Single-molecule, single-particle observation of size-dependent photocatalytic activity in Au/TiO₂ nanocomposites

Abstract

Electronic communication between the building blocks of nanocomposites is an important property that affects their functionality with regard to many optoelectronic and catalytic applications. Herein, we report a single-molecule, single-particle approach for elucidating the inherent photocatalytic activity of individual Au nanoparticle-loaded TiO₂ particles using a novel redox-responsive fluorescent dye. A single-particle kinetic analysis of the fluorescence bursts emitted from the products revealed that the photocatalytic activity leading to reduction of the probe molecules is controlled by not only the substrate concentration and excitation intensity but also the Au particle size, and that these factors are intricately interrelated. Furthermore, we discovered that the stochastic photocatalytic events around the millisecond-to-second time scale showed considerable temporal and spatial heterogeneity during photoirradiation, and that they actually originate from the charging/discharging of Au nanoparticles on TiO₂. Our findings represent a significant contribution to the scientific understanding of the interfacial electron transfer dynamics in composite systems, and more fundamentally, in heterogeneous (photo)chemical processes.

---

Introduction

Interfacial electron transfer (ET) is a fundamental process in physics, chemistry, and biology, playing a crucial role in natural and artificial energy conversion systems.¹ For instance, charge transport across the heterogeneous interface of metal–semiconductor and semiconductor–semiconductor nanocomposites has attracted a great deal of research interest, because this process largely governs the performance of photovoltaic devices, batteries, fuel cells, sensors, and (photo)catalysts used for water splitting, organic synthesis, and environmental purification.^2–6 To facilitate ET reactions, possible rate-limiting factors must be addressed and suitably optimized. The incorporation of noble metals with a large work function (e.g., Au and Ag) onto metal oxide supports (e.g., TiO₂ and ZnO) has been shown to significantly enhance the charge separation efficiency via ET from the conduction band (CB) of semiconductor metal oxides to metals on the surface. This results in the Fermi energy level of the loaded metal in the photostationary state (E_F^*) increasing and eventually reaching equilibrium with that of the metal oxides being subjected to photoirradiation.⁷ This concept was further developed by Kamat et al., who found that the E_F^* increased with a decrease in the diameter of the Au nanoparticles (d_Au) added to the TiO₂ nanoparticles using 3-mercaptopropionic acid as the bifunctional linker. They attributed this observation to the quantum size effect of the Au nanoparticles (3 < d_Au < 8 nm).^8,9 In addition, Tada et al. reported a converse trend in which the E_F^* of Au nanoparticles directly deposited on TiO₂ increased with d_Au in the range of 3 < d_Au < 13 nm. They concluded that the predominant mechanisms involved in determining E_F^* are either the solvation of charged metal nanoparticles in polar media or the quantum size effect in less polar media.¹⁰

Although a number of studies have been conducted to elucidate the mechanism of photoinduced ET reactions in these composite systems, the directional flow of photogenerated charge carriers through the interface has still not been fully explained because of the inherent static and dynamic heterogeneities in these systems. In fact, recent spectroscopic studies conducted using advanced microscopic techniques have demonstrated that chemical reactions occurring at the surface of a solid are intrinsically heterogeneous and closely related to the structural dispersions and spatial distributions of reactive sites, as well as surface restructuring dynamics, conformations of adsorbates, and electronic interactions between adjacent components.^11–22 However, to the best of our knowledge, no single-molecule studies have described the interfacial ET dynamics in nanocomposite systems.

In this study, we investigated the photocatalytic redox reaction over Au nanoparticle-loaded TiO₂ (Au/TiO₂) particles at the single-molecule, single-particle levels using a novel redox-responsive fluorescent dye. This investigation revealed the temporal and spatial heterogeneities of the interfacial ET process on an individual particle. The single-particle kinetic analysis of the fluorescence characteristics of the product revealed that the substrate concentration, UV light intensity, and structural characteristics of Au nanoparticles are important parameters governing the photocatalytic activity of Au/TiO₂. Furthermore, we discovered that the stochastic photocatalytic events actually originate from the charging/discharging of Au nanoparticles on TiO₂. Overall, this study provides a great deal of valuable information that is impossible to obtain from ensemble-averaged experiments because each particle behaves differently.

Results and discussion

Bulk photocatalytic experiments

The fluorogenic reaction is based on our recently developed ET-involved reduction of boron-dipyrromethene compound, 3,4-dinitrophenyl-BODIPY (DN-BODIPY, Fig. 1).²¹ The reduction of DN-BODIPY proceeds along the following general pathway: (i) the nitro (NO₂) group at the para-position is first reduced by two electrons to form a nitroso (NO) group, (ii) the NO group immediately accepts another two electrons to form a hydroxylamino (NHOH) group, and (iii) the NHOH group finally becomes an amino (NH₂) groupvia a slow two-electron reduction.^23,24 Because the two NO₂ groups greatly reduce the lowest unoccupied molecular orbital (LUMO) energy level of the benzene moiety introduced at the meso-position of the BODIPY core, the BODIPY fluorescence is significantly quenched by an intramolecular ET from the excited fluorophore to the nitro-substituted benzene moiety prior to reduction.^25,26 However, after one of the NO₂ groups of DN-BODIPY is reduced preferentially, the other NO₂ group neutralizes the increased electron density owing to the produced NHOH or NH₂ group, which dramatically suppresses the intramolecular ET process. Therefore, by accepting electrons, non-fluorescent DN-BODIPY (Φ_fl ≈ 10⁻⁴ in methanol) can be reduced to form the highly fluorescent 4-hydroxyamino-3-nitrophenyl-BODIPY (HN-BODIPY, Φ_fl = 0.50 in methanol), which provides the basis for studying the interfacial ET process on individual nanoparticlesviadetection of the fluorescence of the produced HN-BODIPY at the single-molecule level.²¹

Fig. 1 Photocatalytic generation of fluorescent HN-BODIPY from non-fluorescent DN-BODIPY.

