Tomasz J. Antosiewicz*ab and
S. Peter Apellb
aCentre of New Technologies, University of Warsaw, Banacha 2c, 02-097 Warsaw, Poland. E-mail: tomasz.antosiewicz@uw.edu.pl
bDepartment of Applied Physics and Gothenburg Physics Centre, Chalmers University of Technology, 412-96 Göteborg, Sweden
First published on 16th December 2014
Noble metals have recently been shown to drive direct photocatalytic reactions in which they both provide hot electrons via the localized surface plasmon resonance (LSPR) and the catalytically active site. Catalytic reactions are also possible on other metals such as platinum or rhodium which, however, exhibit rather poor plasmonic properties (low field enhancements, low resonance quality factors) and their LSPR for nanometer sized particles occurs in the UV, an unfavourable effect when considering sunlight as a photon source. By coupling the LSPR response of catalytic metal nanoparticles to that of a silver nanoparticle we can excite a hybridized resonance that matches the spectral characteristic of the light source and light absorption in the catalytic metal is enhanced by up to one order of magnitude. This is shown for a number of catalytic metals and is further discussed for model Drude and Drude–Lorentz materials. These results provide guidelines for designing catalytic metal nanostructures which absorb the solar spectrum very efficiently.
An LSPR involved in a photocatalytic process may contribute to it indirectly – by transferring energy from the excited resonance to materials in the vicinity of the metal nanoparticle11 – or via a direct process in which metal nanoparticles serve both as the plasmonic resonator and the catalytic material.12 The efficiency of plasmon induced catalysis is limited by various factors, however, the one considered here is related to the electromagnetic interaction between light and the metal resonator. The transfer of energy can occur either radiatively, by scattering of enhanced electromagnetic fields, or nonradiatively, a process in which energetic electrons are generated.13 These two processes, in principle, compete against each other and depending on the requirements, it is usually beneficial to maximize the desired channel. Yet, despite this trade-off, in any plasmon-assisted process it is of paramount importance to excite the plasmon as efficiently as possible.
In this work we analyse the plasmonic properties of heterometallic structures (nanoparticles) involved in direct photocatalysis: specifically, the focus is placed on poor plasmonic metals. These, in contrast to good metals (e.g. Au or Ag), have high losses in the visible and consequently low resonance quality factors and offer low field enhancements. The reason for choosing early transition metals rather than coinage ones is that, in general, noble metals are not the best catalytic metals14 (for simplicity, when mentioning transition metals we address those except for Au and Ag). That being said, due to their low losses noble metals offer very large field enhancements and as a result are ideal model materials to study direct plasmon-assisted photocatalysis, as recent review articles illustrate.13,15,16 On the other hand, transition metals, which usually exhibit a poor plasmonic response, are known to catalyse many reactions17,18 and could benefit from plasmon-assisted photocatalysis. The LSPR for a given size of a particle made from a transition metal is usually blueshifted in comparison to silver or gold. Placing them on a substrate will redshift the response, however, catalytic metal nanoparticles have, in general, their LSPR in the UV. Thus, one major disadvantage of plasmon-assisted photocatalysis on transition metals is a lack of spectral overlap between the solar spectrum (for sunlight energy harvesting) and the plasmonic response of such metals. Hence, it is prudent to look into possible ways of shifting the plasmonic activity of transition (poor plasmonic) metals (e.g. rhodium,19 vanadium,20 platinum,20 and ruthenium21) to the red so that efficient interaction with sunlight will be possible. A simple way is to use high-index substrates, yet this approach has limitations related to the available refractive indices. Another approach, the one investigated in this work, is to couple plasmonic resonances of adjacent nanoparticles and make use of the new hybridized resonances. We should note, that the complexity of any photocatalytic process, which includes light absorption, electron–hole separation and surface oxidation–reduction reactions, means that a simple enhancement of one of the steps may not translate into improved catalysis. However, if (plasmon-assisted) light absorption is the limiting step, we expect that enhancing it will improve the overall efficiency, assuming all other factors remain unchanged. Furthermore, the presented approach can be viewed not only as absorption enhancement, but also as shifting the optical activity of nanoparticles from one part of the spectrum into another.
