Letter pubs.acs.org/NanoLett
Scalable Method for the Fabrication and Testing of Glass-Filled, Three-Dimensionally Sculpted Extraordinary Transmission Apertures Sameer S. Walavalkar,* Pawel Latawiec, Andrew P. Homyk, and and Axel Scherer Applied Physics Department, California Institue of Technology, MC 200-36, Pasadena, California, United States S Supporting Information *
ABSTRACT: This Letter features a new, scalable fabrication method and experimental characterization of glass-filled apertures exhibiting extraordinary transmission. These apertures are fabricated with sizes, aspect ratios, shapes, and side-wall profiles previously impossible to create. The fabrication method presented utilizes top-down lithography to etch silicon nanostructures. These nanostructures are oxidized to provide a transparent template for the deposition of a plasmonic metal. Gold is deposited around these structures, reflowed, and the surface is planarized. Finally, a window is etched through the substrate to provide optical access. Among the structures created and tested are apertures with height to diameter aspect ratios of 8:1, constructed with rectangular, square, cruciform, and coupled cross sections, with tunable polarization sensitivity and displaying unique properties based on their sculpted side-wall shape. Transmission data from these aperture arrays is collected and compared to examine the role of spacing, size, and shape on their overall spectral response. The structures this Letter describes can have a variety of novel applications from the creation of new types of light sources to massively multiplexed biosensors to subdiffraction limit imaging techniques. KEYWORDS: Plasmonics, extraordinary transmission, silicon nanofabrication, biosensing, templated fabrication
T
Bethe−Bouwkamp model. A wealth of theoretical work has been conducted to explain this result and it has been shown that the enhanced transmission stems from a combination of effects related to waveguiding within apertures and the confinement of guided waves at the upper and lower metal surfaces.4−9 While several detailed theoretical investigations have been performed, the relative difficulty in finding a scalable fabrication method for these apertures has limited the experimental investigations into this phenomenon. Current methods for producing extraordinary optical transmission (EOT) apertures tend to rely on focused ion beam (FIB) milling as the primary fabrication scheme. In this approach, an appropriate metal is deposited onto a transparent substrate and a focused beam of gallium ions is used to mill away portions of the film.1,10−13 This method has been applied with great success in seveal experiments,1,14 however, the
he interaction between light and subwavelength-sized apertures has been an area of detailed study since extraordinary optical transmission (EOT) was first demonstrated 15 years ago.1 The classical theory explaining the behavior of light in subwavelength holes made in infinitely thin, perfect conductors was initially proposed by Bethe and then refined further by Bouwkamp.2,3 This theoretical approach predicted a power law scaling of the transmission through the holes with respect to ratio of hole radius to wavelength. Specifically it took the form of T(r, λ) ∝ (r, λ)4 where r is the hole radius and λ is the wavelength of interest. However, with the maturation of fabrication technology this simple, monotonic transmission relationship has been challenged. When arrays of holes with radii smaller than the wavelength of visible light are made in films of certain noble metals (e.g., silver or gold) enhancements in transmission, not predicted by theory, are observed. This augmented transmission at certain wavelengths was considered extraordinary as it produced transmission efficiencies several orders of magnitude higher than expected by the © 2013 American Chemical Society
Received: October 31, 2013 Revised: November 27, 2013 Published: December 5, 2013 311
dx.doi.org/10.1021/nl4040576 | Nano Lett. 2014, 14, 311−317
Nano Letters
Letter
Figure 1. Fabrication and testing of EOT apertures. (a) Fabrication flow: (i) Oxidation of silicon pillars to form transparent silicon dioxide templates. (ii) Sputtering of gold to provide a plasmonic material. (iii) Gold reflow causes the material on the side-walls to bead on top of the pillars. (iv) PMMA is spun on as a protective layer and the template pillars are snapped off mechanically. (v) A window is opened on the backside silicon dioxide. (vi) An optical window is cut through the wafer from the back to the front. (b) Experimental setup: White light from a halogen source is passed through a collimator, an optional polarization filter and impinges on the sample. Transmitted light is collected by an objective and measured either by a camera or a liquid nitrogen cooled grating spectromter.
