Optical routing and sensing with nanowire assemblies

Results and Discussion

Figs. 1 and 2 document several experiments that were performed with a single ribbon in various combinations with GaN and ZnO nanowires. Fig. 1a shows a GaN nanowire (130 nm in diameter × 65 μm in length) that was coupled to the SnO₂ nanoribbon (240 nm × 260 nm × 460 μm) with the micromanipulator. A magnified scanning electron microscopic view of the GaN–SnO₂ junction (Fig. 1c, Inset) indicates that the two structures are in physical contact over a distance of ≈2 μm. This staggered-bonded configuration provides good optical coupling between the cavities and some degree of interwire adhesion (via electrostatic forces), which aids in the construction of multiwire networks. Typical power transfer efficiencies of ≈50% were measured for nanoribbons of similar size linked in a staggered geometry. These measurements also represent an upper bound on the coupling efficiency between ZnO or GaN wires and SnO₂ ribbons, which should show slightly lower efficiencies because of the mismatch in refractive index, size, and cross-sectional shape between wires and ribbons. We found that this staggered geometry did not cause appreciable scattering of light at the ribbonribbon interface. Therefore, the fraction of the guided power not transferred from wire/ribbon to ribbon is lost at the end facets of the primary waveguide. We have also tested several alternative coupling geometries. Butt-end coupling is less effective than staggered coupling because the emission from these subwavelength waveguides is poorly collimated. If two ribbons are crossed instead of staggered, the coupling losses decrease with shallower intersection angles, which has also been observed recently for crossed CdS nanowires (7). It is also possible to detect the transfer of light between ribbons that are weakly linked across an air gap of up to several hundred nanometers (data not shown).

To demonstrate the routing of continuous wave light, we excited the GaN nanowire with the focused beam of a HeCd laser operating at 325 nm. Band-edge PL from the GaN cavity was channeled through the SnO₂ ribbon to emerge primarily at its far end. A fraction of the light was also scattered by imperfections along the length of the ribbon (i.e., attached particles or macroscopic step edges). Propagation losses through nanoribbons of this size are ≈10 dB/cm for wavelengths from 400 to 550 nm (smaller ribbons show higher losses of up to 80 dB/cm at 475 nm). Far-field spectra collected from the output end of the ribbon (Fig. 1f) show that the broad PL band of GaN is imprinted with the mode structure of the SnO₂ cavity during its transit. This mode structure is not longitudinal (Fabry-Perot) in nature, as it is for shorter nanowires (8); instead, it is likely caused by transverse modes that depend on ribbon shape and cross-sectional dimensions.

By pumping the GaN wire above its lasing threshold (≈100 μJ/cm², see Fig. 6, which is published as supporting information on the PNAS web site) with pulsed UV excitation, we were able to send single optical pulses from the nanowire laser through the nanoribbon waveguide (Fig. 1e). The spectrum of several thousand accumulated pulses (Fig. 1f) shows a series of sharp modes with a full width at half of maximum of 0.8 nm. These are the Fabry-Perot type lasing modes of the GaN wire resonator, modulated in intensity by the ribbon cavity. We have obtained similar results with junctions between ribbons and lasing ZnO nanowires. Moreover, it is possible to simultaneously guide the output of two (or more) nanolasers by coupling multiple ZnO and GaN wires to the same ribbon (Fig. 2), establishing the possibility of performing all-nanowire nonlinear wave mixing within single nanocavities. This is experimental verification that coherent optical pulses can be directly transferred between nanowires and steered hundreds of micrometers from their source. With high-frequency electrical pumping, nanowire laser/waveguide combinations could be used to transduce and shuttle packets of electro-optical information within future computing and communications devices.

Because diffraction losses in a subwavelength cavity increase markedly with wavelength, a ribbon waveguide preferentially confines the bluer portion of any nonmonochromatic beam. As a result, ribbons act as short-pass filters with cutoff wavelengths that are determined by their cross-sectional dimensions and overall length. We assembled a simple network consisting of four ribbons of different sizes to show how such a structure may be used to separate colors (Fig. 3). When excited at 325 nm, the large ribbon that forms the stem of the network emits white light composed of two broad SnO₂ PL bands centered at 495 and 590 nm (Fig. 7, which is published as supporting information on the PNAS web site). Varying amounts of the stem emission then flows into the three shorter and consecutively thinner branch ribbons, separating the white light into green, aqua, and blue components (ribbons 1–3). Alternatively, monochromatic red light launched into the stem was effectively guided only by the stem ribbon, whereas green light was guided strongly (weakly) by the largest (smallest) branch and blue light passed through all four channels (Fig. 3 b–e). Although this color filtering effect works only in short-pass mode, and so cannot, for instance, isolate the pure red component of a white beam, it may prove useful in tasks such as providing local excitation for fluorophores with narrow absorption bands, such as quantum dots.

