Optical design of the ESPRESSO spectrograph at VLT

Transcription

1 Optical design of the ESPRESSO spectrograph at VLT P. Spanò* a, D. Mégevand b, J.M. Herreros c, F.M. Zerbi a, A. Cabral d, P. Di Marcantonio e, C. Lovis b, S. Cristiani e, R. Rebolo c, N. Santos f, F. Pepe b a INAF - Osservatorio Astronomico di Brera, V. Bianchi 46, I Merate, Italy; b Univ. de Genève, Obs. Astronomique, 51 Ch. Maillettes, 1290 Versoix, Switzerland; c Istituto de Astrofisica de Canarias, Via Lactea s/n, La Laguna, Tenerife, Spain; d Univ. de Lisboa, Estr. do Paço do Lumiar 22, Lisboa, Portugal; e INAF - Osservatorio Astronomico di Trieste, V. Tiepolo 11, I Trieste, Italy; f Univ. do Porto, Centro de Astrofisica, Rua das Estrelas, Porto, Portugal; ABSTRACT ESPRESSO, a very high-resolution, high-efficiency, ultra-high stability, fiber-fed, cross-dispersed echelle spectrograph located in the Combined-Coudè focus of the VLT, has been designed to detect exo-planets with unprecedented radial velocity accuracies of 10 cm/sec over 20 years period. To increase spectral resolution, an innovative pupil slicing technique has been adopted, based onto free-form optics. Anamorphism has been added to increase resolution while keeping the physical size of the echelle grating within reasonable limits. Anamorphic VPH grisms will help to decrease detector size, while maximizing efficiency and inter-order separation. Here we present a summary of the optical design of the spectrograph and of expected performances. Keywords: High-resolution spectrograph, echelle grating, volume-phase holographic grating, anamorphism, pupilslicer, free-form optics, exo-planets 1. INTRODUCTION Since '90, when very large telescopes (diameters >8m) like Keck and VLT became operative, many efforts have been put to fully exploit their large collecting area to resolve tiny spectral features onto faint, deep objects. HIRES 1 on the Keck telescope and UVES 2 on the VLT Kueyen unit telescope are worldwide recognized as the two best examples of possible solutions, even if they differ in many aspects: the former maximized the area of the dispersive element with a 300mm collimated beam dispersed by a 3x1 mosaiced R-2.8 echelle replica, the latter, while keeping the collimated beam at a lower 200 mm value, increased the blaze angle of the 2x1 mosaiced, monolithic, echelle grating at the maximum available value (76 deg, corresponding to a R4 echelle) in order to maximize the spectral resolution. This asked for an innovative white-pupil configuration to minimize shadowing effects and large anamorphism due to the steep angle hitting the echelle grating. Moreover, it helped to keep the size of both the cross-disperser and the all-dioptric camera optics within reasonable size. Upcoming 30-40m class ELT telescope size fight against scaling laws for spectrographs 3, asking for collimated beams larger than 60cm. Then, echelle grating area must exceed 1m 2, making them extremely heavy, complex to be replicated or aligned, and expensive. Moreover, all the collimation optics start to be comparable to 1-2m telescope optics. Even on smaller 8-10m class telescopes, spectral resolution cannot be increased, without breaking the same scaling laws. In order to circumvent such a hard limit, new approaches must be found, keeping into account state-of-the-art technologies, like free-form optics and newly available dispersive devices, like slanted volume-phase holographic (VPH) gratings. ESPRESSO spectrograph optical design has been derived from ideas developed for CODEX 4, the Cosmic Dynamic EXperiment (now re-called "COsmic Dynamics and EXo-earth experiment") by Delabre and Dekker 5 (ESO). It cleverly combine pupil slicing technique with anamorphism and slanted-vph gratings to achieve very high resolving powers on the 42m European Extremely Large Telescope. *paolo.spano@brera.inaf.it; phone ; fax ;

