Localized surface plasmons and hot electrons

Chemical Physics 445 (2014) 95–104

Contents lists available at ScienceDirect

Chemical Physics journal homepage: www.elsevier.com/locate/chemphys

ChemPhys Perspective

Localized surface plasmons and hot electrons Kyle Marchuk, Katherine A. Willets ⇑ Department of Chemistry, University of Texas at Austin, Austin, TX 78712, United States Department of Chemistry, Temple University, Philadelphia, PA 19122, United States

a r t i c l e

i n f o

Article history: Available online 28 October 2014 Keywords: Hot electron Localized surface plasmon Photocatalysis Photodiodes Water splitting Super-resolution imaging

a b s t r a c t The ability of plasmonic devices to generate hot electrons has the potential to move chemical manufacturing outdoors by harnessing photon energy and converting it to useful chemical energy. By using localized surface plasmons to generate hot carriers in noble metal nanostructures, visible light can produce energetic electrons (or holes) which drive chemical reactions or create a light-induced photocurrent. Within this Perspective, we look into recent theory of plasmonic hot electron generation and how the underlying nanoparticle structure influences both the number and energy of the hot carriers produced. Applications in photodiodes and photocatalysis are highlighted to demonstrate potential device opportunities for plasmon-generated hot electrons. Super-resolution imaging studies, in which the location of hot carrier production in hybrid plasmonic-semiconductor devices is spatially localized to <10 nm, are also presented. Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction The collective oscillation of surface conduction electrons in metallic nanoparticles and nanostructures induced by the electric field component of light, known as a localized surface plasmon [1], has led to a variety of applications in areas such as biological imaging [2], chemical sensing [3], and enhanced surface-detection techniques including surface-enhanced Raman scattering (SERS) [4] and metal-enhanced fluorescence (MEF) [5]. The localized surface plasmon resonance of these nanoparticles is dependent upon the size, shape, and material that comprise the particles and is tunable across the visible and near-infrared spectrum [6]. Surface plasmon properties have been utilized in many devices and applications including biosensing and nanomedicine [7–15], plasmon rulers [16–20], steam generation [21–23], photovoltaics [24,25], small molecule gas sensing [26–29], single molecule detection [30–32], tip-enhanced spectroscopies [33–36], and waveguiding [37]. Upon excitation, surface plasmons can either decay radiatively as re-emitted photons [38] or nonradiatively, leading to local heating [39,40] and/or energetic electrons [41]. In most applications related to sensing and spectroscopy, the nonradiative decay of plasmons has been considered a hindrance that negatively affects the performance of plasmonic devices by limiting the plasmon ⇑ Corresponding author. Tel.: +1 512 471 6488. E-mail address: [email protected] (K.A. Willets). http://dx.doi.org/10.1016/j.chemphys.2014.10.016 0301-0104/Ó 2014 Elsevier B.V. All rights reserved.

lifetime and thus the strength of the enhanced fields at the nanoparticle surface [42]. Recently it has been shown that energetic electrons, referred to as ‘‘hot’’ electrons, can be useful in a variety of applications including photodetection/photovoltaic devices [24,43], photocatalysis [43,44], and surface imaging [45]. In the most straightforward definition, a hot electron has energy that is above the distribution of available electron energies in the material as described by Fermi–Dirac statistics; put another way, the electron appears to have a higher effective temperature than is expected based on thermal equilibrium, which renders it ‘‘hot’’ [46]. A simplified view of hot electron excitation by a photon with energy  hxp is depicted in Fig. 1. Hot electrons with varying energies relative to the Fermi level EF can be produced in this scheme, with the maximum hot electron energy given by Emax = EF +  hxp. Recent experimental and theoretical work has shown that plasmon excitation can promote enhanced formation of hot electrons in noble metal nanoparticles [47–51]. As such, significant interest in characterizing the relationship between plasmon excitation and hot electron generation in metallic nanoparticles has arisen within the field, as demonstrated by the recent reviews by Clavero [43], Baffou and Quidant [52], and Kale et al. [53], with new device geometries and materials being explored in order to favor hot electron production. Within this Perspective, we will focus on describing the processes, applications, and challenges that arise from harnessing hot electrons generated by the decay of surface plasmons on metal nanoparticles.

96

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104

The picture of hot electron excitation shown in Fig. 1 implies that electron energies can be accessed over a wide range from EF to Emax = EF +  hxp. However, in reality, the actual available energies of hot electrons (and hot holes) that can be generated depend on many parameters, including the size of the nanoparticle and the hot carrier lifetime. To understand the distribution of hot carriers (both electrons and holes) that can be produced by photon excitation, Govorov and coworkers performed calculations in which both bulk plasmon excitation in a gold film and localized plasmon excitation in a 10 nm gold nanocube are compared, as shown in Fig. 2A and B, respectively [50]. In both figures, dashed lines show the distribution of electron energies at thermal equilibrium, as described by Fermi–Dirac statistics, while solid lines illustrate the change in the electron (and hole) populations as a function of energy upon photon excitation. In the case of bulk plasmon excitation (Fig. 2A), momentum conservation limits the energies that can be accessed by the hot carriers, resulting in a cluster of hot electron and hole energies around the Fermi energy of the metal. On the other hand, hot electrons generated by localized surface plasmon excitation can access the full range of energies from EF to Emax. In this case, the electrons can interact with the walls of the nanoparticle, which relaxes the momentum conservation rules, and allows electrons to access higher energies relative to the Fermi level. Based on this argument, the size of the nanoparticle becomes important, with smaller nanoparticles favoring production of more high energy hot electrons, despite the fact that larger nanoparticles will produce more hot electrons (e.g. electrons with energy above EF) overall [50,51]. To illustrate this last point, Fig. 3 shows recent work from Nordlander and coworkers in which the distribution of hot carrier energies in silver nanospheres is plotted as a function of both the lifetime of the hot carrier as well as the nanoparticle size [51]. The lifetime of the hot electron (s) collapses the time scales of the different mechanisms by which the hot electron can decay (e.g. electron–electron, electron-surface, and electron–phonon scattering) into a single value, ranging from 0.05 to 1 ps. For smaller nanospheres (15 nm diameter, Fig. 3A), more high energy hot electrons (relative to EF) are produced, especially for longer carrier

Ó Copyright (2014), Elsevier.

2. How ‘‘hot’’ is hot? Predicting hot carrier energy distributions

Fig. 2. (Solid lines) Plasmon-generated hot carrier distribution excited by (a) bulk plasmon in a thin gold film and (b) localized plasmon in a 10 nm gold nanocube. The dashed curve shows the equilibrium Fermi distribution of electrons at room temperature. Reprinted with permission from Ref. [50].

lifetimes (s = 0.5, 1 ps). On the other hand, the overall number of hot electrons increases as the carrier lifetime gets shorter, albeit with energies clustered closer to the Fermi level. In the case of a 25 nm diameter nanosphere (Fig. 3B), the number of hot electrons is increased relative to the smaller nanosphere for a given carrier lifetime, but there are significantly fewer high energy hot electrons. Both the work from Govorov and coworkers and the work from Nordlander and coworkers highlight an important point when considering plasmon-generated hot electrons: it is not just the number of hot electrons produced but also the energy (the relative ‘‘hotness’’ if you will) that is dictated by the properties of the underlying nanostructure. To address this, Nordlander and coworkers defined a figure of merit which represents the number of hot electrons produced per plasmon with energy above a certain threshold (typically defined relative to the energy of the plasmon excitation,  xp). Using this definition, they showed that while nanoshells proh duce more than twice as many hot electrons as solid nanospheres of comparable size, the figures of merit with energy thresholds set above both 20% and 50% of the original plasmon excitation energy are roughly identical between the two different structures. This balance between number and energy of hot electrons is a critical point when thinking about using plasmon excitation to enhance hot electron production and utility. 3. Hot carrier lifetimes

Fig. 1. Schematic of hot electron and hot hole generation by photons with energy  hxp. Electrons are excited from occupied energy levels to energies above the Fermi level, EF, leaving hot holes behind.

