Very Wide Integral Field Unit of VIRMOS for the VLT: Design and Performances

Transcription

1 Header for SPI use Very Wide Integral Field Unit of VIRMOS for the VLT: Design and Performances. Prieto 1,O.LeFèvre 1,M.Saisse 1,C.Voet 1, C. Bonneville 1 1 Laboratoire d Astronomie Spatiale, Marseille, France ABSTRACT: This paper presents the VLT-VIMOS Integral Field spectroscopy Unit. This unit allows to observe a very large 54"x54" field on one edge of the VIMOS instrument multi-object field. This unit contains 6400 sets of µlenses-fibers-µlenses, producing the equivalent of a 72arcmin x 0.67 arcsec slit projected on the sky. Two spatial resolution (0.66" and 0.33") are offered, coupled with the low and high spectral resolution of VIMOS. The design philosophy, technological choices and the first test results of the assembled unit are presented. Keywords: Integral field, Fiber, micro-lenses, VIRMOS, VLT 1 INTRODUCTION: For the development of the instrument Visible Multi-Object Spectrograph (first of the two instruments called VIRMOS), we are developing an integral field spectroscopy unit (hereafter IFU). VIMOS is a wide field imaging multi-object spectrograph capable to observe large numbers of objects simultaneously (see Le Fèvre et al. 1 ). It offers 4 fields of view of 7x8 arcmin (~14x16 arcmin), with spectral resolution from 300 up to This instrument will be installed on the uropean Southern Observatory - Very Large Telescope at Paranal in Chile. The overall specifications for the VIMOS IFU are given in table 1. Field of View Large field (Low spectral resolution, R=300) 1 x1 Small field (High spectral resolution, R~2500) 30 x30 Spatial Resolution High 0.7 Low 0.35 ntrance / xit aperture Beam F/15 Table 1 : Specification for the IFU 2 TCHNICAL DFINITION: 2.1 System analysis In this section, we discuss the definition of the system, including the critical points, and the parameters that could be adjusted. After the requirements definition, we describe how to define each part of the system (entrance microlenses, output microlenses,...). Then, we inject realistic parameters in order to define the system. 2.2 Requirements Input data are as follows : F/15 input F/15 output 0.7 arcsec entrance spatial resolution (or closer possible) 1 x1 field of view (or as close as possible) minimize the crosstalk between channels minimize the size of the spectrograph entrance slit to maximize spectral resolution 2.3 ntrance µlens calculation : This optical system is used to image the telescope pupil on the fiber core. This ensures that all light coming from the sky does enter in the fiber. The first µlens plane is in the telescope focal plane in order to sample the field.

2 Telescope focal plane Lens Index : n Φ L Φ Pupil image T :F/15 Fiber core e Fiber entrance angle : = Ε Φ = T e n (q. 1) The entrance angle in the fiber is : Figure 1 :ntrance Lens configuration nφ e The curvature radius is given by : R = n 1 f ' n where f ' = e l = (converted in air) (due to the fact that the focal plane of the lens has to be on the second face) We deduce that : R = n 1 e (q. 2) n From (1), we deduce that : n e = Φ T (q. 3) with (2) : another relation : ( n 1) Φ R = T (q. 4) Φ l = Φ (q. 5) T The µlens is entirely defined by (q.1) and (q.4). T is the telescope aperture and is given. The only choice is Φ Ε =the pupil image dimension (which is smaller than the fiber core dimension Φ F, typically 0.95xΦ F ).

3 The (q.5) defines the µlens dimension (therefore the spatial resolution) where is given the entrance fiber aperture,the telescope aperture T, and the image pupil dimension Φ. 2.4 Output lens calculation : In this case, we have to adapt the fiber emission cone to the VIMOS entrance aperture. Here we image the exit of the fiber at infinity. Figure 2: output lens configuration o Let s define the ratio K=, which is the ratio between the fiber output aperture and the fiber input aperture. f f ' = Φ T and we have : e' = nf' = n Φ (q.6) f t f R ' = then we deduce : R = ( 1 n) 1 n Φ f T (q.7) now let s deduce the slit diameter : Φ Φ s = f' K = K f T (q.8) so we deduce : Φ S T = (q.9) KΦ f (q.6) and (q.7) are defining the µlens. The lens aperture, and therefore the fiber speed are defined in (q.9) 2.5 Field coverage :

