1. Introduction

The ability to write fiber Bragg gratings (FBGs) into each single-mode core of a multicore fiber (MCF) offers benefits in both research and commercial applications. One major goal of astrophotonics [1] is to suppress the unwanted atmospheric emission lines in near-infrared astronomical observations. The lines (mostly from OH radicals) are at least 5 orders of magnitude brighter than the interline continuum, so it is desirable to eliminate this light before it can be scattered during dispersion [2]. FBGs were demonstrated in 2004 to be the most compact and robust method of suppressing these emission lines in ground-based telescopes. In order to do so without losing information about the target source, the FBG reflection characteristics must match those of the emission lines: Bragg wavelengths of 1.4-1.8 µm, notch depths of up to 40 dB and notch widths on the order of 150 pm [3, 4].

FBGs can only be written into single-mode fibers (SMFs), but light from celestial sources cannot couple efficiently into these fibers and is instead collected with multimode fibers (MMFs). We have already demonstrated the use of complex aperiodic FBGs in SMFs fed via photonic lantern transitions from MMFs in the GNOSIS instrument [4]. This is an expensive option that is not sustainable when the spatial mode count is large. Compared to arrays of individual SMF gratings, the MCF grating is compact, resistant to damage and easy to taper into a photonic lantern; it also eliminates the requirement for a large (> 500) number of splices between the single-mode ports of the lantern and the individual SMF Bragg gratings (SMFBGs), as required by GNOSIS. MCFBGs are planned for inclusion in the next-generation atmospheric suppression instrument PRAXIS [5].

However, the fabrication of MCFBGs presents significant technical challenges, as inscription must be uniform across all cores for accurate line suppression. For a writing system consisting of a single laser beam and phase mask, the standard writing method is insufficient to produce good quality MCFBGs due to the spatial variations of the field at each core within the fiber. In order to write FBGs into an MCF, each core must be inscribed with identical index variations; otherwise, each core will have a different reflection/transmission profile and therefore the fiber’s ability to selectively and strongly eliminate a given narrow frequency band is compromised. In order to compensate for the effects of field variation producing different gratings in each core, one of two options is required: the first is to modify the characteristics of the fiber to exactly negate the effects of variations. A group at the Naval Research Laboratory achieved this in a 4-core fiber by having an asymmetrical distribution of cores and altering their photosensitivities so that the regions of greater UV exposure would be less responsive [6]. The other option, which we discuss in this paper, is to modify the incoming field so that all cores are evenly illuminated. This option is preferable as it allows creation of MCFBGs from easily manufactured basic fiber designs. It is also easier to scale to different core numbers and layouts without requiring calculations of the necessary variations in photosensitivity.

In this paper we describe a method of creating uniform MCFBGs from fibers with symmetrically positioned, identical cores. The work detailed here covers the case of a 7-core fiber but can theoretically be extended to any number.

2. Fabrication of MCFBGs

2.1 Design of MCF

Achieving identical FBG responses in all cores of a MCF requires careful design and fabrication. Light in all cores must see an identical Bragg condition, comprised of Bragg wavelength, notch width and response strength. Even assuming the same grating index modulation imprinted on all cores, obtaining the same Bragg response in all cores will require them to be optically identical and isolated.

The optical properties of the individual cores must be uniform from core to core and along the length of the imprinted grating index modulation. A core-to-core variation in either refractive index or core diameter will give rise to differences in the propagation constants (β) of the cores, hence producing a spread of Bragg wavelengths amongst them. Depending on the severity of this spread, the combined FBG response from all the cores could range from a broadened and weakened FBG response, to a collection of isolated Bragg notches in the best and worst case scenario respectively. Hence the optical quality of the MCF is a very important factor in the fabrication of MCFBGs.

MCF parameters such as core diameters, numerical aperture NA, and core-to-core separation are also important for achieving quality narrow deep FBG notches in every single core. If adjacent cores of the MCF are too strongly coupled, this core-to-core coupling will affect the β of the individual cores. This coupling will produce a splitting of the β of the MCF supermode (i.e. the combination of the individual cores with a fixed amplitude and phase relationship). If the splitting of the MCF supermode is too large then the transmission spectra of the FBG gratings written across the cores will be unacceptably broadened. Mode coupled theory tells us that a large core-to-core separation and high numerical apertures NA of the cores will reduce supermode splitting. However, large NA cores will increase the sensitivity to core-to-core variations in the MCF fabrication; and the large pitch will increase the overall size of the MCF, thus strengthening the conditions for a homogeneous and equal grating index modulation across all MCF cores. In practice, for a successful MCFBG device, compromises have to be made depending on the specific application. For extremely narrow deep MCFBG notches to be written in a standard FBG facility the overall size of the MCF fiber has to be kept to a minimum due to the limited laser power and beam size. Furthermore, the NA of the independent cores has to be large to avoid coupling between the adjacent cores. The photosensitivity of the cores is increased by including germanium as a dopant.