As described above, the non-radiative decay of plasmons gives rise to single-electron excited states, which can undergo a variety of paths to relaxation that, in turn, determine the hot carrier lifetime. The plasmon lifetime is determined by a loss of coherence in the collective electron oscillation, which can be due to electrons

97

Ó Copyright 2014 American Chemical Society.

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104

Fig. 3. Hot carrier distributions in silver nanospheres with diameters of 15 nm (a) and 25 nm (b). The distribution of hot electrons (red lines) and hot holes (blue lines) generated per unit of time and volume as a function of their energy are plotted. Four different hot carrier lifetimes (s) ranging from 0.05 to 1 ps are shown. The frequency of the external illumination is fixed to 3.65 eV, which corresponds to the plasmon frequency and therefore to the maximum absorption. Zero energy refers to the Fermi level. Reprinted with permission from Ref. [51]. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

scattering off the nanoparticle surface, defect sites (both internal and external) or other electrons and is found to be on the order of 5–20 fs [54]. Once coherence is lost, the remaining hot electron–hole pair has its own associated lifetime that is dictated by electron–electron, electron-surface, and electron–phonon scattering processes within the material. These various relaxation pathways each have their own associated timescales: electron– electron scattering, <100 fs; electron–phonon scattering, 1–10 ps, and phonon–phonon (100 ps) interactions [55–57]. Because the hot electron may exist beyond the lifetime of the plasmon, it is possible to use these hot electrons for photoreactions and in other device applications, as described in the next section. To probe the lifetimes of hot carriers, one must capture them before they relax to thermal equilibrium; one way to accomplish this is to create a Schottky barrier between the metal and a semiconductor material, which allows electron transfer to occur if the energy levels are sufficiently matched (Fig. 4). If the energy of a

Fig. 4. Depiction of metal/semiconductor Schottky barrier. Hot electrons with high enough energies can overcome the Schottky barrier (uSB) and be injected into the semiconductor conduction band (Ec). uM is the work function of the metal, XS is the electron affinity of the semiconductor, Eg is the bandgap of the semiconductor, and Ev is the valence band of the semiconductor.

hot electron is above the Schottky barrier, it can be injected into the conduction band of the semiconductor. A properly designed electron capture device has a much smaller barrier energy (uSB) compared to that of the semiconductor bandgap (Eg) [58,59]. This allows the plasmon-based devices to utilize visible and near-infrared light to generate conduction band electrons in the semiconductor, at energies where direct excitation of the band gap does not occur. Post electron-injection, the metal is left positively charged; therefore a hole-transporting material (HTM) or electron-donating solution is needed to be in contact with the metal to keep charge balance and sustain an electrical circuit. Investigations into the timing of hot electron processes are typically accomplished through ultrafast pump/probe experiments. For example, injection of plasmon-generated hot electrons into TiO2 can be time-resolved by tracking transient absorption spectra [60]. Tachiya’s group used ultrafast visible-pump/infrared-probe femtosecond transient absorption spectroscopy to determine that electron injection into the TiO2 was completed within 50 fs, while the charge recombination (e.g. back electron transfer) depends on the size of the TiO2 nanoparticle and happens within 1.5 ns [61,62]. Based upon the photoexcited electron relaxation pathway timescales, the researchers concluded that the electron injection process occurs before or during the thermalization process. As mentioned above, a HTM or an electron-donating solution is needed to balance charge and minimize hot electrons returning to the nanoparticle; thus, it is important to have rapid restoration of charge neutrality before back-electron transfer of the hot electron from the semiconductor to the metal can occur. Investigations into various electron-donating species have been undertaken [63–65]. Comparisons were made among [Fe(CN)6]4 , Fe2+, and ferrocenecarboxylic acid as electron donors to gold particles on TiO2, and it was determined that Fe2+ performed best, regenerating charge neutrality to the nanoparticles within 500 ns and thereby inhibiting a significant fraction of back electron transfer from the TiO2 [64]. The researchers further shortened the timescale over which charge neutrality is restored by entombing the gold–TiO2 particles in a polyethylene oxide matrix containing the I /I3 redox couple. This system achieved 20 ns charge neutralization of the gold, further improving the efficiency of hot electron transfer from the gold to the TiO2 [65]. The timing measurements described above were made using transient absorption spectroscopy requiring femtosecond laser

98

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104

pulses. As an alternative approach, Link and coworkers [66] used the plasmon linewidth to calculate the hot electron transfer time and efficiency from gold nanoparticles into graphene. By comparing the spectral linewidths of Au nanorods on quartz to nanorods on graphene, linewidth broadening was observed on the graphene-supported nanorods relative to their glass-supported counterparts that could not be accounted for by intrinsic bulk and radiation damping. This additional broadening was assigned to electron transfer from the nanorods into the graphene with an electron transfer time of 160 ± 30 fs calculated from the spectral linewidth as determined by a Lorentzian fit to the spectral data. However, in comparison to the ultrafast measurements described above, this strategy cannot discriminate between forward- and back-electron transfer between the gold nanorod and the graphene; nonetheless, it provides a straightforward strategy for determining whether charge transfer processes are occurring between the different materials.

4. Applications of plasmon-generated hot electrons in devices Having discussed the range of available hot carrier energies as well as the lifetime of the hot carriers, we now turn to various devices that have successfully utilized plasmon-generated hot electrons. An important aspect of increasing device efficiencies is to maximize the branching ratio between the nonradiative and radiative decay pathways of the plasmon. The size of the nanoparticles is well known to affect the plasmon resonance wavelength, but will also influence the relaxation pathways. For example, Au and Ag particles 20–40 nm in diameter are known to predominantly relax through radiative processes [67] while nonradiative processes dominate for smaller nanostructures [68]. This affects the electron-harvesting efficiencies as was demonstrated by Tatsuma and coworkers, who loaded various sizes of gold nanoparticles onto a nanoporous TiO2 film and studied the photocurrent produced by this system [69]. In this work, the maximum photocurrent was recorded when using 15 nm gold particles compared to 40 and 100 nm particles, indicating that the smaller nanoparticles led to greater charge production and injection into the semiconductor. In addition to size, the composition of the nanoparticles is important to increasing the nonradiative decay. Kasemo and coworkers compared the extinction and scattering intensities of lithography-fabricated nanodisks and showed that Pt and Pd nanodisks undergo strong nonradiative decay processes throughout the investigated range of disk diameters (38–530 nm), while radiative decay channels dominated in Ag for nanodisks larger than 110 nm [70]. Thus, the traditional plasmonic materials of gold and silver, which have been studied extensively for sensing and spectroscopy applications, may find competition from other metals for hot electron applications. An additional challenge with understanding the role that hot electrons play in realistic devices is that there are often multiple decay mechanisms at play, and the importance of hot electron production relative to other mechanisms is not always clear. For example, when plasmonic nanoparticles are integrated into photovoltaic devices, the subsequent performance improvement is attributed to locally-enhanced electromagnetic fields (which increase the absorbance of the cell) and light scattering effects (which increase the effective path length–or ‘optical thickness’–of the cell) [24]. Direct collection of plasmon-generated hot electrons is most likely a low efficiency process in these types of devices. Moreover, the different decay channels for plasmons—particularly, the competition between hot electrons and local heating—must be carefully considered. For example, Halas and coworkers showed that double stranded DNA attached to gold nanoparticles would dissociate under light excitation, but with different mechanisms

