Two-photon luminescence microscopy of large-area gold nanostructures on templates of anodized aluminum

Using linear reflection spectroscopy and far-field two-photon luminescence (TPL) scanning optical microscopy, we characterize highly enhancing, large-area gold nanostructures formed on porous templates made by anodization of aluminum with either oxalic acid or phosphoric acid. These templates are formed by a newly developed, stepwise technique making use of protective top oxide layers facilitating continuously tunable interpore distances. The upper, porous alumina layers are subsequently removed and the remaining embossed barrier layer is used as template for the sputtered gold, where the density of gold particles covering the sample is adjusted by regulating the sputtering conditions. We observe spatially averaged field intensity enhancement (FE) factors of up to ~ 2 5.2 10 ⋅ and bright spots in the TPL-images exhibiting maximum FE factors of up to ~ 2 14 10 ⋅ which is the largest estimated FE from any hitherto examined structures with our setup. We relate this large-area massive FE to constructive interference of surface plasmon (SP) polaritons scattered from the densely packed, randomly distributed gold particles and directly correlate this particle density with the strong and broad SP resonances as well as the magnitude of the FE factors. The average FE and the position of high enhancements in the TPL-images are dictated by the excitation wavelength, and the structures could evidently serve as versatile structures facilitating practical molecular sensing. ©2010 Optical Society of America OCIS codes: (240.6680) Surface plasmons; (310.6628) Subwavelength structures, nanostructures; (260.3910) Metal optics; (290.4210) Multiple scattering; (240.4350) Nonlinear optics at surfaces; (180.5810) Scanning microscopy. References and links 1. D. J. Jeanmaire, and R. P. Van Duyne, “Surface Raman spectroelectrochemistry: Part 1. Heterocyclic, aromatic and aliphatic amines adsorbed on the anodized silver electrode,” J. Electroanal. Chem. 84(1), 1–20 (1977). 2. M. Fleischmann, P. J. Hendra, and A. J. McQuillan, “Raman spectra of pyridine adsorbed at a silver electrode,” Chem. Phys. Lett. 26(2), 163–166 (1974). 3. M. G. Albrecht, and J. A. Creighton, “Anomalously intense Raman-spectra of pyridine at a silver electrode,” J. Am. Chem. Soc. 99(15), 5215–5217 (1977). 4. K. Kneipp, Y. Wang, H. Kneipp, L. T. Perelman, I. Itzkan, R. R. Dasari, and M. S. Feld, “Single molecule detection using surface-enhanced Raman scattering (SERS),” Phys. Rev. Lett. 78(9), 1667–1670 (1997). 5. G. T. Boyd, T. Rasing, J. R. R. Leite, and Y. R. Shen, “Local-field enhancement on rough surfaces of metals, semimetals, and semiconductors with the use of optical second-harmonic generation,” Phys. Rev. B 30(2), 519– 526 (1984). 6. E. J. Sánchez, L. Novotny, and X. S. Xie, “Near-field fluorescence microscopy based on two-photon excitation with metal tips,” Phys. Rev. Lett. 82(20), 4014–4017 (1999). 7. K. Kneipp, H. Kneipp, I. Itzkan, R. R. Dasari, and M. S. Feld, “Surface-enhanced Raman scattering and biophysics,” J. Phys. Condens. Matter 14(18), R597–R624 (2002). #129458 $15.00 USD Received 2 Jun 2010; revised 19 Jul 2010; accepted 21 Jul 2010; published 27 Jul 2010 (C) 2010 OSA 2 August 2010 / Vol. 18, No. 16 / OPTICS EXPRESS 17040 8. P. J. Schuck, D. P. Fromm, A. Sundaramurthy, G. S. Kino, and W. E. Moerner, “Improving the mismatch between light and nanoscale objects with gold bowtie nanoantennas,” Phys. Rev. Lett. 94(1), 017402 (2005). 9. P. Mühlschlegel, H. J. Eisler, O. J. F. Martin, B. Hecht, and D. W. Pohl, “Resonant optical antennas,” Science 308(5728), 1607–1609 (2005). 10. A. Hohenau, J. Krenn, F. Garcia-Vidal, S. Rodrigo, L. Martin-Moreno, J. Beermann, and S. Bozhevolnyi, “Spectroscopy and nonlinear microscopy of gold nanoparticle arrays on gold films,” Phys. Rev. B 75(8), 085104 (2007). 11. K. Sarychev, and V. M. Shalaev, “Electromagnetic field fluctuations and optical nonlinearities in metaldielectric composites,” Phys. Rep. 335(6), 275–371 (2000). 12. A. Mooradian, “Photoluminescence of metals,” Phys. Rev. Lett. 22(5), 185–187 (1969). 