Control of plasmonic nanoantennas by reversible metal-insulator transition

We demonstrate dynamic reversible switching of VO2 insulator-to-metal transition (IMT) locally on the scale of 15 nm or less and control of nanoantennas, observed for the first time in the near-field. Using polarization-selective near-field imaging techniques, we simultaneously monitor the IMT in VO2 and the change of plasmons on gold infrared nanoantennas. Structured nanodomains of the metallic VO2 locally and reversibly transform infrared plasmonic dipole nanoantennas to monopole nanoantennas. Fundamentally, the IMT in VO2 can be triggered on femtosecond timescale to allow ultrafast nanoscale control of optical phenomena. These unique features open up promising novel applications in active nanophotonics.

So far, the effects of the VO 2 IMT on plasmonic nanostructures have been studied only in the far-field 20 , so that understanding and active control of the near-field interaction by the VO 2 domains has been elusive. Here, we present an experimental study of nanoscale interactions of plasmonic structures with VO 2 undergoing the IMT using scattering-scanning near-field optical microscopy (s-SNOM) 21 , which images local vector near-fields with minimal perturbation, indispensable for the study of nanoplasmonic phenomena [22][23][24][25] . Image formation in s-SNOM relies on the effective polarizability of tip-sample complex, allowing image contrast that is based on local dielectric environment, which is ideal for nanoscale imaging of IMT.

Results
The model system comprised an array of identical infrared plasmonic nanoantennas: gold nanorods fabricated by e-beam lithography on a 100 nm VO 2 film grown on a [100] R Si substrate (Fig. 1a). The dimensions of the rods (~2510 nm × 232 nm × 30 nm) were selected to be near-field resonant at mid-infrared frequencies (10.7 μ m vacuum wavelength). Near-field optical images were acquired using a commercial s-SNOM system (neaspec.com). A linearly polarized CO 2 laser is focused on the tip-sample interface at an angle of 45° to the sample surface (Fig. 1b). The scattered field is detected by phase-modulation (pseudo-heterodyne) interferometry, yielding topography, amplitude, and phase images [23][24][25][26] .
First, we investigate the temperature-dependent emergence of metallic nanodomains during the IMT. The temperature was controlled by a heater, and a p-polarized (in the yz plane)) excitation laser was tuned to 10.7 μ m, with the p-polarized detection ( Fig. 1 and Methods). Figure 1 shows the third harmonic (of the tip oscillation frequency) amplitude (A 3 ) images of VO 2 and the Au nanorod array as a function of temperature. These images reveal the emergence of an anisotropic network of bright domains as the IMT progresses with increasing temperature. While the onset of these local metallic domains appears random, the stripes in fact evolve by connecting to the existing domains in crystallographically preferred directions until a quasi-uniform rutile phase emerges at high temperatures. In the s-SNOM, higher local polarizability in the sample results in stronger near-field optical contrast; hence these near-field images faithfully represent temperature-driven formation of metallic domains in the VO 2 film 19 .
The formation of stripe phases was noted soon after the discovery of VO 2 . Subsequently, a link was inferred between the stripe phase and substrate strain in high-quality single crystals and thick (250 nm) epitaxial grown by ion-beam assisted sputtering on TiO 2 [100] R surfaces 27,28 . Stripe formation is also correlated with the power of a near-IR pump laser (1.56 μ m), close to the surface-plasmon resonance of the VO 2 .
Here, the VO 2 films are thinner (100 nm), polycrystalline, and have no epitaxial relationship to the Si substrate. Nevertheless, the stripe phase appears during the transition from monoclinic to rutile, suggesting that it may occur simply because of localized, in-plane (xy) strains that develop at the film surface as individual grains of VO 2 begin to change phase, without any reference to the substrate. This possibility is further supported by the very small height of the stripes seen in the present experiment (Fig. 2).
The IMT metallic stripes appear with localized near-uniform spacing in the near-field amplitude image at an intermediate phase coexistence temperature. These are accompanied by correlated topographic modulation as clearly shown in Fig. 2. Topography ( Fig. 2a) and near-field amplitude ( Fig. 2b) images are taken at excitation laser wavelength, λ = 10.7 μ m and temperature, T = 344 K, along with the line profile sketches (Fig. 2c). The topographic variation in our case is smaller (0-3 nm) and the correlation weaker compared to what was observed on a TiO 2 substrate (0-5 nm) 27 . The correlation of the topographic line profile with the periodic stripes is due to a structural change in VO 2 during MIT, which results from the modulation of the rutile and the monoclinic axes.
To directly visualize plasmonic modes of the antennas and their interaction with the IMT of the VO 2 film, we implement in-plane polarization-selective excitation (s-excitation, i.e. polarized along the y direction) and in-plane detection (s-detection), which is referred to s/s imaging. Figure 3 shows topography, third-harmonic s/s near-field amplitudes, and phase images of the antennas on the VO 2 film for different temperatures. The four IR antennas are nominally identical (Fig. 3a), making it possible to compare the effects of VO 2 IMT on them. The amplitude images (Fig. 3b-d) show bright and dark optical contrast due to the coexisting insulating (dark) and metallic (bright) phases affecting the nanoscale dielectric environment of the antennas. The metallic phase begins to form randomly with increasing temperature. As a result, portions of the antennas are located partly on the metallic and partly on the insulating phases of VO 2 as observed in the amplitude images ( Fig. 3b-d). The amplitude images allow one to see the change of nanoscale field magnitudes and the metallic phase formation. At the same time, the near-field phase images Fig. 3e-h) are less sensitive to material contrast but allow one to follow the dipolar mode modification on each of the four antennas due to IMT. They display strong phase contrast at the rod ends. The VO 2 regions exhibit very weak phase contrast as shown in Fig. 3e-h, which is independent of excitation or detection polarizations.
At room temperature, all antennas display identically the expected pronounced dipolar phase contrast at their ends, as shown in Fig. 3e. At higher temperatures, all antennas whose one end is situated on the metallic phase turn from dipole to monopole as evident for Rod 1 (Fig. 3f-h), Rod 3 (Fig. 3f), and Rod 4 ( Fig. 3f-g). At even higher temperature, when the amplitude image shows that most of the film is in metallic phase (Fig. 3d), both dipole and monopole antenna modes of Rods 2, 3 and 4 turn off (Fig. 3h) completely. An interesting case is Rod 2: despite the middle part of the rod sitting on the metallic phase ( Fig. 3c), it still retains its dipole characteristics (Fig. 3g) since both ends are on the insulating grains. It only turns off at higher temperature when the entire antenna is situated on metal (Fig. 3h). These results are interpreted in schematics shown in Fig. 3i-l. This interpretation is supported by numerical calculations performed using the finite difference time-domain (FDTD) simulations (Lumerical Inc.,) shown in Fig. 4, which are in excellent qualitative agreement with experiment. Schematic of the s-SNOM experimental setup, which allows polarization-controlled simultaneous imaging of IR plasmonic antenna modes together with the phase spatial evolution of VO 2 phase transition in amplitude and phase. The coordinate system is positioned so that the y axis is directed along nanorods and the z-axis is normal to the plane of the nanostructure. (c) Temperature-dependent nearfield amplitude images reveal IMT via forming initial metallic nanodomains, which grow and connect to stripes and quasi-uniform metallic phase.
Further tracking of active dipole-to-monopole transformation of plasmonic antennas can be performed using s-excitation and p-detection (s/p) cross-polarization selective imaging see Fig. 5. Panel a displays topography, third harmonic optical near-field amplitude and phase images of an Au antenna on the VO 2 substrate at room temperature. The amplitude image displays a stronger optical contrast at rod ends and the phase image shows π phase difference between the rod ends. The amplitude and phase images both exhibit the signature of a dipolar mode of a plasmonic rod expected from S/P cross-polarized excitation/detection experimental method. Figure 5b shows amplitude contrast of the Au dipolar mode simultaneously with the metallic domain at the onset of phase transition at T = 341 K. Here, the amplitude optical contrast at rod ends is masked by the bright metallic domain contrast of the VO 2 film, and is not clearly distinguishable in the amplitude image. In contrast, the phase image (Fig. 5c) distinctly discriminates the Au plasmonic rod from the metallic background of VO 2 film. As temperature increases (T = 344 K), the metallic phase grows and a portion of one side of the rod sits on the metal and the other side sits on the insulator. The phase image between the rod ends indicates dipole (at T = 296 K and 341 K) to monopole (at T = 344 K) to off (at T = 348 K) transformation of the nanoantenna.
In summary, we have shown the first experimental evidence that near-field local optical processes in plasmonic nanostructures can be directly and actively controlled by nanodomains in VO 2 film as it undergoes the IMT. Depending on the precise location of the nanoantennas with respect to metallic and insulating domains in the VO 2 film on the scale of 15 nm or less, the IMT reversibly transforms infrared plasmonic dipole antennas to monopole antennas or switches them off. We envision that such dynamic active control of the nanoscale interaction of light with nanostructured materials, which can potentially be ultrafast, will open up diverse applications in nanooptics and the related technologies.