based on particle shape: for gold nanoshells, the dissociation appeared to be hot-electron mediated, while for gold nanorods, thermal release was the dominant mechanism [71]. Thus, when designing devices that exploit plasmon-generated hot electrons, it is important to be aware of other mechanisms that could compete both favorably and unfavorably with the collection and use of the hot electrons. To investigate the different mechanisms by which plasmons can assist in photocatalysis, Tsai and coworkers used X-ray absorption spectroscopy to study the effect of gold nanoparticles deposited on ZnO nanorods [72]. They found that plasmon excitation generates hot electrons which can be injected into the conduction band of the ZnO for use in water splitting, similar to Fig. 4. However, their Xray data also revealed that plasmon-enhanced electromagnetic fields at the gold nanoparticle surfaces created vacancies in the conduction band of the ZnO near the gold/ZnO interface, which promotes transport of the plasmon-excited hot electrons from the gold into the semiconductor [72]. Thus, the plasmon excitation has two net positive effects: (1) producing hot electrons for injection into the semiconductor and (2) promoting electron transfer to the ZnO through perturbations in its electronic structure due to local electromagnetic field enhancements. This observation highlights one of the truly unique benefits of plasmon excitation, in that different decay pathways can offer synergistic improvements in device performance and efficiency. 4.1. Photodiodes Perhaps the most straightforward device for utilizing plasmongenerated hot electrons is a photodiode, which has the primary function of converting light into current. Park and coworkers constructed a planar diode using gold island films deposited on top of TiO2, which were then heated to induce a structural rearrangement of the gold and the emergence of a localized surface plasmon resonance [73]. They found that devices using the plasmonic gold island films showed an increased photon-to-electron conversion efficiency when the device was illuminated at wavelengths near the plasmon resonance, while devices using planar thin gold films showed no plasmonic response and no efficiency improvement. Thus, plasmon excitation within the gold island film created more hot carriers relative to a thin gold film, leading to improved device performance. To improve the wavelength specificity and polarization response of plasmonic photodiodes, Halas and coworkers fabricated gold nanorod optical antennas on n-type Si (Fig. 5A and B) [58,59]. By using optical antennas, a polarization-dependent device was created that could be tuned to a specific wavelength range in the NIR by simply changing the nanorod geometry. Fig. 5C shows the photocurrent response as a function of wavelength for devices using gold nanorods with differing length. The spectral response of the photocurrent shifts to the red, tracking the shift in the plasmon resonance as the length of the nanoantenna increases. By using rationally-designed plasmonic components, photodiodes can be designed to have highly wavelength- and polarization-specific responses, simply by adjusting the architecture of the device. Many hot electron capturing devices have a physical geometry where the nanoparticle is in planar contact with the semiconductor; this typically means that the momentum of the hot carriers is directed parallel to the semiconductor instead of toward the interface. One way to increase the quantum efficiency of extracting the hot electrons is to embed the plasmonic nanostructures within the semiconductor [74,75]. Halas and coworkers embedded gold nanowire belts into n-type Si and found that even a relatively minor implantation of 5 nm of the gold into the Si could produce increased device responsivities [75]. Embedding the plasmonic nanoparticles in the semiconductor matrix also allows for

99

for the Advancement of Science.

Ó Copyright (2011) The American Association

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104

Fig. 5. An optical antenna-diode for photodetection. (A) Representation of a single Au resonant antenna on an n-type silicon substrate (B) Scanning electron micrograph of a representative device array prior to ITO coating. Imaged at a 65° tilt. (C) Experimental photocurrent responsivity for nine different antenna lengths (points). Solid lines: Fit of the Fowler response to the data. Reprinted with permission from Ref. [58].

additional layering enhancements, as shown by Moskovits and coworkers who embedded layers of gold nanospheres in TiO2 and measured photocurrents that tracked the plasmon resonance of the embedded gold [74]. Other hot electron device configurations have also been investigated to increase efficiency such as tandem devices with double Schottky barriers (TiO2/Au/Si and TiO2/Au/TiO2) [76], and a metal–insulator-metal device that improves photocurrent through the generation of hot electrons that can inject or tunnel through the insulator [77]. In addition, graphene has also been explored as an electron carrier [78,79]. Since graphene has no bandgap, researchers were able to directly inject hot electrons generated in the plasmonic nanoparticles to the conduction band of the graphene sheet. As our understanding of the relationships between plasmon wavelength, nanoparticle geometry, and available hot electron energies continues to improve, we expect new devices with even better hot carrier generation and collection efficiency to continue to develop. 4.2. Photocatalysis Harvesting solar energy and converting it into accessible and storable forms has been a long-standing challenge in alternative energy production. Many applications in photocatalysis have focused on the generation of hydrogen, as it allows for a convenient form of energy storage and can be attained through the electrolysis of water. Currently, many designs for water electrolysis depend upon the electron–hole pair production in semiconductors, which suffer from high rates of charge-carrier recombination and have bandgaps that are not accessible to the bulk of solar photons. Additionally, many broadband semiconductor absorbers such as Si, CdS, CdSe, etc., tend to photocorrode through the water oxidation half-reaction that produces O2 [80–85]. To circumvent this corrosion, only a few materials such as TiO2, WO3, and CeO2 can be used as photoanodes, which have limited photoconversion efficiency due to their large bandgap [86–89]. As a result, plasmonic nanoparticles have become attractive as materials additives to promote the electrolysis of water. One of the best-studied materials for plasmon-mediated photocatalysis is the Au/TiO2 system, particularly for the hydrogen reduction reaction [90–94]. While TiO2 is catalytic for many reactions on its own, ultraviolet excitation is required to excite electrons from the valance band into the conduction band, limiting