deposited material limits the aspect ratio and metal film thickness of such fabricated apertures. The fabrication method developed in this paper uses the CMOS technologies refined over the past 60 years to produce and subsequently characterize EOT apertures with previously unattainable shapes, spacing, sizes, and aspect ratios opening up new experimental methods and applications. This new technique utilizes electron-beam lithography and plasma etching to sculpt silicon nanopillars that are fully oxidized to transparancy. These three-dimensional structures then serve as templates on which gold is deposited and reflowed to create novel glass filled plasmonic apertures. This newly developed fabrication method is summarized below in Figure 1a and presented in much greater detail in the Supporting Information. The fabrication of the nanopillar templates follows Walavalkar et al.19 The cross-sectional shapes of the aperture template were defined by electron-beam patterning of a layer of poly(methyl methacrylate) (PMMA). Thirty nanometers of aluminum oxide was sputtered into these shapes using an aluminum target and a reactive sputtering process with a 5:1 Ar/O2 gas composition.20 Lift-off was performed in dicholoromethane leaving the aluminum oxide hard-mask with the cross-sectional shape of the aperture. The mask pattern was transferred into the silicon using an ICP-RIE with a mixed mode, pseudo-Bosch etch chemistry.20 During this etching process, in some devices vertical sculpting of the aperture sidewalls was incorporated following the procedure found in Walavalkar et al.19 Etched structures were oxidized at 1000 °C for eight hours in a dry oxygen ambient as shown in Figure 1a,i. The high temperature and extended oxidation time ensured that there was no silicon core present in the vertical structures.21 A 2 nm titanium wetting layer was sputtered onto the structures, followed by 250 nm of gold (Figure 1a,ii). The samples were then heated in a rapid thermal annealer (RTA) to 650 °C in 45 s, and held at temperature for 7 min in a forming gas (5% H2:95% N2) atmosphere. The sample was actively cooled in the RTA, using a continuous nitrogen purge, back to
dimensions, aspect ratios, and geometries of structures fabricated with this method are limited by the spot-size and beam-width to which the milling ions can be focused. Furthermore, as a consequence of the physical sputtering of the FIB, material that is removed from the bottom of a trench is redeposited over the entire sample during milling, including, on the side-walls of the structure being milled.15 This unintended outcome makes it difficult to accurately control the dimensions of fabricated apertures and limits the height-to-diameter aspect ratio of holes drilled in metal films to less than 2:1. Recent work has shown some improvement in this area using a helium ion FIB15 but the fact remains that FIB-produced structures are neccesarily serial and would be impossible to fabricate in a scalable fashion. Lastly, and of most concern, is gallium contaimination of the metal during milling; experimental studies have shown the incorporation of milling ions with atomic fractions up to 50% of the host material up to 200 nm away in the lateral direction from a milled trench.16 It is critical to control or understand how this contamination changes the permittivity of the metal, since the surface plasmon mode is almost completely contained within 50 nm of the surface, yet there have been no detailed studies to date investigating this effect. Traditional, top-down approaches to defining apertures have also been used to varied amounts of success.17,18 However, the lack of an appropriate plasma-enhanced, gas-phase etchant of plasmonic metals such as gold or silver has meant that subtractive lithographic processes still must rely on physical sputtering to transfer e-beam patterned masks into the metal, running into the same dimensional and aspect ratio problems seen in FIB milling. Other approaches utilzing wet-etchants have found only limited success due to the isotropic nature of the etching and relatively harsh chemical properties of acidic metal etchants. “Lift- off” processing17 has provided a useful alternative to these methods but the traditional rule of thumb that the polymeric layer be three times the height of the 312
dx.doi.org/10.1021/nl4040576 | Nano Lett. 2014, 14, 311−317
Nano Letters
Letter
Figure 2. Spectral response of apertures with various diameters and spacings. (a) Data is from arrays of apertures on the same chip at 500 nm center to center spacing for a variety of aperture diameters. Note the blue shift of the primary transmission peak as the aperture size decreases. Scaling (back to front): × [1, 1, 2.5, 8, 10, 35 100 350]. (b) Data is from arrays of apertures of 170 nm diameter at five spacings between 500 and 1000 nm. Note that the overall line shape of the transmitted light changes based upon the spacing of the apertures. Inset: Camera image of a an array of 105 nm diameter apertures at 500 nm spacing.
