A Comparison of Photocatalytic Activities of Gold Nanoparticles Following Plasmonic and Interband Excitation and a Strategy for Harnessing Interband Hot Carriers for Solution Phase Photocatalysis.

Light driven excitation of gold nanoparticles (GNPs) has emerged as a potential strategy to generate hot carriers for photocatalysis through excitation of localized surface plasmon resonance (LSPR). In contrast, carrier generation through excitation of interband transitions remains a less explored and underestimated pathway for photocatalytic activity. Photoinduced oxidative etching of GNPs with FeCl3 was investigated as a model reaction in order to elucidate the effects of both types of transitions. The quantitative results show that interband transitions more efficiently generate hot carriers and that those carriers exhibit higher reactivity as compared to those generated solely by LSPR. Further, leveraging the strong π-acidic character of the resulting photogenerated Au+ hole, an interband transition induced cyclization reaction of alkynylphenols was developed. Notably, alkyne coordination to the Au+ hole intercepts the classic oxidation event and leads to the formation of the catalytically active gold clusters on subnanometer scale.

Introduction

Plasmonic noble metal nanoparticles (such as those of Au, Ag, or Cu) exhibit strong light absorption through excitation of a localized surface plasmon resonance (LSPR). The strong light absorption, robust nature, recyclability, and large surface areas of these nanoparticles render them candidates as photocatalytic materials. Therefore, these nanoparticles have recently been examined as photocatalysts for selective chemical synthesis by harvesting visible light to drive reactions at ambient conditions.[1,2] Although the catalytic mechanism is not fully understood, hot carriers (here referred to as high kinetic energy carriers) generated from surface plasmon excitation have been strongly implicated for driving some catalytic reactions.[1,3,4] These plasmonic noble metal nanoparticles also have strong absorption on the high-energy side of their plasmon resonances due to intrinsic interband transitions (Figures B and 2A). Limited studies of gold nanoparticles supported on metal oxides show higher photocatalytic activities in oxidation reactions for interband transitions as compared to that of LSPR. The higher potential for the interband transition results from the creation of a hot hole with a high reduction potential in the d band and a hot electron, to some extent, in the sp band.[5−7]

Figure 1

Carriers generated after photoexciting metallic nanoparticles at the LSPR and interband transition. (A) A classical description of LSPR shows oscillation of collective electrons driven by the electric field (left). The LSPR decay generates a population of hot carriers at various energies (right). The integral energy of both e– (red area) and h+ (blue) occupation is equivalent to a single photon energy. A quantum mechanical description of LSPR uses red arrows to represent a superposition of multiple isoenergetic, single-electron transitions when one photon is absorbed. (B) Interband transition results from absorption of a higher energy photon (left) and directly generates an e––h+ pair (right).

Figure 2

Photoinduced fast etching of GNRs in an FeCl3 solution at room temperature. (A) Optical spectra of colloidal CTAB capped GNRs in water with depiction of the interband and LSPR absorptions. (B) Spectral progression of a typical reaction under 450 nm excitation. (C) Representative TEM images and average particle sizes showing tip etching of GNRs in B. (D) Linear dependence of λlong vs GNR length (data in Table S3). The fitting does not propagate the errors in length to the final errors of intercept and slope values.

To gain insights into how hot carriers are generated under specific excitation wavelength, recent theoretical work predicted that photoexcitation of Au or Ag nanoparticles with photon energy below the interband energy can create hot s electrons or hot s holes in the intraband. However, when the photon energy is above the interband energy threshold, more hot carriers are generated from direct excitation of d–sp transitions even though these carriers do not have as high kinetic energy as those produced in the intraband transition.[8] These findings imply that photoexcitation at surface plasmon resonance, which usually has lower energy than the interband transition, does not produce hot carriers efficiently. This is not surprising when considering how the energy of an absorbed photon is divided in these two excitation regimes. In a classical description (Figure A, left), the LSPR photon energy is split between carriers in the sp band and their distribution among all continuous energy states approximately follows a Fermi–Dirac distribution after the LSPR decays.[1] Consequently, the probability of having high-energy carriers is very low (Figure A, right). A more rigorous description is a quantum mechanical picture of superposition of many single-electron, quantized energy transitions (Figure A, right).[4] When considering momentum conservation in this discrete energy picture, the hot carrier generation is relatively weak because there is only a limited joint density of states in the particles with finite momentum that are matched with the LSPR wave vector.[8] In contrast, interband transitions efficiently create hot carriers because the photon energy is concentrated to a hot electron–hole pair (Figure B, right). In this regard, an important question is raised: is the ability to generate low populations of higher kinetic energy carriers (i.e., hotter carriers) by LSPR responsible for the observed light-driven catalysis by plasmonic metal nanoparticles, as compared to interband transitions, which produce a larger number of “cooler” hot carriers? In this work, we provide evidence to support the latter mechanism.