the utility of this material for photocatalysis [95–98]. However, by incorporating gold nanoparticles, visible photons create plasmon-generated hot electrons in the gold, which are then transferred into the conduction band of the TiO2, similar to Fig. 4; this allows lower energy visible light to drive catalytic reactions, such as hydrogen reduction, on the surface of the TiO2. Even within this simple framework, many variables can impact the efficiency of hydrogen evolution on Au/TiO2 composites, including nanoparticle size, percentage of nanoparticle loading within the TiO2 matrix, calcination temperature, and pH [90]. Recent work by Wei and coworkers suggests that the nanoparticle size dependence of hydrogen evolution efficiency within the Au–TiO2 system is due to a change in the reduction potential of the transferred electrons, with larger nanoparticles producing hot electrons that are better able to overcome the overpotential barrier for hydrogen reduction [91]. Although this seems counter to the theoretical results above which suggest that smaller nanoparticles should produce higher energy hot electrons [48–51], the authors explain their results by suggesting that the larger nanoparticles produce more hot electrons, which leads to overfilling of the TiO2 conduction band, and allows electrons to access higher energies and more facile transfer to nearby H+. The influence of locally-enhanced electromagnetic fields, which are typically more strongly enhanced on larger nanoparticles, may also impact the conduction band of the TiO2, similar to the results of Tsai described above [91]. Of the two half-reactions involved in the splitting of water, hydrogen reduction is considered simpler due to protons only needing to accept electrons at the electrode surface. Water oxidation to produce molecular oxygen, on the other hand, is mechanistically more challenging as it requires four positive holes and the formation of a new O–O bond [99,100]. To overcome this rate limiting step, an oxygen evolution catalyst (OEC) can be incorporated into the device. For example, Moskovits and coworkers created a gold nanorod array capped with TiO2, which then had a cobalt/ borate OEC electrochemically deposited on the particles [101]. Including this OEC produced significantly more O2 compared to the Au/TiO2 alone. Importantly, the photocurrent response tracked the plasmon resonance of the Au nanorods, indicating that plasmon excitation remained an important contributor to the overall increase in efficiency of this multi-component catalytic system. Shortly thereafter, the authors followed up with a standalone device that used Pt nanoparticles attached to the TiO2 layer as the reduction catalyst (Fig. 6) [102]. In this geometry, the

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104

Ó Copyright (2013) Nature Publishing Group.

100

Fig. 6. (a) Schematic of an autonomous water splitting device showing the inner gold nanorod where hot carriers are produced, the TiO2 cap decorated with platinum nanoparticles, which functions as the hydrogen evolution catalyst, and the Co-OEC material deposited on the lower portion of the gold nanorod, which functions as the oxygen evolution catalyst. (b) Corresponding transmission electron micrograph (left) and magnified views of the platinum/TiO2 cap (top right) and the Co-OEC (bottom right). (c) Energy level diagram superimposed on a schematic of an individual unit of the plasmonic solar water splitter, showing the proposed processes occurring in its various parts and in energy space. CB, conduction band; VB, valence band; EF, Fermi energy. Reprinted with permission from Ref. [102].

plasmonic nanostructure can behave as both a photoanode and a photocathode. As with the water oxidation results above, the reduction of hydrogen also shows an enhanced photocurrent under visible excitation that tracks with the plasmon resonance spectrum of the gold nanorod supports, indicating that plasmon excitation increases photocatalytic efficiency for both half-reactions involved in the electrolysis of water. Incorporating other optically important materials such as photonic crystals and quantum dots can also help improve the overall efficiency of devices. For example, a new design consisting of gold nanoparticle-decorated TiO2 photonic crystals generated nearly twice as much current from water oxidation compared to gold nanoparticle-decorated TiO2 nanorods under visible excitation; without the gold nanoparticles, no current was observed for illumination with wavelengths above the band gap of the TiO2 [103]. The authors ascribed the improved performance with the photonic crystal to the ‘‘slow light’’ effect, which allowed the incident light to effectively be tuned into resonance with the plasmon of the nearby gold nanoparticles, creating more plasmon-generated hot electrons (and holes) to be used for water oxidation. More recently, a new device consisting of TiO2 nanorod arrays coated with CdS quantum dots and gold nanoparticles has enhanced the photonto-electron conversion efficiency over an extended wavelength range between 325 and 725 nm [104]. Below 525 nm (the plasmon resonance cutoff) the gold nanoparticles acted as an electrical connection between the quantum dots and the TiO2, allowing charge to migrate from the photo-excited quantum dots to the conduction band of the TiO2. At longer wavelengths where surface plasmons are resonant, the gold nanoparticles directly inject plasmon-generated hot electrons into the TiO2. These studies highlight the ability to design multi-component devices, which achieve enhanced photoconversion efficiency by exploiting multiple mechanisms beyond plasmon excitation. Although water splitting is the most common method for hydrogen production, hot electron-mediated production of hydrogen can also be formed from organic molecules such as ethanol [105] and isopropanol [106]. Again, it was shown that many factors influence the rate of hydrogen evolution using these plasmonicsemiconductor hybrid materials. For gold on TiO2, the size of the gold nanoparticle along with the size and nature (anatase vs. rutile) of the TiO2 support can influence the hydrogen evolution rate from ethanol [105], while Au/TiO2 Janus particles exhibit higher rates of hydrogen production from isopropanol compared to typical Au core/TiO2 shell nanoparticles [106].

In addition to hydrogen generation, plasmonic nanoparticles have been used to photocatalyze a variety of reactions such as dissociation of H2 [107], photodecomposition of methylene blue [108], methyl orange [109], and sulforhodamine-B [110], oxidization of CO [111] and ethylene [112], conversion of CO2 to hydrocarbons [113], reduction of nitroaromatic compounds [114] and many others [115–120]. In all cases the rate of reaction increased when illuminated with wavelengths that excite the surface plasmon resonance. Most systems utilize gold nanoparticles as the plasmonic material, which absorbs significantly in the blue/UV region of the spectrum, causing electrons to undergo an interband transition from 5d to 6sp [1,121,122]. This joint effect can be used in tandem with surface plasmon hot electron generation to photocatalytically drive chemical reactions such as the reduction of a variety of nitroaromatic compounds on ZrO2 supported gold nanoparticles [114]. Many of the plasmon-supporting substrates useful for photocatalysis are also excellent SERS substrates, allowing reactions to be followed in real time through changes in the SERS spectrum. In a previous review by Sun and Xu [120], the authors discuss plasmon-mediated surface-catalyzed reactions revealed by SERS spectroscopy. More recently, the reduction of Fe3+ to Fe2+ on cyanoterminated Ag nanoparticles was monitored through SERS spectroscopy and was determined to be facilitated through plasmonmediated hot electrons [123]. Under 632.8 nm illumination, the authors see a discernable difference between the frequency of the –CN stretch when either Fe3+ to Fe2+ is interacting with the pendant nitrogen of the cyano group. Under 514.5 nm excitation, the frequency of the –CN stretch shifts over time in the presence of Fe3+, corresponding to the reduction of Fe3+ to Fe2+ due to plasmon-excited hot electrons. A corresponding shift is not observed when the Fe2+ is introduced, indicating that hot holes cannot drive the reverse oxidation reaction. The authors rule out local heating effects by running the reaction at 77 K and show that the kinetics of the reduction reaction are strongly wavelength-dependent, indicating that hot electrons are, in fact, the dominant mechanism for the photoreduction of the Fe3+. 5. Mapping hot carriers on single plasmonic nanoparticles: super-resolution imaging One interesting outcome of the calculations from Nordlander and coworkers described previously is that they predicted that hot electrons would be spatially localized on the surface of

101

Ó Copyright (2013) American Chemical Society.

