow NP growth, because it is not possible to separate it from the refractive index. Thus, we found that the most appropriate way to define the converted fraction for the second region is to consider it as a function of the energy of the plasmon peak position, En. Therefore, α(En) = 1 − (Enf − En)/(Enf − En0), where Enf is the final plasmon energy, in this case the minimum observed at approximately 200 °C in Figure 8b, while En0 is the initial plasmon energy (the maximum En at ∼80 °C). The evolution of the converted fraction versus time was analyzed in the framework of five kinetic models of the most common solid state reaction mechanisms (see details in SI).77 The kinetic parameters n and m were adjusted in the search of the best linear correlation between the left and right terms of eq 6.77,78 The best correlation corresponds to the Jander equation 13145



Article

that accounts for three-dimensional diffusion limited particle growth (D3). From the slope, the activation energy of the process can be estimated. The value obtained was (68 ± 5) kJ/ mol (see SI, Figure S14). However, the validity of the Jander model is very restricted, and therefore it is usually replaced by the more appropriate Ginstlein and Brounshtein (G-B) model (D4), also corresponding to 3D diffusion controlled growth.78 The result of such an analysis is shown in Figure 9. The

Figure 10. Results of the thermo-ellipsometry performed on TF Au10RS sample at 20 °C/min: (a) the results in the fitting of the thickness and the Cauchy parameters, and (b) the Lorentz parameters describing the LSPR.

°C (see Figure 8) are found; however, a sudden change in all of the fitting parameters was observed to happen at a critical temperature of approximately 525 °C, characterized by a remarkable thickness decrease and a significant red-shift of the plasmon band from 2.086 eV (595 nm) to 1.958 (633 nm). An increase in the amplitude of the plasmon band (and the coupled An parameter) as well as a reduction in its broadness are also detected. An additional characterization of this sample by SEM (Figure 11) suggests a clear explanation for this observation: large particles of approximately 50 nm in size, that were not present

Figure 9. Kinetic analysis for the second region of Figure 8. The Pearson Correlation Coefficient (PCC) and the values of n and m are indicated in the figures. The activation energy (Eact) was calculated from the slope of the linear fitting. The error arises from the dependence with the fitting range.

activation energy obtained for this model was (49 ± 5) kJ/mol. This value is much lower than those reported by Parker et al. for the Au/TiO2 system under ultra-high-vacuum conditions19(260−330 kJ/mol) but higher than the ones (10 ± 2) kJ/mol found by Yang et al. for the same system under CO atmosphere.22 The strong dependence of the gas environment on the sintering kinetics is well-known; indeed, the major difference between the cited authors was attributed to the CO oxidation.71 Similar differences were also reported by Turba et al. for the case of Au deposited on GaN under CO atmosphere (27.4 kJ/mol).79 In our work the kinetic experiments were performed under 1 bar, dry air. To the best of our knowledge, activation energies of Au growth on comparable nonmesoporous titania coatings under air atmosphere are not reported, which poses difficulties in making an exact comparison between our kinetic data and the previous reports. Notwithstanding, the low activation energy obtained here suggests that confinement and curvature effects due to the mesopore system might also have a role in the transport kinetics, as analyzed in detail in ref 18. In the treatment of the diffusive sintering mechanism there are two limiting cases to be considered: (a) growth is limited by diffusion, and (b) growth is limited by the phase boundary reaction. The very good fit of our kinetic data to a diffusiondetermining mechanism supports the idea that diffusionlimiting growth is an accurate description of our system. Yet, it must be remembered that although the particle growth is a 3D diffusional process, it is not possible to solve if the mass transport is through monomers (atoms on the gas phase) or particle diffusion. In the temperature range of T2, it is clear however that NPs grow anisotropically, and that the NP size is limited by the mesopore topology. Thermal Treatment to High Temperatures. More interesting is the extension of thermoellipsometry experiments to a higher final temperature (Figure 10). In this case a heating ramp of 20 °C/min was set up to a final temperature of 600 °C. Profiles very similar to those found for temperatures below 350

Figure 11. FE-SEM image of the surface of the gold loaded TF film before and after the thermo-ellipsometry measurement at a final temperature of 600 °C. 13146



