Characterization of hexabundles: Initial results
J. J. Bryant, J. W. O'Byrne, J. Bland-Hawthorn, S. G. Leon-Saval
aa r X i v : . [ a s t r o - ph . I M ] A p r Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 9 August 2018 (MN L A TEX style file v2.2)
Characterization of hexabundles: Initial results
J. J. Bryant ⋆ , J. W. O’Byrne , J. Bland-Hawthorn , and S. G. Leon-Saval , School of Physics, The University of Sydney, NSW, Australia 2006; Institute of Photonics & Optical Science, The University of Sydney, NSW, Australia 2006;
ABSTRACT
New multi-core imaging fibre bundles – hexabundles – being developed at the Uni-versity of Sydney will provide simultaneous integral field spectroscopy for hundreds ofcelestial sources across a wide angular field. These are a natural progression from theuse of single fibres in existing galaxy surveys. Hexabundles will allow us to address fun-damental questions in astronomy without the biases introduced by a fixed entranceaperture. We have begun to consider instrument concepts that exploit hundreds ofhexabundles over the widest possible field of view. To this end, we have compared theperformance of a 61-core fully-fused hexabundle and 5 lightly-fused bundles with 7cores each. All fibres in the bundles have 100 µ m cores. In the fully-fused bundle, thecores are distorted from a circular shape in order to achieve a higher fill fraction. Thelightly-fused bundles have circular cores and five different cladding thicknesses whichaffect the fill fraction. We compare the optical performance of all 6 bundles and findthat the advantage of smaller interstitial holes (higher fill fraction) is outweighed bythe increase in modal coupling, cross-talk and the poor optical performance causedby the deformation of the fibre cores. Uniformly high throughput and low cross-talkare essential for imaging faint astronomical targets with sufficient resolution to disen-tangle the dynamical structure. Devices already under development will have between100 and 200 lightly-fused cores, although larger formats are feasible. The light-weightpackaging of hexabundles is sufficiently flexible to allow existing robotic positionersto make use of them. Key words: instrumentation: miscellaneous:hexabundles – techniques: miscellaneous– methods: observational – instrumentation: spectrographs – techniques: imaging spec-troscopy.
Current and planned cosmological surveys in the opticaland infrared have fundamental limitations. Multi-fibre [e.g.2dF (Colless et al. 2001) and SDSS (York et al. 2000)] andmulti-slit [e.g. DEIMOS (Faber et al. 2003) and VIMOS(Le Fevre et al. 2003)] surveys have or will amass large cata-logues of galaxies in order to deduce their global properties.However these suffer from biases introduced when a fixedangular size aperture such as a single fibre is used to ob-serve galaxies, irrespective of the size, distance or morphol-ogy (see for example, fig.8 in Ellis et al. 2005). On the otherhand, integral field units (IFUs) like TEIFU (Murray et al.2000) and GMOS (Allington-Smith et al. 1997) or imageslicers such as SINFONI (Eisenhauer et al. 2003) and NIFS(McGregor et al. 2003) spatially sample the spectra, givingmorphological and dynamical information. However, current ⋆ E-mail: [email protected] (JJB)
IFUs are restricted in the number of objects that can be ob-served. The ideal is to combine multi-object spectroscopy(MOS) positioning technology with IFUs. Hexabundles cando exactly that.Hexabundles have up to many hundreds of closely-packed fibres to allow spatially resolved spectroscopy. Theyhave several practical advantages. Firstly, they are not assensitive as single-fibre devices to seeing and positioning er-rors and are unaffected by seeing losses. Secondly, they canbe used with AO-corrected or natural seeing and in the op-tical or infrared. Thirdly, there is no need for microlens ar-rays, and the plate scale can be changed simply with a singlemacro lens. Fourthly, they can be used with conventional fi-bre positioning technology as imaging bundles to obtain spa-tial information in a survey of many thousands of galaxies.The resulting scientific gains over existing large galaxy sur-veys would then include the ability to investigate AGN trig-gering and feedback including outflows, galaxy merger ratesand merger-induced processes, the substructure in gravita- c (cid:13) J. J. Bryant et al. tional lenses, stellar populations and abundance gradients,as well as tracing the build up of dark matter, stellar massand angular momentum in galaxies. Decomposition of bulgeand disk components would be possible for thousands ofgalaxies. Hexabundles are particularly effective for galaxieswith asymmetries, multiple components, mergers and sub-structures