Fig. 1 (a) Two coupled metallic disks are used to investigate resonant energy transfer and absorption enhancement in one of the resonators. The bottom disk is made from a good plasmonic metal, silver, to couple efficiently light into localized plasmons and enhance absorption in the top disk made from poor plasmonic metals: Pt, Rh, Ru, and V. The radii of the disks are r1 and r2 for the bottom and top disks, respectively, their thickness is h and separation d. (b) Real and (c) imaginary parts of permittivity of the metals.22 Notice the large variation of permittivities and a very low imaginary part of silver in comparison to other listed metals. |
In order to quantify absorption tunability of plasmonic dimers it is first necessary to calculate the absorption baseline of its constituent elements. In Fig. 2 we plot the absorption efficiency (absorption cross section normalized to geometrical cross section) of disks illuminated normally. The absorption spectrum of silver (Fig. 2a) is characterised by a sharp LSPR up to an amplitude of 7 that redshifts from ca. 400 to 700 nm, while at shorter wavelengths interband absorption manifests itself. In comparison to silver, the other metals have weaker – up to 2 for Pt, 2.5 for V, and 3.5 for both Rh and Ru – and broader resonances – between 2 and 4 times. However, their most important characteristic is that their spectra are shifted to the blue with respect to silver. This is indeed a signature of many poor plasmonic metals as was demonstrated recently.29,30 Of the metals considered here disks 10 nm in radius have their resonances at or slightly below 200 nm with only vanadium exhibiting it at ca. 300 nm. The spectral location of their resonance is connected directly to the real parts of their permittivities. Let us recall that in the quasistatic case a sphere in vacuum exhibits a resonance when its permittivity ε = −2. Note, that εRu and εRh become −2 at ≈200 nm, coinciding with what is seen in Fig. 2c and d. Permittivities of Pt and V are redshifted with respect to Ru, hence their absorption spectra are not so far in the ultraviolet. The difference between Pt and V arises from losses, which are larger for Pt causing a slight blueshift.
In Fig. 3a a rhodium disk is placed on top of a silver disk with r1 = 80 nm. Maximum absorption enhancement occurs at ca. 650 nm for the smallest r2 values and decreases as the Rh disk size increases. The case for small r2 resembles that of a large optical antenna driving a smaller one at a frequency that is off the latter one's resonance. Hence, absorption enhancement is greatest at the wavelength at which the bottom silver disk (the large optical antenna) resonates. However, due to the two disks being in contact with one another, the efficiency decreases rapidly when r2 increases and simultaneously experiences a blue shift. The system exhibits only one plasmon resonance supported by a metal which is neither Ag nor one of the poor plasmonic metals, but rather resembles a disk with material properties representative of a weighted sum. A single, yet heterogeneous metal disk is also the reason for the blue shift – when a plasmonic disk thickens, its resonance shifts to the blue. When the Rh disk is moved away from the silver resonator (d = 10 nm and 20 nm in Fig. 3b and c, respectively), the maximum enhancement decreases. The reason for this is a lowering of the coupling efficiency between the disks due to increased separation. Increased separation between the disks causes a red shift of the absorption enhancement (which follows the Ag resonance) as r2 increases due to coupling between the disks and the resulting anticrossing of the two hybridized modes. For decreasing r1 the Ag LSPR and the absorption enhancement blueshift, yet these previous observations hold.
Fig. 4 presents total absorption enhancement in Rh, defined as
(1) |
Fig. 4 Wavelength-integrated absorption enhancement [eqn (1)] in rhodium dimer element due to coupling to the silver dimer element for different separations d and Ag disk radii r1 as a function of the Rh disk radius r2. Solid lines mark r1 = 50 nm, dashed – 60 nm, dash-dotted – 70 nm, and dotted – 80 nm; symbols indicated separation of 0, 10, and 20 nm for triangles, diamonds, and circles, respectively. At d = 0 nm enhancement saturates at 7 for small Rh disks and decreases with increasing r2. When the Rh disk moves away from the Ag one maximum enhancement decreases while its maximum shifts toward larger radii. |
In Fig. 5 we show the absorption enhancement for all considered poor plasmonic metals only for d = 0 nm for the smallest and the largest used Ag disk, because the enhancement for larger separations (d > 0 nm) is smaller than for d = 0 nm. Out of the four metals the largest enhancement, Ru is the metal which exhibits the largest absorption enhancement by a considerable margin – 9-fold enhancement for r1 = 50 nm (Et over the next best material – Rh – is almost 50% larger) and 13-fold for r1 = 80 nm (twice better than Rh) for disks on the order of 20–30 nm in diameter. This advantage is also observed for other r1 values. Rh, Pt, and V show weaker, but stable enhancements that are qualitatively similar and equal, at maximum, 6, 5, and 4, respectively. Table 1 summarizes all maximum enhancements.
