Photothermal Heating of Plasmonic Nano-antennas: Influence on Trapped Particle Dynamics and Colloid

Transcription

1 Photothermal Heating of Plasmonic Nano-antennas: Influence on Trapped Particle Dynamics and Colloid Distribution AUTHOR NAMES: Steven Jones*, Daniel Andrén, Pawel Karpinski, and Mikael Käll AUTHOR ADDRESS: Chalmers University of Technology, Fysikgränd 3, , Göteborg, Sweden. Fabrication of samples. The following procedure was followed to fabricate the various bowtie nano-antennas used in this study. A simplified sketch of the antenna design is shown below in Figure 1. The bowtie width and height were maintained at 70 nm and 30 nm respectively, while the length was varied from 120 nm to 240 nm in 30 nm steps. The gap at the center of the antenna was designed with a nominal distance of approximately 20 nm; however, due to subsequent fabrication steps (annealing), this distance would enlarge, to compensate for this the structures were designed such that their optical properties were insensitive to gap distance. Substrates of fused silica were consecutively washed in isopropanol and acetone under sonication for five minutes, respectively. Hereafter, a plasma etching step was applied (O2 at 50W and 250mTorr for 30s) to rid the surface of possible organic contamination. Electron beam resist (PMMA A3 MCC NANO 950k) was spin-coated at 5000rpm to a thickness of 120nm and baked on a hotplate at 180 C for 5 minutes. A 20nm thick, sacrificial layer of chromium was deposited 1

2 on the surface to enable grounding of the substrate. Electron beam lithography (JEOL JBX9300FS) was used to expose the nano-antenna patterns with 100kV beam voltage and 2nA of current. After development in MIBK:IPA (1:3) and a short O2-plasma treatment (50W at 250mTorr for 5s) an adhesion layer of 1 nm of titanium followed by 30 nm of gold was deposited on the substrate. A lift-off process was performed in hot acetone and followed by 10 minutes annealing at 150 C to improve crystallinity and optical properties of the nano-antenna. Figure 1 Sketch of bowtie nano-antenna. Antenna design parameters, including width, length, height, and gap distance. After fabrication, a representative selection of fabricated structures was chosen for analysis under a scanning electron microscope (Zeiss Supra 55 - EDX). The final bowtie size characteristics are presented in Figure 2 and Table 1. In general, the length (gap) of the bowties are smaller (larger) than the designed shape due to annealing; while the widths of the bowtie antennas are systematically larger than designed due to overexposure of the resist along the long edges. 2

3 Figure 2 Measured bowtie dimensions and symmetry. a) Length, width, and gap size of the fabricated bowtie structures plotted against the design length. Dots indicate the mean values, while the colored error bars indicate the total range of values measured. Dashed lines indicate the design parameters. b) and c) symmetry of bowtie structures, showing the length and width of the top triangle plotted against the same parameter of the bottom triangle. All colors used correspond to the bowtie color convention shown in a) and Figure 2 of the main text. Table 1 Fabricated bowtie parameters. The length, width, and gap distance for each (design) length of nanoantenna was measured with a scanning electron microscope. For structure size, the mean, standard deviation (σ), and % difference (with respect to design length) are provided. Length [nm]: (design) Length - mean σ % difference Width mean σ

4 % difference Gap mean σ Methods: The following subsections are included within the methods section: Confirmation of anti-stokes thermometry; Single nanoantenna photothermal heating and thermometry; Plasmonic optical trapping; and Thermophoresis. Confirmation of anti-stokes thermometry. To verify the technique of anti-stokes thermometry on plasmonic nanostructures, a confirmation experiment was performed using a thermally controlled microscope stage (Janus ST-500-UC). This thermal stage was inserted into an upright optical microscope (Nikon Ti80 Eclipse) for a reflection style microscope geometry. The thermal stage was operated with air as the surrounding medium, and was able to vary the temperature of the sample from 300 K to 450 K. A 633 nm thermometry laser (Melles Griot LHP) was coupled into a microscope objective (Nikon CFI Plan Apo Lambda 20x) and was significantly defocused so as not to induce any heating, while simultaneously probing a large number of nanoantenna for thermometry. The anti-stokes emission from the plasmonic nanoantenna was separated from the Rayleigh scattered light via a dichroic mirror (Semrock 545/650 nm BrightLine) and short pass filters (Semrock 633 nm RazorEdge) before being recorded by a spectrometer. The spectrometer used for observing the anti-stokes signal (Andor Shamrock 303i) was fiber coupled to the microscope. The grating used consisted of 600 lines/mm at 500 nm blaze and was detected by a CCD camera (Andor idus 416) cooled to -60 C; this setup provided an overall 4