4 We have to adapt the field of view of each entrance microlens to the fiber acceptance cone. If we want to acquire the entire field, we have to fit the larger dimension of the lens within the fiber acceptance cone. microlens Acceptance cone fiber Fieldinfront of the lens Part of the field lost (out of the acceptance cone) Figure 3: The input microlens adaptation at the fiber cone acceptance If we favor the dimension of the lens (which is the spatial resolution of the system), and if the acceptance cone doesn t fit the microlens dimension, we will lose field. Figure 4 shows the field curvature lost with a square aperture µlens dimension Microlens aperture Projection of the fiber acceptance cone on the field plane Field lost Figure 4: Field lost due to the adaptation of a square lens to a fiber With a square aperture, the lost field fraction is given by : lost = 1 1 π ρ ρ ρ arcsin 2 ( ρ ρ where : ρ is the ratio : a/d (a : dimension of the square, d : diameter of the acceptance cone projection) 1 2 ρ 1 Figure 5 shows this effect for various fiber numerical aperture

5 µlens dimension versus relative coverage field lost (square lenses) µlens dimension in arcsec 1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 0,00 0,05 0,10 0,15 0,20 0,25 0,30 relative field coverage lost Fiber speed F/3 F/2.5 F/3.75 Figure 5: µlens dimension versus the relative coverage field lost For 5% loss in field, and an aperture of 3, the sky resolution will be 0.66 arcsec with a square microlens aperture. Discussion : The figure n 6 describes the system definition logical chain. The four entrance values are : The pseudo-slit dimension at the entrance of VIMOS (Φ S : Figure 2: output lens configuration) The fiber core dimension (Φ f : Figure 2: output lens configuration) The acceptable crosstalk on the CCD plane The acceptable loss in field coverage Figure 6: System Definition Logical Chain

6 In order to compare different system configuration, the following table summarizes some example configuration data. Pseudo-slit size (arcsec) (page 3) Pseudo-slits separation (µm) µlens f/ # acceptable field lost (percentage) spatial resolution (arcsec) number of fiber field of view (arcsec) 39x39 46x46 36X36 42x42 45x45 54x54 41x41 50x50 50x50 60x60 47x47 56x56 Table 2: performance comparison for differents configurations Where : Slit size is the dimension of the slit at the entrance of the spectrograph. The slit has a circular shape. Slits separation is the distance in the mask plane between each slit. This distance has a minimum of 100 µm (dimension of the fiber). This distance changes the crosstalk on the CCD plane between each fiber. µlens aperture is the numerical aperture of fiber entrance µlenses Field lost is the percentage of field you accept to lose due to the bad adaptation of the square lens to the fiber acceptance cone (see Chapter : 2.5) spatial resolution is the resolution on the sky of the entrance µlenses Field of view is the total field of view of the IFU After trade off between scientific requirement and technical feasibility, we choose the following configuration: The technical description gives: Pseudo-slit size (arcsec) (page 3) 0.86 Pseudo-slits separation (µm) 100 µlens f/ # 3 acceptable field lost (percentage) 5 spatial resolution (arcsec) 0.67 number of fiber 6506 field of view (arcsec) 54x54 ntrance µlenses pitch 384 µm ntrance µlenses focal length 1.38 mm Fiber speed F/3 Fiber Core Dimension 100 µm (140µm of cladding) Number of fibers 6400 Cumulated fiber length 19 km Output µlenses pitch 600 µm Output µlenses focal length 1.55 mm

