Extremely Large Telescope (ELT) Million Element Integral Field Unit (MEIFU) Final Report

Transcription

1 Extremely Large Telescope (ELT) Million Element Integral Field Unit (MEIFU) Final Report Simon Morris, Robert Content, Marc Dubbeldam, David Robertson, Cedric Lacey and Ray Sharples (University of Durham, Astronomy Instrumentation Group) AURA contract No GEM00303 Version date: 01/11/02 11:10 Contents 1 Executive Summary Science Case Deep Field Science Re-ionisation The development of Dark Matter Halos Galaxy Formation Feedback (winds from star forming galaxies, gas stripping in groups and clusters of galaxies) Identification of X-ray, MIR and Submm targets Telescope/Instrument Field of View Physics of high z galaxies from resolved spectroscopy Serendipity Examples of Complementary Surveys possible at other wavebands (text from R. Ivison) Some Numbers for the MEIFU deep fields Additional Science Instrumentation Concept and Trade Analysis with Alternative concepts IFU vs. IFTS Pros Cons Budget Collections Mass Throughput Cost Representative S/N Calculations Natural Seeing AO corrected and diffraction limited Fluorescent Lyman Alpha Emitters Optical Design Introduction and commentary on Telescope Interface Best telescope design for the instrument General description Field of view and spectral length Basic optical design Derotator Page 1 of 78

2 5.3 Pickoff Mirrors Fore-optics Lenslet Arrays Spectrograph Mechanical Design Introduction Subsystem Design Image Slicer Assembly Re-Imaging Optics Micro-Lens Arrays Spectrograph Instrument Design Moving Gravity Vector Cassegrain Instrument Fixed Gravity Vector Cassegrain Instrument Issues needed to be addressed for IR version of ELT MEIFU Science Optical Mechanical Technical Issues Identified as needing investment to reduce risk for Optical ELT MEIFU Appendices MathCAD calculation of Fluorescent Lyman α Fluxes and Surface Brightnesses MathCAD calculation of expected Sensitivity, and also comparison with IFTS performance Discussion of Tolerances Surface Quality: 2 λ (P - V) Installation tolerances Structural Deformations Positioning Accuracy of Interchangeable Optics Thermal Issues Milestone 1 Powerpoint Milestone 2 Powerpoint Milestone 3 Powerpoint Page 2 of 78

3 List of Figures Figure 1: Ellis et al (astro-ph/ ) Image of lensed cusp in A2218 showing circled two faint images of background object....7 Figure 2: Ellis et al (astro-ph/ ), Keck ESI spectrum of the two images showing the sharp blue edge and red wing expected for Lyman α emission at z=5.6. Horizontal bands are residual sky emission lines...8 Figure 3: Becker et al (astro-ph/ ), SDSS QSOs illustrating the Gunn- Peterson trough blueward of the Lyman α emission line Figure 4: Becker et al (astro-ph/ ). Evolution of the optical depth in neutral hydrogen on the IGM as a function of redshift, suggesting that reionisation of Hydrogen was completing at z= Figure 5: Dave et al. 1999, (ApJ, 511, 521) Column density of neutral hydrogen through a 200 km s -1 slice of the LCDM simulation volume at z=3,2,1.5,1,0.5, and 0 (from top left to bottom right panels). Red corresponds to N H = cm -2, orange to N H = cm -2, yellow to N H = cm -2, white to > cm -2. Illustrating the predicted relationship between gas (orange and red), and galaxies (white) Figure 6: Pie-diagrams for the galaxies and absorbers in the LOS to the QSO pair Q A,B (Morris et al. 2002, In prep). Filled triangles mark higher column density absorber, empty triangles are lower column density absorbers. Filled circles mark galaxies within 500 kpc of the QSO LOS, empty circles galaxies outside of that cylinder Figure 7: Shapley et al. 2001, (astro-ph/ ). Composite spectra for Lyman break selected galaxies, grouped according to relative strengths of emission and absorption features Figure 8: Frye et al. 2001, (astro-ph/ ) diagram showing the evidence for both outflow of the Lyman α emitting gas and also the ISM gas producing the UV absorption lines Figure 9: Subaru image of M82 illustrating the possible large differences between the continuum (blue/white) and emission line (red = Hα) morphologies of rapidly star forming galaxies Figure 10: Crampton et al. 2002, (astro-ph/ ). Serendipitous discovery of a case of galaxy-galaxy lensing. This geometry allows an accurate estimate of the mass distribution of the lensing galaxy Figure 11: Baugh et al. Semi-analytic Model for galaxy formation, Greyscale=dark matter, Dots=galaxies, colour=galaxy colour Figure 12: Semi-analytic surface density prediction, escape fraction fixed at Figure 13: Semi-analytic surface density prediction, with escape fraction calculated from dust model Figure 14: Semi-analytic prediction of the Observed I band magnitudes for a Lyman alpha selected sample Figure 15: Spatial Correlation Function in co-moving coordinates for Lyman alpha emitters. Assumed constant fixed escape fraction. Single flux range. Points = Lyman alpha emitters, Lines = dark matter...22 Figure 16: Spatial correlation function for two flux ranges Page 3 of 78

4 Figure 17: Cartoon of sample selection effects for continuum and emission line selected samples Figure 18: Some estimates for UV background versus redshift, HM = 1996 Haardt and Madau, Q = 2002 Haardt QSO only, G = 2002 Haardt QSO + Galaxies...33 Figure 19: Predicted Lyman alpha flux from optically thick and optically thin fluorescent emitters Figure 20: Predicted Lyman alpha surface brightness from optically thick and optically thin fluorescent emitters Figure 21: Angular size of chosen charactaristic radius (50 kpc) as a function of redshifts for the consensus cosmology...34 Figure 22: Predicted space density of optically thick fluorescent Lyman alpha emitters.35 Figure 23: PSF comparison for Cassegrain versus Richey-Chretien GSMT designs...37 Figure 24: Initial field mosaicing concept...39 Figure 25: Second mosaicing concept Figure 26: Comparison of two mosaicing concepts Figure 27: Schematic of Focal Planes Figure 28: Schematic of Optical Path...43 Figure 29: GSMT Fixed Gravity Cassegrain Environment Figure 30: Optical Derotator Figure 31: Pickoff Mirrors Figure 32: Pickoff Mirrors Figure 33: Fore-Optics Figure 34: Fore-Optics Figure 35: Fore-Optics optical design Figure 36: Fore-Optics PSF...51 Figure 37: Fore-Optics detail Figure 38: Microslice concept drawing...53 Figure 39: Microslice lenslet arrays Figure 40: Spectrograph Design Figure 41: Image Slicer Assembly Figure 42: Re-Imaging Optics Assembly (1st Re-Imaging Mirror not shown)...59 Figure 43: Micro-Lens Arrays Figure 44: Spectrograph Figure 45: Grism Wheel Assembly...63 Figure 46: Telescope Interface and Available Instrumentation Envelope...66 Figure 47: Baseline Instrument Lay-Out (Structure not shown) Figure 48: Moving Gravity Vector Cassegrain Instrument Lay-Out...69 Figure 49: Fixed Gravity Vector Cassegrain Instrument Lay-Out...70 Figure 50: Fixed Gravity Vector Cassegrain Instrument Envelope...72 Page 4 of 78

5 List of Tables Table 1: Summary Table of Instrument Properties... 6 Table 2: A sampling of the observed surface densities found for Lyman α emitters sorted by redshift. (1) ergs cm -2 s -1 (2) number arcmin -2 ( z=1) Table 3: Listing of FOV for Various Instruments Table 4: Throughput Calculation Table 5: Steps from 8m to 30m...38 Page 5 of 78

6 1 Executive Summary We present the results of a study funded by the AURA NIO. We develop a detailed science case for a wide field optical IFU instrument to be placed on the Cassegrain focus of the GSMT 30m telescope. The unique science niche is found to be in deep field exposures producing data cubes sensitive to extremely low emission line fluxes and surface brightnesses. Such observations will find star forming objects that will be undetectable in broad band continuum observations, and which would hence never be targeted for multi-slit spectroscopic follow-up. We present an optical design which would deliver a roughly 3x3 arcminute field with complete spatial coverage, with 0.1 arcsecond sampling. A wavelength range from 0.6 to 1µm can be observed, critically sampled at a spectral resolution of ~1750 in three exposures. A summary table of the instrument properties is given below: Table 1: Summary Table of Instrument Properties Wavelength Coverage micron in 3 settings Image Quality 50%EE in 0.11 arcsec Image Sampling 0.1 arcsec/pixel Microslice size 0.2 arcsec x 1.2 arcsec FOV 2.68 arcmin x 3.01 arcmin Number of Spectral pixels 600 pixels per setting Number of Microslices 164,832 Number of Spectrographs 24 Spectral Resolution ~1750 Spatial Resolution Depends on telescope+ao image quality 2 pixel sampling = 0.2 arcsec. We also show a mechanical packaging concept that would support this optical design at the Cassegrain focus of the GSMT. In appendices we present detailed signal-to-noise calculations for the instrument, and also calculations of the expected line flux from optically thick gas excited by the UV background. We also present a preliminary analysis of the tolerances for the instrument. Finally we also include the three milestone powerpoint presentations that documented the development of the instrument. Page 6 of 78

