Tuning surface plasmons in interconnected hemispherical Au shells

We present a new approach for making interconnected hemispherical shells by stripping Au from templates of anodized aluminum, where the metal thickness can be adjusted without affecting the outer radius of curvature, film roughness and the sharpness of the hemisphere contact areas. This provides increased understanding of the surface plasmon resonances (SPRs) observed for Film-On-Nanospheres (FONs) by decoupling these parameters, which are coupled in the case of FONs. Investigating the influence of the shell thicknesses on the spectral positions of SPRs for FONs involves a dielectric core with a fixed radius encased by a metal film with adjustable thickness. By performing linear reflection spectroscopy, we demonstrate a wide tunability of the SPR by tailoring the inner hemisphere diameter, while keeping the outer diameter fixed. Deposition of extra Au on top of thick, previously stripped hemispherical shells isolates optical response contributions from Au grainand islandmediated roughness, and unsharpening contact areas in form of decreasing LSPR quality factor. Two-photon luminescence scanning optical microscopy of shells with different thicknesses, applying several different laser wavelengths, is exploited to map local electromagnetic hot spots and correlate the high field enhancements with the linear reflection spectroscopy


Introduction
Continuing advancements in the fabrication of regularly patterned metal nanostructures allow controlled and fast production of smaller and sharper features, which are made to cover increasingly larger sample areas, homogeneously.Reproducible tailoring of nanostructure dimensions in conjunction with the proven ability of metal nanostructures to support localized and propagating surface plasmon (SP) resonances, with explicit energies and attractive spatially localized electromagnetic field enhancements, is opening new avenues for sensor designs based on for example surface-enhanced Raman spectroscopy (SERS) and for single molecule detection [1][2][3].Low-loss propagation of SPs can be supported by sharp V-shaped grooves thereby facilitating control of the direction of propagation for use in optical devices [4,5], and SPs can be focused close to nm-sized metallic objects [6] thus leading to extreme light concentration in nm-sized regions, often referred to as electromagnetic hot-spots.A number of linear and nonlinear optical effects have been enhanced based on concentration of light [7,8] and potential applications of plasmonic nanostructures are anticipated within broadband plasmonic light-harvesting devices [9][10][11].The field intensity enhancement (FE) effects associated with localized SP resonances (LSPRs) are both in magnitude and spectral position strongly dependent on size, shape, material, surrounding medium and degree of aggregation of the nanoparticles.One elegant approach for tailoring the LSPRs is to utilize the configuration of a dielectric core coated with a metallic nanoshell together having extraordinary optical properties that differ considerably from those of solid Au nanoparticles including a sensitive dependence of the LSPR on the radii of the core and shell [12][13][14].Another particular and widely exploited class of field enhancing substrates which has proven to be efficient for detection of a large number of different molecular species by SERS is based on noble metal films deposited over close-packed arrays of polystyrene spheres, known as "film-on-nanosphere" substrates, often abbreviated FONs [15][16][17][18][19].A number of reports have considered the effect of tailoring the metal film thickness and roughness of FONs for the purpose of controlling the spectral position of the plasmonic resonances [15][16][17][18][19], similarly to the case of