The photocatalytic activity of Au/TiO₂ particles was first evaluated by ensemble-averaged spectroscopy. We selected three types of Au/TiO₂ photocatalysts with the same crystalline anatase form of TiO₂ and the same loading amount of Au, but different d_Au: 5 nm, 8 nm, and 14 nm Au/TiO₂ (Fig. S1†). The structural and optical properties of the selected compounds are listed in Table 1. As the Au particle size increased, the mean number of Au nanoparticles loaded per TiO₂ particle (N_Au) decreased significantly. As shown in Fig. S2†, Au/TiO₂ reduced DN-BODIPY to HN-BODIPYvia the reduction pathway in Ar-saturated methanol under UV light irradiation, where the photogenerated valence band holes (h⁺) in TiO₂ are efficiently scavenged by methanol.^27,28 These findings were similar to those observed for bare TiO₂ particles,²¹ but all of the tested Au/TiO₂ samples showed much higher photocatalytic reduction ability than TiO₂. This difference was likely because the loaded Au nanoparticles greatly enhance the charge separation within the nanocomposites by collecting electrons from TiO₂.

Table 1 Structural and optical properties of Au/TiO₂ particles prepared under various heating conditions and adsorption equilibrium constants (K_ad) of DN-BODIPY

Sample	Annealing conditions		Properties of Au/TiO2			K ad ^/ μM	γ eff ^/ ^s ^−1	n ^c
	T/°C	t/h	λ max/nm	d Au /nm (RSD, %)	N Au ^a
a N Au represents the mean number of Au nanoparticles loaded per TiO2 particle. b Obtained from Fig. 5D. γeff represents the combined reactivity of all surface catalytic sites. c n is the slope of the log(<τoff>s^−1)–log(IUV) plot obtained from Fig. 6A.
TiO2	N/A	N/A	N/A	N/A	N/A	0.5 ± 0.1	0.55 ± 0.07	0.22
5 nm Au/TiO2	300	0.5	530	5.5 ± 1.2 (21.8)	21.9 ± 5.7	1.1 ± 0.2	0.61 ± 0.13	0.06
8 nm Au/TiO2	600	4.0	544	8.1 ± 1.5 (18.5)	10.4 ± 2.8	1.0 ± 0.2	0.57 ± 0.11	0.11
14 nm Au/TiO2	700	4.5	572	14.1 ± 3.2 (22.7)	3.0 ± 1.5	0.8 ± 0.2	0.62 ± 0.15	0.24

---

Single-particle observation of photocatalytic reduction

Total internal reflection fluorescence microscopy (TIRFM) was used to investigate the ET-induced photocatalytic reduction of DN-BODIPY over individual Au/TiO₂ particles. To obtain individual Au/TiO₂ nanoparticles, well-dispersed methanol suspensions of Au/TiO₂ with very low concentrations were spin-coated on the cleaned cover glasses, which resulted in only one particle being present in a 12.5 μm × 12.5 μm area. Here, the term “single (or individual) Au/TiO₂ (nano)particle” indicates one Au/TiO₂ particle integrated with several Au nanoparticles present on the surface of one TiO₂ nanoparticle.

Fig. 2A shows typical fluorescence images captured for a single 8 nm Au/TiO₂ particle in Ar-saturated methanol containing DN-BODIPY (2.0 μM) under 488 nm laser and UV irradiation. Individual particles showed a number of fluorescence bursts that had signals higher than the background (also see Fig. 5A). Control experiments confirmed that Au/TiO₂ (or TiO₂), DN-BODIPY, and UV excitation are essential to the generation of fluorescence bursts. The locations of the fluorescence bursts, which were determined by fitting two-dimensional Gaussian functions to the distribution of fluorescent spots, are likely distributed over the particle (see the red dots in the transmission image of Fig. 2A). Moreover, the fluorescence lifetimes of the in situ generated bursts over individual Au/TiO₂ particles were measured by combining confocal microscopy with a time-correlated single-photon counting (TCSPC) system. The fluorescence decay profiles were well-fitted by the biexponential function: a₁·exp(τ/τ₁) + a₂·exp(τ/τ₂). As shown in Fig. 2C, the fluorescence burst (component 2) depicted in red in Fig. 2B exhibited a much longer lifetime than the background signal (component 1) from DN-BODIPY in solution (see blue mark in Fig. 2B). These findings suggest that such a sudden increase in intensity corresponds to the generation of the fluorescent product (i.e., HN-BODIPY). However, when compared to free HN-BODIPY molecules in the bulk solution (τ₂ = 3.7 ns), these in situ-generated products on the Au/TiO₂ surface showed much shorter lifetimes (τ₂ = 1–2 ns). This difference may have been due to the intermolecular ET from the excited BODIPY chromophore to the TiO₂ and/or Au nanoparticles.²¹ Although the fluorescence quenching and/or enhancement of HN-BODIPY by Au nanoparticles might be considered, detailed studies to clarify the underlying mechanisms are beyond the scope of this article.

Fig. 2 (A) Optical transmission of a single 8 nm Au/TiO₂ particle immobilized on a cover glass and fluorescence images of the same particle in Ar-saturated DN-BODIPY solution (2.0 μM in methanol) under 488 nm laser (I₄₈₈ = 0.1 kW cm⁻²) and UV irradiation (I_UV = 0.5 W cm⁻²). The fluorescence images display a series of successive images exhibiting on and off events. The acquisition time of an image was 50 ms. The scale bars are 500 nm. The red dots in the transmission image indicate the location of fluorescence bursts. The accuracy of the location was about 20–30 nm. (B) Typical trajectories of fluorescence intensity (black) and lifetime (green, bin time is 500 ms) observed for a single 8 nm Au/TiO₂ particle in Ar-saturated DN-BODIPY solution (2.0 μM, methanol) under 485 nm laser (I₄₈₈ = 2 kW cm⁻²) and UV irradiation (I_UV = 0.5 W cm⁻²). (C) Fluorescence decay profiles of the burst emission and background emission in the time regions indicated by red and blue in panel B, respectively. Black lines indicate biexponential curves fitted to the data.