5 spectral resolution of less than 0.3 nm. Because of the large number of particles being measured in parallel in this configuration, relatively short acquisition times (<10 s) could be used. Several different gold nanostructures were measured (all designed to resonate near 633 nm), showing a good temperature extraction response for: bowties, rectangular antenna, nanorod agglomerates, and nanoapertures in thin gold films. Slight deviations from perfect agreement were consistently observed at higher stage temperatures; these deviations are consistent with the expected sample temperature when natural convection is considered. Single nanoantenna photothermal heating and thermometry. For the case of single plasmonic nanoparticle photothermal heating and thermometry, a similar setup to the one described above was used. In this case, the thermally controlled stage was replaced with a piezo-electric stage (Mad City Labs Nano-LP200) providing precise alignment capability as well as dark field imaging. A 1064 nm heating laser (Cobalt Rumba 1000) was incorporated into the setup and was coupled into the microscope objective (Nikon S Plan Fluor, 60x, 0.7NA) with a dichroic mirror (Thor Labs DMSP1000R). The 1064 nm heating laser was tightly focused with the microscope objective, while the 633 nm probe laser was slightly defocused to reduce the incident intensity, while making alignment between the heating and probe laser easier. The samples were prepared by creating a liquid cell (depth 120 microns) surrounding the plasmonic nano-antenna which was then inserted onto the microscope stage. Once the heating and probe laser were aligned (with respect to one another), they were then focused onto the nanoantenna (observed with dark field imaging, camera: Andor ixon). Precise alignment was obtained by maximizing the 1064 nm light scattered by the antenna (at low power) via adjusting the position of the antenna with the piezo-electric stage. After alignment, the signal was observed for an extended period to ensure there was no appreciable drift in the setup. 5

6 Once the setup had been aligned the heating laser was turned off, an initial recording of the anti- Stokes spectra was observed. Before further measurements, the power of the probe laser was varied, and the anti-stokes spectra recorded to ensure that negligible heating in the plasmonic structure was induced by the probe laser. Next, the heating laser was increased in power linearly and at each power the anti-stokes signal from the probe laser was recorded. Typically, these spectra required an acquisition time of 200 s to provide a suitable signal to noise ratio at large wavenumbers. Plasmonic optical trapping. To observe the ability for these plasmonic structures to enhance optical forces and act as plasmonic traps, 200 nm (diameter) fluorescent polystyrene beads (ThermoFisher Scientific R200) were introduced at low concentration to the system. These beads had the advantage that they were easily observable as single particles under excitation from a mercury lamp with a fluorescent filter cube (Nikon TRITC filter cube). During this experiment, the 1064 nm laser was focused onto a plasmonic antenna, and a nearby polystyrene bead would eventually diffuse towards the antenna at which point the 1064 nm laser was turned on and a particle could be trapped by the plasmonic structure and the particle motion within the trap recorded as a video. By analyzing the recorded videos and fitting the observed fluorescent bright spot with a Gaussian profile, an estimate for the particle central location at each image was obtained enabling particle tracking while being trapped. This allowed us to observe the degree of confinement of each trapped particle (characterized by the standard deviation of position) as a function of 1064 nm laser power, and bowtie length. The same procedure was performed on a stationary bowtie (with no fluorescent particles) under darkfield illumination, to measure the degree of motion of a stationary object compared to those of the trapped polystyrene beads; this analysis showed that the motion of a stationary particle was approximately half that of the 6

7 minimum confinement observed for a trapped polystyrene bead, indicating that the observed confinement of trapped particles is due to Brownian motion of the trapped fluorescent nanospheres in an optical potential well, and not system vibrations. It was possible to trap the 200 nm polystyrene spheres using only the focused 1064 nm laser, albeit at very high incident intensities. The trapped particle motion observed for trapping with a plasmonic antenna was significantly less than that observed for regular optical trapping without a plasmonic antenna, and the maximum power used to trap with an antenna was much less than was required to trap using only a focused laser beam; this indicates the influence of the antenna over the trapped particle and that it is indeed a plasmonic enhancement of optical forces which enable the trapping at the relatively low powers used. Thermo-plasmonically induced thermophoresis. To study thermophoretic effects in our plasmonic system a dilute solution (0.01% wt/vol) of 100 nm (diameter) fluorescent polystyrene spheres (ThermoFisher Scientific R100) was used. A wide-field fluorescence imaging geometry was used to capture a large region surrounding the nanoantenna (field of view μm 2 ). The nanoantenna was positioned using the piezoelectric stage to be in the center of the image and aligned with the 1064 nm laser. A monochromatic camera (Andor ixon) was used to quantitatively observe the fluorescent signal. The intensity of the fluorescent signal detected by the camera is a 2D projection of the 3D concentration profile where the contribution of out of plane fluorescent particles to the observed signal is determined by the numerical aperture of the imaging objective. To suppress the contribution from out of plane particle (thereby increasing the effective signal to noise ratio) a thin sample cell was constructed using large polystyrene microspheres to create a thin chamber which was filled with the diluted fluorescent particles. The chamber height was measured by adjusting the z-position of the piezo stage and observing the range over which in- 7