Results and Discussion

Photoinduced Etching of Gold Nanorods

Recent research on photocatalysis of plasmonic nanoparticles has shifted heavily toward elucidating catalytic activity originating from LSPR and away from catalysis induced by interband transitions. In contrast, there is a dearth of experimental assessment of these two excitation pathways and the resulting mechanism of catalysis. Herein, the photocatalytic activities of gold nanoparticles following LSPR and interband excitation are compared in two reactions: the oxidative etching of gold nanoparticles with FeCl3 and the cyclization of 2-(phenylethynyl)phenol. Colloidal gold nanorods (GNRs) and spheres (GNSs) were chosen as photocatalysts to avoid any effect from supporting or heterojunction materials. Moreover, GNRs have longitudinal LSPR absorption that is well separated from the interband transition absorption, allowing the evaluation of only the contribution of LSPR to the photocatalysis. Comparison of the catalytic performance of the particles as a function of the excitation wavelength supports the conclusion that interband transitions provide for greater catalytic activity than direct LSPR.

In the first reaction, the etching rates of GNRs and GNSs by a FeCl3 solution in the dark and LSPR and interband excitations were compared. GNRs and GNSs gave similar results in this reaction; therefore GNRs were selected for examination. Previous studies have shown that, under dark conditions, etching occurs selectively at the tips of the GNRs and follows eq because the gold ion product creates a strongly coordinated complex with halide ions so that the reduction potential of Au+/Au is lower than that of Au3+/Au and Fe3+/Fe2+.[9] To quantify the reaction rate, the blue-shifting rate of the longitudinal LSPR absorption was measured (Figure S2). This absorption peak is an effective probe for reaction rate as it is well separated from absorption of other reactants and products in solution. TEM and UV–vis spectroscopy proved a proportional dependence of this blue-shifting rate to reaction rate due to the selective etching at the GNR tips (see the Supporting Information). A first order rate constant of kdark = 3.41 × 10–7 s–1 was measured. The reaction showed first order dependence on [Fe3+] and an activation energy of 2.87 kcal/mol. These observations are consistent with homogeneous reaction with low absorption of Fe3+ on the GNR surface.

In order to study the light driven reaction, a monochromatic light source from a xenon lamp equipped with narrow band-pass filter (10–12 nm fwhm) was used and the reaction was monitored via a UV–vis spectrometer after various irradiation times (Figure B). TEM inspection showed a similar tip etching profile as in the reaction performed in the dark (Figure C). Upon photoexcitation, the etching reaction occurred through a pathway that generates soluble Au3+ as an intermediate that further etches the GNRs (see eqs S8 and S9 in the Supporting Information). However, the overall reaction still followed eq . The reaction rate was again measured via the blue-shifting rate of the longitudinal LSPR absorption (Figures D, 3A, 3B). At first glance, the etching rate appears faster under photoexcitation. Noticeably, interband excitation resulted in the highest etching rate while the rate following longitudinal LSPR excitation showed only an incremental difference as compared to kdark (Table S1). Kinetics studies confirmed the first order of [Fe3+] in the rate equation and the linear dependence of the reaction rate to the excitation power (Supporting Information). Moreover, the increase in local temperature around the GNRs is very small (<1 K), and the rise in temperature of the reaction solution is just a few degrees above room temperature (Table S1) under our experimental conditions, and therefore the photothermal enhancement of the reaction rate can be ignored.[10,11] Taken together, these observations support the hypothesis that photoexcitation is responsible for promoting the etching reaction. It is important to note that the self-catalysis and the simplicity of this reaction provide a good model to understand the catalysis of hot carriers.

Figure 3

Kinetic studies of photoinduced fast etching of GNRs in an FeCl3 solution at room temperature. (A) Illustration of that photoexcitation promotes fast etching, and the reaction rate is measured via the blue-shifting rate of the longitudinal SPR absorption (Supporting Information). (B) The etching rates at various wavelengths and absorbed photon powers. (C) Reaction rate normalized to each single photon absorbed, with respect to optical spectra of GNRs. (D) Proposed mechanism of the etching reaction.