Conversely, each decrease in intensity marks the disappearance of the fluorescent product from the surface of the nanoparticle. Our previous experiments showed that photobleaching or blinking of single HN-BODIPY molecules occurs on much longer time scales under similar laser intensities,²¹ which suggests that the observed sudden decreases in intensity are attributed to either the consecutive reduction or the dissociation of adsorbed HN-BODIPY. Although the occurrence of the fluorescence bursts generated from one Au/TiO₂ nanoparticle is stochastic in nature, their statistical properties should provide valuable information regarding the kinetic mechanism of photocatalysis over Au/TiO₂ at the single-particle level.

Temporal and spatial heterogeneities of photocatalytic activity

Fig. 3A shows the time profile for the turnover rate of the on-off cycle every 10 s (TOF) over one Au/TiO₂ particle during 10 min of UV irradiation (see Section S4 for the method used to calculate TOF†). It is clear that the TOF values over the same particle vary temporally, which is indicative of the fluctuation in the photocatalytic activity of one particle at different times. To quantify the fluctuation in activity, the mean value of the turnover rate (<TOF>) and its relative standard derivation (RSD) over each particle were calculated as shown in Fig. 3B. Based on the statistical analysis of 10 different particles, most of the bare TiO₂ particles showed lower activity (<TOF>, 0.23–0.60 s⁻¹) and greater temporal variation (RSD, 30%–51%) in the photoreduction of DN-BODIPY than Au/TiO₂ samples. Moreover, bare TiO₂ particles possess a much broader distribution in activity. Indeed, approximately 20% of the TiO₂ particles did not show any detectable photocatalytic activity (for more than 20 individual particles examined), while approximately 10% of the TiO₂ particles showed excellent activity (e.g., see Fig. 3B). These results suggest that there was heterogeneous activity among particles, which is always masked in the ensemble-averaged measurements. Conversely, almost all of the Au/TiO₂ particles could effectively photocatalyze the reduction of DN-BODIPY, implying that Au nanoparticles can preferentially trap electrons from TiO₂ and then deliver them to DN-BODIPY. In other words, the reduction of DN-BODIPY occurs more easily on the loaded Au nanoparticles. The surface of TiO₂ has a number of electron trapping sites with different energies and structures, which results in an extremely broad distribution in the reaction rates;^22,29–31 therefore, three different sized Au/TiO₂ samples showed smaller fluctuations in activity than bare TiO₂ particles (Fig. 3B). Consequently, the precise positions of fluorescent spots that appear on the surface of single particles were analyzed.

Fig. 3 (A) Fluctuation of TOF over a single 5 nm Au/TiO₂ particle ([DN-BODIPY] = 2.0 μM, I_UV = 0.5 W cm⁻²). (B) Distribution of <TOF> and its RSD over 10 different individual particles for TiO₂ (black), 5 nm Au/TiO₂ (blue), 8 nm Au/TiO₂ (green), and 14 nm Au/TiO₂ (red).

Fig. 4A and 4D show the trajectories of the fluorescence intensity obtained for single TiO₂ and 14 nm Au/TiO₂ particles in Ar-saturated DN-BODIPY solution (2.0 μM in methanol) under 488 nm laser (I₄₈₈ = 0.1 kW cm⁻²) and UV irradiation (I_UV = 0.5 W cm⁻²), respectively. Because the fluorescent spots from individual bursts are spatially distributed on the surface of the particle (Fig. 4B and 4E, see the corresponding colors in panels A and D, respectively), the locations of the fluorescence bursts were determined with an accuracy of about 20–30 nm by employing a two-dimensional Gaussian distribution.^13,14 Interestingly, as shown in Fig. 4C and 4F, when compared to those for a single TiO₂ particle (over 5 centers), a single 14 nm Au/TiO₂ particle had a limited number of reactive centers (2–3 centers) (see red circles in panel F), being equal to the number of Au nanoparticles loaded onto one TiO₂ particle. A similar tendency was observed when over 10 individual particles were examined. These results suggest that the reduction of DN-BODIPY over Au/TiO₂ occurs more easily on the surface of Au nanoparticles than unmodified TiO₂ surfaces.

Fig. 4 (A, D) Fluorescence intensity trajectories obtained for single TiO₂ (A) and 14 nm Au/TiO₂ (D) particles in Ar-saturated DN-BODIPY solution (2.0 μM, in methanol) under 488 nm laser (I₄₈₈ = 0.1 kW cm⁻²) and UV irradiation (I_UV = 0.5 W cm⁻²). (B, E) The locations of fluorescence bursts determined using centroid analysis of each fluorescent spot obtained for single TiO₂ (B) and 14 nm Au/TiO₂ (E) particles. The integration time per frame was 50 ms. See the corresponding colors in panels A and D for panels B and E, respectively. (C, F) The spatial distribution of reactive sites determined by fitting a two-dimensional Gaussian function to the fluorescence spot distributions (dots in panels C and E) collected from identical single TiO₂ (C) and 14 nm Au/TiO₂ (F) particles. Note that the locations of the burst with a short duration, i.e., a small number of images, were determined by centroid analysis. The red circles in panel F indicate the highly active reaction centers. The particle size of TiO₂ is about 150 nm.

[S]-Dependence of τ_off and τ_on

To clarify the fundamental principles governing the photocatalytic activity of Au/TiO₂, we examined the substrate concentration ([S])-dependent photocatalytic behavior at the fixed UV light intensity (I_UV) of 0.5 W cm⁻². In the single-particle fluorescence intensity trajectory (Fig. 5A), the actual photocatalytic events could be separated into two characteristic durations, τ_off and τ_on, where τ_off is the characteristic time prior to the formation of fluorescence products on the TiO₂ or Au/TiO₂ and τ_on is the characteristic time for which persistent emission is exhibited, which might be related to the consecutive reduction or dissociation of fluorescent products. As shown in Fig. 5B and 5C, respectively, the distributions of τ_on and τ_off are well fitted with a single-exponential decay function (R² > 0.97). A noticeable deviation at longer on time (e.g., τ_on > 4 s in Fig. 5C) would be partially due to the multiple burst events.