core-shell structures.Several factors should, however, be taken into account when investigating the consequence of an increasing metal film thickness, which first of all might lead to an increasing outer shell radius of curvature.As known from e.g.Mie theory on solid spheres [12], an increase in sphere diameter will cause a red-shift of the LSPR wavelength.On the other hand, for a core-shell with a fixed outer radius of curvature, increasing the shell thickness will lead to a blue-shifting LSPR.Adding to that, the fact that deposition of increasing metal thicknesses will lead to further grain nucleation, metal island formation and gradual clogging of the sharp and narrow crevices at the intersphere junctions [17], it is evident, that separating out the individual contributions to the experimentally measured reflectance is a very complicated and complex task.
Self-organizing porous alumina (Al 2 O 3 ) films on aluminum (Al) [20][21][22] can be exploited as nanostructured templates for large-area ordered [23] and randomly distributed metallic nanoparticles [1,8] or incorporated in e. g. metamaterials [24].We have recently reported a novel fabrication scheme exploiting the low adhesion between noble metals and anodized Al template surfaces by applying Epoxy resin in a stripping procedure [Fig.1] [25].Welldefined, hexagonally ordered, and highly SERS-active inter-connected hemispherical shell noble metal nanostructures are rapidly prepared by this technique [26,27] thereby providing a simple and unique fabrication procedure for large-area template-assisted production of morphologically tunable nanostructures with excellent stability.
The morphology of these stripped nanostructures resembles the morphology of the successful FONs albeit having sharper inter-sphere gaps, constituting sites for electromagnetic hot-spots.Very importantly, the stripping procedure provides a means of fabricating FON-like structures where the shell thickness can be tailored without changing the outer shell radius of curvature.Furthermore, given that the stripped shells inherit the surface roughness of the template, the shell thickness can be increased without promoting grain nucleation, island formation or changes of the inter-sphere junctions, and due to the morphology of the template, the shell thickness can be made thinner than 40 nm without compromising the film quality.Owing to these properties of the technique, the isolated effect of the different parameters on the optical properties when tuning the shell thickness of FON-like structures can be systematically studied by employing linear reflection spectroscopy [1,8] and two-photon luminescence scanning optical microscopy (TPL-SOM) [28][29][30][31].We demonstrate the possibility of tuning the LSPRs in the entire wavelength range from 535 nm to 700 nm by combining the tunability facilitated by tailoring the shell radii of curvature with the sensitive resonance dependence on shell thickness for a fixed outer shell radius.Atomic force microscopy (AFM) investigations were conducted in ambient air using a Nanowizard from JPK Instruments, or a Dual Scope DS 45-40 AFM from DME, and scanning electron microscopy (SEM) investigations were done in a Hitachi S-4800 field emission SEM.We have previously provided detailed descriptions on the procedure for preparation of Al foils and fabrication of Al/Al 2 O 3 templates by anodization [1,8,23] as well as how they can be exploited for metal stripping of inter-connected hemispherical shells with tailored shell radii of curvature [26,27] ., respectively.The 20 nm film clearly suffers from cracks in the Au film, but when thicker Au layers are deposited, the film is crackfree and a grainier Au surface forms.This does, however, not affect the smooth stripped surface, or the radius of curvature, as is evident in (g) and (h).

