Letter pubs.acs.org/NanoLett
Unravelling the Dependence of Hydrogen Oxidation Kinetics on the Size of Pt Nanoparticles by in Operando Nanoplasmonic Temperature Sensing Kristina Wettergren,† Anders Hellman,†,‡ Filippo Cavalca,§ Vladimir P. Zhdanov,†,∥ and Christoph Langhammer*,† †
Department of Applied Physics and ‡Competence Centre for Catalysis, Chalmers University of Technology, 41296 Göteborg, Sweden § Center for Electron Nanoscopy, Technical University of Denmark, 2800 Kongens Lyngby, Denmark ∥ Boreskov Institute of Catalysis, Russian Academy of Sciences, Novosibirsk 630090, Russia S Supporting Information *
ABSTRACT: We use a noninvasive nanoscale optical-temperature measurement method based on localized surface plasmon resonance to investigate the particle size-dependence of the hydrogen oxidation reaction kinetics on model supported Pt nanocatalysts at atmospheric pressure in operando. With decreasing average nanoparticle size from 11 down to 3 nm, the apparent reaction activation energy is found to increase from 0.5 up to 0.8 eV. This effect is attributed to an increase of the fraction of (100)-facet and edge and corner sites and their increasingly important role in the reaction with decreasing particle size. KEYWORDS: Indirect nanoplasmonic sensing, hydrogen oxidation, kinetics, particle size dependence, platinum
T
Here we demonstrate an optical temperature measurement approach that exploits the intrinsic temperature sensitivity of localized surface plasmon resonances (LSPR) of metal nanoparticles3 as nanoscale optical temperature probes to minimize the aforementioned issues. In particular, we show that the LSPR of Au nanoparticle sensors, separated from the active Pt catalyst nanoparticles by a 10 nm dielectric layer, can be used to efficiently measure the temperature of the active catalyst nanoparticles in operando. By tracking the temperatureinduced shifts of the LSPR wavelength it becomes possible to measure the temperature changes of a small amount of catalyst nanoparticles that cover only a few percent of a 2D support material of choice. Notably, the measurement is noninvasive due to the nanometer size of the used temperature probes (very small thermal mass) and the remote readout by low-intensity visible light. We investigate the particle size dependence of the apparent activation energy for hydrogen oxidation on a silica-supported Pt model catalyst under operando conditions as a model system to demonstrate and validate the approach. This is one of the generic reactions in heterogeneous catalysis (see, e.g., earlier studies on polycrystalline wires,4,5 foil,6 single crystals,7,8 and a field electron microscope tip;9 see also ref 10 for the heterohomogeneous reaction regimes and ref 11 for the
he time scale characterizing heat redistribution or transfer in nanoscale objects and their more-and-more miniaturized supports is extremely short.1,2 Accurate temperature measurements in such systems are thus a significant experimental challenge because standard temperature probes typically operate at longer time scales and/or feature orders of magnitude larger sizes and therefore potentially severely affect the temperature of interest. However, being able to accurately measure temperature at the nanoscale is critical in view of emerging materials, devices, and processes that are based on nanoparticles and structure. One specific field where nanosized objects play a key role, and where accurate temperature readout is very important, is heterogeneous catalysis. Typically, the catalyst temperature is measured at macroscopic length scale by using a temperature probe (e.g., a thermocouple), which may be invasive due to its much larger thermal mass compared to the catalytically active nanoparticles. This means that the measured temperature might not be identical to that of the catalyst nanoparticles themselves. Moreover, the presence of such macroscopic probes can significantly disturb the uniform reactant flow and thus mass transport and reaction energetics. Consequently, locally, where the temperature measurement takes place, this may create a situation that is significantly different from the rest of the system, hence making the measured temperature not fully representative. Nevertheless, conventional sensors based on thermocouples are widely used in catalysis to determine the reaction rate as a function of temperature and/or reactant mixture. © XXXX American Chemical Society
Received: October 21, 2014 Revised: November 23, 2014
A
dx.doi.org/10.1021/nl504042u | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