13. G. T. Boyd, Z. H. Yu, and Y. R. Shen, “Photoinduced luminescence from the noble metals and its enhancement on roughened surfaces,” Phys. Rev. B Condens. Matter 33(12), 7923–7936 (1986). 14. M. R. Beversluis, A. Bouhelier, and L. Novotny, “Continuum generation from single gold nanostructures through near-field mediated intraband transitions,” Phys. Rev. B 68(11), 115433 (2003). 15. A. Bouhelier, M. R. Beversluis, and L. Novotny, “Characterization of nanoplasmonic structures by locally excited photoluminescence,” Appl. Phys. Lett. 83(24), 5041 (2003). 16. J. Beermann, S. M. Novikov, T. Søndergaard, A. E. Boltasseva, and S. I. Bozhevolnyi, “Two-photon mapping of localized field enhancements in thin nanostrip antennas,” Opt. Express 16(22), 17302–17309 (2008). 17. A. Hohenau, J. R. Krenn, F. Garcia-Vidal, S. G. Rodrigo, L. Martin-Moreno, J. Beermann, and S. I. Bozhevolnyi, “Comparison of finite-difference time-domain simulations and experiments on the optical properties of gold nanoparticle arrays on gold film,” J. Opt. A, Pure Appl. Opt. 9(9), S366–S371 (2007). 18. J. Beermann, I. P. Radko, A. Boltasseva, and S. I. Bozhevolnyi, “Localized field enhancements in fractal shaped periodic metal nanostructures,” Opt. Express 15(23), 15234–15241 (2007). 19. P. Nielsen, S. Hassing, O. Albrektsen, S. Foghmoes, and P. Morgen, “Fabrication of large-area self-organizing gold nanostructures with sub-10 nm gaps on a porous Al2O3 template for application as a SERS-substrate,” J. Phys. Chem. C 113(32), 14165–14171 (2009). 20. P. Nielsen, O. Albrektsen, S. Hassing, and P. Morgen, “Controlling inter-particle gaps in self-organizing gold nanoparticles on templates made by a modified hard anodization technique,” J. Phys. Chem. C 114(8), 3459– 3465 (2010). 21. A. B. F. Martinson, J. W. Elam, J. T. Hupp, and M. J. Pellin, “ZnO nanotube based dye-sensitized solar cells,” Nano Lett. 7(8), 2183–2187 (2007). 22. H. Masuda, and K. Fukuda, “Ordered metal nanohole arrays made by a two-step replication of honeycomb structures of anodic alumina,” Science 268(5216), 1466–1468 (1995). 23. H. Masuda, K. Yada, and A. Osaka, “Self-ordering of cell configuration of anodic porous alumina with large-size pores in phosphoric acid solution,” Jpn. J. Appl. Phys. 37(Part 2, No. 11A), L1340–L1342 (1998). 24. S. Ono, M. Saito, M. Ishiguro, and H. Asoh, “Controlling factor of self-ordering of anodic porous alumina,” J. Electrochem. Soc. 151(8), B473–B478 (2004). 25. W. Lee, K. Nielsch, and U. Gösele, “Self-ordering behavior of nanoporous anodic aluminum oxide (AAO) in malonic acid anodization,” Nanotechnology 18(47), 475713 (2007). 26. K. Schwirn, W. Lee, R. Hillebrand, M. Steinhart, K. Nielsch, and U. Gösele, “Self-ordered anodic aluminum oxide formed by H2SO4 hard anodization,” ACS Nano 2(2), 302–310 (2008). 27. W. Lee, R. Ji, U. Gösele, and K. Nielsch, “Fast fabrication of long-range ordered porous alumina membranes by hard anodization,” Nat. Mater. 5(9), 741–747 (2006). 28. G. Mie, “Beiträge zur Optik trüber Medien, speziell kolloidaler Metallösungen,” Ann. Phys. 330(3), 377–445 (1908). 29. C. Sönnichsen, T. Franzl, T. Wilk, G. Von Plessen, and J. Feldmann, “Plasmon resonances in large noble-metal Clusters,” N. J. Phys. 4, 931–938 (2002). 30. K. L. Kelly, E. Coronado, L. L. Zhao, and G. C. Schatz, “The optical properties of metal nanoparticles: The influence of size, shape, and dielectric environment,” J. Phys. Chem. B 107(3), 668–677 (2003). 31. S. S. Aćimović, M. P. Kreuzer, M. U. González, and R. Quidant, “Plasmon near-field coupling in metal dimers as a step toward single-molecule sensing,” ACS Nano 3(5), 1231–1237 (2009). 32. A. Pors, O. Albrektsen, S. I. Bozhevolnyi, and M. Willatzen, “The optical properties of a truncated spherical cavity embedded in gold,” Excerpt from the Proceedings of the COMSOL Conference, Milan (2009). 33. E. C. Le Ru, and P. G. Etchegoin, “Rigorous justification of the |E| enhancement factor in Surface Enhanced Raman Spectroscopy,” Chem. Phys. Lett. 423(1-3), 63–66 (2006). 34. J. Beermann, and S. I. Bozhevolnyi, “Microscopy of localized second-harmonic enhancement in random metal nanostructures,” Phys. Rev. B 69(15), 155429 (2004). 35. J. Beermann, and S. I. Bozhevolnyi, “Two-photon luminescence microscopy of field enhancement at gold nanoparticles,” Phys. Status Solidi 2(12), 3983–3987 (2005).