from a flat, aperture-free, portion of the suspended gold/glass membrane and dividing by the lamp spectrum, measured without the presence of a sample. Both the background transmission and lamp spectrum were collected under conditions identical to those used to measure the transmission through the apertures. This was done to ensure that the corrected data could be used to provide an accurate absolute transmission measurement. Apertures were fabricated with a variety of shapes, spacings, and side-wall profiles. As an initial test to show the feasibility of the fabrication and testing method an array of circular apertures arrays was constructed on a single chip with a gold film 250 nm thick. The arrays featured aperture diameters from 30 to 200 nm and aperture spacings from 250 to 1000 nm. Figure 2a shows scaled, absolute transmission data for apertures of various diameters with a 500 nm center to center spacing. The inset shows a camera image of 105 nm array of apertures at 500 nm spacing. For a fixed pillar spacing, we observed a clear blue shift in peak transmission with decreasing diameter. The trend is consistent with the expected shift in surface plasmon resonance as the planar surfaces progress from a regularly “corrugated” structure to a flat metal surface as the diameter and therefore influence of the aperture “corrugation” decreases. This effect has been previously observed13 and theoretically justified in a coupled-mode context.23 The effect of aperture spacing on the shape of the measured transmission can be seen in Figure 2b. As the spacing between apertures is increased from 500 to 1000 nm the peak transmission wavelength blue shifts and the overall transmission curve changes shape. The blue shift can be explained, once again, as the tendency of the SPR to move toward that of an “uncorrugated” surface as the apertures move further apart and therefore have less effective influence on the composition of the metallic surface. For a fixed aperture spacing, the transmission line shape remains qualitatively similar between different aperture diameters, while varying aperture spacing exhibits a much greater influence on the transmission spectrum. In accordance with the coupledmode model of EOT,13,23 this indicates that the spacing of the apertures determines the SPR resonances on the upper and lower surfaces and therefore sets the overall line shape of the transmission. In addition to this effect, the characteristics of the
room temperature in 2 min. This reflow step served two purposes: the gold on the flat, glass substrate coalesces from multiple smaller grain sizes into larger polycyrstaline structures via Ostwald ripening and the gold on the sidewalls of the nanostructures wicks to the top, solidifying as large beads and leaving the sidewalls clean (Figure 1a,iii). A thin layer of PMMA is applied to protect the surface of the chip and the pillars are mechanically cleaved (Figure 1a,iv). Any remnants of the pillars are detached from the chip when the PMMA is removed in dichloromethane. A brief CF4/CHF3/Ar etch is performed to ensure that the glass and gold surface is flush. Finally, an opening is aligned and patterned on the backside oxide and an optical window is cut through the chip with XeF2 (Figure 1a,v−vi). It is important to note that at the elevated temperatures during the reflow process there can be titanium diffusion along gold grain boundaries toward a free surface.22 In the case of our fabrication, the titanium wetting layer would be pulled away from the glass pillar surface and either coalesce in the bead above the pillar (where it would be snapped off) or be intermixed with the thick gold layer. It is estimated that this effect serves to minimize the influence of the titanium as the electric field of the transmitted EOT mode is confined within a skin depth of the glass−gold surface and sees little of the bulk gold and hence little of the diffused titanium. Detailed theoretical studies, combined with spatially informative electron backscatter, energy dispersive X-ray spectroscopy (EDAX), can be performed in the future to better understand the influence of the titanium diffusion on the overall permittivity of the gold