Optical Absorption Engineering in Stacked Plasmonic Au–SiO2–Pd

Aug 23, 2012 - Here we explore the possibility of efficient and spectrally widely tunable optical absorption engineering based on heterometallic optic...
0 downloads 0 Views 4MB Size
Download PDF
Letter pubs.acs.org/NanoLett

Optical Absorption Engineering in Stacked Plasmonic Au−SiO2−Pd Nanoantennas Carl Wadell,‡ Tomasz J. Antosiewicz,‡ and Christoph Langhammer* Department of Applied Physics, Chalmers University of Technology, SE-412 96 Göteborg, Sweden S Supporting Information *

ABSTRACT: The nonradiative decay of a localized surface plasmon through absorption of a captured photon and excitation of an energetic electron−hole pair is a potentially very effective way to enhance chemical reactions on metal nanoparticle surfaces, so far limited to Ag (and Au). Here we explore the possibility of efficient and spectrally widely tunable optical absorption engineering based on heterometallic optical nanoantennas. They consist of an optimized antenna element made of Au (or Ag) and a catalytically active second metallic element separated by a thin SiO2 layer. Specifically, we find that stacked Au−SiO2−Pd nanodisk antennas exhibit pronounced local absorption enhancement in the catalytic Pd particle. The effect is caused by efficient power transfer from the Au disk, exhibiting a narrow low-loss resonance and acting as an antenna collecting photons, to the Pd disk due to strong coupling between the two. The Pd element thus acts as receiver that efficiently dissipates energy into electron−hole pairs owing to efficient coupling to intra and interband transitions. In fact, the energy transfer is found to be so effective that the absorption efficiency at a given wavelength can be enhanced up to 6 to 9 times, and the total absorption integrated over a wide spectral range (400−900 nm) up to 2-fold, depending on the antenna dimensions. This finding suggests a novel route toward highly efficient plasmon-enhanced catalysis on widely selectable catalytic metal particle surfaces not limited to the “classic” plasmonic metals Au and Ag. KEYWORDS: Localized surface plasmon resonance, absorption enhancement, palladium, nanoantenna, plasmon enhanced chemistry

W

created hot/energetic electrons for photodetection14 and for bond-breaking and formation in adsorbates on an adjacent semiconductor15−18 or directly on the plasmonic metal nanoparticle surface.19,20 For the latter application, mechanistically, it is proposed that the LSPR donates hot electrons into empty (LUMO) adsorbate states to create a reactive negative ion species on the particle surface. All experimental efforts in this direction have so far been focusing on Ag as the plasmonic metal due to its catalytic activity for certain reactions. In a broader catalysis context, however, the surfaces of both the “favorite” plasmonic materials, Au and Ag, are rather uninteresting due to their weak interactions with adsorbates.21 As a solution to combine the strong surface plasmon excitations in Au (or Ag) with the intrinsic catalytic activity of other metals, in this Letter we propose a novel and generic strategy of plasmonic absorption engineering by means of heterometallic nanoantennas. We show, experimentally and theoretically by finite-difference time-domain, FDTD, simulations and an analytical coupled mechanical oscillator model, how a Au−SiO2−Pd nanodisk stack (“nanosandwich”, see Figure 1) can be used to engineer and significantly enhance light absorption (and, consequently, electron−hole pair

hen electromagnetic radiation interacts with solid materials it can either be absorbed, ultimately leading to a transformation of the energy carried by photons into heat if not harvested otherwise, or be reflected/scattered back from the surface. At the nanoscale, localized surface plasmon resonance (LSPR) constitutes a very interesting phenomenon invoked when metal nanoparticles interact with light at nearvisible frequencies through collective resonant excitation of its electrons.1 After excitation, LSPRs can decay either radiatively by re-emission of a photon (scattering) or nonradiatively via (hot) electron−hole pair formation (absorption).2−4 Most of the scientific efforts devoted to the utilization of LSPR have so far been focused either on the exploitation of locally strongly enhanced electromagnetic fields for sensing5,6 and enhanced spectroscopy7,8 applications or on the ability of plasmonic structures to serve as a tool for light manipulation at the nanoscale via the scattering channel.9 Hence, the second decay process, light absorption involving hot electron−hole pair formation, has received much less attention. Nevertheless, cancer therapy based on hyperthermia10,11 and temperaturetriggered drug release,12,13 where heat generation associated with light absorption through plasmon excitation in colloidal metal nanoparticles is successfully being utilized, can be put forward as prominent examples. More recently the absorptive LSPR decay channel has come into the spotlight due to the exciting possibilities to utilize the © 2012 American Chemical Society