Fig. 5 Wavelength-integrated absorption enhancement [eqn (1)] for touching dimer disks (d = 0 nm) as function of the top disk radius r2 for (a) a silver resonator with r1 = 50 nm and (b) 80 nm. Despite a large variation in permittivities (up to one order of magnitude) the absorption enhancement of the various metals is relatively similar. Despite an increase of the size of the Ag nanoantenna absorption enhancement in Rh, Pt, and V remains almost unchanged and only increases in Ru. |
r1 | Transition metal | |||
---|---|---|---|---|
Pt | Rh | Ru | V | |
50 nm | 5 | 6 | 9 | 4 |
60 nm | 5.5 | 6.5 | 10.5 | 4.5 |
70 nm | 5.5 | 7 | 12 | 4.5 |
80 nm | 5.5 | 7 | 13 | 4.5 |
Another factor determining absorption enhancement is the metal permittivity, whose influence we investigate using Drude and Drude–Lorentz materials. Despite this simplification, it is possible to identify qualitative dependencies which are useful in discussing real metals. This analysis is conducted using the coupled dipole approximation for one disk made from a low loss Drude metal and the second disk made of a poor plasmonic metal. In Fig. 6a we use Drude dispersion for the poor metal disk (plasma frequency ħωp from 6.5 to 14.5 eV, damping ħγp from 10 meV to 4 eV). A large ωp shifts the poor plasmonic metal disk's resonance out of the solar spectrum, decreasing its absorption in the optical regime. Coupling it to a good resonator attuned to the solar maximum increases absorption by more than an order of magnitude. For low damping absorption enhancement Et ≈ 1 when ωp is small, but increases sharply for increasing plasma frequency. When γp increases, the resonance smears out and Et becomes relatively insensitive to the plasma frequency.
Fig. 6 Wavelength-integrated absorption enhancement [eqn (1)] under solar illumination in a poor plasmonic metal disk coupled to a silver disk modelled as a Drude material with ħωAg = 9 eV and ħγAg = 65 meV, and background permittivity 3.7 (LSPR at 2.1 eV). (a) The poor plasmonic metal is modelled as a Drude metal (ωp, γp). For low damping absorption enhancement increases (nonlinearly) with ωp; it is largest when the poor plasmonic metal disk's resonance blueshifts out of the solar maximum. Large damping smoothes out the enhancement. (b) Drude–Lorentz permittivity of the poor plasmonic disk with ħωp = 11 eV and ħγp = 0.5 eV and variable Lorentz resonance frequency ωl and line width γl. Absorption enhancement is almost constant when the Lorentz pole is far away from the LSPR. When the pole is close in frequency to the LSPR, small γl causes Et to be low, while for large γl Et is large. |
The second plot, Fig. 6b, shows absorption enhancement when a Lorentz term (resonance position ωl, damping γl) is added to the dispersion relation of the poor plasmonic metal. In this case absorption enhancement depends mostly on the overlap between the silver disk's LSPR (at 2.1 eV) and the Lorentz contribution to permittivity. When they are spectrally separated, Et is insensitive to γl. When they overlap, Et for a narrow Lorentzian is small, because initial absorption of an uncoupled poor plasmonic disk is already large. On the other hand, when γl is large only a negligible part of the Lorentzian overlaps with the LSPR and the potential for increase is large.
These observations help elucidate the variations in absorption enhancement calculated for real metals. Out of Pt and Rh, which exhibit a relatively smooth permittivity, larger Et is observed in Rh, which has a larger plasma frequency and slightly larger Im(ε) than Pt. A comparison between Ru and V, both metals with very strong and discrete interband transitions in the visible, shows that Ru, having a larger γp and stronger Lorentzians of the two metals, is the metal which is characterized by a larger Et.
Increased absorption in catalytic metals comes from coupling energy into them from incident light via the LSPR of the silver disk. One factor affecting this process is coupling of energy into the silver nanoantenna, this being quantified by the extinction efficiency which exhibits a maximum for a given radius. This energy should then be coupled into the adjacent catalytic nanoparticles with the distance between the two elements being another factor. The larger d is, the less efficient is the coupling and the peak absorption enhancement decreases with d.