including high redshift galaxies that are not yetformed into spheroids or disks but consist of clumpy merg-ing components which would be poorly sampled with a sin-gle fibre spectrum. Differential binning of the fibres can alsobe applied, particularly for more distant objects where thebrightness decreases rapidly away from the nucleus.A 100,000 galaxy survey out to a redshift of 0.2 willbe possible with hexabundles in the proposed FIREBALLinstrument on ESOs Very Large Telescope (VLT). Currentlythe FLAMES instrument uses the OzPoz positioner to place132 single fibres across a 24 ′ field, which then feed into theGIRAFFE spectrograph. The proposed FIREBALL upgradewould involve fifty hexabundles, with around 100-cores eachsampling ∼ . ′′ .Hexabundles are not limited in the number of cores ineach bundle, beyond the physical size of the bundle holder.So far we have made bundles up to 61 cores, but larger de-vices of several hundred cores would be suitable for multi-object positioners. In anticipated applications, the core willhave 0 . − . ′′ per core. Then 61-core and 367-core bun-dles would have imaging areas 4 − ′′ and 9 − ′′ acrossrespectively.The main trade-off with the design of hexabundles isto have the largest fill-fraction (ratio of core area to totalbundle area) possible without compromising the optical per-formance. The fill-fraction is affected by both the claddingthickness and how the fibres are packed together. Fibres thatare fully-fused together have smaller gaps between cores giv-ing a fill-fraction of over 90%, but are significantly distortedfrom a circular shape. On the other hand, lightly-fused fibresremain circular but have larger gaps between them, resultingin fill-fractions less than 87%.This paper describes the performance tests of our firstfully-fused and lightly-fused hexabundles with an aim to as-sess the trade-offs of fill-fraction versus optical performance.Section 2 describes the hexabundle devices, while the exper-imental method and data reduction are in section 3. Sections4 and 5 have the results for the fully-fused 61-core hexabun-dles and the lightly-fused 7-core hexabundles respectively.Final conclusions on the comparison of the different typesof hexabundles is in section 7. Six new hexabundle devices have been characterized. In eachdevice, one end has multimode fibres with reduced claddingthickness, fused into a single element, which is the hexabun-dle. The fibres at the other end are loose and have normalcladding thickness. The first of these hexabundles consists of61 fully-fused fibres without interstitial holes (Fig. 1) givinga fill-fraction of > µ m and 125 µ m respectively, before the claddingwas etched away over a 2 − Figure 1.
An image of the 61-core fully-fused hexabundle takenwith non-uniform illumination to show the shape and position ofthe cores. Cores 1 −
16, 17 −
31, 32 −
46 and 47 −
61 were groupedtogether at output on four glass plates. Each of the cores are100 µ m in diameter. The cores are distorted from circular in orderto increase the fill-fraction. were then fused (further details of the manufacturing pro-cess can be found in Bland-Hawthorn et al. 2011). The re-sulting hexabundle has a diameter of ∼ µ m. On output,the 61 fibres are lined up in V-grooves between glass plateswith three rows of 15 and one row of 16 fibres. The fibresin adjacent grooves in the same glass plate are separated by500 µ m.The remaining five hexabundles have seven lightly-fusedfibres (with small interstitial holes) with cladding thick-nesses of 1, 2, 4, 6 and 8 µ m. The main difference to theprevious bundle is that the fibres are circular rather than dis-torted in shape (see Fig. 2), resulting in lower fill-fractions.In these bundles, the output fibres are loose and weremounted into 35 parallel V-grooves on a plate with care-ful attention not to apply pressure to the fibres. We aimto investigate the effect of circular cores on the optical per-formance and consider the trade-off with fill-fraction. Thedifferent cladding thicknesses are used to assess the idealbalance between minimising cross-talk while maximising fill-fraction. Each of the bundles were tested for cross-talk, relativethroughput and mode-dependent losses. It was crucial forthe cross-talk tests in particular, to ensure that no straylight was getting into any core other than the intended tar-get. Originally laser light was focussed into the individualfibre (output) end of the fibres while the hexabundle endwas imaged with the camera. This has the advantage thatthe fibres could be separated so that no stray light from thelaser could go into adjacent fibres. However, the light coming c (cid:13) , 000–000 haracterization of hexabundles: Initial results Figure 2.