Contrasting earlier applications of the same INPS platform,17 here we are exploiting its intrinsic temperature sensitivity, rather than the sensitivity to local dielectric changes typically used in nanoplasmonic bio/chemosensing experiments.18−20 Typical optical extinction spectra of a blank INPS temperature nanosensor without catalyst measured at two different temperatures in the 4% hydrogen/oxygen mixture in Ar are shown in Figure 2a. As the temperature increases, the extinction peak of the Au nanoparticles broadens and its maximum shifts to longer wavelengths. This temperatureinduced peak shift is predominantly a consequence of the thermal expansion of the Au nanoparticles upon heating.3 Figure 2b shows the linear relationship between the spectral position of the extinction peak of the Au nanoparticles (expressed as “peak shift”, Δλ), and the global temperature of the sample measured by a thermocouple (to calibrate the sensor response) on the backside of the INPS chip (as shown in Figure 1; for other setups, the calibration can be done in other suitable ways). The high stability of the INPS chips at these elevated temperatures and the consequent reproducibility of this kind of measurements is evident by the basically identical response in five subsequent temperature ramps. These results thus clearly demonstrate that plasmonic nanoantennas in general, and INPS sensors in particular, can be efficiently used as nanometric optical thermometers to measure temperature locally. With our current setup, the INPS sensors have a temperature resolution of approximately 1 K as derived from the measured T-sensitivity factor, Δλ/K, of 0.0195 nm/K (Figure 2b) and a maximal Δλ-resolution of 0.01 nm achievable by using the analysis procedure by Dahlin et al.23 After having established this optical nanoscale temperature measurement, we now turn to the kinetic experiments for the hydrogen oxidation reaction. Typical INPS response curves during the heating ramp in pure Ar and in the 4% hydrogen− oxygen mixture for sensors decorated with Pt nanoparticles are shown in Figure 3d for three particle sizes (2.96, 3.77, and 10.1 nm). What is plotted is the peak position, λ, signal versus the gas temperature at the sample position measured by a thermocouple in direct contact with the backside of the sample during a heating ramp in Ar. As for the blank sample shown
kinetics of this reaction on supported Rh). The reaction intermediates have been extensively studied under ultrahigh vacuum conditions (see, e.g., references in ref 4 and references in the discussion of the results below) and have recently been investigated both by DFT calculations12−14 and microcalorimetry.15 Nevertheless, and despite its relatively simple mechanism, the full understanding of its kinetics, in particular its particle-size dependence, is still lacking. Partly, this is related to the fact that under suitable conditions the hydrogen oxidation reaction on Pt is perhaps the fastest in heterogeneous catalysis, which complicates its conventional investigation. To the best of our knowledge, our work is thus the first report on particle size effect for the kinetics of this reaction. In our experiments, we have utilized the indirect nanoplasmonic sensing (INPS) platform16 to measure the temperature changes of Pt model catalyst nanoparticles during reaction under operando conditions at atmospheric pressure (Figure 1).
Figure 1. Schematic depiction of the used flow-reactor setup with spectroscopic readout, including a close-up of the INPS-chip, schematic and SEM, featuring the Pt catalyst nanoparticles and plasmonic Au nanodisks serving as temperature sensors on one side and a conventional thermocouple on the other side.
Figure 2. Principle of optical temperature measurements using plasmonic nanosensors. (a) Extinction spectra of a blank (no catalyst) INPS-chip measured at two different temperatures. The inset shows the shift in peak position, Δλ, induced by the change in temperature. (b) Peak position shift Δλ plotted continuously as a function of the sample temperature during a linear heating ramp (0.1 K/s) in a 4% hydrogen/oxygen mixture in Ar carrier gas. Note the clear linear scaling between chip temperature and the plasmon peak position, as well as the reproducibility of the sensor response for five subsequent temperature ramps. B
dx.doi.org/10.1021/nl504042u | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
Figure 3. (a−c) Representative TEM images of Pt catalyst nanoparticles with increasing average size obtained by sintering in 4% oxygen at 823 K. (d) Peak position, λ, for INPS-chips decorated with Pt catalyst particles of the three different sizes shown in (a−c), plotted as a function of temperature during a linear temperature ramp both in inert environment (Ar) and in a 4% H2 + O2 mixture ([H2]/([H2] + [O2]) = 0.45). Note that the temperature on the x-axis (denoted gas temperature) was measured by the thermocouple on the backside of the chip during the ramp in inert Ar atmosphere (i.e., no reaction). The curves for the different particle sizes were offset by 4 nm for clarity. (e) INPS response, obtained by subtracting the λ-curves measured in Ar from the respective ones obtained in the H2 + O2 mixture, plotted as a function of the true local catalyst temperature, Tcat. The latter is obtained by translating the λ-signal obtained during the reaction shown in panel (d) into a local temperature measurement by relying on the temperature calibration of the INPS sensor. What we obtain in this way is a light-off curve, which on the y-axis shows the local increase in catalyst temperature (expressed as Δλ that is proportional to T) due to the exothermic hydrogen oxidation reaction and on the x-axis shows the true local catalyst temperature during this reaction.
