Interfacing optical fibers with plasmonic nanoconcentrators

1.1 Conventional fiber tips: properties and limitations

One of the first approaches for achieving confinement of light at fiber tips involves using silica fiber tapers with subwavelength diameters [40], which has found widespread applications in photonics, including sensing [11], nonlinear light generation [41], telecommunications [42], imaging [17], and particle trapping [43]. Chemical etching [44], [45] and heating-and-pulling techniques [46] can produce fiber-based deep-subwavelength tips with extremely smooth surfaces. Although silica tapers support a single mode at visible wavelengths down to a waist diameters of ~100 nm (Figure 2A), shrinking them further causes the effective index of the mode to approach unity, indicating leakage of light into free space and preventing subwavelength light concentration below the diffraction limit (Figure 2B). Such “nanospikes” are often used as high-efficiency fiber couplers, e.g. in high-index contrast hybrid fibers for nonlinear applications [47], and also self-align when placed inside large-area hollow-core photonic crystal fibers [48], [49]. This is due to the two to three times enhancement of the intensity in a region close to the fiber tip (~λ/2 spatial confinement), making them suitable for other optomechanical applications such as optical tweezing and particle trapping [43]. For imaging applications, it has been shown that such tips allows spatial resolution in the order of ~70 nm by passing light to the fiber twice [17].

Figure 2:

Inherent limits of conventional subwavelength fiber-based nanotips.

(A) Calculated effective index of the fundamental linearly polarized mode a silica nanofiber (inset) as a function of its local radius. For small radii, light leaks into free space (effective index=1) and cannot be confined. (B) Finite element calculations of the intensity (log scale) for a mode propagating toward the apex of a tapered silica nanotip. Black line: edge of the taper. (C) Calculated real (blue) and imaginary (red) parts of the effective index of the fundamental linearly polarized mode a silica nanofiber that is embedded in gold (inset) as a function of its local radius, corresponding to a typical aperture probe. For small radii, the mode cuts off and becomes evanescent, leading to low throughput (1.3×10⁻⁵) through the aperture (diameter: 10 nm). (D) Finite element calculations of the intensity (log scale) of the mode propagating down a tapered aperture probe. Yellow line: edge of the metal-coated taper. Scale bar (B and D), 200 nm. The radius of the fiber at the tip/aperture in B or D is 10 nm. All calculations are performed at λ=600 nm.

One obvious pathway to prevent the leakage of radiation into free space relies on coating the fiber with metal and leaving a nanoaperture at the end-face of the fiber, forming the so-called aperture probe. The diameter of the aperture in fact defines the spatial resolution achievable, so that raster scanning an illuminated sample close to the surface enables imaging with subwavelength resolution. To date, this type of approach is the most widely used method for near-field scanning optical microscopy, owing to its simplicity and ease of integration into most setups, and is readily commercially available (http://www.nanonics.co.il/our-products/nsom-snom-probes). However, one major drawback of this device is its comparably low performance in both collection and delivery operation. This can be understood by considering the properties of the aperture mode as the hole diameter is reduced (Figure 2B): for taper diameters down to ~100 nm, a propagating mode is still supported. However, for smaller diameters, the modes cut off at a significant distance before reaching the end of the fiber and only its evanescent tail reaches the actual aperture at the fiber end, leading to low optical throughput, typically in the order of 10⁻⁵ [20]. Although this can be improved by two orders of magnitude by designing the aperture probe such that its mode resonantly couples to the evanescent mode in the tapered region [50] or by placing a monopole antenna close to the aperture [51], better strategies are required to achieve nanoscale spatial confinement accompanied by large-field enhancements and percentage conversion efficiency.

1.2 New strategies for fiber-compatible nanoconcentration

1.2.1 Conical tips

Stockman addressed a fundamental challenge in nano-optics, namely the efficient delivery of optical energy to the nanoscale, orders of magnitude below the diffraction limit in free space [52]. This is achieved via SPPs propagating on a tapered metallic nanowire (i.e. nanocone) with apex size of <10 nm. When travelling toward the apex, the propagating SPPs are adiabatically slowed down and show an increasing localization to the nanotip in all three dimensions, with modal areas of (λ/10)². Note that although for a cylindrical metal waveguide several modes are supported, only the radially polarized (RP) modes (in terms of the transverse electric field) possess this property of increasing field localization, with all other modes either cutting off or transversely dissipating energy for small lateral tip dimensions [53], [54], similar to tapered dielectric waveguides. RP SPPs are also often referred to as the short-range (SR) surface plasmons and are analogous to the asymmetric plasmonic modes (in terms of the magnetic field) in planar metallic films. They are fundamentally different from long-range (LR) surface plasmons, which are lower-loss symmetric modes. In case of metal films, modal splitting between the SR and LR SPPs emerges for film thicknesses below 100 nm, above which only single-interface plasmon modes exist [54]. In cylindrical wires, this splitting occurs at diameters approximately 10 times larger than the mentioned film thickness, significantly relaxing fabrication and material constrains.