layer. For the reflow temperatures and durations used in this study, it is not expected for there to be any significant interdiffusion of the silicon dioxide layer and the gold/titanium film.22 Fabricated samples were tested in an upright, transmission microscope under broad-band illumination as depicted in Figure 1b. The input light was collimated and passed through an optional polarization filter and onto the underside of the sample. Light was collected with a 100×/.45NA objective and either sent to a camera or a grating spectrometer coupled to a liquid nitrogen-cooled CCD (Acton SP2300i). Transmission data was corrected by subtracting the background transmission 313
dx.doi.org/10.1021/nl4040576 | Nano Lett. 2014, 14, 311−317
Nano Letters
Letter
individual apertures sets the coupling between the upper and lower metallic surfaces and provides an intensity or spectral modification to the overall line shape by convolving its transmission properties with the SP resonances. Figure 3 supports this multifaceted, coupled-mode model. This figure compares the transmission curves measured for a
Figure 4. Transmission spectra of uniquely shaped and distributed apertures. Inset images show the shape of the apertures; from top to bottom they are cross-shaped (scale bar 250 nm), interspersed 4:1 and 2:1 rounded rectangles (scale bar 500 nm), and coupled pips (scale bar 250 nm).
transmission in the 700−900 nm range. It is hypothesized that this new transmission window stems from the interaperture coupling through the metal striphowever, detailed theoretical studies should be performed to verify this hypothesis. In this vein, these types of structures can serve as an important test-bed for the study of “slow-wave” strip guided modes as coupling mechanisms between the SPR excited on the upper and lower metal surfaces.24 Further investigations into the effects caused by coupling across this metal strip can be conducted by tuning the width of the metal strip. More detailed studies can also be done to understand the link between the polarization of the input light and the excitation of the symmetric and antisymmetric modes characteristic of such insulator−metal−insulator (IMI) structures. The middle curve in Figure 4 corresponds to the middle SEM image, a 500 nm array of interspersed rectangular apertures of two different aspect ratios. Essentially, this configuration combines the 2:1 aspect ratio, rounded rectangular aperture array measured in Figure 3 with a similarly spaced but orthogonally polarized array of 4:1 aspect ratio rectangular apertures. Comparing this curve to the curve measured for 2:1 rounded rectangular apertures in Figure 3 indicates that the broad resonance between 500 and 700 could most likely be decomposed into sharper transmission peaks characteristic of the 2:1 and 4:1 aspect ratio apertures. The broader transmission peaks can also stem from the perturbation of the square lattice of apertures by the introduction of the second set of apertures. There is a great deal that can be studied in this two-apertures system and the relative roles of lattice perturbation or aperture characterisistics in determining the final transmission curve can be established through further sets of experiments that exploit the polarization sensitivity of such apertures. Looking forward, such a configuration can be used to generate illuminating spots for fluorescence correlation spectroscopy or standard index contrast microscopy with illumination densities that exceed the diffraction limit. Consider a 500 nm × 500 nm “hyper-pixel” subdivided into an N × N grid of apertures that have high transmission at different input polarization or wavelengths. Consequently, an individual or a particular collection of apertures in each hyper-pixel could be used as a spatially unique subdiffraction illumination source. In
Figure 3. Comparison of aperture shapes. Data is from square, circular, and rounded rectangular apertures with identical spacing. Note that the spectral response across each curve is similar but the relative intensities at the transmission maxima are modified by the cross sectional shape of the aperture. Inset images show the shape of the apertures.
