IMACS User Manual

A User’s Manual for IMACS: Version March 2016 – including dewar swap updates in 2018

• IMACS Description
• Characteristics & Planning
• Preparing to observe
• IMACS Observing

Welcome to IMACS, the Inamori-Magellan Areal Camera & Spectrograph.  IMACS offers a wide range of imaging and spectroscopic options, and as such lends itself to both standard and novel observational programs.  Because of the large number of modes in which IMACS can be used, it is unlike any imaging spectrograph, including GMOS, DEIMOS, and VIMOS, all of which are limited to a a few spectral and imaging modes — sometimes only one. As a result of the versatility of IMACS, this User Manual is itself wide ranging and contains much more information than many observers will need to know. At the same time, it will not cover important details and methods that will be appropriate to more custom observational programs, including the specific “cookbooks” that are available to lead an observer through important tasks like object aquisition and mutlislit mask alignment.  However,  If a specific mode of operation is not discussed here or in the cookbooks, observers should not assume that the IMACS software, or the interface software for the Baade telescope, can be readily modified to accommodate the desired observations.  In such cases, observers should query the IMACS Team and/or the Magellan Support Staff before showing up at the telescope.  Requests for novel modes, if they can be supported, may require weeks or even months of advance notice.  Because the Magellan operation is a lean one, some things that IMACS could do will not be supported by software and/or the Magellan Support Staff. Again. if you read about it here, it can be done without special requests of the Team or the Staff.  Requests for clarification of the standard modes are encouraged — after a careful reading of the available documentation, please. 

A thorough discussion the design goals, construction, and implementation of IMACS can be found at http://www.jstor.org/stable/pdf/10.1086/658908.pdf

The User Manual is divided into four parts — use the collapsible “outline” menu at right to navigate.  The first describes IMACS, in brief and then in detail; the second covers capabilities, characteristics (including performance), and other considerations that are relevant to planning programs; the third is about preparing for IMACS observations; and the fourth is a tutorial in making common observations.

As the IMACS PI, I have added some recommendations, denoted by my initials.  These are my opinions, of course, and others may see things differently.

     – Alan Dressler

1. A DESCRIPTION OF IMACS

A. IMACS: A Thumbnail Sketch 

The Inamori-Magellan Areal Camera and Spectrograph, IMACS, was built at Carnegie Observatories to provide a wide-field imaging and spectroscopic capability for the Magellan Baade Telescope.  A principle design goal for IMACS was to provide a versatile spectrograph that would support many types of scientific investigations with minimal compromise in the quality and sensitivity of any of these diverse modes. Pictures of the development work can be seen at the IMACS project website .

IMACS offers a choice of two cameras — f/2 and f/4 — which provide two different scales and spectral resolutions covering the range of 20 < R < 20,000.  As of March 2012 there are two 8K x 8K CCD mosaic cameras, one dedicated to each foci. In addition to wide-field imaging and a range of low- to medium-resolution spectroscopic modes, IMACS has four ‘accessory’ modules: (1) a 2 x 1000 fiber-fed integral field unit (IFU) built by Durham University; (2) a mutli-object echelle mode (MOE); (3) a full-field tunable filter (MMTF); and (4) an image-slicing reformator for dense mutlislit coverage over a 4 x 4 arcminute field (GISMO).


B. IMACS: A More Detailed Description

Basic Component Layout

IMACS is a reimaging spectrograph mounted permanently on the Magellan Baade Telescope. It is fed by the telescope’s 6.5m primary, f/11 Gregorian secondary, and flat tertiary mirrors, and a corrector.  This optical train delivers an unvignetted field out to R = 12 arcmin with 9% vignetting at R = 15 arcmin. The all-transmitting (all-spherical-optics) collimator in IMACS contributes very little in the way of additional aberrations (FWHM < 0.10 arcsec). The Gregorian secondary was specifically designed to feed the IMACS collimator to produce this excellent performance — in effect, the Gregorian secondary is the first optical element in the IMACS collimator. Further detail on the Baade Telescope can be found here

The rotating tertiary mirror in the telescope directs the beam to IMACS on the west Nasmyth platform, or to Fourstar, the wide-field niear-IR on the east Nasmyth platform, or to the auxiliary folded mounting the spectrograph FIRE (Folded-port InfraRed Echellette). A typical time of 15 minutes is required to make one of these changes. The wider field of IMACS requires a radially mounted, integral ADC (atmospheric dispersion compensating) corrector. Prior to May 2004 IMACS was used without a corrector, which means that data from the first 10 months of operation showed significant aberrations outside of a field radius of 5 arcminutes, and also non-negligible spherical aberration on-axis.

The Optical Layout of the f/4 and f/2 channels of IMACS

The f/4 Channel in IMACS -- Imaging and Spectroscopy

IMACS has two cameras, f/2 and f/4, but unlike the classical double spectrograph divided by colors, the two cameras offer different imaging scales and dispersions, as such, only one mode is used at a time. The all-transmitting (all-spherical optics) f/4.3 "long" camera (known as f/4) uses an overcoated aluminum mirror for direct imaging. The f/4 camera provides a 22-arcmin diameter field which, when combined with the detector, fully illuminates a 15.4 x 15.4 arcmin field at a (central) scale of 0.111 arcsec per pixel. Imaging with the f/4 camera is excellent and suitable for photometric programs.  Image quality of r < 0.45 arcsec FWHM  has been recorded over the entire field in the best seeing that has been available; potentially, the image quality should be r < 0.35 arcsec FWHM over the entire field in the best Magellan seeiing, ~0.25 arcsec. 

Spectroscopy with the f/4 is by means of standard 150 x 200 mm (Richardson) reflection gratings that are placed in the parallel beam produced by the collimator. The giant wheel that carries the dispersive elements in IMACS accommodates any three gratings; these are quickly changed during a night's observing.  Harland Epps' design provides for wavelength coverage of 3650 Å < Lambda < 10000 Å without refocus. The spectral resolutions cover a range of 1-12 Å FWHM (for a 1.0 arcsec slit-width).  A script (click here) calculates the grating angle for a central wavelength and range. The convention that has been adopted is that zero-order is defined as grating tilt = 0 degrees. The IMACS Multi-Object Echelle (MOE, see below) is used (only) with the f/4 camera. MOE provides a spectral resolution of ~0.25 Å.

The IMACS f/4 camera transmits down to 3300 Å with a modest defocus; the typical use is for filters or spectroscopy stopping at about 3500 Å The enhanced blue response of the Mosaic3 CCD camera, installed in 2012, makes IMACS Observations in the UV more efficient than previously. The throughput using the 600-l grating as of 2015 is only 5% at 3600 Å (see graph below). However, this  disappointing value probably reflects the large offset from the ~5000 Å blaze rather than the UV transmission of the f/4 camera or the UV quantum efficiency of the Mosaic3 camera, predicted to be ~50%.  A grating blazed for UV use would likely raise the throughput for UV spectroscopic observations to ~10-15% -- we are in the process of acquiring a 600-l grating blazed at 4000 Å.

To calculate the FWHM spectral resolution in angstroms divide the intended slit width by 0.11 arcsec and multiply by the dispersion for the chosen grating. as listed in the table below. 

Gratings available for the f/4 channel

The f/2 Channel in IMACS -- Imaging and Spectroscopy

The all-transmitting, double-asphere, glass-and-oil-lens f/2.5 "short" camera (known as f/2) works in-line with the IMACS collimator (no reflection) for imaging.  The f/2 focus delivers an 27.4 arcmin diameter field at a (central) scale of 0.200 arcsec per pixel. The field is vignetted in the corners (by the tertiary and its mounting assembly), changing from 0% at R = 12 arcmin to ~10% at R = 15 arcmin. Beyond this, light is cut off by bafflles inside the camera: the corners of the square field of the CCD mosaic camera are missing -- this is a consideration for 'tiling' a field. This vignetted area is also subject to "shadowing" by the two IMACS outboard guiders, which move over 40 degree arcs "above and below" the 27 arcmin field.  The Epps design provides for wavelength coverage of 3900 Å < Lambda < 10500 Å without  refocus.