Fig. 5 (A) A typical fluorescence intensity trajectory observed for a single 8 nm Au/TiO₂ particle in an Ar-saturated DN-BODIPY solution (2.0 μM, in methanol) under 488 nm laser and UV light irradiation. The corresponding fluorescence intensity histogram is shown in the right panel. The green solid and dashed lines indicate the Gaussian-fitted background and the threshold level separating the on and off states, respectively, which corresponds to 3σ, where σ is the standard deviation. (B) Off and (C) on time distributions constructed from over 150 events for 20 different single 8 nm Au/TiO₂ particles ([DN-BODIPY] = 2.0 μM, I_UV = 0.5 W cm⁻²). (D, E) DN-BODIPY concentration dependence of <τ_off> (D) and <τ_on> (E) obtained for TiO₂ (black), 5 nm Au/TiO₂ (blue), 8 nm Au/TiO₂ (green), and 14 nm Au/TiO₂ (red). The solid lines in panel D were obtained from eqn (1), and the fitting parameters are summarized in Table 1. The solid lines in panel E are guides for the eye.

The inverse of the average values of τ_off (<τ_off>) and τ_on (<τ_on>) are plotted against [S] (Fig. 5D). The values of <τ_off>⁻¹ over all Au/TiO₂ samples were found to be much larger than those over bare TiO₂ particles under all tested substrate concentrations. This result is qualitatively consistent with the bulk measurements, in which Au/TiO₂ exhibited excellent photocatalytic activity when compared to the bare TiO₂ (Fig. S2†). Moreover, the <τ_off>⁻¹ values are strongly dependent on [S]. At higher concentrations, adsorption equilibrium is attained more quickly, which facilitates the reduction of DN-BODIPY to the fluorescent products, leading to enhancement of <τ_off>⁻¹. The [S]-dependence of the product formation rate can be described by the Langmuir–Hinshelwood equation:^16,32

where K_ad is the apparent adsorption constant for S, and γ_eff is the apparent rate constant for HN-BODIPY formation, which represents the reactivity of all catalytic sites on each Au/TiO₂ or TiO₂ particle. As summarized in Table 1, the resulting K_ad values increased slightly from 0.8 ± 0.2 to 1.1 ± 0.2 μM⁻¹ as the Au nanoparticle size decreased from 14 nm to 5 nm, but all of the Au/TiO₂ systems showed larger K_ad values than that of the TiO₂ system (0.5 ± 0.1 μM⁻¹),²¹ which is of the same order of magnitude as those for nitrobenzenes.³³ These results imply that DN-BODIPY prefers to adsorb onto the Au surface of Au/TiO₂, and smaller Au nanoparticles exhibit a stronger substrate binding affinity, probably because of a sufficient amount of Au atoms with a low coordination number.³⁴ When these findings are combined with the fact that the locations of the fluorescence bursts are likely distributed on the loaded Au nanoparticles (Fig. 4), it is reasonable to assume that the reduction of DN-BODIPY over Au/TiO₂ is mainly triggered by accepting electrons accumulated in the Au nanoparticles. This assumption is also supported by previous proposals that the reduction of nitrobenzene and oxidation of methanol occur selectively on the surface of deposited metal nanoparticles (Pt or Ag) and TiO₂, respectively.^5,35 In contrast, the DN-BODIPY concentration had a negligible effect on <τ_on>⁻¹, which means that the disappearance of HN-BODIPY from the surface of the nanoparticles does not involve the substrate-assisted steps.¹⁶ Another possible explanation is that the decrease in τ_on caused by the substrate-assisted product dissociation was compensated for by an increase in τ_on owing to the integration of multiple product molecules. It should be pointed out that when K_ad[S] > 1, the role of the substrate can be omitted from the kinetic mechanism. Thus, all subsequent experiments were conducted using 2.0 μM DN-BODIPY to clearly identify other important factors influencing the Au/TiO₂ behavior.

Interrelated I_UV- and d_Au-dependence of τ_off and τ_on

The effects of I_UV on τ_off and τ_on are illustrated in Fig. 6A and 6B, respectively. Because <τ_off>⁻¹ is proportional to the total number of catalytic sites on each Au/TiO₂ particle, which is related to both N_Au and the surface area of one Au particle (A_Au), the catalytic reactivity per Au nanoparticle per surface area (<τ_off>_s⁻¹, s⁻¹ particle⁻¹ nm⁻²) can be used to evaluate the size-dependent activity of single Au/TiO₂ particles:

Fig. 6 UV light intensity dependence of <τ_off>_s⁻¹ (A) and <τ_on> (B) obtained for 5 nm Au/TiO₂ (blue), 8 nm Au/TiO₂ (green), and 14 nm Au/TiO₂ (red) ([DN-BODIPY] = 2.0 μM).

It was found that increasing I_UV leads to an increase in <τ_off>_s⁻¹ owing to the accelerated formation of photogenerated electrons in TiO₂ and enhanced sequential ET from TiO₂ to Au (Fig. 6A). In addition, <τ_off>_s⁻¹ increased as d_Au decreased at a certain value of I_UV. For example, <τ_off>_s⁻¹ of 5 nm Au/TiO₂ was 1.95 times higher than that of 14 nm Au/TiO₂ at an I_UV of 0.015 W cm⁻². However, this difference in <τ_off>_s⁻¹ between 5 nm and 14 nm Au/TiO₂ particles gradually decreased to 1.03 times as the I_UV increased to 0.5 W cm⁻².