Article

Figure 12. Schematic summary of the different processes induced by thermal treatment on Au TF/Si.

before, appear on the film surface after the thermal treatment, indicating an expulsion of the metallic phase from the mesopores. This phenomenon explains the thickness contraction after Au expulsion. The fact that the final thickness is slightly lower than the original one is not surprising, since at these high temperatures the mesoporous structure can suffer a certain collapse due to extended anatase recrystallization.53 Also during the coalescence stage, some damage or cracking on the film structure could be expected. Additionally, the observed particles on the surface are larger than the limit of validity of the dipolar approximation (d < 10 nm), and multipolar excitation should be considered to account for retardation effects. These particles are well-known to resonate at lower energies.35,36 In addition, the broadness corresponding to larger particles, as mentioned before, should become narrower. This phenomenon coincides with the observations displayed in Figure 10. It should be mentioned that in some of the experiments the swelling of the film thickness produced in region 2 was less pronounced or even shows a decrease after the initial expansion stage. This could arise as an expulsion of the metallic phase from the porous framework, perhaps only from the top layers of pores, resembling a segregation process. It is worth mentioning that the dilatation coefficients of both gold and titania (αAu = 14.2 × 10−6 1/K, αTiO2 = 10.2 × 10−6 1/K) are insufficient to explain the changes in thickness, given that at 400 °C, for example, the expansion would be less than 1 nm. Picture of NP Growth in Confinement. The crossed analysis of thermoellipsometry and TEM information sheds light on the mechanisms of transformation of Au nanoparticles included in the mesoporous matrix. At T < 100 °C, water desorption induces a slight blue-shift in the LSPR peak position. Interestingly, NP growth proceeds even at relatively low temperatures, the activation energies being lower than those measured for flat surfaces under nonreactive gas flow. At moderate temperatures, a progressive red-shift and a reduction in its broadness occurs until a critical temperature (T2) at ∼210 °C is reached. Below T2, the initially isotropic particles present a diffusion-controlled growth that leads to elongated Au NP,

the shape of which seems to be guided by the mesopore. At these low temperatures Ostwald-like ripening could be presumed since the energy barrier for NP migration is probably larger. When T2 is reached, the film thickness was observed to increase significantly. This and the TEM images sustain the hypothesis of a restricted coarsening mediated by the porous geometry, swelling the film as a result of a mechanical stress. For higher temperatures, a fast migration process leads first to NP growth until a small fraction of the pores is completely filled, and in a second stage, to the growth of anisotropic chainlike NP arrangements. This morphological change, which is clearly displayed in Figure 5a,b, possibly takes place through migration of smaller NP. If the detailed growth mechanism is to be ascribed, at a certain point this will lead to colliding particles that will rapidly coalesce to form larger ones. However, these processes find a restriction due to the limited pore size, opposing a mechanical strength that produces the swelling of the film and could explain the observed increase in thickness (Figure 8a). This situation initiates the third region. As a result of the opposing forces, spheroidal particles are no longer possible, so the particles filling the pores match their ellipsoidal shape. This can explain the sudden increase in the plasmon broadness, associated with the presence of oblate spheroids. Particle migration and coalescence seems to continue at the higher temperatures of the fourth region (T > 250 °C), connecting particles in adjacent pores that end up forming rods or bone-like stretched particles inside the porous framework at T = 350 °C. High resolution TEM images support the particle migration mechanism, revealing the planar defects and multiple twins, characteristic of coalescence.71,80−82 It can be seen that colliding particles retain their crystalline structure, but recrystallization might also occur (see Figure 5d−g and additional images on SI). The spatial connectivity and the topology of the anisotropic nanoparticles are directed by the mesopore structure. Finally, a higher critical temperature at ∼525 °C was found to correspond to the collapse of the mesostructure and the rejection of the metal phase, producing 13147



Article

larger (30−40 nm) Au particles on the surface of the film. The different stages are schematically represented in Figure 12. The kinetic analysis indicates a 3D growth mechanism limited by diffusion, probably more related to Ostwald ripening at T < 210 °C and through NP attachment for higher temperatures. Migration could be favored by the high curvature structure of the mesoporous support since the temperatures required to alter the nanostructure of the particles are lower than the ones usually reported. The observed profiles are similar for different filling fractions of the films and relatively independent of the heating ramps (see SI).