The SR-SPP behavior can be understood in terms of its phase velocity v_p=c/n_eff as a function of wire diameter. Figure 3A shows the phase velocity for a gold wire (ε_Au=−8.4398+1.4108i [55]) in air (ε_Air=1), at 600 nm, which asymptotically tends to zero when travelling toward the apex of the nanotip. The subwavelength nature of the plasmonic mode can be regarded as a reduction in the diffraction limit, which is associated with a large effective index of the SPP, Δx ~ λ/2n_eff, with the spatial confinement being limited only by the size of the tip itself. The corresponding losses are shown in Figure 3A (red) and are between 3 and 20 dB/μm. Although large, they are typically not particularly detrimental due to large intensity enhancements and comparably short sample lengths (Figure 3C), so that large conversion efficiencies to/from the tip can be achieved; for example, Chen et al. [56] reported a 70% collection efficiency at 780 nm for a 6 nm diameter tip using a single emitter embedded in a substrate placed 5 nm from the tip apex.

Figure 3:

Fiber-compatible schemes for plasmonic nanoconcentration at the end-faces of optical fibers.

(A) Phase velocity normalized to the speed of light in vacuum (blue) and modal loss (red) of the RP plasmonic mode (inset) for a gold nanowire surrounded by air as a function of local radius. (B) Calculated electric field amplitude of the RP electric mode as a function of the tip radius, showing plasmonic enhancement. (C) Finite element simulations of the intensity for a mode propagating toward the apex of a gold nanocone (apex radius: 2.5 nm). (D) Phase velocity normalized to the speed of light in silica (n=refractive index of silica; blue) and modal loss (red) of the linearly polarized mode supported by a metal-dielectric-metal structure as function of the silica slab width w. The structure consists of a silica slab surrounded two gold nanolayers (gold thickness: 100 nm). (E) Calculated electric field amplitude as a function of the waveguide width w and constant gold thickness through the central slice of the tapered MIM structure (dotted line), showing plasmonic enhancement. (F) Finite element calculations of the spatial distribution of the intensity for a mode propagating down the tapered MIM waveguide in the dotted plane in (E). Scale bar (C and D), 200 nm. All calculations are performed at λ=600 nm.

2.1 End-fire coupling and mode matching: toward efficient nanowire plasmon excitation

The most common ways to couple fiber and nanocone modes rely on end-fire coupling, adiabatic modal conversion, and phase matching. The most practical and direct way of exciting SPPs, particularly in the context of fiber optics, is via end-fire coupling. However, as a result of the inherently different modal characteristics of SPPs (which can have modal areas of ~100 nm²) and fiber modes (typical modal areas are 10–100 μm²), associated coupling efficiencies are typically low. An example fiber-plasmonic end-fire interface is shown in Figure 4A: a step-index fiber is aligned with the symmetry axis of a silver nanowire. Using typical fiber parameters (silica cladding, 4 μm diameter, doped core Δn=8×10⁻³), the modal overlap [60] between the radially fiber polarized mode (Figure 4B, left) and the SR-SPP of a silver nanowire with 1 μm diameter (Figure 4B, right) is ~17% at λ=633 nm. This results in an overall poor conversion of light from the photonic to the plasmonic mode with most of the energy being lost, as shown in the 3D simulations of Figure 4C. This comparably simple structure is quite challenging to fabricate even using splice-and-pressure-assisted melt filling of gold, whereas the issues of core/nanocone alignment and air gaps at the splice point introduce additional losses and scattering at the fiber/nanowire junction.

Figure 4:

Schematics of typical structures that use end-fire coupling to excite the plasmonic modes.