Linear reflection spectroscopy
The film of inter-connected Au hemispherical shells could facilitate excitation of propagating SP modes being focused in the grooves and crevices, analogous to channel plasmons in Vshaped grooves [4].Furthermore, these nanometer-sized sharp features could support prominent LSPRs in the optical range, which along with the propagating SP modes will give rise to substantial field enhancements in electromagnetic hot-spots.
The optical properties of the fabricated Au hemispherical shells were investigated by linear reflection spectroscopy [8,26]  to ~601 nm by tailoring the hemispherical shell thicknesses.From Mie theory it is wellknown that plasmon resonances of a metal shell enclosing a dielectric core can be carefully tailored by adjusting the shell thickness and radius [12].The D = 450 nm samples have a larger outer shell radii of curvature and by adjusting shell thicknesses in the same range as for the D = 270 nm samples (L* = 40-300 nm), SP resonance tuning in another extended spectral range from ~618 nm up to ~705 nm is accomplished [Fig.5(d)].Consequently, using this fabrication scheme, ordered Au-nanostructures can be customized to support plasmon resonances covering a large part of the visible spectrum, thereby allowing design of SERS active surfaces yielding enhancements at wavelengths where laser sources are available and molecules are Raman active.In the case of FONs and other FON-like nanostructures, where increasing the amount of deposited metal will lead to a simultaneous change in surface roughness, a change of the outer radius of curvature and change the sharpness of the inter-sphere junctions, the LSPR tuning will be a superposition of several spectrally counteracting contributions [17][18][19].The stripping technique presented in this paper, on the other hand, enables tuning of the shell thickness without altering any of these parameters, and consequently, the reflection measurements of samples with different shell thickness will show LSPR shifts caused by only one contribution; the shell thickness.Employing the described fabrication procedure, crack free Au hemispherical shells could not be fabricated with thicknesses less than L* = 40 nm, while there was no obvious limit in maximum thickness.Classical Mie calculations on spherical core-shell particles can be utilized to study the case where the outer shell radius of curvature is fixed and the core radius is varied systematically [Fig.5(b), (c), (e), (f)] corresponding to increasing the spherical Au shell thickness.All calculations of the extinction cross-section, Q ext , were carried out using the optical constants for Au from [32], a refractive index of the surrounding medium equal to n medium = 1, and a core refractive index equal to that of the Epoxy resin (n core = 1.519 at 589 nm).This very simple approach is used to explain the observed resonance trends (with respect to changes of film thickness and particle size), though it does not take into account the interaction between neighboring particles and the differences between the optical properties of a spherical and hemispherical shell.
To estimate the outer shell radius of curvature, r 2 , for each of the samples (D = 270 nm and D = 450 nm), AFM measurements [inset Fig. 5(b)] were carried out on the templates along the path sketched with a yellow line in Fig. 2(c).By fitting a sphere to the template hollow, the radii of curvature were estimated to r 2 ~135 nm and r 2 ~205 nm for the D = 270 nm sample and the D = 450 nm sample, respectively.
From the full Mie-calculation of the extinction for a spherical core shell with r 2 = 135 nm (corresponding to the D = 270 nm sample) and different shell thicknesses, L, it is inferred that the extinction has contributions from a broad dipole resonance centered around ~750 nm, and a comparably sharp quadrupole resonance tunable in a wavelength range similar to the measured tunable range for the D = 270 nm samples, i. e., from ~535 nm to ~601 nm [Fig.5(b)].The contribution from the isolated quadrupole moment [Fig.5(c)] shows a very strong resemblance with the measured reflection curves [Fig.5(a)], both illustrating that increasing the shell thickness beyond L~100 nm will have little or no effect on the optical properties of shells with a fixed outer diameter of r 2 = 135 nm (D = 270 nm).For relatively large nanoparticles, the scattering and absorption contributions to extinction become similar in strength with scattering being most pronounced for the dipole moment contribution.When using a high N.A. objective for illumination and collection of light in a reflection spectroscopy setup, a significant part of the light scattered by the nanostructure will be recollected by the objective.Consequently, the absorption contribution will show up more clearly in such reflection spectra [8], thereby also favoring the higher order modes, including the quadrupole mode.
Full Mie calculations of core-shells with a fixed outer diameter of r 2 = 205 nm (corresponding to the D = 450 nm sample) and different shell thicknesses, reveal a strong and broad fundamental dipole resonance beyond the visible range around λ = 1250 nm (not shown).The resonance related to the quadrupole moment, appears in the visible range and can, as was the case for the r 2 = 135 nm sample, be tuned in the same range as the stripped hemispherical shells with D = 450 nm [Fig.5(d)] with an increasing redshift as the shells are made thinner.The comparably weak octupole moment expected from full Mie calculations (not shown) does not show up in the measured reflection spectra, but may in combination with the dipole moment be responsible for the relatively broad measured reflection spectra.By plotting the shell thickness, L, as a function of the wavelength of maximum extinction for different outer shell radii of curvature [Fig.5(f)], the sensitivity of the tunability with respect to the shell thickness is clearly illustrated with a surprisingly good correspondence between experiments and calculations.Taking into account the distribution of hemispherical shell diameters on the stripped surfaces and the very simple model used to explain these trends, the   Linear reflection spectroscopy measurements of samples with different roughnesses according to Fig. 7(a-d).The spectra are normalized, to emphasize the degrading LSPR quality factor of samples with post-deposited Au compared to the smooth stripped hemispherical shell.The inset illustrates the very small red-shift probably caused by partial coverage of the inter-sphere gaps or small changes in outer hemisphere radius of curvature.

