How Does a Single Pt Nanocatalyst Behave in Two ... - Peng Chen

Jan 25, 2012 - app app. (3b). Figure 2. (A) Dependence of the single-particle turnover rate v on the ..... (32) Xu, W.; Kong, J. S.; Yeh, Y.-T. E.; Ch...
0 downloads 0 Views 4MB Size
Letter pubs.acs.org/NanoLett

How Does a Single Pt Nanocatalyst Behave in Two Different Reactions? A Single-Molecule Study Kyu Sung Han,† Guokun Liu,† Xiaochun Zhou, Rita E. Medina, and Peng Chen* Department of Chemistry and Chemical Biology, Cornell University, Ithaca, New York 14853, United States S Supporting Information *

ABSTRACT: Using single-molecule microscopy of fluorogenic reactions we studied Pt nanoparticle catalysis at singleparticle, single-turnover resolution for two reactions: one an oxidative N-deacetylation and the other a reductive Ndeoxygenation. These Pt nanoparticles show distinct catalytic kinetics in these two reactions: one following noncompetitive reactant adsorption and the other following competitive reactant adsorption. In both reactions, single nanoparticles exhibit temporal activity fluctuations attributable to dominantly spontaneous surface restructuring. Depending on the reaction sequence, single Pt nanoparticles may or may not show activity correlations in catalyzing both reactions, reflecting the structure insensitivity of the N-deacetylation reaction and the structure sensitivity of the N-deoxygenation reaction. KEYWORDS: Single-particle catalysis, platinum nanoparticles, single-molecule imaging, deacetylation and deoxygenation, structure sensitivity

M

Most metal nanoparticles can catalyze a multitude of chemical transformations, for example, Pt nanoparticles can catalyze both oxidative and reductive reactions.14 For a particular type of nanoparticle, its activities in catalyzing these different reactions can be correlated to each other or have little correlation, because different reactions may occur at different surface sites on the same nanoparticles. Considering the inherent inhomogeneity of nanoparticle catalysts, a fundamental question then arises: how would the catalytic behavior of a single nanoparticle be correlated between different reactions? An answer to this question will contribute to understanding the structure−activity correlation of nanoparticle catalysts across a variety of chemical transformations. Along this line, here we study the catalysis of individual Pt nanoparticles in two different reactions, one an oxidative N-deacetylation reaction and the other a reductive N-deoxygenation reaction, using singlemolecule microscopy of fluorogenic reactions (Figure 1A,B). We prepared colloidal Pt nanoparticles by reducing chloroplatinic acid (H2PtCl6·(H2O)6) with sodium borohydride (NaBH4) in aqueous solution that also contains citrate ions, following reported procedures (see Supporting Information Section 1 for details).38 The citrate ions here are weak ligands, helping stabilize the Pt nanoparticles. This preparation does not use strong-binding capping ligands or polymers to stabilize Pt nanoparticles, thus alleviating ligand passivation of nanoparticle surfaces for catalysis. The resulting Pt nanoparticles are 4.6 ±

etal nanoparticles are perhaps the most important industrial catalysts. They can catalyze many chemical transformations and have applications ranging from chemical synthesis to pollutant removal and to energy production and storage.1−9 Yet they are structurally inhomogeneous, even with the state-of-the-art colloidal synthesis that is capable of producing single-crystal nanoparticles with well-defined shapes.4,10−13 Moreover, because of their nanometer dimension, they are structurally dynamic, especially on their surfaces and under reaction conditions where the constantly changing adsorbate−surface interactions can further induce dynamic surface restructuring.14−20 These structural inhomogeneities and dynamics present an inherent challenge to characterizing and understanding the catalytic activity of nanoparticles fundamentally, as individual particles differ from one to another and from time to time; and it becomes necessary to study their catalysis at the single-particle level in real time. Significant progress has been made lately by several groups in studying nanoparticle catalysis at the single-particle level using methods including localized surface plasmon resonance spectroscopy,21−24 single-molecule fluorescence microscopy,25−27 and electrochemical detection.28−31 Our group has used singlemolecule fluorescence microscopy to study Au nanoparticle catalysis at single-particle, single-turnover resolution; these studies revealed intricate interplays between inhomogeneous reactivity, selectivity in parallel reaction pathways, dynamicsurface-restructuring coupled catalytic dynamics, and reactantconcentration-dependent surface switching in nanocatalysis.26,32−37 © 2012 American Chemical Society

Received: October 18, 2011 Revised: January 21, 2012 Published: January 25, 2012 1253

dx.doi.org/10.1021/nl203677b | Nano Lett. 2012, 12, 1253−1259

Nano Letters

Letter

Figure 1. (A) Experimental scheme of single-molecule fluorescence measurements of catalysis by individual nanoparticles using total-internalreflection fluorescence microscopy of fluorogenic reactions. (B) Pt-nanoparticle-catalyzed oxidative N-deacetylation of amplex red by H2O2 and reductive N-deoxygenation of resazurin by N2H4. Both reactions lead to formation of the fluorescent resorufin. (C) TEM image of Pt nanoparticles. Inset: diameter distribution of Pt nanoparticles; NPs = nanoparticles; Gaussian fit gives an average diameter of 4.6 ± 0.9 nm. (D) Exemplary fluorescence intensity versus time trajectories of single Pt nanoparticles under catalysis for (1) the oxidative N-deacetylation reaction at 5 μM amplex red and 200 mM H2O2 and (2) the reductive N-deoxygenation reaction at 0.2 μM resazurin and 1 mM N2H4. Time resolution: 20 ms. τ is the microscopic reaction time between sequential reaction events, which are manifested by the fluorescence intensity bursts.