Received: June 12, 2012 Revised: August 14, 2012 Published: August 23, 2012 4784

dx.doi.org/10.1021/nl3022187 | Nano Lett. 2012, 12, 4784−4790

Nano Letters

Letter

assembly of charged polystyrene nanospheres to produce an evaporation mask. Au, SiO2, and Pd layers are deposited trough the mask using physical vapor deposition. The mask is subsequently removed by lift-off in acetone and left on the sample surface are the sandwich-like nanostructures, arranged in an amorphous array lacking long-range order as shown in Figure S1 in the Supporting Information (SI). As evident from the SEM pictures in Figure 1, the resulting bimetallic nanoantennas have the shape of truncated cones with decreasing top-diameter for increasing total structure height. This particular effect, rendering the top metallic nanodisk smaller than the bottom one, is created by the (HCL-typical) successive decrease in diameter of the holes in the evaporation mask upon material deposition. The latter effect is a result of material deposition onto the rims of the nanoholes in the mask and the subsequent shrinkage of the latter, which yields a corresponding decrease in the diameter of the growing nanoparticle during evaporation. Hence, in a typical stacked bimetallic nanoantenna studied here, the bottom Au nanodisk has constant dimensions dictated (i) by the diameter of the polystyrene nanospheres used and (ii) the thickness of the deposited Au layer. In addition to the preceding two points, the dimensions of the top Pd nanodisk, however, depend on (iii) the thickness of the SiO2 separating layer (decreasing diameter for increasing SiO2 thickness) and (iv) on the thickness of the deposited Pd layer. For all the experiments and calculations presented in the main text, the dimensions of the bottom Au disk are kept constant at a mean diameter DAu = 140 nm and thickness tAu = 20 nm. The thickness of the Pd disk is kept constant at tPd = 20 nm while the thickness of the SiO2 separating layer is varied systematically between 0 and 30 nm. The optical properties of the nanoantennas, that is, their extinction, scattering, and absorption efficiencies (optical cross sections normalized by the surface area of the nanosandwich defined by the larger of the two disks in the stack if not specifically stated otherwise), were measured in a Varian Cary 5000 spectrophotometer. The absorption efficiency was measured using an integrating sphere accessory Varian DRA 2500 with a sample center mount for transflectance measurements. The samples were mounted in the center of the integrating sphere at a 9° angle from the normal to the incident light beam to ensure that the specular scattering contribution is measured. The scattering efficiency was obtained from the extinction and absorption measurements by relying on the optical theorem. Experimental Results and Electrodynamics Simulations. Figure 1 contains a series of measured extinction spectra for Au−SiO2−Pd stacked nanoantennas with increasing SiO2 separating layer thickness ranging from 0 to 30 nm. The spectra for lone Au and Pd disks with diameters of 140 nm are also shown for comparison. Varying the spacer thickness provides, as previously demonstrated for Au−SiO2−Au nanosandwiches,24 a route to tune plasmon coupling by tailoring the separation between the individual disks in the stack. Interestingly, for the heterometallic Au−SiO2−Pd system at scrutiny here, the formation of hybridized bonding (low energy) and antibonding (high energy) modes is not as clearcut, as we will discuss in detail below. At this point, we only briefly discuss the trends seen in the extinction spectra. For 0 nm spacer layer thickness (direct contact between the Au and the Pd disk) the extinction spectrum is dominated by a single, relatively broad peak with a maximum around 690 nm. The peak resembles, both in terms of spectral position and line

Figure 1. Heterometallic nanoantennas composed of stacked Au and Pd nanodisks with a SiO2 spacer layer. The top image shows a schematic illustration of the nanostructures, a “nanosandwich”, considered here. They consist of a bottom Au nanodisk with constant diameter of 140 nm and thickness of 20 nm, a SiO2 spacer layer with thickness varied systematically from 0 to 30 nm, and a top Pd disk with constant thickness of 20 nm and systematically decreasing diameter for increasing SiO2 thickness. The lower left part of the figure shows SEM images of individual nanoantennas, taken at a 70° tilt angle on sample analogues fabricated on a Si wafer for higher conductivity, for SiO2 thicknesses of 0, 15, and 30 nm, respectively. The drawings correspond to schematic illustrations of the structures seen in the SEM images. The lower right part of the figure shows experimentally measured extinction spectra for amorphous arrays of heterometallic stacked nanoantennas with different SiO2 thicknesses, as indicated to the right of the plot. The corresponding extinction spectra for lone Au and Pd nanodisks, both with 140 nm diameter, are also shown for reference.