To test the limits of intercavity optical coupling, we assembled four nanoribbons into a rectangular grid (46 μm long ×≈25 μm wide) featuring X-junction vertices with small contact areas (<0.15 μm²) (Fig. 4a and Inset). The structure was designed with one long channel for light input and seven short output channels that could be monitored simultaneously. As shown in Fig. 4b, direct excitation of the input channel triggered emission from all seven of the ribbon outputs, with the following intensity distribution: 1 ≫ 6 > 4 ≈ 7 > 3 > 5 > 2. This sequence is exactly the one we would expect after considering the trajectory of the incoming light and the intensity of scattering at the four ribbonribbon junctions. The light trajectory is important here because the low reflectivity of their end facets makes ribbons poor resonators (with an ideal finesse of ≈1.3). As such, most photons do not make multiple passes and light flow is highly directional. The right-angle intersections present significant obstacles to intercavity waveguiding by total internal reflection. At the same time, they act as quasi-isotropic scatterers that feed light between ribbons. Ribbon-to-ribbon losses, although nearly maximized in this geometry (<5% coupling efficiency), are still low enough for the activation of channels 2 and 3, which require photons to negotiate two right-angle junctions and transit three separate cavities. When we added a ZnO nanowire laser to the input channel and used it to launch light into the grid, emission was detected from all channels but 2 and 3; the number of injected photons was simply too small to illuminate the parallel ribbon in this case. Nanowire grids have already been used to implement rudimentary electronic logic (9). Integrated optical logic (10–12) and all-optical switching (13, 14) present exciting prospects, and our results show that grids of nanowires should be capable of routing signals for such tasks.

Because of their high refractive indices (n ≥ 2), the nanoribbons and wires described here function well as waveguides in water and other liquids. This is a considerable advantage over subwavelength silica waveguides (15), which cannot efficiently confine visible light in liquids because of a low dielectric contrast (n _silica ≈ 1.45). Waveguiding in liquids is especially important for integrated on-chip chemical analysis and biological spectroscopy in which small excitation and detection volumes are required. Fig. 8, which is published as supporting information on the PNAS web site, shows that the guided PL spectrum of a thin ribbon (in this case, 365 nm × 105 nm × 265 μm) resting on the oxide surface broadens to longer wavelengths when it is covered by pure water. Such a redshift would be anomalous for a fiber with a cladding of homogeneous refractive index, where one expects the replacement of air (n = 1) by water (n = 1.33) to increase losses and result in a blueshift of the mode cutoff. However, when a slab or strip waveguide exists in an asymmetric cladding environment (that is, when n _waveguide > n _subtrate > n _cover), as it does here, raising the index of the cover reduces its asymmetry with the substrate and improves confinement (16, 17). Numerical simulations show an increase in the effective guide index with a decrease in substrate-cover index profile, strongly supporting this view (see Fig. 9, which is published as supporting information on the PNAS web site). We note that the spectra of ribbons too thick to show a PL cutoff in air were unaffected by immersion in water.

Ribbon waveguides can sense molecules, proteins, or larger biological entities in solution by means of either an emission or absorption mechanism. In the former, a ribbon provides local excitation for fluorophores passing through the cone of scattered light at its output end, and the emission is collected by a fiber or microscope. To demonstrate this fluorescence scheme, we embedded the tip of a ribbon in a ≈3-pL droplet of 1 mM Silosuper Pink B [i.e., Rhodamine 6G (R6G) laser dye] in 1,5-pentanediol (n = 1.45). Blue light (442 nm) launched into the far end of the ribbon resulted in strong fluorescence from within the droplet, where the R6G emission mapped out the spatial intensity distribution of the waveguide output as a cone of light (Fig. 5a and Insets). A fraction of this fluorescence was captured by the ribbon cavity and guided back to its far end, demonstrating that these waveguides are capable of routing signals both to and from liquids. Spectra acquired from both ends of the ribbon are presented in Fig. 5b. The guided fluorescence is red-shifted and somewhat sculpted by its passage through the ribbon, but there is little trace of the heavy mode imprinting evident, for instance, in Fig. 1f.