2 Here we will give an overview of the optical design choices selected to match scientific and technical requirements. More details about the ESPRESSO overall system and the project status can be found in other paper within this same proceeding 6,7, Science drivers As defined in the name, the Echelle SPectrograph for Rocky Exoplanet and Stable Spectroscopic Observation, ESPRESSO, will aim to: (a) measure high precision radial velocities (RV) for search for rocky planets; (b) measure the variation of physical constants; (c) analyze the chemical composition of stars in nearby galaxies. Mainly, the first two scientific requirements can be translated into a quest for extreme spectroscopic and radial-velocity precision, approximately improving by one order of magnitude current state-of-the-art instruments, like HARPS spectrograph 9, mounted on the ESO 3.6m telescope in La Silla. A stability budget has been set up, taking into account all possible sources of RV drifts, from the telescope and the Coudè train to the fibers, the scrambler, the calibration sources, and the spectrograph itself. On the spectrograph side, some important technical drivers have been fixed. The spectrograph will have no moving functions, placed inside a vacuum chamber within a many-layers thermally controlled room. Monolithic devices, like the echelle 1x3 mosaic replica of R4 echelle grating onto a common Zerodur blank substrate has been preferred. The last main scientific driver will ask for maximum efficiency compatible with all other constraints. Generally speaking, stability fights against efficiency, so a careful trade-off must be made. The VLT telescopes offer the possibility to feed the spectrograph with light coming from anyone of the four UTs, or to simultaneously inject light coming from all telescopes together, thus allowing to operate ESPRESSO with a 16m-equivalent telescope, enabling high-resolutions spectroscopy of very faint sources. A much larger science case has been studied for ESPRESSO, properly described by Pepe 6 et al. (2010). 2. OPTICAL DESIGN Both high spectral resolution and efficiency requirements can be met despite the large size of the telescope and the 1 arcscec field of the instrument. At the entrance of the spectrograph an anamorphic pupil slicing unit (APSU) shapes the beam in order to compress the beam in cross-dispersion direction but not main-dispersion direction, where resolving power is achieved. In the latter direction, however, the pupil is sliced and superimposed on the echelle grating to minimize its size. The rectangular white-pupil is then re-imaged and compressed by the anamorphic VPH grism. Given the wide spectral range and the required efficiency, two large 90x90 mm 2 CCD detectors are required to record the ful spectrum. Therefore, a dichroic beam splitter separates the beam in a blue and a red arm which in turn allows to optimize each arm for image quality and optical efficiency. The cross-disperser has the function of separating the dispersed spectrum in all its spectral orders. In addition, an anamorphism is re-introduced to make the pupil square and to compress the order width in cross-dispersion direction, such that the inter-order space is maximized. Table 1. Main spectrograph parameters in the three observing modes. Parameter Standard 1-UT 4-UT Ultra-High Res. 1-UT Telescope aperture 8m 16m 8m Spectral Resolution Wavelength Range nn nn nn Aperture on sky 1.0 arcsec 4x 1.0 arcsec 0.5 arcsec Sampling (average) 3.5 pixels 4.0 pixels (binned x2) 2.1 pixels Spatial sampling 7.0 pixels 4.0 pixels (binned x4) 3.5 pixels Simultaneous reference Yes (no sky) Yes (no sky) Yes (no sky) Sky subtraction Yes (no sim. ref.) Yes (no sim. ref.) Yes (no sim. ref.) Total Efficiency 12% 12% TBD Instrumental RV precision <10 cm s -1 ~1 m s -1 <10 cm s -1 Detector area 6kx6k, 15 µm pixels 6kx6k, 15 µm pixels 6kx6k, 15 µm pixels