7 2 Science Case We describe in detail the science case for a wide field optical IFU to be used on a 30m class ground based telescope like the GSMT. The case includes an in depth analysis of the use of such a telescope-instrument combination for studying galaxy formation, including some sophisticated modelling, and also lists some of the other potential areas where such a combination might make important discoveries. The niche for such an instrument lies in Hubble deep field type of science, where comparatively small regions of the sky are studied in exquisite detail down to extremely low flux levels. We start by exploring what can be mined from such deep data cubes. Some of this material was also used for the case for an 8m class version of this instrument, but the modelling and flux limits discussed are all newly estimated for a 30m, and imply substantial possible gains from the larger collecting area. 2.1 Deep Field Science In this section we describe the simultaneous science goals that can be achieved by performing a series of deep spectroscopic observations with the MEIFU instrument. As we will describe below, such data will push forward our understanding of the epoch of reionisation, the process of galaxy formation, the effects of feedback from galaxies on the IGM and the dynamics of high redshift galaxies. The primary science driver for our instrument concept is the detection and measurement of high redshift emission line objects. To a large extent this is unexplored territory. Some hints as to what we might expect to find come from the rare, fortunate objects that lie behind a lensing cusp of a rich cluster of galaxies. Such an object is described by Ellis et al. 2001, (astroph/ ), and is shown in Figure 1 and Figure 2. Figure 1: Ellis et al (astro-ph/ ) Image of lensed cusp in A2218 showing circled two faint images of background object. Page 7 of 78

8 Figure 2: Ellis et al (astro-ph/ ), Keck ESI spectrum of the two images showing the sharp blue edge and red wing expected for Lyman α emission at z=5.6. Horizontal bands are residual sky emission lines. With a Lyman α luminosity of ~10 42 ergs s -1 at z=5.6, this object is currently unique, as it would lie well below the flux limits of the current narrow band emission line surveys. If, as many people suspect, it is in fact part of a large population of hierarchical fragments which will eventually merge to form the present day population of galaxies, then our deep exposures will find thousands of these. Such large samples will be needed to measure their correlation function and hence confirm or refute the above hierarchical model. In Table 2, we list the current knowledge of the surface densities of Lyman α emitters. As can be seen, the surveys are limited to flux densities of a few times ergs cm -2 s -1, yielding surface densities of a few per square arcminute per redshift interval of 1. Current models predict that by pushing deeper we should see densities of several hundred per square arcminute per redshift interval of 1. Table 2: A sampling of the observed surface densities found for Lyman α emitters sorted by redshift. (1) ergs cm -2 s -1 (2) number arcmin -2 ( z=1) -1 z delta z flux lim (1) Surface Density (2) reference year E Stiavelli et al E Warren & Moller E Moller & Fynbo E Steidel et al E Kudritzki et al E Cowie & Hu E Cowie & Hu ? 6.9 Cowie & Hu x E Rhoads et al large? 2.3 Dawson et al E Rhoads&Malhotra 2001 In the following sections we outline the science questions we would like to address with this huge data sample of line emitting objects. At the end we present some more detailed justification of our predictions of the expected numbers of detections. Page 8 of 78

9 2.1.1 Re-ionisation In a very exciting recent development, Becker et al (astro-ph/ ) show that there is good evidence that the tail end of the reionisation of hydrogen in the universe has been detected at redshift 6 (see Figure 3 and Figure 4). Figure 3: Becker et al (astro-ph/ ), SDSS QSOs illustrating the Gunn-Peterson trough blueward of the Lyman α emission line. Figure 4: Becker et al (astro-ph/ ). Evolution of the optical depth in neutral hydrogen on the IGM as a function of redshift, suggesting that reionisation of Hydrogen was completing at z=6. If confirmed, this means that the last stages of reionisation are observable in Lyman α in the optical (i.e. shortward of 1 µm with MEIFU). A tally of the line emitting objects present during that epoch should allow us to answer once and for all whether the ionization is dominated by UV radiation from star formation or by non-thermal sources such as AGN. An exciting additional possibility is that we may be able to detect high column density neutral hydrogen after reionisation simply from fluorescent Lyman α emission excited by the UV background. I.e. we would detect gas clouds with no intrinsic UV sources (see Page 9 of 78

10 Bunker et al. 1998, ApJ, 116, 2086 for an example calculation, and the modelling section below). Such detections will allow us to measure the UV background as a function of redshift. This will be a very challenging observation that will require very good understanding of the background in the data The development of Dark Matter Halos A major science goal that can be addressed with the MEIFU deep field data is the development of dark matter haloes. This is probably the strongest prediction of hierarchical models. Measuring the velocity dispersion of bound groups is much stronger than the simple correlation function analysis. The distribution of objects as a function of mass and its evolution with redshift (N(M,z)) is a very strong test of hierarchical assembly - more so than correlation functions which have issues of biasing etc. Measuring haloes through rotation curves is problematical: the galaxies are very small, there is the possibility for non-gravitational motions, and one is never sure how much of the halo is being sampled. Statistical measures of velocity dispersions of companion galaxies are much stronger (essentially giving the bi-dimensional separation-velocity correlation function). There are several reasons why this is well done with an IFU: Unlike rotation curves where one can estimate the major axis position, the "companions" for this experiment will be at all positions, and may well not be known beforehand. Also, since we just want an accurate mean velocity of each companion, IFU measurements can avoid slit-sampling problems. To do this we need a reasonable field, and good velocity accuracy (20-50 kms -1 ). Wavelength coverage gives one a multiplexing gain in redshift space Galaxy Formation By combining the emission line detections from MEIFU described above with high dispersion spectroscopy of background QSOs (such as the LOS studies for the ESO large absorption line programme) it will be possible to tie together the low column density gas and the locations where star formation has started. A recent attempt at predicting this relationship is shown in Figure 5, where expected regions of star formation are shown as white, where the gas filaments are shown as orange and red. Substantial evolution in the distribution is expected over the redshift range from 7 to the present day. One can see the prediction is that the gas settles into sheets and filaments and then drains into the regions of star formation. This model is largely untested at present. Page 10 of 78

11 Figure 5: Dave et al. 1999, (ApJ, 511, 521) Column density of neutral hydrogen through a 200 km s -1 slice of the LCDM simulation volume at z=3,2,1.5,1,0.5, and 0 (from top left to bottom right panels). Red corresponds to N H = cm -2, orange to N H = cm -2, yellow to N H = cm -2, white to > cm -2. Illustrating the predicted relationship between gas (orange and red), and galaxies (white). In a few lucky cases there are multiple background QSOs bright enough to study with UVES within a few arcminutes on the sky. A low redshift example (using HST spectroscopy on the QSOs) is shown in Figure 6. Page 11 of 78

12 Figure 6: Pie-diagrams for the galaxies and absorbers in the LOS to the QSO pair Q A,B (Morris et al. 2002, In prep). Filled triangles mark higher column density absorber, empty triangles are lower column density absorbers. Filled circles mark galaxies within 500 kpc of the QSO LOS, empty circles galaxies outside of that cylinder. One can begin to see supporting evidence for the above picture, but the sample of star forming objects is woefully small and only extends over a very small redshift range. With MEIFU we expect to cover all redshifts from zero to the QSO redshift, discovering all objects with emission lines either in Hα, [OIII], [OII] or Lyα Feedback (winds from star forming galaxies, gas stripping in groups and clusters of galaxies) Of course, the gas flow between the IGM and star forming regions is certainly not all one way. There are many examples of local starburst galaxies driving strong outflows (see Figure 9). Detections of metals in relatively low column density IGM clouds (Songaila and Cowie 1996AJ, 112, 335) also suggest that a large fraction of the IGM has been affected by such winds. Recent results now indicate that a large fraction of the Lyman break selected galaxies are also driving powerful winds. In Figure 7 we show some mean spectra of such galaxies from Shapley et al. 2001, (astro-ph/ ). Large velocity shifts are seen between the emission and absorption lines in these mean spectra, strongly suggesting high velocity outflows. Page 12 of 78

13 Figure 7: Shapley et al. 2001, (astro-ph/ ). Composite spectra for Lyman break selected galaxies, grouped according to relative strengths of emission and absorption features. In Figure 8 we show results from Frye et al. 2001, (astro-ph/ ) which demonstrate that probably not only the Lyα emitting gas, but also the ISM producing the UV absorption lines is also in outflow. Figure 8: Frye et al. 2001, (astro-ph/ ) diagram showing the evidence for both outflow of the Lyman α emitting gas and also the ISM gas producing the UV absorption lines. (The line of argument is that the UV absorption line EW is produced by a large number of optically thick clouds that only partially cover the continuum sources. Thus the EW is really a measure of the velocity spread of the absorbers, and its correlation with the Lyα velocities indicates the velocity fields are related.) The MEIFU deep field data will uncover thousands of such objects, allowing statistical analysis of their properties. Such Page 13 of 78