It is interesting to note that, except for 5 nm Au/TiO₂, I_UV affects the time required for HN-BODIPY to disappear from the Au/TiO₂ surface (Fig. 6B). In particular, the τ_on value for the 14 nm Au/TiO₂ sample greatly increased from 0.58 ± 0.02 s to 1.0 ± 0.1 s as I_UV increased from 0.015 W cm⁻² to 0.5 W cm⁻². If the duration of the on time is ascribed to the consecutive reduction of fluorescent HN-BODIPY to non-fluorescent reduced species, i.e., 4-amino-3-nitrophenyl-BODIPY,²¹ the excitation of Au/TiO₂ with weak UV light should prolong the on time of the fluorescence bursts. Because this was not the case in the present study, it is possible to exclude the contribution of the HN-BODIPY reduction to the fluorescence trajectories. Furthermore, it was noted that the burst events, especially for 14 nm Au/TiO₂ at strong UV power, included single HN-BODIPY molecules and the integration of multiple product molecules (e.g., the burst at 20 s in Fig. 5A). In other words, a new HN-BODIPY molecule is generated before earlier ones dissociate away from the particle surface. Such multi-level on events directly reflect a multitude of either reduction sites that can undertake photocatalysis in parallel or adsorption sites at which products can remain on the surface before dissociation. The difference between <TOF> and <τ_off>⁻¹ is thus attributed to the multiple burst events involved in the single-molecule turnover trajectory. Because no apparent changes in the photocatalytic activity and size of the Au nanoparticles were observed during photoirradiation (Fig. S4†), the number of adsorption sites would not be affected by UV power. In contrast, the effective reduction sites, which are related to the accumulated electron concentration in Au/TiO₂, may be controlled by I_UV and the electron storage ability of Au particles. Larger Au nanoparticles can store more electrons (Table S1†), and the stronger excitation intensity can induce more electrons to transfer from TiO₂ to Au, both of which would increase the probability of multiple parallel reduction reactions in Au nanoparticles, thereby leading to longer τ_on values. Accordingly, the proportion of bursts with multiple levels to the total number of detected bursts increased from 27% to 64% when I_UV increased from 0.015 W cm⁻² to 0.5 W cm⁻² in the 14 nm Au/TiO₂ system. A striking correlation (R² = 0.98) was also found between τ_on and the storage ability of Au under saturated UV excitation conditions (see Section S6 for details†). In this sense, τ_on cannot simply be determined by the dissociation of HN-BODIPY, but is also influenced by the number of electrons accumulated in Au nanoparticles loaded on TiO₂. Overall, these findings suggest that the effects of I_UV and d_Au on τ_off/on are intricately interrelated and that the photocatalytic behavior of TiO₂ loaded with larger Au nanoparticles is more dependent on I_UV when compared to smaller Au nanoparticles.

Kinetic mechanism of Au charging-controlled photocatalytic reduction

Based on the above experimental results, we propose a possible mechanism for the photocatalytic reduction of DN-BODIPY over Au/TiO₂, as illustrated in Fig. 7. The results confirmed that the DN-BODIPY molecules prefer to be adsorbed on the Au surface rather than the TiO₂ surface (Fig. 5D and Table 1), and their reduction likely occurs on Au nanoparticles (Fig. 4). The accumulation of electrons in the Au nanoparticles increases their E_F^* step by step until the E_F^* exceeds the reduction potential of DN-BODIPY (namely, the charging process of Au),^8–10 after which the electron flow from the charged Au nanoparticles to adsorbed DN-BODIPY molecules leads to the formation of HN-BODIPY concomitantly with turning on fluorescence. The reduction of single or multiple DN-BODIPY molecules adsorbed on the surface eventually depletes the accumulated electrons in the Au, which is followed by a decrease in E_F^* (namely, the discharging process of Au). This decrease in E_F^* is a preceding event before the transition to the off state.

Fig. 7 Schematic of kinetic mechanism for photocatalytic reduction over Au/TiO₂. S and P are the substrate (DN-BODIPY) and fluorescent product (HN-BODIPY), respectively. The blue, brown, and orange spheres show the TiO₂, and the neutral and charged Au nanoparticles, respectively. The charging process occurs during the τ_off until the E_F of a Au nanoparticle raises to a certain level, at which a product is generated and discharging occurs. This event marks the end of τ_off and the start of τ_on. Discharging continues during τ_on to generate more product molecules until the E_F drops to a certain level; the τ_on does not end, however, until all product molecules dissociate from the nanoparticle surface.

To better understand the charging-discharging process of Au nanoparticles influencing τ_off and τ_on, we developed a kinetic model that describes the Au charging-controlled photocatalytic reduction of DN-BODIPY over single Au/TiO₂ particles under substrate-saturating conditions (i.e., K_ad[S] > 1). It should be noted that once any one of Au nanoparticles are changed, photocatalytic reduction of DN-BODIPY is triggered. Consequently, ET through nanocomposite interfaces is resolved into that between single TiO₂ nanoparticles and several individual Au nanoparticles. In light of the photochemical equilibrium state, the following expression for the corresponding reduction in the rate of DN-BODIPY (<τ_off>_s⁻¹) under light-rich conditions was obtained:

where k_et is the ET from the TiO₂ CB to Au nanoparticles, k_rec is the charge recombination rate constant for electron-hole pairs, k_ox is the oxidation rate constant of adsorbed methanol (CH₃OH_ad) by holes, A_Au is the mean surface area of one Au particle, N_Au is the average number of Au nanoparticles per TiO₂ particle, φ is the integrated absorption fraction, [e_CB⁻] is the electron concentration in TiO₂ CB, [R_s] is the concentration of the surface reactive sites that trap the electrons on one Au nanoparticle, which corresponds to the electron storage ability of Au nanoparticles (i.e., the number of electrons per Au nanoparticle needed to equilibrate with the excited TiO₂), and <τ_off>_s⁻¹ is the reduction rate of DN-BODIPY over a single Au particle per surface area under substrate-saturating conditions (see Section S7 for details on the kinetic model†).

