1. Introduction

Optical microfibers have cross-sectional dimensions of the order of a few micrometers with potential importance as building blocks for microphotonic circuits and devices, leading to diverse applications. Recently, several publications have reported the demonstration of evanescently coupled devices such as couplers [1] and optical resonators [2] based on microfibers as well as nanofibers with sub-micron dimensions [3]. These illustrate the versatility as well as the challenges of fabricating functional optical circuits using microfibers. Microfibers can be drawn from standard telecom fibers (such as single-mode fibers) via a wide range of techniques. Examples include flame-assisted taper drawing [4], and CO₂ laser-based tapering [5]. Microfibers are advantageous for photonic circuits as the optical losses associated with these fibers can be significantly lower than those of conventional lithographically-defined photonic devices [6]. This can be attributed to excellent surface smoothness and structural uniformity of the microfibers. Integration of these narrow tapered fibers into photonic systems such as lab-on-a-chip devices [7] would require a suitable technique for forming permanent interconnects between the silica wires, i.e., by splicing them together as is commonly done for standard optical fibers with commercially available fusion splicers. Recent reports have made use of the attractive electrostatic as well as inter-atomic van der Waal’s forces [8] associated with the reduced physical dimensions of the microfibers for forming micro or nanofiber couplers [3] as well as for manipulating the fibers into loops and knots for demonstrating resonances in optical transmission. Fiber-to-fiber optical coupling, using the evanescent field that is strong in very small fibers [2], by putting the fibers in lateral contact with each other or with themselves (as in the case of self-touching loops [9]), does not result in mechanically stable connections. A method that provides high efficiency optical coupling with mechanical robustness is required such as fusion splicing, which has been demonstrated for a miniature fused fiber coupler [10]. Conventional fusion splicers are not designed for splicing together two microfibers as they have no provision for holding in place fibers of such small dimensions, and the magnitude of the arc currents is not optimized for fusing micron-sized fibers. In this paper, we report the fabrication of tapered step-index fibers that are a few μm thick and we demonstrate a low loss splice between two such fibers.

2. Experimental methods

We made our microfibers by tapering based on a standard heat-and-draw technique, using a CO₂ laser as the heat source. A galvanometer-controlled mirror scanned the beam back and forth at 60 kHz over a small section of the fiber creating a uniform ‘hot’ zone, and computer-controlled stepper motors provided constant tensioning in a self-regulated manner [11]. The starting medium for the tapered waveguides was standard telecom fiber (Corning’s SMF-28), which had an initial outside diameter of 125 μm, and the subsequent microfibers had a final outside diameter of about 4 μm. The tapered fibers consisted of an untapered region, a conical taper transition region where the diameter decreased along the length in a continuous manner, and a tapered section that was in the range 4–10 centimeters long which was connected to untapered fiber through another transition region. For achieving relatively uniform diameter along the taper length, the pulling speed as well as the output power of the laser was controlled carefully to optimize the process [12]. For our experiments, we cleaved the fiber at the start and end of the narrow uniform segment and used free-space input and output coupling.

We characterized the microfibers by placing a few centimeters of the fiber on a slab of silica aerogel. Silica aerogels are nanostructured materials derived from sol-gels through a critical-point drying process which gives it high porosity and an exceedingly low index of refraction of ~1.03 [13]. We placed a 4-cm long section of the microfiber on a piece of aerogel and coupled ~3 mW from a He-Ne laser into it. We placed a collimating objective at the output end and determined that the propagation loss is ~ 0.2 dB/cm. We also measured the right angle scattered light emanating from the fiber under these conditions with a CCD camera as is commonly done for channel waveguides [14]. From this, we obtained a propagation loss of ~0.18 dB/cm for our microfiber, which is close to that obtained by the free-space coupling experiment.

 

Fig. 1. Graphical representation of the cutback measurements.