7 3 TCHNICAL DSCRIPTION: 3.1 Input micro-lenses description and performances: As described above, we need an 80x80 µlenses array of 1.3mm of focal length. Because of the UV absorption and the problem of homogeneity of the thickness of epoxy µlenses, we have preferred to select µlenses substrate entirely in silica. As the manufacturing of such arrays is not easy, we choose the solution to couple two sets of cylindrical lenses which are manufactured by LIMO (Germany). (see fig 7). The aspheric shape which can be manufactured by LIMO on these lenses gives us the possibility to reach the equivalent optical quality of single µlenses. Therefore, we have designed the two µlenses array in order to have the best performance possible: the result is very close from the diffraction limit (See Figure 8 and 10). Figure 7: µlenses description The coupling efficiency should be 0.94, if the fiber is exactly at the image of the telescope pupil. The surface quality is good with a roughness RMS of 0.014µm. We measure a dead zone between 2 µlenses less than 30µm. Figure 8: µlenses performance Figure 9: µlens array picture Figure 10: ncircled energy

8 3.2 Fiber Bundle description and performances: Fiber description: The requirements for the fiber are: Transmission range: µm Maximize the robustness of the fiber Fiber core: 100µm After consulting several manufacturers, we choose the silica Ceramoptec fibers. With Hydrogen treatment, the transmission is good in the blue down to 0.37µm. With the drying method, the transmission is also very good up to 1.7µm. The transmission is better than 0.99 for the entire spectrum range. In order to minimize the focal ration degradation (FRD), and the number broken fibers, great care was taken on the fiber constitution. The fiber core is 100µm, which is standard. A double layer of cladding is then applied. The first one is doped silica, up to 140µm of diameter. And the second one is pure silica up to 170µm, this layers giving robustness to the fiber and helps the bundle manufacturing. The last effort was made on the external coating. It is an epoxy double layer coating (standard for Figure 11: fiber description Ceramoptec). The central layer is soft and prevent stress on the fiber, while the outer one is strong and gives robustness to assembly Bundle description: The fiber bundle provides an optical link between an 80x80 square µlens array (Input link) and 80 linear µlens array (of 80 lenses each). The system described in this document has the following schematic function: Alignment of 6400 fibers on a 80x80 square matrix (384µm pitch ) with 5µm rms requirement on the positioning error Distribution of the light through 80 sub-bundles (bundle of 6400 fibers reformatted in 80 bundles of 80 fibers) Align each sub-bundle on a 80 fiber linear array (pitch of 0.6mm) with 5µm rms positioning error requirement Bundles of this size had never been manufactured before anywhere in the world, and very few Figure 12: Fiber Bundle sketches companies have the skill to manufacture bundles of this type. The packaging of this piece was also very important to insure the protection of the fibers, without stress. The manufacturing of this bundle was commissioned to Andaoptec. The more difficult part in this bundle is the fiber matrix arrangement. The requirement in the accuracy of positioning was better than 5 µm rms versus a grid with the pitch of the µlenses (384µm). As the manufacturing process of the µlenses is very precise with a pitch very well maintained all along of the substrate, the bundle has to have exactly the same pitch,

9 otherwise the fibers at the edges will not be facing exactly the back of the µlenses. Figure 13 shows the measured coupling efficiency between the fiber and the µlenses versus the error of position. Because of a technical difficulty, the bundle manufactured does not have the exact pitch requested. An error of 30µm at the edges has been measured. In order to have a complete fiber position map, we use a photographic film to store the position on a usable media. After, we used 2 very accurate digital scanners (one commercial: thanks to Mr. Rasigni from Universite d Aix Marseille III, the other the MAMA, dedicated to astronomical plate scans: thanks to Mr. Jean Guibert from Observatoire de Paris Meudon). Both gave similar results on the position of the fiber cores. In order to correct the overall scale effect, a simple spherical lens can be used. By putting this lens in front of the µlenses, we change the position of the telescope pupil in front of the IFU head. Therefore, the angle of view of the pupil from the µlens change from the center field to the edge, and the position of the pupil in the focal plane does change as well. Choosing the right shape of the lens, we could in principle correct all smooth variations of the local pitch. However, a compensation of more than 50µm should give an average angle of incidence too important in comparison to the acceptance angle of the fiber. In order to have a very good quality of the surface of the corrector lens, we kept the surface spherical. In addition, in order to save the number of glass-air interfaces, we are gluing a thin spherical lens on the first face of the µlenses substrate. The result increases the coupling efficiency by 10% in average. Table 3 presents the result of the commissioning of the bundle: coupling efficiency 1 0,95 0,9 0,85 0,8 0,75 0,7 0,65 0,6 0,55 couplingefficiencyversus centeringerror 0, centeringerror inµm Figure 13: coupling efficiency versus centering error in µm Specifications Bundle alone Bundle + Lens rror rms 5µm 17µm 11µm Dead fibers Fibers w/ coupling > 85% Fibers w/ coupling < 85% Fibers w/ coupling < 70% Fibers w/ coupling < 60% Fibers w/ coupling < 50% Dead Fibers + C<50% Table 3 : Characteristics of the fiber bundle Figure 14: image of the actual fiber positions at the bundle entrance

