Because optical interconnects are promising technologies for overcoming bottlenecks in computing [1] such as clock rates, input/output counts, and cooling, there has been considerable research on the implementation of optical wiring into computers [2–8]. One of the challenges of using optical interconnects is the difficulty of creating optical couplings between optical devices. Although edge coupling [9] and grating coupling [10] techniques have been developed, these require precise alignment at a number of coupling points, raising the system cost. To solve this, we propose an optical coupling technique based on a self-organized lightwave network (SOLNET), which makes use of self-focusing between write beams in photoinduced refractive-index increase (PRI) materials such as photopolymers.

Optical waveguide formation in photopolymers was first reported by Frisken [11]. A self-written waveguide was formed by the emission of a write beam from an optical fiber core into a photopolymer. Further research work has since been conducted in this field using photopolymers [12–15], glasses [16], and photorefractive crystals [17].

Meanwhile, in SOLNET [18], an attractive force between a plurality of write beams in PRI materials is utilized. This allows construction of self-aligned optical waveguides between misaligned optical devices and operates as an optical solder. A simplified SOLNET fabrication process has also been developed as reflective-SOLNET (R-SOLNET), which uses reflecting elements such as mirrors and wavelength filters [19]. Recently, the use of R-SOLNET with luminescent targets was proposed and was demonstrated both theoretically [20] and experimentally [21].

In R-SOLNET using a luminescent target, optical devices such as optical fibers, waveguides, laser diodes, light modulators, tunable wavelength filters, or photodetectors are placed into a PRI material as shown in Fig. 1. The PRI material is typically based on a photopolymer, photosensitive glass, or photorefractive crystal with a refractive index that increases upon write beam exposure. A luminescent target is deposited at the active site of one of the optical devices.

 

Fig. 1. Concept of R-SOLNET with luminescent targets.

Download Full Size | PDF

When a write beam is introduced into the PRI material from the other optical device, the luminescent target is excited by the write beam and generates luminescence. In the region where the write beam and the luminescent emission from the target overlap, the refractive index increases more rapidly than in the surrounding region. The write beam and the emission from the luminescent target interact and merge through a self-focusing effect that constructs a self-aligned optical coupling (R-SOLNET) between the optical devices. This occurs automatically, even where misalignment and core size mismatch exist between the optical devices. Although the luminescent target remains permanently attached to the optical device, signal beams with wavelengths in the near-infrared region can propagate through the target where its absorbance is low.

In a previous study on R-SOLNET using luminescent targets, three issues were found. A first issue was the poor transparency of the luminescent target used. Because tris(8-hydroxyquinolinato)aluminum (Alq3) powder dispersed in polyvinyl alcohol was used for the target, the probe beams were scattered by the powder dispersed in the structure. This prevented us from measuring the coupling efficiency. A second issue was the balance of the sensitivity of the PRI material to different wavelengths of light. The photopolymer (a composite of Norland Optical Adhesive NOA65 and NOA81) used in the previous experiments had a high sensitivity in the ultraviolet region, decreasing at longer wavelengths. When a 405-nm write beam and an Alq3-containing target, having peak luminescence $\sim 500\mathrm{nm}$, were used, a large wavelength separation of $\sim 95\mathrm{nm}$ existed between the write beam and the luminescence. The sensitivity to the luminescence was therefore very low compared with that to the write beam, reducing the targeting effect of the Alq3 target. A third issue was stability of the PRI material. Norland Optical Adhesive is not a typical material for optical waveguides, but an adhesive. It is preferable to use an optical waveguide material with better long-term stability for the R-SOLNET to improve the durability of R-SOLNET used with optical waveguides in optical interconnects.

In this Letter, we fabricated nonscattering luminescent targets located at the core edges of the optical fibers. The luminescent targets contained the dye coumarin 481, which has a luminescence peak at 470 nm. A 448-nm light source was used for write beams to reduce the wavelength separation between the write beam and the luminescence to $\sim 22\mathrm{nm}$. These conditions enabled us to measure coupling efficiency between optical fibers connected by R-SOLNET and reduce the sensitivity imbalance between the write beam and the luminescence. For the PRI material, a novel photosensitive organic/inorganic hybrid material, Sunconnect, was used. The refractive index of Sunconnect increases from $\sim 1.59$ to $\sim 1.60$ through write beam exposure. Sunconnect was developed by Nissan Chemical Industries, Ltd., as an optical waveguide material with long-term stability [22,23]. R-SOLNET experiments were performed using optical fibers with core diameters of $\sim 50\mathrm{\mu m}$, denoted by 50-μm fiber, and $\sim 10\mathrm{\mu m}$, denoted by 10-μm fiber. The former was an ordinary graded-index multimode fiber and the latter a single-mode fiber.

