Localized and Propagating Plasmons in Metal Films with Nanoholes

Publication Date (Web): March 13, 2013 ... In contrast to Al films where the localized plasmons can be excited from the near-IR to the UV, no plasmon ...
7 downloads 10 Views 2MB Size
Letter pubs.acs.org/NanoLett

Localized and Propagating Plasmons in Metal Films with Nanoholes Markus Schwind,* Bengt Kasemo, and Igor Zorić Department of Applied Physics, Chalmers University of Technology, SE-412 96 Göteborg, Sweden ABSTRACT: The occurrence of plasmon resonances in thin (∼20 nm) Al and Au films, perforated with nanoholes, was studied. In both metals, two plasmon resonances were observed: (i) A surface plasmon polariton mode associated with a maximum in extinction and (ii) a localized resonance in the nanohole associated with a minimum in extinction. By varying the diameter of the nanoholes, the scaling of the peak positions of the plasmon resonances was determined as a function of hole diameter. In the large nanohole limit, the plasmon peak positions depend only on the nanohole diameter being independent of the material. On the other hand, for small nanoholes the plasmon peak positions are material and size dependent. In contrast to Al films where the localized plasmons can be excited from the near-IR to the UV, no plasmon resonances were observed for Au at energies above the interband threshold (2.4 eV). The interaction between a distinct interband transition in Al at 1.5 eV and the localized plasmon resonance is considered in detail. We observe for the first time experimentally a noncrossing behavior of the interband transition and the localized plasmon resonance. The energy (size) dependence of surface plasmon peak width, being a measure for the decay/damping of the latter, is very different for the two metals. This can be explained by considering the different decay mechanisms active in the two metals. Apart from these basic plasmonics results, we test the potential of using the shifts of the plasmon resonances in perforated Al films to follow the atmospheric oxidation/corrosion kinetics of Al. The results are quantified by model calculations. The obtained kinetic law for the oxide growth is in good agreement with a previous XPS study on plain Al films. This suggests that the nanohole-induced plasmon resonances can be a sensitive and simple measure for Al corrosion and metal corrosion in general. KEYWORDS: Localized surface plasmon resonance, surface plasmon polariton, gold, aluminum, nanoholes, oxidation sensing

N

following changes of their LSPRs;19,20 by monitoring LSPR shifts, the solid−liquid phase transition of (Sn) nanoparticles21 or the insulator−metal transition in, for example, vanadium dioxide VO222 can be sensitively characterized. The common denominator in these studies is the fact that the LSPR supporting metal itself is reactive under the corresponding reaction conditions. In addition, the LSPR shifts are not only due to optical changes in the nanoenvironment but also due to changes of the nanoparticle itself like size, shape, band structure, or even temperature.21,23,24 However, despite these promising applications, metals other than Au are yet seldom used in plasmonics and when this is done they are in most cases coming in the form of nanoparticles. Nanohole-based sensing, which has been proven so effective for biosensing,1,25 has so far hardly been explored using metals other than Au. Optically thick Cu26 and Al films27 perforated with nanoholes have, however, previously been explored for their SERS activity. Apart from SERS, Liu et al. studied the electroluminescence enhancement due to perforated Al cathodes in OLEDs.6 In the present work, we characterize the optical resonances of nanoscale holes in thin Al films. We compare these findings to the optical resonances of Au films perforated with similar holes. Furthermore, we demonstrate that fundamental scaling laws for the plasmon spectral peak positions and plasmon peak width,

anoplasmonic sensing1,2 is today one of the major research branches of plasmonics besides related disciplines such as conventional surface plasmon resonance sensing,3 surface-enhanced Raman scattering (SERS),4 optoelectronics,5,6 and metamaterials.7,8 Nanoplasmonic sensing is based on the detection of localized surface plasmon resonance (LSPR) shifts in metallic nanostructures induced by changes of the optical properties of the nanoenvironment. As sensor structures Au nanoparticles are by far the most common ones, often − in biosensing − coated with surfactants or biomarkers/antibodies acting as an active and selective binding surface for biomolecules on the otherwise inert Au nanoparticles.9 It has been shown in many cases that more elaborate nanoparticle synthesis and/or nanostructure engineering, can greatly increase LSPR peak shifts or decrease its linewidths and thereby increase the sensitivity of the plasmonic sensor. Besides examples like nanorice,10 nanostars,11 and nanoeggs,12 one common approach is to use nanometric holes in thin (tens of nanometers) Au films,13−15 which exhibit optical resonances qualitatively similar to surface plasmon polaritons excited in Au films3 and LSPRs observed in Au nanoparticles.16 For plasmonic sensors using nanoparticles, not only classical plasmonic materials like Au and Ag, but also other metals have been successfully applied due to their specific properties.17 For example, hydrogen storage can be studied using plasmonic Pd nanoparticles that form PdHx in a hydrogen atmosphere;18 the corrosion of Al nanoparticles can be directly monitored by © 2013 American Chemical Society