Up until June 2006 the f/2 camera field suffered from a tilted field that limited image quality to 0.45 arcsec in the field center and 0.60 arcsec at the field edges.  The Mosaic1 CCD camera was tilted to compensate this; following this correction, a signficant amount of coma and astigmatism was also found, in excess of the design. These were substantially reduced in 2008 by translation and piston of the last element in the camera. The combination of camera tilt -- now the Mosaic2 camera -- is producing very good images over most of the f/2 field, with a small degradation  of elliptical (but still relatively tight) images at some parts of the field's edge.  In the best seeing (FWHM = 0.25 arcsec) the f/2 camera would produce images of 0.40 arcsec (2 pixels) or better over most of the field.  In more common but still excellent seeing of FWHM = 0.50 it will deliver 0.55" or better images over most of the field, with a few patches of 0.60" at the edge.  Median seeing of FWHM = 0.65 arcsec will be minimally degraded by the camera, consistent with original design specification.

From IMACS commissioning (10/2003) through 2006 the baffling required for the f/2 field was not in place.  This has been corrected and, although baffling of the f/2 field may not be not perfect, it appears now to be adequate for general imaging programs, including accurate photometry --- flat fielding is effective to the few percent or better level.

Spectroscopy with the f/2 camera is done with a grism, a transmission grating replicated on a prism in order to pass the 1st order at zero deviation. The IMACS grisms were also manufactured by Richardson.  Grisms are inserted into the parallel beam generated by the collimator. Two of the five available grisms can be mounted in the dissperser wheel. The available grisms produce spectral resolutions of 6-10 Å FWHM for a slit width of 1.0 arcsec, corresponding to spectral resolutions of R=500-1200 at the blaze wavelength. In spectroscopic mode the full-area of the CCD mosaic is illuminated; that is, after accounting for the loss due to vignetting at the position of the object, the spectrum is uniformly diminished by that amount from the center out to the corners. 

The parallel light of the beam is the idea place for narrow-band filters, because light from each part of the field travels through the full aperture of the filter.  Narrow-band imaging can also be done with the Maryland-Magellan Tunable Filter (MMTF), a tunable Fabry-Perot etalon that can only used with the f/2 camera.

In 2005 Scott Burles produced a low-dispersion prism (LDP) for use in IMACS.  Prisms offer the highest throughput of the IMACS spectral modes, but at a low and rapidly varying red-to-blue spectral resolution of 20 < R <150.  In 2008 a second-generation UDP (uniform dispersion prism) designed and built by Steve Shectman was added: the UDP enables determination of spectro-photometric redshifts of galaxies with an accuracy of order 1% for thousands of galaxies in a single setup. The UDP is described below.

To calculate the spectral resolution in angstroms of a planned observation, divide the intended slit width by 0.20 and multiply by the dispersion for the chosen grism. Reference emission-line spectra for He, and Ar arcs spectra for IMACS dispersers in long slit mode are available in the IMACS spectral atlas, and also at the telescope, under "Calibrations_and_Tools" on each of the datadisks on the observer workstations.  Grisms available for the f/2 channel

* Due to a manufacturing error, the ruling on the 400/mm grism is slightly rotated with respect to the prism. This results in a leftward shift of the dispersed images. Observers should be aware that targets more than approximately 12.3 arcmin left of the center (as viewed in the QL tool or in maskgen) will be shifted off of the detector mosaic.  

Either camera, f/2 or f/4, can be used for (1) single object spectroscopy using the slit-viewing feature of the centerfield guider, (2) for multislit spectroscopy using laser-cut masks, and (3) with the Durham-IMACS Integral Field Unit (IFU). These modes are described in more detail below.  It is not necessary to use a slit mask when observing a single object over an area of 10 arcsec or less!  Direct slit viewing provides a particularly efficient way of making flux calibrations with standard stars. (However, as is explained below, this f/2 spectroscopic mode cannot be used if the grisms are rotated to the N&S orientation, because the slits will not be perpendicular to the dispersion.)

Observers with a program of f/4 single-slit spectroscopy have noted that once they position an object or objects along the slit, the image seems to shift from this position once a spectrum is taken. That is, the pixel positions in the area move to the right or left from where they were in the direct image.  (The position of objects relative to the slit itself does not change.) This relates to exact alignment of the mirror used for imaging and that of each spectral grating -- technically, the roll angle of these elements.  A misalignment of only 0.02 deg of the grating produces a shift of about 10 pixels with respect to imaging.  Anything larger that than 10 pixels indicates that the grating has moved in its cell with respect to roll angle (probably by the loss of shims) and must be re-aligned.  Although shifts of up to 100 pixels have been reported, even such large shift does not effect the quality of the data -- the angle from normal is still very small.  However, a sizeable shift is a problem for setup, for example, making sure that objects do not disappear in a CCD chip gap.

The IMACS Mosaic CCD Cameras

The IMACS Mosaic1 detector  served both the f/2 and f/4 focus until 2008.  Mosaic2 was installed in 2008 to serve the f/2 camera, and Mosaic3 -- a rebuild of Mosaic -- was installed at f/4 in 2012.  Mosaic2 and Mosaic3 each use eight thinned CCDs manufactured by E2V to produce a 8192 x 8192 pixel mosaic. These CCDs are all 2K x 4K x 15-micron devices; Mosaic2 and Mosaic3 use CCDs with different antireflection coatings.   The following table shows the average quantum efficiency (in percent) for the 8 chips in Mosaic2 and Mosaic3 (standard E2V of Mosaic2 and the E2V deep depletion CCDs used in Mosaic3 -  errors are the standard deviations of the mean values). 

As of December 2017, Mosaic2 is on f/4 and Mosaic3 is on f/2. 

 Quantum Efficiency of CCD Detectors

The CCD configuration of Mosaic2 and Mosaic3 is the same.   With two available CCD cameras, the observer is free to switch between foci during observing -- the change is simple and rapid. The operational aspects of this are discussed below. For the defunct Mosaic1 the CCDs were parallel to within ~5 pixels, with of gaps of about 0.93 mm = 12.4" (62 pixels); the gaps were about the same between both the short axes and long axes of the CCDs.  The Mosaic2 camera CCDs are aligned to within 1-2 pixels. The gap between the short sides of the CCDs are 0.86 mm = 11.4 arcsec (57 pixels).  The gaps between the longer sides are larger, about 1.38 mm = 18.4 arcsec (92 pixels). The Mosaic3 camera CCDs are also aligned to within 1-2 pixels, but the gap spacings are different. The gap between the short sides of the CCDs are 1.08 mm = 8.0 arcsec (72 pixels).  The gaps between the longer sides are about 1.37 mm = 10.1 arcsec (91 pixels).

Readout time, gain, and read noise (as of Aug. 2018) are as follows:

Mosaic3 (f/2) FAST readout (recommended)  -- 82 seconds   (2x2 binning -- 29 seconds)

Mosiac3 (f/2) SLOW readout -- 144 seconds   (2x2 binning -- 46 seconds) 

Mosaic2 (f/4) FAST readout (recommended) -- 81 seconds   (2x2 binning -- 29 seconds)

Mosaic2 (f/4) SLOW readout -- 144 seconds   (2x2 binning -- 46 seconds)

The Mosaic2 and Mosaic3 CCD cameras are normally run in 1x1 or 2x2 binning for science observations, and there is a 4x4 "Snap" exposure mode that reads the CCDs in 4x4 binning for checking setups.  Any combination of binning can be used, for example, 2x3, 3x4, 1x3, etc.

The Three IMACS Guiders

The Centerfield Guider (CFG) can be inserted, by either the observer or the TO, to a fixed position at the center of the IMACS field (on the optical axis of the telescope). There is also a second off-axis position which allows the centerfield guider to be used for slit-viewing spectroscopy, described below.

The two outboard guiders -- Principal (PG) and Shack-Hartman (SHG) -- move in an arc at the edge of the f/2 science field, well outside the f/4 field.  Using the instrument rotator's standard setting with north up and east to the left, the PG is to the north and the SHG is to the south.  These guiders have field diameters of 105" and they sweep in ~40-deg arcs (at a field radius of 30 arcmin), each covers about 30 sq arcmin of sky.  These patrol areas are diagramed on the GMAP display that the Telescope Operator (TO) uses to set up these guiders; GMAP shows the location and orientation of the IMACS field, and the PG and SHG patrol fields and locations. 

Guiding and Shack-Hartmann active-optics operations utilizing these three guiders are the responsibility of the TO, as is normal with Magellan operations.