Introduction
Strongly localized optical fields in the vicinity of metal nanostructures have been intensively investigated due to promising applications within surface enhanced sensing with a special reference to surface enhanced Raman spectroscopy (SERS) [1][2][3] facilitating detection of few or even single molecules [4].The huge field enhancement (FE) associated with resonant interactions in the metal nanostructures involving localized and propagating surface plasmons (SPs) is both in magnitude and spectral position strongly dependent on size, shape, material, surrounding medium and degree of aggregation of the nanoparticles.Several FE studies have been performed dealing with various configurations of metal nanostructures ranging from individual pointed particles [5,6] to their pair- [7][8][9], periodic - [10] and random [11] types of ensembles.Two-photon photoluminescence (TPL) from metals was previously described [12,13] with spatially resolved TPL studies [14] and near-field imaging [15] being lately used for characterization of local field enhancement.We have previously employed two-photon luminescence scanning optical microscopy (SOM) in the systematic investigation of FE effects achieved with individual metal nanostrips [16], periodic metal nanoparticles [10,17], and fractal shaped metal nanostructures [18].
Large-area nanostructures with immense active surface sensing areas open up new avenues to meet the demand for sensors with high molecular sensitivity.Electron beam lithography (EBL) and focused ion beam (FIB) methods are other nanofabrication techniques far superior in defining the desired size, shape and pattern of nanoparticles, arbitrarily.However, as the dimensions of photonic structures continue to decrease, the extensive processing time and precision needed in EBL and FIB for the fabrication of metal nanostructures over large areas, make these techniques expensive, time-consuming, and in this respect inferior to parallel fabrication techniques.In consequence, "bottom-up", selforganizing nanotechnology techniques with comparably very low fabrication and time expenses per area have advanced as an elegant and prevalent alternative to the aforementioned "top-down" techniques.
Self-organizing porous alumina (Al 2 O 3 ) attracts increasing attention as substrates for large-area ordered or randomly distributed metallic nanoparticles [19,20] incorporated in e.g.sensors and solar cells [21].By carefully tailoring the temperature, anodization time, and anodization voltage for a given electrolyte during mild anodization (MA), the dimensions of the hexagonally ordered porous oxide can be controlled over a wide range by a two-step method [22][23][24][25].Further control of the dimensions is gained by the use of hard anodization (HA) techniques [26,27] in which the anodization voltage is considerably increased resulting in a two orders of magnitude higher current density.Especially during HA, the formation of a protective oxide layer on the aluminum (Al) surface and ramping of the anodization voltage might be necessary in order to avoid breakdown related effects [25,27], such as oxide thickening and deformation, occurring after typically one hour of anodization.
Growing µm-long pores when using phosphoric acid as anodization electrolyte is notoriously difficult due to the likely occurrence of breakdown, but by adapting an earlier published ramping procedure used for HA [20], we here report how ~15 µm long pores can be grown routinely in phosphoric acid.We utilize porous layers fabricated by either HA in oxalic acid or MA in phosphoric acid as hexagonally ordered templates for large-area gold nanostructures, which are characterized with scanning electron microscopy (SEM) and reflection spectroscopy.Correlating these measurements with the obtained TPL intensity enhancement data leads us to conclude, that the high FE factors can be attributed to large electromagnetic fields in the vicinity of the individual particles and constructive interference of SP polaritons scattered from the densely packed gold particles on the fabricated templates.

Sample preparation
The details of the preprocessing of Al sheets by electropolishing, annealing, ultrasonic cleaning as well as the applied anodization conditions and equipment for basic MA and HA with oxalic acid as electrolyte can be found in [19,20].In the present article, both oxalic acid and phosphoric acid electrolytes are utilized (step 1, Fig. 1) to provide porous Al 2 O 3 templates with interpore distances ∆ 2 of ~270 nm and ~500 nm, respectively.A selective etch (step 2) removes the Al 2 O 3 entirely and leaves behind scallop-shaped hollows separated by an interpore distance of ∆ 2 .Depending on the experimental settings during subsequent gold sputter-coating (step 3), the gold will organize with either differing degrees of small particles or as an almost smooth gold film covering the hollows.MA with phosphoric acid for periods longer than 1 h (and consequently the formation of pores longer than ~1 µm) is notoriously difficult due to the occurrence of breakdown in the electrolyte [24].Both for the MA in phosphoric acid as well as HA in oxalic acid, an ~0.4 µm thick protective porous oxide layer with an interpore distance of ∆ 1 was therefore initially formed on the Al surface by 8 min of MA in 0.3 M oxalic acid at a constant applied voltage of 40 V.The protective oxide layer works as a means of avoiding an inhomogeneous current density across the substrate surface, and thereby enables uniform film growth under HA conditions.Following this preanodization procedure, the samples either remained in the 0.3 M oxalic acid in order to obtain templates with interpore distances of ∆ 2 ~270 nm by HA or they were transferred to an anodization setup containing 0.1 M phosphoric acid where interpore distances of ∆ 2 ~500 nm could be realized by MA.  1) Annealed and electropolished Al is firstly pre-anodized in oxalic acid followed by anodization with either phosphoric acid or HA in oxalic acid, which forms an Al2O3 pore layer with interpore distances of ∆2 ~500 nm or ∆2 ~270 nm, respectively.(2) This pore layer is etched down to the pore bottom and subsequently the bottom is sputter-coated with gold under different deposition conditions in step 3.