circular, rounded rectangular, and square shaped apertures scaled by overall aperture area in a 500 nm spaced square lattice. The underlying line-shape envelope is comparable between aperture shapes but the relative intensities of the specific transmission peaks are different from curve to curve. This follows from previous models and experiments13,23 that indicate that the resonantly transmitted modes through square, circular, and rectangular holes are different for individual apertures. However, the overall properties of extraordinary transmission are set by the interactions between apertures. Specifically, the hole spacing sets up SP resonances on both surfaces of the metal membrane that are linked through the evanescent coupling of the apertures. The shape, composition, and thickness of each aperture determines the degree of coupling between the SPR modes resonant on the metallic surface. We can see this effect of aperture shape on coupling between the faces of the metal membrane in Figure 3 by comparing the relative intensities of the transmitted light at 650 and 825 nm for square, circular, and rounded rectangular apertures. While the circular apertures transmit the light with roughly equal efficiency at 650 and 825 nm the square-shaped apertures do a much better job of coupling the light at 825 nm. The templated fabrication method presented in this paper also lends itself to the fabrication of novel aperture shapes. Figure 4 shows the transmission spectra of three aperture shapes that have been previously difficult or impossible to create with FIB lithography. The bottom-most curve corresponds to a one micrometer spaced array of the apertures shown in the bottom-most SEM image. These double “pips” are weakly coupled through a metal gap of approximately 60 nm. Comparing this curve to similarly sized apertures of one micrometer spacing (purple curve in Figure 2b) we can see that the addition of the second pip opens a new band of 314
dx.doi.org/10.1021/nl4040576 | Nano Lett. 2014, 14, 311−317
Nano Letters
Letter
Figure 5. Polarization dependence of aperture arrays. (a) Transmission data from 4:1 aspect ratio (long axis 200 nm), rounded rectangle (SEM inset) for variation in polarization of incident light. Note that 0° polarization is along the long axis of the rectangle. (b) Polar plot of polarization response of aperture array. The red circles track the response at 620 nm and the blue squares track the response at 480 nm. Note that the 480 nm data has been scaled by a factor of 5 to compare to the 620 nm data. (c) Demonstrating polarization color change. The image on the left shows the transmission through rectangular and round apertures shaped in the Caltech logo for polarization along the long axis of the rectangular apertures. The second image is of the same aperture array with the polarization aligned along the short axis of the rectangular apertures, “turning on” the orange flame.
this manner there would be N2 unique spatial illumination configurations for each hyper-pixel; from which, one could reconstruct spatial information about the sample beyond the diffraction limit. The final transmission curve in Figure 4 is that of a tapered, cross-shaped aperture (shown in the upper-most inset). Similar FIB fabricated structures have been investigated in the past25 and the transmission curve measured here shares similar characteristics with those previously reported. Specifically, we see multiply peaked transmission in the 450−700 nm range and a cutoff between 700 and 900 nm. Data presented in previous experiments25 for air-filled, cross-shaped apertures shows multiply peaked transmission between 600 and 800 nm and a cutoff between 800 and 1000 nm and further indicates that there is a second band of transmission present in the 1000− 1500 nm range; we ascribe the tail of increased transmission in the 900−1000 nm range in our data to be the start of such a band. In our case, however, since the apertures are glass rather than air, the spectral features that we see are blue shifted due to the lower index contrast between glass−gold versus air−gold. These cross-shaped apertures are of great interest in terms of creating unique lenses for the manipulation of both the near and far-field shape of transmitted light.26 By expanding upon the planar lens design put forth in Lin et al.26 with the fabrication method illustrated in this report, it would be possible to incorporate chirped spacing for control of focal length, asymmetric cruciform shapes to implement polarization sensitivity,7 a larger array to increase the lens size, and tapered side-walls (as explained below) to tune the spectral resonance of the lens. This templated fabrication method also facilitates the construction of apertures with polarization sensitivity. Figure 5a is a plot of the transmission measured through a set of rectangular apertures with a 4:1 length to width aspect ratio (shown in the inset) where the long axis was 200 nm. To show the polarization dependence of these structures the input light was passed through a rotatable polarization filter; zero polarization was set with the electric field of the input light along the long axis of the aperture. Maximal transmission was measured when the input electric field was polarized along
