A low-loss polymer medium to interconnect 2 single mode optical fibers is developed and characterized. It consists of a so-called self-written waveguide (SWW) formed by illuminating a photosensitive polymerization mix with light emanating from the fiber, after which the exposed part polymerizes. Depending on the material system used, this waveguide can have a step index or graded refractive index profile. The fabrication process and its effect on the waveguide performance are explained using an empirical model and afterwards experimentally verified. This approach enables easy process monitoring and optimization, effectively resulting in total insertion losses below 0.3 dB for a single mode fiber-SWW-fiber transition at 1550 nm.
© 2014 Optical Society of America
Self-written waveguides (SWWs) are formed by immersing an optical fiber end-face in a photosensitive polymerization mix and illuminating the mix with light emanating from the fiber, after which the exposed part of the mix polymerizes. Mostly (near-)UV light is used for this process, although visible [1, 2], and near-infrared light curing of the polymer structure has also been reported . This polymerization reaction causes the refractive index (RI) to increase locally, thus creating a “self-written” polymer optical waveguide structure, aligned with the fiber’s end-face. Depending on the type of fiber from which the light is launched (i.e. multimode or single mode fiber), the UV-sensitive material and the exposure parameters, the properties of the polymer SWW structures differ.
Several researchers have reported this process for a variety of applications, where the SWW is typically used as an intermediate structure connecting two optical entities. Because of the simpler fabrication process, mostly multimode SWWs have been reported, e.g. for creating sensors , coupling structures [2, 5–7], fiber tips , lasers , and fiber-to-fiber connections . Within this last category, also the use of single mode SWW structures has been studied because of the popularity of single mode (SMF-28) optical fibers in telecom applications. In order to avoid excessive insertion losses in such a fiber-to-fiber connection, the intermediate polymeric SWW structure should be precisely optically matched to the fiber, which occurs for specific RI profiles of the polymer medium between the 2 optical fibers, which will be illustrated in the current paper.
Some studies have already reported on the experimental characterization of these fiber-SWW-fiber connections [10–12], and on modeling the underlying chemistry and evolution of the RI during photopolymerization [13–16], and even on modeling the formation process of an SWW [17–19]. Therefore, the aim of the current paper is not to model the underlying photopolymerization principles, but to link the observed RI evolution during the SWW fabrication process with the optical performance of these structures. To this end, we have constructed a simple empirical model that helps to optimize the SWW fabrication procedure in a practical way.
Two different types of commercially available material systems were studied, i.e. organically modified ceramic (Ormocer, micro resist technology GmbH) and acrylate materials (NOA 68, Norland Products Inc.). Ormocer materials are widely used for optical interconnects because of their low absorption losses (around 0.5 dB at 1550 nm), ease of use and high stability . NOA 68 is sold as an optically clear UV curable adhesive but has also been used for holographic recording. It has been reported that a permanent RI modulation can be inscribed when exposed with a spatially modulated UV beam .
In the case of Ormocer, a combination of 2 material variations was used, i.e. OrmoCore and OrmoClad, to form the core and the cladding of the SWW. In the case of NOA 68 on the other hand, the single material was used to fabricate the SWW, but relying on UV exposure with a spatially varying intensity, the required RI difference between the core and cladding was achieved. Both approaches resulted in consistent and low total insertion losses below 0.3 dB for single mode fiber-SWW-fiber transitions at 1550 nm. However, both material systems have their advantages; the 2-material system using Ormocer was selected because it allows an easy study of the polymerization curing kinetics since the formation of the core and cladding is decoupled, while the single material system using NOA 68 was selected because it involves a more practical technology for use in industrial applications. Analogously, the empirical model and analysis presented below describing the SWW formation is focused on the 2-material system first, and afterwards the principle is also applied for the single material system.
2. Step-index SWW based on Ormocer
2.1. Material and SWW formation principle
To create an Ormocer based SWW, first the core is fabricated and in a second step it is surrounded with a lower RI material to act as the cladding. For the cladding, the commercially available OrmoClad variant of Ormocer is used and for the core, an in-house prepared mix of OrmoClad and OrmoCore (with different RI) is used to achieve a desired core-cladding RI difference, and this mixture is hereafter referred to as “OCore”. The OCore material was prepared by mixing OrmoClad and OrmoCore, for 2 hours at room temperature. After mixing, the material was left to rest for 1 day to let the mixing induced air bubbles escape. To achieve a 0.005 higher RI compared to OrmoClad, the OCore mixture was prepared in a 65:35 weight% OrmoClad: OrmoCore ratio.