Download Full Size | PDF

We note that this microfiber loss is about 10,000 times higher than the fiber from which it is drawn, i.e., Corning’s SMF 28 (which has a propagation loss of ~1.8 dB/km at 633 nm). This may be attributed to effects of contamination from the local environment or degradation occurring from the outer surface of the microfiber being exposed to the ambient surroundings. Also, as has been studied for nanowires in a previous publication [15], such fibers may develop nanoscale cracks or fissures if they are not surrounded by a protective cladding. Since microfibers can exhibit a significant evanescent field on the surface [2], the optical perturbations as a result of these mechanical faults can result in an increased overall loss of the device. Our experiments were not carried out in a clean room environment, which increased the vulnerability of the fibers to damage. Electron microscope pictures of our microfibers show the presence of sub-micron scale particles attached to the surface, and these can also contribute to transmission losses. This effect can be used to advantage in sensing applications.

We carried out some preliminary experiments for fusing together two microfibers using oxy-butane and MAPP (methylacetylene-propadiene) gas torches, however, the excessive convective air currents prevented accurate fiber alignment [16]. Next, we tried using an electric micro-arc that we generated between two tungsten fusion splicer tips and noted that this approach had different problems. Once the micro-arc was generated, charge accumulations developed on the exposed microfiber tips resulting in strong repulsion between them, which hindered the alignment necessary for the splice to occur [16]. Excessive degradation of the microfibers also occurred due to the fiber tips attracting dust and other forms of contamination. A viable alternative was to use a CO₂ laser, which, at a wavelength of 10.6 μm, is strongly absorbed by silica [17], and offers significant advantages such as adjustable control over the size of the fiber’s ‘hot’ zone by controlling the focused spot size of the beam, and directionality of the beam. Furthermore, since the fiber is heated exclusively by radiation, no residue is created on the fiber surface [18], and since the surrounding air is not directly heated by the laser, air convection currents are reduced. We aligned the microfibers using microscope cover slides along with glass cover slips that were cut into a desired shape to hold the two microfibers on XYZ translation stages. Once the fibers were aligned under a microscope objective connected to a CCD camera, a 3 mm wide beam from a CO₂ laser was focused down to roughly 0.8 mm using a ZnSe-coated lens with a focal length of 180 mm. Initially, we aligned the fibers end-to-end and focused the CO₂ beam on it with a power of approximately 5 watts and observed that the splices obtained this way lacked sufficient robustness and had a tendency to break at the joint. We attributed this to a lack of control over the quality of the cleaved end faces as well as over the general alignment of the ends of the fiber in the presence of atmospheric turbulence.

 

Fig. 2. Log plot of the right-angled scattered light intensity as a function of length for two spliced microfibers. The red line represents a least squares fit of slope 0.19 dB/cm. Inset shows an image of the microfiber(s) guiding light at 633nm.

Download Full Size | PDF

We adopted an alternate technique wherein instead of end-coupling the microfibers, we overlapped a section of the two fibers with an overlap length of approximately 300 μm. This approach offered a crucial advantage, namely, the microfibers tended to adhere to each other owing to strong electrostatic forces thus reducing the movements of the fiber segments suspended in air. The overlapping also resulted in a more efficient absorption of power from the CO₂ beam as the effective volume of glass in the path of the beam was greater, which meant that the beam power could be comparatively lower (~ 2–3 watts) reducing the probability of thermal damage of the microfibers. After the required alignment, it was observed that once the fiber segments absorbed enough radiation to reach their melting point, the microfiber tip that directly faced the beam, melted partially and bent down under the action of gravity to form a sharp bend. The power of the CO₂ as well as the lasing duration had to be controlled precisely to avoid the microfiber from breaking off completely. The second fiber tip softened and fused onto the first fiber at the point where bend occurred, thus forming the splice. We found that the splices obtained from this unobvious approach were significantly more mechanically robust, and in addition did not introduce significant splice loss. We characterized these splices via two methods: cut-back technique, and imaging the right-angle scattered light. We fused together two microfibers in the manner described earlier and once again, placed a section of the fiber containing the splice on an aerogel substrate. A standard way of characterizing a splice is in the transmission mode by the ‘cutback’ technique [19] wherein light is injected at one end of the fiber containing the splice and the transmitted power is detected at the output end. The fiber is then subsequently excised to shorter lengths with great care and the transmitted power is measured again. The ‘cutting back’ of the fiber is repeated till the splice is eliminated from the fiber segment and the propagation loss is estimated by the following relation;