Received: January 26, 2013 Revised: March 4, 2013 Published: March 13, 2013 1743

dx.doi.org/10.1021/nl400328x | Nano Lett. 2013, 13, 1743−1750

Nano Letters

Letter

previously demonstrated for nanoparticles of different materials,17 also apply to nanoholes. The noncrossing behavior of localized interband transitions and LSPRs, which has been theoretically described recently,28 is for the first time observed in an experimental study. Previously, only metals characterized by an interband transition threshold had been studied.29 In addition to these basic aspects, we investigate the performance of nanoholes in Al films as oxidation sensors. The nanoholes may, in this case, be used as probes of film properties, rather than of hole properties, employing simple transmission (or reflection) measurements. In the latter work, the plasmon resonances in Al thin films with holes are used to follow the atmospheric corrosion of Al films. Nanometric holes in gold and aluminum thin films (height h = 20 nm) were prepared by colloidal lithography, described in detail elsewhere.30 In brief, the fabrication consists of the following main steps: (i) A glass substrate is positively charged by the deposition of a monolayer of polyelectrolyte (polydiallyldimethylammonium chloride). (ii) Negatively charged polystyrene spheres are then adsorbed onto the surface. On the one hand, their charge ensures sticking on the surface by electrostatic attraction and, on the other hand, prevents the formation of aggregates as the spheres are repelling each other. This leads to an adsorption pattern with characteristic relatively narrow distribution of nearest neighbor distances, that is, short-range order, but no long-range order. (iii) Then, 20 nm of Au or Al are deposited by physical vapor deposition (at a base pressure of ∼2 × 10−7 Torr). (iv) Finally, the polystyrene spheres are removed using adhesive tape. The resulting nanohole diameter is varied by using different sizes of the polystyrene spheres. Figure 1 displays a schematic of the sample structure (Figure 1A), representative SEM images (recorded using a FEI Quanta 200 FEG ESEM) of a few selected hole diameters (Au, Figure 1B,C; Al, Figure 1D,E), and the corresponding radial distribution functions (Au, Figure 1F; Al, Figure 1G). The latter show a peak at the characteristic nearest neighbor

distance. The absence of any further peaks in the radial distribution function indicates that the samples only exhibit short-range order but no long-range order. The different hole diameters for the Au and Al samples as well as the corresponding characteristic nearest neighbor distances are listed in Table 1. The optical response of the Au and Al thin films perforated by nanometric holes was characterized by extinction measurements using a Varian Cary 500 spectrophotometer. Figure 2A shows the extinction spectra of a plain Au film (black spectrum) and from one perforated by nanoholes 85 nm in diameter (light blue spectrum). The optical spectrum of the Au film can be divided into two regions: (i) Below 2.4 eV, the metal behaves very much like a free electron metal and can be described by a Drude dielectric function. The extinction in this regime decreases for increasing photon energy. (ii) Above 2.4 eV, various interband transitions occur, which leads to a significant increase of extinction. Thus, a minimum in extinction occurs at the threshold of the interband transitions (marked by “IT” in the graph). For the Au film with nanoholes, one additional minimum and maximum are observed. As described in detail by Sannomiya et al.,31 the maximum in extinction corresponds to the excitation of a surface plasmon polariton mode (SPP) whereas excitation of a localized surface plasmon resonance (LSPR) mode inside the nanoholes leads to the additional minimum in extinction (see the schematics in Figure 2B). The SPPs are excited when the periodicity, that is, the nearest neighbor distance in the case of short-range order particle ensembles, matches the SPP wavelength. The condition for SPPs can be calculated by solving the dispersion relation for the specific material and film thickness. It is noteworthy that in case of an ideal metal, the SPP only depends on the holes’ nearest neighbor distance but not on the holes’ size/geometry. It has been confirmed experimentally that the extinction maximum, associated to the SPP, remains at essentially the same spectral position when the hole size or shape was changed but the spacing between the holes remained the same.31 On the other hand, the extinction minimum strongly depends on the geometry of the nanoholes, which is in agreement with the picture of a localized excitation in the nanohole.31,32 In Figure 2C, the extinction spectra of a plain Al film (black spectrum) and an Al film with nanoholes 84 nm in diameter (pink spectrum) are shown. Aluminum is well described by a Drude-like behavior all throughout the spectral region from the near-infrared to the deep-UV and the extinction increases gradually with increasing photon energy. Only at around 1.5 eV, can a small feature be recognized. This increase in extinction is caused by an interband transition (IT) localized in a narrow energy range around 1.5 eV. It is attributed to transitions between a pair of parallel bands around the Σ axis on the Γ−K−W−X plane of the Brillouin zone.33 In addition to this feature, the Al film with nanoholes, just like the Au film with nanoholes, shows one additional extinction maximum and minimum. In analogy to Au, we attribute these to the SPP in the perforated film and to the LSPR in the nanoholes, respectively. This is supported by previous studies, which report the occurrence of extraordinary transmission enhancement, which is caused by the coupling of SPPs on the two sides of optically thick films to the localized excitation in the nanoholes, in Al films.27,34 So far, Al films with nanoholes have, however, not been studied in the thin film limit. In order to explore systematically the properties of the plasmon resonances in Au and Al films, the hole diameter has