To enhance the targeting effect, the sensitivity of the PRI material to the luminescence should be comparable with, or higher than, that to the write beam. As Fig. 2 shows, in Sunconnect the absorption spectrum has a peak in the ultraviolet region, and the absorbance decreases at longer wavelengths, indicating that the sensitivity decreases at longer wavelengths. The photoluminescence spectrum of Sunconnect containing coumarin 481, which is used as the luminescent target, has a peak around 470 nm. Then, for a 405-nm write beam, the wavelength separation is 65 nm, and for a 448-nm write beam, the wavelength separation is 22 nm. The 448-nm write beam is therefore preferable for reducing the sensitivity imbalance.

 

Fig. 2. Absorption spectrum of Sunconnect and photoluminescence spectrum of the coumarin 481/Sunconnect luminescent target excited by the write beam at 448 nm.

Download Full Size | PDF

Figure 3 shows optical waveguides grown from 50-μm fibers in Sunconnect. The refractive index contrast between the core and cladding is expected to be $\sim 1.60/1.59$. The write beam power required to form a 1.5-mm-long optical waveguide is almost two orders of magnitude larger for the write beam wavelength, ${\lambda }_w$, of 448 nm than for ${\lambda }_w$ of 405 nm. This can be attributed to the lower absorbance of Sunconnect at 448 nm as shown in Fig. 2.

 

Fig. 3. Optical waveguides grown from 50-μm fibers in Sunconnect by write beams with wavelengths of (a) 405 nm and (b) 448 nm.

Download Full Size | PDF

In Fig. 3 a straight waveguide is shown that was formed using the 448-nm write beam. A nonuniform widened waveguide was formed for the 405-nm write beam. This difference can be explained in terms of the differing write beam absorption in the PRI material as indicted in Fig. 4. If there is large write beam absorption, the write beam may propagate only a short distance and form a short waveguide. After the waveguide formation has taken place, the absorbance in the waveguide region decreases because of decomposition of the sensitizers so that another stage of waveguide formation can start. By repeating this process, a long waveguide is formed in a stepwise manner. This may result in nonuniformity of the waveguide shape. Conversely, for small write beam absorption, the write beam can propagate over a longer distance, allowing uniform self-focusing to occur. This results in the formation of a straight waveguide with a more uniform shape. Low absorbance of the PRI material is also desirable for the formation of R-SOLNET using luminescent targets because the write beam can reach the target without attenuation by the medium so that strong luminescence can be generated.

 

Fig. 4. Schematic of the influence of write beam absorption in PRI materials on waveguide shape.

Download Full Size | PDF

To deposit a luminescent target onto the core of an optical fiber, we used the process shown in Fig. 5(a). The fiber edge was coated by coumarin 481 dissolved in Sunconnect at concentration of $\sim 0.1\mathrm{wt}.\%$. The coating was exposed to ultraviolet light from the fiber core, which cured the region just on the core. The luminescent target was obtained by wet etching the coating to remove the noncured part using a mixture of isopropyl alcohol/4-methyl-2-pentanone. Figure 5(b) shows a photograph of a luminescent target emitting luminescence by excitation at 448 nm.

 

Fig. 5. (a) Illustration of the process for luminescent target fabrication at the core edge of a fiber. (b) Photograph of a luminescent target under excitation at 448 nm.

Download Full Size | PDF

Figure 6 shows an R-SOLNET grown between a 50-μm fiber core and the edge of an opposite fiber in Sunconnect. The fiber-to-fiber distance, $L$, is 250 μm, and the lateral misalignment, $d$, is 30 μm. A 405-nm write beam is introduced from the fiber on the left. The trace of the waveguide induced by the write beam grows from the left-hand side. Another trace of the waveguide, induced by light reflected from the edge of the right-hand-side fiber, also develops. The two traces finally merge to form the R-SOLNET.

