Multifunctional Plasmonic Film for Recording Near-Field Optical Intensity

Jul 14, 2014 - film for recording optical intensity in the near field. The plasmonic ... structures as plasmonic optical tweezers and show sequesterin...
0 downloads 0 Views 419KB Size
Download PDF
Letter pubs.acs.org/NanoLett

Multifunctional Plasmonic Film for Recording Near-Field Optical Intensity Brian J. Roxworthy,† Abdul M. Bhuiya,† V. V. G. Krishna Inavalli,‡ Hao Chen,‡ and Kimani C. Toussaint, Jr.*,‡ †

Department of Electrical and Computer Engineering and ‡Department of Mechanical Science and Engineering, University of Illinois at Urbana−Champaign, Urbana, Illinois 61801, United States S Supporting Information *

ABSTRACT: We demonstrate the plasmonic equivalent of photographic film for recording optical intensity in the near field. The plasmonic structure is based on gold bowtie nanoantenna arrays fabricated on SiO2 pillars. We show that it can be employed for direct laser writing of image data or recording the polarization structure of optical vector beams. Scanning electron micrographs reveal a careful sculpting of the radius of curvature and height of the triangles composing the illuminated nanoantennas, as a result of plasmonic heating, that permits spatial tunability of the resonance response of the nanoantennas without sacrificing their geometric integrity. In contrast to other memory-dedicated approaches using Au nanorods embedded in a matrix medium, plasmonic film can be used in multiple application domains. To demonstrate this functionality, we utilize the structures as plasmonic optical tweezers and show sequestering of SiO2 microparticles into optically written channels formed between exposed sections of the film. The plasmonic film offers interesting possibilities for photonic applications including optofluidic channels “without walls,” in situ tailorable biochemical sensing assays, and near-field particle manipulation and sorting. KEYWORDS: Plasmonics, optical data storage, optical tweezers, particle sorting, plasmonic film, plasmonic optical trapping hotographic film as we know it is credited to the discoveries of the 19th-century French inventor Joseph Nicéphore.1 His pioneering work on heliography paved the way for the eventual use of light-sensitive photographic emulsion using silver halide crystals. The advent of film has had a profound impact on the way humans interact with light, from art and entertainment to health and scientific discoveries, and served as inspiration for optical data storage that is a hallmark of the “digital age”.2 Similarly, the current “nanotechnology age”, heralded by nanometer-sized devices, has given rise to radical improvements in photonic technologies. Driven chiefly by the capabilities of plasmonic and photonic crystal devices, nanophotonics has found myriad applications in biochemical sensing,3 plasmonic optical tweezers,4−6 metasurface holography,7,8 and optofluidics.9 Recently, the plasmonic properties of Au nanorods have been explored for applications in highdensity optical data storage.10,11 In particular, Zijlstra and coworkers showed that photothermal heating induced with a pulsed optical source can be used to selectively tune the aspect ratio of nanorods embedded in a polymer matrix. This method enables wavelength and polarization-sensitive data storage in three spatial dimensions with a minimum pixel size of ∼500 nm.10 This technique demonstrates that the response of a plasmonic system can be controlled locally using thermally driven morphological changes. However, the process of

P

© XXXX American Chemical Society

embedding the Au particles in a matrix effectively limits this approach to data storage applications. In this Letter, we present an approach to record the nearfield optical intensity using arrays of Au bowtie nanoantennas (BNAs) supported on SiO2 pillars. Contrary to previous methods, our scheme is applicable to not only plasmon-based data storage but a multitude of other applications. We refer to this system as “plasmonic film” and show that the localized surface plasmon resonance (LSPR) of the pillar-supported BNAs (pBNAs) can be tuned over 100 nm in the visible spectral region using subtle, photothermally induced morphological changes in the Au particles. Owing to reduced thermal conduction compared to substrate-bound devices, plasmonic film is capable of recording of the optical intensity history using low power (μW), continuous-wave (CW) illumination. Importantly, plasmonic film retains functionality for applications beyond data storage because the pBNAs can interact with a changing local environment. This novel feature is attractive for on-chip photonic applications, whereby the local plasmonic response can be tailored, for example, in biochemical sensing or plasmonic optical trapping applications. To this end, we show that the pBNAs can be used for plasmonic optical trapping of Received: May 13, 2014 Revised: July 11, 2014