We have previously provided experimental details concerning the fabrication of porous Al 2 O 3 by HA in oxalic acid, including the build-up of protective oxide layers, and ramping of the anodization voltage in a step-wise fashion, allowing gas bubbles accumulated in the electrolyte container to diffuse away, while the generated heat at the pore bottom is dissipated in the cooled electrolyte [20].In the present work, we use a similar fabrication method for the HA with oxalic acid, and adapt the voltage ramping technique for MA with phosphoric acid as well.In the case of MA with phosphoric acid, the voltage was ramped from 40 V to 190 V at a rate of 15 V/min and the anodization voltage could subsequently be kept constant at 190 V for prolonged time without any breakdown related effects.In this work, the templates were made by MA with phosphoric acid for 10 h.The anodization results in vertically aligned pores in the oxide, with the bottom of the pores separated from the aluminum by a thin Al 2 O 3 layer known as the barrier oxide (Fig. 1).Sub-micron length Al 2 O 3 pores will form on top of the Al already within the first few minutes depending on the applied electrolyte.By prolonging the anodization time, the length of the pores will increase and the hexagonal ordering at the bottom of the pores will extend into larger domains whereby a more homogeneous template can be fabricated.Accordingly, the pores were grown to lengths of ~30 µm and ~15 µm for the HA in oxalic acid and the MA in phosphoric acid, respectively.The SEM micrograph (inset, Fig. 1) depicts a porous double-layer where both the smaller pores (interpore distance ∆ 1 ~100 nm) formed during the pre-anodization in oxalic acid as well as larger pores (interpore distance ∆ 2 ~500 nm) in an underlying porous layer created by a subsequent anodization in phosphoric acid is seen.This final interpore distance at the bottom of the pores can be tuned in the range of 405-500 nm simply by adjusting the anodization voltage in the range of 160-195 V in the case of phosphoric acid in step 1 [23].A similar tailoring of the interpore distance can be conducted in the range of 220-380 nm by tuning the voltage between 100 and 170 V for the HA in oxalic acid.Etching of the topmost porous Al 2 O 3 layer was carried out in a mixture of 5% w/w phosphoric acid and 1.8% w/w chromic acid at 70 °C for 16 h.Owing to a very high selectivity of the etchant, the Al 2 O 3 will be removed from the substrate without modifying the underlying Al structures even for prolonged etching times.
Formation of gold particles on top of the hexagonally ordered pore bottoms was realized by using a Fine Coat Ion Sputter Model JFC-1100 from JEOL with the sample placed at an angle of either 30° or 60° between the normal of the sputter target and the sample normal.The surface morphology of the nanostructures was investigated with a JEOL JSM-6490LV SEM.From an SEM-micrograph depicting a crack in the porous oxide fabricated by a preliminary 8 min anodization in oxalic acid and a subsequent MA in phosphoric acid at 190 V for 10 h (Fig. 2) it is inferred that forming a ~0.4 µm thick protective oxide layer and employing a steadily ramped voltage until 190 V is reached, clearly has resulted in the growth of approximately 10 to 15 µm long pores in the Al 2 O 3 .The crack was made deliberately by bending a sample, with the purpose of showing a cross section of the barrier oxide and the pores running in parallel with the double arrow marking the Al 2 O 3 layer.Individual pores, that were not broken during the bending of the sample, can be tracked almost all the way to the end of the pores at the barrier oxide layer.In the case of the porous layer fabricated by HA in oxalic acid, the SEM micrographs (not shown) look very similar except for a smaller interpore distance of ~270 nm and a thicker oxide layer of ~30 µm instead of 15 µm [20].Increased magnification of the sample in the marked rectangle (upper right inset Fig. 2) reveals the shape of the barrier oxide embossing the Al/Al 2 O 3 with hexagonally ordered scallop shaped hollows, which are uncovered (Fig. 3) when the porous oxide is removed by etching (step two in Fig. 1).From the SEM micrographs, the average interpore distances are estimated to be 500 ± 30 nm and 270 ± 20 nm for the samples exposed to MA in phosphoric acid and HA in oxalic acid, respectively.At each joining of three adjacent pores a small, white apex protrudes, which is most clearly seen for the sample anodized with HA in oxalic acid (Fig. 3b).Some apexes appear particularly bright in the SEM micrograph at locations where the apexes merge and the hexagonal symmetry is broken.Gold nanostructured samples (Fig. 4) were prepared by sputter-coating the fabricated templates after removal of the porous oxide.Depending on the amount of deposited gold and the angle between the normal of the sample surface and the sputter target, randomly distributed gold nanoparticles cover the scallop-shaped hollows in the sample surface with differing particle densities and size distributions.The amount of deposited gold and the angle between the normals are indicated at the lower left, and henceforth in this article the fabricated substrates will be referenced by their sub-figure name, e.g., A, B, C, D or E. The templates used for substrate A, C and D were made by MA with phosphoric acid and do therefore have a larger underlying interpore distance as compared to the substrates shown in B and E made by HA in oxalic acid.The density of nanoparticles is lowest on substrates D and E on which the lowest amount of gold has been deposited.For these two substrates, the gold covers the underlying Al template with a relatively smooth film though to some degree in favor of the apexes and saddle points (the ridge connecting the apexes) between adjacent pores due to the angled substrate surface with respect to the sputter target.For substrate B and C, the amount of deposited gold has been increased providing the surface of C with a significantly higher density of randomly distributed gold nanoparticles compared to D and E. Due to the increased shadowing effect caused by the decreased interpore distance and oblique deposition angle, the gold has in particular organized on the apexes and saddle points between adjacent pores on B, causing small nanospikes to protrude from the surface.Sample A was tilted further with respect to the normal of the sputter target during the gold deposition.Particles with small inter-particle gaps have formed on the apexes and saddle points while the scallop-shaped hollows are densely packed with comparably smaller gold nanoparticles.In summary, the number and size of the gold nanoparticles covering the templates can be varied by regulating the amount of sputter-coated gold and angle between the sample and the sputter target, while the interpore distance can be tuned by appropriate choice of electrolyte and anodization conditions.