14 large samples are needed, as one has to break the initial sample into several bins in redshift, luminosity, environment and spectral type. As is well know, this rapidly whittles down a seemingly huge sample into a small number of objects per bin Identification of X-ray, MIR and Submm targets As is further emphasized in the section below describing complementary observations in HI, CO and submm continuum, deep field studies of the sort MEIFU is designed for benefit greatly from observations at all wavelengths. Conversely, several of the current regions studies to great depths at other wavelengths would benefit from MEIFU followup. In Table 3 we list the fields of view of some of the current and near future instruments that are, or will be, producing such deep field data. Table 3: Listing of FOV for Various Instruments Telescope/Instrument Field of View Chandra 17x17 arcmin XMM 30 arcmin diameter HST ACS 3.3 x 3.3 arcmin SIRTF 5.1 x 5.1 arcmin NGST Imaging 4 x 4 arcmin SCUBA 2.3 arcmin diameter ALMA 12 arcsec diameter, but will mosaic The intent of the table is to show that many of these deep fields require wide field followup. While MOS or deployable IFU approaches may prove adequate in some cases, two factors make a wide field IFU very attractive. First, any optical emission may well be significantly displaced from the peak of emission at some other wavelength. Secondly, as is discussed in the section below on resolved spectroscopy, the emission line morphology of any candidate identification may not match that of the continuum making placement of a slit a guessing game, and even placement of deployable IFUs a challenge. Thirdly, some of the positions of the sources at other wavebands are uncertain by several arcseconds. As an example, searching for extended Lyman alpha blobs similar to those found by Steidel et al associated with SCUBA detections would be of great interest. As a final comment, we emphasise that by careful choice of field for deep surveys, we can produce these identifications at the same time as pursuing all the other science goals listed in this section. The identifications are essentially free Physics of high z galaxies from resolved spectroscopy It is clear from many low redshift objects (for example M82, shown in Figure 9) that the emission line morphology can bear little resemblance to that of the continuum. This is a serious problem for slit spectroscopy. Page 14 of 78

15 Figure 9: Subaru image of M82 illustrating the possible large differences between the continuum (blue/white) and emission line (red = Hα) morphologies of rapidly star forming galaxies. The deep field MEIFU observations proposed in this section will produce spatially resolved spectroscopy for all the objects in the field with high enough surface brightness. At lower redshifts, where lines such as Hα and [OII] can be measured we expect that a reasonable fraction of our detections will yield well-defined rotation curves allowing reliable mass estimates. At higher redshifts, where we are observing Lyα, we expect far fewer such objects, but can switch our science focus to the study of outflows described above Serendipity A factor that is hard to quantify and generally therefore extremely hard to make a science case for is serendipity. Despite this, many in the MEIFU collaboration firmly believe that the most memorable results from MEIFU will come from discoveries that are currently not anticipated in this proposal. In order to illustrate the enormous scientific multiplex gain from obtaining complete data cubes over large areas of the sky with wide wavelength (and hence redshift) coverage, we show in Figure 10 an example of galaxygalaxy lensing discovered during the course of the Canada_France Redshift Survey (CFRS). Page 15 of 78

16 Figure 10: Crampton et al. 2002, (astro-ph/ ). Serendipitous discovery of a case of galaxygalaxy lensing. This geometry allows an accurate estimate of the mass distribution of the lensing galaxy. In the case of CFRS, two out of 350 galaxies imaged with HST showed such galaxygalaxy lensing, so we expect a large number of these in a MEIFU data cube. Galaxygalaxy lenses are a powerful means of measuring the gravitational potential of the lensing galaxy. While we can quantify how many such objects we expect to find, the serendipitous possibilities from a MEIFU data cube are unquantifiable, yet we believe vast Examples of Complementary Surveys possible at other wavebands (text from R. Ivison) In this section we describe complementary surveys that should be carried out within the same volume of space as the MEIFU deep fields in order to extract the maximum possible science from the data. We show that by observing for comparable time periods with the ALMA, e-vla and ultimately the SKA one can map out the neutral and molecular gas content of the volume to compare with the ionized gas seen in the MEIFU data. The molecular hydrogen (H 2 ) in a galaxy is of particular importance since it fuels star formation and accretion onto AGN. Comparing M(H 2 ) with the dynamical mass allows the determination of the evolutionary status of a galaxy, while a comparison of its dynamical mass with that of a present-day spiral or elliptical points to its descendent at the current epoch. The most straightforward way to measure M(H 2 ) is via CO, the second most abundant molecule in galaxies. The Atacama Large Millimetre Array (ALMA) will provide the principal means of measuring CO, with complementary measurements of lower-j rotational transitions made by e-vla and, eventually, SKA. With 16 GHz of bandwidth and receivers covering the 140-, 230- and 345-GHz bands, ALMA will reach at least 5x10-22 Wm -2 (5 sigma) in an hour; it will require 375 (170, 60) pointings to cover a 5 x 5' field at 345 (230, 140) GHz and will detect several hundred galaxies in several transitions given an hr per pointing (Blain et al. 2000). In the absence of low-frequency receivers for ALMA, e-vla will be crucial: its 8-GHz bandwidth and 1-50 GHz tuning range will provide the means to detect the important CO Page 16 of 78

17 J=1-0 transition (that requiring the least extreme excitation conditions) across the entire redshift range covered for Ly alpha by MEIFU. Its primary beam is frequency dependent: 1.6' FWHM at z=6.7, 2.7' FWHM at z=6.7. To map CO(1-0) in a 5 x 5' field across the z= range thus requires around 22 pointing/tuning combinations. In 300hr, e-vla will reach a minimum depth of 5x10-23 Wm -2 (5 sigma), sufficient to detect tens of galaxies with M(H 2 ) ~ 1x10 11 M, hundreds of galaxies if line ratios typical of low-excitation conditions prove common (Blain et al. 2000; Papadopoulos et al. 2001; Papadopoulos & Ivison 2002). For HI studies one requires high surface brightness sensitivity rather than sub-arcsecond resolution. With baselines up to 100 km, SKA will provide a resolution of 1" at 610 MHz (z = 1.3), so confusion will not be an issue. Such an instrument will be able to detect L * galaxies (HI masses of 6x10 9 M * for H 0 =75) out to z=4.5, well beyond the redshift range where the Universe shows considerable evolution. At a 0.3 x L * mass limit, SKA is anticipated to detect 400 galaxies in a 5x5' field (actually a tiny fraction of its field of view) between z= , over 100 more at an L * mass limit between z= (as well as several thousand between z= at far lower mass limits). One would be in a unique position to trace the evolution of the ISM in galaxies over a substantial fraction of the age of the universe, from the era of strongest evolution and star formation activity until the present. The selection of objects in the field would be entirely based on HI, not on the associated stellar component, and is therefore independent of the effects of extinction, colour and optical surface brightness. The combination of deep, HIselected samples and deep, optically-selected samples will be extremely powerful for studying galaxy evolution over a large range of redshift. In addition to HI content, such a survey would also measure the continuum emission of the galaxies in the field, which is known to be an excellent indicator of the massive starformation rate, independent of the effects of extinction. This information can be used to link the star-formation rates to the HI content of galaxies as a function of redshift and environment. It will also provide an independent estimate of the evolution of the comoving star-formation-rate density to be compared with the optically determined functions Some Numbers for the MEIFU deep fields In section 4.4 we will calculate the flux limits reachable by this instrument in combination with the GSMT. Here we try to estimate what populations of line emitting objects may lie below the current thresholds for detection. The approach we use is semianalytic modeling as described in Benson et al. 2002, MNRAS, 333, 156. As shown in Figure 11, Dark matter models are then combined with semi-analytic approximations for the expected behavior of the gas and stars to make predictions for the star formation rate (SFR) and properties of the galaxies that form in the dark matter potential wells. Page 17 of 78

18 Figure 11: Baugh et al. Semi-analytic Model for galaxy formation, Greyscale=dark matter, Dots=galaxies, colour=galaxy colour The effects of dust are included in the models. In order to predict the Lyman α luminosity of each object, at present only two fairly simple approximations for the escape have been considered. The physics of resonant line transfer in a dusty medium with a complex velocity field is complex, and insufficiently constrained by observation at present, and so the above simplification seems justified. We assume either (a) that all star forming object have the same constant escape fraction for Lyman α photons, and normalize the resulting number counts to match the observed surface density of objects at the bright end, or (b) we assume that the Lyman α is affected in the same way as the starlight by the dust that is modeled (i.e. ignoring the resonant scattering and hence longer pathlengh for escape). In the first case the normalization requires an assumption of a 10% escape fraction. The second assumption actually gives a good match to the number counts without additional normalisation Page 18 of 78

19 Figure 12: Semi-analytic surface density prediction, escape fraction fixed at 0.1. The results for a fixed escape fraction are shown in Figure 12. The dashed lines show the results from the earlier model of Cole et al. 2000, MNRAS, 319, 168, showing how the models have changed. The dotted lines show the current narrow band flux limit and the observed surface density of objects at z~3. The expected flux limit for a 4 night deep field exposure with MEIFU on the GSMT (see section 4.4). The same information from the assumption of diffuse dust obscuration of the Lyman α is shown in Figure 13. Page 19 of 78

20 Figure 13: Semi-analytic surface density prediction, with escape fraction calculated from dust model. One interesting prediction from the two models is a fairly strong fall-off in the surface density of objects above a given line flux as a function of redshift. Table 2 does not really constrain this prediction much, but does show some interesting signs that the fall-off may not be as steep as predicted. In Figure 14 we show the calculated relationship between the measured Lyman α flux and the observed I-band AB magnitude for the objects in the semi-analytic model at redshift 3. In part due to the simplifying assumptions, there is a fairly tight relationship between the two. The solid line shows the median values. One can see that at the expected line flux limit for a 4 night observation, the median I band magnitude for the detected objects will be around Page 20 of 78

21 Figure 14: Semi-analytic prediction of the Observed I band magnitudes for a Lyman alpha selected sample. Finally in Figure 15 and Figure 16, we show the predicted clustering properties of objects selected on the basis of their Lyman α flux. As expected, the Lyman α emitters are strongly biased relative to the dark matter, with this bias factor expected to rise steeply as one goes back in redshift. This effect is similar to that seen for the Lyman break selected objects. Page 21 of 78