(Supporting Information Figure S6). However, because we perform our optical measurement at the center of the INPS chip and, as discussed below, focus on the kinetically controlled regime where heat dissipated by the reaction is small, the temperature gradients across the illuminated area are negligible in our analysis. By subtracting the corresponding INPS signals measured in Ar from the INPS signals obtained in the hydrogen−oxygen mixture, we thus obtain the local temperature increase of the Pt catalyst particles induced by the dissipated heat from the 2H2 + O2 → 2H2O reaction. At low catalyst temperatures, the reaction is slow and the reaction rate (and the associated dissipated chemical power, which we measure as a temperature increase) is kinetically limited. In this regime, the reaction rate can be described by an Arrhenius expression, r = ν exp(−Ea/kT), where ν is the preexponential factor, and Ea is the apparent activation energy for the reaction. Because the supply of reactants to the surface is eventually limited by diffusion, further increasing the temperature causes the system to transition to mass transport controlled conditions, yielding the observed typical sigmoidlike curve of reaction rate versus temperature, a light-off curve.25 We find in Figure 3d that the transient regime occurs at ca. 430 K (gas temperature) for the 10.1 nm particles, at ca. 400 K for the 3.8 nm particles, and at 388 K for the 2.9 nm particles. The decrease of the light-off temperature with
above, also the INPS signals from sensors with Pt catalyst scale linearly with the gas temperature in inert atmosphere. The signal in the hydrogen−oxygen mixture, however, is distinctly different and characterized by a pronounced sigmoidlike deviation from the linear temperature dependence seen in pure Ar. This behavior is determined by the mass- and heattransfer processes during reaction. With increasing temperature, the reaction heat becomes appreciable as the reaction rate increases according to the Arrhenius law. Thus, the INPS chip temperature increases above the gas temperature in the reactor. In this situation, due to the specifics of the geometry of our reactor and sample holder (Figure 1 and Supporting Information Figure S4) the reaction heat generated by the catalyst on the INPS chip can be transferred both directly and indirectly (via the stainless steel sample holder) to the gas. Because of the high heat capacity and thermal conductivity of the chip and sample holder in the reactor, the indirect channel dominates (see the detailed analysis in the Supporting Information). In this case, the asymmetry of the holder with respect to the sample results in the appearance of temperature gradients along the INPS chip, primarily from the top of the chip toward the holder. Our measurements indicate that the maximum difference in surface temperature along the chip may be up to 25 K in the mass transport limited regime with increasing steepness of the gradient toward the holder C
dx.doi.org/10.1021/nl504042u | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
decreasing particle size is mainly an effect of the variation in total platinum surface area (because the total platinum volume for the samples is constant for all samples, the total platinum surface area will vary when the nanoparticle size changes), as shown in Figure S7 in the Supporting Information. Figure 3e shows the Δλ-signal obtained by subtracting the measurement in Ar from the one in the reactant mixture plotted versus the optically measured catalyst temperature, Tcat. This temperature was determined from the INPS peak position, λ, signal obtained in operando by converting it into a local temperature using the T-sensitivity calibration factor determined from Ar heating ramps shown in Figure 2b. It becomes now even clearer that the system exhibits the expected sigmoidlike temperature behavior dictated by a kinetically controlled regime at low temperature, a transient light-off regime that is shifted to higher temperature for increasing particle size, and a diffusion-limited regime at high temperature with a decreasing maximal temperature for increasing particle size. Moreover, a decrease in Δλ-signal (Figure 3e) and, accordingly, reaction rate is observed in the mass-transport limited regime while one could expect a plateau in this case. In an attempt to understand this observation, one should bear in mind that the catalyst temperature in this regime depends on the interplay of the mass transport to and heat conduction from the sample. Both processes depend in turn on geometry and temperature of the system. The observed slight decrease of the Δλ-signal with increasing temperature in the mass-transport limited regime thus seems to indicate that for our (complex) sample geometry the heat conduction becomes more efficient with increasing temperature than the mass transport. The fullscale analysis of this aspect is, however, beyond the scope of this work where we focus on the kinetically controlled regime. Finally, we also note that in principle the LSPR signal may originate not only from a temperature change but also from a change in the chemical state of metal nanoparticles or the support material.16,17 However, in this work our quantitative analysis has been focused on the regime where the surface of the catalyst is predominantly covered by atomic hydrogen, that is, the catalyst is (and remains throughout the entire process) in the reduced/metallic state. Under such circumstances, a potential reactant surface coverage-related shift17 of the LSPR signal is not expected. For the same reason, oxygen “spillover” from the catalyst to the support (which could induce a change of the dielectric properties there and thus convolute the LSPR signal) is also negligible. For further analysis, we now focus on the kinetically controlled regime of our data and perform an Arrhenius analysis to obtain the size-dependent apparent activation energies of the reaction. Specifically, taking into account that Δλ (Figure 3e) is proportional to the reaction rate, we construct the dependence of ln(Δλ) on 1/T, where T is the temperature measured by INPS, Tcat. This procedure is valid if the temperature gradients along the INPS chip are negligible, which, as discussed above, is the case in the kinetically limited region of the light-off curve. The resulting apparent activation energies for the different particle sizes are shown in Figure 4 (representative Arrhenius plots are shown in Supporting Information Figure S8). The indicated data points and error bars correspond to the average value obtained from three consecutive measurements on the same sample (three subsequent ramps in the gas mixture). The apparent activation energy for the hydrogen oxidation reaction on a (reduced) Pt wire26 has been included for comparison.