formation) in catalytically active metals such as Pd. We find absorption as the dominant decay channel even at NIR wavelengths with absorption efficiencies (that is optical cross sections divided by the projected particle area) above 4. Notably, this is measured for individual antenna dimensions exceeding 100 nm for which in homogeneous plasmonic structures scattering is the dominant decay channel.22 Our study thus constitutes a general blueprint for absorption engineering in heterogeneous plasmonic nanostructures featuring different materials combinations with Au or Ag as the receiver and another (lossy) metal as the dissipating (electron− hole pair creating) and catalytically active element. The made observations for the specific example presented here, that is, Pd as the catalytic material, can be directly extended to other catalytically active transition metals (which are all characterized by large imaginary parts of their dielectric functions at nearvisible frequencies) such as Pt, Ni, Ru, Re, Co, or Fe. Sample Fabrication and Optical Characterization. Stacked Au and Pd disks separated by a SiO2 layer, see Figure 1, were fabricated onto borosilicate glass substrates by holemask colloidal lithography, HCL.23 In brief, HCL uses self4785

dx.doi.org/10.1021/nl3022187 | Nano Lett. 2012, 12, 4784−4790

Nano Letters

Letter

width, pretty much a linear combination of the individual Au and Pd disk spectral response. This is in agreement with what can be seen from the near-field plot in Figure S2 in the Supporting Information and the phase profiles in Figure 5, that is, that the resonance basically corresponds to a single dipole. Increasing the spacer layer thickness gives rise to a slight blue shift of the observed peak combined with a significant broadening. For further increasing spacer layer thickness, a very broad second feature appears in the NIR region of the spectrum, resulting in a double peak structure. Furthermore, a gradual blue shift and decrease in the intensity of the originally dominant high-energy peak can be seen. Simultaneously, the broad low-energy resonance moves to shorter wavelengths, sharpens, and eventually dominates the spectrum for spacer layer thicknesses larger than 12.5 nm. For further increasing spacer layer thicknesses, the peak position and line width more and more approach and resemble the resonance observed for an array of lone 20 nm thick Au nanodisks. In a brief qualitative analysis (to be detailed below) of these observations, we find that the high-energy peak, visible for thin spacer layers and seemingly corresponding to the antibonding mode in the Au−Au case,24 carries predominantly a signature of the Pd disk in terms of spectral position and width. The second resonance, corresponding to the bonding mode in the Au−Au case and appearing at lower energy for larger spacer thicknesses, predominantly carries the signature of the Au disk resonance. Figure 2a,c,e shows the experimentally measured extinction, absorption, and scattering efficiencies, respectively, as a function of the Au−Pd disk separation. To determine the optical efficiencies, that is, the optical cross sections normalized by the projected area of the bottom disk of the structure, SEM images of the stacked nanoantenna arrays were used. Also shown in Figure 2b,d,f are the corresponding efficiencies obtained from finite-difference time-domain (FDTD) simulations of the structures. For the simulations, the sandwiches, which include the slanted sides as seen in the SEM images, are placed in an effective medium of ε = 1.44 to take into account the spectral red shift caused by the substrate. The permittivities of the two metals are modeled using a Drude−Lorentz model fitted to experimental data obtained by Johnson and Christy.25,26 For the SiO2 spacer layer, a permittivity equal to 2.1 is assumed. We use a uniform spatial mesh Δr = 1 nm and temporal step Δt = Δr/2c, where c is the speed of light in vacuum. Overall, we find an excellent agreement between the simulated and experimental data, both in terms of absolute efficiencies, and spectral position and width of the maxima. For thin spacer layers, the broad, relatively weak Pd-like highenergy resonance is nicely reproduced in the simulated extinction efficiency contour plot (Figure 2b). For increasing spacer thickness the narrower and stronger Au-like low-energy resonance appears with maximum extinction efficiency around 6 for both experiment and simulation. We now focus our further analysis on the light absorption in our sandwich structures. Looking at the absorption efficiencies in Figure 2c,d reveals a value of almost 4 for the low energy resonance (for another experimental sample series with DAu = 110 nm see Figure S3 in the Supporting Information). This is a surprising result in view of the fact that this resonance is mainly attributed to the Au disk and that individual Au nanodisks (not in a sandwich) of the same size typically have small absorption efficiencies of the order of 1.5 or less (see Figure S4 in the Supporting Information).