Linear reflection spectroscopy
The SP resonances of the fabricated structures were investigated by reflection spectroscopy using a UV-Vis-NIR lightsource DH-2000 from Ocean Optics.The light was focused onto the substrate and collected by a 0.85 numerical aperture (N.A.) objective having a ~50 µm diameter focus.Reflection spectra were recorded with an Avaspec-2048 Fiber Optic Spectrometer from Avantes.A flat gold reference was fabricated by evaporating 200 nm of gold onto a piece of Si wafer with a Cryofox 600 Explorer from Polyteknik.
The reflection spectra measured from the fabricated samples (Fig. 4) exhibit a deep and broad resonance which becomes significantly more pronounced for the samples with the highest density of randomly distributed gold nanoparticles.From the SEM micrographs it is inferred, that A has the highest density of nanoparticles followed by B and C which is in accordance with the corresponding depths in the reflection spectra, with A being the deepest.It is well known, that increasing the size of metal nanoparticles beyond the quasi-static limit will lead to a red-shift of the localized SP resonance and an increase in linewidth [28][29][30].The large width of the SP resonances for the spectra of A, B and C, we relate to inhomogeneous broadening caused by contributions from a size-distribution of particles with differing resonances.Furthermore, particle-particle SP interaction is heavily influenced by the interparticle gap-size [31].
According to Mie theory [28] the absorption will be far dominant over scattering for small gold nanoparticles with diameters less than ~0.1 µm.An N.A. of 0.85 in air will collect light within a cone angle of ~116 degrees (the doubleangle), enclosed by the lens.In consequence, light scattered into the ~65% of the 2π solid angle on the reflection side of the sample will not contribute to the dip representing the resonances in the reflection curves.Samples mainly consisting of larger particles/structures with a higher degree of scattering will therefore have a less pronounced dip in the reflection curve when using this setup.The large density of small particles primarily contributing to the absorption on A will accordingly express itself as a pronounced dip in the reflection curve.The scattering from B will be more pronounced due to the (on average) larger nanoparticles covering its surface, but this scattering will be less influential on the overall measured dip in the reflection curve.This regardless of the fact, that the local FE on B might be larger.Sample D and E both have comparably (with respect to A, B and C) lower densities of gold nanoparticles covering the surface and consequently, sample D and E exhibit much shallower dips in the reflection curves.Simulations have shown that scattering from truncated spherical gold cavities (much resembling the present scallop-shaped hollows) with diameters larger than 0.1 µm will be far more pronounced as compared to absorption while an increase in the radius of curvature for the spherical cavity will bring about a broader and spectrally red-shifted optical response [32].Hence, the contribution from the isolated gold covered hollows might (due to the applied reflection spectroscopy setup) easily be masked in the absorption from small gold nanoparticles.The global minima (555 nm and 525 nm, respectively) in the reflection curves measured on D and E are mutually shifted by ~30 nm.Due to the relatively few particles on D and E, the reflection spectra can for these samples be considered as almost the isolated contribution from the smoothly gold covered scallop-shaped hollows, where the spectral red-shift for D could be ascribed to its larger pore diameter of ~500 nm as compared to ~270 nm for E. Inhomogeneous broadening due to distribution in size of pore diameters could lead to the broad spectra.With the purpose of evaluating the magnitude and spatial dependence of the FE across the fabricated samples, TPL-SOM measurements were conducted with an experimental reflection geometry setup (Fig. 5), which has previously been described in detail [18].The TPL is excited by a linearly polarized, mode-locked Ti:sapphire laser (model: Tsunami, Spectra Physics) delivering pulses with ~200 fs of duration, a repetition rate of 80 MHz at a tunable fundamental harmonic (FH) wavelength in the range of 730-800 nm with spectral line width of ~10 nm.The laser light passes an optical isolator (OI) in order to prevent reflections into the laser cavity, and a half-wave plate (λ/2) in combination with a prism polarizer (P) is employed to adjust the polarization and power of the light focused onto the sample.The SOM setup is based on a commercial Nikon microscope applying normal incidence focusing with a Mitutoyo infinity-corrected long-working distance 100 × objective (L) and a N.A. of 0.7.A computer-controlled piezoelectric xy-translation stage moves the sample horizontally underneath the objective in steps down to 50 nm and with an accuracy of ~4 nm.The same objective is used for collection of the FH and TPL emitted in the direction of reflection, after which a wavelength selective beamsplitter (WSBS) separates the FH and TPL photons.Both FH and TPL