5. In confined systems, like the one studied here, the coarsened particles can mimic the internal porous morphology, giving rise to a completely different system formed by anisotropic particles which are difficult to synthesize otherwise. 6. Other phenomena such as particle rejection from the porous framework occur, and are determined by the interconnectivity of the mesoporosity and the typical particle size. 7. The thermal stability of the mesoporous oxide should also be taken into account since it limits the maximum temperature at which coarsening treatments can be performed. We state that further research should be performed to establish the influence of the mesopore structure, like the pore and neck sizes, the role of the atmosphere, the composition of the metallic NPs, and the physical chemistry nature of the metal/oxide interface. In summary, we have shown the thermal evolution of a nanocomposite film system through different techniques, focusing on the use of ellipsometry as a powerful in situ technique for the determination of the optical properties in real time. Thermo-ellipsometry can be used to detect critical temperatures at which the system change and the results could be explained according with coarsening and plasmon resonance theories. This work provides a framework and a predictive tool for similar “metal NP@mesopore” systems in which thermal behavior is relevant, as in the case of metal loaded solar cells, plasmon-based sensors, supported metallic catalysts, and selfassembled multilayered systems. These findings are also relevant for the development of synthesis procedures of nanocomposite systems by a “dynamic nanocasting” route, in which the thermal treatments and their influence on growth kinetics can be used as an additional tool to control and direct the nanostructure of the metal phase, using mesoporous scaffolds to template the NPs shape and anisotropy.

■

CONCLUSIONS Cross utilization of TEM, UV−visible spectroscopy, and in situ thermoellipsometry permits a clear view of the thermal evolution of a nanocomposite system formed by Au NPs included within a mesoporous titania thin film. A relatively simple model can account for the changes in the LSPR when these nanocomposites are submitted to thermal treatment in the range 25−550 °C, which are due to NP growth. Significant changes in particle shape, from spherical to ellipsoidal, take place for samples treated at temperatures below 200 °C. At higher temperatures, elongated metallic particles presenting twins and planar defects appear aligned inside the pore channels, suggesting a restricted coalescence of the initially isolated particles. Based on the real time thermo-ellipsometry coupled with modeling, a mechanism of thermal evolution is proposed. The evolution steps imply (a) water desorption (T < 100 °C), (b) NP growth (100−210 °C), and (c) extensive coarsening mediated by the pore geometry, starting at a critical temperature of ∼210 °C. NP coalescence for T > 250 °C is directed by the mesopore array topology, and probably achieved by particle migration and oriented attachment. At a critical temperature of ∼525 °C, the mesostructure collapses, and larger Au NP (30−40 nm) are expelled to the surface of the film. The growth mechanism proposed here seems to be the starting point of the formation of a mesocrystal by a kind of “confinement-induced attachment”. Particle migration seems to be the main coarsening mechanism, although Ostwald-like ripening can also be present in the lower temperature range. Interestingly, the observed phenomena are similar for different filling fractions of the films and relatively independent of the heating ramps. Based on our observations and previous works by other groups we conclude the following general issues concerning metal NPs supported in mesoporous oxides: 1. Optical changes can be observed even at low temperatures due to water desorption or any other possible interface reaction (e.g., oxidation, molecule detachment, etc.) that leads to modification of the NP dielectric environment. 2. First stages of reconstruction of the metallic phase occur at temperatures lower than the Tamman temperature and also lower than the observed critical temperatures for reconstruction of NPs supported on flat oxide surfaces, suggesting an influence of the confined environment provided by the mesopores. 3. Ostwald ripening and particle migration can both contribute to the coarsening mechanism; however, the identification of twins and stacking faults in the reconstructed particles are indicative of particle migration and coalescence. 4. The atmosphere under which the reconstruction takes place can significantly affect the activation energies of the different coarsening mechanism.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional TEM images, details of the TEM analysis, EDS spectra and quantification of the Au/Ti fraction in the TFAu10RS sample, FE-SEM images of the film surface before and after the continuous heating experiments, XRR curves showing the critical angle after different isothermal treatments, pictures of the experimental setup used in the ellipsometric studies, an example of the fitting accuracy, the linear average used for the substrate temperature correction, the thermoellipsometry profiles for samples loaded with different number of reduction steps and under different heating rates, and a reconstruction of the absorbance spectra from the ellipsometric data, calculated through eq 3. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Author Contributions

All authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest. 13148



■

Article

(17) Bartholomew, C. H. Mechanisms of Catalyst Deactivation. Appl. Catal. A: Gen. 2001, 212, 17−60. (18) Wynblatt, P.; Gjostein, N. A. Supported Metal Crystallites. Prog. Solid State Chem. 1975, 9, 21−58. (19) Parker, S.; Campbell, C. Kinetic Model for Sintering of Supported Metal Particles with Improved Size-Dependent Energetics and Applications to Au on TiO2 (110). Phys. Rev. B 2007, 75, 035430. (20) Lai, X.; Goodman, D. W. Scanning Microscopy Studies of Metal Supported on TiO2 (110): Morphology and Electronic Structure. Prog. Surf. Sci. 1998, 59, 25−52. (21) Han, P.; Goodman, D. W. Controlling the Size and Distribution of Pd-Au Nanoparticles on TiO2 (110). J. Phys. Chem. C 2008, 112, 6390−6397. (22) Yang, F.; Chen, M. S.; Goodman, D. W. Sintering of Au Particles Supported on TiO2 (110) during CO Oxidation. J. Phys. Chem. C 2009, 113, 254−260. (23) Parker, S. C.; Campbell, C. T. Reactivity and Sintering Kinetics of Au/TiO2(110) Model Catalysts: Particle Size Effects. Top. Catal. 2007, 44, 3−13. (24) Mitchell, C. E. J.; Howard, A.; Carney, M.; Egdell, R. G. Direct Observation of Behaviour of Au Nanoclusters on TiO2 (110) at Elevated Temperatures. Surf. Sci. 2001, 490, 196−210. (25) Stone, P. Scanning Tunnelling Microscopy Investigation of Sintering in a Model Supported Catalyst: Nanoscale Pd on TiO2 (110). Chem. Commun. 1998, 2, 1369−1370. (26) Jak, M. J. J.; Konstapel, C.; van Kreuningen, A.; Verhoeven, J.; Frenken, J. W. M. Scanning Tunnelling Microscopy Study of the Growth of Small Palladium Particles on TiO2 (110). Surf. Sci. 2000, 457, 295−310. (27) Akita, T.; Tanaka, K.; Kohyama, M.; Haruta, M. Analytical TEM Study on Structural Changes of Au Particles on Cerium Oxide Using a Heating Holder. Catal. Today 2007, 122, 233−238. (28) Walsh, M. J.; Yoshida, K.; Gai, P. L.; Boyes, E. D. In-situ Heating Studies of Gold Nanoparticles in an Aberration Corrected Transmission Electron Microscope. J. Phys. Conf. Ser. 2010, 241, 012058. (29) Simonsen, S. B.; Chorkendorff, I.; Dahl, S.; Skoglundh, M.; Meinander, K.; Jensen, T. N.; Lauritsen, J. V.; Helveg, S. Effect of Particle Morphology on the Ripening of Supported Pt Nanoparticles. J. Phys. Chem. C 2012, 116, 5646−5653. (30) Banerjee, R.; Crozier, P. A. In Situ Synthesis and Nanoscale Evolution of Model Supported Metal Catalysts: Ni on Silica. J. Phys. Chem. C 2012, 116, 11486−11495. (31) De Marchi, G.; Mattei, G.; Mazzoldi, P.; Sada, C.; Miotello, A. Two Stages in the Kinetics of Gold Cluster Growth in Ion-Implanted Silica During Isothermal Annealing in Oxidizing Atmosphere. J. Appl. Phys. 2002, 92, 4249. (32) Bechelany, M.; Maeder, X.; Riesterer, J.; Hankache, J.; Lerose, D.; Christiansen, S.; Michler, J.; Philippe, L. Synthesis Mechanisms of Organized Gold Nanoparticles: Influence of Annealing Temperature and Atmosphere. Cryst. Growth Des. 2010, 10, 587−596. (33) Ouyang, R.; Liu, J.-X.; Li, W.-X. Atomistic Theory of Ostwald Ripening and Disintegration of Supported Metal Particles Under Reaction Conditions. J. Am. Chem. Soc. 2013, 135, 1760−1771. (34) Lepoutre, S.; Smått, J.-H.; Laberty, C.; Amenitsch, H.; Grosso, D.; Lindén, M. Detailed Study of the Pore-Filling Processes During Nanocasting of Mesoporous Films Using SnO2/SiO2 as a Model System. Microporous Mesoporous Mater. 2009, 123, 185−192. (35) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light by Small Particles; Wiley, 1983. (36) Kreibig, U.; Vollmer, M. Optical Properties of Metal Clusters; Springer, 1995. (37) Persson, B. N. J. Polarizability of Small Spherical Metal Particles: Influence of the Matrix Environment. Surf. Sci. 1993, 281, 153−162. (38) Hövel, H.; Fritz, S.; Hilger, A.; Kreibig, U.; Vollmer, M. Width of Cluster Plasmon Resonances: Bulk Dielectric Functions and Chemical Interface Damping. Phys. Rev. B 1993, 48, 18178−18188. (39) Kreibig, U.; Gartz, M.; Hilger, A.; Neuendorf, R. Interfaces in Nanostructures: Optical Investigations on Cluster-Matter. Acta Metallogr. 2000, 11, 1335−1342.