Since the reaction rate constant increased linearly with the photon flux, the rate constant, kexcitation, was normalized per single photon absorbed so that the rate can be compared properly across excitation wavelengths (Figure C, section I.1 in the Supporting Information describes how to quantify the absorbed but not the scattered photons). Strikingly, the rate constant was 3 to 4 orders of magnitude higher for the interband transition as compared to the longitudinal LSPR. The rate in the region of 480 to 600 nm excitation did not show a catalytic enhancement due to the transverse LSPR. In fact, the rate in this region was still very high as compared to the longitudinal LSPR due to the tail of the interband absorption starting at ∼600 nm.[12] The quantum yield also exhibited a similar trend (Figure S10A). As mentioned in the previous paragraph, the high energy photon does not provide enough thermal energy to activate the reaction, regardless of the photon energy; therefore the increase of the reaction rate is mainly due to the higher efficiency of generating carriers. Finally, in order to probe the perspective of observing this reaction under sunlight, we simulated the reaction rates at each wavelength after accounting for the solar spectrum and the GNR optical absorption (Figure S10B). The calculated rates remained a few orders of magnitude higher activity for the interband excitation.

A stepwise mechanism for the etching reaction is proposed (Figure D). The first absorbed photon promotes an interband transition that results in a hot electron that reduces Fe3+ on the surface of the GNR to give an Fe2+ product and effectively leaves a positive charge on the GNRs. This step is repeated three times to produces an Au3+ product. The Au3+ further reacts with the GNRs to give the final product Au+. This mechanism is supported by our detection of Au3+ intermediates, the observation of Au3+ further etched GNRs immediately after stopping irradiation, and that hole scavengers (acetone, methanol) resulted in decreased reaction rates (Supporting Information). With this mechanism of hot carriers catalyzing the reaction, the aforementioned reaction rates and quantum yield measurements indicate a significantly better efficiency of generating hot carriers for catalysis via interband transition as compared to direct surface plasmon excitation. These observations and conclusions beg the next question: can this mechanism be harnessed effectively for catalysis in liquid solution?

Photoinduced Gold Nanoparticle Catalysis

Over the past decades, several studies involving light driven oxidation[13−20] and reduction[21−28] reactions catalyzed by semiconductor supported gold nanoparticles (GNPs) excited at their plasmon resonance have been described.[29] In these studies, a semiconductor support is often required to perform the electron–hole separation. Notably, the probability of interband transition, based on the intensity of optical absorption, accounts for roughly 50% of the overall optical absorption at the transverse LSPR frequency of GNRs or at the single LSPR frequency of GNSs. Therefore, we hypothesized that photocatalysis activities observed at the plasmon frequency in some of these studies might originate from the hidden tail of the interband transition similar to those observed in the etching study described above.

Photoinduced nanoparticle-catalyzed redox reactions are proposed to occur by a mechanism involving transfer of a hot electron on the surface of the semiconductor bulk, resulting in formation of a hole at the metal. In accord with this classic operational mode and the mechanism proposed above for carrier generation in GNPs via interband transition, we envisioned improved generation of hot carriers for catalysis. More specifically, we posited that the photogenerated Au+ hole would have strong π-acidic character,[30−33] which could be leveraged for alkyne activation to intercept the oxidation event and drive to downstream catalysis.