Figure 1D shows time trajectories of fluorescence intensities from two Pt nanoparticles, each undertaking one of the two catalytic reactions. The trajectories contain many short fluorescence intensity bursts, each burst marking the generation of a product molecule resorufin, that is, one catalytic turnover. For a single Pt nanoparticle, the time τ between the appearance of a burst and that of the subsequent one is the microscopic reaction time for each product generation reaction; the lengths of τ are stochastic, but its statistical properties, such as its averages and distributions, are well-defined by the underlying reaction kinetics. From these fluorescence turnover trajectories, the turnover rate v of a single Pt nanoparticle can be obtained easily by counting the number of bursts per unit time; and v also equals ⟨τ⟩−1, where ⟨ ⟩ denotes averaging. With the capability of quantifying the turnover rates (v) of individual Pt nanoparticles, we examined how v depends on reactant concentrations to probe the kinetic mechanisms of the two catalytic reactions. When averaged over many nanoparticles, v shows a dependence on the reactant concentrations in both catalytic reactions (Figure 2A,B). Yet, the two catalytic reactions show distinct differences in how v changes with reactant concentrations. For the oxidative N-deacetylation reaction, the single-particle turnover rate v exhibits saturation kinetics with increasing concentration of the reactant amplex red while the other reactant H2O2 is kept constant and at a large excess (Figure 2A). In contrast, for the reductive Ndeoxygenation reaction, when the concentration of the reactant resazurin is increased while that of N2H4 is kept constant, v

0.9 nm in diameter, having faceted shapes of cube, tetrahedron, or cuboctahedron (Figure 1C). These Pt nanoparticles can catalyze two fluorogenic reactions (Figure 1B, and Supporting Information Section 3, Figures S2 and S3): (1) the oxidative Ndeacetylation of amplex red (AR), a nonfluorescent molecule, to resorufin, a highly fluorescent molecule, by hydrogen peroxide (H2O2), and (2) the reductive N-deoxygenation of resazurin (RZ), another nonfluorescent molecule, to resorufin by hydrazine (N2H4). The strong laser-induced fluorescence signal of the catalytic product resorufin in both reactions allows its detection at the single-molecule level. To monitor single Pt nanoparticles catalyzing either of these two reactions, we dispersed Pt nanoparticles on a quartz slide inside a microfluidic channel and flowed in the reactant solution with various reactant concentrations (0.3 to 10 μM amplex red and 200 mM H2O2 for the oxidative Ndeacetylation reaction; and 0.001 to 1 μM resazurin and 0.025 to 8 mM N2H4 for the reductive N-deoxygenation reaction, Figure 1A). The Pt nanoparticles are immobile due to nonspecific interactions with the quartz slide. Using a totalinternal-reflection fluorescence microscope, an electric multiplying charge coupled device camera operating at a 20 ms frame rate, and a 532 nm laser to excite resorufin fluorescence (Figure 1A), we recorded movies of fluorescence bursts on individual Pt nanoparticles; each fluorescence burst signals the generation of a catalytic product molecule resorufin on a single Pt nanoparticle. 1254

dx.doi.org/10.1021/nl203677b | Nano Lett. 2012, 12, 1253−1259

Nano Letters

Letter

where vAR is the single-particle turnover rate for the oxidative N-deacetylation reaction of amplex red; keff is the single-particle catalytic rate constant, representing the reactivity of an entire Pt nanoparticle; AR stands for amplex red and O for H2O2; and KAR and KO are the respective reactant adsorption equilibrium constants. Under saturating H2O2 concentrations, eq 1a reduces to

vAR = keff

v RZ = keff

KRZKR [RZ][R] (1 + KRZ[RZ] + KR [R])2

(2)