8 plane fluorophores could be observed; this corresponded to a cell depth of approximately 25 microns. For data acquisition, first a reference image was taken with no laser excitation of the antenna; this was used to normalize the subsequent image for the illumination profile of the mercury lamp used to excite the fluorescent particles. Then subsequent images were recorded after the 1064 nm laser was turned on; each of these signals was divided by the reference image to highlight relative changes in the observed fluorescence intensity distribution around the antenna. By monitoring the fluorescent image, it was possible to observe when the solution had reached steady-state conditions with no appreciable further changes in the overall intensity profile. The signal was then recorded for 20 images at an acquisition time of 200 ms each to get the steady state fluorescence intensity profile used in further analysis. Next, the radial average (measured with respect to the antenna location) of the fluorescence intensity distribution was calculated. This procedure was repeated for five different 1064 nm laser powers for which the steady state temperature increase had been previously measured using anti-stoke thermometry. This entire procedure was performed on five different antennas with the same nominal length of 180 nm, the data shown in Figure 6c of the main text corresponds to one of these trials. To obtain the 3D concentration profile that results in each of these observed fluorescence radial profiles, the temperature dependence of the fluorescence intensity should be accounted for. The fluorescence intensity temperature dependence was measured using a temperature-controlled stage and was observed to be relatively weak as expected. A linear fit was applied to the measured intensity values giving the following temperature dependence of the fluorescent signal of the polystyrene beads: I F = T (which has been normalized to the intensity at T = 300 K). This temperature dependence corresponds to a fluorescence intensity decrease of 8

9 approximately 2.7% for a temperature increase of ΔT = 10 K above ambient temperatures. In the subsequent analysis, the fitting is applied for radial distances from the antenna greater than 2 microns, where the temperature increase reaches a maximum value of less than 5 K. Regardless of the small impact on the fluorescence intensity of the particles in this region, this temperature dependence has been accounted for in all subsequent analysis. To account for out of plane fluorophores, the spatial light collection efficiency function (CEF) of the imaging objective must be calculated. The light CEF of microscope objectives has been studied previously in the context of single particle flow cytometry experiments 1,2. In these works, geometric optics, wave optics, and Monte-Carlo ray-tracing simulations have been shown to match well with experimental results. A modified geometric optics blur-circle approach was chosen here for its simplicity and versatility. Under this approximation, any isotropic point light source displaced from the image plane will project itself as a uniform intensity circle onto the image plane which is detected by the collection optics. The size of the projected circle that is effectively collected is determined by the distance the light source is displaced from the image plane as well as the effective numerical aperture of the collection optics. This function is summarized for the case of the collection efficiency through a rectangular aperture in the image plane by an objective with numerical aperture NA = n med sin ψ by following equation 2 : CEF(x, y, z) = v max 1 2π(1 cos ψ) dv x x 2 + (v y) 2 v min w z [ (x 2 + (v y) 2 + (w z) 2 ) 1 ] 2 w max (v) w min (v), with: 9

10 w min (v) = max [ h 2, z (x2 tan 2 ψ (v y) 2 ) 1 2], w max (v) = min [ h 2, z + (x2 tan 2 ψ (v y) 2 ) 1 2], and v min = max [ d 2, y x tan ψ], v max = min [ d 2, y x tan ψ], for a rectangular slit with width d and height h. This equation calculates the degree of overlap between the projected blur circle and the rectangular slit in the image plane while constraining the size of the blur circle such that only angles which can be accepted by the NA of the objective are included. For a lossless medium, this effectively corresponds to equal contributions from all z- planes for a cone of acceptance determined by the NA of the objective although spatial resolution decreases drastically as the light source move out of the focus plane. This approach assumes that the effective numerical aperture of the collection optics does not vary with distance from the image plane. In ref 1, it was observed experimentally that the effective numerical aperture of microscope objectives did not significantly change over distances on the order of 100 microns from the focal plane. In order to use the CEF and temperature dependence to extract the 3D concentration profile from the observed 2D images, a least squares fitting algorithm was used. This is done by assuming the concentration profile in the space studied follows a distribution given by equation 10 of the main text. The temperature profile was calculated with finite element simulations, and a general form of the Soret coefficient temperature dependence was used, given by: 3 S T (T) = S T [1 exp ( T T T 0 )], 10