3 Figure 1. Spectrograph optical layout. 2.1 Design drivers Image quality must be as uniform as possible everywhere inside the accepted aperture of the spectrograph in order to give well behaved PSF everywhere onto the detector. Efficiency of the spectrograph must be kept as high as possible, in order to maximize the overall ESPRESSO optical throughput. High-transmission glasses and coatings are preferred, wherever possible. All spurious stray-light inside the spectrograph chamber must be avoided or minimized to reduce contamination onto spectra. Non radioactive materials must be selected near the detector to reduce detector noise. The instrument shall have a fixed optical layout. No moving parts have been foreseen inside the spectrograph, to maximize the stability and repeatability of the instrument performances. Moreover, no active components (like motors, or light sources, or actuators) will be placed inside to avoid any thermal load generated inside the spectrograph itself. The whole spectrograph must fit inside the available space at the Combined Coudè Laboratory room. A compact layout will help to reduce the size of the vacuum vessel and insulation assembly. Accessibility of optical components during alignment must be taken into account during design phase, to facilitate the integration and handling of optical components. 2.2 Overall system The optical layout of the spectrograph is given in Figure 1. It consists of two channels, partly sharing optical components. The wavelength splitting happens after the main dispersion done by the echelle grating, in order to reduce the size and cost of this large component. Light enters the spectrograph through fibers at the entrance aperture of the anamorphic pupil slicer unit (inside the dotted rectangle in the design). This device will have two functions: elongate the starting circular pupil into an elliptical one, with a magnification factor of 3:1, and then cut it into two pieces (slicing factor 2x). Then it overlaps each sub-pupil in order to decrease the effective anamorphism from 3x to 1.5x. This will reduce the size of the echelle grating by the same factor. Two images are created for each object into the entrance aperture, asking for increased interorder separation at the detector level, to avoid overlapping spectra. Light is collimated by a three-mirror anastigmat (TMA) onto the echelle grating. The collimated pupil is an elliptical one, 300 mm in the dispersion direction, and 200 mm in the spatial direction. This allows increasing spectral resolution, controlled by the number of ruled grooves only, without increasing the size along the spatial direction. After dispersion into high echelle spectral orders, light is refocused by the same TMA onto an intermediate focal plane. Of course, all spectral orders are overlapped at this stage and require a cross dispersion to be separated onto the final focal plane. Two additional mirrors will create a white-pupil. In order to decrease the size of the

4 camera optics, this second pupil will be smaller than that one at the echelle, being only 120x80 mm. A wavelength split is acted by the dichroic mirror (cut-off wavelength ~515 nm), working in quasi-collimated light in order to improve its spectral performances. A small incidence angle will help to better control its cutoff wavelength. Each channel will have separate corrector lenses, cross-dispersions and camera optics. Two detectors will be mounted at the two focal planes. The optical axes of the cameras have been made horizontal, to make cameras more stable to vibration or gravity load stresses. 2.3 Pupil slicer The spectrograph will be fed by different fibers, depending on the observing mode. Fiber shape, size and placement inside the available aperture can freely vary. Image quality of the fibers inside this aperture has been optimized in order to be very homogeneous. This will help during the alignment, too. The corrected field of view of the entrance aperture is 1.5 mm along the spectral direction and 0.5 mm along the spatial direction. The entrance aperture will be processed by the anamorphic pupil slicer. Different solutions were investigated. A first solution was based onto an all-mirror solution, with no chromatism, and the highest efficiency. However, it suffered from a quite limited corrected field of view. Unless a very large system is built, the physical size of the entrance aperture should be limited to an area less than 0.5x0.5 mm 2. Moreover, image quality varies quite a lot. A second solution was developed to increase the corrected field of view. It is still based onto free-form optics, i.e. parabolic cylinders, but now they are used on-axis along their powered axis, while the no-powered axis will be used slightly off-axis, improving the image quality. Collimation of light coming from fibers is done by a triplet that creates a 20 mm circular pupil onto the first parabolic cylinder mirror. Two cylinder mirrors with different focal lengths will introduce a 1:3 anamorphism. Pupil slicing is done with squared doublets, put side by side. Two flat folding mirrors will redirect light coming from the two sub-pupils onto a mirror stack that will overlap them at the exit pupil of the system. The mirror stack is composed of two cylinder mirrors oriented in such a way that they can be easily mounted together, aligned, and fixed. No aspherical terms have been added onto these small mirrors. Each object inside the rectangular area of the 1.5x0.5 mm 2 entrance aperture is reimaged into two elongated images onto square 3x3 mm 2 exit apertures. The two reimaged apertures are slightly separated by a dead gap, to avoid scattering and diffraction effects at the edges of the mirror stack. Image of the Entrance Entrance Aperture Figure 2. Anamorphic pupil slicer unit. Top: optical layout. Bottom: the entrance aperture and its double image.