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104

Fig. 7. (a) TEM images of Au-tipped CdS nanorod heterostructures. (b) Schematic illustrating two distinct photocatalysis mechanisms with the opposite direction of hot carrier transfer based on excitation wavelength. In Mechanism A, hot electrons are produced in the gold tip and transferred to the CdS, while in Mechanism B, hot electrons are produced in the CdS and transferred to the Au tips. (c-d) Super-resolution mapping of single Resorufin turnover events on a Au-tipped CdS nanorod heterostructure under (c) 532 nm excitation (mechanism A) and (d) 405 and 532 nm excitation (mechanism B). Each point shows where a single turnover event occurs and the local charge environment (positive, excess holes, red data; or negative, excess electrons, blue data). In (d) the 532 nm laser is needed to excite fluorescence from the resorufin product, while the 405 nm light drives the charge transfer (mechanism B). Reprinted with permission from Ref. [137]. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

individual nanostructures [51]. However, most hot electron studies are performed at the bulk array level, which reports an ensemble average and is not sensitive to either inter- or intra-particle heterogeneities. To study particle-to-particle heterogeneity, single nanoparticle studies are required, and reports can be found in the literature in which electrochemical detection [124], surface plasmon spectroscopy [28,125,126], and single-molecule fluorescence microscopy [127–133] are all used to study catalytic processes at the single particle level. However, to understand the spatial localization of hot electrons in single-particle catalysis studies, super-resolution/super-localization microscopy techniques are required. In super-resolution imaging, diffractionlimited images of single emitters are fit to a model function, such as a 2-dimensional Gaussian, to calculate the location of the peak emission (or emission centroid), with localization precision to a few nanometers [133–136]. To date, super-resolution imaging of hot electrons requires indirect detection, based upon a change in an optical signature induced on a local reporter molecule upon interacting with a hot electron. For example, Majima and coworkers used a redox responsive Bodipy-based probe, which turns from non-fluorescent to fluorescent upon reduction, to compare reactions on TiO2 nanoparticles and gold nanoparticle-decorated TiO2 nanoparticles [136]. Individual Bodipy reduction events were recorded as a burst of diffraction-limited emission, which could be spatially localized by fitting the emission to a 2-dimensional Gaussian function. The sites of the localized emission could then be compared to the shape of the underlying nanoparticles using scanning electron microscopy (SEM). Using this approach, they found that single-molecule

turnover events occurred indiscriminately across the TiO2 only nanoparticles, while on the Au-decorated TiO2 nanoparticles, the turnover events were confined to distinct regions of the nanoparticle surface. Using correlated SEM, the authors observed that the turnover events appeared to happen within tens of nanometers of the TiO2/gold interface, although the resolution of the SEM prevented the Au nanoparticles from being well-resolved. Nonetheless, the clear difference in local Bodipy reactivity between the bare TiO2 and Au-decorated TiO2 nanoparticles suggest a mechanism mediated by the gold interacting with the semiconductor. The authors speculate that either hot electrons from the gold are transferred into the TiO2 and used to reduce the Bodipy dye or locally-enhanced electromagnetic fields interact with defects in the TiO2 and change the local band structure, allowing direct excitation of the TiO2 bandgap. Fang and coworkers studied gold/semiconductor electron transfer mechanisms through the super-resolution mapping of fluorescent turnover events on CdS nanowires capped with gold nanoparticles (Fig. 7A) [137]. Previous transient absorption spectroscopy studies by Lian and coworkers had shown wavelength-dependent charge transfer between CdS nanowires and gold nanoparticle tips: when the CdS was directly excited by blue light, the electron in the conduction band could be transferred to the gold nanoparticle, as shown as Mechanism B in Fig. 7B [138]. On the other hand, at wavelengths resonant with the gold nanoparticle plasmon, a hot electron is created, which then transfers to the conduction band of the semiconductor, as shown in Mechanism A in Fig. 7B. To study these electron transfer reactions at the single nanoparticle level, Fang and coworkers used a non-fluorescent

102

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104

Amplex Red reporter molecule which can be oxidized to fluorescent Resorufin product in the presence of hydrogen peroxide using the CdS-Au heterostructure (Fig. 7A) as a photocatalyst [137]. Upon Resorufin production, the fluorescence ‘‘on’’ time (or time the product spends adsorbed to the catalyst) depends on the local nanoenvironment of the catalyst: when excess electrons are present, the Resorufin will quickly desorb, while when excess holes are present, the fluorescence ‘‘on’’ time increases dramatically. Using these trends and localizing the site of Resorufin emission through the 2-dimensional Gaussian fitting procedure, the authors can spatially distinguish both the site of Amplex Red oxidation and the local charge environment. When 532 nm excitation is used, Mechanism A should dominate, leading to excess hot holes localized on the gold nanoparticle and hot electron transfer to the CdS nanowire. The resulting superresolution image of Resorufin reduction events under these conditions (Fig. 7C) shows this exact trend: most events are localized towards the ends of the heterostructure, but the Au tips show a more positive nanoenvironment, indicative of excess holes, while more negative environments associated with hot electrons are located closer to the gold/CdS interface. Under 405 nm illumination, the situation reverses to Mechanism B, where the CdS transfers electrons to the Au, leaving excess holes on the CdS; this trend is also observed in the super-resolution images where more negative nanoenvironments are found on the Au tips and more positive nanoenvironments along the length of the CdS wire (Fig. 7D). These studies not only verify the bulk transient absorption results of Lian and coworkers, but also reveal the spatial dependence of hot carrier production, which is highly localized in the case of plasmon excitation (Fig. 7C).

6. Looking towards the future While plasmonic hot electron devices have achieved much in a relatively short amount of time, they still need to become more efficient in converting light to hot carriers, more tunable to specific electron potentials, and need to be made out of more earth abundant materials to become widespread. Paramount to achieving these goals is the development of new materials, given the expense of Au and Ag. Alloys [139] and materials such as transition-metal nitrides [140,141] are currently being investigated to expand the capabilities of plasmonic devices. In addition, the carrier concentration of conducting oxides such as aluminum-doped zinc oxide (AZO) and indium-tin oxide (ITO) can be tuned by changing the doping concentration [142]. Efforts are being undertaken to understand the effect dopant distribution within ITO nanocrystals have on the optoelectronic properties [143]. One challenge with semiconductor plasmonics is that the resonances are all in the infrared. Thus, the resulting hot electrons will be lower energy than their gold or silver counterparts, where plasmons can be easily tuned into the visible spectrum of light. While this energy restriction is not a problem for photodiodes, where new materials for infrared sensing and detection would be welcome, it does present a problem for catalysis, where having large amounts of energy to drive reactions of interest is critical for success. Infrared absorption also limits the utility of these materials for solar applications, given the steep drop-off in intensity at longer wavelengths. However, the semiconductor plasmonics field is still in its relative infancy, so as new materials are discovered and fabrication/synthesis strategies are refined, the utility of this new class of plasmonic materials should continue to grow. While the use of noble metals and rare earth elements is acceptable in small-scale applications, the use of these materials is impractical for large-scale products such as commercial photovoltaics and photocatalysts. Aluminum is coming back into the light