(A) State-of-the-art few-mode fiber (core radius: 2 μm, index difference: 8×10⁻³) that interfaces a plasmonic nanowire (radius: 1 μm). (B) Calculated mode profiles for the RP modes of fiber (left) and SPP (right). (C) Propagation simulations, showing that the incoming field is partially reflected and scattered, with only a small portion being coupled to the nanowire mode. (D) Schematic of the geometry proposed in Ref. [56], which showed improving end-fire coupling. Here, a thin silica dielectric fiber (radius: 342 nm) located in air interfaces with a gold nanowire (radius: 164 nm). (E) Calculated mode profiles for the corresponding RP modes of fiber (left) and SPP (right). (F) Propagation simulations showing that most of the light is coupled to the plasmon mode (up to 99% [56]).

Chen et al. [56] showed in simulations that this can be significantly improved using a silica nanofiber in air that interfaces a silver nanowire (schematic in Figure 4D). Here, simple modal overlap integration between the dielectric mode (Figure 4E, left) and the SPP mode (Figure 4E, right) is not sufficient to predict the coupling efficiency, as a result of the nontrivial modal interaction at the boundary. Indeed, finite element simulations show almost 100% conversion efficiency from the fiber mode to the SPP (Figure 4F). It was also found that “rounding the edges” of the nanowire at the silica/gold boundary improved the coupling efficiency significantly. This kind of interface thus provides the first pathway to efficiently couple light to nanoscale dimensions and vice versa; it was also predicted that the tapered silver nanowire potentially allows to collect 70% of the light from a point source placed 5 nm below the nanowire. Note, however, that fabricating this particular structure is extremely challenging in terms of obtaining large aspect ratio tapers and precisely producing the required dielectric fiber and nanowire diameters with nanometer precision (including rounded edges). Also placing the nanowire at the center of the dielectric cylinder with such small diameter is crucial, as small deviations from the optimal position would reduce the coupling efficiency significantly.

2.2 Adiabatic modal conversion of SPPs on metal-coated fiber tapers

Metal-coated fibers shaped into a cone terminating into a sharp apex can also be used to convert the RP mode propagating “inside” the fiber taper to the RP plasmons on the “outer” film that are superfocused at the sharp apex (Figure 5A, inset). This was shown theoretically by several groups [62], [63], [64], [65] and later demonstrated experimentally [61], [66]. In essence, it relies on adiabatically transferring the RP dielectric mode into the plasmonic mode. Figure 5A shows the effective indices of the RP modes for a metal-coated tapered fiber. As the radius of a silica nanofiber decreases, the effective index asymptotically would tend to unity (Figure 2A). In this case, however, the dielectric-like mode anticrosses with the (external) SR-SPP mode, which has the same properties as an SR-SPP of a solid gold wire (green dashed line). The two modes couple strongly and split at the anticrossing point, so that the dielectric mode (Figure 5B) transitions into the external SPP mode (Figure 5C). Thus, the “internal” RP fiber mode (Figure 5B) transitions to a guided “external” SPP (Figure 5C) outside the metal film (metal/air interface; Figure 5A, red line; Figure 5D shows full 3D FE simulations of this transition). This coupling process can also be interpreted as a nanofiber equivalent of the planar Kretschmann configuration [27], where phase matching has been achieved between the radial SPP and the fiber core mode [61].

Figure 5:

Superfocusing of SPPs on a metal-coated optical fiber tip.

(A) Real part of the effective index for the relevant modes supported in a typical experimental configuration [61]. The red curve represents the RP dielectric (internal) mode (intensity distribution at R=1 μm shown in B) of the fiber that anticrosses with the (external) RP plasmonic mode (shown in C at R=0.3 μm; dashed green line), enabling plasmonic adiabatic tapering. (D) 3D finite element simulation illustrating the evolution of the intensity (|E_z|²) of the mode that incidence from the right into the fiber tip. Inset: field enhancement at the plasmonic tip due to the axially polarized mode. (E) Example of a gold-coated fiber tip (gold thickness: 30 nm, plasmonic tip radius: 100 nm) excited by an RP mode (λ=780 nm). Top: SEM image of the fiber structure. Also shown are polarization resolved optical microscope images of the light emitted from the tip. Middle/bottom: horizontally/vertical oriented polarizer between tip and camera (adapted from Ref. [61]). Reprinted with permission from Ref. [61]. Copyright 2015 American Chemical Society.

Energy transfer is expected to be more efficient in the infrared with a reduced confinement and slightly less efficient in the visible but with tighter localization [65]. The full coupling process subtly depends on shape of the tip, taper angle, modal transition length, and metal film thickness: up to 40% tip transmission can be reached in the visible. These results warrant a comprehensive study of the tradeoff between the long taper lengths and by the adiabatic condition [67] and the short taper lengths required to reduce plasmonic losses to acceptable levels.