The 7-core lightly-fused hexabundles, shown withdifferent magnifications. In each horizontal panel, the claddingthickness after etching away part the cladding is listed to the leftand the core-to-core separation is listed to the right. The imagesshow the bundles before the interstitial holes were filled with soft,low refractive index glue. In each of these bundles the cores are100 µ m diameter. These lightly-fused bundles are each the samesize as the central 7 cores of the 61-core bundle shown in Fig. 1. out of the multimode fibres at the hexabundle has a specklepattern that is larger than the size of the core. This is be-cause the different modes focus in slightly different planesoutside of the polished bundle face so that all of the light outof the fibre cannot be focussed at the same time, resultingin an imaged spot size that has an apparent diameter of ap-proximately 125% of the core diameter. An example of thiseffect is shown in Fig. 3. In any adjacent fibre it is impos-sible to distinguish light from this oversize speckle patternfrom light leaked into the fibre through cross-talk.The final method adopted, as shown in Fig. 4, used anOriel LED light source which was focussed into a 50 µ m corefibre with a 60 µ m diameter including cladding. This fibrewas then butt-coupled to the 100 µ m hexabundle cores (re-versing the direction of the light through the fibres com-pared to the previous method - this is the direction thehexabundles will be used in astronomical applications). Wemounted the butt-couple (input) fibre in a V-groove on a Figure 3.
This picture of an earlier prototype bundle (Skov-gaard, private communication) shows that when coherent light isinput into the output multi-mode fibres of a hexabundle, the lightcoming out of the hexabundle has a speckle pattern that at bestfocus, appears larger than the core of the fibre. The spot at thelower right is an obvious reflection. narrow plate that was attached to a tilt/rotation stage offto the side of the hexabundle mount. This minimised ob-struction of the area around the hexabundle face so that thefibre could be adjusted to align perpendicular to the hex-abundle face. The narrow plate could then slot in from theside between the hexabundle and a microscope with the fibrebending abruptly away to the side. This allowed the micro-scope to be positioned close enough to image the hexabundleand input fibre at high enough magnification to be sure theinput fibre was centred on a chosen fibre core within thehexabundle. The input fibre needed to be less than 100 µ mfrom the hexabundle face to prevent light straying into ad-jacent fibres. The small outer diameter of the butt-coupledfibre was essential in order to not obscure the view of thehexabundle cores. Two other microscopes mounted aboveand side-on were used to ensure that the fibre was perpen-dicular to the hexabundle face in both axes. The maximumerror contribution to the NA from the input alignment was ± . . µ m) then witha Bessel red filter (centred on 0 . µ m) between the LEDsource and the focussing assembly. Time variation in theinput light intensity plus variations in the SBIG camera re-sponse, were quantified with repeated images through onecore of the bundle at time periods ranging from secondto hours. The maximum variation in resulting integratedcounts was < .
0% with typical values of ∼ . c (cid:13) , 000–000 J. J. Bryant et al.
Figure 4.
Equipment setup to test the hexabundles. Clockwise from top left: The light from an LED source passes through a filter holderand is focussed into a 50 µ m core fibre. This fibre is butt-coupled in turn to each core of each hexabundle. The five metal tubes lined upare the lightly-fused 7-core hexabundles (the fully-fused 61-core hexabundle was in the same position when being tested). An xyz stageallows the bundle cores to be accurately aligned with the butt-couple fibre. An alignment microscope allows us to see which hexabundlecore the input fibre is butt-coupled to. Top and side microscopes are used to check the butt-couple fibre is within 100 µ m of the bundleface. The view from the top microscope is inset, showing the butt-couple fibre aligned with a hexabundle. The second butt-couple fibrein this image was a spare, and not used. Each of the 5 bundles has 7 fibres coming out, giving 35 fibres mounted in V-grooves 2 mmapart on an xyz stage. However, when the fully-fused bundle was being tested, the 61 output fibres were aligned on four glass plates onthe xyz stage. The output fibres are imaged through a pair of matching lenses onto a camera. was measured by centring the input light on each fibre. Forcross-talk measurements, the light was centred on a corewhile the surrounding fibres were imaged. FITS images fromthe SBIG camera were processed using the iraf astronomi-cal data reduction software (Tody 1986). Firstly each imagewas flat fielded. The background was fitted and subtracted in iraf radprof using an annulus with inner radius approxi-mately 3 times the maximum radius of the fibre image. Thenthe total integrated counts were measured in iraf radprof by fitting the barycentre of the fibre image and selecting afixed aperture size for all fibres that was large enough toinclude all the output from the fibre. The target fibre andthe surrounding fibres were measured through the same sizeaperture.We also aimed to compare the performance of differ-ent fibres within the same bundle in terms of focal ratiodegradation (FRD) or mode-dependent losses. FRD (e.g.Carrasco and Parry 1994) is a loss of optical entropy or non-conservation of ´entendue due to mode mixing as light propa-gates down the fibre. This modal mixing causes the openingangle of the output beam to be larger than that of the inputbeam resulting in a wider encircled energy distribution (seeFig. 