Figure 2. Extinction, absorption, and scattering efficiencies of stacked Au and Pd disks, separated by a SiO2 spacer layer plotted as a function of the Au−Pd disk separation, that is SiO2 layer thicknesses. Panels a,c,e are experimentally measured values, whereas panels b,d,f are values obtained from FDTD simulations. Note that the color scale corresponds to different absolute efficiencies for extinction, absorption, and scattering, respectively. Clearly, the plasmonic coupling of constituent disks in a heterometallic nanoantenna does not produce a clear bonding and antibonding mode, rather one single clear resonance peak is observed for most disk separations. Note the large absorption cross-section (compared to the scattering cross-section) at long wavelengths.

As the next step, we try to localize where in the sandwich structure light is actually being absorbed by using the FDTD simulation tool. This approach is well motivated by the found excellent agreement between experiment and simulation. Figure 3a shows the simulated absorption efficiency corresponding to the fraction of light absorbed in the Pd nanodisk in the sandwich as a function of spacer layer thickness (the absorption efficiency corresponding to the fraction of energy absorbed in the Au disk is shown in Figure S4 in the Supporting Information). It is now interesting to compare this to the absorption efficiency of a Pd disk with identical dimensions alone (Figure 3b). Surprisingly, light absorption in the Pd disk in the sandwich at the point of maximal absorption around 775 nm for ca. 20−25 nm SiO2 spacer thickness, is strongly enhanced by a factor of approximately 6, as seen in Figure 3c. Furthermore, as shown in Figures S5 and S6 in the Supporting Information, enhancement factors of up to 9 are predicted for smaller Au antenna diameters. This result, while seemingly an effect of shifting the energy of the resonance, has in fact profound implications for total absorption, as will be shown below. The key point here is the fact that the maximum 4786

dx.doi.org/10.1021/nl3022187 | Nano Lett. 2012, 12, 4784−4790

Nano Letters

Letter

absorption efficiency is increased for a disk separation of ca. 20 nm from about 1 to 2.5 over a rather wide spectral range. To be able to directly compare the efficiencies in Figure 3 panels a and b as done above, we also note that we have used the same projected nanodisk area to calculate the absorption efficiencies for all Pd nanodisk sizes (we recall that the Pd disk in the stack shrinks for increasing spacer layer thickness), namely the area of the Au nanodisk at the bottom of the stack with a diameter of 140 nm. This implies that the absolute absorption efficiency in the smaller Pd disks at larger spacer layer thicknesses actually is even underestimated up to 2-fold for the largest spacer layer thickness. We also note that in the simulations for the lone Pd disks shown in Figure 3b, we have taken the shrinkage into account via a virtual increase of the spacer layer (reflected by the gradual blueshift of the absorption peak). Hence, the absorption efficiency is underestimated in the same way as for the Pd disks in the sandwich and thus does not affect the enhancement factor. As the next step in our analysis we calculate the total (that is, integrated over the whole considered wavelength range from 400 to 900 nm) normalized power absorbed in the sandwich, in the top Pd disk and in the bottom Au disk as a function of spacer layer thickness (Figure 3d). This is done assuming a flat intensity profile for the illumination in the investigated spectral range and normalizing to the power absorbed by a lone Au disk (this value is the only one that remains constant with respect to disk separation). For comparison, the normalized absorbed power by a lone Au (equivalent to unity because of the normalization) and Pd nanodisk is also shown in the figure. As clearly seen, for most disk separations the total integrated absorbed power in the Pd disk is greatly enhanced by the presence of the Au nanodisk. For the particular sandwich dimensions with DAu = 140 nm considered here, the enhancement peaks for separations around 20 nm and the integrated total absorbed power over the considered wavelength range is around 2 times that of a Pd disk alone. Examples for other dimensions (in particular an experimental data series for DAu = 110 nm) are shown in the Supporting Information and illustrate very clearly that the spectral range in which the absorption enhancement occurs can be tuned widely by simply adjusting the dimensions of the stacked heterometallic nanoantenna elements. We now turn to a more detailed analysis of the local distribution of light absorption in the stacked nanoantennas for different spacer layer thicknesses and at different wavelength. In Figure 4, we show the normalized absorption projected on the cross section through a sandwich nanoantenna. Clearly, for all considered wavelengths, absorption is predominantly taking place in the Pd top disk with a maximum, as seen above, for a spacer layer thickness of ca. 20 nm (third column) and in the wavelength range between 750 and 800 nm. For the Pd disk placed directly on the Au one (left column), absorption in Pd is moderate and spread over a large wavelength range, corresponding to a low and relatively broad absorption peak. With an increase in the spacer thickness to 10 nm, absorption increases considerably in amplitude and width in comparison to the first discussed case. Next, for 20 nm of SiO2 absorption blue shifts and becomes maximal, however, the width of the peak remains quite large. Increasing the distance between disks to 30 nm lowers maximum absorption, as well as decreases the line width at the long-wavelength side of the spectrum. Plasmon Coupling in Stacked Heterometallic Nanoantennas. To scrutinize the physics of the observed