signals are subsequently directed through appropriate filters, F1 (transmission above 600 nm and absorbing any TPL generated before the microscope) and F2 (transmission range ~350-550 nm) and enhanced through PMT's with the tube for TPL being connected with a photon counter giving typically ~20 TPL dark counts per second (cps).Twodimensional images mapping and correlating the spatial variation of the FH and TPL signals are obtained pixel-wise by recording the signals at respective positions during a raster scan of the sample surface.The FH and TPL resolution at full width of half-maximum in the images is limited by the laser focal spot size at the sample surface and are ~0.75 µm and ~0.45 µm for FH and TPL, respectively.The focused light gives rise to an average intensity of ~5 kW/cm 2 and a peak intensity of ~0.31 GW/cm 2 at the surface when conducting measurements with an average illumination power of 50 µW on the sample surface.All presented measurements were conducted employing a scan speed of 20 µm/s, a step size ~143 nm over a scan area of 10 µm × 10 µm and an integration time of 100 ms.The power was for all measurements adjusted so as to avoid deformation of the sample by thermal damage, and the polarization of the analyzer (A 1 ) was parallel to the excitation polarization.The TPL emitted at the sample will contain a broad spectrum of wavelengths and polarizations interacting with ensembles of nanostructures by multiple scattering events, which in turn will enhance the TPL, although not preserving information about the incident FH polarization.In this work, the excitation polarization and detected FH and TPL polarizations are kept constant.However, it has previously been shown [34] that changing excitation polarization, when probing randomly distributed nanostructures, gave rise to similar intensities and statistics of TPL bright spots, but being driven by different FH eigenmodes, the constructive interference takes place at other locations.
FH and corresponding TPL images (Fig. 6) recorded on substrate B with an incident power of 50 µW on the sample and an excitation wavelength of 735 nm showed the highest average TPL signal and the TPL image is densely covered with bright spots corresponding to areas with high TPL intensity.Despite of the excitation wavelength at 735 nm being relatively far from the spectral position of the global minimum in the reflection curves, it was the shortest achievable wavelength still having a stable mode from the tunable Ti:Sapphire laser, and an appreciable number of photon counts could be detected.Unfortunately, the highest achievable resolution with the experimental setup is not sufficient to resolve the hexagonal structure.
It has previously been shown, that the magnitude of the TPL signal will be strongly dependent on the density of gold nanoparticles due to the SP polaritons' scattering being stronger for high particle density samples as compared to the low-density case [34].In order to correlate the reflection measurements as well as the TPL intensity to the density of gold nanoparticles, we have ranked the samples according to the average TPL intensity (sum of all photon count values at each pixel divided by the total number of pixels), the density of gold nanoparticles, and the level of dip in the reflection curve at a wavelength of 735 nm, respectively (Tab.1).A strong correlation between the estimated nanoparticle density, the depth in the reflection curve at 735 nm, and the average TPL signal collected from the samples is established.The reverse ranking of A and B for the TPL signal could either be caused by a particularly large TPL signal caused by lightning rod effect from the nanospikes on substrate B or be related to a possibly more pronounced absorption from the small particles on A relative to the absorption from B. High density and average TPL intensity bright spots were found on images recorded on A (not shown) and C (Fig. 8) but despite the interpore distance being considerably larger compared with substrate B, the resolution is still not sufficient for the setup to image the hexagonal structure.Due to the comparably lower gold particle density on substrates D and E, they are to a much larger extent characterized by a few very bright spots in the TPL image (Fig. 7), while having a comparably lower average intensity which expresses itself in a large standard deviation, σ.Experimentally, the level of TPL enhancement can be objectively evaluated by taking into account the area and incident power producing the TPL signal [8].The intentionally flat gold film also used for the reflection measurements was here utilized as reference when carrying out TPL measurements to estimate the magnitude of the FE factor α, which has previously been carried out according to the following relation [16]: where S struct is the number of TPL photon counts from the fabricated structures measured with an average FH laser power struct P focused onto a gold structured sample area of A struct .
Similarly, S ref and ref P are the number of photon counts and the average focused laser power, respectively, when conducting the measurement on the smooth gold reference with a FH focus area of A ref .