2.2. Empirical model
The empirical model consists of a cylindrical polymer waveguide core (i.e. the SWW) located between the perpendicularly cleaved end-faces of 2 SMF-28 standard telecom single mode optical fibers and was implemented and simulated in Lumerical FDTD Solutions (Fig. 1). The fundamental TE mode (at 1550 nm) is launched from the leftmost fiber, propagates through the central SWW section and finally enters the rightmost fiber where a monitor is placed to record the field. The amount of light which couples back into the fundamental fiber mode (of the rightmost fiber) is then obtained from the data in this monitor to determine the total transmitted power (T) into the rightmost fiber. The insertion loss expressed in dB is then calculated as 10 log10 (1/T). The goal of this simulation is to determine the insertion losses as a function of RI difference between core and cladding of the SWW which is linked to a specific time instance during the polymerization process.
It is assumed that upon UV-illumination through the fibers, the polymerization and therefore also the RI increase of the SWW core (ncore) occurs uniformly. Since the non-exposed, surrounding material or cladding remains unpolymerized and exhibits a lower RI (nclad), a waveguide is formed between the end-faces of the SMF-28 fibers. Depending on the exact RI difference (Δn = ncore - nclad, which evolves during UV-illumination), the fiber-waveguide-fiber insertion loss differs greatly because of a better or worse optical mode matching between the fiber and polymer waveguide section. This phenomenon allows inspection of the SWW polymerization process accurately by monitoring this insertion loss over time. After the formation of the SWW core, the unpolymerized surrounding material is removed and a different cladding material is applied. This material is then exposed to a uniform UV-illumination, similarly leading to insertion loss variations as the material increasingly polymerizes.
The wavelength of operation was 1550 nm (telecom wavelength); the diameter of the SWW core, 6 µm, was obtained from experiments as explained below, and the RI of all materials used in the model were obtained from in-house measurements. To record the RI of these materials (at 1550 nm) before and after polymerization, a separate setup (see Fig. 2(a)) was used inspired on the concept reported by A. Cusano et al. . The measuring principle is based on the RI dependent Fresnel reflections at a perpendicularly cleaved fiber end-face embedded in the polymer material under test. Uniform power densities of 7 mWcm−2, 7 mWcm−2 and 2 mWcm−2 (wavelength 365 nm) were applied for Ormoclad, OCore and NOA 68 respectively. These are the same parameters used during the SWW fabrication. The results in Fig. 2(b) show a RI of 1.506 and 1.511 for the OrmoClad and OCore material in unpolymerized state (at 1550nm), both increasing with 0.01 after polymerization.
These values were used as input data for modeling the insertion loss of the fiber-SWW-fiber structure during polymerization. In a first step, the polymerization of the SWW core was simulated by keeping the cladding RI constant at 1.511 (unpolymerized OCore material) and sweeping the core RI from 1.511 to 1.521 (completely polymerized OCore material). In a second step, the fabrication of the SWW cladding was simulated by keeping the already fabricated core RI constant at 1.521, but sweeping the cladding RI between 1.506 (unpolymerized OrmoClad material) and 1.516 (completely polymerized OrmoClad material). These simulations were performed for different SWW lengths and the resulting graphs depicting the fraction of transmitted optical power T are shown in Fig. 3.
During the core fabrication, it is clear that the minimum insertion loss of about 0.07 dB (98.4% transmission) is achieved when the partially polymerized SWW reaches a RI which is 0.004-0.005 higher than the surrounding unpolymerized material, slightly depending on the SWW length. In this case, a maximum overlap is achieved between the optical modal profile in the SMF-28 fiber and in the polymer waveguide section. However, upon completion of the core formation process, the insertion loss rises again because the RI difference becomes too high to support ideal single mode propagation. Although the SWW is still single mode up to a RI difference of 0.012, there clearly is a significant influence of the increasing mode mismatch for higher RI differences. Nevertheless, after substituting the unpolymerized OCore material with OrmoClad, the insertion loss finally evolves to a minimum value owing to the ideal RI difference (0.005) of the selected core and cladding materials upon complete polymerization. Also note that for shorter SWWs, the maximum in the T curves as a function of the RI difference is broader, resulting in a larger processing window, meaning that shorter SWWs are more tolerant to process variations in RI, provided the pre-alignment of the fibers is perfect.