The effect of post-deposition of Au
An increased roughness attributable to small Au particles with a distribution of sizes and shapes is known to yield absorption and scattering of light over a broad spectral range.The less sharp grooves and crevices in combination with a broad absorption and scattering from the increased roughness leads to a decreasing LSPR quality factor, which is clearly illustrated by plotting the normalized 1-R/R ref spectra for each of the samples with post-deposited Au [Fig.7(e)].An increasing shell thickness of L R only induces a very small (~5 nm) red-shift of the LSPR [inset Fig. 7(e)] which can be rooted in the partial coverage of the otherwise very sharp grooves and crevices or be caused by a change in the outer hemisphere radius of curvature.However, judging from the size of this red-shift, and the fact that increasing L R beyond 25 nm does not red-shift the resonance further, the change in outer curvature radius must be almost negligible.Conclusively, tuning the shell thickness of stripped hemispherical shells facilitates precise tuning of the LSPR in a spectral range dependent on the shell radius of curvature.For the FONs, however, the unsharpened inter-sphere gaps and increasing roughness leading to a weaker LSPR and a lower quality factor, will also heavily influence this tuning.Hence, the stripped, comparably much smoother hemispherical shells investigated in this study, are expected to have attractive applications within e. g.LSPR-spectroscopy.

Two-photon luminescence SOM of stripped hemispherical shells
An excellent technique for sensitive probing of the extreme light concentration in the vicinity of nm-sized features is provided by two-photon luminescence scanning optical microscopy (TPL-SOM) in which the magnitude of the obtained TPL signal from metals depends quadratically on the local field intensity [28,29].We have previously employed TPL-SOM in systematic investigations of FE effects obtained close to individual metal nanostrips [30] as well as periodic [31] and randomly distributed metal nanoparticles [8].The TPL-SOM setup (see [8] for a detailed description) was utilized with the purpose of evaluating the magnitude and spatial (lateral) variation of the local FE across the stripped hemispherical shells with D = 450 nm, which according to the reflection measurements [Fig.5(d)] exhibit increasing resonance strength at a fixed wavelength of 725 nm as the shells become thinner.The TPL is excited by a linearly polarized, mode-locked Ti:sapphire laser (model: Tsunami, Spectra Physics) delivering pulses with ~200 fs of duration and a repetition rate of 80 MHz.The laser has a stable mode of operation at tunable excitation wavelengths in the range of λ = 720-1000 nm with a spectral line width of ~10 nm.The laser light is focused onto the sample at normal incidence using a 100 × objective having an N.A. of 0.7.The same objective is used for collection of the linear reflection and TPL photons, which are separated by a wavelength selective beamsplitter and subsequently directed through appropriate filters, and enhanced through photo multiplier tubes with the tube for TPL being connected with a photon counter.All investigations were conducted employing a step size of ~100 nm, a scan area of 6 × 6 µm 2 with a scan speed of 20 µm/s, and an integration time of 100 ms.The polarization of the analyzer was (for both linear reflection and TPL detection) parallel to the excitation polarization, and the laser power was adjusted to avoid deformation of the sample by thermal damage.Simultaneous mapping of the spatial variation of the linear reflection and TPL signals is performed pixel-wise by recording the signals at respective positions during a raster scan of the sample surface, thereby achieving a resolution at full width of halfmaximum of ~0.75 µm and ~0.45 µm for linear reflection and TPL, respectively.The linear reflection image of the sample with D = 450 nm [Fig.8(a)], measured at an excitation wavelength of 725 nm, reveals a large number of dark areas corresponding to areas on the sample with high scattering and absorption at this particular wavelength.Similarly, the matching TPL image (over 6 × 6 µm 2 ) [Fig.8(b)] is densely covered with bright spots corresponding to a high TPL signal, and further examination of the two images leads to the conclusion, that the two images are of approximately opposite contrast.I. e., the high TPL yield in certain areas correlates to the high absorption and scattering of the excitation light in the same area, which in turn indicates enhancement of the local electromagnetic field.This fact is further corroborated by investigating the two images along identical scan lines [Fig.8(c)], where a large TPL signal in most cases is accompanied by a low linear reflection, i. e., a high absorption and scattering.Estimating the magnitude of the scan-area averaged FE factor, α, is carried out according to the following relation [31]:

Summary
In

Fig. 1 .
Fig. 1.(a) After annealing and electropolishing Al foils are anodized in oxalic acid or phosphoric acid, and a porous oxide is formed.(b) Selective etching of the porous Al2O3 leaving behind hemispherical embossed hollows in the Al surface.(c) Au deposition of different thicknesses.(d) Sandwiching the samples in (c) between two pieces of microscope glass slide with Epoxy resin.(e) Mechanical stripping of the Au layers from the Al template surfaces.(f) The sample Au layers with hemispherical shells constitutes fresh substrates; ready for e. g.SERS, and the Al templates can be re-used in step (b).Shells stripped from the same type of template have the same outer radius of curvature regardless of Au thickness.