11 here, S T is the high temperature limit of the Soret coefficient, T is the local temperature, T is the temperature value at which the behavior changes from thermophilic to thermophobic, and T 0 determines how strongly the Soret coefficient depends on temperature. For each measurement point in the observed radial intensity plot, a cone of acceptance defined by the numerical aperture was projected through the medium and the particle concentration in this acceptance cone was integrated using the CEF to give the value measured by the detector. This process is summarized in Figure 3. Figure 3 Image projection analysis procedure. a) Schematic overview of the image projection profile. For each point along the radial distance the concentration of particles (determined by the thermophoretic forces) is integrated over the acceptance cone according to the CEF of the objective. b) Overview of the fitting algorithm used to extrapolate the concentration profile from the observed fluorescence profile. Initially a 3D concentration profile is assumed for a test value of the Soret coefficient; next, the fluorescent intensity of each region is adjusted for the temperature dependence; the light CEF is then applied to find the corresponding 2D projection; the experimental data is then compared with the 2D projection. This process is repeated iteratively and a least squares regression is used to determine the best fitting parameters and therefore the most suitable concentration profile. The free parameters used in fitting the concentration profile to the projected fluorescence distribution are those from the Soret coefficient (above), as well as the temperature increase at the antenna is allowed to vary within the experimental uncertainty. The values of the Soret coefficient 11

12 obtained from this fitting procedure (see Figure 4) are well within the expected range (see main text), and are consistent with the value obtained for similar particle sizes from other works 4. Figure 4 Distribution of Soret coefficient values from fitting. Histogram of Soret coefficient values obtained from the fitting procedure. The range given on the x-axis is the range of values for the Soret coefficient that is generally expected for a particle with a diameter of 100 nm (see main text and ref 5 ). The y-axis has been normalized to an integrated value of 1. Simulations: All optical and thermal simulations were performed using the COMSOL 5.2a finite element method software package; specifically, the wave optics, and heat transfer packages for the optical and thermal components respectively. To create an accurate 3D representation of the bowtie antenna, the structures were modeled using the SolidWorks computer aided design suite. This enabled the modeling of realistic structures that had curved features which closely resembled those observed in the electron microscope images of the fabricated antenna. These geometries were then imported into the COMSOL framework for further analysis. 12

13 For the optical simulations, a perfectly matched layer was used on the boundaries of a spherical geometry to minimize reflections from the interface. The distance from the bowtie to the exterior boundaries was kept to a minimum of λ, although typically larger simulations were used. The mesh element size was kept to a maximum size of λ/5n med in all regions; however locally surrounding and within the bowtie a considerably smaller mesh size was used (5 nm within the antenna and immediately exterior); additionally, boundary element mesh layers were used on the surface of the antenna to ensure very small mesh elements locally where the electric fields were the largest. Thermal simulations used a spherical simulation region, with a radius ranging from 100 μm to 1 mm depending on the length scale studies. The exterior boundary temperature was kept at a fixed value of T = K; the large simulation radius was required to ensure that the boundary condition did not appreciably affect the temperature near the antenna. The mesh elements locally surrounding the antenna were kept to a maximum 5 nm to ensure accurate representation of the physical geometry, towards the boundaries the mesh elements approached 10 μm. In both thermal and electrodynamic simulations, the mesh elements were varied in size to ensure the obtained results converged to a solution that did not change appreciably with further mesh element size reduction. References: (1) Goodwin, P. M.; Ambrose, W. P.; Martin, J. C.; Keller, R. a. Spatial Dependence of the Optical Collection Efficiency in Flow Cytometry. Cytometry 1995, 21 (2), (2) Enderlein, J.; Ambrose, W. P. Optical Collection Efficiency Function in Single-Molecule Detection Experiments. Applied optics 1997, 36 (22), (3) Iacopini, S.; Piazza, R. Thermophoresis in Protein Solutions. Europhysics Letters 2003, 63 (2), (4) Würger, A. Hydrodynamic Boundary Effects on Thermophoresis of Confined Colloids. Physical Review Letters 2016, 116 (13),

14 (5) Piazza, R. Thermophoresis: Moving Particles with Thermal Gradients. Soft Matter 2008, 4 (9),