Eqn (3) clearly represents a strong dependence of <τ_off>_s⁻¹ on A_Au, N_Au, [R_s], and I_UV. Specifically, the charging process of Au nanoparticles should be among the most fundamental factors controlling the ET characteristics, which provides the dependence of <τ_off>_s⁻¹ on I_UV with a predicted power coefficient of 0.5 through [e_CB⁻]. The linear and square root dependences of the reaction rate on light intensity in TiO₂ photocatalysis have been observed by many research groups. Although the reaction rate is independent of light intensity, it is limited by mass transport at very high light intensities.³⁶ As expected, <τ_off>_s⁻¹ is dependent on I_UV and the correlation coefficients of the regression straight lines indicate a satisfactory linearity between log(<τ_off>_s⁻¹) and log(I_UV) (R² > 0.90) as shown in Fig. 6A. Nevertheless, the obtained slope values (n) were considerably lower than the predicted limit (n = 0.5), implying that the reduction of DN-BODIPY is controlled not only by the photons, but also by the mass transfer. Similar phenomena have been observed during the photocatalytic degradation of oxalic acid on the TiO₂ layer and the photostimulated adsorption of oxygen on ZrO₂.^37,38 Nevertheless, the extent of I_UV-dependence of <τ_off>_s⁻¹ (i.e., n) differed notably among the three different sized Au/TiO₂ samples, being 0.06, 0.11, and 0.24 for 5 nm, 8 nm, and 14 nm Au/TiO₂, respectively (Table 1). In fact, it has reported that the rate of the photocatalyzed reaction can also be described by eqn (4), and that the power coefficient n varies with substrate concentration (c), approaching unity as c → ∞ and conversely n → 0 if c → 0.^38,39

It has also been shown that n is related to the substrate flow rate or oxygen flux, and that enhancement of the photocatalysis rate by I_UV increases as the oxygen concentration or substrate flow increases.³⁷ According to the model of diffusion-limited kinetics developed by Roeffaers and co-workers,¹⁴ however, it is likely that there is no significant depletion of substrate from the bulk solution by catalysis under the present experimental conditions.

Similarly, <τ_off>_s⁻¹ is greatly limited by the interfacial ET process between TiO₂ and Au nanoparticles (i.e., k_et[e_CB⁻][R_s]) under substrate-saturating conditions; thus, it is reasonable to assume that the interdependence of <τ_off>_s⁻¹ on I_UV and d_Au primarily arises from the different electron storage ability of Au nanoparticles as shown in eqn (5):

As d_Au increases from 5.5 to 14.1 nm, A_Au increases from 95.0 to 624.3 nm², and the electron storage ability of Au nanoparticles (q), which increases in proportion to A_Au,⁴⁰ is enlarged significantly from 154 to 1014 electrons particle⁻¹ (Section S6†). That is, [R_s] increased from 154 to 1014 sites per Au nanoparticle. Since the number of surface trapping sites for electrons on 5 nm Au is relatively limited, additional photon flux could not efficiently increase the rate of ET from TiO₂ CB to Au nanoparticles; accordingly, n approaches zero (0.06). In contrast, 14 nm Au nanoparticles provide many more surface reactive sites, which results in their photocatalytic activity being more sensitive to I_UV (n = 0.24). This interpretation is further supported by the positive correlation between n and q (R² = 0.98, Fig. S6†).

Efficiency of photocatalytic reduction on Au/TiO₂

In terms of multiple molecules involving-on events, τ_on reflects the concentration of electrons temporarily accumulated in Au nanoparticles, and thus should be combined with τ_off to assess the photocatalytic ability of Au/TiO₂. Here, we define the <τ_on>/<τ_off> (being equal to <τ_on>_s/<τ_off>_s) obtained under light-controlled conditions (i.e., K_ad[S] > 1) as the single-particle activity factor to rationally evaluate the inherent photocatalytic activity for the reduction over individual Au/TiO₂ particles. Since <τ_off>⁻¹ represents the formation rate of HN-BODIPY and <τ_on> is proportional to the number of simultaneously generated HN-BODIPY molecules, the values of <τ_on>/<τ_off> can reflect the rough amount of produced HN-BODIPY per unit of time.

Fig. 8 illustrates the relationships between d_Au, I_UV, and <τ_on>/<τ_off>. The most interesting finding is the opposite dependence of the photocatalytic activity on d_Au that was observed at different I_UV, which makes the reaction mechanism more complicated. Although a similar contradiction was observed in previous studies,^9,10 their proposed mechanisms in terms of the quantum size and solvent polarity effects on E_F^* cannot fully explain our results. We believe that our findings originate from the size-controlled charging process of the Au nanoparticles leading to the interdependence of τ_off/on on I_UV and d_Au. Because smaller Au nanoparticles require substantially fewer electrons to raise their E_F, the charging process can be accomplished rapidly, even upon weak UV excitation, while the charging of larger Au nanoparticles would be greatly limited because of the relatively insufficient photogenerated electrons in TiO₂. Such superiority causes Au/TiO₂ particles with smaller d_Au values to show better photocatalytic activity at low I_UV. Simultaneously, the limited surface trapping sites for electrons on smaller Au particles results in the photocatalytic behavior of 5 nm Au/TiO₂ being less dependent on I_UV than Au/TiO₂ with larger d_Au values, as discussed above. However, intense UV excitation can produce excess electrons in TiO₂ to efficiently induce Au charging; hence, it greatly reduces the difference in τ_off for Au/TiO₂ with various d_Au (Fig. 6A). Once the reaction attains the new equilibrium E_F^* state, larger Au nanoparticles would shuttle more electrons to reduce multiple DN-BODIPY molecules at one time, leading to a greatly prolonged τ_on (Fig. 6B). Therefore, the rapid charging process accompanied with an extended discharging time will cause enhanced photocatalytic activity for Au/TiO₂ with increasing d_Au at high I_UV.

Fig. 8 d _Au and I_UV dependence on <τ_on>/<τ_off> values obtained for 5, 8, and 14 nm Au/TiO₂.

Conclusions

In conclusion, we investigated the photocatalytic redox reaction of the dinitrobenzene moiety over single TiO₂ or Au/TiO₂ particles using single-molecule fluorescence imaging techniques with a redox-responsive fluorescent probe, DN-BODIPY. Examination of the location of fluorescent spots emitted from the products using super-resolution fluorescence microscopy with single molecule sensitivity allowed us to map the distribution of the reactive centers on the surface of the particles. All of the tested Au/TiO₂ samples showed much higher photocatalytic reduction ability than TiO₂, because the loaded Au nanoparticles greatly enhance the charge separation within the nanocomposites. Statistical analysis of the single-molecule TOFs revealed that the Au/TiO₂ particles exhibit smaller activity fluctuation when compared to bare TiO₂. The dependence of fluorescence on and off times (τ_on and τ_off) on the substrate concentration, UV light intensity, and size of Au nanoparticles was also examined and evaluated in terms of the Au charging-controlled reaction mechanism developed in this study. Our single-molecule, single-particle approach provides further insight into the mechanism by which (photo)catalytic reactions at a heterogeneous interface occur, as well as the transport characteristics of electrons in composite systems consisting of semiconductors, metals, and organic compounds.

