We present a coherent fiber bundle comprising over 11,000 doped silica cores separated by an air-filled cladding. The fiber is characterized, and its imaging quality is shown to be a substantial improvement over the commercial state of the art, with comparable resolution over an unparalleled spectral range.
Published by The Optical Society under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.
Optical fibers have long been used to guide images from one location to another. Over the years, many methods have been developed to achieve higher resolution imaging. Some methods rely on more complex principles, such as phased array  or speckle correlation imaging , while others pursue simpler methods, such as moving the cores of a coherent fiber bundle closer together. In this Letter, we present a new development, to the best of our knowledge, of coherent fiber bundle technology that provides a considerable leap in imaging performance.
The proximity of cores in a fiber bundle is limited by the point where their evanescent fields begin to interact, causing light to be coupled between them, blurring the image [3,4]. There are three approaches that are commonly used to mitigate intercore coupling: increase the index contrast between core and cladding , make the cores dissimilar to their nearest neighbors [3,5,6], or have additional index step “trenches” [7,8].
Many imaging fibers commonly used for bronchoscopy consist of a pure or fluorine-doped silica glass cladding surrounding a random array of highly doped cores of slightly differing diameters to reduce coupling as much as possible. These types of fiber have been shown to perform well at short visible wavelengths, but performance of a given fiber degrades rapidly towards longer wavelengths when intercore coupling becomes significant [3,4].
Attempts to extend the functional wavelength range of imaging fibers into the infrared have included both antiresonant hollow core structures  and capillary arrays with an internal metal coating . In both cases, the number of cores has been limited to well under a thousand due to the fabrication techniques used. Chalcogenide fiber bundles have been reported with over 8,00,000 cores providing effective thermal imaging , but such soft glasses are far more challenging to work with than silica, preventing their widespread use.
Our previous research into fabricating solid imaging fibers has led to square arrays of cores drawn from germanium-doped silica telecom preforms, carefully stacked to ensure no two cores with the same diameter are neighbors . Though these have cost advantages over commercial imaging fibers, their imaging resolution is comparable.
Since the late 1990s, the index step between air and silica has been used in photonic crystal fibers (PCF) to achieve index guidance with a high numerical aperture (NA), allowing light to be confined extremely tightly [12,13]. It has long been recognized that the performance of imaging fibers could be greatly increased by incorporating these types of “air-clad” structures to minimize intercore coupling; however, traditional stack and draw techniques would limit them to only a few hundred cores . Imaging using such a structure made by drilling holes in a polymer rod and drawing it down to a fiber diameter has been reported , but with fewer than 100 cores, image quality is limited. Recently, an imaging fiber utilizing the index step between air and silica was reported . It consists of a solid silica outer jacket and a random (but longitudinally coherent) distribution of silica and air in the core region of the fiber, relying on Anderson localization for image guidance . Although this fiber provides evidence for the benefits of a silica–air waveguide, its imaging resolution is practically limited to group 4 of a USAF test chart primarily by the transverse localization beam radius.
This report describes an optical fiber with an array of 11,343 solid glass cores regularly arranged in an air-filled silica capillary lattice, and demonstrates its imaging performance to be superior to that of a state-of-the-art commercial fiber produced by Fujikura (FIGH-30-650S) by maintaining high resolution over a far broader spectral range.
Our fabrication process is similar to the multistage unjacketed stack-and-draw technique used to make other superlattice PCF structures . First, the basic elements of the structure (capillaries and cores) were drawn to a 1.85 mm outer diameter. The capillaries were drawn from F300 silica glass tubes with an inner-to-outer diameter ratio 0.928, and the cores were from a set of three telecom-grade preforms, each with a different diameter germanium-doped step-index core of refractive index 1.467 (1.450 cladding index). The ratios of core to cladding in these preforms were 0.35, 0.42, and 0.48. Using doped material in this way prevents the effective indices of nearest neighbor cores from being identical, reducing intercore coupling.
These canes and capillaries were stacked to form the 57-core array shown in Fig. 1(a), where the letters A, B, and C each represent a core with a distinct radius of doped inner core. This stack was drawn unjacketed to a flat-to-flat thickness of 1.2 mm.
One hundred ninety-nine of the 57-core canes were then stacked in a hexagonal array, inserted in a silica jacket of inner/outer diameters of 19/25 mm, and packed at the edges with pure silica canes to produce the preform with 11,343 cores shown in Fig. 1(b). This was drawn down to 4.3 mm diameter canes before finally being drawn to fibers of several different sizes.
In the final fibers, less than 5% of the cores are nonguiding, primarily due to fusing with the outer jacket, or with each other.