2. Planning IMACS Observations: Capabilities, Characteristics, and Considerations

A.  Imaging with IMACS

The Choice of Mosaic CCD Cameras  

With two Mosaic CCD cameras, both the f/2 and f/4 focal positions are always available for observations.  There are many considerations that fold into the choice of f/2 or f/4, but field and spectral resolution are the two major drivers. The f/4 camera has a field of 236 sq arcmin (0.066 sq deg) with uniformly excellent image quality (down to 0.30"-0.35" FWHM in the best seeing). Its fine sampling of 0.11 arcsec/pix takes advantage of the best seeing for both imaging and spectroscopy. The fully-illuminated, unvignetted, square field is well-suited for tiling an area of sky. (Note: because of the gaps between the chips, the "active area" is slightly less at 230 sq arcmin).

A new feature as of 2015 makes it easier to make imaging observations that produce a mosaic of the field that covers the chip gaps.  A script executed from the f/2 or f/4 camera window will run a series of exposures with offsets that will cover the chip gaps. Minimally, 2 exposures will cover all but a small square, but the choice of 3 exposures will cover the gaps completely, and the 4 or 5 exposure options will insure that the depth covered will be close to uniform across the complete field of either camera. More detail on this can found in the "Observing with IMACS" section that follows this section describing IMACS.Field layout for the f/4 and f/2 channels of IMACS

The f/2 camera accesses the largest possible field, 670 sq arcmin, 656 of which are covered by CCD pixels, or 0.186 sq deg). Because the corners are not illuminated, the f/2 camera is less suitable than the f/4 for tiling an area of sky. Furthermore, at some positions the Principal and SH guider occult the science field and shadow a 1-2 arcminutes arc top and/or bottom. (Since their positions are not fixed with respect to the f/2 camera, this will make flatfielding difficult at the camera edges, top and bottom.  The E2V CCDs are "flat" enough to recover important data, however.) Note that carefully chosen guide stars, making use of the full accessible area, will usually avoid this problem.  Image quality of the f/2 camera is also excellent for its field size: with seeing of 0.30 FWHM the f/2 camera delivers images of 0.45 arcsec FWHM or better. In particular, the central region R < 8 arcmin, can produce images of 0.30" FWHM in the best seeing, however, these images are undersampled by the 0.20 arcsec pixels. The image shows the performance of the camera in the best seeing conditions.

 Map of the image size for the f/2 camera (left) and the psf, described as encircled energy for each channel (right).

The Choice of Filters

IMACS has two identical filter servers, one for each camera. They are "jukebox style" with 15 black-anodized aluminum frames designed to hold 6.5-inch (165 mm) square filters up to 0.47-inch (12 mm) thick. Each camera includes a basic complement of filters --- Bessel B, V, R, & CTIO-I, and a spectroscopic filter (full optical band) that is necessary to equalize optical path length, and a single set of Sloan g,r,i,z filters that can be requested for use in either filter server. Many wide-band blocking filters and several narrow-band filters are available as well, and there are plenty of vacant positions for these or for user-provided filters.

Filter transmission curves are provided at IMACS Filters

Filter specifications are provided at http://instrumentation.obs.carnegiescience.edu/imacs/optical/filters.txt

 Selected Filters for f/2 or f/4

Pixel Binning Options

Because of the fine sampling of IMACS, particularly the f/4 camera, many observations can and should be made using 2x2 rebinning. (The binning options are 1,2,3,4... in any x,y combination.) Rebinning reduces data quantity considerably, of course. Furthermore, some applications, such as centerfield slit-viewing spectroscopy, only illuminate one or two of the eight CCDs, and many setup frames will be done in subraster mode where data volume is greatly reduced. Subraster setups can be created to store only the parts of the array that are illuminated in slit viewing mode. Note that subraster files used for alignment of multislit masks should always be stored in the "full" frame format to be used with the IRAF programs ialign and ifalign.

Data Archiving

 readout of the full 8K x 8K mosaic camera (in 1x1 binning) generates 128 Mbytes 16-bit unsigned integers and is stored as integer FITS files. Disk storage on the two available observer workstations greatly exceeds what can be gathered on an observing run.  Data can be transferred directly to the observer's laptop computer via sftp, stored on DVDs using Llama, or transferred to a personal USB or FireWire disk.

Performance: Throughputs in Imaging Mode

The anti-reflection coatings of the IMACS lenses are typically 1-2% per surface: through the optical train of either camera, the losses are about ~30% -- including the collimator. Losses from absorption in the glasses are small, less than 3%. Filter transmissions over their bandwidths vary widely, from 60-90%. Those with the broadest bandwidth are interference filters with 90% or greater throughput.  The quantum efficiency of the E2V CCDs peaks at approximately 90%, but the response over the full range is very different for the two cameras. as shown in the diagram above. The f/4 Mosaic3 camera uses CCDs with an extended blue and red response at the expense of the visible.  Losses in the telescope -- the three reflective surfaces, primary, secondard, and tertiary, and ADC/Corrector -- total about 55-65%, depending on the freshness of the coatings. Together, the throughput of IMACS in imaging mode (for both f/2 and f/4) is 30-40% for either f/2 or f/4, depending on  the wavelength and filter, not including atmospheric losses.

The following table uses the convention of recording the magnitude at which IMACS direct imaging yields 1 e- per second for the Bessel B, V, R, the Cerro-Tololo I, and the Sloan u, g, r, i, z filters.  Values are including the telescope but not the atmosphere: to correct for the atmosphere, subtract 1.0 from the expected airmass of the observation and multiply by 'extinction per airmass' (in the table) and subtract from the values below.  Photometric Zero Points for IMACS  (June 2018 - new configuration)*

* BVRI magnitudes are on the Vega system and u'g'r'i'z' are AB magnitudes.   

Photometric Calibration: The IMACS Shutters

Many imaging programs require absolute photometric calibration. There are many Landolt Standard Fields that will provide suitable brightness and color standards for most programs (http://www.cfht.hawaii.edu/ObsInfo/Standards/Landolt/). It is convenient to take in-focus images of these fields, however, in typical seeing it may not be possible to avoid saturating the CCD with the brighter standard stars in the Landolt fields. An intentional defocus of 500 microns (the units of the focus readout) will generally produce unsaturated donut-shaped images a few arcseconds in diameter. Generally it is a good idea to keep the standard stars at 13th mag or fainter for f/2, 11th mag or fainter for f/4, even with 1x1 binning.

The IMACS shutters provide help in this regard: they are capable of very short, very accurate exposures. The IMACS shutters are driven by linear motors, each of which have closed-loop feedback. Each shutter has two blades that alternatively open and close the exposure. Because the time trajectories of each blade closely match, even short exposures are guaranteed to have equal and accurate exposure length over the field. In pre-commissioning tests IMACS exposures of 1 sec were shown to be accurate to 2% over the entire field of the CCD mosaic camera.  For certainly, exposures should be 2 seconds or longer.

B.  Spectroscopy with IMACS 

The Choice of Channels -- f/2 or f/4?

For spectroscopy, the choice between f/2 and f/4 is driven by field, spectral dispersion, and (less often) spatial resolution.  The field of the f/2 camera is approximately 2.5 times larger, and its spatial resolution is a factor of two less (in each direction).  The spectral resolution is typically a factor of 3-5 higher in f/4 compared to f/2, although there is overlap --- the 150 l/mm grating produces about the same spectral dispersion per pixel as the 300 l/mm grism, but note that to actually achieve comparable resolution requires a slit approximately a half-as-wide in f/4 mode. 

Programs targeting faint galaxies or stars greatly benefit from high multiplexing.  This favors large field and relatively low spectral dispersions, so for these the f/2 will generally be the choice.  Often multiple tiers of spectra are used to increase the multiplexing factor to 500-1000 objects per field, or higher -- prism mode (see below) has been used to observe 2000-4000 objects in a single exposure.

For a point source or single object extended as much as 15 arcmin, long-slit spectroscopy at f/4 is a natural choice, although multislit spectroscopy with f/4 --- particularly for brighter objects (m < 20) -- where higher spectral dispersion is an option -- can be a powerful use of IMACS.  Unless the spectral range can be narrowed down substantially using a filter, these are likely to be single tier, that is, a single spectra only along the dispersion direction for any given spatial position.

The wavelength coverage of the f/4 extends several hundred angstroms further into the ultraviolet than the f/2. The spectral dispersions for f/4 can be much higher with the f/4 camera, all the way up to the Multi-Object Echelle mode of R ~ 20,000.