Figure 5E (top) shows a typical tapered fiber tip coated by gold. In delivery mode, it was shown that light coupled into the fiber is localized at the tip and polarized along the fiber axis, confirming the excitation of the SPP mode (Figure 5E, middle: polarizer between sample and camera is along the fiber axis; bottom: polarizer between sample and camera is perpendicular the fiber axis). In collection mode, the light polarized along the taper tip axis was collected and converted into an RP mode in the fiber, which was applied to detecting the transversal fields of scattered modes.

This scheme is flexible in terms of design and highly compatible with fiber technology; however, tips with radii smaller than ~125 nm are difficult to achieve as a result of the granular nature of gold, a limitation that should be addressed for it to be competitive with commercial systems in terms of resolution. Finally, it should be noted that, for efficient operation in delivery mode, an isolated TM₀₁ is required in the fiber; this being a higher-order mode, long lengths of fiber result in random mode scrambling to other modes that do not excite the SR-SPP, resulting in a decrease in performance.

2.3 Gold-filled nanobore optical fiber

The issue of precisely aligning metal nanowires with the core of a dielectric fiber (see Figure 4) can be addressed using pressure-assisted melt filling (PAMF) [68]. With this technique, it is possible to interface metal nanowires (e.g. gold) with diameters of a few hundred nanometers and lengths of several centimeters within a glass capillary or fiber containing one or more nanochannels in its core and/or cladding. The ductility of gold enables the immediate generation of nanotips at fiber end-faces, which possess an apex size of ~10 nm, via a straightforward fiber cleave [69]. In this approach, the tip is formed naturally and uses a quasi-bottom-up strategy, which does not require complex top-down nanofabrication technologies, which struggle to reach dimensions of 10 nm or below.

This technique was recently applied to an microstructured optical fiber including a central gold filled nanochannel (or nanobore) [70], [71], leading to a perfect and monolithic alignment between fiber core and gold nanowire. Based on this geometry, it was recently proposed and experimentally verified that a fully integrated fiber-based near-field nanoprobe can deliver light to nanoscale dimensions (apex size of <10 nm) even after propagating hundreds of micrometers. Such lengths are normally prohibitive for SR-SPP propagation as a result of losses. However, using the dielectric-like hybrid mode, it is possible to couple to the SR-SPP, albeit with a lower efficiency, using the so-called “hybrid mode-assisted” SPP excitation scheme. The device is shown in Figure 6A: a cylindrical silica step-index fiber that incorporates a central nanochannel is partially filled with gold and cleaved at an appropriate location to produce a nanotip, on which the light is propagating and focused. Figure 6A (inset) shows an example tip apex (<10 nm). The device functions by exciting the two RP modes in the gold-filled section – the first being dielectric (most of the energy in the dielectric) and the other plasmonic (most energy near the gold surface). As the plasmonic-type mode fully decays after a few micrometers, only the dielectric-like mode “survives” and via its tails in the metal excites the SR plasmonic mode at the fiber end-face, which then gets focused on the tip. The hybrid nature of the dielectric mode thus relaxes the fabrication constrains, allowing functional devices with nanowire lengths between tens and hundreds of micrometers. However, as only a small portion of the energy is near the gold surface for this mode, the conversion efficiency is in the order of 0.1% [70].

Figure 6:

Nanobore fiber-based near-field probe consisting of a gold nanowire inside the fiber core with protruding gold nanotip.

(A) Schematic of the monolithic near-field nanoprobe excited by either RP or AP mode. The magenta dot at the front of the tip indicates the intensity enhancement. The inset shows an SEM of a typical gold nanotip, with the sub-10 nm tip region highlighted in green. (B) Illustration of the operating principle: the input core mode (dark green) excites two hybrid EMs (dual-mode region) that propagate in the gold-filled region. The SR-SPP mode has high loss and rapidly decays, so that the TM₀₁ is the dominant mode at the fiber end-face. The RP field on the gold surface is focused at the apex via the SR-SPP travelling down the tip; the field in the dielectric core diffracts into free space. (C) Microscopic image of side scattered light for an RP input state for a short (left) and long (right) nanowire inside the fiber. Highlighted are the empty (blue) and nanowire (orange) regions. The scattered light originating from the nanotip is indicated by magenta arrows. (D) Detailed view of the measured scattered light when the input is RP (top) or AP (bottom). (E) Also shown is the scattered light with a polarizer inserted between sample and camera that is either axially (top) or transversely oriented (bottom). The orientation of the polarizer with respect to the nanotip axis is indicated by white double-headed arrows. All data are for λ=650 nm. Adapted from Ref. [69]. Reprinted with permission from Ref. [70]. Copyright 2017 American Chemical Society.