A

dx.doi.org/10.1021/nl501788a | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

silica mirospheres. Remarkably, particles are preferentially trapped in optically written channels in the film and spontaneously conform to the shape of the channel, which enables passive particle sorting in an optofluidic channel “without walls”. The plasmonic film comprises arrays of pBNAs that are fabricated by a combination of electron beam lithography (EBL), electron-beam evaporation, and reactive ion etching (RIE) processes. First, a plasma-enhanced chemical vapor deposition system is used to deposit 500 nm of SiO2 on top of an SiO2 substrate that is coated with a 25-nm-thick, indium tin oxide (ITO) layer (CEC080P, Praezisions Glas & Optik GmbH, Germany). Subsequently, the BNAs are patterned via EBL using a 100-nm-thick poly(methyl methacrylate) photoresist layer on top of the deposited SiO2.4 We find that the presence of the ITO layer is crucial to the EBL patterning, although it is covered with a thick SiO2 layer. The EBL process is followed by deposition of a 5 nm Cr adhesion layer, 50 nm Au, and a 5 nm Ni protective coating prior to liftoff, which is performed by soaking the sample in acetone for 30 min. The Ni layer protects the Au during the final RIE process in which a Plasmatherm RIE system is used with 35 mTorr pressure, 90 W power, and 70 sccm CF4 flow rate to achieve an ∼40 nm/min etch rate of the SiO2. The fabricated pBNAs have 35 nm gaps with a 425 nm array spacing and pillars with a height of 500 nm. The functionality of plasmonic film is derived from photothermally induced morphological changes in the Au particles. Heating of the pBNAs is generated by focusing a 660 nm wavelength (λ), CW laser onto the nanostructures with a collar adjustable, 0.6 numerical aperture (NA) objective (Olympus LUCplanFLN 40×), which produces an approximately Gaussian intensity distribution with an e−2 focal radius of ∼670 nm; the laser is polarized along the bowtie tip-to-tip axis, which henceforth is referred to as horizontal polarization. As the input power is increased from 100 μW (the minimum power required to observe changes in the pBNAs) to 4 mW, which corresponds to dosages from 0 to 2.8 mW μm−2, the color of the exposed region visibly changes (Figure 1a). Here, ∼10 × 10 μm patches of pBNAs are exposed by scanning the focused beam over the film in square pattern. Given the 670 nm focal spot, this process exposes at most two rows of antennas at a time, which yields an effective pixel size of 850 nm. However, the intrinsic limit of the current design, the 425 nm array spacing, can be easily accessed by using a higher-NA exposure. Moreover, the minimum pixel size can be further reduced by fabricating a smaller array spacing in the EBL process. The resulting color change of the pBNAs occurs due to the modification of the geometrical parameters of the Au particles including the tip-radius of curvature and triangle height, which in turn modifies the gap size and produces nearly spherical particles at high dosage (Figure 1b). This process can be understood by examining the temperature profile (T) resulting from thermoplasmonic heating, which we calculate using a commercial software package (COMSOL Multiphysics). PBNA heating is a coupled electromagnetics-heat transfer problem governed by the following system of equations ∇ × (∇ × E) − k 02ε E = 0

(1)

∇·( −κ ∇T + ρc pT u) = q

(2)

Figure 1. Plasmonic film dosage map. (a) Optical image of the plasmonic film showing exposed areas with dosages ranging from 0 to 2.8 mW μm−2. Numbers show the dosage used and the area exposed on the film; the scale bar is 20 μm. (b) “Top” and “tilted” scanning electron micrographs (top and bottom rows, respectively) of the exposed areas with dosage values; scale bars are 200 nm.