The Mosaic2 CCD camera is operated in  "normal" orientation (spectra along the long-axis of the CCD detectors) at the f/2 focus.  "Nod & Shuffle" observations -- spectral dispersion along the short-axis of the CCDs -- are made at f/2 by rotating the grism of choice. This request must be specified on the Setup Form submitted well in advance, and the orientation cannot be changed during the night.  Observations with the prism dispersions, LDP and UDP have been done in N&S exclusively.  Because gratings cannot be rotated like the grisms, the CCDs in the f/4 Mosaic1 CCD camera are oriented to make N&S observations.  Although N&S observations are in fact much less common,  the capability for N&S observations has the negative consequence of running spectra across 4 CCDs and 3 chip gaps, instead of 2 and 1. 

Multislit Spectroscopy, Long-slit Spectroscopy, or Centerfield Slit-viewing Spectroscopy?

IMACS has been designed principally for wide-field multislit spectroscopy. This mode is intended for faint objects where multiplexing is a necessity, and/or for objects of sufficiently high spatial density (> 50 per sq deg) that program efficiency can be substantially increased. Multislit spectroscopy requires considerable advanced preparation and a production cost $200 per IMACS slitmask. Multislit spectroscopy is also a feature of MOE, the multi-object echelle, as described below in the section below, "The IMACS Accessories."  MOE is capable of delivering a 9-order spectrum at R~20,000 for ~7-8 objects, or a single order for ~50 objects.

Long-slit spectroscopy of up to 27 arcmin can be carried out with IMACS using one of the multislit mask blanks that has been cut by laser for this purpose. The long slit needs to be made up of ~10 sections to maintain the structural integrity of the mask, which means there are ~1 arcsecond bridges about every 2 arcminutes.  Observers will be charged for any first-time, cut-to-spec long slit, but several long-slit slitmasks with widths ranging from 0.7" to 2.0" are kept on hand.

Traditional slit-viewing spectroscopy, typically for single objects, is also supported by IMACS. This is accomplished by sending the IMACS Centerfield Guider to a location 2 arcmin off the optical axis.  In  Slit-View mode (as opposed to Shack-Hartmann mode of the Centerfield Guider) an arcminute-long multi-width slit, and a hole of 7" diameter,, are viewed by the TV guider. (A field mask which occults (somewhat imperfectly) the remaining IMACS field will be deployed automatically.)  In "slit-viewing" spectroscopic mode a grating or grism can be left in the beam as the observer proceeds from object to object, without using one of the Mosaic CCD cameras for acquisition, and without reconfiguring IMACS (except for optional grating or grism changes).  The slit-viewing mode is particularly useful for flux or radial velocity calibrations of standard stars. Details of operating in this mode are described below.

Performance: System Throughput

The following plots show the throughput of the f/2 and f/4 channels, including the losses from the Baade telescope optics.  The throughput is defined as the fraction of photons striking the primary mirror that are detected -- atmospheric losses are not included. Throughput degrades following the recoating or cleaning of the primary and secondary mirrors at a rate of ~1% per month, as described in the figure label. 

The throughput for spectroscopic mode follows the discussion above "Performance: The Throughput in Imaging Mode" except that spectral dispersers result in a further loss of about 20% for the f/2 grisms and 30% for the f/4 gratings.  Throughput in spectroscopic mode therefore ranges between 25-30% in f/2 and 15-25% in f/4.  Individual measured performance is excluding atmospheric losses, are tabulated below.

Values plotted in the following figure were made in August 2016 and therefore represent the middle of a coating/cleaning cycle. After each coating/cleaning cycle the system is subject to ~1% loss per month, with a ~13% loss of sensitivity by the end of the cycle. Since the present values were measured in the middle of a coating/cleaning cycle, expected throughput is  ~6% higher at the beginning of the cycle, and 6% lower at the end.

Throughput of the f/2 channel, including telescope, with its 5 available grisms. *

      *Throughput measurement shown in this plot  correspond to the new configuration - after the dewar swap (Aug. 2018). 

  Throughput of the f/4 channel, including telescope with its 8 available gratings. *

                               *Throughput measurement shown in this plot correspond to the new configuration, after the dewar swap (June 2018).  

Scattered Light and Ghost Images

In multislit exposures with many hundreds of slits that add up to a significant sky area), scattering of 1-2% of the total light passing through the slits can be seen as a low-level diffuse background over the field. With respect to ghost images from bright objects, these are rare in IMACS, thanks to aggressive baffling inside the collimator and camera barrels.  When seen, they contain ~2% of the light in an object (for spectroscopy, a bright zero-order spectrum) spread over a 40 arcsec diameter disk with an 80 arcsec halo. More information on scattering and ghosting can be found in S6.9 and S6.10, respectively, in http://www.jstor.org/stable/pdf/10.1086/658908.pdf.

Fringing in the Near-IR

Fringing is moderate in the Mosaic2 and Mosaic3 CCDs.  Mosaic2 shows a fringe amplitude of <1% peak-to-peak from Lambda < 7500 Å, grows to 2-3% in the 8800 Å  < Lambda < 8300 Å region, and crests at ~5% for Lambda > 9000 Å.   Mosaic3 shows the same behavior, but the amplitude reaches ~6% at Lambda ~ 9200 Å and is growing, perhaps reaching 10% at 9500 Å (consistent with its "deep depletion" design).  Because these are modest amounts, as far as fringing goes, in practice the fringe signal is well removed with spectroscopic flat fields, despite the fact the flat field spectrum is not likely to be a good match to the spectrum of object + sky.

The Choice of Reflection Gratings and Grisms

IMACS carries an assortment of 3 reflection gratings and 2 grisms for a night's observing, chosen from a present complement of 6 gratings and 4 grisms https://www.lco.cl/?epkb_post_type_1=imacs-specs These are mounted on the disperser server, a large wheel that also carries the reflecting mirror for f/4 imaging. The IMACS Echelle (MOE) and tunable filter (MMTF), when requested, are also carried by the disperser server.

A filter mount is also available for the 'disperser server,' a very large wheel that moves grisms into the collimated beam for use in the f/2 channel.  The fixture takes a standard IMACS 6.5-inch square filter and mounts in place of one of the dispersers or the f/4 mirror. The beam is the preferred location for relatively narrow-band filters (Delta Lambda < 100 Å FWHM). It is also possible to mount a filter in front of a grism, however, this is by special arrangement with the Magellan staff.  The MMTF -- a six-inch aperture tunable etalon, is also inserted into the collimated beam, but its use precludes the use of dispersing elements in series.  MMTF is available for general use, in collaboration with PI Sylvain Veilleux (see IMACS Accessories, below).

Observers specify their choice of these units before the beginning of observations, that is, they will not be swapped during a night's observing. Grating exchanges can usually be made without removing the grating-tilt-mechanisms from the disperser wheel, which has made it easier to support grating changes between observing nights.

Prism Mode: LDP & UDP

The f/2 channel can be used to in a highly multiplexed, high-throughput, low-spectral-resolution mode using prisms instead of gratings or grisms as the disperser element.  This capability, pioneered by Scott Burles for the PRIMUS galaxy survey, passes the zero-deviation light dispersed only by the wedge of the prism -- no rulings are involved.  The R ~ 30 dispersion produces very short spectra, of course, with the result that thousands of objects can be targeted, as seen below  in the mask design by Shannon Patel and Dan Kelson.  A disadvantage is the strong dependence of the dispersion on wavelength, shown in the graph below.  Burles mitigated this affect by building a prism of several layers of glasses with different indices or refraction, and the idea was taken further in a design by Steve Shectman for a uniform dispersion prism, which -- as the graph shows -- levels of for the reddest wavelengths.  This is particularly advantageous because the night sky in the near-IR is lit by bright OH sky emission, as shown in the sample spectra from both prisms. The UDP's feature of a modest but near-constant dispersion much improves sky subtraction as well photometric calibration compared to the LDP.  The low dispersion of prism mode requires nod-and-shuffle observations to achieve acceptable sky subtraction.

Mask Design by D. Kelson for the CSI survey using the UDP (left).  Comparison of the format and plot of dispersion for LDP and UDP (right).

