In an unusual intersection of fiber optics and astronomy, a collaboration of British and Australian scientists has yielded a technique that they believe could revolutionize ground-based infrared astronomy. Bright and narrow spectral lines from atmospheric OH interfere with night-sky astronomy in the 1- to 2-µm region, overwhelming all but the strongest celestial signals. The scientists have designed and demonstrated a unique filter that may be capable, after refinements, of filtering out these OH noise lines.The researchers from the University of Bath in the UK and from the Anglo-Australian Observatory in Epping and Redfern Optical Components in Eveleigh, both in Australia, have shown that complex fiber Bragg gratings (FBGs) can provide the spectral selectivity to filter out the OH lines. But FBGs work well only in single-mode fibers. In a multimode fiber, each mode interacts differently with the grating and is reflected at a different wavelength. The result is a broadening of the FBG’s reflection spectrum, with its exact shape dependent on which modes of the fiber are excited.The light collected by an infrared telescope is inherently multimode, however, and can be efficiently coupled only into a multimode fiber. The problem, therefore, is how to implement in a multimode fiber an FBG whose spectral characteristics are as precise as those of one in a single-mode fiber.Merely splicing a multimode fiber to a single-mode one is not a viable solution because the coupling loss — ultimately mandated by the second law of thermodynamics — is too great. The single-mode fiber can accept only one “mode’s worth” of light from the multimode fiber, and the rest is lost.Figure 1. By coupling the modes of a multimode fiber into multiple single-mode fibers, the superior performance of the latter can be obtained with a multimode input and output. Images ©OSA. But theoretically, lossless coupling between two multimode systems is possible, if each has the same number of degrees of freedom. If these two systems are a multimode fiber and a number of single-mode fibers, and if they are coupled with a gradual taper so that the transition is adiabatic, the modes of the multimode fiber can evolve into supermodes of the single-mode fiber system. FBGs in each of the single-mode fibers will exhibit the same precise reflection spectrum. A second taper can couple the single-mode fibers back into a multimode fiber (Figure 1). The system thus will have multimode input and output, but the spectral characteristics will be those of an FBG in a single-mode fiber. Figure 2. The scientists fabricated a gradually tapered coupler to assure that the transition from multimode to single-mode fibers was adiabatic. The investigators fabricated the gradual taper with a technique they developed for connecting normal fibers with photonic crystal fibers (see “Holey Fibers Connect to Conventional Fibers with Low Loss,” Photonics Spectra, August, page 84). They inserted 19 single-mode fibers into 19 holes in the preform of a photonic crystal fiber and then drew the preform into a fiber, leaving the end of the preform undrawn (Figure 2). The core of the resulting multimode fiber is composed of material from all 19 single-mode fibers.After forming the taper from the multimode fiber to multiple single-mode fibers, the scientists spliced an FBG onto each of the single-mode fibers, which they combined in another taper (Figure 1). The spectral response of the system was as good as that of a single FBG (Figure 3).Figure 3. The spectral response of the experimental system with 19 single-mode fibers coupled to the multimode photonic crystal fiber (solid line) and the average response of the 19 individual fiber Bragg gratings (broken line) are shown. The zero-loss value of the experimental system was adjusted for comparison with the individual gratings and does not correspond to a measured zero loss. The system was far from lossless, however, because the multimode fiber supported far more than 19 modes. The extra modes that could not be coupled into the single-mode fiber array were lost at the interface. Nonetheless, the researchers believe that the spectral response establishes the validity of this proof-of-principle experiment.