Figure 3. Systematic analysis of absorption enhancement in the Pd antenna element. (a) Absorption efficiency of the Pd disk in the sandwich structure as a function of Au−Pd disk separation taken from FDTD simulations. (b) Absorption efficiency of lone Pd disks of the same size as in (a), but without the underlying Au disk, also from FDTD simulations. The physical cross section used to calculate the efficiencies was the same both in (a,b), that is, a disk with diameter 140 nm. (c) Ratio of the absorption cross-section of Pd in the sandwich shown in (a) to the lone Pd disk shown in (b). Note both the large maximum enhancement and the wide wavelength range for which the enhancement is larger than unity. (d) Total power absorbed from an incident beam with a flat power spectrum in the sandwich structure (total sandwich, purple; only in the Pd disk, cyan; and only in the Au disk, green) compared to the total power absorbed in a lone Au disk with diameter 140 nm (blue) and a lone Pd disk (red) with the same diameters as in the sandwich for given SiO2 thicknesses. The data were calculated from FDTD simulations in the 400−900 nm range and normalized to the power absorption in the lone Au disk. Note that the majority of power is dissipated in the Pd disk with a maximum of about 80% for a disk separation of 18 nm and that the Pd disk in the structure absorbs twice more than alone. 4787

dx.doi.org/10.1021/nl3022187 | Nano Lett. 2012, 12, 4784−4790

Nano Letters

Letter

Figure 4. Normalized (with the highest observed value for the investigated system) absorption in the sandwich structure for spacer thicknesses 0 nm (far left column), 10 nm (middle left), 20 nm (middle right), and 30 nm (far right). Light is incident from the top with the electric field in the plane of the figure. The bottom Au disks absorb only a small amount of energy when compared to the Pd disks. Maximum absorption in Pd occurs for long wavelengths and for SiO2 layers thicker than 10 nm and thinner than 30 nm.

Figure 5. Phase profiles in the sandwich structure for spacer thicknesses 0 nm (far left column), 10 nm (middle left), 20 nm (middle right), and 30 nm (far right). For 0 nm SiO2, the sandwich behaves like a dipole with the fields in-phase in both disks. As the disk distance increases a “bonding-like” mode (maximum phase shift close to 2π/3) appears for a broad wavelength range between 650 and 850 nm and follows the maximum Pd absorption seen in Figure 4.

corresponding to a “bonding mode” is thus dominant for long wavelengths, however, it is not clearly peaked and is spread out over a large wavelength range from 650 to 850 nm. Within it the phase shift reaches a maximum of −2 for tSiO2 = 30 nm and occurs approximately when the absorption is maximal. In the short wavelength range, where we do not observe a peak for large spacer thicknesses, the phase distribution is complex. The center of the Au disk is in antiphase with its sides, see for example, 550 nm for tSiO2 = 30 nm. Hence, here the top Pd disk is coupled to the central part of the Au disk, while the sides of the Au disk remain under the influence of the incident field. In summary, despite the lack of sharp bonding and antibonding modes as seen for the homometallic case the dipolar coupling between the two disks is very clear and the fundament for the final step in our analysis. Coupled Oscillator Model. As the last part of our analysis, we show that the considerable absorption increase in the Pd disk can be understood mechanistically by analyzing a simple case of two damped coupled harmonic oscillators with different damping constants. The latter are chosen to mimic the most important properties exhibited by the nanodisks in our stacked optical nanonatenna. One of the mechanical oscillators has a