Two different sizes of the air-clad fiber were investigated for the purpose of this report: a 535-μm-diameter variant with a core spacing of roughly 3.5 μm, and packing fraction of 0.3—both similar to a leading commercial fiber, Fujikura’s FIGH-30-650S, which was used for comparison, and a 335-μm-diameter fiber with a core spacing of for high-resolution, short-wavelength applications. The cross section of the larger fiber can be seen in the scanning electron microscope images in Fig. 2.
The imaging quality characterization of the fiber was done in three experiments, each using approximately one meter of air-clad fiber, coiled to a radius of around 15 cm.
USAF test targets  were used to assess the achievable resolution of the fibers. This was done in a zero working distance imaging setup where the target was back-illuminated by a supercontinuum white light source. Bandpass filters were used at the output end of the fibers to produce the images in Figs. 3 and 4.
Figure 3 shows a set of these images taken across a range of wavelengths from 500 to 1000 nm with both the 535 μm air-clad fiber and FIGH-30-650S. Group 7 is shown to the top right of each image, with element 1 of group 6 at the bottom. In all images taken using our air-clad fiber, element 3 of group 7 is clearly visible. This corresponds to a resolution of 161.3 line pairs per millimeter (LP/mm), or 3.1 μm linewidth. With the FIGH-30-650S, element 3 of group 7 is only visible using light of wavelength 700 nm or less, and even element 1 of group 6 (64 LP/mm, or 7.81 μm) is lost past 800 nm, as has been shown in previous publications .
Figure 4 shows the exceptionally high resolution achievable using these fibers as even element 6 of group 7 (228.1 LP/mm, 2.19 μm linewidth) is clearly visible using the 353 μm air-clad fiber and 500 nm illumination.
To demonstrate the long wavelength characteristics of the air-clad fiber, Fig. 5 shows images of group 6 at 1600 nm using the 535 μm version and an infrared camera. Although there is loss of resolution due to increased coupling, the rate at which this occurs with increasing wavelength is far lower than with conventional fibers. When using the FIGH-30-650S, cross talk at 1600 nm is so strong that the cores appear randomly illuminated.
The NA of a selection of the fiber’s cores was found using a knife-edge measurement. The edge of the beam profile was taken as the point where the power drops to of its peak. This was done at two wavelengths by coupling a supercontinuum light source into single cores and filtering the output with 500 and 1000 nm bandpass filters. The NA is and at 500 and 1000 nm, respectively .
Due to the high NA of the fiber’s cores, their many slight differences, and the fiber’s relatively large minimum bend radius compared to telecom fibers, achieving sufficient bend loss to carry out a standard mode cutoff measurement was impossible. Instead, we can state that single-mode performance was observed in images taken using the 535 μm fiber and a light of wavelength longer than 1400 nm.
Using the cutback method, the losses of the 535 and 353 μm fibers were measured to be less than 0.26 and 0.33 dB/m, respectively.
From the images and data presented so far, it is difficult to tell how effective the strategy of using three families of doped cores to mitigate coupling is in this fiber. By coupling 1200 nm light into the circled core of the 335 μm fiber, the image in Fig. 6 was produced.
In a conventional fiber bundle, the cores adjacent to the one being excited undergo the most cross talk, whereas in this image the brightest cores are instead the next-nearest neighbors. This is because they possess a common doping profile, indicating that arranging families of cores in this way has a strong influence over the coupling characteristics of the structure. Also of note is the drastic reduction in cross talk across the double capillary boundary of the sublattice stack (yellow hexagon).
Coupling between cores depends on propagation distance as well as core separation and index step. Image resolution can therefore be further improved by locally decreasing the core separation at the distal end, if the affected length is short enough. This can be achieved, without compromising the overall length of an endoscope, by tapering a larger fiber.
An air-clad imaging fiber of outer diameter 600 μm was heated and stretched in a small flame . The narrowed region was cleaved to form a tapered tip 9.5 cm long with a diameter of 310 μm at the end face, attached to 80 cm of 600-μm-diameter fiber (Fig. 7). The reduced core separation was 1.2 μm, and the airholes have survived the process.
Group 7 of a USAF test target placed at the tapered end was imaged via a 500 nm bandpass filter [Fig. 7 (right)]. Compared to the smaller uniform fiber at the same wavelength (right half of Fig. 4), the elements of group 7 are sampled by several of the tapered fiber’s cores across each line, compared to only one or two with the untapered fiber. The combination of better sampling and the magnifying effect afforded by the taper leads to better resolved images.