Figure 4. Size dependence of the apparent activation energy for hydrogen oxidation on Pt nanoparticles. The values shown are derived by Arrhenius analysis based on the local catalyst temperature, Tcat, determined by INPS. The dashed line corresponds to hydrogen oxidation apparent activation energy on a (reduced) Pt wire.26
As the first observation we see a pronounced dependence of the apparent activation energy on particle size, with an increasing barrier for smaller particles. Furthermore, we find that the derived apparent activation energies for the largest particle sizes approach the bulk value obtained for a reduced (metallic) Pt wire.26 To elucidate the results of our measurements, we recall the conventional scheme of the reaction under consideration4,27 H 2,gas + 2* ⇆ 2Hads
(1)
O2,gas + 2* → 2Oads
(2)
Oads + Hads ⇆ OHads + *
(3)
OHads + Hads → H 2Ogas + 2 *
(4)
OHads + OHads → H 2Ogas + Oads + *
(5)
where * is a vacant adsorption site. In the kinetically controlled regime, the specifics of this reaction is that step (1) is close to equilibrium, steps (3−5) are rapid, and thereby the reaction rate is limited by O2 adsorption [step (2)] on the surface covered primarily by hydrogen. Accordingly, the apparent reaction activation energy is a fraction of the activation energy of hydrogen desorption.4 Thus, the particle-size dependence of the reaction activation energy can be associated with that of the activation energy of hydrogen desorption. The latter activation energy may depend on the particle size due to the surface tension-induced compression of nanoparticles. For CO adsorption on Pd nanoparticles, this effect was scrutinized in recent density functional theory (DFT) calculations.28 For the Pt-particle sizes used in our work, provided the same effects as determined for CO on Pd would apply, the activation energy for hydrogen desorption is thus expected to increase with increasing particle size. However, because we experimentally observe an opposite trend for our system, an additional effect must be more important, namely the significantly different properties of facet sites and undercoordinated sites at edges and corners of the Pt nanoparticles. In particular, the O2 sticking D
dx.doi.org/10.1021/nl504042u | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
rather than of the entire catalyst system. This, in turn, makes it possible to efficiently and noninvasively measure the reaction heat of catalytic reactions under operando conditions. Using this new experimental approach, we found distinct particle-size dependence of the kinetics of the hydrogen oxidation reaction on Pt nanoparticles with increasing apparent activation energies for decreasing particle size, which can be understood as an increasingly important contribution of the (100) facet and undercoordinated sites at particle edges and corners to the total reaction rate for decreasing size. To the best of our knowledge, our work constitutes the first report of this category for this reaction. In a broader context, our new experimental approach and the obtained results help to investigate and understand structure sensitivity of catalytic reactions. Although this subject has long attracted attention in heterogeneous catalysis,36 its understanding is still limited. We also note that other aspects of the kinetics of heterogeneous catalytic reactions can be studied with superior spatial resolution using our approach to optically and noninvasively measure the temperature of catalyst particles and/or the support at the nanoscale. Methods. Indirect Nanoplasmonic Sensing (INPS) Platform Fabrication. The INPS sensor “chips” were prepared on borofloat glass substrates (15 mm × 15 mm × 1 mm) that were covered by an amorphous array of Au nanodisks (average diameter d = 80 nm and height h = 20 nm) by means of hole-mask colloidal lithography.21 The average distance between the nanodisks was ∼200 nm. After fabrication, the Au nanodisks were annealed at 888 K in air for 3 h to allow recrystallization and to attain a structure that is thermally stable. They were then covered by a 10 nm thick RF-sputtered SiO2 layer (FHR MS150 sputter system), which both serves as a protection layer to prevent alloying of the Au nanosensors with the catalyst (that is deposited on top) at elevated temperatures and to mimic the support material of a real catalyst. After SiO2 deposition, the INPS sensors were again annealed at 888 K in air for 36 h for further thermal stabilization. For sequential kinetic measurements, we prepared 10 identical INPS samples with the same amount of Pt catalyst by simultaneously e-beam evaporating 5 Å of Pt (deposition rate 0.05 nm/s at a base pressure