Experimental section

Synthesis of DN-BODIPY

8-(3,4-Dinitrophenyl)-1,3,5,7-tetramethyl-4,4-difluoro-4-bora-3a,4a-diaza-s-indacene (DN-BODIPY) was synthesized as previously reported.²¹

Preparation and characterization of Au/TiO₂

A series of Au/TiO₂ particles (5, 8, and 14 nm Au/TiO₂) were prepared by the deposition-precipitation method under various temperatures and times using HAuCl₄ and TiO₂ (Ishihara Sangyo, A-100, 100–200 nm particles) as the raw materials.⁴¹ Detailed descriptions of the morphology (TEM images), crystal structure (powder XRD pattern), and optical properties (steady-state UV-visible diffuse reflectance spectra) of the synthesized Au/TiO₂ particles are given in the Supporting Information (Fig. S1†).

Ensemble-averaged spectral measurements

Steady-state UV–visible absorption and diffuse reflectance spectra were measured using UV–visible–NIR spectrophotometers (Shimadzu UV-3100 and Jasco V-570, respectively). Steady-state fluorescence spectra were measured using a Hitachi 850 fluorescence spectrophotometer.

Single-molecule fluorescence measurements by TIRFM

The experimental setup was based on an Olympus IX71 inverted fluorescence microscope. The details of the experimental setup are described in the Supporting Information†. To obtain isolated TiO₂ or Au/TiO₂ particles, well-dispersed methanol suspensions containing small amounts of TiO₂ or Au/TiO₂ powders were spin-coated onto clean cover glasses. The surface density of the particles on the cover glasses was very low, with only one particle being present in a 12.5 μm × 12.5 μm area. The position of the TiO₂ or Au/TiO₂ particles immobilized on the cover glass was determined from the transmission image obtained by illuminating the samples using a halogen lamp (Olympus, U-LH100L-3). A circular-polarized light emitted from a CW Ar ion laser (Melles Griot, IMA101010BOS; 488 nm, 0.1 kW cm⁻² at the glass surface) was reflected toward a second dichroic mirror (Olympus, DM505) using a first dichroic mirror (Olympus, RDM450), which reflected wavelengths longer than 450 nm and was transparent to wavelengths shorter than 450 nm. The laser light passing through an objective lens (Olympus, UPLSAPO 100XO; 1.40 NA, 100×) after reflection by a second dichroic mirror was completely reflected at the cover glass-methanol interface. This resulted in generation of an evanescent field, which made it possible to detect a single fluorescence dye molecule. For excitation of the TiO₂ or Au/TiO₂ particles, the 365 nm light emitted by a LED (OPTO-LINE, MS-LED-365) and passing through an ND filter was passed through the objective. The light intensity through the objective, immersion oil, and cover glass was measured using a power meter (Ophir, Nova II) equipped with a PD300-UV head. The fluorescence emission from the fluorescent products generated over a single TiO₂ or Au/TiO₂ particle on the cover glass was collected using the same objective, after which it was magnified by a 1.6× built-in magnification changer, passed through a bandpass filter (Semrock, FF01-531/40-25) to remove the undesired scattered light, and then imaged using an electron-multiplying charge-coupled device (EM-CCD) camera (Roper Scientific, Cascade II:512). The images were recorded at a frame rate of 20 frames s⁻¹ and processed using ImageJ (http://rsb.info.nih.gov/ij/) or OriginPro 8.1 (OriginLab). All experimental data were obtained at room temperature. To determine the locations of the reactive site distributed on the surface, the precise positions of the fluorescent spots were analyzed for each image using the ImageJ software with a SpotTracker plugin.⁴²

Single-molecule fluorescence measurements by confocal microscopy

Confocal fluorescence images were taken using an objective-scanning confocal microscope system (PicoQuant, MicroTime 200) coupled with an Olympus IX71 inverted fluorescence microscope. The samples were excited through an oil-immersion objective lens (Olympus, UAPON 150XOTIRF; 1.45 NA, 150×) using a 485 nm pulsed laser (PicoQuant, LDH-D-C-485) controlled by a PDL-800B driver (PicoQuant). The emission from the sample was collected using the same objective and detected by a single photon avalanche photodiode (Micro Photon Devices, PDM 50CT) through a dichroic beam splitter and bandpass filter (Semrock, FF01-531/40-25). All experimental data were obtained at room temperature. The details of the experimental setup and procedures are described in the Supporting Information†.

Acknowledgements

N. W. gives thanks to Prof. Lihua Zhu at Huazhong University of Science and Technology and the CSC (China Scholarship Council) program for their support. T. M. gives thanks to the WCU (World Class University) program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (R31-10035) for their support. This work has been partly supported by a Grant-in-Aid for Scientific Research (Projects 22245022, 21750145, and others) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of the Japanese Government.