5). The modal coupling means that lower-order modeswill “couple” to higher-order modes which have a larger an-gle at the output beam, hence a larger NA. In order to com-pare the consistency of the performance of different fibres within the same bundle, it was therefore important to en-sure that all modes were filled on input. When the coresare distorted, under-filling the NA on input may mean thelight couples into different modes at the input for differentfibres, making it invalid to compare the distribution on out-put. Therefore we have filled the NA of the fibres to exciteall possible modes, so that the output energy distributionacross the output image in the back-focal plane can be di-rectly compared for each fibre. The encircled energy profileat back-focus, will therefore be broader for higher modalcoupling.The camera was shifted to a known back-focal distance;selected to give sufficient pixels across the output fibre im-age to accurately fit an encircled energy profile. Using theknown pixel size of the camera, an NA value was calculatedfrom sin θ of the cone angle. This gives the effective out-put NA at each point across the image of the fibre. ThisNA value is used in the graphs to follow. After flat field-ing, iraf radprof was used to fit the centre position alongwith the background and peak counts in each image profile.These initial values were then input into iraf pprofile ,and the radially-collapsed encircled energy profile was fittedinteractively. An iterative procedure was used to determinethe background value to subtract and the peak counts inthe profile until the background was flat and the encircledenergy levelled at 100%. The encircled energy values could c (cid:13) , 000–000 haracterization of hexabundles: Initial results then be matched to the NA calculated for each radial pixelposition in the image. The resulting plots of NA vs encircledenergy will therefore be shifted to the right when there areworse mode-dependent losses (larger mode-dependent losseswill give higher FRD). The encircled energy profile of each of the fibres in the61-core hexabundle, was measured against the output NA.Fig.6 shows the encircled energy profiles through both theBessell blue (0 . µ m) and red (0 . µ m) filters. The profilesthrough the red filter are at a lower NA than those throughthe blue filter because the shorter wavelengths have moremodes and therefore a higher NA. While the average profilesof 16 cores get worse as the range of core numbers gets fur-ther from the bundle centre, there is a very large variation inencircled energy profiles from fibre-to-fibre, which will giveinconsistent performance across the bundle for imaging andspectroscopy.In order to quantitatively compare the encircled energyprofiles, the drop in encircled energy compared to the centralfibre was calculated for a nominal NA of 0.20. This NA ischosen because it is less than the total size of the outputcone (output NA at 100% encircled energy). The maximumNA is truncated by mode-dependent losses out of the fibre,however the distribution of encircled energy within that conewill also be broadened by modal coupling, and more modalcoupling is indicative of more FRD. Therefore, by selectingthe encircled energy loss at an NA of 0.2, we are measuringthe broadening of the profile due to modal coupling. Theresultant drop in encircled energy ranges from less than 1%up to more than 25% for different fibres in the bundle. At90% encircled energy, the up-conversion of NA is 0.10 forthe worst fibre in both the blue and red filters.To see if the deformation of each fibre affects the encir-cled energy profiles, the deformation was measured by fittingcircles to the image shown in Fig. 1. The largest circle thatdoes not include any cladding, and the smallest circle thatincludes all the core were fit to each core. The ratio of thediameters of these circles was used to define a ‘deformationratio’. While this measure is subjective, in Fig. 7, the plotfor deformation ratio against encircled energy drop showsa correlation with Spearman rank correlation coefficient of0.34 (95% probability). Therefore, the fibres with the highestdeformation ratio tend to have the largest encircled energydrop, indicating the highest modal coupling. Cross-talk was measured for 14 random fibres in the bundleby measuring the integrated counts in each adjacent fibrewhile the light was centred on the target fibre. The losswas not equally distributed into the adjacent fibres, withmost target fibres showing the majority of the cross-talkinto one or two of the adjacent fibres while in some casesthe counts were more than 100 times lower in other adjacentfibres. This was due partly to significant differences in thecontact area and cladding width between different fibres. Such differences arise from the distortion of the fibres in thefully-fused bundle. The total loss into adjacent fibres waspredominantly in the range of 0 . − .