Figure 1. (A) A schematic figure of the metal films perforated with nanoholes of diameter D and height h. The films are supported on borofloat glass. (B,C) SEM images of selected Au films with nanoholes. (D,E) SEM images of similarly perforated Al films. (F,G) Plots of the radial distribution functions, obtained from the SEM images shown for Au and Al samples, respectively. One peak associated with the characteristic nearest neighbor distance can be observed. The fact that the intensities for distances larger than that are very close to one shows that the nanoholes only exhibit short-range order. 1744

dx.doi.org/10.1021/nl400328x | Nano Lett. 2013, 13, 1743−1750

Nano Letters

Letter

Table 1 Au Al

hole diameter D (nm) char. nearest neighbor dist. (nm) hole diameter D (nm) char. nearest neighbor dist. (nm)

513 1017 479 934

309 598 275 636

187 308 260 564

166 285 188 381

144 261 131 309

112 191 118 254

85 144 103 252

70 129 84 188

61 127

proportional to 1/D. The solid lines reflect a linear fit to the experimental data points in this regime and through the origin. For the SPP mode, we observe a slope of 449 eV nm for Au and 390 eV nm for Al and, for the LSPR mode, a slope of 322 eV nm for Au and 260 eV nm for Al. (ii) The second regime is characterized by material dependent plasmon energies. Clearly, aluminum features plasmon resonances even in the UV whereas in gold the peak position asymptotically approaches a value of about 2 eV. Figure 3C shows that the linear fits of the SPP mode in regime (i) for both metals are identical within the experimental error, that is, the SPP peak position in regime (i) is only size but not material dependent. The same can be seen for the LSPR mode in Figure 3D. These two scaling regimes, material independent plasmon energies for large diameters and material dependent ones for small diameters, have been previously observed for Au and Al nanoparticles.17 Not only the observation of the two regimes, but also the slopes of the 1/D-dependence are very close for LSPRs in films with nanoholes and nanoparticles, which shows once again the fundamentally similar character of the localized plasmon excitation in these two geometries. For nanoparticles, the scaling of the two regimes can be described in general by the electrostatic spheroid theory in the modified long wavelength approximation (MLWA), which we have previously applied successfully to describe LSPRs in Al nanoparticles.19,20 Principally, ignoring all influences by the surrounding medium, the LSPR energy is determined by the plasmon energy in the bulk material as well as the geometry of the nanoparticle. For spheroid particles in the large particle diameter limit, Zorić et al.17 showed how the bulk plasmon energy cancels out and the LSPR energy only depends on 1/D. Thus, the LSPR energy looses its material dependent contribution, which is reflected very well in the experimental results for nanoparticles17 as well as for nanoholes (this study, cf. Figure 3). In a simplified picture, this can be rationalized by considering the retardation time, τ = D/c with c the speed of light, needed to polarize the nanoparticle or nanohole. As the plasmonic cavity (particle or hole) becomes larger, the plasmon energy red shifts and the period of the collective electron oscillations approaches the retardation time τ. According to the Heisenberg energy-time uncertainty relation, the plasmon energy then becomes equal to (h/τ) = (hc/D), that is, it scales with 1/D. The other side of the scaling relation between plasmon energy and diameter D is dominated by the bulk plasmon energy, which is material dependent. It is therefore not surprising that Al shows plasmon resonances further in the UV than Au as the bulk plasmon energy is much larger for Al. Likewise, this can be explained by considering the IT threshold for d-band transitions in Au at around 2.4 eV. In order to closer examine the interplay between interband and plasmon excitations, their influence on each other, as observed in the extinction spectra, is plotted in Figure 4. On the x-scale, the expected nanohole LSPR energies, as obtained from the 1/D-scaling, are plotted (the LSPR energies have been obtained from the linear fit displayed in Figure 3). The y-scale

been varied. The different hole diameters and characteristic nearest neighbor distances are listed in Table 1. Figure 2D shows the optical spectra for a Au film as well as for the Au films perforated with nanoholes of different diameters. For reasons of clarity, the nanohole spectra have been vertically offset by 5% from each other. Both the extinction maximum, associated with the SPP on the film surfaces, and the extinction minimum, attributed to the LSPR in the nanohole, shift to higher energy with decreasing hole diameter. The hole diameter alone should not have any influence on the SPP. However, using standard colloidal lithography, the nearest neighbor distance changes along with the hole diameter (the nearest neighbor distance is generally about twice the hole diameter). In this study, we therefore do not focus on the differentiation between the two types of modes. This has been the subject of previous studies.31,32 Instead we make use of the spectrally shifted modes to study material dependent scaling laws as well as the interaction with the modes and interband transitions. It can, for example, be noted that the plasmon resonances never occur at photon energies above 2.4 eV, that is, above the interband threshold in Au, which indicates strong damping of the former by the latter. Furthermore, the IT threshold slightly shifts to the blue when the plasmon energy approaches it. This observation will be addressed in more detail below. Figure 2E shows the results of a systematic variation of the hole diameter in Al films. Once again, the spectra are offset by 5% for reasons of clarity. For Al, the plasmon resonances can be observed from the near-infrared up to the UV region. They occur, in contrast to Au, on both sides of the IT at 1.5 eV, which is due to the localized (in energy space) nature of this interband transition in Al. It can be noted that the IT position in nanohole perforated Al films deviates considerably from the IT position in the plain film when the plasmon resonances are close to 1.5 eV. Obviously, we have here a situation where a collective excitation (LSPR) interacts strongly with a single particle excitation (IT) when the two are at the same energy. Here, we actually observe for the first time experimentally a noncrossing behavior, which has recently been studied theoretically.28 This issue will be considered in detail further down the text. Figure 3 shows the scaling of the peak position (plasmon energy) of the SPP and LSPR associated with the nanoholes for Al (Figure 3A) and Au (Figure 3B) as well as of the SPP and LSPR for both metals combined (Figure 3C,D, respectively). The plasmon peak position is plotted against 1/D, that is, the reciprocal of the nanohole diameter. As previously stated, the SPP does not actually depend on the hole diameter but it rather scales with the nearest neighbor distance of the nanoholes. For the samples in this study, the nearest neighbor distance changes in line with the hole diameter, being about twice the hole diameter, thus, to be able to compare the scaling of the two modes directly, also the SPP mode has been plotted against 1/ D. For both metals, two regimes can be recognized. (i) For large diameters, that is, small 1/D values, the peak position is 1745