The MCF used for this study has 7 cores with diameter d = 5.5 µm, NA = 0.174, core-to-core separation s = 35 µm, and overall fiber diameter D = 125 µm. The fiber has been designed for low-loss telecommunication applications, and has a low core-to-core coupling and very good core uniformity for dispersion management.

2.2 Standard procedure for single-notch SMFBGs

The standard method for inscribing an FBG into an SMF uses a phase-mask (PM), a diffractive optical element to spatially modulate the UV light. The PM is placed between the UV laser source and the photosensitive fiber. The PM causes the UV beam from the laser to split amongst numerous diffraction orders which interfere while still overlapped for a small extent behind the PM. This interference pattern causes a modulated change in the index of refraction of the fiber core, resulting in a grating. This procedure can be modified to induce both chirping and apodisation into the resulting grating.

The writing system uses a 244 nm frequency doubled Ar + laser with an average of 90 mW power delivered to the PM. The UV laser beam is focused on the PM horizontally with a cylindrical lens to increase the intensity, resulting in a 30 μm beam width for a beam height of 700 μm. To write gratings, the fiber was exposed to the phase mask output while the stage moved perpendicular to the beam. All gratings have a length of 40 mm.

2.3 Modified procedure for single-notch MCFBGs

There are many experimental parameters involved in fabricating efficient FBGs across multiple cores. In order to enhance the UV photosensitivity of the cores, the MCF must be hydrogenated for an extended period. The hydrogenation chamber was maintained at 300 bar and at room temperature (25°C) for 2 weeks before writing gratings. This exposure to hydrogen gas is sufficient for our UV laser (90 mW average power delivered at the PM) to inscribe gratings with 30 dB suppression.

A key issue is to ensure uniform grating properties across the cores of the MCF. A laser beam incident on the fiber is focused by the circular cladding, leading to uneven illumination across the cores [7, 8]. As a result, each core has a different Bragg wavelength and the fiber as a whole does not reflect effectively at a single wavelength. Our solution to this problem is to place the MCF into a capillary tube with the illuminated side polished to an optically flat surface. Figure 1 shows the results of beam propagation simulations [9] comparing the UV field inside the fiber when the laser beam is directly incident on the cladding and when the polished capillary is introduced. Without the capillary, the UV beam narrows and the power at each core is unequal. Even for a 7-core fiber with diameter 125 µm, where all cores are within the region of exposure, the focusing of intensity results in uneven exposure. Adding a polished capillary to the system improves the MCFBG quality by making the field strength more uniform between cores, as the beam within the cladding becomes flatter and the focusing of intensity does not occur. In both simulations, the fiber cores are aligned at an angle with respect to the incoming beam in order to ensure no core lies within another core’s shadow. These and related simulations are presented in more detail in [8].

 

Fig. 1 (Top left) UV power inside a side-illuminated 7-core fiber. Solid black lines represent air/glass boundaries; solid black circles represent individual cores. (Top right) UV power inside the same fiber when placed inside a capillary tube with polished side towards the incident beam. (Bottom left) Photograph of polished capillary tube tapered to inner diameter 140 µm. (Bottom right) Photograph of 7-core fiber.

Download Full Size | PDF

The unwanted effects of lensing are also demonstrated experimentally. Figure 2 shows the response of cores in an MCFBG written without a polished capillary tube. There is minimal overlap between the wavelengths reflected by each core, and the notch widths and depths vary due to unequal exposure. The cutoff at −36 dB is due to the limited dynamic range of the camera; the transmission values are scaled relative to this.

 

Fig. 2 (Left) Transmission profiles of individual cores of an MCFBG written without any compensation for lensing by the cladding. (Inset) Diagram of core numbering.

Download Full Size | PDF

We have already shown [8] that a close match in size between capillary and fiber provides the best field flattening as well as preventing movement and rotation of the fiber during writing. However, the large costs associated with manufacturing custom-sized tubes have limited our selections to off-the-shelf sizes provided by Polymicro and Postnova Analytics. The capillary tubes pictured in Fig. 1 were heat-tapered in our laboratories with a Vytran glass processing machine (GTX-3000) in order to reduce the inner diameter to 140 µm so they closely matched the sizes of the fibers. Due to the small diameter variations ( ± 5 μm) along commercially available capillary tubes, their size cannot be smaller than an average inner diameter ~10 μm larger than the cladding thickness. Capillaries must be tapered prior to polishing, as the tapering equipment is only suited to cylindrical samples. Polishing the capillaries consisted of two stages. First, they were ground down to the desired wall thickness of 50 µm with a benchtop lapping machine (Logitech PM5) using Al₂O₃ as an abrasive. They were then polished to restore the flat surface to optical quality.

Gratings were written by removing a section of the MCF’s protective coating and sliding a polished capillary onto the target area. The fiber and capillary were then mounted on a moving stage as close as possible to the phase mask. The flat polished surface of the tube was positioned parallel to the surface of the phase mask and the surfaces were separated by small pieces of aluminum foil to prevent damage to the phase mask. More detail on the writing method can be found in [10]. After writing, gratings were annealed in an oven at 110°C for 20 hours to stabilize their Bragg wavelengths. The fibers were recoated with a Vytran PTR-200-ARL to protect the sensitive MCF claddings from dirt and breakages.