Operating Features of the Disperser ServerWithin the constraints listed above, once the gratings or grisms are chosen are chosen, and the disperser server is loaded, the observer's control of these devices is rapid and efficient. Exchanging gratings or grisms takes 30-60 seconds. The tilts of the gratings are encoded: the central wavelength and wavelength intervals of each grating as a function of encoder position are easily calculated (Link). Setting a new grating tilt usually takes about 30 seconds. (All IMACS mechanical functions can be exercised simultaneously.) Grating tilts positions are repeatable to +/- 2 pixels, but this refers to the accuracy of setting the tilt once a grating tilt mechanism is inserted into the beam. Once grating tilts are established, they will remain fixed -- it is not necessary to zero the tilt in order to install or retract a grating. (Homing the grating tilt by initialiizating the device is, however, recommended if there is any indication that the tilt is not correct.) The caveat to this is that the mechanisms themselves are not inserted with perfect repeatability; shifts of 5 pixels are common even in proper operation, and may be considerably larger if there is a problem with the latching of the mechanism.  If a specific pixel location is required for a particular wavelength, this can only be achieved after the grating is installed, by adjusting the tilt.

The wavelength coverage and central wavelength of a grism is fixed. Because the grisms are transmissive dispersers instead of reflective the angle of 'latch up' is much less important. Also, the grism mounts have no moving parts and are lighter than the grating mounts.  Unsurprisingly, then, they usually re-insert with good repeatability, typically 1-2 pixels in the dispersion direction.

Standard 'Stare' Spectroscopy or Nod-and-Shuffle?The default orientation for f/2 slit-spectroscopy, whether long slit or multislit, is to have the dispersion along the long axis of the CCDs -- in this way there is only one gap in the spectral direction.   Except for the most demanding of applications, the 'staring mode' is the preferred mode of spectroscopy with the f/2 camera.  A highly-effective sky subtraction algorithm designed by Dan Kelson, included in the COSMOS reduction software, allows this more traditional mode of spectroscopy to be competitive with the Nod-and-Shuffle technique -- at least for IMACS.

Nod-and-shuffle (N&S) spectroscopy is a highly-regarded method of taking mutlislit data, especially for very faint objects at wavelengths beyond 7500 Å. Chopping the spectra between opposite ends of their slits, at the same time shuffling the charge back and forth on the CCD and nodding the telescope in the opposing direction, results in excellent sky subtraction, even at relatively low spectral dispersion. The use of short slits contributes to a larger mutliplexing factor, and the excellent sky subtraction makes it feasible to study of objects whose brightness is a very small compared to the night sky. In order to observe in this mode, the spectral dispersion direction  must be along the short-axes of the CCD detectors.  The charge is clocked in the spatial, rather than the dispersion, direction. N&S is the default orientation of the f/4 camera, while N&S is accomplished at f/2 by rotating the grism by 90-deg.  One caution is that, because in comparison with conventional spectroscopy the exposure time is cut in half for each part of the exposed CCD, it is important to make sure that the sky level in electrons is well above the read noise squared.   The IMACS Cookbooks contains a full description of making N&S observations with IMACS.

Guiding is done differently with N&S.  In order to accomodate the nods from one telescope position to another, the target (a software cross) is moved back and forth on the IMACS Principal Guider image, and the telescope is moved the same amount.  The SH guider itself does not move.  The Shack-Hartmann guider is turned off every-other-exposure because it would need to move to keep the SH star centered in its (physical) aperture.  Long nods > 50 arcsec will not work because the star in the Principal Guide camera must stay on its field of 105" diameter.

A detailed description of nod-and-shuffle spectroscopy can be found in Glazebrook and Bland-Hawthorn (2001, PASP, 113, 197).

Recommendation:Although Nod-and-Shuffle has been shown to be extremely effective over the small field of some spectrographs, for example, the Gemini GMOS, the very wide field has made the critical alignment and precision of subsequent shifts challenging.  The PI recommends standard mode over N&S for most observing programs. An exception is LDP and UDP observations, for which the benefits of accurate sky subtraction greatly outweigh loss of flat-fielding in the N&S method.  --- AD

Stability and Repeatability of IMACS Slitmasks

The principal use of IMACS is multislit spectroscopy.  The multislit masks for IMACS are 26-inch diameter shallow-curved dishes of 0.010-inch-thick stainless steel into which slits are cut by a laser.  A software package maskgen, written and maintained by Ken Clardy (clardy@obs.carnegiescidnce.edu) is provided to users for designing these masks. Programs and necessary star catalogs are available at each of the Magellan Partner institutions, and there is a web-based interface to the star catalog that facilitates the use of maskgen by observers outside the Magellan Consortium.  The programs are easily ported to a variety of platforms.

The actual cutting of slit masks, done with a laser milling machine at Las Campanas, requires about an hour of machine time, including setup, for a typical mask, one with ~300 slits.  IMACS holds 6 of these slitmasks, which has proven adequate for the great majority of multislit programs, but some observing programs might require swapping slit masks once during the night.  Although the operation to change the masks complement is a relatively simple operation that costs only about 30 minutes of observing time, the personnel are generally not available to do so.  Special requests will be considered; the request should be made well in advance, on the Observing Setup form.  

A 'multi-box' mask cut to match to match the positions of ~500 stars in a a high-quality astrometric field produced typical decentering of 0.15 arcsec from the target box positions. Residuals reached as large as 0.4 arcsec in a few areas of the periphery of the field, most too close to the mask edge to be useful for spectroscopic observations (because about half of the spectrum would fall outside the detector.  See http://www.jstor.org/stable/pdf/10.1086/658908.pdf, S6.3, for more details.

Observers are required to upload the SMF files that are needed to cut the masks (see https://www.lco.cl/?epkb_post_type_1=how-to-submit-imacs-mask-files) well before the date of observation  -- read the policy at: Link

Mask insertion is often repeatable at the 1-2 level. However, there are still occasional cases of a differences as large as 3-5 pixels.  Very good repeatability is common when a mask is removed and reinserted at a given rotation of IMACS, and the larger differences occur when the very different orientations are compared, for example, when a mask imaged in the afternoon is compared to an on-sky image at a much different rotator angle.  For the f/2 camera, good repeatability means that the coordinates of slits or apertures cut into the mask become fixed with respect to CCD pixels.  However, for the f/4 camera an additional complication is that the reflecting mirror does not repeat its position to within a few pixels, so the position of slits or apertures moves on the detector with each mirror insertion. 

To take account of the imperfect repeatability of the slit masks and mirrors, the alignment procedure includes aIRAF program ifalign that adjusts the target positions to reflect the final position of the mask and/or mirror at the time of observation.  The full alignment procedure is discussed below.

Flexure Between the Guider Cameras and the Focal Surface (slit)

A design goal of IMACS was to provide negligible flexure between the guide cameras and the focal plane.  This flexure has been measured to be less than 0.5 arcsec for all orientations (rotator angle), which means that -- once an alignment is made -- it will hold indefinitely through all sky positions.  An earlier problem with rotation guiding (using both the PG and SHG  to guide the rotator in the same way the telescope guides in azimuth and altitude) has been resolved.  It is now well demonstrated that multislit setups hold for at least several hours of tracking -- at least. Most observers still recheck the alignment on an hourly basis, but performance since 2012 indicates that this is unnecessary.  The reliability of rotator guiding also means that long direct-imaging exposures (for example, for narrow-band imaging) will also show no discernible drifts.  Note that the issue here is keeping objects square on the slits, or otherwise aligned.  Possible shifts between the focal plane and the Mosaic CCD cameras -- through flexure in the spectrograph -- is another matter, relevant mainly for data reduction.  As discussed next, it is also negligibly small for standard observations.

Recommendation:When alignment issues are critical, arrange your program to avoid tracking within a few degrees of the zenith, or re-align the slitmask afterwards. --- AD

Flexure Between the Focal Surface (slit) and the Mosaic CCD Cameras

All spectrographs exhibit flexure between the focal plane and the detector, but the flexure in IMACS is exceptionally low for such a large instrument. For a full rotation of the spectrograph, images move 2-4 pixels for the f/4 camera and 1-2 pixel for the f/2 camera. Most of this is repeatable (elastic), so it is taken out with one of two sets of lookup tables (direct, spectroscopy) that are used to position a piezo-driven x.y stage in the Mosaic CCD Camera.  After correction, the complete range of residual flexure for a full rotation of the instrument is +/- 0.5 pixels for the f/4 camera and +/-0.25 pixels for the f/2 camera, equivalent to +/-0.05 arcsec and +/- 0.3 A (typical) for both cameras. (Most observations use only a sector of the possible rotation, so these numbers can be considerably smaller.)  Both Mosaic2 and Mosaic3 include flexure stages that correct most of this expected flexure. The bottom line is that flexure from the focal plane to the Mosaic CCD camera is remarkably small and should not be an issue for standard programs. For direct imaging, in particular, errors in guiding caused by the mechanical separation (load path) of the IMACS components (guiders, optics, CCD camera, etc.) should be small even compared to the best seeing.  