short axis of the aperture as expected from previous theoretical models and experimental results.12,23 Sinusoidal variation of intensity with respect to polarization angle can be seen at the locations of maximal transmission (e.g., 620 and 775 nm). The red circles plotted in Figure 5b show the normalized intensity variation of the transmission peak at 620 nm as a function of polarization angle. The data displays the traditional two-lobed structure of a sinusoidally varying phenomenon. In addition, the polar plot shows that there is a second type of mode present with orthogonal polarization, represented by the blue squares. As expected from previous work,23 the modes supported with electric fields polarized along the long axis would be those of short wavelengths. This is a consequence of the “TM requirement” of propagating surface plasmons in nonmagnetic materials.23 Therefore, if an SPR can propagate through the aperture with its electric field polarized along the long axis, then it must be of a wavelength short enough to be supported on the surface of the short axis of the rectangle. This data tracks the transmission at 480 nm and has been normalized and scaled by a factor of 5 for visibility. It is important to note that for both sets of modes there is some transmission of light even when the polarization is 90° out of phase; this is a consequence of the finite size of the structures as well as the coupling between out of phase polarization modes due to scattering at interfaces. Exploiting this fabrication technique to create apertures with arbitrary spacing, size, thickness, and polarization sensitivity can have a variety of applications. A simple demonstration of a variable color filter is shown in the panel c in Figure 5. The yellow field of the Caltech logo is made of 150 nm diameter apertures with 750 nm spacing while the flame is constructed of 2:1 aspect ratio rounded rectangles with a short axis of 100 nm and a spacing of 625 nm. When the polarization of the input light is aligned along the long axis of the rectangles both the field and the flame pass light-shaded yellow; when the polarization as aligned along the short axis of the rectangles the flame “turns on” to its orange color. Included in the Supporting Information is an animation showing sequential images taken while rotating the polarization filter showing the “turn on” and “turn off” of the flame under different input light 315
dx.doi.org/10.1021/nl4040576 | Nano Lett. 2014, 14, 311−317
Nano Letters
Letter
aperture on coupling one metallic surface to the other. Furthermore, the use of asymmetrically tapered apertures (as in the top curve/image) allows for the creation of conditions where traveling SPR modes are excited on one side of the metallic membrane but are not guided on the other side. In such a case, the SPRs would be guided along one surface and evanescently coupled to the opposite surface where they would be trapped at the mouth of the aperture, evanescently tailing off into the space above the aperture. Such modes could have a variety of novel applications for biosensing or for a spatially localized form of total internal reflection fluorescence (TIRF) imaging. A simple biosensing example could involve the antibody functionalization of the tips of such apertures with the intent of capturing a fluorescently tagged protein. The confinement of the light to the aperture mouth would ensure that freely floating tags would not be excited, while captured tags would land within the evanescent tail and would luminesce. To a camera monitoring this experiment the surface of the chip would appear dark as the pumping laser light is confined to the surface, as tags would bind to the antibodies apertures would begin to “light-up”. The work presented in this Letter features the scalable fabrication of EOT apertures with size, aspect ratio, shapes and side-wall profiles previously impossible to create. Utilizing a top-down fabriction method silicon was sculpted and oxidized to provide a transparent template around which gold was deposited and reflowed. Among the structures created and tested were apertures with diameters as small as 30 nm through 250 nm thick gold films, apertures with custom cross-sectional profiles, apertures with tunable polarization sensitivity and apertures with unique properties based on their sculpted sidewall shape. Transmission data from these aperture arrays was collected and compared to examine the role of spacing, size, and shape on their overall spectral response. Tapered and sculpted apertures were presented to show that the unique shape-dependent transmission properties of these structures could be used to modify the spectral response of aperture arrays. The structures presented here could have a variety of novel applications; The initial oxidation of the silicon structures could be done at a lower temperature resulting in “selfterminating” oxidation. Such structures would have a thin (