2% (0.005-0.06dB)with one fibre at 4.2% with the blue filter, and one at 4.4%(0.2dB loss) in the red filter. The total cross-talk could notbe evaluated in the outer row of the hexabundle because wehave shown that the loss is unevenly distributed to adjacentfibres and therefore could not be assessed where there wereno adjacent fibres on the outer edge. As the outer row offibres include some of the most distorted profiles, the totalcross-talk is likely to be above 0.2dB for some of those fibres.For example, the cross-talk from fibre 56 into just 55 and57 is already a total of 0.17dB alone and therefore would bemuch higher if the losses could be measured on all sides.In Fig. 7 there is an anti-correlation (with Spearman-rank test correlation coefficient of -0.65, and 99.9% probabil-ity) between the drop in encircled energy and the through-put. Throughput losses are a combination of higher cross-talk in the more distorted fibres and modal stripping (indica-tive of worse FRD). The encircled energy drop is not mea-sured at the edge of the output light cone but at NA= 0 . >
99% probability). It can also be seen thatthe throughput varied by a factor of five between each ofthe 61 cores which would have an impact on the accuracyof photometry in extended astronomical sources. It is clearthat the more distorted fibres have lower throughput, worsecross-talk and higher encircled energy drop.
It was also noticed that when the input light was directedat the cladding between fibre cores, that light was then dis-persed through the entire bundle. In an astronomical ap-plication, the component of an extended source that hit thecladding would result in all positional information being lostand the spectrum of that component of the object beingadded to every fibre across the field. By centring the lighton the cladding between fibres 11 and 12, then measuringthe counts in 6 fibres that were not adjacent, the countswere 0 . − .
2% of the counts in fibre 12 when the lightwas directed into fibre 12. Fibre 12 was chosen because thethroughput counts in that fibre are average compared to thecounts in each of the other 61 fibres. Based on an estimate ofthe cladding area covered by the light in this test comparedto the total cladding area, up to 25% of the light in anyone fibre could be scattered from cladding elsewhere in thebundle. This fraction would lead to unacceptable confusionin the spectroscopic imaging of an extended astronomicalsource.The original aim of fully-fusing the bundle was to in-crease the fill-fraction to >
90% . However, based on thisperformance, the gain in coverage across an object being c (cid:13) , 000–000 J. J. Bryant et al.
NumericalApertureNA=sin NA Encircled Energy E n c i r c l e d E n e r gy NAInput light optical fibre θ θ Figure 5.
FRD increases the output cone angle θ . Therefore the comparative FRD in the fibres within a bundle can be assessed fromplots of encircled energy vs numerical aperture (NA), where NA is sin θ . Worse FRD shifts the profile to the right (dashed line). Whenmore light is coupled to higher-order modes, more of the encircled energy is at higher NAs, such that the shape of the encircled energyprofile will broaden or shift to the right. E n c i r c l ed E ne r g y NA (output) 0 20 40 60 80 100 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 E n c i r c l ed E ne r g y NA (output) 0 20 40 60 80 100 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 E n c i r c l ed E ne r g y NA (output) 0 20 40 60 80 100 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 E n c i r c l ed E ne r g y NA (output)
Figure 6.
Fused fibre bundle results. (Top) The encircled energy vs numerical aperture (NA) at output for the central fibre (solid line)and the average of all the core numbers 1 −
16, 17 −
31, 32 −
46 and 47 −
61 (short dashed, long dashed, dotted, dash-dot respectively).(Bottom) The same profiles but for the individual fibres 1 −
16, to show the scatter around the average values. The plots on the left arethrough the blue (0 . µ m) filter and those on the right are through the red (0 . µ m) filter. The arrow indicates how we defined encircledenergy drop from the central fibre to the worst fibre. A short horizontal line at NA=0.15 indicates the measurement error in the profilesat that NA, primarily due to focussing the output from each multimode core onto the camera. c (cid:13) , 000–000 haracterization of hexabundles: Initial results Figure 7.
Fused fibre bundle results.