5 2.4 Collimator Collimation optics have been designed and optimized in order to fulfill different constraints and requirements: (a) give enough image quality to not reduce resolving power by more than a few percent level; (b) stay within a compact space envelope; (c) make manufacturing of optical components feasible (with regards about their test setup for acceptance, and the alignment plan) and cost-effective. It consists of 5 mirrors and a corrector lens. First three mirrors (M1, M2, M3) act like a three-mirror anastigmat, even if after optimization of the full collimation optics it is not a perfect TMA anymore. Light dispersed by the echelle is recollected and focused near an intermediate focal plane, where another mirror (M4) acts as a pupil mirror. The last mirror (M5) and the corrector lens will create a white pupil just after the corrector lens, where a cross-disperser is placed. Only two mirrors are off-axis portions of aspherical mirrors (M2, M3), while all other mirrors and the corrector lens are made with spherical surfaces only. The wavelength splitting is done between the last mirror of the collimator and the corrector lens. There, the beam is near collimated, improving spectral performances of the dichroic. The angle of incidence has been minimized, to reduce complexity of the multilayer deposition. The size of the substrate is 360 mm. Due to its dimension, in order to avoid bending of the substrate, a quite large thickness has been selected on a low-cte and high-transmission material (Fused Silica). A small wedge has been added to avoid ghosts onto the focal plane. The orientation of the wedge is working in the same plane of the cross-dispersion, but the variation in cross-dispersion is almost negligible (in the Red arm only). Dichroic beam splitter Figure 3. Optical layout of the collimator optics, for the Red arm only. 2.5 Echelle and VPH cross-dispersers Two quite different dispersers will be used along the optical path to create both the main dispersion at high diffracted orders (R4 echelle grating), and two VPH-grisms for cross-dispersion, each one optimized for the Blue and Red arms. The main disperser is a R4 (blaze angle 76 deg), 31 gr/mm line density, 1x3 mosaic echelle grating. Formerly, it was a 1x4 replica of a 420x210 mm 2 master ruling. However, current manufacturing capabilities can deliver up to 1x3 mosaics. The ruled area will be 1270x210 mm 2, well matching the projected sub-pupils over the echelle surface. Two optimized cross-dispersers have been designed, based onto VPH gratings coupled with an entrance prism to introduce a anamorphic magnification. Fused Silica and BK7 were selected as materials for the two prisms, in the Blue and Red arm, respectively, delivering slightly different anamorphism (1.5x in the Blue, 1.7x in the Red). VPH peak efficiency up to 85% can be routinely obtained.

6 2.6 Camera optics and CCDs Each channel has optimized F/3xF/3 (~F/2 circular) camera optics, based onto different glasses, to optimize efficiency and image quality across the whole fixed spectral format. Cameras have been optimized as stand-alone subsystems, in order to simplify their construction and functional test. No moving parts have been foreseen. Focus adjustment will be done only during integration. Typical image quality across the full field of view is very high, with monochromatic spot diameters better than 1 pixel (rms) and <4 pixels (peak-to-valley). Axial and lateral colours were not corrected. Then focal plane for a fixed spectral format will be tilted by a small angle (<2-3 degree). Good anti-reflection coatings must be applied to enhance throughput and reduce possible ghost contaminations. If 0.5% average losses will happen at each air-glass interface, then overall throughput will be above 90% (in the blue) and 85% (in the red), taking into account for glass internal transmission losses, and coating absorption. The Blue camera design is based onto a 5 single lenses (S-FSL5Y, PBM18Y, S-FPL51Y, PBM18Y, Fused Silica), with only one aspherical surface at the first lens, and the last lens acting as cryostat window. Lens diameters are less than 270mm. Transmission losses are minimized due to careful selection of the glasses. These glasses are known to have good internal transmission down to 365 nm (i-line glasses) and good homogeneity in large blanks, being used in photolithographic machines. The Red camera consists of 6 single lenses, with only one aspherical surface on the first lens. The last lens is used as cryostat window. Well known glasses have been selected (S-FPL53, PBM8Y, PBM2Y) to improve efficiency and manufacturability nm nm mm 18 mm 36 mm 45 mm Figure 4. Top: Optical layout of the Blue (left) and Red (right) cameras. Bottom: spot diagrams as function of field of view and wavelengths. Boxes are 4x4 pixels wide.