On the basis of this hypothesis, 2-(phenylethynyl)phenol was chosen as a model substrate, and the corresponding cyclization reaction has been previously studied using gold nanoparticle catalysts and chemical oxidants.[34] CTAB-coated gold nanospheres with 40 nm diameter were initially used as catalysts in the cyclization. Additional CTAB was necessary for stabilizing the catalyst and solublizing the alkynylphenol into the aqueous media. The GNS-catalyzed reaction at 60 °C under 26 W household fluorescent compact light (CFL) irradiation produced 2-phenylbenzofuran in excellent yield (99%). On the other hand, blank and dark experiments in the absence of the catalyst and light irradiation respectively gave only trace amounts of the benzofuran product (<5%). Given these observations, the influence of incident light on the performance of cyclization was studied using xenon light with various wavelengths ranging from 390 to 660 nm with normalized incident photon numbers (Figure A). The studies found that the reaction proceeded with poor chemical conversions under the irradiation of light with wavelengths ranging from 480 to 660 nm. Notably, the low conversion at LSPR suggests that the cyclization was not driven effectively by the local surface plasmonic effect. On the other hand, nearly quantitative yields of the desired benzofuran were obtained when irradiating the reaction mixture in the near-ultraviolet region (390 and 420 nm). The same trend in reaction conversions was obtained when using GNSs with various diameters as well as GNRs with different aspect ratios (see the Supporting Information). In addition, a nearly linear relationship between the conversion of phenol and the power of the incident light was established (Figure B). These results advocate for a mechanism wherein hot carriers from interband excitation have a dramatic impact on the performance of cyclization. We noted that the reaction conversion showed drastic change when moving to near-ultraviolet excitation and that this trend does not correlate proportionally to the interband absorption as in Figure A. While a clear explanation for the phenomenon awaits further study, we posit that the reaction may have a threshold at a particular photon energy for the formation of the catalytically active clusters.

Figure 4

Photoinduced gold nanoparticle catalysis. (A) UV–vis spectrum of GNSs (diameter = 40 nm) and the catalytic activities of aforementioned GNSs using light source at various wavelength. Number of incident photons was normalized. (B) The power dependence of reaction yield using GNSs as catalyst and xenon lamp with 390 nm filter. (C) Plot-time–conversion studies under both “light always on” (red) and “light on/off” (blue) conditions; gray shadow means during this period the light was turned off and reaction was under dark conditions. 23W CFL was used as light source. (D) The major Au2 cluster formed in situ in supernatant and detected by HR-ESI mass spectroscopy. See HR-ESI assignment of other Au2 and Au4 clusters in the Supporting Information.

In order to gain further insight into the mechanism of the photocatalysis, plot-time–conversion studies were performed under both “light always on” and “light on/off” conditions (Figure C). A reaction-induction period with “light on” of 20 min was observed, after which the product was steadily formed even without further light irradiation. Examination of the supernatant (see the Supporting Information) revealed that during the irradiation period a soluble catalyst was formed in situ from the GNPs. Most notably, several subnanosized Au2 and Au4 clusters complexed with the phenol substrates were identified in the supernatant solution by HR-ESI mass spectroscopy (Figure D). In accord with observations of Corma[35] and Hutchings,[36] these subnanosized clusters exhibit striking reactivity as catalysts in the cyclization reaction. In terms of scope, alkynylphenols with steric or electronic modifications on both arenes were found to be compatible under the same reaction conditions, affording the corresponding benzofurans in good yields (see the Supporting Information).

By analogy to the iron-promoted etching of GNPs, a mechanism involving initial interaction of alkynylphenol on the GNPs’ surface was proposed (Figure ). This alkynylphenol substrate acts as the electron acceptor (much like Fe3+) for the hot electron produced from the light-driven interband transition. Subsequent oxidation on the Au+ hole and strong alkyne coordination to gold generates the corresponding nanoclusters that are leached into the solution. Compared to our previous catalysis[34,37−43] in the use of dendrimer capped GNPs with iodonium oxidants, the model phenol substrate plays a key role of not only an oxidant of GNPs but also a stabilizing ligand in the forming cluster.

Figure 5

Proposed mechanism of the leaching process of GNPs.

Conclusions

While localized surface plasmon resonance (LSPR) provides a powerful platform for nanoparticle catalysis, our studies suggest that in some cases interband transitions should be considered as an alternative mechanism of light-driven nanoparticle catalysis. The benefits already demonstrated by plasmonic nanostructures as catalysts provided the impetus for examining complementary activation modes based on the metal nanoparticle itself. Leveraging these transitions has the potential to provide a means to highly active catalysis modes that would otherwise be challenging to access. For example, only a few methods have been reported for the preparation of highly active metal catalysts on a subnanosized scale, thus limiting their exploitation and study in catalysis.[44] This work suggests a novel and facile strategy for the formation of highly active gold nanocluster catalysts by light illumination of the interband transitions in the presence of the appropriate substrate. Finally, these studies implicated interband transitions, rather than LSPR, as important for the specific reactions and catalysts under study and should be applied judiciously beyond that. Indeed, it is likely that the activation mode will depend on catalyst features (metal, size, support) and the reaction and, therefore, warrants careful evaluation.