dx.doi.org/10.1021/nl400328x | Nano Lett. 2013, 13, 1743−1750

Nano Letters

Letter

Figure 3. The scaling of the SPP and LSPR in the nanohole with the inverse of the nanohole diameter are shown for (A) Al and (B) Au. For both metals, two regimes can be recognized: (i) For large diameters, small 1/D, the plasmon resonance proportional to 1/D (the solid lines are linear fits to the data points in this regime). (C,D) The SPP and LSPR, respectively, in this regime are material independent (deviations are most likely due to too few data points in this regime). (ii) For small diameters, large 1/D, the plasmon resonance is material dependent. In the case of Al, plasmons can be excited even in the UV, whereas in the case of Au, the plasmon resonance asymptotically approach the interband threshold.

Figure 4. (A) Interaction of the plasmon resonance and the interband transition in the case of Au. Along the x-axis, we plot the LSPR energy at which the plasmon resonances would occur if they were not material dependent, that is, the energies obtained by the linear fit shown in Figure 3B. The plasmon resonances in Au asymptotically approach the interband threshold which is influenced slightly and shifts by about 50 meV to the blue. (B) The interaction of the localized interband transition in Al and the plasmon resonances. A noncrossing behavior can be observed were the LSPR converges into the interband transition and vice versa. The black lines are obtained from a recent theoretical study describing this noncrossing behavior in general and using reasonable model parameters for Al film.28

Figure 2. (A) The extinction spectra for a solid Au film (black) and an Au film perforated with nanoholes (blue). The Au film shows a minimum in extinction owing to the interband transition (IT) threshold at ∼2.4 eV. Two additional features are caused by plasmon resonances in the perforated Au film. The maximum in extinction is associated with the SPP and the minimum in extinction is attributed to the LSPR in the nanohole. (B) Schematic charge distribution for the two plasmon modes, SPP and LSPR. (C) displays extinction spectra for an Al film (black) and Al film with nanoholes (pink). The film shows a small feature at the localized interband transition. The perforated film shows, as for Au, two more features attributed to the different modes. (D) Extinction spectra for an Au film and Au films with nanoholes of various diameters. (E) Extinction spectra for an Al film and Al films perforated with nanoholes of different diameters. In both (D) and (E), the spectra are offset by 5% from each other for reasons of clarity. It can be seen that for both Au and Al, the plasmon resonances shift to the blue with decreasing nanohole diameter (and nearest neighbor distance).

displays the actual energies at which the LSPR, the IT threshold in Au, and the localized IT in Al are observed, that is, were there no IT excitations and no material dependence, the curves would just be straight lines. For Au, in Figure 4A the previously seen asymptotic behavior can be observed (cf. Figures 2D and 3B). As has been noted 1746

dx.doi.org/10.1021/nl400328x | Nano Lett. 2013, 13, 1743−1750

Nano Letters

Letter

Figure 5A,B shows the dependence of plasmon peak width on the plasmon energy for nanohole perforated Au and Al