For an RP input, the side-scattered light showed a bright spot at the fiber end-face, which is unaffected by the length of the nanowire (Figure 6C), in support of the hybrid mode-assisted excitation scheme. Additionally, the fiber scatters the most when the input is RP (Figure 6D) and the least when it is azimuthally polarized (AP), in agreement with expectations. Finally, it was shown that the light at the apex is axially polarized (Figure 6E), as expected for SPP focusing. This indicates that the scattered light originates from the SPP mode and reaches nanotips with apex dimensions below 10 nm.

The hybrid mode-assisted technique, although practical in terms of fabrication and monolithicity, is limited in coupling efficiency due to the large mode mismatch between dielectric and plasmonic modes as a result of the small dimensions of the nanowire compared to the fiber core. Additionally, end-fire coupling does not allow for the selective excitation of a desired mode and requires the appropriate input polarization state to be prepared. As we now show, this issue can be overcome in a directional coupling scheme.

2.4 Optical fiber with wire satellite next to the core

An alternate technique for the efficient excitation of SPPs is plasmonic directional coupling, which requires phase matching between neighboring waveguides. Inspired by previous planar plasmonic directional coupling designs [72], [73], it was suggested to design a hybrid fiber plasmonic focusing device that is compatible with standard optical fibers and can be fabricated using PAMF [74] (Figure 7A). The hybrid device is formed by two parallel running waveguides, whose modes have the same effective index when their separation is infinitely large (phase-matching condition); when the waveguides are brought together, their separation determines the strength of the hybridization (i.e. the difference between the two supermode effective indices Δn_eff) and thus the plasmonic coupling length L_c, which equals half the beat length of the two eigenmodes (EMs), L_c=λ/2Δn_eff. The modal evolution along the coupler is well described by an EM analysis, whereby the total electric and magnetic fields in the dual mode region (Figure 7A) are expressed as a superposition of the two supported hybrid EMs. This EM analysis confirmed that the excitation of the plasmon mode is typically quite efficient (in the order of 40%) and broadband (λ=1.4–2 μm) over a short coupling length of only a few micrometers.

Figure 7:

Simulated characteristics of the fiber-based hybrid plasmonic directional coupler.

(A) 2D scheme of the couplers including all relevant modes. The dielectric core mode excites two hybrid EMs in the dual waveguide region, which propagate and beat to excite an SR-SPP after a defined propagation length at the end of the device that is nanofocused using the tip. (B) Real part of the effective indices of the hybrid modes of a fiber-based plasmonic directional mode coupler for the excitation of the RP plasmonic mode on the nanowire (core diameter: 1 μm, wire diameter: 0.5 μm, edge-to-edge separation: 0.5 μm, light blue: core index=1.67, dark blue: silica cladding, yellow: gold satellite). (C) SEM image of a two-hole capillary to be filled with a high-index dielectric (core) and gold (wire). Scale bar, 1 μm. (D–G). Spatial distributions of the Poynting vector of the hybrid plasmonic directional coupler that includes a nanotip at characteristic spatial positions (λ=1.55 μm) [74]. (D) Distribution around the coupling region (central xz-plane) calculated using approach discussed in Ref. [74]. The dashed line indicates the coupling length. Finite element simulations at the end of the fiber (E) and at the end of the apex (F: xz-plane and G: xy-plane; both 1 nm away from the apex). Adapted with permission from Ref. [74].

The example of interest is a realistic equivalent fiber-like superfocusing device consisting of two parallel cylindrical wires (one dielectric and one gold) embedded in silica, where the gold nanotip forms naturally during the fiber cleave. To provide a sense of the geometric parameters required, at λ=1.55 μm, the SR-SPP mode of a 500-nm-diameter gold nanowire phase matches with the fundamental core mode of a step-index fiber waveguide of 1 μm diameter and refractive index n_core=1.67, which is compatible with current fabrication techniques of hybrid optical fibers [2]. As an example, Figure 7C shows a two-hole capillary to be selectively filled with a high-index dielectric material and gold. Mode evolution simulations (Figure 7D) predict a remarkable power transfer of ~53% after a coupling length of L_c=22 μm. By cleaving at this location to form a conical tip – so that the first part of the wire is embedded in silica and the second part tapers in down a nanoapex in air – such a mode then undergoes plasmonic concentration (5 nm tip radius) over a length of 0.5 μm (Figure 7E–G) with spot sizes of ~10 nm (Figure 7E and F).