2.3. Experimental verification
Since insertion losses are easy to measure, these simulations can be linked to experimental data and used for process optimization. However, to allow a direct comparison, the simulated insertion loss curves need to be expressed as a function of time, which is achieved by approximating the RI increase during UV-exposure by an exponential dependence with time :
A fiber-based setup was constructed for recording the insertion loss dynamically and in real-time during the different SWW fabrication steps, as illustrated in Fig. 4.The current study is limited to characterizing at the popular 1550 nm telecom wavelength due to the availability of suited fiber-pigtailed components. All fiber connectors were angled polished (FC/APC type) to avoid back-reflections and they were cleaned, fixed and then kept unaffected for the complete duration of the experiment to minimize measurement inaccuracies.
Figure 4(b) illustrates the different steps in the SWW formation process using the 2-material system with Ormocer. (1) For calibration, the 2 fibers in the test region were joined using a fusion splice process and the corresponding insertion loss value was set to 0 dB, serving as a reference for all successive measurements. (2) Then, the splice was removed, the fibers were recleaved, their end-faces separated by a certain distance, e.g. 50 µm and the unpolymerized OCore material was applied in between. (3) At this point, the SWW fabrication was initiated by irradiating the OCore material with 1 µW, 405 nm laser light emanating from both fibers for 10 s. (4) The unpolymerized OCore material was subsequently dissolved and removed using acetone. (5) Finally, OrmoClad material was applied surrounding the SWW structure and (6) polymerized using a uniform UV flood exposure (Hamamatsu LC8 with 365 nm filter, 30 s at 7 mWcm−2).
2.4. Results and discussion
The evolution of the insertion loss during the formation of a 50 µm and 100 µm long Ormocer based SWW is plotted in Fig. 5, together with the corresponding simulated curves. During the polymerization of the core, a very similar trend is observed between the experimental and modeled data, and the main difference is a slight offset in insertion loss which can be attributed to non-perfectly aligned fibers in the experiment. Furthermore, microscope and SEM images of the SWW (when the unpolymerized OCore material was removed), as shown in Fig. 6, show a smooth cylindrical structure with a 6 µm core diameter as assumed for the theoretical model. Note that for capturing this image, a longer SWW structure was prepared allowing easier visualization. Figure 7 shows a millimeters long structure, generated using the same parameters, but without the second fiber, illustrating the uniformity of the fabricated SWWs along their length.
During the polymerization of the cladding, again experiment and simulation yield similar curing kinetics (see Fig. 5(b)), although the offset in insertion loss is more pronounced. In the previous case of the core formation, only the initial alignment of fibers resulted in slightly higher insertion loss as predicted, while in this case, also the exact RI of the already formed core is important. Indeed, if the RI difference between core and cladding is slightly different than expected, this will have an effect on the insertion loss, as can be observed from the sensitivity of the simulated curve, especially for large Δn (see Fig. 3(b)). Remark that in Fig. 5(b) and Fig. 5(d), only the first 6 s of the cladding formation is plotted to zoom in on the relevant part of the process, while the total UV exposure time was 30 s, after which final insertion losses below 0.3 dB were obtained, both for 50 µm and 100 µm long SWW connections. Comparing this value with the minimum achievable insertion loss in case of perfect alignment (i.e. about 0.07 dB according to the simulations), the insertion loss due to fiber misalignment was limited to maximum 0.23 dB in all cases.
These experiments illustrate that monitoring the insertion loss during core and cladding formation can provide information on the instantaneous RI of the polymer and therefore also on the curing kinetics of this self-written waveguide polymerization process. In search for a structure with the lowest possible insertion loss, tuning of RI is indeed important and in this context it is advantageous to be able to tune core and cladding separately as described above using the Ormocer 2-material system. However, a major drawback of this technology is the need to remove the uncured (core) material to replace it with another (cladding) material, which is not practical in many industrial applications. This can be overcome with the single material approach using NOA 68 material, as described in the following section.