3. Results

To measure the spectra of all cores rapidly and reliably, the MCFBGs were fed with a tunable laser source (SWS15101, JDS Uniphase, 1460 nm to 1600 nm, 0.01 nm resolution) which was automated using NI LabVIEW to scan through wavelengths near 1550 nm and the output imaged by a Xenics infrared camera (XEVA-1429, 320x256 pixel, 16 bit resolution) in order to measure the optical throughput at each core as a function of wavelength. After 5 scans, the results were averaged to produce a set of spectra, one for each core.

In Fig. 3, we show the response of individual cores. These were measured after the grating was annealed in an oven, which accounts for the overall lower wavelength compared to Fig. 1; any variation between cores is maintained in this process. We find that 6 out of 7 cores are well aligned with the central core shifted slightly by 100 pm to shorter wavelengths. Since our goal is to generate a deep narrow notch from the combined responses of the cores, this is undesirable as any deviations compromise the combined performance.

 

Fig. 3 (Left) Transmission profiles of all gratings in the 7-core fiber. The wavelengths of reflection of the outer six cores overlap centered at 1548.25 ± 0 0.01 nm. The grating response of Core #1, which is located in the center of the fiber, is offset towards shorter wavelengths by around 100 pm. (Inset) Diagram of core numbering.

Download Full Size | PDF

Figure 4 shows the results of adding individual core spectra to produce an overall spectrum. Two cases are shown: an MCFBG with 6 cores arranged in a ring (central core removed) and the full 7-core fiber (including central core). The 6-core combined signal shows the desired behavior for effective OH suppression [11]. The result for 7 cores shows that including the central core reduces the overall filtering performance of the MCFBG, with only 8 dB suppression at the Bragg wavelength. This is because the central core does not provide any suppression at this wavelength, instead allowing the light to pass freely. If we reject the central core and consider only the response of the well-matched outer cores, the fiber then rejects light at the Bragg wavelength to a depth of at least 36 dB and a width of 30 pm.

 

Fig. 4 Comparison of the total throughput of the 7-core fiber (blue) with the throughput of only the outer 6 cores (green).

Download Full Size | PDF

If we choose to use the existing fiber and block the central core, this results in an overall 1 dB loss in throughput, which is undesirable in astronomical applications due to the small number of available photons. If instead we write gratings into a 6-core MCF which lacks the center core of the current design, the result will be a MCFBG with filtering capabilities equal to individually fabricated SMFBGs. However, the absence of a central core in this solution introduces difficulties in the manufacture of low loss multicore photonic lanterns from such a fiber [7].

Data from previous experiments with MCFs shows that the central core consistently shows a shorter Bragg wavelength than surrounding cores. We investigated the effects of hydrogenation at higher temperatures and pressures (25°C and 380 bar, also 80°C and 380 bar) with the aim of increasing the hydrogen concentration at the center core, but this did not improve the wavelength alignment of this core. The most likely explanation is that there is no significant shadowing across the 7 cores (as this would have led to poorly matched responses in the outer 6 cores), but that the central core has different optical properties. We conjecture that this small difference in Bragg wavelength response is due to the middle core having a different effective refractive index to the outer cores. This difference could arise from the remaining small amount of core-to-core coupling. Such crosstalk will affect the effective index of the outer cores differently to the central core due to the fiber geometry. The central core is surrounded by 6 identical cores at the same distance, while each outer core only has 3 neighbors.

To shift the spectrum in the center core, we may either increase the strength of UV exposure at that core or increase the amount of germanium doping in that core. The existing shorter Bragg wavelength in the central core corresponds to a lower modal effective index than the surrounding cores. If we accept the above hypothesis that the coupling conditions are affected by geometry, the mode confinement is reduced in the central core due to the stronger coupling, creating the observed reduction in effective refractive index. Therefore we aim in future experiments to eliminate the difference in response between the inner and outer cores. We will examine methods to change the refractive index compositions of individual cores and the geometry of MCFs to optimize them for writing Bragg gratings. Once we have full control of these properties, we will expand our techniques to larger MCFs with greater core numbers.

4. Conclusions

We have demonstrated techniques for creating MCFBGs in fibers with symmetrically positioned, identical cores. The most significant improvement comes from modifications to the writing process. The associated costs are small and only minimal changes to the optical setup are required. Multicore fibers are very important to the future of photonics [1, 12]. Our specific goal is to insert an identical FBG across all tracks within a multicore fiber. These can then be drawn down at either end to form a multimode fiber [13, 14]. As the field stands today, photonics is almost entirely limited to single-mode action in single-mode fibers which limits most applications to bright sources, in particular, powerful lasers. If it were possible to achieve single-mode performance within a multimode fiber, this would open up the field to many new applications.

The type of gratings described in this paper will be implemented in PRAXIS [5] and other astrophotonic instruments under development. In addition, this project marks a significant step towards the development of multimode photonics. Advancements in this field will benefit all applications in which photonic technology is used, allowing devices to become more compact without sacrificing performance.