where RZ stands for resazurin and R for N2H4; the rest parameters have similar definitions as in eq 1a. Fitting the data in Figure 2A,B with eqs 1b and 2 gives keff = 0.047 ± 0.009 s−1 particle−1 and KAR = 1.7 ± 1.2 μM−1 for the oxidative Ndeacetylation reaction, and keff = 0.02 ± 0.03 s−1 particle−1, KRZ = 20 ± 11 μM−1, and KR = 0.0025 ± 0.0014 μM−1 for the reductive N-deoxygenation reaction. As the data in Figure 2A,B are averaged over many particles, all the values here reflect the averaged properties of many Pt nanoparticles. We could not do the H2O2 concentration titration of the single-particle catalysis experiment for the oxidative N-deacetylation reaction because at low concentrations H2O2 is unstable during the course of experiments (∼1 h), but ensemble kinetics show the expected saturation kinetics when H2O2 concentration is titrated (Supporting Information Figure S4B), as predicted by eq 1a. With the above kinetic mechanisms we can now quantify the catalytic activity of individual Pt nanoparticles in both reactions and determine their activity distributions. From the mechanism, we can derive the probability density functions f(τ) of the microscopic reaction times τ for both the oxidative Ndeacetylation reaction (fAR(τ)) and the reductive N-deoxygenation reaction (f RZ(τ)) (see Supporting Information Section 2)

initially increases until it reaches a maximum before decaying at higher resazurin concentrations; similar behavior is observed when the N2H4 is increased while the resazurin concentration is kept constant (Figure 2B). The kinetics of the two catalytic reactions can both be interpreted by the Langmuir−Hinshelwood mechanism for heterogeneous catalysis,39 but with key differences in how the two reactants of each catalytic reaction adsorb onto the surface sites of Pt nanoparticles. For the oxidative N-deacetylation reaction, the adsorption of the two reactants, amplex red and H2O2, follows a noncompetitive model, in which they adsorb onto different types of surface sites on a Pt nanoparticle. Consequently, the turnover rate v follows saturation kinetics when the concentration of one reactant is increased while the other is kept constant. This type of Langmuir−Hinshelwood kinetics is described quantitatively by the following equation (see Supporting Information Section 2)39

KAR K O[AR][O] (1 + KAR [AR])(1 + K O[O])

(1b)

On the other hand, for the reductive N-deoxygenation reaction, the adsorption of the two reactants, resazurin and N2H4, follows a competitive model, in which they adsorb onto the same type of Pt nanoparticle surface sites. Consequently, the turnover rate v decays when one reactant is at very high concentration and dominates the surface site occupation, making the other reactant unavailable for reaction. This type of Langmuir−Hinshelwood kinetics is described quantitatively by the following equation (see Supporting Information Section 2)39

Figure 2. (A) Dependence of the single-particle turnover rate v on the amplex red concentration for the Pt-nanoparticle-catalyzed oxidative N-deacetylation reaction. [H2O2] = 200 mM. (B) Dependence of v on the resazurin and N2H4 concentrations for the reductive Ndeoxygenation reaction. When resazurin concentration was titrated, N2H4 was kept at 1 mM; when N2H4 concentration was titrated, resazurin concentration was kept at 0.1 μM. Each data point in (A,B) is an average from the fluorescence turnover trajectories of >50 nanoparticles, with sem as the error bar. Solid line in (A) is a fit with eq 1b with keff = 0.047 ± 0.009 s−1 particle−1, KAR = 1.7 ± 1.2 μM−1, and those in (B) are global fits with eq 2 with keff = 0.02 ± 0.03 s−1 particle−1, KRZ = 20 ± 11 μM−1, KR = 0.0025 ± 0.0014 μM−1. AR, amplex red; RZ, resazurin; R, N2H4. (C) Distribution of keff from individual Pt nanoparticles for the oxidative N-deacetylation reaction. Solid line is a Gaussian fit with center at 0.036 ± 0.002 s−1 particle−1 and full-width-at-half-maximum (fwhm) of 0.038 ± 0.005 s−1 particle−1. Inset: distribution of τ from a single fluorescence turnover trajectory at [AR] = 5 μM and [H2O2] = 200 mM; solid line is a single exponential fit with a decay constant keff = 0.09 ± 0.01 s−1 particle−1. (D) Distribution of kapp from individual Pt nanoparticles at [RZ] = 0.2 μM and [N2H4] = 1 mM for the reductive N-deoxygenation reaction. Solid line is a Gaussian fit with center at 0.024 ± 0.001 s−1 particle−1 and fwhm of 0.018 ± 0.001 s−1 particle−1. Inset: distribution of τ from a single fluorescence turnover trajectory; solid line is a single exponential fit with a decay constant of kapp = 0.031 ± 0.003 s−1.

vAR = keff

KAR [AR] (1 + KAR [AR])

KAR K O[AR][O] (1 + KAR [AR])(1 + K O[O]) ⎛ ⎞ KAR K O[AR][O] τ⎟ exp⎜− keff (1 + KAR [AR])(1 + K O[O]) ⎠ ⎝

fAR (τ) = keff

[AR] →∞

=

[O] →∞

fRZ (τ) = keff

keff exp(− keff τ)

(3a)

KRZKR [RZ][R]

(1 + KRZ[RZ] + KR [R])2 ⎞ ⎛ KRZKR [RZ][R] exp⎜⎜ − keff τ⎟ 2 ⎟ (1 + KRZ[RZ] + KR [R]) ⎠ ⎝

= kapp exp(− kapp τ)