4.1 Fiber-based photonic-plasmonic circuits

Another avenue for coupling fiber modes to SPPs was demonstrated by Li et al., who integrated a silver nanowire directly onto the end-face of a tapered optical fiber [77]. They adapted existing techniques [78] for coupling to plasmonic nanowires and nanocircuits via optical fibers, which relied on the nanowire resting on a substrate, to produce nanowires that are monolithically attached to the fiber in free space. The individual silver nanowires were fabricated by chemically reducing silver nitrate with ethylene glycol and polyvinyl pyrrolidone using a self-seeding process [79]. This produces nanowires of nominally 200 nm diameter and several tens of micrometers in length. The fiber tapers were made with a regular standard flame-heated technique down to a 300 nm waist. When a silver nanowire was brought in contact with the tapered fiber using tip micromanipulation, it remained attached to fiber due to van der Waal forces, possessing sufficient mechanical strength for many applications. Figure 9A shows a typical SEM of the resulting hybrid fiber plasmonic device.

Figure 9:

Silver nanowire integrated onto the end-face of a tapered optical.

(A) SEM image of the fabricated structure [77]. (B) Microscope side image in the situation light at a wavelength of 785 nm is coupled into the device. (C) 3D finite element simulations showing the Poynting vector magnitude distribution in the vicinity of the nanowire. Here, a linearly polarized input mode is capable of exciting an LR plasmonic mode on the nanowire. The color bar is clipped to 25% of the maximum value. (A and B) Adapted with permission from Ref. [77].

Figure 9B shows experimental side images of the light scattered by the silver nanowire (right) when light (λ=785 nm) is coupled into the fiber (left) with fiber/plasmon coupling efficiency and propagation losses of 92% and 0.41 dB/μm (corresponding to the expected loss of LR plasmons for such nanowire thickness and wavelength). The high efficiency of the coupling mechanism can be understood as follows: as the linearly polarized mode of the optical fiber taper cuts off (Figure 2), the light can efficiently couple into the linearly polarized LR plasmon of the attached silver nanowire, establishing a nearly adiabatic transition from the dielectric optical fiber mode into the plasmonic mode of the silver nanowire.

Although the authors suggest that this might be useful for near-field imaging, optical sensing, and nanoscale endoscopy, the achievable confinement is not better than the diffraction limit, due to the excitation of the linearly polarized SPP mode, possessing an effective index that approaches unity in air. As a result, the nanoconcentration that can be achieved is limited. However, the proposed fiber-integrated nanowire platform is undoubtedly attractive for fiber-compatible ultra-compact photonic interconnects and devices, with Mach-Zender interferometers [77], photonic-plasmonic nanocircuits [80], and quantum-dot emitters [81] already reported. Most recently, it was shown that this type of hybrid nanoscale plasmonic waveguide can be used as a quantum probe that maintains the quantum polarization entanglement state [82] and has the potential for highly sensitive measurements on the subwavelength scale, thus bridging free space quantum optics with conventional fiber-based nanophotonics [83].

4.2 LSPRs on a fiber end-face

Another approach for collection mode operation relies on another kind of plasmonic mode being excited on the nanowire tip, namely the LSPR. LSPRs are “nonpropagating” collective oscillations of the surface electrons of metals around sharp (subwavelength) features in metallic nanostructures that are excited by external fields and show strong local field enhancements. Typically, their scattering properties are determined by shape and the (negative) permittivity of the metal and the (positive) dielectric; for deep-subwavelength particles, the resonance peaks occur at the wavelength where the real parts of the permittivity of the materials involved have the same magnitude, which for silver and gold occurs at visible wavelengths. The large field intensities and the strong dependence of the resonance on the surrounding medium make them an excellent candidate for nanoconcentration.

