1. Introduction

During the past four decades the propagation loss of single-mode fibers has been progressively reduced towards the fundamental limits imposed by Rayleigh scattering. This highly impressive technological achievement, driven by the increasing demands of telecommunications, has resulted in transmission loss values as low as 0.148 dB/km at 1550nm [1], and has only been possible through the continuous refinement of the fabrication process and improved understandings in glass science. The development of fiber taper technology also benefited initially from the growth in demand for optical components within telecommunications and more recently as a result of their increased application within fiber sensors and lasers [2–4]. Recently, the generation of broadband supercontinuum light has been added to the long list of taper applications thanks to the relatively high nonlinearity per unit length and unusual dispersion properties that are possible in these components [5].

Most of the applications to date have required tapers with micron scale diameters and few attempts to fabricate submicron structures [6–11] have been reported. Silica nanotapers with diameters in the range of several tens to several hundreds of nanometres have previously been fabricated using a variety of methods, however most of these have exhibited an irregular profile along their length and this has limited their usefulness for optical applications. In fact high diameter uniformity and low surface roughness are fundamental requirements to achieve submicron low-loss optical fiber tapers [6,12–13].

A recent publication [6] reported a two-step drawing process for the fabrication of low-loss submicron tapers and the measurement of their losses at 1.55μm and 633nm. A standard silica optical fiber was firstly drawn to a micron size using a flame and this taper was then drawn down to submicron dimensions by pulling the taper around a heated 80μm sapphire rod. This method allowed the fabrication of a taper with radius r~440nm and loss ~0.3dB/mm at 1.55μm. At 633nm, where the light is more confined and less sensitive to diameter uniformity and surface roughness, a loss of ~1.8dB/mm for tapers as small as 190 nm in diameter was reported. No loss measurements were reported for the smallest of the fabricated samples which had a diameter of ~50nm.

These excellent results, which have shown that tolerable levels of loss for many device applications can indeed be achieved in nanotapers, have generated renewed interest in the field.

In this paper we show that it is possible to fabricate long nanotapers with similar diameters and more than ten times smaller losses than previously reported with the two stage process using a conventional fiber coupler/tapering rig.

2. Experimental set-up

Our set-up is a simple coupler fabrication rig but with several relatively straightforward improvements to achieve more stable taper diameter control. A telecom fiber (outer diameter D~125μm, numerical aperture NA~0.12, cut-off wavelength λ_c~1250nm) is pulled by two translation stages each with submicron precision. A small region of the fiber is heated by a flame of millimetre dimensions fed by oxygen and isobutane. This small flame was scanned (over lengths of several tens of mm) along the fiber, to produce tapers with an extremely uniform waist diameter and taper transitions of well defined length and shape [14]. A constant hot-zone approach [14] was used for tapers as long as 40mm in order to ensure low insertion and extraction losses in the taper transitions. A hot zone with a length that changes with taper extension [14] was used to produce longer tapers (>40mm) in order to prevent the transition regions becoming unmanageably long. As an example, a uniform taper ~110mm-long with r~430nm, fabricated using a linearly increasing hot-zone [14], showed a loss ~10^-2 dB/mm. The typical lengths of the taper transitions used in this work were of the order of several tens of millimetres in order to ensure good adiabaticity and hence low loss. To avoid contamination of the fiber surface, the metallic piping used to deliver the gas was cleaned with hydrochloric acid, filters were inserted in the gas delivery line, and the best available gas purity was used (zero grade, 100% pure for both Oxygen and Isobutane). Two sets of regulators were used to ensure a constant gas flow and guarantee uniform taper heating during all stages of the process. The whole fabrication rig was housed in a plexiglas box to avoid air turbulence.

The total loss of the taper (fiber input facet to fiber output facet) was measured both during the fabrication process (dynamic) and at the end of the process once the burner had been removed (static). Light at 1.55μm from a laser diode was launched into the fiber and the transmitted signal was measured by connecting the output fiber to an InGaAs detector using a bare fiber adaptor. At least 1 m of coated fiber was left on the collection arm between the taper and the detector to eliminate any unwanted forward scattering from the taper launch. The temporal power stability of this measuring system was checked by analysing the transmission in an untapered fiber fixed on the fabrication rig for more than one hour prior to the taper manufacture. Power fluctuations of less than 0.1% over one hour were recorded.

3. Loss measurement

Figure 1 shows examples of loss measurements on standard telecom fibers pulled to a radius of 130–150nm. The filled circles represent the loss of the previous state-of-the-art silica tapers recently reported in the literature [6]. Figure 1 clearly shows the improvements that can be achieved using a clean, stable flame to fabricate the nanotapers. The measured losses are consistently smaller than those previously reported across a wide rage of taper diameters. The benefit is >10dB at the smallest diameters, meaning that the loss per unit of length is at least ten times smaller for our fibers. (It should be noted that the measured loss is an overestimation of the nanotaper loss per unit length since our estimate also includes contributions from the insertion and extraction loss of the nanotaper. The accurate fabrication of short adiabatic tapers could minimise these contributions.)