where, α is the optical attenuation (measured in dB/length), P _initial is the output power measured prior to shortening the length of the fiber, P _cut-back is the output power after performing the cut-back, and ΔL is the length of the excised segment. This experiment had to be carried out with utmost care, given that our fiber had dimensions corresponding to just a few microns and was extremely susceptible to physical damage that could seriously affect the transmission measurements. Figure 1 depicts the results from our cutback experiments and the plot shows that the splice contributed a maximum loss of ~0.32%. As depicted in Fig. 2, we also measured the scattered light intensity decay along the length of the fiber, the slope of which yielded an overall optical attenuation of ~ 0.19 dB, which is essentially identical to that of a single microfiber strand and indicated that the presence of the splice did not introduce significant losses.

 

Fig. 3. Simulated 3D-FDTD propagation results for two successfully spliced microfibers with a bent segment at the joint (black arrow indicates its location). The power contained in the fundamental mode as well as the total power has been plotted along the propagation length. The picture on the left is a microscope image of the successfully spliced microfibers.

Download Full Size | PDF

Loss estimations may be also be deduced from numerical methods such as the coupled mode theory [17] and the beam propagation method (BPM). We have used a commercially available BPM software (BeamPROP, RSoft Inc.) based on a three-dimensional finite difference time domain (3D FDTD) method to study the mode propagation through our microfibers as also to gain insight into the impact of the splice morphology on the optical properties of the fiber. Figure 3 shows the results of a simulation performed on two microfibers joined together with the bent segment at the juncture, similar to our experimentally obtained splice. The simulation was carried out by launching a fundamental Gaussian mode at the input of the fiber. As can be seen from the contour plot, at the microfiber junction, some mode-mixing and other perturbations occur due to the presence of the added structure at the intersection between the two fibers, however, the total loss (after propagating to the end of the second fiber) is only 0.3%. This can be explained on the basis of the length of interaction between two optical elements that are in physical contact with each other. Essentially, the situation is akin to that of two silica wires that intersect at angles close to 90° and barring some induced scattering, suffer minimal cross-talk owing to a lack of sufficient region of coupling between the two.

 

Fig. 4. Simulated 3D-FDTD propagation results for two improperly spliced microfibers with a ‘globule’ junction (black arrows indicate splice location). The pictures on the left are bright-field and dark-field images of the spliced microfiber(s) guiding light at 633 nm. The power contained in the fundamental mode as well as the total power has been plotted along the propagation length.

Download Full Size | PDF

We carried out another splice by adopting the more conventional technique wherein the two fiber tips are pressed against each other while softened in order to compensate for any facet irregularities (generically referred to as ‘end-stuffing’) and we noted that it was extremely hard to avoid the formation of a globule at the joint between the microfibers in order to obtain splices with sufficient mechanical strength. We simulated the modal propagation on two microfibers that were joined together through a thicker, spherical region of 10 μm diameter (structurally similar to the observed splice) and the resulting plots of the power contained in the fundamental mode as well as the total power revealed much higher optical attenuation introduced by the spliced region compared to our previous splice (Fig. 4). These results show that the overlap method that we adopted for fusing together two tapered fibers produces splices with greater mechanical strength as well as significantly lower losses (10x) compared to the conventional end-stuffing procedure that is typically used for single-mode optical fibers.

3. Conclusions

We have developed a methodology to splice together two narrow step-index fiber waveguides with diameters in the few-micrometer regime using a CO₂ laser. Our approach results in splices that are mechanically robust and exhibit minimal losses, which may assist further development and integration of tapered silica fibers into microphotonic circuits. To date, published reports of devices constructed from microfibers have been based on evanescent wave guiding which can provide coupling efficiencies that are at most ~90% under the best conditions [20]. The ability to realize low-loss splices for microfibers opens up several possibilities for fabricating a range of devices whose performance can be significantly enhanced by overcoming structural instabilities related to evanescent wave-coupling and facilitate greater micro-scale integration. Our results may further the development of all-fiber integrated microphotonic platforms. We are currently fabricating several different kinds of such devices, and the results will be reported elsewhere.