before, not only does the excitation of IT above the threshold prevent any occurrence of plasmon resonances beyond the IT threshold energy but also the plasmon resonances influence the IT and the threshold shifts to slightly higher energies as the plasmon resonance approaches it. This is in line with competing electron excitation channels: (i) collective electron excitation (plasmons) and (ii) single electron excitations (interband transitions). Because of the large oscillator strengths of the interband transitions in Au above 2.4 eV, the plasmon resonances are influenced greatly, whereas the IT threshold is only shifted by ca. 50 meV (cf. inset Figure 4A). The influences of the IT threshold and plasmon excitations has been previously addressed not only in Au nanostructures but also for Cu nanoeggs.29 Far more interesting is, however, the interaction between an IT localized to a narrow energy interval and LSPR excitations. This topic has been the subject of a recent theoretical study.28 Pakizeh derived a noncrossing rule for localized IT and LSPRs based on quasistatic theory. Figure 4B shows for the first time this noncrossing behavior experimentally. The localized nature of the IT in Al makes it possible to excite plasmons not only at energies lower but also at energies higher than the IT energy. The plasmon resonances in Al are for the plotted spectral region (0.5−2.5 eV) still in the material independent regime and generally well described by the linear fit displayed in Figure 3A. Around 1.5 eV, the plasmon energies, however, deviate slightly from this linear trend. Even more noticeable are the deviations of the IT from 1.5 eV as observed in the plain Al film. The black solid lines in Figure 4B have been obtained from the general description when LSPR and IT are interacting as derived in ref 28. (eq 3 in that reference; for the Lorentzian term describing the interband excitation, the IT energy as observed in the plain Al film and gain G0 = 0.1 was used; damping was neglected for reasons of simplicity). The good agreement between the theoretical model and our experimental results is striking but may not be too surprising as the dielectric function (and thus the optical properties governing interband transitions and plasmon excitations) of Al is well described by a Drude term plus a Lorentzian accounting for the localized IT, which has been the basis for the theoretical description.28 In addition to the peak position of the different Au and Al films with nanoholes, also their plasmon peak widths give insight into fundamentals of nanoplasmonics. The plasmon peak widths are determined by different regimes that may or may not apply to certain metals and in certain spectral regions.17 The plasmon peak width is a measure for the damping/decay of the plasmon resonance, that is, it is inversely related to the lifetime of the plasmon excitation (if inhomogeneous broadening can be excluded). Zorić et al.17 consider three general contributions to the overall peak width in order to explain the observed trends for the plasmon peak widths: (i) Intraband damping caused when individual electrons scatter and leave the collective plasmon oscillation; (ii) interband damping results from the decay of the plasmon into the single electron excitation when the interband transition energy matches the plasmon energy; and (iii) radiation damping describes the loss of energy when the dipole induced by the plasmon oscillation radiates. Whereas intraband damping hardly shows any energy dependence and can thus be treated as a constant addition to the overall plasmon peak width, interband damping and radiation damping are energy and material dependent, leading to characteristic scaling of the plasmon peak width for different materials.

Figure 5. (A,B) The dependence of the plasmon peak width of the SPP mode on its peak position for Au and Al, respectively. The observed features due to the corresponding plasmon damping mechanisms as discussed in detail in the text.

films. In this figure, only the energy dependence of the line width of the SPP mode is plotted. Both the LSPR mode and the SPP mode follow roughly the same trend, however, the spectral position of the features in the SPP mode agree better with the observations reported for nanoparticles.17 At this point, it can only be speculated that this may be an indication that the SPP mode, as it is obviously bound to the film, is the major promotor of material dependency for plasmon resonances in Al films with nanoholes. It would intuitively not be surprising if the fact that the nanohole obviously does not contain any metal renders the LSPR localized in the nanohole less prone to the damping mechanisms described above. On the other hand, the nanohole LSPR cannot be free from any material dependency either due to the following reasons: First, the oscillating electrons stem from the metal and are not detached from it either and second, the SPP and LSPR modes are coupled, thus, any material dependence of the SPP should to some extent be reflected in the LSPR. For Au, Figure 5A, the peak width is smallest around 2.1 eV plasmon energy. For higher and lower energies, the peak width increases. In the studied interval, the highest value is observed close to 1 eV. In the case of Al, the peak width is smallest just above 1 eV. For energies higher than that, the peak width increases steplike to almost three times the lowest value and rises further for very high energies. These trends can be explained by considering the aforementioned mechanisms: The occurrence of interband damping obviously requires the presence of band-spacings matching the plasmon energy so that the plasmon oscillation can decay into the IT. For Au, this is fulfilled above the IT threshold. The peak width increase for plasmon energies above 2.1 eV is therefore attributed to the fact that the plasmon energy approaches the IT threshold. For energies lower than 2.1 eV, radiation damping leads to an increase in peak width. For Al, in the considered spectral region only the localized IT at 1.5 eV is of importance. As can be seen in Figure 5B, above 1.5 eV, the plasmon peak width is much increased compared to below this value. Zorić et al.17 observed a steplike increase of the plasmon peak width right at 1.5 eV. This is not observed in the present study as there are simply no data points so very close to 1.5 eV. The increase in peak width for energies above 3 eV, corresponding to very small nanohole diameters, is most likely due to inhomogeneous broadening (broadening because not all nanoholes in the film have the same dimensions and the 1747