is the material density, cp is the heat capacity, and u is the velocity distribution of the fluid (air) surrounding the pBNAs, which in this case is assumed to be zero. Electromagnetic excitation of the BNAs, described by eq 1, leads to ohmic losses that generate a heat power density given by q = 1/2Re[J·E*], where J is the current density, Re[] represents the real part, and * is the complex conjugate.12 The electromagnetic calculations are performed in a domain with dimensions of x × y × z = 425 × 425 × 1000 nm wherein periodic (Floquet) boundary conditions are placed along the x and y boundaries to simulate an infinite array and perfectly matched layers on the z boundaries prevent spurious reflections in the computational domain. This domain is embedded in a semi-infinite SiO2 substrate with dimensions of 20 × 20 × 20 μm for the heat transfer calculation with an ambient T = 20 °C condition on the lower surface and “open” boundaries elsewhere to allow heat to diffuse out of the computational domain. The resulting temperature distribution in the pBNAs is shown in Figure 2a for input powers of 100 μW and 500 μW (shown in parentheses). It is notable that the structures are

where E is the electric field, k0 = 2π/λ, ε is the materialdependent relative permittivity, κ is the thermal conductivity, ρ B

dx.doi.org/10.1021/nl501788a | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

Figure 2. Simulated pBNA heating. Temperature distribution in the pBNAs for 100 μW input power (0.1 mW μm−2) and 500 μW input power (0.4 mW μm−2). The peak temperature for the latter is shown in parentheses. (b) Temperature distribution for substrate-bound BNAs and a 500 μW input power.

heated to over 200 °C with only 100 μW input power. Such large temperatures are attributed to the fact that the Au particles are lifted off the substrate, which normally acts as a heat sink, and therefore heat conduction away from the Au is significantly reduced. This effect is evident in Figure 2b, which shows the temperature distribution of substrate-bound BNAs excited with 500 μW. Evidently the temperature rise is an order of magnitude smaller when the antennas are attached to the substrate, which signifies the importance of the pillar structure to the functionality of plasmonic film. Despite significant heating of the pBNAs, the maximum temperatures in both cases are significantly below the bulk melting temperature of Au (1067 °C),13 which suggests that no structural change will take place. However, it has been shown that surface melting in nanoscale metallic particles, which is enhanced near highly curved regions, can occur at temperatures significantly lower than the bulk melting temperature.13−15 As such, the observed increase in tip-radius of curvature from ∼15 to 20 nm and reduction of triangle height from ∼120 to 115 nm for 100 μW input power is attributed to surface melting. As the input power is increased, the Au temperature exceeds the melting point and the metal is pulled into a spherical shape by surface tension.14 The visible optical changes (Figure 1a) in the array are a direct result of geometric changes in the antennas (Figure 1b), which in turn shift LSPR peak of the pBNAs. The optical response of each ∼10 × 10 μm region of the exposed film is assessed by measuring the spectral reflectance R raw R=1− max{R raw } (3)

Figure 3. Plasmonic response of pBNAs. (a) Experimental spectral response showing the shift in the LSPR peak as a function of dosage. (b) Corresponding theoretical results. Numbers represent the dosage in mW μm−2.

sensitivity of the film and the overall large degree of tunability of the LSPR that is achievable using low input optical power. To verify the relationship between observed LSPR shifts and geometrical changes of the nanoantennas, we calculate theoretical power reflectance curves by modeling the exposed pBNAs using the observed structural changes, including radii of curvature, triangle height, or sphere radius (where applicable). Using plane wave excitation with normal incidence, the spectral location of the peak LSPR shows close correspondence with experimental data (Figure 3a). However, the experimental data exhibit a peak in the range of 600−700 nm that is not present in the calculated results. This peak can be attributed to the Rayleigh anomaly (RA), which occurs for specific combinations of array periodicity, wavelength, and angle of incidence (AOI). The RA results in the diffraction of incident light at grazing angles along the plane of the nanoantennas and produces sharp dips in reflected power.17,18 Given that we calculate the normalized reflectance using eq 3, these features appear as peaks in both theoretical and experimental spectra. Performing the reflectance calculations as the excitation AOI is swept through 0 to sin−1 (NA) ≈ 37° reveals not only the presence of this peak for AOI > 25° but also a redshift in its spectral location with increasing AOI (Supporting Information Figure S1). Experimentally, the focused supercontinuum source comprises all angles up to 37°, and therefore the measured reflectance line shape is due to an interference between the broad dipolar LSPR and an effective average of the RA over the range of angles. Additional slight deviations of the theoretical LSPR wavelengths compared to experimental values result from the fact that Au nanoparticles are represented as triangular cylinders (or spheres in the case of the pBNAs exposed with >1.4 mW μm−2), which does not fully capture the complex geometry of the exposed pBNAs. In effect, the LSPR shift records the history of the intensity exposure in analogy to the chemical changes in silver halide