Methods
Near-field microscopy. The microscope is a commercial s-SNOM system (neaspec.com), which has been described in detail elsewhere. A probing s or p linearly polarized CO 2 laser is focused on the tip-sample interface at an angle of 45° from the sample surface. The scattered field is acquired using a phase modulation, or pseudoheterodyne interferometry. The background signal is suppressed by vertical tip oscillation at the mechanical resonance frequency of the cantilever (f 0 ~ 285 kHz) and demodulation of the detector signal at higher harmonics nf 0 (commonly n = 2, 3) of the tip resonance frequency. The combined scattered field from the tip and the reference beam pass through a linear polarizer which further selects the p/s polarization of the measured signal for analysis.       After development of the resist, 50 nm of gold was deposited by thermal evaporation; the remaining resist was removed by lift-off in warm acetone.

Numerical calculations.
Experimental results are theoretically interpreted with the aid of finite difference time-domain (FDTD) simulations (Lumerical Inc., lumerical.com). For all simulations each Au rod has dimensions l = 2512 nm, w = 232 nm and h = 30 nm. These dimensions were averaged from topography scan measurements. Each particle is simulated atop a VO 2 substrate. The optical excitation source used for the simulation is a mid infrared (10.7 μ m) plane wave. The simulation is performed by assuming a uniform VO 2 film with complex dielectric constant ε = 4.9 for the monoclinic insulating phase and ε = − 35 + 119i for the rutile metallic phase at λ = 10.7 m 30 . The simulation is performed using both s/s and s/p polarization selective excitation-detection methods.