1. Introduction

Plasmonics allows for enhanced local electromagnetic fields using metal nanostructures. These enhanced local fields are naturally appealing for nonlinear optical processes. Many works have studied the nonlinear optical response of metal nanostructures fabricated both by top-down [1–8] and bottom-up [9–12] approaches. Of particular interest is the nonlinear optical response of metal nanoparticles [13,14]. For example, these may be incorporated in other media for multi-photon lithography [15]. Symmetry is an important feature in second harmonic generation (SHG), and the structural properties of metal nanoparticles are commonly altered to achieve broken symmetry with greater SHG [16] and the related property of directional enhancement of the nonlinear response [17].

Past works have studied the nonlinear optical response of silver nanoparticles, even mapping down to the single nanoparticle level [18]. Silver nanoprisms are particularly interesting because of their single-crystal structure, the low loss of silver, the sharp tips of the nanoprism, and their lack of inversion symmetry. Silver nanoprisms of less than 100 nm in size, however, do not have a plasmonic resonance at the near-infrared wavelengths of a Ti:Sapphire laser, the most common source for ultra-fast measurements. To achieve the resonance at these wavelengths requires some tuning mechanism [19–24]. This can be achieved by top-down fabrication of multi-resonant optical antenna structures. For silver nanoprisms, and other particles fabricated by bottom-up methods, we propose the plasmon hybridization approach to tune the resonance to that of the laser source.

Plasmon hybridization refers to coupling between metal nanoparticles [25–27], or nanoparticles to other metal nanostructures (such as a metal film) [28–35], to tune the optical response. For example, it has been shown that the plasmonic resonance of silver nanoparticles can be tuned by various amounts by spacing them off from a gold film with a spacer of various thicknesses [36]. The spacer layer thickness can also be tuned a posteriori by voltage controlled oxidation [37]. Here we are particularly interested in the hybridization between a metal nanoparticle and a thin metal film that supports short-range modes and gives precise tuning of the lowest order resonance [31]. A thin metal film is advantageous because it can transmit light; e.g., in applications where the film is deposited on top of a photoresist layer.

In this work, we use plasmon hybridization between colloidally synthesized silver nanoprisms and a 10 nm thick gold film to tune the plasmonic resonance around the peak wavelength of the fundamental laser source. At the peak wavelength, we obtain three orders of magnitude enhanced second harmonic generation, as compared to the far off-resonance condition of a large spacer layer, or having no metal film at all.

2. Silver nanoprisms in colloidal solution

2.1 Synthesis of silver nanoprisms

The silver nanoprisms were synthesized by the light-assisted conversion method [38,39] of silver nanospheres to nanoprisms. As a typical synthesis process, aqueous solution of AgNO₃ (0.1 mM, 100 ml), (204390, Aldrich Chemicals), and trisodium citrate (0.3 mM), (S2990, ACP Chemicals Inc.), was prepared in presence of air with a moderate stirring rate (120 rpm). Next, NaBH₄ solution (50 mM, 100 ml), (7420-1, Caledon Laboratories Ltd.), was injected to the system. Following this, Bis (p-sulfonatonphenyl) phenylphosphine dehydrate dipotassium salt (BSPP) (5 mM, 2 ml), (698539, Aldrich Chemicals), was dropped into the solution over 2 min. The system was irradiated with a 24 W halogen lamp (ser. 700, Sunnex Inc.) for 72 h, while the optical extinction of the samples from the solution was monitored as a function of time with a Cary 5 UV-VIS-NIR spectrometer.

2.2 Extinction evaluation of silver nanoprisms in aqueous solution

Figure 1(a) shows that after 17 h irradiation, the initial silver nanospheres with 400 nm plasmon resonance were converted to nanoparticles with plasmon resonance at 600 nm. Continuing irradiation for longer intervals, the extinction resonance shifted to longer wavelength, while the reaction was terminated after 72 h irradiation with final nanoprisms with 87 ± 13 nm edge length (uncertainty is standard deviation), 6 nm radius of curvature at the edges, and 12 nm thickness, and the plasmon resonance at 680 nm. The inset of Fig. 1(a) shows a scanning electron microscope (SEM) image of a typical silver nanoprism after 72 h irradiation. In this work, all hybrid structures and samples were prepared using the same solution of silver nanoprisms with a maximum extinction at 680 nm. Figures 1(b,c) show an atomic force microscope scan (Agilent 5500) in AC mode of the silver nanoprisms dispersed on a silicon substrate. The height was measured to be 12 nm (see Fig. 1(d)).

 

Fig. 1 (a) Extinction of silver nanoparticles in aqueous solution as a function of irradiation time. Inset: Scanning electron microscope image of a typical silver nanoprism after 72 h irradiation. (b,c) Atomic force microscope (AFM) characterization of silver nanoprisms spin coated on a silicon substrate. (d) The height of a typical silver nanoprism was measured to be 12 nm with the AFM.