10 Figure 15 : Coupling efficiency (map after correction, histogram before and after correction) The performances of the linear exit modules were in specifications, except for a few modules which had to be redone due to a heating process of glue polymerization. The redone modules are perfect now. Although this bundle is not fully in specifications, we think that the results achieved are already remarkable: this is the largest bundle made to this type of accuracy, and the final performances are acceptable for the astronomical user. This unit has therefore been accepted at Andaoptec. From the experience gained in this unique development, we are convinced that a bundle of this size can be manufactured with fiber positioning performances better than 5 microns rms. The NIRMOS IFU bundle that we are developing will be manufactured capitalizing on this experience. We want to thanks Daumants Pfafrods and Aldis Vanags from Andaoptec for their dedicated work on this complex unit Output µlenses: The output µlenses are linear array of 80 µlenses. They are glued to each module of 80 fibers. The technology and quality of manufacturing is the same than for the entrance array Prism and pupil adaptation: The beam has to be folded to enter the spectrograph, and the exit pupil of the µlenses array (at infinity) needs to be adapted to the spectrograph entrance pupil (4 meter before the focal plane). Prisms 240mm long have been manufactured, the last surface being a spherical shape.

11 4 OVRALL PRFORMANCS: The overall integration is being completed at this time. Definitive measurement of the performances are not yet available. Table 5 shows the last estimation of performance based on our lab measurements of all single components. IFU Specifications IFU final unit with Bundle + lens corrector ntrance µlenses Fiber Bundle Coupling 0.92 (for 5µm error) 0.87 (for 11 µm error) Fiber transmission Output optics Spectro Coupling Total Table 4 : overall performances 5 PHOTOGRAPHIS: Figure 16: IFU Head (input link) 80 x 80 fibers with a pitch of 384µm +/- 5µm rms (in the support for µlenses alignment (during integration)

12 Figure 17: Fiber Bundle with the input link and 3 of the 4 mechanical support of 80 exit modules of 80 fiber each Figure 18: the focal plane adaptation plus folding prism, folding the exit beam into the spectrograph, and adapting the pupil for the 240mm pseudo-slit (16 entrance slit in VIMOS) 6 ACKNOWLGDMNTS We thank the VIRMOS project team members for their help in our common effort to build the best instrument possible. The workshop and mechanical designers at the Laboratoire d Astronomie Spatiale from Marseille, are thanked for all their effort, help and advise, in particular Louis Castinel, and Pascal Dargent). A special thank to Gabriel Moreaux who has masterfully taken care of the delicate micro-lenses and prisms gluing. Three students from the cole de Physique de Marseille have worked on this project: Yves Secretan, Catherine Martinelli, and Remy Parmentier, they are warmly acknowledged. We would like to thanks with a special attention Mr Didier Bouchet from Ceramoptec Paris for his openmind very helpful for our very special application of fiber technology. 7 RFRNCS 1. Le Fèvre et al., these proceedings 2.. Prieto & al, 1997, A Wide-Field Integral Spectroscopy Unit for the VLT-VIRMOS, Fiber optics in Astronomy III ASP conference Series, Vol 152, G. Avila, S. D Odorico, 1991, Fiber Optics inastronomical instruments at SO, Fiber optics in Astronomy II, ASP conference Series, M. Tecza, N. Thatte, 1997, SINFONI: a High-Resolution Near-Infrared Imaging Spectrometer for the VLT, ASP conference Series, Vol 152, 1998