Recommendation: Before taking a slit or mask image for measuring positions to align sources, make sure that Flexure control is ON and properly applied. This means that there are no flexure control error messages in the IMACS message GUI. Another way to check this, is to confirm that the "desired" flexure values reported in the IMACS terminal window match the values reported in the HardHat window.  -- KB

Observations with the IFU, however, have shown a stead drift of about 0.2 arcsec per hour.  Of course, the IFU does not experience "slit losses," so this is a relatively benign effect, but some observers would prefer to hold their exposures to the same pixel patterns for several hours.  The problem is irreparable because the unit is much heavier than a slit mask (which hold position very well -- see the next section), so the likely explanation for this is that the IFU itself is shifting slightly as it rotates with respect to gravity.

C. The IMACS "Accessories" – IFU, MOE, MMTF, GISMO

The following accessories will not normally be installed in IMACS, unless requested in the observer's setup form:

Durham-IMACS IFU

The IFU displaces three slots in the slitmask cassette, so observational programs can combine both mutlislit and IFU observations. During observations the IFU is inserted like any other slitmask. At the present time, installation of the IFU is a daytime changeover. A detailed description of the IFU and its use can be found here.

MOE – Multi-Object Echelle

MOE is an accessory that transforms IMACS into a echelle spectrograph with a resolution of ~0.3 Å (R ~ 20,000) that includes a significant multiplexing capability.  MOE on IMACS permits crossed-dispersed spectra to be obtained over the 15x15 arc minute field of the f/4 camera, accomplished by placing unit that includes a grating and cross-dispersing prism in the IMACS grating wheel.

The optical layout of MOE (top) and a schematic representation of the spectral format (bottom). 

Because of the large number of detector pixels in the 8K x 8K array, MOE can in principle record echelle spectra covering the full optical wavelength range for up to 15 targets at a time.  Each object can be chosen from anywhere within a 1' x 15' (spatial direction x dispersion direction) area of the sky. In practice, targets are not ideally spaced, and at least 3 alignment holes are required, so typically 7-8 objects can be observed with full spectral coverage. By limiting the wavelength range with a bandpass filter (so each spectrum contains fewer echelle orders), a larger number of targets can be observed simultaneously.  For example, A. Williams used a 4800--7800 Å blocking filter to obtain 4-order spectra of 20 red giant stars in the Carina dwarf spheroidal galaxy. Masks designed by J. Simon using a z-band filter to isolate the reddest MOE order (containing the Ca triplet lines) contained up to ~80 stars. By using MOE in combination with the GISMO image slicer (see below), 16 targets -- one for each 0.5' x 2' rectangle selected from the central 4' of the IMACS field -- can be observed at high spectral resolution in denser fields or for less extended objects.

For a 0.5 arcsec-wide slit MOE provides a resolving power of R=20,800 in the center of the field.  Each echelle spectrum is composed of 10 orders (9 full + 2 half) from roughly 3400 Å < Lambda < 9500 Å.  The wavelength coverage is complete near the center of the field up to a wavelengths below 8200 Å, while the length of the reddest order may slightly exceed the size of the CCD mosaic. Central wavelengths of the MOE orders are listed below:

                                   MOE wavelengths by spectral order

Table notes:

Central wavelength -- the calculated wavelength at the array center for each order.  These wavelengths are slightly different than those in the spectral format plot above because of the revised blaze angle, but the difference should be negligible for almost all purposes.

Wavelength of peak efficiency -- the wavelength of observed peak efficiency of the MOE grating, measured in the laboratory, for a=59.5, b=14.5 (see discussion below).  Actual peak flux for a MOE spectrum depends on additional factors.

Based on the grating peak efficiency wavelengths, the manufacturer estimated the actual blaze angle to be 37.96 degrees rather than the requested value of 36.5 degrees.

The spectral resolving power of MOE depends on slit width, location of the slit in the field, and position within a particular order.  For slits not near the center of the field in the dispersion direction, the angle of incidence onto the grating, a, is different than at field center.  This anamorphic magnification effect results in a different projected slit width onto the CCD for off-center slits and effectively changes the spectral resolving power (see plot of projected slit width).  For slits at large positive Y positions on the mask (large a) a 0.5 arc sec slit may be under-sampled; for slits at such positions, one may choose a wider slit (e.g., 0.6 arc sec).  Because the angle of dispersion off the grating, b, changes with position along the order (i.e., with wavelength) the spectral dispersion changes along each order as well.  This serves to increase the resolving power (l/Dl) for larger b, although the slit projection also changes slightly with b, counteracting the resolving power change by a small amount (a few percent).  Therefore, while the spectral resolution in the middle of the spectrum (for a centrally located slit) is R=20,800, the resolution over the rest of the spectrum varies from 19,020 to 22,550, going from the smallest to largest angle of dispersion, respectively (equivalently, smaller to larger wavelengths within an order).  These effects also change the blaze angle. 

Projected Slit Width versus Field Position

An observation of the standard star EG21 obtained in 2012 has been used to determine the throughput of MOE.  Including the telescope and spectrograph, the measured efficiency is 7.7% at 6400 Å.  This speed translates to a S/N=50 per pixel for a V=17.5 magnitude star in 3 hours under good observing conditions. MOE's multi-object capability can more than compensate for this factor-of-two lower speed compared to single-object spectrographs of comparable dispersion.

A detailed description of MOE and its use can be found in Sutin & McWilliam (2003, SPIE 4841, 1357). 

MMTF – Maryland-Magellan Tunable Filter

The MMTF -- Principal Investigator, Sylvain Veilleux (U. of Maryland) -- is a Fabry-Perot that serves as a narrow-band filter, tunable in both central wavelength and transmission bandpass.  The etalon operates in low orders (close plate settings) to provide a large "monochromatic spot" (region of constant wavelength) across the field of view.  Its tunability, sensitivity, narrow bandpass, and wide field make this a unique capability applicable to a wide range of programs. 

The MMTF is mounted in the IMACS diperser server (see above) and is used with the f/2 camera.  The basic characteristics of the etalon are: (1) a monochromatic spot diameter ranging from 7 to 11 acrminutes (larger for wider bandwidths); (2) a bandpass tunable from 6-25 Å; and (3) wavelength coverage 5000 Å< Lambda < 9200 Å --- the actual wavelengths depend on blocking filters.  The measured 6σ sensitivity to a emission line pont source at 6600 Å  is 3 x 10^-17 ergs s^-1 cm^-2 for a 1 hour exposure with a 25 Å bandpass.  This sensitivity can be improved by fequency switching the etalon in synchronization with charge  shuffling in the  CCDs of the mosaic array.  This observing mode averages over temporal variations of the atmosphere and the instrument. The MMTF achieved first light in 2006 and was comissioned during the 2007A observing semester.  Observers can propose to use the MMTF in collaboration with Sylvain Veilleux and the MMTF team.  The MMTF website is http://www.astro.umd.edu/~veilleux/mmtf/. 

MMTF images of NGC 1365 in the stellar continuum, emission, and emission-line ratios.

GISMO – Image-Slicing Mutlislit Option for IMACS 

GISMO is an image-slicing reformatter that installs into the slitmask area of IMACS, at the focal surface of the Baade telescope.  Its purpose is to slice up an approximately 4' x 4' area at the center of the IMACS f/4 field into 16 sectors and reimage them back in the focal plane at the same scale and f-ratio, but covering the entire f/4 field. GISMOS's unique capability is to allow multislit spectroscopy of densely packed fields -- more than 5 targets per sq arcmin -- whose spectra would necessarily interfere in the central area, but will not once the slices have been separated.  This results in a multiplexing gain of a factor of 5-8. This makes GISMO idea for such fields as the centers of galaxy clusters, including, for example, faint arcs in distant clusters, MW globular clusters, LG dwarf galaxies, and globulars and planetary nebulae in nearby galaxies.  GISMO can also be used in a combination long-slit and multislit mode for mapping extended objects, and in combination with MOE to increase the resolution to R ~ 20,000.  Planning observations with GISMO, and more discussion of setup and operations are found in the Observing Cookbooks.

GISMO format on a rich galaxy cluster field, including sky image of one of the sixteen reimaged slices with slits. 