3. Graded-index SWW based on NOA 68
3.1. Material and SWW formation principle
The rationale of this approach is to write the core using UV light emanating from the fiber end-faces and performing a flood exposure to form the cladding. Due to the specific composition of the NOA 68 material, the final RI profile after polymerization is correlated with the UV light intensity distribution during photopolyemerization. This material property has been studied extensively for application in holographic recording and it was reported that this RI modulation can be caused by a different degree of monomer conversion or by diffusion of monomers (material transport) between regions exposed with different UV light intensities [21, 23, 24]. This mechanism can therefore also be exploited for the fabrication of SWWs. Alternatively, a solution consisting of two kinds of polymers with different polymerization reaction mechanisms could be used for the fabrication of SWWs without a washing step , but this generally requires dedicated, non-commercially available materials.
In our approach, the UV light beam emanating from the SMF-28 fiber combined with the UV flood exposure give rise to a non-homogeneous UV light exposure yielding a graded-index RI region between the core and cladding. Since this graded-index region ensures the waveguiding in the polymer, no substitution step is needed for this material. We have found no evidence that the OrmoClad and OCore materials exhibit this property, therefore requiring a material substitution process to obtain a sufficient RI difference, which consequently leads to a step-index RI distribution in the waveguide region.
3.2. Empirical model
Since the RI profile in the NOA 68 is related to the exposure intensity distribution, the near-field profiles at the end-face of the SMF-28 fiber guiding the 405 nm light were recorded using a Spiricon SP620U beam profiler with C-NFP adaptor using a 40x objective. Figure 8 shows these near-field profiles for 1 µW (used for fabricating the Ormocer based SWWs) and 10 µW (used for fabricating the NOA 68 based SWWs) of optical power measured at the fiber end-face. The SMF-28 fiber is not single mode at the used 405 nm exposure wavelength (the cut-off wavelength of the fiber is about 1260 nm) and therefore the effect of higher order modes is clearly visible. Although the beam has no ideal Gaussian profile, it was found that the higher order modes generated in the fiber reach a stable distribution and these profiles were reproducible during different experiments.
The influence of the graded index profile on the insertion losses through the polymer section was determined from simulations. An analogous model as described above was employed but now with a graded RI profile in the SWW region. This RI profile was determined by fitting a Gaussian curve through the actually measured RI data which is shown below. Then, simulations were performed for varying amplitude of this Gaussian profile, i.e. the difference in RI in the center of the waveguide compared to the RI of the cladding, far away from the center.
The resulting graphs depicting the evolution of the transmission T as a function of varying RI difference are plotted in Fig. 9. The maximum RI difference of 0.016 corresponds to the difference between completely polymerized and unpolymerized NOA68 material. It can be seen that the optimum RI difference was 0.006-0.007, again slightly depending on the SWW length.
3.3. Experimental verification
The insertion losses as a function of process time were similarly recorded using the setup shown in Fig. 4, but steps (4) and (5) were omitted and step (3) and (6) were executed simultaneously. Through both fibers, 10 µW, 405 nm laser light was launched for 30 s and simultaneously the complete structure was polymerized using a uniform UV flood exposure (Hamamatsu LC8 with 365 nm filter, 30 s @ 2 mWcm−2)
Using these optimized exposure parameters for fabricating the graded index region, the core-cladding difference becomes optimum, obtaining high-quality waveguides with low insertion losses, comparable to the Ormocer material approach.
3.4. Results and discussion
Since the core and cladding are formed from the same material and therefore fabricated simultaneously in this approach, it is more difficult to predict the evolution of the core-cladding RI difference as a function of exposure time during polymerization. Consequently, the recorded data is more difficult to compare with the optical simulations, which only take into account the insertion loss as a function of the RI difference. However, the UV power density launched through the fibers (about 12 W/cm2) is much higher than the UV power density used for the flood exposure (2 mW/cm2) and therefore the time scale on which the core is formed is shorter than the time scale on which the cladding is formed in the polymer material. Therefore, the effect of both processes can be separated to a certain extent, when analyzing the experimental data, see Fig. 10.Again, a dip in the insertion loss is observed and occurs after about 1 s of exposure, when the waveguide core-cladding RI difference for the NOA 68 material passes through the optimal value. Afterwards, the insertion loss increases again since the RI difference with the almost unpolymerized cladding becomes too high and consequently the optical mode mismatch with the SMF-28 fiber increases. Then, after about 3 s of exposure, the loss decreases since the slower flood exposure process starts polymerizing the cladding so that the core-cladding RI difference gradually reduces to the optimal value and finally leads to a minimum insertion loss value. The RI difference, which determines the optical insertion loss, depends therefore on a combination of the UV dose launched through the fiber and dose used for the flood exposure. Since there is a link between the RI difference and the insertion loss, the SWW structure can be optimized in situ during fabrication: increasing the UV flood exposure dose or the dose launched through the fiber will decrease and increase the core-cladding RI difference respectively.