Download Full Size | PDF

3. Fabrication of hybrid systems

Figure 2(a) shows a schematic of the hybridized system containing the silver nanoprisms, spacer layer, thin gold film and glass substrate. Commercial substrates were used consisting of a 10 nm thick gold film with a 2 nm Ti adhesion layer on glass (AU.0100.CSS, Platypus Technologies). The commercial substrate had a flat and continuous film, even for 10 nm thickness, and we could achieve similar quality only by evaporating Au films with the substrate temperature elevated to 200C or more; however, the remainder of this work is based entirely on samples prepared with the commercial substrates. The spacer layer was created by spin coating poly methyl methacrylate (950 PMMA A2, MicroChem) at 3500 rpm for 90 sec with varying anisole (296295, Sigma-Aldrich) solvent concentration to achieve varying post-baking thicknesses between 5 nm and 20 nm [40]. The PMMA on substrate was baked for 5 min at 180C. Following this, the silver nanoprisms were spin coated onto the substrate at 700 rpm for 90 sec (1 drop of the undiluted solution). The surface coverage density was 120 nanoprisms per 60 µm², assuring a good statistical average in the SHG measurements below.

 

Fig. 2 (a) Schematic of silver nanoprisms with PMMA spacer layer over a 10 nm thick Au film adhered to a glass substrate with a 2 nm Ti layer. (b) Schematic of scattering measurement setup. WLS = white light source, obj = microscope objective lens. (c) Scattering measurement for three hybrid structures with different PMMA spacer layer thicknesses (shown in legend). Green dotted line: Ti:Sapphire spectrum. (d) Scattering simulation results for silver nanoprisms for the corresponding spacer thicknesses.

Download Full Size | PDF

4. Scattering of hybrid system

4.1 Scattering experimental setup

Figure 2(b) shows a schematic of the scattering measurement setup. Halogen white light (LS-1-LL, Ocean Optics Inc.) was focused into the glass side of the sample by a 20 × microscope objective (Mitutoyo Plan Apo, Edmund Optics Inc.), allowing for waveguiding in the glass substrate. The scattered light at the surface of the sample, where the nanoprisms and metal film were located, was collected normal to the surface by an optical fiber (0.22 NA) and the spectrum was recorded using a spectrometer (QE65000, Ocean Optics Inc.).

This measurement setup allowed for recording the scattered light from the nanoprisms. The setup was aligned to ensure that minimal light is coupled into the optical fiber when there are no nanoprisms. For a glass slide only, there is negligible scattering for the well-aligned setup; however, there is still some broadband scattering observed due to the roughness of the Au film for the case of the gold-over-glass structure.

4.2 Scattering FDTD simulation

The proposed structure was simulated using a commercial finite-difference time-domain simulation package (FDTD ver. 7.5.7, Lumerical Solutions Inc.) to estimate the scattered power and local field enhancement. For accurate modeling of the structure, a mesh size of 1.5 nm was used. The simulation domain was terminated with a perfectly matched layer for minimal reflection. The complex permittivities of gold and silver were modeled using the experimental data of Johnson and Christy [41] and Palik [42], respectively. For the PMMA spacer layer, a refractive index of 1.5 was used. The silver nanoprisms were modeled with a rounded edge radius of 6 nm. The source was polarized along the axis of symmetry of the silver nanoprism in the x direction (see Figs. 5(a) and (c)). To calculate the scattered power, we employed the formalism of total field scattered field. A set of two-dimensional power monitors was used, enclosing the nanoprism. The total power exiting this closed surface was the scattered power. Also, we used a two-dimensional frequency domain field profile monitor to determine the near-field intensity distribution.

4.3 Scattering measurements

Figure 2(c) shows scattering experimental measurements for three different spacer layer thicknesses. The Ti:Sapphire laser spectrum is shown with a dotted line green line, to compare the peak position of light source and the hybridized resonances. Figure 2(d) shows the FDTD simulation results for silver nanoprisms for the corresponding spacer thicknesses. It is noted that the linewidth of the scattering spectra for the experimental spectra in Fig. 2(c) is broader than that of the numerical simulations in Fig. 2(d). This is typical due to differences in the nanoprism size variations (of the order of 10 nm), edge sharpness, and surface roughness of the Au film.

Figure 3 shows the experimental and simulation scattering peak wavelength for the hybrid plasmonic system (Ag-PMMA-Au-Ti-glass) as a function of PMMA thickness for additional thicknesses. Good agreement is seen between the measured scattering peak, and the one obtained by FDTD calculations. It is clear from Fig. 3 and Figs. 2(c,d) that we can tune the plasmonic resonance through the peak wavelength of the Ti:Sapphire laser (808 nm) by plasmon hybridization. Since the linear response gives the largest scattering at the peak wavelength of the Ti:Sapphire laser for a PMMA thickness of 10 nm, we expect that the local field enhancement, and hence the SHG, will be greatest for that thickness too.

 

Fig. 3 Peak scattering wavelength as a function of PMMA spacer layer thickness from experiment and FDTD simulations.