Since, the fabricated gold nanostructures in this work cover the entire anodized surface corresponding to an area of ~5 cm 2 they will indeed also fill out the entire focused spot of the exciting laser.Thus, the areas A ref and A struct will be considered equal and are omitted in the expression for the FE factor.Using TPL counts measured on the smooth gold reference (86 cps at 8 mW of FH incident power at λ = 735 nm after subtraction of dark counts) we obtain the maximum and average FE factors listed in column 3 and 4 in Table 2.The maximum FE in the case of sample B gives: which is also (together with the enhancement for D) the highest signal from the five fabricated structures.Even the calculated maximum FE factors should, however, be considered as an average over the entire TPL focus, and the high TPL signal could thus originate from few or even single nanometer-sized particles locally on the sample surface.In that case, the actual maximum FE at these sites could be orders of magnitude larger, which would be particularly appealing with reference to, e.g., SERS, where the electromagnetic surface enhancement of the Raman signal is generally proportional to the FE factor squared [33].The calculated average FE at the excitation wavelength of 735 nm is highest on sample B ( For sample D and E, the average number of cps is only a few times larger than the number of dark counts for the system, and hence, the calculated FE will be less precise.It should be born in mind that the calculated FE factors are estimated on the basis of TPL measurements carried out on intentionally flat gold films.The magnitude of the TPL signal from a given reference is highly dependent on the roughness of the gold film and consequently, the employed fabrication procedure.Hence, a reference fabricated by evaporating gold on a microscope glass slide will be considerably rougher than a similar gold film evaporated on a Si wafer or cleaved mica and accordingly a higher TPL signal might be detected from the microscope slide.FE factors estimated on the basis of a TPL reference signal from smooth gold films on Si wafers or mica will therefore be comparably higher.A comparison between the average and maximum FE factors reveal no obvious correlation, which is made clear from sample D having a very large maximum FE and at the same time having a comparably very low average FE.We believe this to be caused by the extremely high signal originating from "hot spots" at perhaps only a single or few particles and hence, it will not be directly correlated to the density of small gold particles on the sample surfaces.Other structures previously examined with the setup have shown very high maximum FE factors (a maximum of 2 20 10 ⋅ cps at 0.02 mW of incident FH average power on a sample at 720 nm wavelength) [35].Careful attention should, however, be paid, when comparing FE factors from different articles due to the previously mentioned influence from the utilized smooth gold references, and the assumptions made regarding the size of A struct in α ≈ ⋅ in [35] is somewhat lower than the maximum FE achieved on sample B in this work.In addition to the high maximum and average FE on the high particle density samples (A, B and C), the large active sample area of the structures fabricated in the present work, compared to structures fabricated by serial topdown techniques such as EBL or FIB, facilitates practical sensing over a relatively wide spectrum of excitation wavelengths since no horizontal alignment of the laser focus and a small micrometer-sized area is required.According to the reflection measurements (Fig. 4) the strength of the plasmonic resonances will be radically reduced when exciting with FH wavelengths above 750 nm as indicated by a local maximum in the reflection curves around 800 nm.The effect of exciting the samples with optical wavelengths being increasingly off-resonance was investigated by collecting four consecutive TPL images of the same area on each sample (shown for sample C in Fig. 8).The images were recorded for excitation wavelengths of (1) 735 nm, (2) 770 nm, (3) 800 nm and repeated at (4) 735 nm, (not shown) in order to ensure, that no deterioration caused by thermal damage of the sample surface had taken place.Neither significant change in the average TPL intensity nor any absence of bright spots was observed when comparing the measurement performed with 735 nm before and after the measurements at 770 nm and 800 nm.There is a slight drift in the piezo crystal, which causes a horizontal shift of the image when the three consecutive scans are performed.The coloured ellipsoids in the TPL images mark characteristic spots making it easier to track the intensity change at a given spatial position, when the excitation wavelength is changed.The red ellipsoid in the TPL image obtained at 735 nm (Fig. 8a), contains a very bright spot at the outer left.When increasing the wavelength to 770 nm, an additional spot at the right end of the ellipsoid emerges, while the spot at the left almost disappears when increasing the wavelength to 800 nm.This change indicates the excitation of localized SP with different resonance wavelengths at spatially separated structures.The change in average FE for each sample when changing the excitation wavelength from 735 nm to 800 nm is summarized in the sixth column in Table 2 by calculating the FE reduction factor β = α 735 /α 800 .In the calculation of the reduction factor, the difference in transmission through the optical system and the increased spot size at 800 nm has been accounted for by performing measurements of the TPL intensity from the smooth gold reference also at 800 nm.The reduction in intensity at 800 nm is largest for A and then followed by B and C. From the reflection measurements (Fig. 4) it is seen, that A has the most steeply ascending curve from 735 nm towards 800 nm and a large reduction factor is indeed expected for this substrate due to excitation off-resonance.Similarly, B and C have the second and third largest β, respectively, which is in accordance with the less steep ascend for these two curves for λ > 735 nm.The size of the FE factors for sample D and E is only slightly reduced owing to the relatively smaller change in the depth in the reflection curves and hence a smaller reduction in the strength of the plasmonic resonances.Furthermore, the number of average TPL counts for D and E was only 3-6 times larger than the number of dark counts, making estimation of the reduction factor less precise compared with sample A, B and C.