Fig. 2 .
Fig. 2. (a) SEM of hexagonally ordered oxide pores prior to selective etching.(b) Crosssectional SEM of aligned pores terminating in the bottom with a hemispherically shaped barrier oxide responsible for embossing the subjacent Al.(c) The embossed hemispherical hollows in the template formed after selective etching of the Al2O3.Protruding apexes and saddle points are marked to point out the template structures responsible for the creation of deep and sharp grooves and crevices in the Au film stripped from the template at a later stage in the fabrication.The yellow line marks the path of the AFM line scans [inset Fig. 5(b)]

Fig. 3 .
Fig. 3. SEM of 30° tilted samples with hexagonally ordered, inter-connected hemispherical Au shells (L* = 300 nm) stripped from templates with an average interpore distances of (a) D = 270 nm and b) D = 450nm.(c) The backside of an Au flap formed by cracking a sample in two.It is clearly inferred how the film is comprised of inter-connected hemispheres.(d) High-angle SEM of the sharp and well-defined grooves at the Au inter-sphere junctions.(e) 3D AFM visualization of a hexagonal cell of hemispherical shells with D = 450 nm.x:y:z scale 1:1:1.In short, pretreated 1 mm thick Al foils were anodized using either phosphoric acid or oxalic acid thereby forming vertically aligned oxide pores [Fig.2] with interpore distances of approximately D = 270 nm and D = 450 nm, respectively.The pores terminate with a hemispherical barrier oxide separating the porous oxide from the subjacent Al while embossing the Al surface with hemispherical hollows [Fig.2(b)].These hollows are exposed [Fig.1(b) and 2(c)] by selective Al 2 O 3 etching, thus completing the template fabrication.Adjacent hemispherical hollows are separated by a sharp saddle ridge connecting two apexes which form where three contiguous pores join [Fig.2(c)], and each of these nanometer-sized features will induce sharp grooves and crevices, when the inverted structures are formed later in the fabrication process.Various Au layer thicknesses, L*, (nominal thickness measured with a quartz crystal thickness monitor) were deposited on the templates by e-beam evaporation [Fig.(1c)], but due to the template morphology, and the angling of the template surface with respect to the evaporation source (deposition angle of 22.5°) the actual deposited metal layer thicknesses especially on the hollow side-walls will be significantly lower than the measured nominal thickness.When referring to the deposited metal layer thickness as measured by the thickness monitor we therefore use "L*" to clearly distinguish it from the actual hemispherical shell thickness, L. The metal covered template was sandwiched between two pieces of microscope glass slide with Epoxy resin and cured at 65 °C for at least two hours [Fig.1(d)].The metal film is mechanically stripped from the template[26] [Fig.1(e)]transferring the hemispherical Au shell structures to the cured Epoxy resin on one of the glass slides [Fig.1(f)] and leaving the template ready for another deposition-stripping cycle.As the stripped Au film will inherit an inverted version of the template surface morphology, the hemispherical shell center-to-center distance in the Au film will accordingly be equal to the interpore distance, D, on the template, i.e.D = 270 nm and D = 450 nm [Fig.3(a) and 3(b), respectively].SEM of the back side of an Au film flap released at the fractured edge of a sample cut in two halves [Fig.3(c)] confirms that the film indeed consists of inter-connected

Fig. 4 .
Fig. 4. (a)-(d) SEM images of templates with D = 450 nm where Au thicknesses of (a) L* = 20 nm, (b) L* = 40 nm, (c) L* = 60 nm, and (d) L* = 300 nm have been deposited.(e)-(h) Hemispherical shells stripped from samples in (a)-(d), respectively.The 20 nm film clearly suffers from cracks in the Au film, but when thicker Au layers are deposited, the film is crackfree and a grainier Au surface forms.This does, however, not affect the smooth stripped surface, or the radius of curvature, as is evident in (g) and (h).
using a 0.85 numerical aperture (N.A.) objective to focus light and investigate the amount of reflected light from the stripped samples (R) and a flat Au reference (R ref ) fabricated by stripping 300 nm Au from a pristine Si wafer.The reflection measurements performed on the D = 270 nm sample [Fig.5(a)] demonstrate the very attractive possibility of continuous tuning of the optical resonances in the range ~535 nm #156045 -$15.00USD Received 4 Oct 2011; revised 2 Dec 2011; accepted 9 Dec 2011; published 21 Dec 2011 (C) 2012 OSA