References

1. M. Grätzel and J. E. Moser, Electron Transfer in Chemistry, Vol. 5, Wiley-VCH, Weinheim, 2001.
2. P. V. Kamat, J. Phys. Chem. C, 2007, 111, 2834–2860.
3. P. V. Kamat, J. Phys. Chem. C, 2008, 112, 18737–18753.
4. A. Corma and H. Garcia, Chem. Soc. Rev., 2008, 37, 2096–2126.
5. H. Tada, T. Kiyonaga and S. Naya, Chem. Soc. Rev., 2009, 38, 1849–1858.
6. D. V. Talapin, J.-S. Lee, M. V. Kovalenko and E. V. Shevchenko, Chem. Rev., 2010, 110, 389–458.
7. Y. Nakato, M. Shioji and H. Tsubomura, Chem. Phys. Lett., 1982, 90, 453–456.
8. V. Subramanian, E. Wolf and P. V. Kamat, J. Phys. Chem. B, 2001, 105, 11439–11446.
9. V. Subramanian, E. E. Wolf and P. V. Kamat, J. Am. Chem. Soc., 2004, 126, 4943–4950.
10. T. Kiyonaga, M. Fujii, T. Akita, H. Kobayashi and H. Tada, Phys. Chem. Chem. Phys., 2008, 10, 6553–6561.
11. B. M. Weckhuysen, Angew. Chem., Int. Ed., 2009, 48, 4910–4943.
12. M. B. J. Roeffaers, B. F. Sels, H. Uji-i, F. C. De Schryver, P. A. Jacobs, D. E. De Vos and J. Hofkens, Nature, 2006, 439, 572–575.
13. M. B. J. Roeffaers, G. De Cremer, J. Libeert, R. Ameloot, P. Dedecker, A.-J. Bons, M. Bueckins, J. A. Martens, B. F. Sels, D. E. De Vos and J. Hofkens, Angew. Chem., Int. Ed., 2009, 48, 9285–9289.
14. G. De Cremer, M. B. J. Roeffaers, E. Bartholomeeusen, K. Lin, P. Dedecker, P. P. Pescarmona, P. A. Jacobs, D. E. De Vos, J. Hofkens and B. F. Sels, Angew. Chem. Int. Ed., 2010, 49, 908–911.
15. G. De Cremer, B. F. Sels, D. E. De Vos, J. Hofkens and M. B. J. Roeffaers, Chem. Soc. Rev., 2010, 39, 4703–4717.
16. W. Xu, J. S. Kong, Y.-T. E. Yeh and P. Chen, Nat. Mater., 2008, 7, 992–996.
17. W. Xu, H. Shen, Y. J. Kim, X. Zhou, G. Liu, J. Park and P. Chen, Nano Lett., 2009, 9, 3968–3973.
18. X. Zhou, W. Xu, G. Liu, D. Panda and P. Chen, J. Am. Chem. Soc., 2010, 132, 138–146.
19. K. Naito, T. Tachikawa, M. Fujitsuka and T. Majima, J. Am. Chem. Soc., 2009, 131, 934–936.
20. T. Tachikawa and T. Majima, J. Am. Chem. Soc., 2009, 131, 8485–8495.
21. T. Tachikawa, N. Wang, S. Yamashita, S.-C. Cui and T. Majima, Angew. Chem., Int. Ed., 2010, 49, 8593–8597.
22. T. Tachikawa and T. Majima, Chem. Soc. Rev., 2010, 39, 4802–4819.
23. P. Gao, D. Gosztola and M. J. Weaver, J. Phys. Chem., 1988, 92, 7122–7130.
24. A. Corma, P. Concepcion and P. Serna, Angew. Chem., Int. Ed., 2007, 46, 7266–7269.
25. A. Loudet and K. Burgess, Chem. Rev., 2007, 107, 4891–4932.
26. T. Ueno, Y. Urano, H. Kojima and T. Nagano, J. Am. Chem. Soc., 2006, 128, 10640–10641.
27. C.-Y. Wang, J. Rabani, D. W. Bahnemann and J. K. Dohrmann, J. Photochem. Photobiol., A, 2002, 148, 169–176.
28. A. A. Ismail, D. W. Bahnemann, I. Bannat and M. Wark, J. Phys. Chem. C, 2009, 113, 7429–7435.
29. S. A. Haque, Y. Tachibana, R. L. Willis, J. E. Moser, M. Grätzel, D. R. Klug and J. R. Durrant, J. Phys. Chem. B, 2000, 104, 538–547.
30. J. Nelson and R. E. Chandler, Coord. Chem. Rev., 2004, 248, 1181–1194.
31. W. R. Duncan and O. V. Prezhdo, Annu. Rev. Phys. Chem., 2007, 58, 143–184.
32. W. Xu, J. S. Kong and P. Chen, J. Phys. Chem. C, 2009, 113, 2393–2404.
33. J. L. Ferry and W. H. Glaze, J. Phys. Chem. B, 1998, 102, 2239–2244.
34. R. Van Hardeveld and F. Hartog, Surf. Sci., 1969, 15, 189–230.
35. H. Tada, T. Ishida, A. Takao, S. Ito, S. Mukhopadhyay, T. Akita, K. Tanaka and H. Kobayashi, ChemPhysChem, 2005, 6, 1537–1543.
36. D. F. Ollis, E. Pelizzetti and N. Serpone, Environ. Sci. Technol., 1991, 25, 1522–1529.
37. J. Krysa, L. Vodehnal and J. Jirkovsky, J. Appl. Electrochem., 1999, 29, 429–435.
38. A. V. Emeline, A. V. Rudakova, V. K. Ryabchuk and N. Serpone, J. Phys. Chem. B, 1998, 102, 10906–10916.
39. A. V. Emeline, V. Ryabchuk and N. Serpone, J. Photochem. Photobiol., A, 2000, 133, 89–97.
40. A. Toyota and T. Sagara, Electrochim. Acta, 2008, 53, 2553–2559.
41. T. Kiyonaga, T. Mitsui, M. Torikoshi, M. Takekawa, T. Soejima and H. Tada, J. Phys. Chem. B, 2006, 110, 10771–10778.
42. D. Sage, R. Neumann Franck, F. Hediger, M. Gasser Susan and M. Unser, IEEE Trans Image Process, 2005, 14, 1372–1383.

---

Footnote

† Electronic Supplementary Information (ESI) available: Movie showing the single-molecule, single-particle fluorescence imaging of the reduction of DN-BODIPY on 8 nm Au/TiO₂ (Movie S1). Experimental details including the descriptions of the movie (S0), the experimental methods (S1), the preparation and characterization of Au/TiO₂ (S2), the ensemble-averaged photocatalytic activity (S3), the turnover rate of off-on cycle (S4), the influence of photoirradiation on the Au particle size (S5), the size dependent of electron storage ability of Au nanoparticle (S6), and the kinetic analysis of Au/TiO₂ photocatalytic reactions (S7). See DOI: 10.1039/c0sc00648c/

---