22 Figure 15: Spatial Correlation Function in co-moving coordinates for Lyman alpha emitters. Assumed constant fixed escape fraction. Single flux range. Points = Lyman alpha emitters, Lines = dark matter. Page 22 of 78

23 Spatial Correlation Function in co-moving coordinates for Lyman α emitters. Assumed constant fixed escape fraction. Two flux ranges Points = Lyman α emitters Lines = Dark Matter Figure 16: Spatial correlation function for two flux ranges Log (Number Mpc -3 ) Lyα emitters, 25% complete? What are we selecting? LBG, 80% complete? Log(SFR M /yr) Figure 17: Cartoon of sample selection effects for continuum and emission line selected samples. Page 23 of 78

24 In Figure 17 we try to illustrate how the Lyman break and Lyman α selected samples relate to one another. To do this, we introduce the idea of a star formation rate function analogous to the more commonly used luminosity function. Lyman break selection will generally pick out the most luminous objects with the highest star formation rates, and while missing a few of the most heavily obscured objects (expected to be the strong SCUBA sources), it will be fairly complete. Selecting on line emission will allow us to detect objects with much lower star formation rates, but will also be much more sensitive to the effects of obscuration, leading to a lower completeness. We note that this predicted lack of completeness is itself an interesting number that can be measured in the region where the LBG and emission line selection overlaps. If the above cartoon does reflect reality, it also shows that line selection will be an extremely powerful way of testing whether hierarchical formation models are in fact correct. 2.2 Additional Science In this section, we present a short shopping list of other science areas where a wide field optical IFU on a 30m class telescope might make exciting measurements. This list is certainly not complete, and also would deserve detailed analysis along the same lines as the deep field case above. Nearby Galaxies: surveys with the SAURON instrument at the WHT have demonstrated the power of IFU observations of the cores of nearby early type galaxies. Extremely complex relationships are being found between the stellar and gas dynamics. De-coupled cores, or hidden stellar disks are being found, along with distinct gas motions. These measurements are telling us about the formation of the galaxy cores, and the interactions between the infalling gas, star formation and also any supermassive black hole. This work could be extended to much lower surface brightnesses with a GSMT MEIFU, allowing us to probe the inner regions of disk galaxies in great detail, and to battle through a large fraction of the dust in these objects. Other areas where a wide field IFU would be powerful are in the study of nearby interacting galaxies, or in the physics of galactic superwinds (like that seen in M82, Figure 9). Stellar Science: Regions of comparatively low obscuration in star formation regions can be studied in the optical, as has been amply demonstrated by HST. Bipolar outflows, and interactions between these and the rest of the ISM, are an area currently of great interest. It seems unlikely that all the questions in this field will have been answered in the next few years, suggesting that a GSMT MEIFU could make a contribution here. Another possibly fruitful science area for MEIFU would be in crowded stellar environments such as towards the galactic centre, or rich star clusters and globular clusters. The red focus of the current instrument may not be an ideal match to this science case, but would allow access to the strong CaII triplet lines in the red. Page 24 of 78

25 3 Instrumentation Concept and Trade Analysis with Alternative concepts The above deep field science case was the original motivation for considering a wide field optical IFU, and has since been iterated back and forth with the proposed instrument design to obtain a better match between the two. In the process of doing this, several other instrument concepts were considered. Given a requirement for an instrument which will deliver a filled data cube with complete coverage of a region of sky in RA, Dec and wavelength, one is left with a choice between scanning instruments (scanning long slit, Fabry-Perot, Imaging Fourier Transform Spectrometers (IFTS)), and IFUs. More technically challenging routes might include imagers with some wavelength sensitivity, such as Superconducting Tunnel Junctions, but they will not be considered further here. At some level, the trade between an IFU and a scanning system is almost entirely one of cost. For a given area on the sky, if one can afford to build an IFU (and one can achieve reasonable throughput), then it will always be better to use an IFU as all the photons of interest are being collected all the time. This obviously comes at a considerable extra expense in detector area relative to the scanning options. That said, a two port IFTS also collects all the photons of interest all the time, albeit with the wavelength information encoded in a slightly complex way. The scanning process in this case consists of taking observations with a number of different optical path differences and seeing how the photons switch from one output port to the other. As a result of the above, it was considered worthwhile to compare the signal-to-noise achieved with and IFTS and to compare this with that from an IFU. For this we used a large part of the same MathCAD spreadsheet material used for the signal-to-noise calculation described in section 4.4, and so we will defer detailed description of the calculation method to there. Here we will just report the results. 3.1 IFU vs. IFTS Calculation summary: (see mathcad spreadsheet ELT_meifu.mcd or ELT_meifu.pdf for details) Exposure 3x4x8 hours (~10 5 seconds), sky I=19.9 IFTS 45% sky-to-hard-disk throughput 50% of object flux in 0.6x0.6 arcsec box IFTS wavelength domain S/N 5σ Detection for 3x10-19 ergs cm -2 s -1 for IFU same flux => S/N=0.034 for FTS fuzzy AO 0.2x0.2 arcsec box IFU S/N=15, FTS S/N=0.10 Diff Lim AO 0.006x0.006 arcsec box IFU S/N=108, FTS S/N=3.5 The above leads to the following list of advantages and disadvantages of an IFTS compared with an IFU for this deep field science case. Page 25 of 78

26 3.1.1 Pros Huge reduction in detector costs High throughput Relatively insensitive to readout noise Cons Very poor spectroscopic S/N performance in background dominated regime Non-intuitive noise properties As a result we have dropped consideration of an IFTS for this application. We note that the science case and the environment of the IFTS proposal for NGST (which was also dropped) seemed to us to be much more competitive and for that case the pros outweighed the cons in our view. Page 26 of 78

27 4 Budget Collections 4.1 Mass Instrument (Moving Cassegrain) Image Slicer Assembly 42 kg Re-Imaging Optics 24 x 92.2 kg = 2,214 kg Micro-Lens Assemblies 24 x 22.7 kg = 546 kg Spectrographs 24 x kg = 22,891 kg Structure 17,129 kg Total 42,822 kg Instrument (Fixed Cassegrain) Image Slicer Assembly 42 kg Re-Imaging Optics 24 x 92.2 kg = 2,214 kg Micro-Lens Assemblies 24 x 22.7 kg = 546 kg Spectrographs 24 x kg = 22,891 kg Structure 11,011 kg Total 36,704 kg Re-Imaging Optics 1 st Re-Imaging Mirror Assembly Mirror 0.1 kg Support 0.1 kg Subtotal 0.2 kg Fore-Optics Assembly Optical Elements (comprising Lens Elements 1-4 incl.) 18.5 kg Support Structure 23.4 kg Subtotal 41.9 kg Filter Assembly Filter 3 x 0.6 kg = 1.9 kg Filter Tray 1.7 kg Subtotal 3.6 kg Aft-Optics Assembly Optical Elements (comprising Lens Elements 5-8 incl.) 14.1 kg Support Structure 20.9 kg Subtotal 35.0 kg Support Structure & Drive Mechanism 11.6 kg Total 92.2 kg Page 27 of 78

28 Micro-Lens Assembly Micro-Lens Assembly Micro-Lens Array 4 x 2.5 kg = 10.1 kg Structure 12.6 kg Total 22.7 kg Spectrograph Collimator Assembly Optical Elements (comprising Lens Elements 1-3 incl.) Support Structure Subtotal Grism Wheel Assembly Interchangeable Optics Optical Elements (comprising Lens Elements 4-7 incl.) Grism Support Structure 50.5 kg 27.8 kg 78.3 kg 48.3 kg 12.5 kg 15.3 kg Subtotal 3 x 76.0 kg = kg Grism Wheel Subtotal kg Camera Assembly Optical Elements (comprising Lens Elements 8-15 incl.) kg Support Structure kg Detector kg Subtotal kg Spectrograph Support Structure & Drive Mechanism kg Total kg 4.2 Throughput An approximate throughput is calculated below. Transmission of the glasses in the lenses has been included. The throughput calculation uses the quoted values for the reflectivity of any mirrors and losses at each lens surface. The absorption losses were calculated from the optical design using ZEMAX. At this stage we have not included any wavelength dependence in the calculation, and so the numbers can be taken as a guide to throughput near the middle of the instruments wavelength range. Table 4: Throughput Calculation reflectivity of mirrors 0.97 transmission per lens surface 0.99 Surface losses element nmirror nlens tot trans Page 28 of 78

29 derotator pickoff foreoptics filter microlens spectrograph Grism 0.65 CCD 0.85 Absorption losses in glass Fopreoptics 0.98 Microlenses 1.00 Spectrograph 0.95 Total Throughput 0.26 We note that if VPH grisms could be used, a significant increase in efficiency would be possible. S. Barden kindly modelled the bluest grism for us, concluding that for the 243 l/mm case, with a 38 micron thickness and a index modulation, efficiencies in first order near 90% would be possible. However, VPH grisms with lower line densities than this (as are needed for the redder grisms) may not be possible. 4.3 Cost Below is a very approximate cost estimate. The savings due to mass production of 24 identical spectrographs was considered using the learning curve formalism kindly provided by L. Stepp. This formalism suggests that the saving is not a large factor, leading to the estimated cost of 24 spectrographs being roughly 71% of the cost of what one would estimate by multiplying the cost of the first by 24. We have also used the estimated cost of $22 per kg for large fabricated machined structures provided by L. Stepp to check that our estimate for the instrument support structure is reasonable. Obviously decisions between the moving and fixed gravity cassegrain structure (detailed in the mechanical design section below) may make substantial differences to this element of the cost. The spectrograph designs included cost as part of the design process, leading to their (relative) cheapness. However, in summary, we have rather little confidence in the cost estimate below. Lens Estimates (Scaled from 8m design) per spect nspec total lens cost off the shelf 3, ,000 custom 5, ,600 cost per run nrun total coating cost Page 29 of 78