A number of examples of fibers with gold nanospheres and nanoparticles have been reported [31], [32], [34], [35]. However, although these provide high-performance sensing functionality for large areas, they do not provide high spatial resolution. An alternative approach relies on a slight modification of the geometry considered in Section 2.3 (Figure 10A): a few-mode step-index fiber interfaces a tapered silica capillary that incorporates a gold nanowire. The tapered portion is obtained by PAMF, a silica capillary, splicing it to the multimode fiber, and wet etching the capillary to achieve a taper, and finally cleaving 1 cm away from the splice point. Remarkably, when this procedure is used, a nanoscale “tip-on-a-tip” (Figure 10C and D) emerges where the gold nanowires terminate, forming an ellipsoid that possesses sub-50 nm radius [69]. The nanoscale confinement of the LSPR formed at the small tip then enables probing the near field with deep-subwavelength resolution. As a proof-of-principle, an experiment was conducted to probe an evanescent field from a prism (Figure 10E): when the tip approached the prism, the LSPR was excited, which scattered part of its energy into the capillary, from which the light was coupled into the multimode fiber. Note that the tip supports several LSPRs at visible wavelengths (example simulations are shown in Figure 7B) due to elongated shape of the tip, allowing the SPPs to propagate between the base and the apex of the tip. This motion effectively forms a standing wave pattern, similar to the modes of a Fabry-Perot and thus the LSPR. As a result, the LSPRs depend on subtle features of the nanoscale tip, making the prediction of the resonance peak to some extent unpredictable. However, because fabrication is simple, many LSPRs can be fabricated on very short timescale, making sample-to-sample variance less critical, as the tip resonances are always located in the visible part of the electromagnetic spectrum (Figure 10F), making such a device particularly interesting for tip-enhanced Raman spectroscopy and near-field microscopy.

Figure 10:

Fiber-based nanoprobe relying on an LSPR at the end of an integrated gold nanowire.

(A) Schematic of the fiber-based near-field nanoprobe (yellow: gold nanowire; red: core of multimode fiber). (B) Microscope image of the scattered light of the plasmonic resonance when the nanoscale resonator is probing an evanescent field (right: end section of the nanoprobes; yellow dashed line: boundary of fiber; white dashed line: prism-air interface; left: close-up of the excited plasmonic resonance). (C) SEM image of the gold nanowire protruding from the end of the fiber probe. The dashed green square highlights the nanoscale plasmonic resonator at the nanowire’s end. (D) Close-up of the plasmonic resonator showing the tip-on-the-tip section. The actual resonator is overlaid with a semitransparent blue area for better visibility. The edge of the resonator is highlighted by the dashed green line. (E) Normalized spectral distribution of the light collected at the end of the fiber device in case it probes an evanescent field that is composed of white light. Adapted from Ref. [69]. Reproduced from Ref. [69], with the permission of AIP Publishing.

4.3 Bowtie nanoantennas on a fiber end-face

Although deep-subwavelength nanoapertures suffer from low transmission, it is possible to improve their efficiency significantly [24] by slightly modifying its shape into a bowtie (Figure 11A, inset), formed by partially closing the aperture on two of its sides to form a nanogap. Such a “bowtie nanoaperture” (BNA) structure essentially forms a local resonant antenna, which thus supports an intense mode in the gap region from the visible to the infrared when the exciting electric field polarization is parallel along the small gap (Figure 11B), leading to nanoconcentration of light and enhanced transmission.

Figure 11:

Integrated bowtie nanoantenna aperture at the end of an optical fiber.

(A) SEM side image of the tip section formed by a single-mode fiber interfacing with an aluminum-coated polymer tip that possesses a BNA at its end. Inset: top view of the apex with the BNA. Also shown are far-field images of the aperture output in delivery mode at a wavelength of 1.55 μm, with the polarization exciting the fiber-polarized (B) parallel and (C) perpendicular to the gap of the BNA, revealing that the transmitted field is enhanced when the BNA mode is excited. (D) Images obtained by the probe in collection mode when scanning a dielectric grating (periodicity: 700 nm; line width: 200 nm) Adapted with permission from Ref. [84].

Mivelle et al. [84] patterned a BNA on the apex of a metal-coated tapered single-mode optical fiber tip (Figure 1), thus interfacing single-mode optical fibers with near field with improved performance compared to conventional aperture probes. Fabrication relied on growing polymer tips on the cleaved end-face of a single-mode fiber [85], which are coated with aluminum (100 nm thick), the tip of which is subsequently patterned by FIB milling. Using a rounded radius of curvature (500 nm) at the tip, the propagating mode is always guided to/from the BNA (preventing the evanescent regime of Figure 2C), improving the mode matching between the deep-subwavelength BNA mode and the larger fiber mode, leading to improved transmission/collection efficiency. Calculations showed an intensity enhancement of up to 350 in the 45 nm antenna gap at λ=1.55 μm, and a collection efficiency of 20% for a dipole source oriented parallel to the gap placed in close proximity to the BNA is obtained. For a dipole perpendicular to the gap, collection efficiency drops by ~170 times. Transmission experiments confirm this strong polarization discrimination (Figure 9B and C); collection experiments were used to image a grating with subwavelength periodicity (Figure 9D).