Drop in percentage encircled energy compared to the central fibre vs total cross-talk (top left),throughput (top right) and deformation ratio (bottom left), along with deformation ratio vs cross-talk (bottom right). In all panels thedot and star symbols are for the R − band (0 . µ m) and B − band (0 . µ m) respectively. The maximum errors are shown by the errorbars on each plot. In the top left plot, the cross-talk errors are smaller than the point sizes in all but the 2 highest points, where it isslightly larger than the point. imaged, is outweighed by the confusion in the spatial ori-gin of the light at the spectrograph due to cross-talk andcladding scatter. Also, inaccuracies in photometry will re-sult from the extreme differences in throughput across animage.The next step was to consider the improvement in thisperformance if the hexabundle was lightly-fused so that thefibres remain circular at the expense of a slightly smallerfill-fraction. We had five hexabundles made with cores that are lightly-fused. These have the advantage that the fibres are notdistorted and therefore is expected to perform more con-sistently. In order to compare cross-talk for a given fill-fraction, they were made with cladding thicknesses of 1, 2,4, 6 and 8 µ m (see Bland-Hawthorn et al. 2011, for a discus-sion on the issues surrounding the manufacture of bundleswith these cladding thicknesses). The aim was to test howthin the cladding could be before the cross-talk became toolarge. Fig. 8 shows the encircled energy profiles for the lightly-fused bundles. The profiles of each of the 7 cores in the 2, 4and 8 µ m clad bundles agree within errors, indicating consis-tent modal coupling performance across the bundle. This isin contrast to the large variation between cores in the fully-fused bundle (see Fig. 6). Each of the profiles in Fig. 8 havelower NA than (sit to the left of) the best performing fi-bre (central fibre) of the fully-fused bundle, indicating lowermodal coupling (and therefore lower FRD) in the lightly-fused bundles. The 1 µ m cladding bundle has a distributionof encircled energy profiles for the 7 cores that is slightlylarger than errors, which may be due to mode stripping in-duced by the higher cross-talk. In this case the central corehas worse NA than the other cores as would be expectedfrom mode stripping based on the symmetry of the bundle. For each of the lightly-fused hexabundles, the relativethroughput was measured for each of the 7 fibres, and thecross-talk was measured by summing the losses into the sur-rounding 6 fibres when light was directed down the centralcore.Table 1 shows the cross-talk performance through the B − and R − band filters compared to the fill-fraction. While c (cid:13) , 000–000 J. J. Bryant et al. E n c i r c l ed E ne r g y NA (output) 0 20 40 60 80 100 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 E n c i r c l ed E ne r g y NA (output) 0 20 40 60 80 100 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 E n c i r c l ed E ne r g y NA (output) 0 20 40 60 80 100 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 E n c i r c l ed E ne r g y NA (output) 0 20 40 60 80 100 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 E n c i r c l ed E ne r g y NA (output) 0 20 40 60 80 100 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 E n c i r c l ed E ne r g y NA (output) E n c i r c l ed E ne r g y NA (output)
B−band (0.45 microns) R−band (0.65 microns) E n c i r c l ed E ne r g y NA (output)
Figure 8.
Unfused fibre bundle results.
The encircled energy vs numerical aperture (NA) at output for the lightly-fused bundles with1, 2, 4 and 8 µ m cladding thicknesses. Results through the B − band and R − band filter are in the left and right columns respectively. Thesolid line is the central core while the surrounding cores are the dotted and dashed lines. A short horizontal line at NA=0.15 indicatesthe measurement error in the profiles at that NA, primarily due to focussing the output from each multimode core onto the camera. the hexabundles with 4, 6 and 8 µ m clad fibres had no cross-talk down to the limit we could measure in our images, theyalso have a lower fill-fraction. The cross-talk was worse forthe 1 µ m clad bundle than for the 2 µ m clad bundle, as ex-pected. In the anticipated uses of these hexabundles (e.g.FIREBALL) the fibres would be 0 . − . ′′ /core - compara-ble to the best seeing at an excellent telescope site of 0 . ′′ , oraverage seeing at a good telescope site of 0 . ′′ . If the core isthe size of the FWHM of a gaussian seeing profile, then 50%of the encircled energy will be in the central core and ∼
45% in the adjacent the surrounding cores. The 2 µ m clad bundlehas a total cross-talk of 0 .
39% (at 0 . µ m), which is < µ mclad bundle (with up to 4 .