30 Quotes for lens coating 1, ,600 total lens cost 259,200 currency conversion $365,515 Sample costs from PNS (used for scaling) US $ Design Cost (Phase A and B) $41,000 Optics (Main camera/collimator) $54,000 Optics (Filters/grating) $57,000 Mechanical $53,000 I&T $20,000 total cost excluding detectors $225,000 ELT MEIFU Estimates cost nunit total Design Costs $200,000 1 $200,000 Optics (cam/coll per spec) $365, $8,772,356 Optics (filters/gratings/lenslets per spec) $50, $1,200,000 Mechanical per spec $100, $2,400,000 I&T per spec $40, $960,000 CCD $20, $5,760,000 Controller $20, $960,000 Total $20,252,356 need to add optical de-rotator and fore-optics and overall mechanical packaging cost to this very uncertain 1,500, $2,115, New Bottom Line total $22,367, Add 25% contingency $27,959, Representative S/N Calculations The details of the signal to noise calculations are given in the mathcad document ELT_meifu.mcd, or the pdf version ELT_meifu.pdf. We consider three cases one with Natural seeing whatever that means, and two others with varying amounts of seeing correction. The fuzzy AO case is in fact what we hope will be the worst case scenario for behaviour of the GSMT, where tip-tilt correction along with some low order AO done on bright stars around the field lead to a moderate level of correction. The diffraction limited AO case we frankly suspect to be impossible to achieve in the optical, but is included as a limiting case. The calculations are for the 5σ detection limit for flux in a Lyman α line emission from an object at z=6. The sky is estimated using a broad band continuum magnitude and averaging this flux across the band. Obviously there will be substantial variations in actual signal=to-noise depending on the line location relative to the bright OH emission features. The calculation can be considered a crude average. Page 30 of 78

31 4.4.1 Natural Seeing Input Parameters: Exposure time of 4 nights i.e. 4x8 hours (~10 5 seconds), sky background I= % sky-to-hard-disk throughput 50% of object flux in a 0.6x0.6 arcsec box All of the line flux in 2 spectral pixels (1.7 nm) Results 5σ Detection for 3x10-19 ergs cm -2 s -1 Z=6 (Observed l=851.2 nm) gives 1x10 41 ergs s -1 luminosity Would need 4x4 nights to survey nm wavelength range (2.45<z<6.49) for full 3x3 arcminute field AO corrected and diffraction limited Assumption as in previous section, except: (a) fuzzy AO with 50% of object flux in 0.2x0.2 arcsec box 5σ Detection for 1x10-19 ergs cm -2 s -1 Z=6 (Observed l=851.2 nm) gives 3x10 40 ergs s -1 luminosity (b) diffraction limited AO with 50% of object flux in 0.006x0.006 arcsec box (50% EE diameter for 30m Airy Pattern at 790 nm) N.B. this also implies a different plate scale to sample this properly. 5σ Detection for 3x10-21 ergs cm -2 s -1 Z=6 (Observed l=851.2 nm) gives 1x10 39 ergs s -1 luminosity The above numbers indicate that GSMT with MEIFU will obtain high spectra resolution (R~1750, 3x600 = 1800 pixel long spectra) information for objects with redshifts between 4 and 7.2 down to limiting line fluxes of ergs cm -2 s -1 over a 3x3 arcminute field (with 0.1 arcsec spatial sampling) in a deep field exposure lasting 3x4 = 12 nights (taking the fuzzy AO case) Fluorescent Lyman Alpha Emitters A second interesting question comes from considering the source of the ionising photons that produce the line emission. In most of the science case above we were considering gas ionised by locally produced UV photons either from high mass recently formed stars or occasionally from non-thermal process around an active galactic nucleus. Now we would like to consider the possibility of detecting ALL of the gas overdensities in the universe after re-ionisation just from the flux excited by the general UV background. This Page 31 of 78

32 calculation has been published by Gould and Weinberg 1996, ApJ, 468, and applied to some data by Bunker, Marleau and Graham 1998, AJ, 116, In the spreadsheet uv_background.mcd or uv_background.pdf, we update and repeat these calculations to explore what integrated fluxes and surface brightnesses we might expect, using the latest estimates for the UV background (kindly provided by Haardt). A summary of the calculation is: Re-Calculate above using most recent estimates for the UV background Algorithm: Assume consensus cosmology Use Haardt and Madau (1996) functional form for UV background strength and shape as function of z, but with latest version for parameters (Haardt, Private Comunication, CUBA code) see Figure 18. Use a Power law approx. to hydrogen photoionisation cross section (simplifies the code, but not necessary). Consider clouds both optically thin and optically thick to UV background as separate cases using Gould and Weinberg formalism. Calculate the cloud luminosities given a characteristic size. Convert these to fluxes and surface brightnesses given the redshift see Figure 19 and Figure 20. Given the known number of absorbers as a function of column density and redshift, and assuming the same characteristic size, calculate volume density of clouds expected satisfying given flux and surface brightness limits see Figure 21 and Figure 22. Page 32 of 78

33 11 ( ) log Γ HM z p s 1 log ( ) Γ Q z p s 1 ( ) log Γ G z p s Figure 18: Some estimates for UV background versus redshift, HM = 1996 Haardt and Madau, Q = 2002 Haardt QSO only, G = 2002 Haardt QSO + Galaxies z p 14 ( ) log F thick R char, z p erg cm 2 s 1 log F thin N char, R char, z p erg cm 2 s 1 ( ) z p Figure 19: Predicted Lyman alpha flux from optically thick and optically thin fluorescent emitters Page 33 of 78

34 17 log log log SB thick ( R char, z p ) erg cm 2 s 1 arcsec 2 ( ) SB thin N char, R char, z p erg cm 2 s 1 arcsec 2 ( ) SB Bunker z p erg cm 2 s 1 arcsec Figure 20: Predicted Lyman alpha surface brightness from optically thick and optically thin fluorescent emitters z p R char ( ) arcsec d A z p Figure 21: Angular size of chosen charactaristic radius (50 kpc) as a function of redshifts for the consensus cosmology. z p Page 34 of 78

35 2 2.2 ( ) log ρ N thick, z p, R char Mpc Figure 22: Predicted space density of optically thick fluorescent Lyman alpha emitters z p The detectability of the fluorescent emission from optically thick gas will clearly depend rather sensitively on the geometry of the gas, but it does seem likely that at least out to redshifts of 3, one will expect to see ALL of the optically thick gas in a given volume. An interesting question will be whether there are any examples of such gas which do not also have their own embedded ionising source from star formation. Page 35 of 78

36 5 Optical Design 5.1 Introduction and commentary on Telescope Interface The proposed instrument was designed around the goal of having a number of spatial elements of the order of 1 million or more. It fills a hole in the range of instruments for integral field spectroscopy in terms of how the 3-D data box is filled with one image. At one extreme is the traditional spectrograph with one slit. It gives only one element in the spatial direction. At the other extreme are the Fabry-Perots, cameras with narrow band filters and Fourier Transform spectrometers. All give one element of information in the spectral direction. In between are the integral field systems but the spectral length in pixels is generally far larger than the number of spatial elements in the spectral direction. The present design is different. The number of spectral pixels is up to 600 while at least 800 slits 0.2 wide would be necessary to cover the whole field. The instrument therefore fills the hole between current integral field systems and instruments such as the Fabry- Perot that give one element of information in the spectral direction. The instrument contains two innovative characteristics, the integral field unit which uses the new concept of microslice systems developed at Durham University, and a low cost spectrograph design also based on a concept developed at Durham University Best telescope design for the instrument The GSMT telescope design proposed in summer 2002 is a standard Cassegrain. The main aberration in the field is coma while the second, astigmatism, is significant only near the edge of the 5.3 field. If the full field design of our ELT MEIFUS is adopted, or the option of changing the magnification for AO is retained, it would be better to remove the coma by slightly changing the design to a Richey-Chrétien. This involves a small change of the primary conic constant from -1 to which is no more difficult to polish compared to a sphere (constant = 0) but gives an OPD P-V difference of 12 λ compared to a parabola. The secondary would have a radius of curvature of mm and a conic constant of See Figure 23 for a PSF comparison. Page 36 of 78

37 Figure 23: PSF comparison for Cassegrain versus Richey-Chretien GSMT designs It would be even better for our instrument if the telescope has a Gregorian design instead of a Richey-Chrétien. One advantage is the sign of the field curvature. We need to put a system of pickoff mirrors in the focal plane that send the light away from the field optical axis. A field curved so that the edges are further away from the telescope than the centre would fit better with our system of pickoff mirrors. A Gregorian would also be better if ground layer AO correction is used with an adaptive secondary because the secondary would be nearer to the conjugate of the ground layer centre. This would lead to a larger corrected field of view. Using the approximation that a single AO mirror corrects for a half atmospheric thickness of r 0 /θ on each side of the mirror conjugate, one can calculate that the field would be 2.53 at 0.8 µm for a Gregorian telescope instead of a 1.68 field with a Richey-Chrétien (r0 = 20 cm at 0.5 µm, ground turbulence layer 300 m thick). The large field would fit well with the design we chose as the case study General description The design of the ELT MEIFUS is based on an extrapolation from the basic design of the 8-m MEIFUS developed at Durham University. Table 5 shows how the basic extrapolation was done in steps. Page 37 of 78