The typical final RI profiles of both the Ormocer and NOA 68 based 50 µm long SWWs were measured after fabrication to verify both polymerization mechanisms. Therefore, a transverse interferometric method (using an Interfiber Analysis IFA-100 optical fiber analyzer) was employed, assuming the SWW is circularly symmetric . The resulting data is shown in Fig. 11, plotting the absolute RI difference between core and cladding as a function of the radial position. The spike in the graph at radial position 0 µm results from the calculation algorithm artifact at this position and is therefore not physical. As expected, a clear difference can be observed between SWWs fabricated using both material systems, illustrating the nearly step-index profile in the case of the 2-material system with substitution step (Ormocer) and the more gradual graded-index profile in the case of the single material system (NOA 68). Furthermore, we found that the RI profiles did not vary significantly along the SWW length, for these 50 µm long structures.
This paper shows that a low-loss self-written polymer waveguide medium can be fabricated to interconnect 2 single mode telecom fibers, based on modeling and optimization of the curing kinetics for either a single material (NOA 68), or 2-material (2 types of Ormocer) system. By monitoring the insertion loss through the waveguide transition medium during fabrication, and comparing with an empirical model, the polymerization progress and refractive index can be estimated. These are 2 parameters defining the optical performance of the polymer waveguide medium and therefore the final optical insertion loss of the fiber-waveguide-fiber structure. The 2-material approach was introduced allowing an easy study of the curing kinetics, the formation and optimization of low-loss self-written waveguide structures, and the single material system was introduced targeting the use in industrial applications. For both approaches, consistent total optical insertion losses below 0.3 dB were obtained for single mode fiber-SWW-fiber transitions at 1550 nm.
This work was conducted in the framework of the project EP2CON, funded in part by the Institute for the Promotion of Innovation by Science and Technology (IWT), Flanders, Belgium. This work is also partly supported by the Research Foundation-Flanders through the project GA04711N.
References and links
1. M. Kagami, T. Yamashita, and H. Ito, “Light-induced self-written three-dimensional optical waveguide,” Appl. Phys. Lett. 79(8), 1079–1081 (2001). [CrossRef]
2. H. Ozawa, Y. Obata, Y. Mimura, O. Mikami, and T. Shioda, “Self-written waveguide connection across diced waveguide gaps,” IEEE Photonic. Tech. L. 18(7), 880–882 (2006). [CrossRef]
3. M. Kagami, T. Yamashita, M. Yonemura, A. Kawasaki, O. Watanabe, and M. Tomiki, “Photopolymer-based three-dimensional optical waveguide devices,” Proc. SPIE 8258, 825812 (2012). [CrossRef]
4. J. S. Young and J. P. Kara, “A self-repairing polymer waveguide sensor,” Smart Mater. Struct. 20(6), 065005 (2011). [CrossRef]
5. K. Myung-Joon, M. Kanda, O. Mikami, M. Yonemura, and M. Kagami, “180 degree light path conversion device with tapered self-written waveguide for optical interconnection,” IEEE Photonic. Tech. L. 22(15), 1126–1128 (2010). [CrossRef]
6. H. Ozawa, Y. Obata, Y. Mimura, O. Mikami, and T. Shioda, “Self-written waveguide connection between VCSEL and optical fiber with 45 degree mirror using green laser,” IEEE Photonic. Tech. L. 18(3), 532–534 (2006). [CrossRef]
7. N. Hirose, T. Yoshimura, and O. Ibaragi, “Optical component coupling using self-written waveguides,” Proc. SPIE 4640, 64–71 (2002). [CrossRef]