7 3.1 Spectral resolution 3. SYSTEM PERFORMANCES The spectral resolving power will be defined by the fiber image projection size and shape onto the CCD, adding all deteriorating factors, like spectrograph image quality, fiber bundle imperfections/defocus, and CCD characteristics (flatness, electro-optical response, sampling). We made a first estimate of all these factors to highlight where limiting factors come from. We assume that all these contributions are gaussian noise sources that will add via RSS summation. The geometric projection can be estimated assuming a uniform illumination pattern of a circular fiber. The equivalent projected slit width can be defined as the FWHM along the spectral direction of this image, weighted along the spatial direction. a simple analytical computation shows that the weighted FWHM for an elliptical area is π/4 ( 0.785) times the diameter along the spectral direction. So, if this diameter is 60 µm, the equivalent slit width (FWHM) is 47 µm. This is equivalent to an increase in spectral resolution of 27%. The spectrograph image quality implies a loss in resolving power due to optical PSF size. It will be dominated by geometrical aberrations, more than diffraction effects. Some margin has been added to nominal values, in order to take into account manufacturing and alignment tolerances. Overall spectral resolution at blaze peak, where echelle anamorphic magnification factor is 1, has been computed as function of the fiber size, as shown in Figure 5 for two different cases: a circular fiber and a rectangular fiber Resolving power Fiber core size along spectral direction on sky (arcsec) Figure 5. Overall spectral resolving power vs. fiber core size. Thick line for circular fibers, dotted thin line for a rectangular fiber. 3.2 Wavelengths, sampling and spectral formats ESPRESSO spectrograph will cover from 380 to 790 nm in two arms. The Blue arm will cover from 380 to 520 nm, the Red from 510 to 790 nm, with a small overlap to guarantee full coverage at the cut-off wavelength of the dichroic beam splitter. No spectral gaps are present till 750 nm between adjacent orders. Above this value, gaps between FSR amounts to a few percent level. Spatial and spectral directions behave differently, due to the pupil slicing and different anamorphic magnifications introduced along the optical path inside the spectrograph. As a first approximation, average values can be computed in the center of the detector area. Different anamorphic magnifications introduced by the echelle grating (along the spectral direction) and the cross-disperser (along the spatial direction) will vary sampling up to 20% (along the spectral direction) and 5% (along spatial direction). Moreover, the two arms slightly differ due to different magnification factors of the cross-dispersers.

8 Figure 6. Blue spectral format. The box on the left represents the CCD area. On the right, detailed view between bluest orders. Figure 6 shows the spectral format in the blue arm. Four spectra are originated from a two fibers (target + sky/calibration) due to pupil slicing. Each order is one FSR long. In the enlarged view, the bluest part of the echellogram is shown to demonstrate that interorder and interspectra spacing is high enough to prevent overlapping spectra. As seen in the zoomed views, free gaps are present between adjacent spectra, belonging to the same spectral order (different fibers), or to different orders. A careful check has been applied everywhere over the CCD area. Interorder separation between adjacent spectra has been computed via ray-tracing. It s defined between the centre of similar spectra. There will be a minimum free gap of 12 pixels between spectral belonging to the same order and of 20 pixels between adjacent orders. 3.3 Image quality A full optical model from the entrance aperture of the spectrograph, before the anamorphic pupil slicer up to the detector focal plane has been built to trace rays and simulate the global performances of the spectrograph image quality. Image quality is very high (<1.5 pixels) everywhere over the free-spectral ranges (FSR), at all interested wavelengths. Above the FSR, image quality is still very good, even if optimization has been done on the FSR only. This value does not include any margin for manufacturing and alignment errors. They will be computed only after a careful sensitivity and tolerance analysis. As a general rule, we can easily take into account a margin factor of 50% on top of nominal values. This is enough to preserve the highest resolving power given by ESPRESSO spectrograph. Figure 7. Spot diagrams for the whole spectrograph optical train. Boxes are 2x2 pixel wide.