38 What Change Field Pixel Length Lens Width before after Total Spectr. arcmin arcsec m m mm Nothing x Pixel(µm) x # spectrog x Detector(K) x F-Ratio Camera x Table 5: Steps from 8m to 30m The spectra proposed here were 200 pixels long. However, the resulting design has a field with a diagonal more than 7 long while the field is 5.3. The design was therefore adapted to better match the science case and properties of the GSMT. The field was reduced while the spectral length was increased. The optical design presented below is for 24 spectrographs with 8k x 8k arrays of 2k x 2k detectors minus the 4 corners, leaving 12 x 2k x 2k detectors per spectrograph. The removal of the 4 corners significantly reduces the maximum field observed by the spectrograph, and hence reduces the spectrograph complexity, while lowering by only 25% the total number of pixels. The spectra were also chosen to be 600 pixels long to accommodate the scientific case, which reduces the field of view. To reduce the difficulty of the design, the bandwidth was limited to 0.6 µm to 1.0 µm. The pixel size was increased to 0.1 giving 0.2 slices and a focal ratio of 1.65, a value that seems fast but is still sufficiently slow Field of view and spectral length For a fixed detector, there is a choice between the number of spectral elements and the spectral length in pixels since the product of the 2 will be equal to the number of pixels used. Figure 24, Figure 25 and Figure 26 show different possible configurations. Page 38 of 78

39 Version 1: expensive 5.3 Aspect ratio of subfields = sky mosaic - 24 spectrographs - D total field = area = arcmin 2 - with 0.1 pixels on detectors of 8K x 8K, spectra 410 pixel long 7.93 Figure 24: Initial field mosaicing concept Page 39 of 78

40 Version 2: less expensive, Trilobite field Detectors made of 12 x 2k x 2k with 0.1 pixels. Easier than 8K x 8K because spectrograph field diameter = 15.3 instead of 19.3 sky mosaic Average field of 2.68 x 3.01 with 600 pixel spectra (diagonal of 4.03 ) Figure 25: Second mosaicing concept Page 40 of 78

41 Two different pickoff optics for a design that maximize field coverage - 24 spectrographs - D total field = pixels on detectors of 8K x 8K (tough ) -Sky mosaic Aspect ratio of subfields = Aspect ratio of subfields = More spectral length - Area = arcmin 2 - Spectra 410 pixel long - More field - Area = arcmin 2 - Spectra 365 pixel long Figure 26: Comparison of two mosaicing concepts The chosen design for this study is on Figure 25. The instrument field does not cover the telescope field completely, but still does cover a large part of it. Larger fields could be obtained by using array detectors of 8k x 8k (with the 4 corners) as shown in the other figures or reducing the spectral length in pixels. All of the proposed field shapes allow sky mosaicing without holes Basic optical design The instrument is made of a train of optical systems: the optical derotator (optional, derotation can be purely mechanical), the pickoff optics, the fore-optics, the lenslet arrays and the spectrographs. Figure 27 shows the different focal planes along the optical axis (assuming no optical derotator). Page 41 of 78

42 Figure 27: Schematic of Focal Planes The telescope image is cut in a series of subfields by the pickoff optics, and then each subfield is magnified by its own foreoptics with a larger magnification in the spectral direction. The lenslets arrays then cut the subfield into microslice images and reimage them, demagnified, at the entrance focal plane of each spectrograph. Figure 28 shows the optical train of each subfield. Page 42 of 78

43 cylindrical foreoptics IFU spectrograph Fore-optics magnify the focal plane anamorphically onto the first (rectangular) lenslet array which acts as the image slicer Cylindrical microlens arrays to divide the input focal plane into micro-slices Second lenslet array demagnifies (anamorphically) the slices onto the input focal plane of the spectrograph and puts the pupil onto the grating Figure 28: Schematic of Optical Path horizontal cylindrical microlens arrays vertical cylindrical microlens arrays 5.2 Derotator The derotation of the instrument can be done mechanically by placing the whole instrument on a turntable (Figure 29) or by a Cassegrain rotator, but the large weight of the instrument may render this option too expensive or imprecise. An optical derotator would then be necessary. Page 43 of 78

44 Figure 29: GSMT Fixed Gravity Cassegrain Environment Figure 30 shows a design scaled from one we made for the VLT. Page 44 of 78

45 flat mirrors plano-covex lens 336 mm Figure 30: Optical Derotator The derotator has a 4.3 field. It is made of one fixed lens in the focal plane and 5 mirrors in a box 540 x 1190 x 1620 mm rotating around the optical axis. Two mirrors are flat, two concave and one convex. The image quality of the derotator is very good even at the edge on the 4.3 circle. For the nearly square field of the ELT MEIFUS, typical 80%EE diameter is Pickoff Mirrors The system of pickoff mirrors in the focal plane is used to cut the field into 24 subfields and redirect the beams in 24 different directions toward the fore-optics of each subfield. Because the subfield is more magnified in the spectral direction than the spatial by the fore-optics, each pickoff mirror is much smaller in the spectral direction than in the spatial. Each pickoff is tilted with respect to the optical axis and has a convex shape. In the present design, the pickoff mirrors are all at an angle of 8 with the optical axis and approximately form 2 planes. This avoids the steps between the sub-mirrors that could mask part of the beams. Because of the telescope field curvature and the fixed pickoff angle, the mirrors are not exactly in focus. However, the defocus only matters at the pickoff mirror edge, where the beam will be truncated. The result is a loss of light of up to 50% along a small band at the edge. The width of that band is equal to the radius of the defocus that has a maximum value of Figure 31 shows how the pickoff mirrors are part of 2 groups each nearly forming a plane, one over and one under the focal plane and how the mirrors are not in the focal plane. Page 45 of 78

46 Figure 31: Pickoff Mirrors With a Gregorian telescope, the field curvature would be in the opposite direction so the mirrors would fit better with the focal plane and the edge defocus would be smaller. Each pickoff mirror sends the beam toward the first fore-optics. Figure 32 shows the shape of the pickoff that is similar in shape to the detector array but compressed in the spectral direction. Page 46 of 78

47 Figure 32: Pickoff Mirrors 5.4 Fore-optics The first optics of the fore-optics is a mirror. The tilt of that mirror can send the beam in any direction. Different configurations are possible. Figure 33 and Figure 34 show a porcupine configuration where the 24 optical axes diverge. Page 47 of 78

48 Figure 33: Fore-Optics Page 48 of 78

49 Figure 34: Fore-Optics In another configuration, the beams would be deviated before hitting the microlens arrays by passing through non-dispersive double prisms so that the spectrographs are parallel to each other. In another approach, the second and third re-imaging optics are reflective along each beam and the beams enter parallel spectrographs. The fore-optics magnify the field of each pickoff mirror onto a microslice array with, in the present design, an anamorphic magnification 4 times larger in the spectral direction than in the spatial. The reason for this anamorphic magnification is that the distance between microslice images on the detector would otherwise be too small in the spectral direction leaving no space for the spectra to extend. The present design uses a toroidal mirror and 4 toroidal doublets (Figure 35). Page 49 of 78

50 3.5 m Figure 35: Fore-Optics optical design The largest optics is 150 x 290 mm. Although a better design is certainly possible, the image quality is good enough to prove the basic principle of the design. The 50% Inslit Energy is 0.04 in the spatial direction and 0.05 in the spectral (Figure 36). Page 50 of 78

51 Figure 36: Fore-Optics PSF A good position for the filter is in front of the smallest doublet (Figure 37). The filter would have a clear aperture of 50 x 280 mm. Page 51 of 78

52 Figure 37: Fore-Optics detail 5.5 Lenslet Arrays The lenslet arrays are of a new concept developed at Durham University. It allows us to pack a much larger number of spectra on the detector of a spectrograph than the usual method. In the classic method, a microlens array of square or hexagonal microlenses is positioned in the focal plane of the telescope. Each microlens reimages the stop of the telescope (a pupil image) onto the focal plane of the spectrograph creating an array of spots with enough space between them for the spectra (Figure 38). Page 52 of 78

53 Figure 38: Microslice concept drawing The problem with this method is that it does not use the detector efficiently. A typical system will have spectra 2-3 pixels wide and a space between spectra of 4-5 pixels to avoid contamination. There is then typically 1 spectrum for 7 pixels in the spatial direction. With our new system, an array in the focal plane of the fore-optics cuts the field and creates an image of the stop of the telescope as in the old system but a second array reimages the focal plane of the fore-optics on the focal plane of the spectrograph. The information in the spatial direction is then saved. Moreover, the arrays are cylindrical instead of spherical permitting to have a different pitch and a different demagnification in the spatial and spectral directions. By making the pitch of the arrays in the fore-optics focal plane much larger in the spatial direction than in the spectral, the arrays cut the field in a large number of rectangular microfields similar to small slits or slices. There is a similarity in the way the arrays work to a slicer system. The new design of integral field unit was therefore called a microslice system. The different focal lengths in the spatial and spectral direction remove the anamorphic factor introduced by the foreoptics to increase the space between microslice images in the spectral direction. The advantage of this system is that every pixel along the microslice in the spatial direction gives a unique spectrum since the spatial information is not lost. Moreover, the distance between spectra can be smaller for the same average contamination since the pixels in the middle of the microslices are further away from the neighbour microslice images than the edge pixels. In the present design, the microslices are 12 pixels long and there are 3 pixels free between neighbouring microslices. This Page 53 of 78