Fig. 5 .
Fig. 5. Linear reflection spectroscopy of Au hemispherical shells with different thicknesses stripped from templates with (a) D = 270 nm and (d) D = 450 nm.Full Mie calculations of the extinction spectra of spherical Au core-shells with fixed outer radius (r2 = 135 nm) and an ncore = 1.519 with various shell thicknesses are shown in (b) along with (c) corresponding calculations of the quadrupole moment.The inset in (b) shows AFM line scans and spheres fitted into the hemispherical hollows.In (d), the red line marks the excitation λ used during the investigations with TPL-SOM.(e) Qext calculations of spherical core-shell quadrupole moments with r2 = 205 nm.(f) The λ of maximum extinction for the quadrupole moment calculated for spherical core-shells with different r2 and shell thicknesses, L. The dashed lines in (a) and (d) mark the measured spectral positions of the resonances for stripped shells with L* = 40 nm and L* = 300 nm.The lines are linked into the other subfigures as a guide to the eye.

#
156045 -$15.00USD Received 4 Oct 2011; revised 2 Dec 2011; accepted 9 Dec 2011; published 21 Dec 2011 (C) 2012 OSA correspondence between the purely classical Mie calculations and the measured reflection spectra seems good for both the D = 270 nm and the D = 450nm sample series.

Fig. 6 .
Fig.6.After stripping smooth hemispherical Au shells with a nominal thickness L* = 300 nm and outer curvature radius of r, from the templates with hemispherical hollows, an additional Au layer is deposited to form a rougher surface with small grains and islands.Keeping in mind the formation of FON surfaces, we investigated the isolated contribution to the optical properties imparted by increased roughness, the possibly changed outer hemisphere radius of curvature and partial coverage of the inter-sphere gaps.This was done by depositing different nominal thicknesses (L R ) of Au on top of the otherwise smooth stripped hemispherical Au shells with evaporation at normal incidence [Fig.6].In order to avoid any contribution from a change in shell thickness, this investigation was carried out on shells with D = 270 nm and a nominal thickness of L* = 300 nm, where further increase in shell thickness for the stripped structures, will cause no significant additional red-shift of the LSPR [Fig.5(a) and 5(f)].Gradual increase of the deposited Au thickness up to L R = 300 nm, leads to increased formation of small Au grains and islands on top of the hemispherical shells thereby clearly contributing to a much rougher surface [Fig.7(a-d)].

Fig. 7 .
Fig. 7. SEM images recorded on stripped hemispherical shells with D = 270 nm on which different thicknesses of Au (LR) have been deposited in order to increase roughness, increase outer radius of curvature and unsharpen the gaps at the inter-sphere junctions.(a) LR = 25 nm.(b) LR = 50 nm.(c) LR = 100 nm.(d) LR = 300 nm.Colors are correlated with the graphs in (e)Linear reflection spectroscopy measurements of samples with different roughnesses according to Fig.7(a-d).The spectra are normalized, to emphasize the degrading LSPR quality factor of samples with post-deposited Au compared to the smooth stripped hemispherical shell.The inset illustrates the very small red-shift probably caused by partial coverage of the inter-sphere gaps or small changes in outer hemisphere radius of curvature.