(1a) 1255

(3b)

dx.doi.org/10.1021/nl203677b | Nano Lett. 2012, 12, 1253−1259

Nano Letters

Letter

Figure 3. (A) Autocorrelation function Cτ(t) of the microscopic reaction time τ from the fluorescence turnover trajectories of single Pt nanoparticles catalyzing the oxidative N-deacetylation reaction at 5 μM amplex red and 200 mM H2O2. The x-axis was converted from the turnover index m to real time using the average turnover time of each nanoparticle, and the data were averaged over >50 Pt nanoparticles. Solid line is a single exponential fit with decay constant of 63 ± 18 s. Inset: autocorrelation function Cτ(m) from the fluorescence turnover trajectory of a single Pt nanoparticle; solid line is a single exponential fit with decay constant of 2.2 ± 0.6 turnovers. (B) Same as (A) but for the N-deoxygenation reaction at 0.2 μM resazurin and 1 mM N2H4. The decay constant of the exponential fit is 39 ± 10 s. Data averaged over >50 Pt nanoparticles. Inset: similar as that in (A). The decay constant of the exponential fit is 5 ± 2 turnovers. (C) Dependences of the activity fluctuation rates on the turnover rates of Pt and Au nanoparticles in catalysis. The activity fluctuation rates are the inverse of the activity fluctuation correlation times determined from the autocorrelation functions Cτ(t). Red and black lines are fits of horizontal lines at 0.017 ± 0.001 and 0.019 ± 0.006 s−1, corresponding to time scales of 59 ± 4 and 53 ± 17 s, respectively. Blue line is for Au nanoparticles of ∼4.6 nm in diameter, extrapolated from the results on 6−14 nm Au nanoparticles in reference32 (see Supporting Information Section 4 for details). The linear dependence of the blue line on the turnover rate for Au nanoparticles reflects the catalysis-induced nature of the activity fluctuations, thus the underlying catalysis-induced surface restructuring; the yintercept gives the time scale of the spontaneous surface restructuring of Au nanoparticles of ∼42 ± 9 s, corresponding to a spontaneous surface restructuring rate of 0.024 ± 0.005 s−1.

In eq 3b kapp = keffKRZKR[RZ][R]/(1 + KRZ[RZ]+KR[R])2. Both probability density functions are exponential distribution functions. Moreover, fAR(τ) reduces to a simple form when the reactant concentrations are saturating (eq 3a). It also follows −1 −1 = 1/∫ 0∞τfAR(τ)dτ = νAR and ⟨τ⟩RZ = 1/∫ 0∞τf RZ(τ)dτ that ⟨τ⟩AR = νRZ, as expected (see eqs 1a, , and 2, and Supporting Information Section 2). Figure 2C inset shows the distribution of τ from the fluorescence turnover trajectory of a single Pt nanoparticle catalyzing the oxidative N-deacetylation reaction at a saturating reactant concentration. This distribution follows an exponential distribution with decay constant keff for this Pt nanoparticle, as shown by eq 3a. For the reductive Ndeoxygenation reaction, the distribution of τ from the fluorescence turnover trajectory of a single Pt nanoparticle also follows an exponential distribution (Figure 2D, inset); the decay constant here is kapp as in eq 3b. By analyzing the distributions of τ of single trajectories, we determined keff and kapp for many Pt nanoparticles and their distributions (Figure 2C,D). Both keff and kapp are distributed over a broad range, indicating large activity heterogeneity among individual Pt nanoparticles in both reactions. We used a heterogeneity index (h, defined as the fwhm of the distribution divided by the average)33 as a measure of the relative spread of the distribution. h for keff and kapp are 106 ± 15% and 75 ± 5%, respectively. This direct quantification of activity heterogeneity among individual Pt nanoparticles is uniquely available from single-particle measurements and is difficult to obtain from ensemble-averaged studies. Besides allowing for direct quantification of the activity differences from one nanoparticle to another, the fluorescence turnover trajectories also enable examining the activity differences from time to time for a single Pt nanoparticle, a unique capability of real-time single-particle catalysis measurements. Our previous studies32,34 have revealed temporal activity fluctuations of single Au nanoparticles of 6−14 nm in diameter, which are attributable to spontaneous and catalysis-induced