dx.doi.org/10.1021/nl400328x | Nano Lett. 2013, 13, 1743−1750

Nano Letters

Letter

spread of nearest-neighbor distances increases). Probably because Al in any oxygen containing environment forms a thin passivating oxide on top of the metal, this inhomogeneous broadening is more pronounced for Al compared to Au (the same has been observed by Zorić et al.17). Previously, we have studied the oxidation of Al nanodisks in air20 and water19 by following the shifts of the LSPR in these Al nanodisks, which provides a simple and very sensitive way of studying the oxidation of Al nanoparticles (and metal nanoparticles in general when other metals are employed). We observed that the results agreed very well with the mechanisms reported for bulk Al oxidation. This may at first sight seem surprising since the small size of the nanometric disks as well as their edges and sides constitute a deviation from a plain metal, which could lead to different oxidation kinetics for the two cases (extended samples versus disks). It seems likely, however, that at least for disks large compared to the average grain size (which has been the case in the previous studies) and with a relatively large width/height ratio, the oxidation kinetics measured on a nanodisk should quite well mimic the kinetics for plain Al films. With these comments as a background, it is actually quite interesting to explore how well an Al film with nanoholes may mimic Al film oxidation. We have therefore followed the atmospheric oxidation/ corrosion of Al films perforated with 144 nm holes (width/ height ratio about 7). The samples were kept in a cleanroom environment at T = 21 ± 1 °C and H = 43 ± 3% relative humidity. The shifts of the plasmon resonances have been measured over a course of 50 days. Figure 6A shows the shifts of the nanohole LSPR plotted against a logarithmic time scale. The nanohole LSPR energy shows a rather linear trend in the graph, that is, assuming a somewhat linear dependence between LSPR energy and oxide thickness the oxidation follows a logarithmic time law as has also been found for bulk Al.35 Figure 6B displays the shifts of the SPP energy against a logarithmic time scale (bullet data points, the lower x-scale (time) applies). The SPP peak shifts about twice as much as the LSPR peak (here, we remind that the SPP samples the whole Al film whereas the LSPR is localized to the hole region), however, the scattering of the data points is also slightly larger so that in terms of signal-to-noise, both measures are rather similar. For Al nanodisks, analytical calculations of the nanodisk LSPR response based on the modified long wavelength approximation (MLWA) for core−shell spheroidal nanoparticles enabled quantification of the shifts observed during the nanoplasmonic sensing experiment, that is, the associated increase of the oxide thickness and refractive index of the oxide could be extracted.19 In the case of nanohole LSPRs, no simple analytical description is available so far, that could enable quantification of the observed data. On the other hand, for the SPP mode shifts due to a growing oxide film can be calculated as described in detail by Dahlin et al.36,37 In brief, the dispersion relations are calculated for different metal film thicknesses that then yield the SPP shifts at the SPP peak position. In this model, the influence of the nanoholes on the dispersion relation are neglected but, as tested in detail by Dahlin et al.,36 this approximation is reasonable as the contribution of the nanoholes in most cases only adds a constant offset. In our case, we have solved the dispersion relations for Al films with layers of different oxide thicknesses. The dielectric function for Al was taken from Palik.38 For the aluminum oxide, a refractive index of 1.73 was used as reported for amorphous aluminum oxide in the applying spectral range.39 The glass substrate below the Al

Figure 6. The atmospheric corrosion of Al films has been followed by measuring the peak shifts of the plasmon resonances associated with the nanoholes in the film. (A) The shifts of the LSPR mode versus time (logarithmic scale). As the data points follow almost a straight line, the peak position shifts logarithmically with time as Al is oxidized. (B) The peak shifts of the SPP mode versus time (the lower x-scale (time) applies). The solid line is obtained from model calculations of the SPP mode for different oxide thicknesses (the upper x-scale (oxide thickness) applies). A very good agreement between theoretical and experimental peak shifts is obtained for the given oxide growth. Thus, the experimentally observed peak shifts can be converted to oxide thickness, as shown in (C). The solid blue line is the logarithmic time law obtained from the model calculation. The yellow square data points correspond to oxide thicknesses measured using angledependent XPS as previously reported20 and the solid yellow line a logarithmic fit to these data points. The order of magnitude of the logarithmic prefactor is the same for both measurements, differing by a factor of 2, which can be attributed to the errors associated with the measurements. This suggests that the SPP mode in perforated Al films is a simple and sensitive measure for Al oxidation/corrosion kinetics.

film and the air above the aluminum oxide were considered as semi-infinite media with refractive indices 1.52 (borofloat glass) and 1.00, respectively. In order to account for the expansion when Al oxidizes, the volume expansion was calculated assuming Al2O3 as the only corrosion product with a density ρAl2O3 = 3.0 g/cm3 for the oxide35 and ρAl = 2.7 g/cm3 for the Al metal.35 The dispersion relations have been calculated for oxide thicknesses hAl2O3 = 0 to 7.5 nm and the corresponding Al film thickness (when the oxide gets thicker the Al film gets thinner) were calculated using the following relation hAl = 20 nm − hAl 2O3