absorption enhancement in the Pd nanodisks in our stacked nanoantennas, we start out by briefly analyzing the plasmon coupling between the Au and Pd elements in the system. As a key point for the following discussion we note that the plasmonic response of Pd nanodisks is characterized by typically rather broad and predominantly absorptive resonances.22,27 We also note that, as indicated earlier, varying the spacer thickness provides a route to tune plasmon coupling, which in a homometallic sandwich has been reported to yield sharp hybridized modes.24 However, for the heterometallic Au−SiO2−Pd system at scrutiny here we find that formation of hybridized bonding and antibonding modes is not as clear-cut due to the very broad nature of the Pd LSPR. This is illustrated in Figure 5 where phase profiles for heterometallic antennas are shown for different spacer layer thickness. Clearly, neither an obvious bonding (antiphase) nor an antibonding (in-phase) mode is observed. Rather, when the Pd disk is placed directly on top of the Au one, the electric fields are for the most part inphase, and the structure behaves like a single dipole. When the SiO2 spacer thickness increases, the disks begin to exhibit a relative phase shift, which is first observed for short wavelengths and subsequently moves toward longer ones. A feature 4788

dx.doi.org/10.1021/nl3022187 | Nano Lett. 2012, 12, 4784−4790

Nano Letters

Letter

the Au mass and only a small amount into the Pd one. For low frequencies below ca. 100 THz, the Pd oscillator in fact even returns energy (coupled into the Pd mass from the Au) to the external force instead of being driven. The dashed lines show the energy dissipated by the masses. As can be seen in the figure, the Pd mass, despite receiving a very low amount of energy from the driving force, still dissipates a very large amount of energy. As the key result of our analysis we find that at the point of maximum dissipation in Pd (ca. 0.8 at ca. 160 THz), 70% of the dissipated energy is “removed” from the external force by the Au resonator and coupled into the nearby Pd element where it is dissipated. This effect can thus be explained simply by the fact that the damping constant γ2 for the Pd oscillator is much larger compared to γ1 of the Au oscillator. In such a system of coupled oscillators, all the power from the external force is basically transferred into the weakly damped oscillator because the transferred power is inversely proportional to the damping constant. At the same time, however, due to the appropriate coupling between the two oscillators the captured power will be predominantly dissipated by the mass with the larger damping. For our experimentally investigated case of heterometallic optical nanoantennas, this translates into the anticipated picture of the energy from the incident light being efficiently “collected” by the Au nanodisk antenna element featuring low damping and large optical cross section, and coupled to the Pd disk featuring large damping, where it is dissipated into energetic electron−hole pairs and ultimately into heat. Summary and Conclusions. In summary, we have demonstrated experimentally and theoretically, by FDTD simulations and an analytical coupled oscillator model, the possibility of efficient and spectrally widely tunable optical absorption engineering based on stacked heterometallic optical nanoantennas comprising one catalytically active element. We propose the use of these antennas for the enhanced creation and utilization of energetic electrons as mediators for a catalytic reaction of adsorbed species on the surface of the metallic and catalytically active antenna element. In this work, the investigated plasmonic nanostructures consisted of a “receiving” element with large optical cross section and low damping, a plasmonic Au nanodisk, and a catalytically active second metallic and dissipating (lossy) element, a Pd nanodisk, separated by a thin SiO2 layer. As the main result, we find a very pronounced local absorption enhancement in the catalytic Pd particle compared to the absorption in an identical Pd nanoparticle alone. The absorption enhancement is provided through very efficient energy transfer from the Au nanodisk to the Pd one in the coupled nanoantenna, as clearly shown by the coupled mechanical oscillator model and FDTD simulations. This energy transfer from the Au to the Pd disk is found to be so effective that the absorption efficiency at a given wavelength can be enhanced up 9 times, depending on specific antenna design. The total absorption in the Pd disk, integrated over 400−900 nm for a flat intensity profile of the incoming light, is enhanced up to 2-fold, again depending on the antenna dimensions. The made observations and conclusions for the specific example presented here, that is, Pd as the catalytic material, can be directly extended to other catalytically active transition metals, characterized by large imaginary parts of their dielectric functions at near-visible frequencies, such as Pt, Ni, Ru, Re, Co, or Fe. We also note that the spectral tunability of the maximal absorption enhancement in the catalytic antenna element can

sharp resonance attributed to low losses (mass 1, damping constant γ1, mimics the Au disk) and the second one has high damping (mass 2, damping constant γ2, mimics Pd; γ2 > γ1). The system is schematically presented in Figure 6a and