dynamic surface restructuring. The time scale of the activity fluctuations, which is also the time scale of the underlying surface restructuring, can be obtained from the autocorrelation function Cτ(m) of the microscopic reaction time τ: Cτ(m) = ⟨Δτ(0)Δτ(m)⟩/⟨Δτ2⟩. Here m is the index of a catalytic turnover in a fluorescence turnover trajectory and Δτ(m) = τ(m) − ⟨τ⟩. In the presence of activity fluctuations, Cτ(m) ≥ 0 and shows an exponential decay behavior; its decay time constant gives the activity fluctuation correlation time. The insets of Figure 3A,B show two such autocorrelation functions Cτ(m), each from the fluorescence turnover trajectory of a single Pt nanoparticle catalyzing either the oxidative Ndeacetylation reaction or the reductive N-deoxygenation reaction at specified reactant concentrations (Figure 3 caption). Both Cτ(m) show exponential decay behaviors, manifesting the temporal activity fluctuations of single Pt nanoparticles. The exponential decay constants are 2.2 ± 0.6 turnovers for the nanoparticle in Figure 3A inset and 5 ± 2 turnovers for that in Figure 3B inset. Cτ(m) of each nanoparticle can then be converted to Cτ(t) in which the turnover index m is converted to real time t using the nanoparticle’s average turnover time. For both catalytic reactions, when Cτ(t) are averaged over many nanoparticles, their exponential decay behavior is preserved (Figure 3A,B). The two corresponding decay time constants are 63 ± 18 and 39 ± 10 s, which are the activity fluctuation time scales and also reflect the time scales of the underlying dynamic surface restructuring at the respective active sites on the Pt nanoparticles. Our observation of the surfacerestructuring-coupled activity fluctuations of single Pt nanoparticles is also consistent with the direct experimental demonstration that Pt-containing nanoparticles can reconstruct dynamically during reactions.15 The inverse of activity fluctuation correlation times gives the activity fluctuation rates. We determined the activity fluctuation rates of Pt nanoparticles at the various reactant concentrations for both catalytic reactions and plotted the fluctuation rates 1256

dx.doi.org/10.1021/nl203677b | Nano Lett. 2012, 12, 1253−1259

Nano Letters

Letter

Figure 4. Correlation plots of the single-particle turnover rates v of individual Pt nanoparticles between two sequential reactions, each lasting for one hour. Each data point is from one Pt nanoparticle. The reaction sequences are (A) first is the oxidative N-deacetylation reaction and second is the reductive N-deoxygenation reaction, (B) first is N-deoxygenation and second is N-deacetylation, (C) first and second are both N-deoxygenation, and (D) first and second are both N-deacetylation. Histograms give the distributions of the single-particle turnover rates in each of the reactions and are fitted with Gaussian distributions (solid lines). The data points that lie outside the 95% confidence level (i.e., 1.96 standard deviations away from the center) of the Gaussian distributions were removed in all plots.

reactions. To do so, we subjected the same set of Pt nanoparticles to the two catalytic reactions sequentially, each reaction lasting for about an hour at fixed reactant concentrations. The reactant concentrations were chosen to have high turnover rates in both reactions based on the titration kinetics in Figure 2A,B. We recorded the fluorescence turnover trajectories for every Pt nanoparticle in both reactions. Figure 4A plots the correlation between the turnover rates of every Pt nanoparticle in these two reactions, where each Pt nanoparticle underwent the oxidative N-acetylation reaction first and then the reductive N-deoxygenation reaction. The turnover rate of each nanoparticle in each reaction is a time-averaged property throughout the entire period each reaction was run. The Pearson cross correlation coefficient ρ is merely 0.11 ± 0.05, that is, close to zero (the error bar here is the probable error of the correlation coefficient, given as 0.6745(1 − ρ2)/√N;41 N is the number of nanoparticles). This small value of ρ indicates little correlation between the two turnover rates, that is, the catalytic activity of a Pt nanoparticle in the later deoxygenation reaction have little bearing on its catalytic activity in the earlier deacetylation reaction. In contrast, when the sequence of the two reactions was reversed, a significantly more positive correlation was observed between the two turnover rates of individual particles (ρ = 0.33 ± 0.04, Figure 4B). This positive correlation indicates that the catalytic activity of a Pt nanoparticle in the later deacetylation reaction now remembered more of its activity in the earlier deoxygenation reaction. As structure determines activity, the presence, or absence, of correlation between the activities of individual Pt nanoparticle in catalyzing two different reactions must be related to the underlying correlation, or the lack of it, between their surface

against the corresponding turnover rates in Figure 3C. The activity fluctuation rates are essentially independent of the turnover rates. This reflects that the underlying dynamic surface restructuring of Pt nanoparticles, which causes the activity fluctuations, is independent of the turnover rate in both catalytic reactions. This is in sharp contrast to the behavior of Au nanoparticles we studied previously:34 their activity fluctuation rate increases linearly with increasing turnover rate because of catalysis-induced dynamic surface restructuring (Figure 3C). Therefore, the dynamic surface restructuring of Pt nanoparticles is largely spontaneous (i.e., thermally driven) under our reaction conditions, and the catalysis-induced effect here is minimal. Consistently, the observed activity fluctuation rates of Pt nanoparticles, which are equivalent to the underlying surface restructuring rates, are about the same in both the oxidative N-acetylation reaction and the reductive N-deoxygenation reaction (0.017 ± 0.001 and 0.019 ± 0.006 s−1, respectively; both are averages of the respective data points across different turnover rates; Figure 3C), as spontaneous dynamic surface restructuring is intrinsic to the nanoparticle and should be independent of the type of the catalytic reactions. Furthermore, the spontaneous surface restructuring rate of Pt nanoparticles is slightly slower than that of Au nanoparticles of similar diameter (∼0.024 ± 0.005 s−1, Figure 3C), consistent with that Pt surfaces are thermodynamically more stable than Au surfaces under similar conditions.40 With an understanding of the kinetic mechanisms for both the oxidative N-deacetylation reaction and the reductive Ndeoxygenation reaction and of how single Pt nanoparticles differ individually and temporally in each reaction, we can now examine how a single Pt nanoparticle behaves in catalyzing both 1257