ACKNOWLEDGMENTS This work was supported by ANPCyT (PICT 1848 and 2087), ECOS-MINCyT binational program (A08E05). Authors thank ABTLUS for funding access to the LNLS (lines XRD2 and D11A-SAXS2). G.J.A.A.S.-I. is a member of CONICET Scientific staff and Gabbo’s. E.D.M. acknowledges a doctoral fellowship from CONICET.

■

ABBREVIATIONS EISA, evaporaction-induced self-assembly; EP, ellipsoporosimetry; FF%, filling fraction of the mesopores; LSPR, localized surface plasmon resonance; MTTF, mesoporous titania thin films; NP, nanoparticles; PSD, particle size distribution; RS, reduction steps

■

REFERENCES

(1) Grzelczak, M.; Pérez-Juste, J.; Mulvaney, P.; Liz-Marzán, L. M. Shape Control in Gold Nanoparticle Synthesis. Chem. Soc. Rev. 2008, 37, 1783−1791. (2) Stewart, M. E.; Anderton, C. R.; Thompson, L. B.; Maria, J.; Gray, S. K.; Rogers, J. A.; Nuzzo, R. G. Nanostructured Plasmonic Sensors. Chem. Rev. 2008, 108, 494−521. (3) Schauermann, S.; Nilius, N.; Shaikhutdinov, S.; Freund, H.-J. Nanoparticles for Heterogeneous Catalysis: New Mechanistic Insights. Acc. Chem. Res. 2013, 46, 1673−1681. (4) Daniel, M.-C.; Astruc, D. Gold Nanoparticles: Assembly, Supramolecular Chemistry, Quantum-Size-Related Properties, and Applications Toward Biology, Catalysis, and Nanotechnology. Chem. Rev. 2004, 104, 293−346. (5) Rycenga, M.; Cobley, C. M.; Zeng, J.; Li, W.; Moran, C. H.; Zhang, Q.; Qin, D.; Xia, Y. Controlling the Synthesis and Assembly of Silver Nanostructures for Plasmonic Applications. Chem. Rev. 2011, 111, 3669−3712. (6) Centi, G.; Perathoner, S. Creating and Mastering Nano-Objects to Design Advanced Catalytic Materials. Coord. Chem. Rev. 2011, 255, 1480−1498. (7) Ren, Y.; Ma, Z.; Bruce, P. G. Ordered Mesoporous Metal Oxides: Synthesis and Applications. Chem. Soc. Rev. 2012, 41, 4909−4927. (8) Prati, L.; Villa, A. Gold Colloids: From Quasi-Homogeneous to Heterogeneous Catalytic Systems. Acc. Chem. Res. 2014, 47, 855−863. (9) Taguchi, A.; Schüth, F. Ordered Mesoporous Materials in Catalysis. Microporous Mesoporous Mater. 2005, 77, 1−45. (10) Luque, R.; Garcia Martinez, J. Editorial: From Mesoporous Supports to Mesoporous Catalysts: Introducing Functionality to Mesoporous Materials. ChemCatChem 2013, 5, 827−829. (11) An, K.; Musselwhite, N.; Kennedy, G.; Pushkarev, V. V.; Baker, L. R.; Somorjai, G. A. J. Preparation of Mesoporous Oxides and Their Support Effects on Pt Nanoparticle Catalysts in Catalytic Hydrogenation of Furfural. Colloid Interface Sci. 2013, 392, 122−128. (12) White, R. J.; Luque, R.; Budarin, V. L.; Clark, J. H.; Macquarrie, D. J. Supported Metal Nanoparticles on Porous Materials. Methods and Applications. Chem. Soc. Rev. 2009, 38, 481−494. (13) Sánchez, V. M.; Martínez, E. D.; Martínez Ricci, M. L.; Troiani, H.; Soler-Illia, G. J. A. A. Optical Properties of Au Nanoparticles Included in Mesoporous TiO2 Thin Films: A Dual Experimental and Modeling Study. J. Phys. Chem. C 2013, 117, 7246−7259. (14) López-Puente, V.; Abalde-Cela, S.; Angelomé, P. C.; AlvarezPuebla, R. A.; Liz-Marzán, L. M. Plasmonic Mesoporous Composites as Molecular Sieves for SERS Detection. J. Phys. Chem. Lett. 2013, 4, 2715−2720. (15) Rafti, M.; Brunsen, A.; Fuertes, M. C.; Azzaroni, O.; Soler-Illia, G. J. A. A. Heterogeneous Catalytic Activity of Platinum Nanoparticles Hosted in Mesoporous Silica Thin Films Modified with Polyelectrolyte Brushes. ACS Appl. Mater. Interfaces 2013, 5, 8833−8840. (16) Granqvist, C. G.; Buhrman, R. A. Ultrafine Metal Particles. J. Appl. Phys. 1976, 47, 2200−2219. 13149