8% cross-talk), approximately atenth of the light in the surrounding cores will be cross-talkcontamination. Therefore, at optical wavelengths, the 2 µ mclad bundle is suitable for applications where the core sizeis comparable to the FWHM, and the 1 µ m clad bundle mayalso be suitable if there are several cores within the FWHMof the seeing profile. In the infrared, longer wavelengths will c (cid:13) , 000–000 haracterization of hexabundles: Initial results Table 1.
Total cross-talk from the central fibre into all of the 6surrounding fibres is listed as both a percentage of the integratedcounts in the central fibre and as a dB loss. Values are shown forthe blue B − band and the red R − band Bessel filters. The firstcolumn lists the cladding thickness, then the fill-fraction (ratio ofcore area to total bundle area) is given for each cladding thickness.Limits listed are set by the image detection thresholds in caseswhere no emission was detected.Clad Fill Cross-talkfract. B (0 . µ m) R (0 . µ m)( µ m) % (dB in brackets)1 0.87 1.4 (0.06) 4.76 (0.21)2 0.84 0.25 (0.011) 0.39 (0.017)4 0.78 < .
01 ( < . < .
01 ( < . < .
01 ( < . < .
01 ( < . < .
01 ( < . < .
01 ( < . cause the cross-talk to be worse and thicker cladding wouldbe required.When the light was directed at the cladding, there wasno detectable transmission of that light into all the coresas was seen with the “cladding scatter” in the fully-fusedbundle (see section 4.3).For the two bundles for which the cross-talk was mea-surable above the noise limit, the variation in the cross-talkvalues for the 6 surrounding fibres was at worst a factor of4. This is significantly lower than the factor of >
100 wefound for the 61-core fully-fused hexabundle. We attributethis to the uniformly circular fibres in the lightly-fused bun-dles which have equal contact areas with adjacent cores.The even, circular-nature of the fibres in the lightly-fused bundles has also resulted in a much more uniformthroughput from fibre-to-fibre. In Fig. 7 the variation inthroughput between different fibres in the fully-fused bun-dle was up to ∼ We have compared the performance of a 61-core fully-fusedhexabundle to that of 5 lightly-fused bundles with varyingcladding thickness. The distortion of the cores in the fully-fused bundle was found to significantly increase cross-talk,reduce throughput and increase mode-dependent losses. Thethroughput varied by a factor of 5 between different coresin the bundle, while the cross-talk reached over 4% in theworst core. The up-conversion in NA from the best (cen-tral) core to the worst was 0.1 at 90% encircled energy. Theperformance of the lightly-fused cores was better in every re-spect, and we attribute that to the uniformly circular cores.The throughput varied by at most 24% and the modal cou-pling results were consistent between the 7 cores in eachbundle within errors. After comparing the cross-talk to thecladding thickness, we found that 2 µ m cladding bundle had < .
4% cross-talk in the R − and B − bands while still hav-ing a fill-fraction of 84%. While the fill-fraction of the fully- Figure 9.
Unfused fibre bundle results.
Normalised integratedcounts (throughput) for each of the 7 fibres in the hexabundleswith 2 µ m and 8 µ m cladding thicknesses. Results for the R − bandand B − band are shown as dots and triangles respectively, and arenormalised to different values to offset them for clarity. An errorbar is given by the vertical black line. The variation in throughputbetween different fibres in the same bundle was less than theerrors for both bundles shown. fused bundle is > ACKNOWLEDGEMENTS
We would like to thank Roger Haynes for valuable discus-sions and input. JJB and JBH are supported by ARC grantFF00776384 which supports the Astrophotonics programmeat the University of Sydney.
REFERENCES
Allington-Smith J. R., Content R., Haynes R., Lewis I.J.,1997, Proc. SPIE 2871, 1284Bland-Hawthorn J. et al., 2011, Optics Express, 19, 2649Colless M. M. et al., 2001, MNRAS, 328, 1039Eisenhauer F. et al., 2003, Proc. SPIE 4841, 1548Ellis S. C., Driver S. P., Allen P. D., Liske J., Bland-Hawthorn J., De Propris R., 2005, MNRAS, 363, 1257Faber S. M. et al., 2003, Proc. SPIE 4841, 1657Le Fevre O., et al., 2003, Proc. SPIE 4841, 1670McGregor P. J., et al., 2003, Proc. SPIE 4841, 1581Murray G. J., et al., 2000, Proc. SPIE 4008, 611Tody D., 1986, Proc. SPIE 627, 733York D. G., et al., 2000, AJ, 120, 1579 c (cid:13)000