 

Fig. 6. Photograph showing an R-SOLNET grown between a 50-μm fiber core and the edge of an opposing fiber in Sunconnect by a 405-nm (100 nW) write beam for $L=250\mathrm{\mu m}$ and $d=30\mathrm{\mu m}$.

Download Full Size | PDF

As shown in Fig. 6, emission from the luminescent target is not observed. Consequently the luminescent target does not have an effect on R-SOLNET formation. The absence of luminescence is explained as follows. Because the sensitivity of Sunconnect is relatively high at 405 nm, as shown in Fig. 2, the 405-nm write beam can be of a low power for SOLNET formation. This power is not sufficient to induce observable luminescence from the target.

Figure 7 shows an R-SOLNET grown between a 50-μm fiber core and a luminescent target on a 50-μm fiber core edge in Sunconnect. The two fibers are placed at $L$ of 250 μm and $d$ of 20 μm. A 448-nm write beam is introduced from the fiber on the left. Bright blue luminescence from the target is observed. As write beam irradiation progresses, the trace of waveguide growth from the core of the fiber on the left can be observed. At the same time, the trace of waveguide growth from the luminescent target on the right is observed, which is induced by luminescence from the target. The two traces finally merge into a single waveguide and form an R-SOLNET based on a luminescent target.

 

Fig. 7. Photograph showing an R-SOLNET grown between a 50-μm fiber core and a luminescent target on a 50-μm fiber core edge in Sunconnect by a 448-nm (20 μW) write beam for $L=250\mathrm{\mu m}$ and $d=20\mathrm{\mu m}$.

Download Full Size | PDF

Figure 8 shows an R-SOLNET grown between a 10-μm fiber core and a luminescent target on a 10-μm fiber core edge, induced by a 448-nm write beam introduced from the left-hand fiber. Luminescence from the target is observed, and the trace of waveguide growth from the fiber on the left and another trace of waveguide growth from the luminescent target are observed. The two traces finally merge into a single R-SOLNET. We show that an R-SOLNET can be formed between 10-μm fibers as well as between 50-μm fibers.

 

Fig. 8. Photograph showing an R-SOLNET grown between a 10-μm fiber core and a luminescent target on a 10-μm fiber core edge in Sunconnect by a 448-nm (20 μW) write beam for $L=190\mathrm{\mu m}$ and $d=9\mathrm{\mu m}$.

Download Full Size | PDF

The coupling efficiency was measured between the 50-μm fibers connected by R-SOLNET with luminescent targets for an $L$ of 300 μm. The probe beam wavelength, ${\lambda }_{\text{Probe}}$, was 856 nm. As Fig. 9 shows, the coupling efficiency decreases with the lateral misalignment between the fibers. Using R-SOLNET the coupling efficiency increases by 10%–30%. The use of R-SOLNET also increases tolerance to lateral misalignment. These results indicate that R-SOLNET with luminescent targets has potential to realize the functions of an optical solder for in-plane couplings and, in principle, for surface-normal couplings such as couplings between waveguide gratings and optical fibers.

 

Fig. 9. Coupling efficiency between 50-μm fibers connected by R-SOLNET with luminescent targets.

Download Full Size | PDF

However, the efficiency of the R-SOLNET coupling ($\sim 80\%$) remains limited, which might be attributed to nonoptimized luminescent target structures. The target shape and reproducibility of target formation may be improved by optimization of the fabrication processes in our future work. Furthermore, coupling efficiency between 10-μm fibers will be evaluated.

In conclusion, we demonstrate self-aligned optical couplings between 50-μm fibers and between 10-μm fibers using R-SOLNET, where the optical waveguides self-organize based on luminescent targets. A novel organic/inorganic hybrid material, Sunconnect, which was developed as a material for optical waveguides, was used as the PRI material. To enhance the targeting effect, adjustment of the sensitivity balance was attempted by using 448-nm write beams and coumarin-481-containing luminescent targets. R-SOLNET was found to widen the tolerance to the lateral misalignment between optical fibers, indicating that it could potentially function as an optical solder.