2ρAl O MAl 2 3

ρAl MAl 2O3

≈ 20 nm − 0.60hAl 2O3 (1)

with MAl and MAl2O3 the molar masses of Al and Al2O3, respectively. The result is shown in Figure 6B (solid line, the x-scale on top (oxide thickness) applies). As can be seen, the agreement between the experimentally obtained peak shifts and the calculated ones is very good when a logarithmic relation 1748

dx.doi.org/10.1021/nl400328x | Nano Lett. 2013, 13, 1743−1750

Nano Letters

Letter

regions. Finally, we applied the plasmon resonances as probes to study the atmospheric corrosion of Al. The shifts of the SPP plasmon resonance can be quantified using model calculations and thus yield information on the actual oxide growth kinetics. The kinetics was compared to data previously obtained by angle-dependent XPS measurements on plain Al films oxidized under the same conditions. The results suggest that the nanoholes do not influence the oxidation kinetics and nanohole based nanoplasmonic sensing can be used as a simple but very sensitive measurement for the corrosion kinetics of aluminum. Reasons why nanoplasmonic sensing would be chosen over other techniques allowing measurements on the plain film are the high sensitivity (4% of a monolayer of oxide could be detected in a nanoplasmonic sensing study on nanodisks19), the simplicity of the measurement and the remote character that allows measurements in basically any optically transparent medium and environment (for example, liquids19 and any kind of gas2) with little requirements on the setup for space and access to the sample (for example, inside UHV chambers21 or gas flow reactors2). In essence, for being as sensitive as nanoplasmonic sensing is, its setup is probably most flexible, portable (using, for example, array spectrometers that can be of the size of portable harddrives) and least expensive compared to the other techniques available today. Further support for this conclusion comes from a recent similar LSPR corrosion study on Cu films.40

between the oxidation time and the oxide thickness is assumed. As was stated before, a logarithmic time law for atmospheric corrosion of Al has been reported previously.35 Above, we reflected how the oxidation kinetics measured for disks may deviate from those measured for Al film due to influences of, for example, size and edges. In the case of oxidation/corrosion sensing using the plasmon resonances associated with the nanoholes in an Al film, one is studying an extended Al film, however, this time the nanoholes and their edges and the walls inside the holes pose a deviation from the plain film. It is therefore interesting to compare the obtained oxidation time law with measurements on plain Al films. The oxide thickness on evaporated plain Al films was obtained from angle-resolved XPS measurements as previously reported.20 The Al films and the Al films with holes studied here were both evaporated in the same vacuum chamber under the same conditions and the oxidation conditions were identical as well. Thus, the oxidation should proceed by the same time law if the nanoholes do not influence the oxidation behavior. Figure 6C correlates oxide increase of a plain Al film obtained by XPS analysis and the oxide increase obtained from the aforementioned quantification of the shifts of the SPP modes in Al films with nanoholes. For the former, a logarithmic fit to the data and, for the latter, the logarithmic time law obtained from the model calculations are shown as well. The order of magnitude of the logarithmic prefactor is the same for both measurements, differing by a factor of 2, which can be attributed to the errors associated with the measurements. It should be noted that the relatively large error in the nanoplasmonic sensing measurement is caused by the fact that the sample had to be removed from the spectrophotometer after recording a data point, as the equipment could not be dedicated to solely this study for 50 days. The majority of the error is simply due to slight differences in sample position for each measurement. The error is greatly reduced when the sample is not moved for the whole experiment (cf. ref 19, for example). To summarize, in this paper we have studied the occurrence of plasmon resonances in thin (∼20 nm) Al films perforated with nanoholes and compared their properties to plasmon resonances in the same nanostructures fabricated from Au. Identical to Au, we find for Al films that the perforation with nanoholes leads to a maximum and minimum in extinction, which are attributed to SPP and LSPR in the nanohole, respectively. The plasmon resonances in Al are observed in the whole spectral range from near-IR to UV. For both metals, they are dependent on the nanohole diameter and/or nearest neighbor distance. In the large diameter limit, the plasmon energy is found not to be material dependent as has been observed for nanodisks, previously.17 In the small diameter limit, the plasmon energy position depends on the material. For Au, no plasmons occur above the interband threshold at around 2.4 eV. For Al, no such upper limit has been observed. In contrast, for Al, plasmons are observed at energies both higher and lower than the localized interband transition at 1.5 eV. The interband transition and the plasmon do interact, and a noncrossing behavior, as described in a recent theoretical study,28 is for the first time observed experimentally. The peak widths of the plasmon resonances, that is, the plasmon damping, are found to be characteristically different for Au and Al. The observed trends can be explained by considering the contributions of different damping mechanism, whose strengths are different for the two metals in different spectral