Fig. 5a also shows strong fluorescence originating from the segment of the ribbon wet by the droplet through capillary action. Here, dye molecules in proximity to the ribbon surface are excited in a subwavelength version of total internal reflection fluorescence (TIRF). In normal TIRF, excitation of a macroscopic waveguide (such as a microscope coverslip) generates an evanescent field of light that decays exponentially with distance from the waveguide surface, limiting the depth of excitation to a distance of ≈100 nm and enabling the local probing of structures such as cell membranes (18, 19). Evanescent wave sensing has also been demonstrated by using conventional optical fibers (20, 21). Because subwavelength fibers can carry a larger fraction of their modal power outside of the core, they enhance the intensity of this evanescent field and increase its penetration depth into the surroundings, making proportionally more power available to excite nearby molecules. Calculations indicate that ≈15% of the electric field intensity exists outside of the ribbon for the wavelength of light used in this experiment. In this case, the radial field intensity decays to 10% of its maximum value by ≈135 nm into the liquid solution. Because TIRF detection sensitivity scales with the fractional power present in the waveguide cladding, 1D nanostructures are promising waveguides for local fluorescence sensing using this approach.

A second way that nanoribbons/wires may be used for optical detection in solution relies on producing an absorption spectrum of molecules located on and near the ribbon surface. Absorbance detection, while inherently less sensitive than fluorescence methods, is applicable to a wider range of molecules and avoids the complications of fluorescent tagging. We launched white PL down a long ribbon (260 nm × 240 nm × 540 μm) onto the midpoint of which a ≈1-pL droplet of 1 mM R6G (α_max = 535 nm) was deposited. Dye molecules in the droplet imprinted their absorption signature onto the propagating double-Gaussian beam, completely quenching transmission through the ribbon around the R6G absorption maximum (Fig. 5d). Considering the dye concentration, droplet size, and spatial extent of the evanescent field, we estimate that <40 attomoles of dye (≈24 million molecules) were probed in this experiment. Because this absorbance approach also uses the evanescent fraction of the guided field, smaller ribbons should again provide greater sensitivity. Other options for improvement include altering the cavity shape to increase the probe length, functionalizing the ribbon surface for selective biosensing, and launching multiple wavelengths for the simultaneous detection of analytes with different electronic transitions. The next steps are to integrate subwavelength nanoribbons/wires into microfluidic devices (22) and apply them as flexible probes in the study of live cells (23).

Conclusions

Chemically synthesized nanoribbon and nanowire waveguides have two unique and potentially useful features for subwavelength photonics applications. (i) Nanowires push subwavelength optical fibers beyond silica. The scores of materials that can now be made in nanowire form include active, passive, nonlinear, and semiconducting inorganic crystals, as well as a wide variety of polymers. Simultaneous manipulation of photons and charge carriers is possible within and between nanowires of different compositions. Also, many of these materials have higher refractive indices than silica-based glasses, permitting light of a given wavelength to be confined within thinner structures for denser integration. This strong confinement enables waveguiding in liquids and makes it possible to extend subwavelength guiding to telecommunications wavelengths using, for instance, ≈300-nm diameter Si or GaP wires (24, 25). (ii) Nanowires are freestanding, mechanically flexible elements that can be manipulated on surfaces or used as mobile probes in fluids. As such, they offer a type of versatility difficult to achieve with lithographically defined structures that are permanently affixed to their substrates.

The disadvantages of nanowire photonics include (i) the paucity of parallel assembly methods for accurately arranging large groups of nanowires into useful structures; (ii) relatively high interwire coupling losses compared with monolithic waveguides formed by lithography [coupling losses could be greatly reduced if branched, multicomponent nanowires were developed to replace the staggered or crossed nanowire cavities used here (26–28)]; and (iii) the lesser geometric perfection of nanowire assemblies relative to the precise shapes and sizes definable with lithography. Geometric imprecision introduces some uncertainty in the resulting light propagation and adds complexity to nanowire experiment/theory comparisons. Despite these limitations, we believe that freestanding, chemically synthesized nanowires represent a unique materials platform for the study and implementation of subwavelength optics.