The largest enhancements are observed for the smallest nanoparticles, which are on the order of (20 nm)3. It is expected, that smaller nanoparticles will exhibit even larger enhancements. They are not considered in this work, as size quantization and surface scattering would have to be included, as well possible changes to other scattering pathways.31,32 Both, important for nanoparticles smaller than ca. 5 nm, induce red- or blueshifts of the LSPR, however, the expected magnitude of the shifts is on the order of a few hundred meV and this will not affect the expected enhancements significantly.33 Surface scattering, on the other hand, will also contribute significantly to an increased width of the LSPR at sizes comparable to 20 nm, potentially affecting the results for (only) the smallest nanoparticles considered here. This effect has been previously included in the bulk damping and will increase it.34 This particular contribution is expected to increase the absorption enhancement, as discussed in connection to Fig. 6a.
Finally, we would like to address the question of what is the ratio of energy deposited in the transition metal element to that of the whole nanostructure. Like the enhancement, it depends on all parameters considered here. For small disks the ratio is on the order of 0.05 to 0.1, while for big particles approaches 0.9. The low fraction of energy dissipated in the smallest nanoparticles (which have the largest enhancement) may seem to negate the approach, however, one has to remember that it is possible to place more small particles near the Ag resonator. In this manner the enhancement efficiency will not be changed significantly, but the total fraction of energy deposited in the increased number of transition metal nanoparticles will also increase.
We shall explore this issue by first considering one dipole in an external electric field E0. Using the Kramers–Kronig properties of the polarizability as a retarded response function and general high-frequency properties of the polarizability, it is possible to show that the total integrated extinction cross section is equal to the instantaneous response of the dipole . The preceding is nothing but the f-sum rule, with V being the volume occupied by the dipole of the system, ωp the plasma frequency, c is the speed of light, and N the total number of electrons in V.
We now perform the same analysis for an interacting system of two dipoles coupled to E0 and each other,2 and focus on one of the dipoles. Since the polarizability vanishes for high frequencies at least as 1/ω2, the coupling dependence in the dipole–dipole interaction term drops out of the integration. Thus the interacting dipole result is the same as for the non-interacting dipole result. Hence, total extinction for every part of a heterometallic nanostructure is constant. Fortunately there are two relaxing conditions in our case: we are only interested in total absorption and, more importantly, only in a predefined frequency region. The first point indicates that we can at the expense of scattering increase absorption. The second implies that, while any increase in one part of the spectrum is balanced by a proportional decrease in another part, appropriate engineering of the nanostructure assures that in the region of interest absorption is only increased and the corresponding decrease occurs outside of it. Hence, expressing our problem in terms of the equation in the preceding paragraph, we set cut-offs on the integration and for carefully chosen limits (frequency range) absorption may be increased considerably despite the fact that the number of electrons does not increase.
This approach is motivated by the fact that tailoring light–matter interactions for virtually any type of solar harvesting application introduces an inherent cut-off in energy that dictates that photons of lower energy will be lost; simultaneously the energy density of the solar spectrum below 300 nm is negligible. These two values, one of which depends on the considered problem (reaction), limit the range of usable energies in solar harvesting applications and the sum rules are no longer strictly valid. Importantly, some metal nanoparticles (especially for plasmon-assisted photocatalysis), depending on their size and the refractive index of the environment, will have their LSPRs in the UV30 where the solar flux is small or nonexistent. Thus, shifting their resonances into the visible is beneficial. It is important to remember, however, that such shifting schemes are not limited to just the solar spectrum but, in fact, can be employed for any given photon source that is mismatched with respect to the LSPR resonance of the catalytic nanoparticles.
The range of possible absorption enhancements was investigated for a variety of geometrical parameters of the heterometallic nanostructure. By choosing an appropriate size of the low-loss nanoantenna the enhancement spectrum of a catalytic metal nanoparticle can be made to overlap with photon energies that are larger than the activation energy of a given process. Such an approach makes optimum use of the illumination source. While the examples here were based on sunlight and employed shifting plasmonic activity from the UV to the visible, in principle it is possible to apply this coupling method to other light sources and match the enhancement profile to their spectrum. This matching would allow for efficient usage of light in plasmon-assisted photocatalysis with an additional benefit of that photocatalysis at surfaces may be more flexible at controlling single elementary step energetics than thermal catalysis.13,21
Finally, it is important to note that photocatalysis is a complicated process involving not only light absorption, but also such steps like electron–hole separation and surface oxidation–reduction reactions and can be limited by any number of effects like recombination or mass transfer. Thus, simply increasing the rate of energy absorption by some factor may not lead to a corresponding increase of the photocatalytic rate. However, if a given photocatalytic system is not constricted by any limits except for light absorption, then all else equal, we expect that increased absorption should translate into greater photocatalytic efficiencies.
This journal is © The Royal Society of Chemistry 2015 |