Figure 1 presents the loss measurement for 3 different nanotapers. Taper 1 was fabricated at a time that the burner gas flow was not completely stabilised. We believe that the increased loss observed at r~480nm arose due to a change in the flow properties at the stage of the tapering and is reflected by a non-uniform nanotaper diameter. Taper 2 was fabricated with a fully stabilized flame and represents the minimum loss that we can currently reliably achieve within our set-up.

 

Fig. 1 Comparison between specific losses of the tapers fabricated in this paper and the data reported in literature [6]. Dynamic and static losses refer to measurements performed during the nanotaper fabrication and after it, respectively.

Download Full Size | PDF

Further data concerning the reproducibility of taper loss for a taper radius of 375nm is presented in Table 1 which shows that the fabrication process is reproducible and emphasises the importance of allowing the flame to properly stabilise before commencing the tapering. A stable flame can halve the total losses.

Table 1. Static loss measurement of nanotapers with r=375nm.

View Table

It is particularly interesting to note the difference in dynamic and static loss at r=130nm. The dynamic loss in tapers 1 and 2 is greater than 2dB/mm, while the static one is approximately 0.9dB/mm. We believe that this difference might be explained by the high bending loss experienced by the fiber tapers during fabrication. Even if the curvature induced in the nanotapers by the burner flame is very small, the light propagating in the nanotapers is very weakly guided and it would leak out in case of any temporary bending due to the gas flow. Presumably also the presence of the flame/gas in the evanescent field could contribute to an additional loss. Such a discrepancy between static and dynamic loss measurements was not observed in taper 3 at r=440nm, where the sensitivity of the nanotaper to bending losses is much reduced. This hypothesis is also supported by the small fiber V-number (V~0.54 for r=130nm), defined as in reference [15]:

In fact the fiber supports only one mode for V-values smaller than the V_cut-off=2.405 and the closer V is to V_cut-off, the more strongly bound the mode is [2,16]. For V<1 the mode is weakly bound and a considerable fraction of the power is propagating outside the guiding structure [17].

4. Beam Propagation

When r is smaller than the wavelength λ of the propagating beam the geometric optics analysis is no longer valid and the description of light propagation requires the exact solution of Maxwell’s equations [18]. Simulations have been performed to solve the exact propagation equations [19] for r=130nm. They showed that in the region of minimum waist more than 99.99% of the modal power propagates outside the silica and the intensity decreases to 1/10 of its maximum value at a distance of as much as ~70μm from the nanotaper. Figure 2 compares the intensity profiles of the modes propagating in nanotapers with waists of r=130nm, 150nm and 500nm. It is clear that the mode propagating in the nanotaper spreads for small diameters. The inset in Fig. 2 shows the diameter at which the intensity decreases to 1/10 of its maximum value versus the nanotaper radius. For r<250nm the intensity spread increases more rapidly than exponentially. In this range of small taper waists, the taper transmission is very sensitive to the local environment. This property might be extremely useful for the development of no contact nanodevices and novel sensors to detect micro/nanoparticles.

 

Fig. 2. Intensity distribution of the propagating mode for three different nanotaper sizes. Inset: Simulations showing the distance from the nanotaper at which the intensity decreases 10dB.

Download Full Size | PDF

5. SEM Picture

In order to check the taper uniformity and its diameter, SEM pictures of the nanotapers were taken. Figure 3 presents a picture of a nanotaper with a diameter of ~320nm. The nanotaper has been fixed to a conductive carbon substrate and has not been coated to minimize uncertainty in the diameter measurements. The picture shows how tapers of submicron size are easily fabricated with this simple set-up.

 

Fig. 3. SEM picture of a nanotaper with 160nm radius.

Download Full Size | PDF

The nanotaper diameter has been sampled along the taper length and shows good uniformity with diameter fluctuations Δd smaller than 10nm measured over a taper length L=20mm, giving an estimate for Δd/L<5·10^-7. The good taper uniformity is a fundamental requirement for low-loss propagation and we believe it is also responsible (together with surface cleanness) for the excellent data presented in Fig. 1.

6. Conclusions

In conclusion we have reported the fabrication of what as far as we are aware is the longest and lowest-loss submicron taper ever produced. The taper has been fabricated in a simple fashion using a conventional taper rig with a ‘clean and stable’ burner flame. The tapers produced this way are found to have excellent uniformity and low-losses and further the possibility of developing nanophotonic components for various application areas including optical communications, sensing, lasers, biology and chemistry.