as a potential hot-electron generating plasmonic material through advances in particle formation. Not only is aluminum the third most abundant element in the earth’s crust, its d-band lies above the Fermi energy allowing plasmon resonances to be extended from the UV throughout the visible. Plasmon resonances have been demonstrated for a variety of particle geometries including spheres [144–146], rods [147,148], discs [149–151], and triangles [152,153], and aluminum nanoparticles have already been integrated with a semiconductor to increase energy capture [154]. The general non-use of aluminum nanoparticles as plasmon hosts has been attributed to the lack of reproducibility in particle formation. While some studies show qualitative agreement between experiment and theory for aluminum nanoparticles [149,155], others have shown a wide discrepancy between predicted and observed results [147,151,152]. Recent work has shown that the plasmon resonance energy of aluminum is determined by the amount of oxide present within the particle, and through careful control in the deposition process, reproducible plasmonic properties can be attained [156]. It is therefore quite hopeful to see the large-scale implementation of plasmon-generated hot electron devices using aluminum nanoparticles in the near future. Currently, a challenge remains in understanding the branching ratios between different plasmon decay mechanisms, and perhaps more importantly, the role each can play in enhancing performance in hot electron devices. As we have described, local electromagnetic field enhancements can facilitate charge transfer between plasmonic nanoparticles and nearby semiconductors, and thus it would be naïve to optimize plasmonic materials to produce the largest possible number of hot electrons at the cost of these other potential benefits. Plasmon-induced local heating, which we have not discussed in detail in this review, is another potential pathway for plasmon decay, which may also have benefits for promoting local reactivity at nanoparticle surfaces. Understanding the role that each of these (potentially) complementary mechanisms play is an important experimental challenge as we move from more empirical to more rational device design. 7. Conclusions The generation of hot electrons in plasmonic devices holds a great deal of promise in the areas of photovoltaics and photocatalysis, particularly due to the ability to tune the plasmon resonance and thus the energies of the hot electrons. Already, energy conversion mechanisms have found many different applications despite being a relatively young field, and there is great potential for reaching even higher energy conversion efficiencies by understanding and optimizing the different plasmon decay mechanisms. We have briefly discussed the mechanisms and timing that are essential to the harvesting of hot electrons to bring to light the necessity of fast hot electron injection before recombination and have shown many examples of increased hot electron generation understanding and improvement in device design. New materials have the potential to extend and tune the spectral range and responsitivities for controlled chemical reactions, while earth abundant materials such as aluminum hold potential to apply hot electron supported plasmonic devices to large-scale commercial applications. Conflict of interest The authors declare no competing financial interest. References [1] K.L. Kelly, E. Coronado, L.L. Zhao, G.C. Schatz, J. Phys. Chem. B 107 (2003) 668. [2] A.S. Stender et al., Chem. Rev. 113 (2013) 2469. [3] S. Zeng, D. Baillargeat, H.-P. Ho, K.-T. Yong, Chem. Soc. Rev. 43 (2014) 3426.

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104 [4] K.A. Willets, R.P. Van Duyne, Annu. Rev. Phys. Chem. 58 (2007) 267. [5] W. Deng, F. Xie, H.T.M.C.M. Baltar, E.M. Goldys, Phys. Chem. Chem. Phys. 15 (2013) 15695–15708. [6] C. Burda, X. Chen, R. Narayanan, M.A. El-Sayed, Chem. Rev. 105 (2005) 1025. [7] S. Lal, S.E. Clare, N.J. Halas, Acc. Chem. Res. 41 (2008) 1842. [8] N.L. Rosi, C.A. Mirkin, Chem. Rev. 105 (2005) 1547. [9] E.C. Dreaden, M.A. Mackey, X. Huang, B. Kang, M.A. El-Sayed, Chem. Soc. Rev. 40 (2011) 3391. [10] E.C. Dreaden, A.M. Alkilany, X. Huang, C.J. Murphy, M.A. El-Sayed, Chem. Soc. Rev. 41 (2012) 2740. [11] C.J. Murphy et al., Acc. Chem. Res. 41 (2008) 1721. [12] P.K. Jain, X. Huang, I.H. El-Sayed, M.A. El-Sayed, Acc. Chem. Res. 41 (2008) 1578. [13] J.N. Anker et al., Nat. Mater. 7 (2008) 442. [14] L.B. Sagle, L.K. Ruvuna, J.A. Ruemmele, R.P. Van Duyne, Nanomedicine 6 (2011) 1447. [15] A.E. Cetin et al., Light Sci. Appl. 3 (2014) e122. [16] C. Sönnichsen, B.M. Reinhard, J. Liphardt, A.P. Alivisatos, Nat. Biotechnol. 23 (2005) 741. [17] B.M. Reinhard, S. Sheikholeslami, A. Mastroianni, A.P. Alivisatos, J. Liphardt, Proc. Natl. Acad. Sci. USA 104 (2007) 2667. [18] P.K. Jain, W. Huang, M.A. El-Sayed, Nano Lett. 7 (2007) 2080. [19] N. Liu, M. Hentschel, T. Weiss, A.P. Alivisatos, H. Giessen, Science 332 (2011) 1407. [20] R.T. Hill et al., ACS Nano 6 (2012) 9237. [21] O. Neumann et al., ACS Nano 7 (2013) 42. [22] Z. Fang et al., Nano Lett. 13 (2013) 1736. [23] O. Neumann et al., Proc. Natl. Acad. Sci. USA 110 (2013) 11677. [24] H.A. Atwater, A. Polman, Nat. Mater. 9 (2010) 205. [25] V.E. Ferry, J.N. Munday, H.A. Atwater, Adv. Mater. 22 (2010) 4794. [26] T. Shegai, P. Johansson, C. Langhammer, M. Käll, Nano Lett. 12 (2012) 2464. [27] A. Tittl et al., Nano Lett. 11 (2011) 4366. [28] N. Liu, M.L. Tang, M. Hentschel, H. Giessen, A.P. Alivisatos, Nat. Mater. 10 (2011) 631. [29] A. Yang et al., ACS Nano 8 (2014) 7639. [30] E.C. Le Ru, P.G. Etchegoin, Annu. Rev. Phys. Chem. 63 (2012) 65–87. [31] K.A. Willets, ChemPhysChem 14 (2013) 3186. [32] M.L. Weber, K.A. Willets, MRS Bull. 37 (2012) 745. [33] M.D. Sonntag, E.A. Pozzi, N. Jiang, M.C. Hersam, R.P. Van Duyne, J. Phys. Chem. Lett. 5 (2014) 3125, http://dx.doi.org/10.1021/jz5015746. [34] B. Pettinger, P. Schambach, C.J. Villagómez, N. Scott, Annu. Rev. Phys. Chem. 63 (2012) 379. [35] B. Pettinger, Mol. Phys. 108 (2010) 2039. [36] S. Berweger, J.M. Atkin, R.L. Olmon, M.B. Raschke, J. Phys. Chem. Lett. 3 (2012) 945. [37] E. Hutter, J.H. Fendler, Adv. Mater. 16 (2004) 1685. [38] C. Sönnichsen et al., Phys. Rev. Lett. 88 (2002) 077402. [39] J.R. Adleman, D.A. Boyd, D.G. Goodwin, D. Psaltis, Nano Lett. 9 (2009) 4417. [40] C.-W. Yen, M.A. El-Sayed, J. Phys. Chem. C 113 (2009) 19585. [41] J. Lehmann et al., Phys. Rev. Lett. 85 (2000) 2921. [42] M.A. El-Sayed, J. Photochem. Photobiol. Chem. 221 (2011) 138. [43] C. Clavero, Nat. Photonics 8 (2014) 95. [44] W. Hou, S.B. Cronin, Adv. Funct. Mater. 23 (2013) 1612. [45] A. Giugni et al., Nat. Nanotechnol. 8 (2013) 845. [46] A.D. Semenov, G.N. Gol’tsman, R. Sobolewski, Supercond. Sci. Technol. 15 (2002) R1. [47] D.B. Ingram, S. Linic, J. Am. Chem. Soc. 133 (2011) 5202. [48] A.O. Govorov, H. Zhang, Y.K. Gun’ko, J. Phys. Chem. C 117 (2013) 16616. [49] H. Zhang, A.O. Govorov, J. Phys. Chem. C 118 (2014) 7606. [50] A.O. Govorov, H. Zhang, H.V. Demir, Y.K. Gun’ko, Nano Today 9 (2014) 85. [51] A. Manjavacas, J.G. Liu, V. Kulkarni, P. Nordlander, ACS Nano 8 (2014) 7630, http://dx.doi.org/10.1021/nn502445f. [52] G. Baffou, R. Quidant, Chem. Soc. Rev. 43 (2014) 3898. [53] M.J. Kale, T. Avanesian, P. Christopher, ACS Catal. 4 (2014) 116. [54] J.-Y. Bigot, V. Halté, J.-C. Merle, A. Daunois, Chem. Phys. 251 (2000) 181. [55] S. Link, M.A. El-Sayed, J. Phys. Chem. B 103 (1999) 8410. [56] C. Bauer, J.-P. Abid, D. Fermin, H.H. Girault, J. Chem. Phys. 120 (2004) 9302. [57] S. Link, M.A. El-Sayed, Int. Rev. Phys. Chem. 19 (2000) 409. [58] M.W. Knight, H. Sobhani, P. Nordlander, N.J. Halas, Science 332 (2011) 702. [59] M. Moskovits, Science 332 (2011) 676. [60] J.B. Asbury, E. Hao, Y. Wang, H.N. Ghosh, T. Lian, J. Phys. Chem. B 105 (2001) 4545. [61] A. Furube, L. Du, K. Hara, R. Katoh, M. Tachiya, J. Am. Chem. Soc. 129 (2007) 14852. [62] L. Du et al., J. Phys. Chem. C 113 (2009) 6454. [63] Y. Tian, T. Tatsuma, J. Am. Chem. Soc. 127 (2005) 7632. [64] Y. Tian, X. Wang, D. Zhang, X. Shi, S. Wang, J. Photochem. Photobiol. Chem. 199 (2008) 224. [65] Y. Tian, X. Shi, C. Lu, X. Wang, S. Wang, Electrochem. Commun. 11 (2009) 1603. [66] A. Hoggard et al., ACS Nano 7 (2013) 11209. [67] U. Kreibig, M. Vollmer, Optical Properties of Metal Clusters, Springer, 1995. [68] S. Linic, P. Christopher, D.B. Ingram, Nat. Mater. 10 (2011) 911. [69] K. Yu, Y. Tian, T. Tatsuma, Phys. Chem. Chem. Phys. 8 (2006) 5417. [70] C. Langhammer, Z. Yuan, I. Zoric´, B. Kasemo, Nano Lett. 6 (2006) 833. [71] R. Huschka et al., J. Am. Chem. Soc. 133 (2011) 12247.