8. S. Jradi, O. Soppera, D. J. Lougnot, R. Bachelot, and P. Royer, “Tailoring the geometry of polymer tips on the end of optical fibers via control of physico-chemical parameters,” Opt. Mater. 31(4), 640–646 (2009). [CrossRef]
9. K. Yamashita, A. Kitanobou, M. Ito, E. Fukuzawa, and K. Oe, “Solid-state organic laser using self-written active waveguide with in-line fabry-perot cavity,” Appl. Phys. Lett. 92(14), 143305 (2008). [CrossRef]
10. K. W. Cheng, M. A. Uddin, H. P. Chan, and S. C. Chan, “Optical alignment tolerances in double-side irradiated self-written waveguide-induced fiber arrays packages,” Opt. Commun. 283(13), 2669–2675 (2010). [CrossRef]
11. O. Sugihara, H. Tsuchie, H. Endo, N. Okamoto, T. Yamashita, M. Kagami, and T. Kaino, “Light-induced self-written polymeric optical waveguides for single-mode propagation and for optical interconnections,” IEEE Photonic. Tech. L. 16(3), 804–806 (2004). [CrossRef]
12. K. Yamashita, T. Hashimoto, K. Oe, K. Mune, R. Naitou, and A. Mochizuki, “Self-written waveguide structure in photosensitive polyimide resin fabricated by exposure and thermosetting process,” IEEE Photonic. Tech. L. 16(3), 801–803 (2004). [CrossRef]
14. K. Dorkenoo, A. J. van Wonderen, H. Bulou, M. Romeo, O. Crégut, and A. Fort, “Time-resolved measurement of the refractive index for photopolymerization processes,” Appl. Phys. Lett. 83(12), 2474–2476 (2003). [CrossRef]
15. K. D. Dorkenoo, F. Gillot, O. Crégut, Y. Sonnefraud, A. Fort, and H. Leblond, “Control of the refractive index in photopolymerizable materials for (2+1)D solitary wave guide formation,” Phys. Rev. Lett. 93(14), 143905 (2004). [CrossRef] [PubMed]
16. T. Yamashita, M. Kagami, and H. Ito, “Waveguide shape control and loss properties of light-induced self-written (LISW) optical waveguides,” J. Lightwave Technol. 20(8), 1556–1562 (2002). [CrossRef]
17. A. Anderson and K. Peters, “Finite element simulation of self-writing waveguide formation through photopolymerization,” J. Lightwave Technol. 27(24), 5529–5539 (2009). [CrossRef]
18. T. Yoshimura, K. Wakabayashi, and S. Ono, “Analysis of reflective self-organized lightwave network (r-solnet) for z-connections in 3-D optical circuits by the finite-difference time-domain method,” IEEE J. Sel. Top. Quant. 17(3), 566–570 (2011). [CrossRef]
19. M. Hocine, R. Bachelot, C. Ecoffet, N. Fressengeas, P. Royer, and G. Kugel, “End-of-fiber polymer tip: Manufacturing and modeling,” Synth. Met. 127(1-3), 313–318 (2002). [CrossRef]
20. U. Streppel, P. Dannberg, C. Wächter, A. Bräuer, L. Fröhlich, R. Houbertz, and M. Popall, “New wafer-scale fabrication method for stacked optical waveguide interconnects and 3D micro-optic structures using photoresponsive (inorganic–organic hybrid) polymers,” Opt. Mater. 21(1-3), 475–483 (2003). [CrossRef]
21. L. Goldenberg, O. Sakhno, and J. Stumpe, “Application of norland adhesive for holographic recording,” Opt. Mater. 27(8), 1379–1385 (2005). [CrossRef]
22. A. Cusano, A. Cutolo, M. Giordano, and L. Nicolais, “Optoelectronic refractive index measurements: Application to smart processing,” IEEE Sens. J. 3(6), 781–787 (2003). [CrossRef]
23. M. R. Gleeson, S. Liu, R. R. McLeod, and J. T. Sheridan, “Nonlocal photopolymerization kinetics including multiple termination mechanisms and dark reactions. Part II. Experimental validation,” J. Opt. Soc. Am. B 26(9), 1746–1754 (2009). [CrossRef]
24. B. Pinto-Iguanero, A. Olivares-Pérez, and I. Fuentes-Tapia, “Holographic material film composed by norland noa 65® adhesive,” Opt. Mater. 20(3), 225–232 (2002). [CrossRef]
25. M. Kagami, T. Yamashita, M. Yonemura, and T. Matsui, “Light-induced self-written optical waveguides,” IEICE Trans. Electron. E90-C(5), 1061–1070 (2007). [CrossRef]
26. A. D. Yablon, “Splice measurement and characterization,” in Optical Fiber Fusion Splicing (Springer Berlin Heidelberg, 2005), pp. 183–202.