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



REFERENCES

(1) Anker, J. N.; Hall, W. P.; Lyandres, O.; Shah, N. C.; Zhao, J.; Van Duyne, R. P. Nat. Mater. 2008, 7, 442−453. (2) Larsson, E. M.; Langhammer, C.; Zorić, I.; Kasemo, B. Science 2009, 326, 1091−1094. (3) Homola, J.; Yee, S. S.; Gauglitz, G. Sens. Actuators, B 1999, 54, 3− 15. (4) Campion, A.; Kambhampati, P. Chem. Soc. Rev. 1998, 27, 241− 250. (5) Collin, S.; Pardo, F.; Pelouard, J.-L. Appl. Phys. Lett. 2003, 83, 1521−1523. (6) Liu, C.; Kamaev, V.; Vardeny, Z. V. Appl. Phys. Lett. 2005, 86, 143501−143501−3. (7) Smith, D. R.; Pendry, J. B.; Wiltshire, M. C. K. Science 2004, 305, 788−792. (8) Shalaev, V. M. Nat. Photonics 2007, 1, 41−48. (9) Olofsson, L.; Rindzevicius, T.; Pfeiffer, I.; Käll, M.; Höök, F. Langmuir 2003, 19, 10414−10419. (10) Wang, H.; Brandl, D.; Le, F.; Nordlander, P.; Halas, N. Nano Lett. 2006, 6, 827−832. (11) Hao, F.; Nehl, C. L.; Hafner, J. H.; Nordlander, P. Nano Lett. 2007, 7, 729−732. (12) Knight, M. W.; Halas, N. J. New J. Phys. 2008, 10, 105006. (13) Dahlin, A.; Zäch, M.; Rindzevicius, T.; Käll, M.; Sutherland, D. S.; Höök, F. J. Am. Chem. Soc. 2005, 127, 5043−5048. (14) Rindzevicius, T.; Alaverdyan, Y.; Sepulveda, B.; Pakizeh, T.; Käll, M.; Hillenbrand, R.; Aizpurua, J.; Gara de Abajo, F. J. J. Phys. Chem. C 2006, 111, 1207−1212. (15) Tetz, K. A.; Pang, L.; Fainman, Y. Opt. Lett. 2006, 31, 1528− 1530. (16) Dahmen, C.; von Plessen, G. Aust. J. Chem. 2007, 60, 447−456. 1749

dx.doi.org/10.1021/nl400328x | Nano Lett. 2013, 13, 1743−1750

Nano Letters

Letter

(17) Zorić, I.; Zäch, M.; Kasemo, B.; Langhammer, C. ACS Nano 2011, 5, 2535−2546. (18) Langhammer, C.; Zorić, I.; Kasemo, B. Nano Lett. 2007, 7, 3122−3127. (19) Schwind, M.; Langhammer, C.; Kasemo, B.; Zorić, I. Appl. Surf. Sci. 2011, 257, 5679−5687. (20) Langhammer, C.; Schwind, M.; Kasemo, B.; Zorić, I. Nano Lett. 2008, 8, 1461−1471. (21) Schwind, M.; Zhdanov, V. P.; Zorić, I.; Kasemo, B. Nano Lett. 2010, 10, 931−936. (22) Lei, D. Y.; Appavoo, K.; Sonnefraud, Y.; Haglund, J.; Maier, S. A. Opt. Lett. 2010, 35, 3988−3990. (23) Liu, M.; Pelton, M.; Guyot-Sionnest, P. Phys. Rev. B 2009, 79, 035418. (24) Langhammer, C.; Larsson, E. M.; Kasemo, B.; Zorić, I. Nano Lett. 2010, 10, 3529−3538. (25) Dahlin, A. B.; Wittenberg, N.; Höök, F.; Oh, S.-H. Nanophotonics 2012, DOI: 10.1515/nanoph-2012-0026. (26) Anema, J. R.; Brolo, A. G.; Marthandam, P.; Gordon, R. J. Phys. Chem. C 2008, 112, 17051−17055. (27) Yang, Z.-L.; Li, Q.-H.; Ren, B.; Tian, Z.-Q. Chem. Commun. 2011, 47, 3909. (28) Pakizeh, T. J. Phys. Chem. C 2011, 115, 21826−21831. (29) Wang, H.; Tam, F.; Grady, N.; Halas, N. J. Phys. Chem. B 2005, 109, 18218−18222. (30) Fredriksson, H.; Alaverdyan, Y.; Dmitriev, A.; Langhammer, C.; Sutherland, D.; Zäch, M.; Kasemo, B. Adv. Mater. 2007, 19, 4297− 4302. (31) Sannomiya, T.; Scholder, O.; Jefimovs, K.; Hafner, C.; Dahlin, A. B. Small 2011, 7, 1653−1663. (32) Koerkamp, K. J. K.; Enoch, S.; Segerink, F. B.; van Hulst, N. F.; Kuipers, L. Phys. Rev. Lett. 2004, 92, 183901. (33) Lee, K.-H.; Chang, K. J. Phys. Rev. B 1994, 49, 2362−2367. (34) Ekinci, Y.; Solak, H. H.; David, C. Opt. Lett. 2007, 32, 172−174. (35) Godard, H. P. Aluminum. In The Corrosion of Light Metals; The Corrosion Monograph Series; John Wiley & Sons, Inc.: New York, 1967; pp 1−218. (36) Dahlin, A.; Zahn, R.; Vörös, J. Nanoscale 2012, 4, 2339−2351. (37) Junesch, J.; Sannomiya, T.; Dahlin, A. B. ACS Nano 2012, 6, 10405−10415. (38) Palik, E. D. Handbook of Optical Constants of Solids; Academic Press: Orlando, 1998. (39) Gils, S. V.; Melendres, C. A.; Terryn, H. Surf. Interface Anal. 2003, 35, 387−394. (40) Schwind, M.; Hosseinpour, S.; Johnson, C. M.; Langhammer, C.; Zorić, I.; Leygraf, C.; Kasemo, B. Unpublished work.

1750

dx.doi.org/10.1021/nl400328x | Nano Lett. 2013, 13, 1743−1750