dx.doi.org/10.1021/nl203677b | Nano Lett. 2012, 12, 1253−1259

Nano Letters

Letter

rates of individual Pt nanoparticles follow a Gaussian distribution; the heterogeneity index, h, defined as the fwhm of this distribution divided by the mean, is a quantitative measure of how the activities of individual particles differ from one another. When a same set of Pt nanoparticles underwent the N-deacetylation and N-deoxygenation reactions sequentially, and regardless of the order of the reactions, h for the Ndeoxygenation reaction was always larger than that for the Ndeacetylation reaction (Figure 5A). This is consistent with that

structures under the two reaction conditions. Yet, the surface of a Pt nanoparticle restructures dynamically during both of these reactions, manifested by the temporal fluctuations of its catalytic activity as discussed earlier. For these Pt nanoparticles of ∼4.6 nm in diameter, the time scale of their dynamic surface restructuring is ∼50 s (Figure 3C caption), which is much shorter than the time (∼1 h) we subjected them in catalyzing each of the two reactions. Therefore, the surface structure of a single Pt nanoparticle in the later reaction must be different from that in the earlier reaction. As a nanoparticle always has many types of surface sites, this difference would be a different composition of various sites on its surface, although the exact nature of the differences and the extent of these differences are unclear from our measurements. TEM showed slight rounding of the facetedness of these Pt nanoparticles after catalyzing these reactions, but no large morphology was observed within our reaction times (Supporting Information Figure S8). Past studies have shown that for a surface-catalyzed reaction, depending on the nature of the chemical bond that is activated in the rate-limiting step, the kinetics of this reaction can depend sensitively on the nanoparticle surface structure (i.e., a structure sensitive reaction) or be insensitive to the structural arrangements of surface atoms (i.e., a structure insensitive reaction).42,43 When catalyzing two reactions sequentially and during the earlier reaction the surface restructures (i.e., the composition of various sites on the surface changes), this restructuring should cause a significant change in the activity of a nanoparticle in catalyzing a later structure-sensitive reaction. Consequently, the activity of a nanoparticle in the later structure-sensitive reaction should have little correlation with its activity in the earlier reaction. This scenario would rationalize the observed little correlation (ρ ∼ 0.11) of the two turnover rates in Figure 4A, where the later N-deoxygenation reaction is presumably more of a structure sensitive reaction. Alternatively, if the later reaction is structure insensitive, the surface restructuring occurred in the earlier reaction would affect less on the activity of a nanoparticle in the later reaction. Consequently, the activity of a nanoparticle in the later structure-insensitive reaction should show some correlation with its activity in the earlier reaction. This alternative scenario would rationalize the significantly more correlation (ρ ∼ 0.33) in Figure 4B, where the later N-deacetylation reaction is presumably more of a structure insensitive one. If our hypothesis is correct that the N-deoxygenation reaction is more of structure sensitive and the N-deacetylation reaction is more of structure insensitive, this hypothesis can predict on the correlation between the two turnover rates of individual Pt nanoparticles if the two sequential reactions are the same. If the two sequential reactions are both the structuresensitive N-deoxygenation reaction, the correlation should still be weak because every Pt nanoparticle undergoes dynamic surface restructuring during the reactions and individual nanoparticles restructure asynchronously and thus differently. If the two sequential reactions are both the structure-insensitive N-deacetylation reaction, there should be a clear correlation because the surface restructuring of every nanoparticle has less effect on its activity. These two predictions were indeed observed with ρ = 0.14 ± 0.05 and ρ = 0.43 ± 0.04, respectively, (Figure 4C,D), thus supporting our hypothesis. Our hypothesis is further supported by analyzing the activity heterogeneity among the individual nanoparticles that underwent two sequential reactions. In each reaction, the turnover

Figure 5. Comparisons of the heterogeneity indices h of the turnover rates among individual Pt nanoparticles in each of the reactions from Figure 4, when (A) the two sequential reactions are different, one the reductive N-deoxygenation of resazurin and the other the oxidative Ndeacetylation of amplex red, and (B) the two sequential reactions are the same, being both N-deoxygenation or both N-deacetylation. The h values should be compared within each panel, not between different panels, as data in different panels are each obtained from a different set of Pt nanoparticles. In each panel, the labels on the x-axis designate the sequence of the reactions, that is, being the first or the second of the two sequential reactions.