Article

(40) Hilger, A.; Tenfelde, M.; Kreibig, U. Silver Nanoparticles Deposited on Dielectric Surfaces. Appl. Phys. B: Lasers Opt. 2001, 73, 361−372. (41) Doremus, R. H. Optical Properties of Small Gold Particles. J. Chem. Phys. 1964, 40, 2389−2396. (42) Kreibig, U. Electronic Properties of Small Silver Particles: the Optical Constants and Their Temperature Dependence. J. Phys. F: Met. Phys. 1974, 4, 999−1014. (43) Link, S.; El-Sayed, M. A. Size and Temperature Dependence of the Plasmon Absorption of Colloidal Gold Nanoparticles. J. Phys. Chem. B 1999, 103, 4212−4217. (44) Link, S.; Wang, Z. L.; El-Sayed, M. A. Alloy Formation of Gold− Silver Nanoparticles and the Dependence of the Plasmon Absorption on Their Composition. J. Phys. Chem. B 1999, 103, 3529−3533. (45) Chung, H. Y.; Xie, H. Y.; Leung, P. T.; Tsai, D. P. Optical Properties of Metallic Nanoshell Composites: The Effects of Temperature and Particle Clustering. Solid State Commun. 2009, 149, 2151−2155. (46) Haruta, M.; Daté, M. Advances in the Catalysis of Au Nanoparticles. Appl. Catal. A: Gen. 2001, 222, 427−437. (47) Haruta, M. Size- and Support-Dependency in the Catalysis of Gold. Catal. Today 1997, 36, 153−166. (48) Jones, M. R.; Osberg, K. D.; Macfarlane, R. J.; Langille, M. R.; Mirkin, C. A. Templated Techniques for the Synthesis and Assembly of Plasmonic Nanostructures. Chem. Rev. 2011, 111, 3736−3827. (49) Fukuoka, A.; Araki, H.; Kimura, J.; Sakamoto, Y.; Higuchi, T.; Sugimoto, N.; Inagaki, S.; Ichikawa, M. Template Synthesis of Nanoparticle Arrays of Gold, Platinum and Palladium in Mesoporous Silica Films and Powders. J. Mater. Chem. 2004, 14, 752−756. (50) Sanchez, C.; Boissière, C.; Grosso, D.; Laberty, C.; Nicole, L. Design, Synthesis, and Properties of Inorganic and Hybrid Thin Films Having Periodically Organized Nanoporosity. Chem. Mater. 2008, 20, 682−737. (51) Soler-Illia, G. J. A. A.; Sanchez, C.; Lebeau, B.; Patarin, J. Chemical Strategies to Design Textured Materials: from Microporous and Mesoporous Oxides to Nanonetworks and Hierarchical Structures. Chem. Rev. 2002, 102, 4093−4138. (52) Berggren, A.; Palmqvist, A. E. C.; Holmberg, K. SurfactantTemplated Mesostructured Materials From Inorganic Silica. Soft Matter 2005, 1, 219−226. (53) Bass, J. D.; Grosso, D.; Boissiere, C.; Sanchez, C. Pyrolysis, Crystallization, and Sintering of Mesostructured Titania Thin Films Assessed by In-Situ Thermal Ellipsometry. J. Am. Chem. Soc. 2008, 130, 7882−7897. (54) Crepaldi, E. L.; Soler-Illia, G. J. A. A.; Grosso, D.; Cagnol, F.; Ribot, F.; Sanchez, C. Controlled Formation of Highly Organized Mesoporous Titania Thin Films: From Mesostructured Hybrids to Mesoporous Nanoanatase TiO2. J. Am. Chem. Soc. 2003, 125, 9770− 9786. (55) Soler-Illia, G. J. A. A.; Angelomé, P. C.; Fuertes, M. C.; Grosso, D.; Boissiere, C. Critical Aspects in the Production of Periodically Ordered Mesoporous Titania Thin Films. Nanoscale 2012, 4, 2549− 2566. (56) Boissiere, C.; Grosso, D.; Lepoutre, S.; Nicole, L.; Bruneau, A. B.; Sanchez, C. Porosity and Mechanical Properties of Mesoporous Thin Films Assessed by Environmental Ellipsometric Porosimetry. Langmuir 2005, 21, 12362−12371. (57) Palik, E. D. Handbook of Optical Constants of Solids; Elsevier, 1998. (58) Losurdo, M.; Bergmair, M.; Bruno, G.; Cattelan, D.; Cobet, C.; de Martino, A.; Fleischer, K.; Dohcevic-Mitrovic, Z.; Esser, N.; Galliet, M.; et al. Spectroscopic Ellipsometry and Polarimetry for Materials and Systems Analysis at the Nanometer Scale: State-of-the-Art, Potential, and Perspectives. J. Nanopart. Res. 2009, 11, 1521−1554. (59) Oates, T. W. H.; Wormeester, H.; Arwin, H. Characterization of Plasmonic Effects in Thin Films and Metamaterials Using Spectroscopic Ellipsometry. Prog. Surf. Sci. 2011, 86, 328−376.