Subsequent experiments at the BNA fiber tips showed that the high-field intensity of the gaps allow for nanoparticle trapping inside the BNA gap [86]. An alternative approach relied on fabricating a bowtie structure, formed by two adjacent touching metal apertures, directly onto a silica fiber core [87]. The authors identified the domain of the core by first high-field etching the cleaved fiber end-face, as the doped section etches faster than the silica cladding. The end-face was then coated with a 80 nm gold film, and two apertures were subsequently milled, which possess a 40 nm gap. The authors were able to trap both 40 and 20 nm polystyrene spheres in the vicinity of the gap. Note that as the nanotip is placed at the end of an untreated single mode fiber with comparably large outer diameter (125 μm), using this tip for scanning microscopy of flat surfaces can be challenging. This issue can be circumvented by first etching or tapering: indeed, a similar BNA design on a tapered fiber (final taper diameter ~1 μm) enabled trapping and manipulation of 50 nm particles using low optical powers [88], allowing to envision applications including noninvasive manipulation of individual nano-objects (e.g. viruses and proteins and nanocrystals) [7].

4.4 Dielectric hemisphere covered in tapered metal film

Another fiber-compatible nanofocusing geometry was suggested by Mason et al. [89]. The structure is formed by a silica fiber with a large spherical tip (sphere radius of 500 nm, which could be produced by adapting existing techniques [90]), as shown in Figure 12A. The subwavelength plasmonic confinement and enhancement in this case is provided by a metal film coating the fiber tip, with a film thickness that gradually reduces toward the apex of the sphere, requiring a few-nanometer film thickness at the tip of the hemisphere.

Figure 12:

Hemispherical fiber-based nanoprobe that consists of a spherical fiber end-face coated by a tapered metal gold film.

The two plots show 2D simulation of the spatial distribution of the electric field in the vicinity of the hemisphere (λ=632.8 nm). The thickness of the gold film changes from 200 to 5 nm toward the center of the hemisphere. An antisymmetric mode (with respect to the transverse electric field) that incident from the left-handed side excites a plasmonic mode that leads to energy concentration at the location of smallest gold film thickness. (A) Distribution of the transverse electric field (E_x). Scale bar, 200 nm. (B) Distributions of the longitudinal electric field component E_z at the end of the hemisphere. Scale bar, 100 nm. Calculations performed by the author, after Ref. [89].

Examples of finite element simulation of the proposed structure and of the actual device operation in the simple 2D case are shown in Figure 12B. The antisymmetric mode (with respect to the transverse electric field E_x) of the uncoated optical fiber excites the inner surface plasmons of the thick gold film for λ=632.8 nm. These propagate toward the fiber tip and constructively interfere at the apex as the gold film tapers, becoming polarized along the z-direction (Figure 12B). Because the gold film thickness is reduced to below skin depth, the electric field can penetrate outside the fiber hemisphere and can probe the surrounding environment. The antisymmetric transverse electric field pattern (Figure 12B) of the nanofilm at the tip indicates that this excitation scheme excites an SR surface plasmon in the thin-film region at the tip. The authors claim that, by engineering the metallic film thickness distribution appropriately, they can achieve energy confinement down to (i.e. ~λ/30) at 633 nm and an intensity enhancement of ~150 with respect to the input.

In principle, this scheme provides a means of using large fiber tip radii (several hundreds of nanometers) and in the absence of protruding metallic tips would thus provide a mechanically robust superfocusing platform, shifting the fabrication challenge to the realization of the tapered metallic nanofilm, which effectively provides the focusing and energy concentration. Although 5 nm gold films are challenging to produce as a result of the emergence of nanoislands below 30 nm, similar 5 nm gold films have been reported [91], [92] and other metals such as niobium yield films of uniform film thicknesses down to 10 nm [93]. Although possessing large losses in the visible, this might provide a viable platform at near-infrared wavelengths. Another issue to be considered is the integration into a near-field microscope, as such devices typically operated using the tuning fork principle (“shaking tip”), whereas the discussed tip might show an unfeasibly high resonance frequency.