Fig. 8 .
Fig. 8. (a) Linear reflection image obtained at a wavelength of 725 nm and a laser power of 0.5 mW on the D = 450 nm sample with shell thickness L* = 40 nm along with (b) corresponding TPL image.(c) From the simultaneously recorded cross-sectional line scans (indicated dotted lines) in the linear reflection and TPL image, it is inferred, that the two images are approximately of opposite contrast.
of time-averaged (within the integration time of 100 ms) TPL photon counts from the fabricated structures and the flat Au reference, respectively, measured with the respective average laser powers struct P and ref P focused onto a Au structured sample area of A struct and a reference area of A ref , respectively.Since, the fabricated Au nanostructures in this work cover the entire surface area corresponding to ~1 cm 2 they will indeed also fill out the entire focused spot of the exciting laser (0.8 µm diameter).Thus, the areas A ref and A struct are considered equal and are omitted in the expression for the FE factor.The FE estimated at an excitation wavelength of λ = 725 nm by measuring and averaging the TPL signal across an area of 6 × 6 µm 2 (not counting in the abnormally large signals from possible defects) on the D = 450 nm shells with different shell thicknesses [Fig.9(a)]clearly reveals an increasing average FE as the shells become thinner.This is in perfect agreement with the linear reflection measurements of the corresponding samples yielding increasingly stronger SP resonance strengths at λ = 725 nm as the shells are made thinner [Fig.5(d)].Consequently, the highest measured average FE of approximately α = 30, is obtained from the shell having a thickness of L* = 40 nm.These calculated average FE factors are considerably lower than the maximum FEs achieved at a number of local hot spots on the stripped surfaces (up to α~45).These FEs might originate from even single nanometer-sized features locally on the sample surface, thereby prompting an actual maximum FE, which is orders of magnitude larger than estimated above.This would be particularly appealing in the context of e.g.SERS, where the electromagnetic surface enhancement of the Raman signal is generally proportional to the FE factor squared[33].To further verify the tunability of the stripped hemispherical shells, TPL-SOM measurements and estimations of the FE were conducted on the (D = 450 nm, L* = 40 nm) sample applying different excitation wavelengths in the range of λ = 725-800 nm [Fig.9(b)].This result is similarly directly correlated with the strength of the resonances at the corresponding wavelengths in Fig.5(d).Even though the limited tunability of the applied laser does not allow investigations at wavelengths below 725 nm where the optimum conditions for FE could have been revealed, the measurements do still corroborate the fact that the SP resonances of the stripped hemispherical Au shells can be carefully tuned by tailoring the shell thicknesses.

Fig. 9 .
Fig. 9. (a) Scan-area averaged FEs at λ = 725 nm calculated by comparing the magnitude of the measured TPL signal from different thicknesses of hemispherical shells stripped from the D = 450 nm templates with the TPL signal from a flat Au reference using Eq.(1).(b) FEs as a function of wavelength for the (D = 450 nm, L* = 40 nm) sample.
conclusion, we have investigated interconnected hemispherical Au shells stripped from embossed templates of anodized Al, and demonstrated how SP resonances can be spectrally tuned over a wide wavelength range by combined fine-tuning of radii of shell curvature and shell thickness.The stripping procedure enables the fabrication of structures resembling FONs but with the possibility of adjusting the shell thickness without changing the outer shell radius of curvature.Furthermore, the stripped shells inherit the surface roughness of the template, and consequently, the shell thickness can be increased without promoting grain nucleation, island formation or changes of the inter-sphere junctions.Due to the morphology of the template, the shell thickness could be made thinner than 40 nm without compromising the film quality, and the isolated effect of the different morphological parameters involved when tuning the shell thickness of FON-like structures could be systematically studied.Deposition of Au on top of thick, stripped hemispherical shells revealed the isolated contribution from Au grains, and island mediated roughness, leading to a decreasing LSPR quality factor.By performing TPL-SOM on shells with different thicknesses and applying several different laser wavelengths, we confirmed the tunability of the SP resonances and directly correlated the obtained field enhancements with the resonance strengths inferred from the linear reflection spectroscopy measurements.Iterative, multiple stripping from the same template without visible degradation facilitates fast production of identical sample sets with continuous spectral LSPR tunability thereby promising various practical applications within bio-sensing and LSPR-spectroscopy as well as deployment as SERS active surfaces, yielding enhancements at wavelengths where laser sources are available and molecules are Raman active.#156045 -$15.00USD Received 4 Oct 2011; revised 2 Dec 2011; accepted 9 Dec 2011; published 21 Dec 2011 (C) 2012 OSA 2 January 2012 / Vol. 20, No. 1 / OPTICS EXPRESS 545