(60) Kubo, S.; Diaz, A.; Tang, Y.; Mayer, T. S.; Khoo, I. C.; Mallouk, T. E. Tunability of the Refractive Index of Gold Nanoparticle Dispersions. Nano Lett. 2007, 7, 3418−3423. (61) Kravets, V. G.; Neubeck, S.; Grigorenko, A. N.; Kravets, A. F. Plasmonic Blackbody: Strong Absorption of Light by Metal Nanoparticles Embedded in a Dielectric Matrix. Phys. Rev. B 2010, 81, 165401. (62) Wan, D.; Chen, H.-L.; Lin, Y.; Chuang, S.; Shieh, J.; Chen, S. Using Spectroscopic Ellipsometry to Characterize and Apply the Optical Constants of Hollow Gold Nanoparticles. ACS Nano 2009, 3, 960−970. (63) Oates, T. W. H. Real Time Spectroscopic Ellipsometry of Nanoparticle Growth. Appl. Phys. Lett. 2006, 88, 213115. (64) Oates, T. W. H.; Mücklich, A. Evolution of Plasmon Resonances During Plasma Deposition of Silver Nanoparticles. Nanotechnology 2005, 16, 2606−2611. (65) Gibaud, A.; Daillant, J. X-Ray and Neutron Reflectivity: Principle and Applications; Springer, 2010. (66) Van der Lee, A. Grazing Incidence Specular Reflectivity: Theory, Experiment, and Applications. Solid State Sci. 2000, 2, 257− 278. (67) Fuertes, M. C.; Marchena, M.; Marchi, M. C.; Wolosiuk, A.; Soler-Illia, G. J. A. A. Controlled Deposition of Silver Nanoparticles in Mesoporous Single- or Multilayer Thin Films: From Tuned Pore Filling to Selective Spatial Location of Nanometric Objects. Small 2009, 5, 272−280. (68) Violi, I. L.; Perez, M. D.; Fuertes, M. C.; Soler-Illia, G. J. A. A. Highly Ordered, Accessible and Nanocrystalline Mesoporous TiO2 Thin Films on Transparent Conductive Substrates. ACS Appl. Mater. Interfaces 2012, 4, 4320−4330. (69) Cullity, B. D. Elements of X-Ray Diffraction; Addison-Wesley, 1978. (70) Smilgies, D. Scherrer Grain-Size Analysis Adapted to GrazingIncidence Scattering with Area Detectors. J. Appl. Crystallogr. 2009, 42, 1030−1034. (71) Harris, F. Growth and Structure of Supported Metal Catalyst Particles. Int. Mater. Rev. 1995, 40, 97−115. (72) Datye, A. K.; Xu, Q.; Kharas, K. C.; McCarty, J. M. Particle Size Distributions In Heterogeneous Catalysts: What Do They Tell Us About the Sintering Mechanism? Catal. Today 2006, 111, 59−67. (73) Kim, T.; Lee, C.