103

[72] H.M. Chen et al., ACS Nano 6 (2012) 7362. [73] Y.K. Lee et al., Nano Lett. 11 (2011) 4251. [74] S. Mubeen, G. Hernandez-Sosa, D. Moses, J. Lee, M. Moskovits, Nano Lett. 11 (2011) 5548. [75] M.W. Knight et al., Nano Lett. 13 (2013) 1687. [76] Y.K. Lee, H. Lee, J.Y. Park, Sci. Rep. 4 (2014). [77] F. Wang, N.A. Melosh, Nano Lett. 11 (2011) 5426. [78] Z. Fang et al., Nano Lett. 12 (2012) 3808. [79] Z. Fang et al., ACS Nano 6 (2012) 10222. [80] H. Flaisher, R. Tenne, J. Phys. Chem. 87 (1983) 3061. [81] S. Cattarin, P. Guerriero, N. Dietz, H.J. Lewerenz, Electrochim. Acta 40 (1995) 1041. [82] C. Bhattacharya, J. Datta, Mater. Chem. Phys. 89 (2005) 170. [83] S. Menezes, F.J. DiSalvo, B. Miller, J. Electrochem. Soc. 127 (1980) 1751. [84] Z. Hens, W.P. Gomes, J. Phys. Chem. B 104 (2000) 7725. [85] M.G. Walter et al., Chem. Rev. 110 (2010) 6446. [86] K.L. Hardee, A.J. Bard, J. Electrochem. Soc. 124 (1977) 215. [87] G.K. Mor, K. Shankar, M. Paulose, O.K. Varghese, C.A. Grimes, Nano Lett. 5 (2005) 191. [88] J. Nowotny, T. Bak, M.K. Nowotny, L.R. Sheppard, Int. J. Hydrogen Energy 32 (2007) 2609. [89] B.D. Alexander, P.J. Kulesza, I. Rutkowska, R. Solarska, J. Augustynski, J. Mater. Chem. 18 (2008) 2298. [90] C. Gomes Silva, R. Juárez, T. Marino, R. Molinari, H. García, J. Am. Chem. Soc. 133 (2011) 595–602. [91] K. Qian et al., J. Am. Chem. Soc. 136 (2014) 9842. [92] Z. Liu, W. Hou, P. Pavaskar, M. Aykol, S.B. Cronin, Nano Lett. 11 (2011) 1111. [93] Y.-C. Pu et al., Nano Lett. 13 (2013) 3817. [94] Z. Zhang, L. Zhang, M.N. Hedhili, H. Zhang, P. Wang, Nano Lett. 13 (2013) 14. [95] O. Carp, C.L. Huisman, A. Reller, Prog. Solid State Chem. 32 (2004) 33. [96] M.R. Hoffmann, S.T. Martin, W. Choi, D.W. Bahnemann, Chem. Rev. 95 (1995) 69. [97] O. Legrini, E. Oliveros, A.M. Braun, Chem. Rev. 93 (1993) 671. [98] A. Mills, S. Le Hunte, J. Photochem. Photobiol. Chem. 108 (1997) 1. [99] R. Nakamura, T. Okamura, N. Ohashi, A. Imanishi, Y. Nakato, J. Am. Chem. Soc. 127 (2005) 12975. [100] C.G. Silva, Y. Bouizi, V. Fornés, H. García, J. Am. Chem. Soc. 131 (2009) 13833. [101] J. Lee, S. Mubeen, X. Ji, G.D. Stucky, M. Moskovits, Nano Lett. 12 (2012) 5014. [102] S. Mubeen et al., Nat. Nanotechnol. 8 (2013) 247. [103] X. Zhang, Y. Liu, S.-T. Lee, S. Yang, Z. Kang, Energy Environ. Sci. 7 (2014) 1409. [104] J. Li et al., J. Am. Chem. Soc. 136 (2014) 8438. [105] M. Murdoch et al., Nat. Chem. 3 (2011) 489. [106] Z.W. Seh et al., Adv. Mater. 24 (2012) 2310. [107] S. Mukherjee et al., Nano Lett. 13 (2013) 240. [108] P. Christopher, D.B. Ingram, S. Linic, J. Phys. Chem. C 114 (2010) 9173. [109] J. Yu, G. Dai, B. Huang, J. Phys. Chem. C 113 (2009) 16394. [110] H. Zhu et al., Chem. Commun. 7524–7526 (2009), http://dx.doi.org/10.1039/ B917052A. [111] W.H. Hung, M. Aykol, D. Valley, W. Hou, S.B. Cronin, Nano Lett. 10 (2010) 1314. [112] P. Christopher, H. Xin, S. Linic, Nat. Chem. 3 (2011) 467. [113] W. Hou et al., ACS Catal. 1 (2011) 929. [114] H. Zhu, X. Ke, X. Yang, S. Sarina, H. Liu, Angew. Chem. Int. Ed. 49 (2010) 9657. [115] K. Kim, K.L. Kim, K.S. Shin, Langmuir 29 (2013) 183. [116] S. Navalon, M. de Miguel, R. Martin, M. Alvaro, H. Garcia, J. Am. Chem. Soc. 133 (2011) 2218. [117] X. Chen, H.-Y. Zhu, J.-C. Zhao, Z.-F. Zheng, X.-P. Gao, Angew. Chem. Int. Ed. 47 (2008) 5353. [118] X. Chen et al., Green Chem. 12 (2010) 414. [119] Z. Zheng et al., J. Mater. Chem. 21 (2011) 9079. [120] M. Sun, H. Xu, Small 8 (2012) 2777. [121] S. Link, C. Burda, Z.L. Wang, M.A. El-Sayed, J. Chem. Phys. 111 (1999) 1255. [122] B. Balamurugan, T. Maruyama, Appl. Phys. Lett. 87 (2005) 143105. [123] K. Kim, S.H. Lee, J.-Y. Choi, K.S. Shin, J. Phys. Chem. C 118 (2014) 3359. [124] X. Xiao, S. Pan, J.S. Jang, F.