9 ESPRESSO spectrograph efficiency Efficiency BLUE RED Extended BLUE (<380 nm) Wavelength (micron) Figure 8. Predicted efficiency of the spectrograph optics (at blaze peak), from the entrance aperture to the focal plane. 3.4 Efficiency Using preliminary estimates of each sub-system efficiency loss, a full efficiency budget has been built. Size of the optical components, and simulated anti-reflection coatings has been taken into account. In the Blue arm efficiency will drop quite strongly below 380 nm. Figure 8 shows the overall spectrograph throughput, from its entrance aperture to the detector focal plane, excluding detector QE. 3.5 Environment effects The spectrograph thermal and pressure behaviour has been simulated, too. Pressure variation will largely affect both the focusing and the alignment of the spectrograph. For a variation of 1.0 atm (from laboratory conditions to vacuum), there will be a defocus of -440 µm (the minus means that the focus shift towards the last camera lens). This can be corrected during the alignment phase, with a defined defocus before to close the vacuum chamber. Moreover there will be a drift of about 2.0 mm in the main-dispersion direction and of 78 µm in the cross-dispersion direction. During the alignment of the echellogram on the CCD, this drift must be added to the spectrum to avoid large vignetting factors at the edges of the echellogram. Effect of temperature variation is three-fold: (a) shift of the spectrum as a whole; (b) varying magnification of the echellogram mainly due to the thermal variation of the camera focal length; (c) defocus. In the main-dispersion direction, thermal drift amounts to about 0.03 pixel/ C. In the millikelvin regime, expected drift is comparable in size with a radial velocity error of about 2 cm/s (1 pixel is about 600 m/s). This effect can be minimized with a careful selection of the echelle blank material CTE. Zerodur CTE is generally below K-1. A value of K-1 will reduce the expected drift in radial velocity by a 10x factor. Along cross-dispersion, up to 0.3 pixel/ C in the Red arm, while being almost negligible in the Blue (<0.01 pixel/ C). Such a difference in due to different materials for the VPH substrate. Along cross-dispersion, drift will not directly affect radial velocity accuracy, but slightly different pixel response between adjacent pixels can influence spectra extraction. This is why temperature must be controlled within 1 mk.

10 Varying magnification is not negligible, being ~0.2 (0.06) pixel/ C along 1 FSR in the Blue (Red) arm. This will translate into a differential drift of the spectral lines at the edges of each FSR. Both calibration and target spectra will be affected by this effect, so no systematic errors in radial velocity accuracy should be expected. Thermal defocus is almost negligible, but high-order aberrations start to increase when temperature will differ by more than 15 C around the average temperature used during the alignment phase. REFERENCES [1] Vogt, S., Allen, S., Bigelow, B., Breese, L., Brown, W., Cantrall, T., Conrad, A., Couture, M., Delaney, C., Epps, H., Hilyard, D., Hilyard, D.F., Horn, E., Jern, N., Kanto, D., Keane, M., Kibrick, R., Lewis, J., Osborne, J., Pardeilhan, G., Pfister, T., Ricketts, T., Robinson, L., Stover, R., Tucker, D., Ward, J., Wei, M., "HIRES: the high-resolution echelle spectrometer on the Keck 10-m Telescope," Proc. SPIE 2198, 362 (1994) [2] Dekker, H., D'Odorico, S., Kaufer, A., Delabre, B., Kotzlowski, H., " Design, construction, and performance of UVES, the echelle spectrograph for the UT2 Kueyen Telescope at the ESO Paranal Observatory," Proc. SPIE 4008, 534 (2000) [3] Schroeder, D., [Astronomical Optics], Academic Press, San Diego, (2000) [4] Pasquini, L., Avila, G., Dekker, H., Delabre, B., D'Odorico, S., et al., "CODEX: the high-resolution visual spectrograph for the E-ELT," Proc. SPIE 7014, 70141I (2008) [5] Spanò, P., Zerbi, F.M., Norrie, C.J., Cunningham, C., Strassmeier, K.G., et al., "Challenges in optics for Extremely Large Telescope instrumentation," Astronomische Nachrichten 327, (2006) [6] Pepe, F., Cristiani, S., Rebolo, R., Santos, N., et al., "ESPRESSO: the echelle spectrograph for rocky exoplanets and stable spectroscopic observations," this Proceeding (2010) [7] Megevand, D., Herreros, J.M., Zerbi, F.M., Cabral, A., et al., "ESPRESSO: projecting a rocky exoplanet hunter for the VLT," this Proceeding (2010) [8] Cabral, A., Moitinho, A., Coelho, J.M.P., et al., "ESPRESSO: design and analysis of Coudè-Train concepts for stable and efficient optical feeding," this Proceeding (2010) [9] Mayor, M., et al., "Setting New Standards with HARPS," The Messenger 114, (2003)