3. Preparing for IMACS Observations

A. Calculating Observing Times

The sensitivities for direct imaging in the standard IMACS filters are given above. An exposure time calculator for imaging observations can be found here. An exposure time calculator for spectroscopic observations is under development as of March, 2016.

B. Designing Multislit Masks 

Multislit masks are cut by a computer-numerically-controlled laser-milling machine that resides in the Astronomer Support Building at Las Campanas Observatory. The Technical Support Group at Magellan is responsible for producing the multislit masks and installing them in IMACS. Generally, up to 6 masks will be loaded into IMACS at the beginning of a night's observing.  A second box of up to 6 additional masks placed can be placed on the Nasmyth platform (top of the stairs) for a swap during the night.  This must be requested in advance. A complete software package maskgen is now available for designing multislit masks. The software has been successfully ported to many different platforms, and help is available when problems are encountered, as explained in the link.  

The observer needs to provide in maskgen a file containing a list of objects and setup stars with sub-arcsecond astrometric accuracy. Relative positional accuracy is more important than absolute, which will be tweaked at the time of observation, although a field position good to a few arcseconds will reduce setup time for a multislit-mask to 5-10 minutes. The list of potential objects can be definitive or can be an overpopulated list from which the software will choose to maximize the number of targets that do not conflict in the spatial or dispersion direction.  Priority can be specified in the list, for example, by object magnitude, but if no priority is given, maskgen will simply prioritize from the top of the list down, eliminating conflicts below as it goes.  Several lists can be entered for the program to use: they will be stacked in order of entry to make one long list.

Priorities can be set in the list such that certain objects are guaranteed to be included in the mask -- these are called "must have" objects. Observers with their own slit mask design can simply enter a list of objects all tagged with the "must have" value.  In either use of the "must have" feature -- for a select subset or for a full arrangement -- observers should recognize that no conflict resolution will be done, that is, the user must eliminate or accept conflicts in this approach.  One way to do this is to run the program with a priority lower than the "must have" value to see if any of the objects are tossed out because of conflicts.  It is also possible to view the conflicts on the display of the mask design the program produces.

Recommendation: The observing catalog input to maskgen must also contain a list of stars that will be used for aligning the mask.  Good results have been obatined by choosing stars between 17th and 19th R-band magnitude. The number to be selected is entered in the maskgen GUI: a set of 10-15 alignment stars covering the full field has been shown to give excellent results, but the alignment procedure will work well with as few as 6.  Note, however, that some stars are lost in "chip gaps" and some turn out to be galaxies, so preparing a mask with only 6 alignment stars is risky. If you have access to photometry and astrometry for your field, it is preferable to choose your own guide stars and alignment stars, rather than relying on the USNO catalog, which includes a significant fraction of galaxies and stars with significant proper motions. --- AD

It is crucial that observers, when creating slitmasks, verify that there are suitable stars for the Principal Guider and the Shack-Hartmann Guider.  Maskgen accomplishes this through the skywin option in contained within the intgui program  To successfully complete the  mask design process requires the SH guider field must contain at least one star 15th magnitude or brighter.  The Principal guide star can be considerably fainter, but to be insensitive to poor seeing, brighter than 16th magnitude is recommended. It is crucial to check the guide stars to make sure they are not galaxies or stars with high proper motion -- the skywin feature of intguilinks to the STScI Digital Sky Survey to bring up a sky image that can rule out these possibilities (ctrl-right-click with the cursor on the star).  For most field centers and field rotations there will not be a problem finding adequate stars; however, it will be sometimes be necessary to either offset the field center or rotate the field around the original center.  Some observers choose to capture the image of skywin to take to the telescope.  This picture will directly match the GMAP monitor display that the Telescope Operator will use to set the telescope.

In theory, a perfectly accurate slit mask requires not only perfect object positions but also knowledge of the field position on the sky at the specific time of observation (to account for atmospheric dispersion), as well as ambient temperature during observation compared to the temperature at which the slitmask was machined.  Generally, for observations are at an airmass of less than 1.5 and slits 0.7 arcsec or wider, these errors will be acceptably small.  The atmospheric dispersion correction makes it unnecessary to align the slits along the paragalactic angle, but the ADC does not compensate for the overall scale change as a field moves from the its median transit to the horizon.  Multislit observations at more than 2.0 airmasses  with the 1/2-degree field of the f/2 camera will show greater errors because of this scale change -- if possible, such high airmass observations should be avoided.  

Maskgen produces ASCII files fieldname.SMF that are used to generate the machine code 'mask cut' file for the laser milling machine (at Magellan).  These are files that the observer uploads to the website slit mask form.  Note that the mask files must be received six weeks prior to the observing run to guarantee that the masks will be cut.  The SMF file contains a line which begins !oc that is the proper format for an observing catalog file that is passed to the Telescope Operator, which ensures that the sky positions of object and guide stars, and the rotator angle, are communicated properly.

Maskgen produces a file fieldname.obw that contains an abridged version of the catalog file for making the mask.  The obw file is used to keep track of the objects that have been targeted and can be used to to avoid duplication of targets in subsequent masks.  The list can be hand-edited to allow duplication.

As descrobed om the "Observing with IMACS" section, the proper way to acquire multislit fields is to use the accurate metrology of the IMACS guiders to correct small errors in telescope pointing.  This process, known as 'centering'  in the GMAP program, sends commands to the telescope that bring the stars to the guiders rather than the other way around.  Verify that the TO is using this procedure, which is unique to IMACS and not the common operating mode for other Magellan instruments.

Full documentation of the many ways to use the maskgen software can be found at maskgen.

C. Preparing for Multislit Observations with MOE

The maskgen software package maintained by Carnegie (http://code.obs.carnegiescience.edu/maskgen) contains options for making multi-slit masks for MOE.  Conventional IMACS masks will result in some spectral overlap, as will slits longer than 5 arcsec.  It is therefore important to use the options specifically for MOE.  When the disperser is set to MOE in maskgen, the filter wavelength range parameters behave differently than with ordinary gratings or grisms: the lower and upper wavelength limits are used to select the order rather than defining the full wavelength coverage.  For example, if one wanted to observe only order 5 (covering ~8300-9800 Å), the lower limit of the filter range must be larger than 8300 Å, and the upper limit must be less than 9800 Å (e.g., [9000, 9300]).  If the lower limit in filter range is set to a smaller value (e.g., 8000 Å) then order 6 will be included as well.  A “standard” MOE slitmask containing slits of different widths at a variety of positions in the field is available for observations of single targets (e.g., standard stars).  This mask can be found in the LCO slitmask database under the name moestd2 and barcode 3181. 

Judicious choice of the slit position angle set in maskgen can significantly improve the packing efficiency of spectra.  The gaps between CCDs in the IMACS f/4 mosaic are approximately 50 pixels.  These gaps should be considered when making the mask, as it is undesirable to have a spectral order or critical spectral line fall in the gap in spectroscopic mode.  It is also undesirable for an alignment hole image to fall in a chip gap in imaging mode.  However, the current version of maskgen does not calculate gap positions.  Note also that it is possible in principle to locate alignment holes outside the f/4 field of view and use the f/2 camera to align.  This technique has not been attempted, but it would make alignment stars in the f/4 area covered by MOE unnecessary, increasing the number of feasible targets per exposure.  This feature is also not presently enabled in maskgen.

Once masks are designed, observing with MOE is essentially the same as standard IMACS f/4 observing for multi-slit spectroscopy, with an identical mask alignment procedure.  A ThAr calibration lamp is mounted at the top of the telescope and can illuminate the flat field screen that deploys in front of the secondary mirror.  This lamp is controlled with the moelamps GUI rather than the GUI used to operate the other arc and quartz lamps.  Some care may be needed to obtain a good spectrograph focus with MOE installed.  If the IMACS collimator is not perfectly aligned (MOE observations in 2015 demonstrated that it is not producing a fully collimated beam), the anamorphic projection effect of the grating results in different camera focus positions for spectral lines along and perpendicular to the dispersion direction (i.e., astigmatism occurs in the spectra).  In such an astigmatic condition -- since only the camera focus can be readily changed -- choosing the best focus in the dispersion direction is recommended, as this will give the highest spectral resolving power.  However, the orders will be slightly wider than for a perfect collimator and camera focus.  If the spatial profiles at the best spectral focus are too asymmetric, a compromise focus value may be preferred.  Note that such astigmatism would not be noticeable in imaging mode or in low-dispersion spectroscopy, and a camera focus offset can only counteract the error in the collimator focus equally in both directions.