-R.F. Fan, A.J. Bard, J. Phys. Chem. C 113 (2009) 14978. [125] C. Novo, A.M. Funston, P. Mulvaney, Nat. Nanotechnol. 3 (2008) 598. [126] M.L. Tang, N. Liu, J.A. Dionne, A.P. Alivisatos, J. Am. Chem. Soc. 133 (2011) 13220. [127] G.D. Cremer, B.F. Sels, D.E.D. Vos, J. Hofkens, M.B.J. Roeffaers, Chem. Soc. Rev. 39 (2010) 4703. [128] P. Chen et al., Chem. Soc. Rev. 39 (2010) 4560. [129] T. Tachikawa, T. Majima, Chem. Soc. Rev. 39 (2010) 4802. [130] W. Xu, J.S. Kong, Y.-T.E. Yeh, P. Chen, Nat. Mater. 7 (2008) 992. [131] W. Xu, J.S. Kong, P. Chen, Phys. Chem. Chem. Phys. 11 (2009) 2767. [132] W. Xu, P.K. Jain, B.J. Beberwyck, A.P. Alivisatos, J. Am. Chem. Soc. 134 (2012) 3946. [133] X. Zhou et al., Nat. Nanotechnol. 7 (2012) 237. [134] M.B.J. Roeffaers et al., Angew. Chem. Int. Ed. 48 (2009) 9285. [135] G. De Cremer et al., Angew. Chem. Int. Ed. 49 (2010) 908. [136] T. Tachikawa, T. Yonezawa, T. Majima, ACS Nano 7 (2013) 263. [137] J.W. Ha et al., J. Am. Chem. Soc. 136 (2014) 1398. [138] K. Wu, W.E. Rodríguez-Córdoba, Y. Yang, T. Lian, Nano Lett. 13 (2013) 5255. [139] P.R. West et al., Laser Photonics Rev. 4 (2010) 795–808. [140] G.V. Naik, J. Kim, A. Boltasseva, Opt. Mater. Express 1 (2011) 1090. [141] G.V. Naik et al., Opt. Mater. Express 2 (2012) 478.

104

K. Marchuk, K.A. Willets / Chemical Physics 445 (2014) 95–104

[142] R.J. Mendelsberg et al., J. Phys. Appl. Phys. 44 (2011) 232003. [143] S.D. Lounis, E.L. Runnerstrom, A. Bergerud, D. Nordlund, D.J. Milliron, J. Am. Chem. Soc. 136 (2014) 7110. [144] T. Stöckli, J.-M. Bonard, P.-A. Stadelmann, A. Châtelain, Z. Für Phys. At. Mol. Clust. 40 (1997) 425. [145] M.G. Blaber, M.D. Arnold, N. Harris, M.J. Ford, M.B. Cortie, Phys. B Condens. Matter 394 (2007) 184. [146] T. Andersson et al., J. Chem. Phys. 136 (2012) 204504. [147] Y. Ekinci, H.H. Solak, J.F. Löffler, J. Appl. Phys. 104 (2008) 083107. [148] M. Castro-Lopez, D. Brinks, R. Sapienza, N.F. van Hulst, Nano Lett. 11 (2011) 4674. [149] C. Langhammer, M. Schwind, B. Kasemo, I. Zoric´, Nano Lett. 8 (2008) 1461. [150] S.K. Jha, Z. Ahmed, M. Agio, Y. Ekinci, J.F. Löffler, J. Am. Chem. Soc. 134 (2012) 1966. [151] I. Zoric´, M. Zäch, B. Kasemo, C. Langhammer, ACS Nano 5 (2011) 2535. [152] G.H. Chan, J. Zhao, G.C. Schatz, R.P. Van Duyne, J. Phys. Chem. C 112 (2008) 13958. [153] A. Taguchi, Y. Saito, K. Watanabe, S. Yijian, S. Kawata, Appl. Phys. Lett. 101 (2012) 081110. [154] T.F. Villesen et al., Nanotechnology 23 (2012) 085202. [155] G. Maidecchi et al., ACS Nano 7 (2013) 5834. [156] M.W. Knight et al., ACS Nano 8 (2014) 834.

Prof. Katherine A. Willets is an associate professor in the Department of Chemistry at the University of Texas at Austin, where she began her academic career in 2007. She received her Ph.D. in Chemistry at Stanford University in 2005, working with W.E. Moerner, before moving to the lab of Richard Van Duyne at Northwestern University, where she worked as postdoctoral researcher for two years. Current research interests include using super-resolution optical microscopy to study surface-enhanced Raman scattering hot spots, ligand binding to nanoparticle surfaces, and electrochemical reactions on plasmonic nanoparticle electrodes. The Willets lab is moving to the department of chemistry at Temple University in January 2015.

Dr. Kyle Marchuk received his Ph.D. from Iowa State in 2013, working on three-dimensional orientational and positional single particle tracking using optical microscopy techniques in the lab of Ning Fang. He joined the Willets lab in the summer of 2013 as a postdoctoral researcher, where he has been working on new ways to study electrochemical reactions at the nanoscale using combined electrochemical and optical microscopy approaches.