54 gives 12 spectra for 15 pixels in the spatial direction. This is almost 6 times more spectra than with the previous lens array system. Figure 39 shows the design of the microslice system for our instrument. Slicing array with vertical microlenses Reimaging array with vertical microlenses Microlens of slicing array with horizontal microlenses Fore-optics field lens Microlens of reimaging array with horizontal microlenses 8 mm Figure 39: Microslice lenslet arrays The first 2 arrays are facing each other and have very large radii of curvature. They cut the field in microslices and adjust the position of the pupil image at the entrance of the spectrograph. The next array reimages the microslices on the spectrograph focal plane in the spatial direction. The last microlens array reimages the microslices in the spectral direction. Another important use of the microlens arrays is visible at the right of the Figure. The beam is not centred on the last microlens so its direction is changed. This is done by slightly reducing the pitch. It allows us to use the microlens arrays to redirect all of the beams from the different microlenses toward the entrance pupil of the spectrograph. This should in principle avoid the need for field lenses. However, a field lens was placed on the fore-optics side to improve the image quality. Note that the microlens are not cylindrical but have a conic constant and even a 6 th degree term. This should not be a problem since the master used to polish the microlenses can be machined to any shape. 5.6 Spectrograph Page 54 of 78

55 The spectrograph design is based on a general concept of low cost spectrograph developed at Durham University. It optimally uses the characteristics of the instrument to reduce the cost. A series of ideas have been applied here: a) the microlens arrays are used to redirect the beams toward the optical axis so that the entrance pupil is not at infinity. This reduces the size of the largest lens in the collimator, usually the largest lens of the spectrograph. b) The relatively small dispersion and spectral length combined with the use of a transmission grating with appropriate associated prisms reduce the aberrations caused by non parallel beams on the grating. This allows all the lenses to work together to cancel each other s aberrations instead of having a collimator and a camera independently designed. c) The grating is changed when changing of bandwidth instead of a complex mechanical system that would displace the camera. As long as the number of bandwidths is small, this is inexpensive. The number of bandwidths in our design is 3, 0.6 µm to µm, µm to µm and µm to 1 µm. The corresponding grism line densities are 243 l/mm, 206 l/mm and 172 l/mm. d) The small bandwidth of the spectra permits to avoid expensive glasses as CaF2 by replacing some of the lenses with the grating when changing of bandwidth. The changing of the lenses is critically important and could have lead to unacceptable aberrations. Fortunately, a design was found that gives a typical 50%EE diameter of 0.11 (1.1 pixels). This result completely proves the theoretical ideas used in the design of the spectrographs. A compromise had to be made however because of the fast focal ratio of the camera and to simplify the design work: 2 doublets must be changed with the gratings leaving 2 gratings and 4 doublets unused at all time. Since the optics are some of the smaller ones, this should not significantly increases the cost. Figure 40 shows the best spectrograph design found so far. Page 55 of 78

56 Volume Phase Holographic Grating Optics to change when changing of bandwidth Figure 40: Spectrograph Design Page 56 of 78

57 6 Mechanical Design 6.1 Introduction The scope of the mechanical design work package is to provide a conceptual mechanical design to illustrate how the instrument may be implemented, and to demonstrate the viability -at least from a mechanical design point of view- of the optical design. No attempt has been made to optimise the mechanical design; this implies that in certain areas the structure may be grossly over-designed and too heavy, whereas in other areas the structure may be inadequate to support the optics to the accuracy required. The mass budget presented in Section 4.1 consequently represents our current best guesstimate (based on experience from previous instrumentation work) and should therefore be treated with caution. It is recommended that future design studies include a full structural analysis of the mechanical design followed by a re-design where necessary. 6.2 Subsystem Design Following on from the optical design format, the following sub-systems were identified: 1. Image Slicer Assembly, 2. Re-Imaging Optics, 3. Micro-Lens Arrays, and 4. Spectrograph. The mechanical design of each of these sub-systems will be discussed in detail in the following sections. The packaging and design of the entire instrument will be discussed in detail in Section Image Slicer Assembly The Image Slicer consists of an assembly of 24 identical pick-off mirrors. Each mirror features a concave spherical optical surface with identical radius of curvature; half of the mirrors are tilted by + 88 about the axis defining the spatial direction, the other half are tilted by - 88 (see Figure 41). The mirror s dimensions are mm in the spatial and spectral directions respectively, giving overall dimensions of mm for the array. The total mass of the Image Slicer Assembly (including the support plate) is approximately 42 kg. Page 57 of 78

58 Figure 41: Image Slicer Assembly An interesting option is to manufacture these mirrors by means of a diamond turning process, where the mirrors are off-set with respect to the machine s rotation axis to implement the required tilt of the optical surface. In this case, an accurately machined blank would be positioned against datum points (dowel pins) on the machining jig to ensure that the optical surface is accurately aligned and positioned with respect to the component s datum surfaces. As all optical components are identical, only one jig is required, which further adds to the accuracy that can be obtained. The only variable would then be the distance of the optical surface with respect to the mirror s back surface (to allow the mirrors to be mounted onto a flat plate), which could easily be achieved by changing this parameter in the machining program. This strategy would then greatly simplify the integration of the Image Slicer Assembly, as the mirrors alignment is achieved by simply stacking the individual components into the support plate. The Astronomical Instrumentation Group at the University of Durham have extensive experience with the manufacture through diamond machining of metal optics for infrared instrumentation and is currently pursuing a development program to extend the application of this technology to the visible wavelength range Re-Imaging Optics The baseline design for the re-imaging optics assembly comprises a toroidal re-imaging mirror and 4 toroidal doublets. The largest optical element measures mm in the spatial and spectral directions respectively. In order to minimise their size (and consequently cost), the filters are located in front of the smallest doublet, which measures Page 58 of 78

59 mm. This provides a natural division of the re-imaging optics assembly into two sections; one section containing two doublets mounted in their lens cell located in front of the filter, and the other section containing the remaining two doublets and associated lens cell behind the filter. The respective lens cells are attached to the main reimaging optics support structure (which essentially is a thin walled box with reinforcements at the attachment points for the lens cells), which also houses the filter exchange mechanism (see Figure 42). Each lens cell provides alignment facilities for the doublets; the main structure provides alignment facilities for both the fore- and aft optics barrels to facilitate the internal alignment of the re-imaging optics assembly. The reimaging optics support structure also forms the interface to the primary instrument support structure and provides the alignment facilities required to align the re-imaging optics assemblies with respect to the other sub-assemblies (i.e. the 1 st re-imaging mirror, micro-lens array and the spectrograph) for each optical path (sub-field). Figure 42: Re-Imaging Optics Assembly (1st Re-Imaging Mirror not shown) As was already highlighted in Section 5.3, the second and third optical elements (i.e. the first and second lens doublets) may be replaced by reflective optics (re-imaging mirrors). As will be discussed briefly in Section 6.3 this may have considerable advantages for the overall instrument design, as it would then be possible to co-align the optical axes of the re-imaging optics assemblies and spectrographs. These would then all be parallel, which would significantly simplify the design of the instrument support structure, and facilitate their subsequent installation into and alignment with the instrument. Alternatively, Page 59 of 78

60 Section 5.3 mentions the use of non-dispersive double prisms to re-direct the light before it enters the micro-lens arrays, which would also allow the spectrographs to be parallel to each other. While these options are mentioned here because they open up promising avenues for future optimisation of the instrument s opto-mechanical design, they have not been investigated in detail. Three filters are provided, one for each wavelength band. The filters have a clear aperture of mm. These filters are mounted in a filter slide, which can be driven in the spectral direction to insert the required filter into the optical path. As the filters and the slide are not particularly heavy, the requirements on the drive mechanism are not expected to be extremely demanding, and a simple stepper motor driven linear slide mechanism is expected to be adequate. Operational redundancy may be provided by a simple mechanical or optical switch, which can be used to confirm that the required filter is inserted into the beam. All toroidal re-imaging mirrors have identical radii of curvature in the spatial and spectral direction. The tilt of the optical surface about the axes defining the spatial and spectral direction, however, is unique for each mirror. Although the re-imaging mirrors are therefore not identical, it may be possible to generate the optical surfaces using diamond machining, but conventional figuring methods, or replication techniques may also be suitable. Each re-imaging mirror will be supported by a conventional three-point mount to enable alignment relative to the slicing mirrors and the rest of the fore-optics assembly. The reimaging mirrors have a diameter of 60.0 mm. The total mass of each re-imaging optics assembly is approximately 92.0 kg for the main optics assembly, and 0.2 kg for each reimaging mirror assembly Micro-Lens Arrays The micro-lens assembly consists of 4 micro-lens arrays with a clear aperture of x mm. Each array comprises either 68 or 101 cylindrical micro-lenses, depending on whether the array divides the field in respectively the spatial or the spectral direction. (see Figure 43). The assumption has been made that the individual micro-lenses will be mounted inside a frame; each micro-lens features accurately machined datum surfaces, which guarantee their alignment relative to the other lenses in the array. The arrays will then be mounted inside the micro-lens support structure (essentially a light-weighted, thin-walled aluminium box), which will provide the facilities for the alignment of the arrays. This structure also forms the interface to the primary instrument support structure and provides the facilities to align the micro-lens assemblies with respect to the other sub-assemblies (i.e. the re-imaging optics and the spectrograph) inside each optical path (sub-field). The mass for each micro-lens assembly is approximately 22.7 kg. Page 60 of 78