where Rraw is the raw reflectance obtained by focusing a spatially coherent, supercontinuum source onto the modified regions. The supercontinuum optical source derives from a photonic crystal fiber (Femtowhite 800, NKT Photonics) pumped using a Ti:sapphire laser with 100 fs pulse length, 80 MHz pulse repetition rate, 800 nm center wavelength, and ∼200 mW average power. The source is coupled into an optical microscope (IX-81, Olympus) equipped with the 0.6 NA objective and spectra are taken using a fiber-coupled spectrometer (USB-2000+, Ocean Optics). As the input laser dosage is increased from 0 to 2.8 mW μm−2, the main LSPR shifts from 670 to 560 nm as both the bowtie tip-radii of curvature and the gap spacing increase (Figure 3a).16 The result of increased absorption in the green part of the visible spectrum is a striking visual change in the appearance of the nanoantenna array that is evident in Figure 1a. Moreover, the ∼35 nm LSPR shift from a 0.1 mW μm−2 exposure demonstrates both the high C

dx.doi.org/10.1021/nl501788a | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

crystals that form images in standard photographic film. With this perspective in mind, we use two methods to demonstrate plasmonic film as a potential optical storage medium: (1) encoding movie frames using a spatial-light modulator (SLM) and (2) using controlled microscope stage movement to encode a desired image. A schematic of the experimental setup for method 1 is shown in Figure 4a, and it consists of a

Figure 4. Exposing plasmonic film. (a) SLM-based exposure whereby a phase hologram, displaying the Fourier transform of the desired image, is projected from the SLM, by a beam-expanding telescope (T), to the back-focal plane of a 0.6-NA objective lens (OBJ). (b) Frames of the plasmonic film-based movie; the scale bar is 8 μm. (c) Scanningstage-based exposure whereby programmed motion of a microscope stage is used to write the “I” logo into the film; the scale bar is 5 μm and the elapsed time is given in the upper-right corner.

Figure 5. Polarization mapping with plasmonic film. (a) Schematic for near-field polarization mapping whereby a vector beam generator (VBG) produces either a radially (shown) or azimuthally polarized vector beam that is subsequently focused using a 0.6-NA lens onto the film; the inset shows a scanning electron micrograph of the film with a 500 nm scale bar. (b) Experimental polarization map for a focused radial (left) and azimuthal (right) vector beam with (c) corresponding theoretical polarization distributions; the scale bars are 5 μm.

reflective SLM placed in a conjugate plane of the back aperture of the 0.6-NA exposure lens by means of a beam-expanding telescope. The SLM used in this experiment is an 8-bit, phaseonly modulator (Boulder Nonlinear Systems, SN7548) operated with Blink software. Holograms that represent the Fourier transform of the final desired image are generated using the iterative Gerchberg−Saxton algorithm19,20 and displayed on the SLM. Movie frames are subsequently recorded using the 660 nm source, which forms final image at the focal plane of the microscope objective. Figure 4b shows a segment of a “walking man” movie sequence created with this approach (Supporting Information Video 1). In the second method, precise microscope stage motion is used to scan a focused Gaussian beam over the film. This approach can be used to encode specific shapes into the film, such as the letter “I” shown in Figure 4c (Supporting Information Video 2), and is also useful for forming long channels or more elaborate shapes, for instance, by using a laser-scanning galvanometer mirror.4 Given that the thermally driven recording process is governed by the polarization-dependent absorption crosssection, plasmonic film can also be used to record the transverse polarization structure of a focused optical beam in the near field. To demonstrate this capability, we expose the film using focused, radially and azimuthally polarized vector beam inputs (Figure 5a). These beams have spatially variant