Figure 6. (a) Two coupled damped harmonic oscillators that mimic the stacked heterometallic nanoantenna structure; the Au disk is represented by mass m1 and spring 1 with spring constant k1, whereas the Pd disk is represented by mass m2 and spring 2 with spring constant k2. The coupling between the two is assured by a third spring with spring constant kc. The entire system is then driven by the external forces F1 and F2. Both oscillators have inherent damping, γ1 and γ2, which, following the properties of Au ad Pd, are small for m1 and large for m2 (γ2 ≫ γ1). (b) The power transferred from the external force into the Au mass is much larger than the one transferred into the Pd mass. However, the Pd mass energy dissipation is up to five times larger than that of Au. Coupling of energy absorbed in the Au and transferred into Pd is responsible for the large absorption increase in Pd. The frequency scale is determined by the resonances of the individual disks.

relations governing the motion of the two masses are given in eqs 1 and 2 x1̈ + γ1x1̇ +

k k1 1 x1 + c x 2 = F1 m1 m1 m1

x 2̈ + γ2x 2̇ +

k k2 1 x 2 + c x1 = F2 m2 m2 m2

(1)

(2)

Here, the spring constants are ki, where i refers to either one of the masses or the coupling spring, losses in each resonator are included by γi, and Fi is the force acting onto mass i. We solve these equations for xi to obtain the power transferred into either of the masses from the external force, as well as to calculate the power dissipated by them. A particular solution of the above equations, corresponding to a 140 nm diameter Au disk and a 100 nm diameter Pd disk, is shown in Figure 6b. The derivation of the model, the coupled dipole approximation, and the specific used values for the different parameters are summarized in the Supporting Information. The power transferred from the external force is shown with solid lines, red for the Au-equivalent mass and blue for Pd. We see that the majority of the energy from the driving force is transferred into 4789

dx.doi.org/10.1021/nl3022187 | Nano Lett. 2012, 12, 4784−4790

Nano Letters

■

be extended to the whole UV−vis−NIR spectral range by using Ag or Al as the receiving antenna element, instead of Au. Thus, our novel absorption-engineering scheme provides a generic and very flexible (in terms of possible material combinations and spectral tunability) platform for the plasmonic enhancement of (photo)catalytic reactions by matching the photon energy of maximal absorption and, consequently, hot electron generation with the specific energies of the LUMO orbitals of an adsorbate molecule on the catalyst surface. The relatively narrow resonances of the stacked heterometallic antennas (compared to the LSPR excitations in the catalyst metals alone28) also facilitate to achieve good overlap with the localized LUMO resonance of the adsorbates. Furthermore we also note that the absorption in the catalytic element can be further engineered by, for example, using Ag instead of Au as the low loss element and by optimizing the shape and relative sizes of the two antenna elements. The latter will be further scrutinized in a forthcoming paper.29 Finally, we notice that our absorption enhancement strategy in principle facilitates exploitation of a second reaction enhancement scheme. While the first scheme, as indicated, targets the enhanced creation and use of energetic electrons as mediators for a catalytic reaction of adsorbed species on the surface of the catalyst, the second scheme may utilize the local heat generation associated with light absorption in plasmonic nanostructures.30 This locally generated heat could then be used to thermally enhance a reaction on the catalyst antenna element surface. In this application the demonstrated wide spectral tunability of maximal absorption enhancement is of key importance since it allows, for example, efficient matching of the wavelength of maximal absorption with the optimal wavelength in terms of irradiated power from a light source like the sun. Furthermore, the use of antenna elements with large ohmic losses (e.g., Ni) offers an additional handle to further maximize the efficiency of the local heating effect.