A technique for the fabrication of air-clad imaging fibers based on index guidance between air and silica has been demonstrated. We have shown that these fibers can be made for high-resolution imaging down to 2.19 μm, or broadband imaging up to 1600 nm of similar resolution to Fujikura’s FIGH-30-650S, but double its maximum functional wavelength. It is possible to optimize these fibers for maximum resolution at the wavelength required by any application (limited only by the transmission spectrum of the silica itself) by drawing the fiber to a different size.
This technology will provide a platform for the development of the next generation of endoscopic techniques, particularly those that rely on near-infrared imaging of fluorescent marker dyes .
The benefits of these fibers to the field of medical imaging are clear, but before they can be exploited, they must first be suitably terminated, with the cladding capillaries blocked at the end face to prevent fluids being drawn inside. We are currently investigating this challenge through the use of low-index blocking materials, such as UV-cured glues and waxes, as well as by splicing short lengths of fiber to the end to act as a window.
The data underlying the results presented in this Letter are available at .
Engineering and Physical Sciences Research Council (EPSRC) (EP/K03197X/1).
We would like to acknowledge our colleagues in the Proteus EPSRC IRC at the University of Edinburgh and Heriot Watt University for discussions and insight into the applications of this technology.
1. A. J. Thompson, C. Paterson, M. A. A. Neil, C. Dunsby, and P. M. W. French, Opt. Lett. 36, 1707 (2011). [CrossRef]
2. A. Porat, E. R. Andresen, H. Rigneault, D. Oron, S. Gigan, and O. Katz, Opt. Express 24, 16835 (2016). [CrossRef]
3. X. Chen, K. L. Reichenbach, and C. Xu, Opt. Express 16, 21598 (2008). [CrossRef]
4. N. Ortega-Quijano, F. Fanjul-Vélez, and J. L. Arce-Diego, Opt. Commun. 283, 633 (2010). [CrossRef]
5. M. Koshiba, K. Saitoh, and Y. Kokubun, IEICE Electron. Express 6, 98 (2009). [CrossRef]
6. K. Saitoh, M. Koshiba, K. Takenaga, and S. Matsuo, Photonics Society Summer Topical Meeting Series (IEEE2012).
7. D. J. Richardson, J. M. Fini, and L. E. Nelson, Nat. Photonics 7, 354 (2013). [CrossRef]
8. S. Zheng, G. Ren, Z. Lin, and S. Jian, Appl. Opt. 52, 4541 (2013). [CrossRef]
9. T. Kobayashi, T. Katagiri, and Y. Matsuura, Opt. Express 24, 26565 (2016). [CrossRef]
10. Y. Matsuura and K. Naito, Biomed. Opt. Express 2, 65 (2011). [CrossRef]
11. B. Zhang, C. Zhai, S. Qi, W. Guo, Z. Yang, A. Yang, X. Gai, Y. Yu, R. Wang, D. Tang, G. Tao, and B. Luther-Davies, Opt. Lett. 40, 4384 (2015). [CrossRef]
12. J. C. Knight, T. A. Birks, P. St.J. Russell, and D. M. Atkin, Opt. Lett. 21, 1547 (1996). [CrossRef]
13. T. A. Birks, J. C. Knight, and P. St.J. Russell, Opt. Lett. 22, 961 (1997). [CrossRef]
14. L. Michaille, C. R. Bennett, D. M. Taylor, and T. J. Shepherd, IEEE J. Sel. Top. Quantum Electron. 15, 328 (2009). [CrossRef]
15. M. A. van Eijkelenborg, Opt. Express 12, 342 (2004). [CrossRef]
16. J. Zhao, J. E. Antonio-Lopez, Z. Zhu, D. Zheng, R. A. Correa, and A. Schülzgen, Sci. Rep. 8, 3065 (2018). [CrossRef]
17. P. W. Anderson, Phys. Rev. 109, 1492 (1958). [CrossRef]
18. M. V. Tse, Z. Liu, L. Cho, C. Lu, P. A. Wai, and H. Tam, Materials 7, 4567 (2014). [CrossRef]
19. “Resolving power chart,” Technical Standard Chart #83-001 (Edmund Scientific Company, 1978).
20. J. M. Khosrofian and B. A. Garetz, Appl. Opt. 22, 3406 (1983). [CrossRef]
21. S. G. Leon-Saval, T. A. Birks, W. J. Wadsworth, P. St.J. Russell, and M. W. Mason, Opt. Express 12, 2864 (2004). [CrossRef]
22. S. Zhu, Q. Yang, A. L. Antaris, J. Yue, Z. Ma, H. Wang, W. Huang, H. Wan, J. Wang, S. Diao, B. Zhang, X. Li, Y. Zhong, K. Yu, G. Hong, J. Luo, Y. Liang, and H. Dai, Proc. Natl. Acad. Sci. USA 114, 962 (2017). [CrossRef]