polarization profiles and both produce two transverse polarization eigenmodes in the near field.21 The vector beams are generated from a linearly polarized Gaussian beam using a vector beam generator (VBG, ARCoptix). The VBG is designed to operate at a wavelength of 780 nm, which is red shifted from the peak LSPR of the BNAs. As such, 10 mW of power, derived from the Ti:sapphire laser, is used to compensate for reduced absorption at 780 nm. In both cases, the film records the spatial mode corresponding to the horizontally polarized component of the focused vector beam (Figure 5b). Close correspondence with theoretical results, which are calculated via Richards and Wolf diffraction methods,22 can be seen in Figure 5c. Here, the horizontally polarized component generates significant heat and induces morphological changes in the Au, whereas for the orthogonally polarized component, the illumination wavelength is off resonance. This condition reduces the effective dosage and mitigates the response of the film. The key advantage of plasmonic film over comparable systems is its ability to retain functionality for purposes other than data storage, owing to its nanoantenna architecture. A particularly attractive application is plasmonic optical trapping in which the spatial tunability of the Au structures can be used D

dx.doi.org/10.1021/nl501788a | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

interact with up to nine 425 nm spaced antennas simultaneously, individual particles are not trapped in the traditional sense by the collimated excitation source.24,25 However, the presence of multiple particles limits the spatial degrees of freedom of individuals within the focal spot, and therefore permits the overall cluster to be manipulated across the array.4 These observations suggest that preferential trapping and particle sorting are the result of larger local gradient forces derived from the sharp tips and smaller gaps in the unexposed pBNAs. In order to elucidate the underlying mechanism, we calculate the optical potential energy

to tailor the local potential energy landscape. This enables novel functionality compared to other nanotweezer systems, such as the formation of optofluidic channels “without walls”. To demonstrate these unique features, we first expose the film in an ambient air environment using a 1.4 mW μm−2 dosage with a focused beam to create predefined trapping channels with a width of ∼3 μm. As a result, the exposed pBNAs are shaped approximately into 85 nm diameter spheres that are separated by an 85 nm gap, whereas the unexposed pBNAs comprise triangles with sharp tips and 35 nm gaps. The array is subsequently immersed in a water-based colloidal suspension of 1-μm-diameter SiO2 particles with the pBNAs placed on the bottom of the fluid cell formed by the SiO2 substrate and a rubber gasket (Life Technologies Coverwell C-18155). The adjacent exposed/unexposed regions are then illuminated with an approximately collimated, 15-μm-diameter excitation beam obtained by focusing the 660 nm, horizontally polarized laser into the back aperture of the 0.6-NA objective. Remarkably, we observe that microparticles are preferentially trapped in the predefined channels; that is, they conform to the geometry of the unexposed pBNAs that are illuminated by the quasiuniform source. This behavior is evident in Figure 6a,b that

U (r0) =

∫r

∞

F(r )dr

0

(4)

at each relevant point r0 in the vicinity of the exposed and unexposed pBNAs (Figure 7a). The optical force data, F(r), is calculated by evaluating the integral of the Maxwell stress tensor over the surface of a 20 nm diameter SiO2 particle as a function of transverse displacement, r, from its equilibrium position.22 Evidently, the optical potential is much steeper for unexposed pBNAs, which results in optical forces up to 6× higher for a given particle displacement compared to the exposed pBNAs (Figure 7b). The corresponding trap stiffness for unexposed pBNAs, taken as a linear fit to the central portion of the force curve, is ∼2.4 fN nm−1, which is also ∼6× that of the exposed pBNAs. Therefore, a particle near the boundary between exposed and unexposed pBNAs will experience a stronger optical force due to the latter. Inasmuch as the heated pBNAs give rise to fluid convection, it is important to consider fluid flow as a potential mechanism for the observed particle dynamics. The local convection is calculated by solving eq 2 simultaneously with the Navier− Stokes’ equations describing the fluid flow ρ0 [u ·∇]u + ∇p − η∇2 u = gρ0 αΔTn ̂

(5)

with ∇·u = 0, where ρ0 is the water density, p is the pressure, η is the kinematic viscosity, α is the thermal expansion coefficient of water, ΔT is the position-dependent temperature distribution relative to the ambient, and n̂ is a unit vector along the direction opposite to gravity.12 Equations 2 and 5 are solved in the presence of two heat sources representing two patches of unexposed pBNAs. These patches surround a region of exposed pBNAs, which are not considered as a heat source because their calculated absorption cross section is