the N-deoxygenation reaction is more of structure sensitive and therefore the structural differences among the individual nanoparticles translate more clearly into their differences in activity, whereas the N-deacetylation is more of structural insensitive and therefore the structural differences among the same set of nanoparticles translate less into the differences in their activity. Consistently, when the two sequential reactions were identical, no significant difference in their h values was observed for a same set of Pt nanoparticles (Figure 5B). Furthermore, our hypothesis that the reductive N-deoxygenation of resazurin is more of structure sensitive is consistent with our previous studies of 6−14 nm Au nanoparticles catalyzing the same reaction (note the previous study used NH2OH as the reductant instead of N2H4).32,34 With decreasing Au nanoparticle size, its specific catalytic rate constant increases significantly,34 following a class II structure sensitivity of surface reactions.42,44,45 Although more studies are needed to elucidate the molecular mechanism of the nanoparticle-catalyzed N-deoxygenation of resazurin studied here, previous computational studies have shown that this class II structure sensitivity is in general associated with σ-bond cleavage in the rate-limiting step.42 In summary, we have studied how single ∼4.6 nm Pt nanoparticles behave in catalyzing two different reactions in real time at the single-turnover resolution using single-molecule microscopy of fluorogenic reaction: one an oxidative Ndeacetylation of amplex red and the other a reductive Ndeoxygenation of resazurin. We found that Pt nanoparticles show distinct catalytic kinetics in these two reactions: the reactants in the N-deacetylation reaction follow noncompetitive adsorption while those in the N-deoxygenation reaction follow 1258

dx.doi.org/10.1021/nl203677b | Nano Lett. 2012, 12, 1253−1259

Nano Letters

Letter

(9) Daniel, M. C.; Astruc, D. Chem. Rev. 2004, 104, 293. (10) Tao, A. R.; Habas, S.; Yang, P. Small 2008, 4, 310. (11) Xia, Y.; Xiong, Y.; Lim, B.; Skrabalak, S. E. Angew. Chem., Int. Ed. 2009, 48, 60. (12) Grzelczak, M.; Pérez-Juste, J.; Mulvaney, P.; Liz-Marzán, L. M. Chem. Soc. Rev. 2008, 37, 1783. (13) Murphy, C. J.; Thompson, L. B.; Chernak, D. J.; Yang, J. A.; Sivapalan, S. T.; Boulos, S. P.; Huang, J.; Alkilany, A. M.; Sisco, P. N. Curr. Opin. Colloid Interface Sci. 2011, 16, 128. (14) Somorjai, G. A.; Li, Y. Introduction to Surface Chemistry and Catalysis, 2nd ed.; John Wiley & Sons: Hoboken, NJ, 2010. (15) Tao, F.; Grass, M. E.; Zhang, Y.; Butcher, D. R.; Renzas, J. R.; Liu, Z.; Chung, J. Y.; Mun, B. S.; Salmeron, M.; Somorjai, G. A. Science 2008, 322, 932. (16) Wang, Z. L. Adv. Mater. 2003, 15, 1497. (17) Hansen, P. L.; Wagner, J. B.; Helveg, S.; Rostrup-Nielsen, J. R.; Clausen, B. S.; Topsoe, H. Science 2002, 295, 2053. (18) Newton, M. A.; Belver-Coldeira, C.; Martinez-Arias, A.; Fernandez-Garcia, M. Nat. Mater. 2007, 6, 528. (19) King, D. A.; Woodruff, D. P. Phase Transitions and Adsorbate Restructuring at Metal Surfaces; Elsevier Science: New York, 1994. (20) Imbihl, R.; Ertl, G. Chem. Rev. 1995, 95, 697. (21) Novo, C.; Funston, A. M.; Mulvaney, P. Nat. Nanotechnol. 2008, 3, 598. (22) Liu, N.; Tang, M. L.; Hentschel, M.; Giessen, H.; Alivisatos, A. P. Nat. Mater. 2011, 10, 631. (23) Tang, M. L.; Liu, N.; Dionne, J. A.; Alivisatos, A. P. J. Am. Chem. Soc. 2011, 133, 13220. (24) Cheng, J.; Liu, Y.; Cheng, X.; He, Y.; Yeung, E. S. Anal. Chem. 2010, 82, 8744. (25) Cremer, G. D.; Sels, B. F.; Vos, D. E. D.; Hofkens, J.; Roeffaers, M. B. J. Chem. Soc. Rev. 2010, 39, 4703. (26) Chen, P.; Zhou, X.; Shen, H.; Andoy, N. M.; Choudhary, E.; Han, K.-S.; Liu, G.; Meng, W. Chem. Soc. Rev. 2010, 39, 4560. (27) Tachikawa, T.; Majima, T. Chem. Soc. Rev. 2010, 39, 4802. (28) Xiao, X.; Pan, S.; Jang, J. S.; Fan, F.-R. F.; Bard, A. J. J. Phys. Chem. C 2009, 113, 14978. (29) Meier, J.; Friedrich, K. A.; Stimming, U. Faraday Discuss. 2002, 121, 365. (30) Chen, S.; Kucernak, A. J. Phys. Chem. B 2004, 108, 13984. (31) Li, Y.; Cox, J. T.; Zhang, B. J. Am. Chem. Soc. 2010, 132, 3047. (32) Xu, W.; Kong, J. S.; Yeh, Y.-T. E.; Chen, P. Nat. Mater. 2008, 7, 992. (33) Xu, W.; Kong, J. S.; Chen, P. Phys. Chem. Chem. Phys. 2009, 11, 2767. (34) Zhou, X.; Xu, W.; Liu, G.; Panda, D.; Chen, P. J. Am. Chem. Soc. 2010, 132, 138. (35) Chen, P.; Xu, W.; Zhou, X.; Panda, D.; Kalininskiy, A. Chem. Phys. Lett. 2009, 470, 151. (36) Xu, W.; Shen, H.; Liu, G.; Chen, P. Nano Res. 2009, 2, 911. (37) Xu, W.; Kong, J. S.; Chen, P. J. Phys. Chem. C 2009, 113, 2393. (38) Eychmuller, A.; Bigall, N. C.; Hartling, T.; Klose, M.; Simon, P.; Eng, L. M. Nano Lett. 2008, 8, 4588. (39) Satterfield, C. N. Heterogeneous Catalysis in Practice; McGrawHill Book Company: New York, 1980. (40) Foiles, S. M.; Baskes, M. I.; Daw, M. S. Phys. Rev. B 1986, 33, 7983. (41) Eells, W. C. J. Am. Stat. Assoc. 1929, 24, 170. (42) Van Santen, R. A. Acc. Chem. Res. 2009, 42, 57. (43) Boudart, M.; Djega-Mariadassou, G. Kinetics of Heterogeneous Catalytic Reactions; Princeton University Press: Princeton, 1984. (44) Che, M.; Bennett, C. O. Adv. Catal. 1989, 36, 55. (45) Henry, C. C., C.; Giorgio, S.; Goyhenex, C. In Chemisorption and Reactivity on Supported Clusters and Thin Films; Lambert, R. M., Pacchioni, G., Eds.; Kluwer Academic: Dordrecht, The Netherlands, 1997; p 117.