Download Full Size | PDF

5. Second harmonic generation measurements

5.1 SHG experimental setup

Figure 4(a) shows a schematic of the SHG detection setup. A Ti:sapphire laser was used to produce ~25 fs pulses with a center wavelength of 808 nm and a repetition rate of 90 MHz. The applied average power was tuned via neutral density (ND) filters, while the maximum average power was limited to 30 mW. The laser light was focused onto the sample with off-axis illumination using a 20 × long working distance microscope objective (Mitutoyo Plan Apo, Edmund Optics Inc.) to produce a 60 µm² spot area. The average power was measured at the same location as the sample. The reflected laser light was blocked with an off-axis metal stop, and the normal emission beam, containing the SHG signal and scattered fundamental beam, was guided to the streak camera (C5680, Hamamatsu) with a 45° cold mirror (FM04, Thorlabs Inc.), passing through a bandpass BG40 filter to remove the fundamental beam with the expense of 30% of second harmonic signal. The streak camera was operated in synchroscan (M5675, Hamamatsu) photon-counting mode with a microchannel plate gain of 6, in a manner similar to our previous work [1,3,5].

 

Fig. 4 (a) Schematic of SHG measurement setup. ND = neutral density filter, obj = microscope objective lens, BG40 = blue-green band pass filter. (b) SHG measured using spectrum analyzer in front of streak camera (not shown in (a)). (c) SHG count dependence on PMMA spacer layer thicknesses. (d) Log-log plot of SHG count vs. incident laser power with linear-fit slope of 2.17 ± 0.20. For SHG, a slope of 2 is expected.

Download Full Size | PDF

5.2 SHG measurements

Figure 4(b) shows the measured second harmonic spectrum, taken from in front of the streak camera using the fiber-probe spectrum analyzer. Figure 4(c) shows the SHG counts as a function of PMMA thickness for 30 mW incident power. The maximum value obtained was 9000 ± 240 (error from standard deviations over 10 runs at different locations on the sample) for a 10 nm PMMA thickness. By comparison, for a sample of silver nanoprisms on a glass microslide alone, the SHG was 14 ± 4. This shows that the enhancement factor of the SHG was approximately three orders of magnitude. Figure 4(d) shows a log-log plot of the SHG versus incident laser power with a slope of 2.17 ± 0.20 for the 10 nm thickness sample. Considering that a slope of 2 is expected for SHG, this is a reasonable result. The PMMA does not show any SHG alone and is expected to give negligible contribution to the SHG signal as compared to the nanoprisms. Furthermore, negligible SHG was found for the gold film alone (counts between 0 and 5).

5.3 Near-field enhancement simulation

To see if the experimentally observed 3-order of magnitude enhancement is reasonable from a theoretical point of view, we considered FDTD simulations. A two-dimensional frequency domain field profile monitor was used to determine the near-field intensity distribution. In Figs. 5(a,c) , the monitor was placed in xy plane, at z = 6 nm (at the center of the silver nanoprism). The other monitor was in the xz plane monitor at y = 0 (see Figs. 5(b,d)).

 

Fig. 5 Near field map of the electric field intensity, (a,b) for a silver nanoprism on 10 nm PMMA, 10 nm Au, 2 nm Ti, glass substrate, at the source wavelength of 808 nm in (a) xy and (b) xz planes. (c,d) shows the same distribution for a silver nanoprism on glass substrate, at the source wavelength of 808 nm in (c) xy and (d) xz planes. The scale bar is logarithmic (base 10).

Download Full Size | PDF

Figures 5(a,b) show FDTD calculations of the near-field intensity (wavelength of 808 nm) for a silver nanoprism hybridized with a 10 nm Au film, with 10 nm PMMA spacer layer. This calculation gives a maximum enhancement of 1420 × the incident intensity. By comparison, we have made a similar calculation for a silver nanoprism on glass and found that the maximum enhancement is 82 × the incident intensity (Figs. 5(c,d)). Therefore, the hybridization process has a net enhancement of 17.3 × in the field intensity, and since the SHG scales as the square, this corresponds to 300 × in the SHG generation, which is half what was seen in the experiments.

While this is reasonable agreement considering the accuracy of numerical calculations, we note that the thin metal film may actually direct the SHG light into the detector, and thereby boost the signal in the experiments [43–45]. We have not accounted for this effect in the calculations. In general, our calculations do not account for the electrodynamics of the system at the second harmonic wavelength. While there is some enhancement from gap plasmons in our structure [46–48], these are not dominant, as can be seen from the field distribution in the calculations. For thinner gaps, significant field enhancements have been reported recently, even departing from local response theories [49].

We also performed calculations for 680 nm wavelength and found an even greater near-field enhancement for the Ag nanoprism on glass (i.e. without hybridization); therefore, the hybridization does not improve the near-field enhancement on resonance. In other words, if a source is available at 680 nm, it would be better to use that source with a non-hybridized system. The main function of the hybridization then was to tune the resonance to the wavelength of the common Ti:Sapphire source.

6. Conclusion

In conclusion, we have demonstrated that plasmon hybridization of metal nanoparticles to a thin metal film is an effective method to obtain SHG enhancement of close to three orders of magnitude. The enhancement comes from fine tuning of the plasmonic resonance to coincide with the excitation source. In the future, it would be interesting to apply these findings to enhanced multi-photon lithography [15], high-harmonic generation [50] and nonlinear sensing [51].