Summary
In conclusion, linear reflection spectroscopy and far-field TPL-SOM, have been exploited to characterize highly enhancing, large-area gold nanostructures formed on porous templates made by anodization of Al with either oxalic acid or phosphoric acid.These templates were formed by a newly developed, stepwise technique making use of protective top oxide layers whereby the interpore distance and pore length can be tuned continuously.The upper, porous Al 2 O 3 layers were removed and the remaining embossed barrier layer was used as template for the sputtered gold, where the density of gold particles covering the sample was adjusted by regulating the sputtering conditions.We observed spatially averaged field intensity enhancement (FE) factors of up to ~2 5.2 10 ⋅ and bright spots in the TPL-images exhibiting maximum FE factors of up to ~2 14 10 ⋅ which is the largest estimated FE from any hitherto examined structures with our setup.We related this large-area massive FE to constructive interference of SP polaritons scattered from the densely packed, randomly distributed gold particles and directly correlated this particle density with the strong and broad SP resonances investigated by the reflection spectroscopy, as well as with the magnitude of the FE factors.The average FE and the position of high enhancements in the TPL-images were dictated by the excitation wavelength, and the structures could evidently serve as versatile structures facilitating practical molecular sensing.

Fig. 1 .
Fig. 1. (1) Annealed and electropolished Al is firstly pre-anodized in oxalic acid followed by anodization with either phosphoric acid or HA in oxalic acid, which forms an Al2O3 pore layer with interpore distances of ∆2 ~500 nm or ∆2 ~270 nm, respectively.(2) This pore layer is etched down to the pore bottom and subsequently the bottom is sputter-coated with gold under different deposition conditions in step 3.

Fig. 2 .
Fig.2.Angled-view SEM of a crack in the porous oxide layer made by MA in phosphoric acid, prior to the etching of the substrate in step 2. The barrier oxide (inset with increased magnification of marked rectangle) embosses the Al/Al2O3, leaving scallop shaped structures (Fig.3) when etched away.

Fig. 3 .
Fig. 3. SEM of the templates of Al/Al2O3 recorded after removal of the porous oxide by selective etching.(a) Structures made by MA in phosphoric acid with interpore distances of ~500 nm.(b) Structures made by HA in oxalic acid with interpore distances of ~270 nm.

Fig. 4 .
Fig. 4. Left: SEM micrographs of substrates with differing amounts of gold nanoparticles covering the bottom of the underlying Al/Al2O3 templates formed by sputter-coating the substrates.The fabrication specifications for each substrate are shown at the lower left.Right: Reflection spectra measured from the substrates A to E. The laser excitation wavelengths at 735 nm and 800 nm (dashed, vertical lines) which were applied in the TPL-measurements and the spectral position of the resonances at ~525 nm and ~555 nm (dashed-dotted vertical lines) for sample E and D, respectively, are indicated.

Fig. 6 .
Fig. 6.(a) FH image obtained at 735 nm excitation and 50 µW laser power on substrate B along with (b) corresponding TPL image.The substrate surface has a relatively high and homogeneous density of nanoparticles (Fig. 4) giving a high avg.TPL intensity of ~890 cps.

Table 1 .
In the left column, the fabricated substrates have been ranked according to the highest (at 50µW) average TPL intensity at 735 nm excitation wavelength.The average TPL intensity and corresponding std.dev.σ are indicated.In the center column, the substrates are ranked according to the estimated density of small particles on the substrate.At the right column the ordering is according to the largest depth in the reflection α 735 even for sample D, which shows the lowest average FE,

Fig. 7 .
Fig. 7. TPL image recorded on sample D, which has a much lower density of gold particles relative to A, B and C, and hence a lower average TPL intensity dominated by a few bright spots.The image was obtained at λ = 735 nm and an incident laser power of 100 µW.

Fig. 8 .
Fig. 8. TPL images from measurements conducted at three different wavelengths on substrate C. (a) 735 nm and 50 µW (b) 770 nm and 141 µW (c) 800 nm and 200 µW.The coloured ellipsoids mark characteristic spots in the image making it easier to track the change in TPL intensity at a given spatial position, when the excitation wavelength is changed.

Table 2 . The maximum TPL intensity collected from the five samples excited at 735 nm and the calculated maximum and average FE at 735 nm and 800 nm excitation. The calculated reduction factor β = α735/α800 is shown in the rightmost column Sample nr. Max TPL intensity at 735 nm, normalized to 50 µW [ ×
[35]ssuming, that the FE in[35]also originates from the entire FH focus, and by applying our smooth gold reference used in the present article (57 cps after subtraction of dark counts at 4 mW of FH incident power at λ = 720 nm) it is possible to compare the achieved FE.The hereby achieved maximum FE of #129458 -$15.00USD Received 2 Jun 2010; revised 19 Jul 2010; accepted 21 Jul 2010; published 27 Jul 2010 (C) 2010 OSA 2 August 2010 / Vol. 18, No. 16 / OPTICS EXPRESS 17050 [Eq.(1)].