competitive adsorption within the model of Langmuir− Hinshelwood kinetics. Large activity heterogeneity is present among individual nanoparticles and the single-particle measurements provide direct quantification of this heterogeneity. In both reactions, individual Pt nanoparticles show temporal activity fluctuations, which are independent of the turnover rate and attributable to dominantly spontaneous dynamic surface restructuring. The time scale of the underlying spontaneous surface restructuring is about tens of seconds, slightly slower than that of Au nanoparticles of similar diameter, consistent with Pt’s more stable surfaces. When catalyzing both reactions sequentially, and depending on the reaction sequence, single Pt nanoparticles may or may not show activity correlations between these two reactions, reflecting that the N-deacetylation reaction is more of a structure insensitive surface reaction and the N-deoxygenation reaction is more of a structure sensitive reaction. The knowledge from single-particle level studies provides fundamental insights into the catalytic behaviors of nanoparticle catalysts, which are complementary to, and often inaccessible in, ensemble-averaged measurements.



ASSOCIATED CONTENT

S Supporting Information *

Materials and methods, additional results, and analyses. 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 †

These authors contributed equally.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This research is funded mainly by the Chemical Sciences, Geosciences and Biosciences Division, Office of Basic Energy Sciences, Office of Science, U.S. Department of Energy (DEFG02-10ER16199), and in part by U.S. Army Research Office (W911NF0910232), the National Science Foundation (CBET0851257), and an Alfred P. Sloan Research Fellowship (P.C.). Rita Medina was a REU student supported by the Cornell Center for Materials Research (CCMR), a NSF funded MRSEC center. We also thank Ivan Keresztes for NMR analysis and Nesha May Andoy for discussions. TEM facility at CCMR is supported by a NSF-MRSEC program (DMR1120296).



REFERENCES

(1) Somorjai, G. A.; Contreras, A. M.; Montano, M.; Rioux, R. M. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 10577. (2) Ertl, G.; Knözinger, H.; Weitkamp, J. Handbook of heterogeneous catalysis; VCH: Weinheim, 1997. (3) Bell, A. T. Science 2003, 299, 1688. (4) Burda, C.; Chen, X.; Narayanan, R.; El-Sayed, M. A. Chem. Rev. 2005, 105, 1025. (5) Chen, M.; Goodman, D. W. Acc. Chem. Res. 2006, 39, 739. (6) Crooks, R. M.; Zhao, M.; Sun, L.; Chechik, V.; Yeung, L. K. Acc. Chem. Res. 2001, 34, 181. (7) Lewis, L. N. Chem. Rev. 1993, 93, 2693. (8) Astruc, D.; Lu, F.; Aranzaes, J. R. Angew. Chem., Int. Ed. 2005, 44, 7852. 1259

dx.doi.org/10.1021/nl203677b | Nano Lett. 2012, 12, 1253−1259