-H.; Joo, S.-W.; Lee, K. Kinetics of Gold Nanoparticle Aggregation: Experiments and Modeling. J. Colloid Interface Sci. 2008, 318, 238−243. (74) Guarise, C.; Pasquato, L.; Scrimin, P. Reversible Aggregation/ Deaggregation of Gold Nanoparticles Induced by a Cleavable Dithiol Linker. Langmuir 2005, 21, 5537−5541. (75) Golunski, S. E. Final Analysis: Why Use Platinum in Catalytic Converters? Platin. Met. Rev. 2007, 51, 162−162. (76) Veith, G. M.; Lupini, A. R.; Rashkeev, S.; Pennycook, S. J.; Mullins, D. R.; Schwartz, V.; Bridges, C. A.; Dudney, N. J. Thermal Stability and Catalytic Activity of Gold Nanoparticles Supported on Silica. J. Catal. 2009, 262, 92−101. (77) Pérez-Maqueda, L. A.; Criado, J. M.; Sanchez-Jiménez, P. E. Combined Kinetic Analysis of Solid-State Reactions: A Powerful Tool for the Simultaneous Determination of Kinetic Parameters and the Kinetic Model Without Previous Assumptions on the Reaction Mechanism. J. Phys. Chem. A 2006, 110, 12456−12462. (78) Khawam, A.; Flanagan, D. R. Solid-State Kinetic Models: Basics and Mathematical Fundamentals. J. Phys. Chem. B 2006, 110, 17315− 17328. (79) Turba, T.; Norton, M. G.; Niraula, I.; McIlroy, D. N.Ripening of Nanowire-Supported Gold Nanoparticles. J. Nanopart. Res. 2008, 11, 2137−2143. (80) McGinn, J. T.; Greenhut, V. A.; Tsakalakos, T.; Blanc, J. Formation of Fault Structures During Coalescence and Growth of Gold Particles in a Fused Silica MatrixI. Acta Metallogr. 1982, 30, 2093−2102. (81) McGinn, J. T.; Greenhut, V. A.; Thomas, T. A Mechanism for Fault Formation in Fine Particles and Implications for Theories of 13150



Article

Annealing Twins in F.C.C. MetalsII. Acta Metallogr. 1982, 30, 2103−2110. (82) Flüeli, M.; Buffat, P. A.; Borel, J.-P. Real Time Observation by High Resolution Electron Microscroscopy (HREM) of the Coalescence of Small Gold Particles in the Electron Beam. Surf. Sci. 1988, 202, 343−353.

13151


Confinement-Induced Growth of Au Nanoparticles Entrapped in

Recommend Documents