■

REFERENCES

(1) Lal, S.; Link, S.; Halas, N. J. Nature Photon. 2007, 1 (11), 641− 648. (2) Kreibig, U.; Vollmer, M. Optical Properties of Metal Clusters; Springer-Verlag: Berlin Heidelberg, 1995; Vol. 25. (3) Stuckless, J. T.; Moskovits, M. Phys. Rev. B 1989, 40 (14), 9997− 9998. (4) Shalaev, V. M.; Douketis, C.; Stuckless, J. T.; Moskovits, M. Phys. Rev. B 1996, 53 (17), 11388−11402. (5) Anker, J. N.; Hall, W. P.; Lyandres, O.; Shah, N. C.; Zhao, J.; Van Duyne, R. P. Nature Mater. 2008, 7 (6), 442−453. (6) Larsson, E. M.; Langhammer, C.; Zoric, I.; Kasemo, B. Science 2009, 326 (5956), 1091−1094. (7) Moskovits, M. J. Raman Spectrosc. 2005, 36 (6−7), 485−496. (8) Tam, F.; Goodrich, G. P.; Johnson, B. R.; Halas, N. J. Nano Lett. 2007, 7 (2), 496−501. (9) Shegai, T.; Chen, S.; Miljkovic, V. D.; Zengin, G.; Johansson, P.; Kall, M. Nat. Commun. 2011, 2, 481. (10) Huang, X.; El-Sayed, I. H.; Qian, W.; El-Sayed, M. A. J. Am. Chem. Soc. 2006, 128 (6), 2115−2120. (11) Loo, C.; Lowery, A.; Halas, N.; West, J.; Drezek, R. Nano Lett. 2005, 5 (4), 709−711. (12) De, M.; Ghosh, P. S.; Rotello, V. M. Adv. Mater. 2008, 20 (22), 4225−4241. (13) You, J.; Zhang, G.; Li, C. ACS Nano 2010, 4 (2), 1033−1041. (14) Knight, M. W.; Sobhani, H.; Nordlander, P.; Halas, N. J. Science 2011, 332 (6030), 702−704. (15) Nishijima, Y.; Ueno, K.; Yokota, Y.; Murakoshi, K.; Misawa, H. J. Phys. Chem. Lett. 2010, 1 (13), 2031−2036. (16) Thimsen, E.; Le Formal, F.; Grätzel, M.; Warren, S. C. Nano Lett. 2011, 11 (1), 35−43. (17) Furube, A.; Du, L.; Hara, K.; Katoh, R.; Tachiya, M. J. Am. Chem. Soc. 2007, 129 (48), 14852−14853. (18) Tian, Y.; Tatsuma, T. J. Am. Chem. Soc. 2005, 127 (20), 7632− 7637. (19) Christopher, P.; Xin, H.; Linic, S. Nature Chem. 2010, 3 (6), 467−472. (20) Linic, S.; Christopher, P.; Ingram, D. B. Nature Mater. 2011, 10, 911−921. (21) Hammer, B.; Norskov, J. K. Nature 1995, 376 (6537), 238−240. (22) Langhammer, C.; Kasemo, B.; Zoric, I. J. Chem. Phys. 2007, 126, 19. (23) Fredriksson, H.; Alaverdyan, Y.; Dmitriev, A.; Langhammer, C.; Sutherland, D. S.; Zaech, M.; Kasemo, B. Adv. Mater. 2007, 19 (23), 4297−4302. (24) Dmitriev, A.; Pakizeh, T.; Kall, M.; Sutherland, D. S. Small 2007, 3 (2), 294−299. (25) Johnson, P. B.; Christy, R. W. Phys. Rev. B 1972, 6 (12), 4370− 4379. (26) Johnson, P. B.; Christy, R. W. Phys. Rev. B 1974, 9 (12), 5056− 5070. (27) Langhammer, C.; Yuan, Z.; Zoric, I.; Kasemo, B. Nano Lett. 2006, 6 (4), 833−838. (28) Zoric, I.; Zäch, M.; Kasemo, B.; Langhammer, C. ACS Nano 2011, 5 (4), 2535−2546. (29) Antosiewicz, T. J.; Apell, S. P.; Wadell, C.; Langhammer, C. J. Phys. Chem. C, submitted. (30) Cao, L.; Barsic, D. N.; Guichard, A. R.; Brongersma, M. L. Nano Lett. 2007, 7 (11), 3523−3527.

ASSOCIATED CONTENT

S Supporting Information *

Derivation of the coupled mechanical oscillator model, additional experimental and simulated data for stacked nanoantennas with bottom diameters of 80 and 110 nm, respectively, and optical near-field plots for stacked antennas with different SiO2 spacer layer thicknesses and wavelengths. This material is available free of charge via the Internet at http://pubs.acs.org.

■

Letter

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Author Contributions ‡

These authors contributed equally.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We acknowledge financial support from the Foundation for Strategic Research project RMA 08 (C.W. and T.J.A.) and the Swedish Research Council project 2010-4041 “Nanoplasmonics for Nanomaterials Science” (C.L.). 4790

dx.doi.org/10.1021/nl3022187 | Nano Lett. 2012, 12, 4784−4790