61 Figure 43: Micro-Lens Arrays Spectrograph A schematic drawing of the spectrograph is given in Figure 44. The presence of the interchangeable optics naturally divides the spectrograph into three sections: 1. the Collimator, which comprises lens elements 1-3 incl. and is located in front of the interchangeable optics, 2. the interchangeable optics, which comprises the grism and lens elements 4-7 incl., and 3. the Camera, which comprises lens elements 8-15 incl. and is located behind the interchangeable optics. The largest optical element (lens element 1) has a diameter of 480 mm. Each spectrograph has a mass of approximately kg. Page 61 of 78

62 Figure 44: Spectrograph Three sets of interchangeable optics, consisting of a grism and associated lenses (lens elements 4-7), are provided - one for each wavelength band. They are mounted onto the grism wheel, which is driven by a stepper motor to enable the appropriate set of optics to be inserted into the optical path (see Figure 45). Each grism is housed in a mounting frame, which provides facilities for the alignment of the grisms relative to each other. Two lens cells contain the lenses positioned before (lenses 4 and 5) and after (lenses 6 and 7) the grism respectively. Where necessary, these cells provide alignment facilities for the lenses (as each cell contains only two single lenses this may not be necessary, which would greatly simplify the design of the lens cells). Both lens cells are mounted to the grism wheel, which provides the necessary facilities for the alignment of the lens assemblies relative to the grisms. A wedge is installed between the wheel and the aft lens assembly to provide the required offset angle of 3.36 degrees between the optics located before and after the grism. Page 62 of 78

63 Figure 45: Grism Wheel Assembly Page 63 of 78

64 It is envisaged that the grism wheel will run on roller bearings located around the wheel s perimeter. The large mass of the grism wheel assembly (~ 250 kg) will require very careful design of the mechanism, particularly in the case where the gravity vector is allowed to vary in direction, as is the case for the moving gravity vector Cassegrain instrument option. Tight tolerances on the machining quality and installation accuracy of the track (which will be integral to the wheel) and rollers will minimise any positioning errors of the interchangeable optics in the radial direction and misalignment (rotations) with respect to the spectrograph s optical path. The drive mechanism will be carefully designed to minimise positioning errors in the tangential direction:- in order to achieve a 25 µm positioning accuracy, the angular error shall not exceed 100 µrad, assuming that the centre of the optical elements is at a distance of 250 mm from the wheel s rotation axis. It may be possible to achieve this positioning accuracy using an open loop control system which drives the wheel with stepper motors through a worm drive in combination with an anti-backlash mechanism. A 16 bit (optical) encoder could be used to provide the necessary feedback should a (more complex) closed-loop control system be required. Alternatively, the alignment errors may be minimised using a ratchet type mechanism, where the wheel is driven back into a hard stop. The main structure, which encases the grism wheel assembly, also supports the Collimator and Camera assemblies, which comprise the optics respectively located in front (lens elements 1-3), and behind (lens elements 8-15) the interchangeable optics. The collimator optics are housed in a single lens cell, which provides alignment facilities for each lens. The camera optics are further divided into two lens groups (the first comprising lenses 8-10, the second lenses 11-15), which are each housed in their individual lens cells, which in turn are housed inside the camera optics barrel. Each lens cell provides alignment facilities for the individual lenses and the camera optics barrel provides facilities to align both cells with respect to each other. A wedge is installed between the spectrograph support structure and the camera optics barrel to provide the required offset angle of 3.36 degrees between the optics located before and after the grism. The main spectrograph support structure provides the facilities required to align the collimator and camera optics assemblies with respect to the rest of the spectrograph. The spectrograph support structure also forms the interface to the primary instrument support structure and provides the alignment facilities required to align the spectrograph with respect to the other sub-assemblies (i.e. the re-imaging optics assembly and microlens array) for each optical path (sub-field). A very coarse, back-of-the-envelope style tolerance analysis was carried out for the optomechanical design of the spectrograph, the results of which are presented in Appendix 9.3. Bearing in mind that no attempt has been made to optimise the mechanical design of the spectrograph, and that a more detailed and comprehensive analysis (using finite element models) will be required to verify the very preliminary conclusions presented here, the analysis does not give any suggestion that the required optical tolerances will be extraordinarily difficult to achieve. A point of particular caution, however, concerns the tolerance on the alignment of the various sub-systems along each optical path (i.e. pickoff mirror, re-imaging optics, micro-lens arrays and spectrograph):- no attempt has been made to estimate the required tolerances, or to verify the performance of the primary instrument support structure. It is recommended that these form part of future studies. Page 64 of 78

65 6.3 Instrument Design The objective of the overall instrument design study was to identify possible solutions for the packaging of the optical sub-assemblies described in the previous sections. The main constraint is formed by the available space envelope for the instrument, which is defined in Figure 46, whilst additional constraints may be placed on the total instrument mass. Whilst every attempt has been made to closely follow the optical design prescriptions presented in Section 5, the baseline optical design revealed some problems which affect its mechanical feasibility (see Figure 47), i.e.: 1. (some of) the pick-off mirrors are vignetted by the fore-optics and the associated support structure; 2. (some of) the re-imaging mirrors are vignetted by the fore-optics and the associated support structure; 3. the spacing between the re-imaging mirrors is zero, thus not allowing for any manufacturing and installation tolerances; 4. the spacing between the first (lens) elements of the fore optics assemblies is less than 10 mm, which does not allow sufficient space for the support structure; 5. there is not enough space to accommodate the filter mechanisms; 6. there is not enough space to accommodate the grism wheel assemblies. This implies that the baseline opto-mechanical design has to be fine-tuned to eliminate the identified problems. This can be achieved, in principle, by changing the orientation angles of the re-imaging mirrors with a view to increasing the space between the optical paths. This will obviously increase the image aberrations and thus deteriorate the image quality. The effects on the optical performance caused by the implemented changes to the optical design have not been quantified at this stage, and the feasibility of the proposed changes will therefore have to be verified during the next phase of the instrument design study. As already highlighted in Section 6.2, various alternative possibilities exist to increase the spacing between adjacent optical paths (i.e. to replace the first two lens doublets in the reimaging optics with reflective optics, or to insert non-dispersive double prisms in front of the micro-lens arrays). Whilst these alternatives are mentioned here because they open up promising avenues for future optimisation of the instrument s opto-mechanical design, they have not been investigated further. Page 65 of 78

66 R ELEVATION AXIS Figure 46: Telescope Interface and Available Instrumentation Envelope Page 66 of 78

67 6.5 m 1.0 m Focal plane 5.3 m Figure 47: Baseline Instrument Lay-Out (Structure not shown) Page 67 of 78

68 Two alternatives instrument configurations were investigated: The moving gravity vector Cassegrain instrument, which is attached directly to the back of the telescope and consequently moves around with the telescope, as it is pointed at different areas of the sky. This means that the mechanical load on the structure constantly varies, as the direction of the gravity vector relative to the instrument changes with the telescope s orientation. The fixed gravity vector Cassegrain instrument, which is mounted onto an instrument table located between the telescope elevation bearings, and consequently only moves in azimuth (but not in elevation) as the telescope is pointed at different areas of the sky (see Figure 46). This means that the mechanical load on the instrument s structure is non-variant, as the direction of the gravity vector relative to the instrument does not change when the telescope is pointed around the sky. The same optical sub-assemblies (i.e. re-imaging optics, spectrograph, etc.) were employed for both instrument options. This ignores the fact that for the fixed gravity vector instrument, the structural design of these sub-assemblies could be further optimised with a view to saving mass. The instrument was designed in a modular fashion to facilitate both integration and test, and serviceability of the instrument. This was preferred over a more integrated approach using e.g. large filter trays and grism trays to change all filters and grisms simultaneously. It was believed that the mechanisms required for this approach would become very large and heavy, and that it would be almost impossible to achieve the stiffness required to maintain the alignment of the optical components. Because the subassemblies could be assembled and tested independently, the modular approach has the additional benefit that (if preferred) the instrument could be integrated in stages; i.e. the instrument would initially be delivered with limited functionality (i.e. with a limited number of spectrographs), but additional spectrographs could be added as and when they become available. A similar rationale applies to the serviceability of the instrument: if one sub-assembly were to fail, it could easily be removed for repair whilst the instrument could remain operational (although with a slightly degraded functionality) Moving Gravity Vector Cassegrain Instrument In the moving gravity vector Cassegrain configuration, the instrument is attached directly to the back of the telescope. The assumption has been made, that in this configuration, the instrument support table (see Figure 46) has been removed to free up the maximum amount of space for the instrument. In order to alleviate the problems highlighted in Section 6.3, the optical paths have been spread out in order to increase the spacing between the sub-assemblies. Also, the optical paths have been re-arranged to allow for a more compact lay-out of the instrument (see Figure 48). This implies that the tilt of the slicing mirrors is no longer identical, which means that it is no longer possible to manufacture these mirrors in a single set-up (but it is - of course - still possible to diamond machine the mirrors). The effects on the optical performance caused by the implemented changes to the optical design have not been analysed at this stage, and the feasibility of the proposed changes will therefore have to Page 68 of 78

69 be verified during the next phase of the instrument design study, should the moving gravity vector Cassegrain instrument be selected for further development. 1.0 m Focal plane 5.3 m 6.0 m Figure 48: Moving Gravity Vector Cassegrain Instrument Lay-Out Page 69 of 78

70 1.8 m Focal plane 2.0 m 11.1 m 0.5 m 6.3 m Figure 49: Fixed Gravity Vector Cassegrain Instrument Lay-Out Page 70 of 78