Reduction of MOE spectra has been successfully performed using standard IRAF routines. However, depending on the number of targets and number of orders per target this route can be laborious because of the large number of CCDs spanned by each spectrum.  Modifications have been made to the COSMOS package in order to handle MOE data more efficiently, but reductions in COSMOS are likely to be straightforward only for experienced COSMOS users.

For observers interested in additional details, we recommend keeping targets close to the center of the MOE field as much as possible to avoid vignetting effects and loss of spectral coverage.  Under all operating conditions the IMACS collimator beam over-fills the MOE grating.  At the field center, approximately 83% of the collimator beam is intercepted, for a vignetting loss of 17%.  The vignetting loss depends on the angle of incidence, a, of the collimated beam onto the grating: a larger angle 'a' over-fills the grating more.  For a slit at the top of the MOE field the vignetting loss is 29.5%, whereas at the bottom of the field of view the vignetting loss is 8.5%. These vignetting losses are in addition to any loss of spectral coverage resulting from parts of spectral orders falling off of the CCD array.

As described above, MOE installs in the IMACS disperser-server in place of one of the three grating tilt mechanisms.  Observers are cautioned that the MOE grating plus cross-disperser is significantly heavier than an ordinary IMACS grating, and in a few recent (2015) instances its increased weight may have led to incorrect positioning of the grating wheel.  It is therefore not recommended to leave MOE in the grating wheel on nights when its use is not expected.

For users interested in the properties of the grating, the ruling density is 245 lines/mm.  Laboratory measurements for a configuration similar to MOE found the grating efficiency to be ~45% for most optical orders, ranging from 40% in blue orders to 52% for the reddest order (order 5: ~52%; order 10: 45%; order 15: 40%).

A detailed description of the MOE and its use can be found in Sutin & McWilliam (2003, SPIE 4841, 1357).http://ads2/astro.puc.cl/cgi-bin/nph-bib_query?bibcode=2003SPIE.4841.1357S&db_key=AST&high=3f7a6c574f22263.

4. Observing with IMACS

A. Some Things You Really Meed to Know.

IMACS is permanently mounted on Magellan Baade.  Generally, it is always "powered up" and the detectors are kept at operating temperature by Cryotiger pumps. Under normal conditions there are no systems to power on. However, before the observer arrives, the on-duty Instrument Specialist will usually have started the software, initialized the mechanical devices, and taken test frames with the Mosaic Camera.

Before the observer has come to the telescope, the on-duty Instrument Specialist will have already received the observer's choices for (1) gratings or grisms, (2) special filters, and (3) an IMACS accessory unit (see above). If multislit masks are part of the program for the upcoming night, the observer will have informed the Instrument Specialist which of the already prepared masks will be used; the masks are usually installed by 3:00 p.m., and often before lunch. The Instrument Specialist will edit the setup files that identify the slitmasks, filters, gratings, and/or grisms that will appear in the pull-down menu of the IMACS-GUI --- these files cannot be edited by the observer. 

For direct imaging programs, observers will usually take a series of flat field exposures with the appropriate filters, using the pupil-mask screen (see description immediately below) and illuminated by lamps mounted on the telescope. Exposures will range from a few seconds to several minutes for blue filters, f/4 1x1 binning -- even with the relatively "hot" quartz-halogen lamp.

"Dome flats" using the pupil screen will generally be sufficient for most imaging programs, but twilight sky flats may be taken to complement these.  The dome flats should well match the illumination of the night sky through the telescope, but the twilight sky flats will by construction be a better match.  If the program includes enough images in one or more bands that can be summed and median filtered, this produces a night sky flat that is essentially perfect, but this is not necessary for most programs. 

Recommendation:It is sufficient to take a few flat fields for each filter, with count levels that are comparable to the sky level of the observations.  Except for rare programs (for example, surface-brightness-fluctuation measurements), nothing is accomplished by taking flat fields with an order-of-magnitude or more in total counts compared to the actual observations.  Also, the overscan region on the frames is, for almost all purposes, sufficient for bias-level subtraction -- bias frames are generally unnecessary for IMACS observations in general. Likewise, dark frames -- which accumulate typically a few counts per hour -- are not necessary for high-quality IMACS imaging.  --- AD

For spectroscopic programs, the observer will usually want to take images of the slitmasks (multislit or longslit) at the same binning as is intended for the nighttime observations.  This should be 1x1 for the f/2 camera but 2x2 for the f/4 camera: there is no point in using the full resolution of the f/4 camera for alignment purposes, because errors in the process (for example, telescope tracking and seeing during exposures) are considerably bigger than errors associated with measuring the positions of alignment stars with ~0.2" pixels.  The mask exposures and alignment pictures should all be taken through a Bessel-R filter or Sloan 'r' filter. These setup mask images are usually exposures of 1-2 seconds taken with the lower-light-level quartz lamp controlled by a power supply at the base of the telescope -- ask the Instrument Specialist. These are used in the first step of the alignment procedure, to produce an xxx.align file with the program icbox. This step is also a double-check of the match-up of masks and locations (in software) in the cassette.

Recommendation: As has just been emphasized, alignment images should be 2x2 for the f/4 camera.  In the same vein, when the seeing is FWHM > 1.2"  (f/2) or FWHM > 0.6" (f/4), little or nothing is gained by taking full 1x1 binned frames for the actual observations -- the data are already well oversampled.  Using 2x2 binning reduces readout time, but the main advantage is minimizes read noise in proportion to signal, a considerable advantage at higher dispersions. In this regard, observers should choose readout speed appropriate to the S/N requirements of the observation -- the slower modes have considerably lower readout noise, and they are not slow in 2x2 binning mode. It is a good idea to keep calibration and observation frames at the same binning, however. This advice is particularly appropriate for imaging, where the readout time will become a substantial fraction of the "shutter-open" time, but is less important for long spectroscopic observations. --- AD

Gratings or grisms can be expected to be where the GUI pulldown menus say they are: a few test exposures at the start of an observing will calm the more obsessive among us. Grating tilts are defined as angles from the zero-order angle, which is defined as 0 degrees to the collimated beam. They can be calculated by using the IMACS Grating Tilt Calculator, where the user enters the grating information and desired central wavelength. In addition to the tilt, the program also returns the wavelength range on the dectector for a slit centered on the field or mosaic array.
 
Tests of any kind can be accomplished with the internal arc-line and continuum (quartz-halogen) lamps in IMACS. These are located in the forward compartment of IMACS; they illuminate the screen that forms the IMACS hatch. Control of the lamps is from the IMACS Observing GUI. Light from the lamps hits the screen at a 45 degree angle and scatters to some extent over the rest of the forward compartment, so these are intended for tests and test calibrations rather than for application to actual observations. The lamp choices are quartz-halogen, neon, neon + mercury, krypton, argon, and xenon.  The krypton and xenon lamps are relatively faint.

The rapidly-deployed screen at the pupil of the telescope is well suited for science calibrations. In addition to the quartz-halogen lamps (settings are QL and QH), this system includes pairs of helium, neon, and argon lamps (controlled separately), and a direct view of hollow-cathode sources in the center of the screen (the area obscured by the secondary in normal operation).  The QL setting is appropriate for direct-imaging flats. A few seconds for I-band flats, and up to a few tens of seconds for B-band, are sufficient to reach half-saturation (recommended) of 32k counts per pixel (unbinned).  Sky flats may be a good alternative for any blue flats, and a necessity for any ultraviolet (user-supplied) filter.

The QL lamp setting and an exposure of 10-20 seconds is appropriate for spectral flats with the grisms, while the gratings, particularly the higher dispersion gratings, require the QH setting. Maximum exposure levels of 20k-30k counts are recommended for flats. Spectral flat fields should be taken immediately after one or a series of spectral observations, before the slitmask, grating, or grism has been retired.  An exposure of 10 seconds will produce ~10k counts per pixel for the 200-l grism (unbinned).  Spectral flats for the 1200-l gratings, particularly in the blue and 2nd order, will require several minutes of exposure, but the user should consider the exposure level of the objects themselves when calculating what number of counts is actually required.  Spectral arc calibrations are typically He + Ne + Ar, but it is sometimes useful to take He alone for identifying a spectral range, or in cases of very low dispersion (e.g. LDP) where the Ne and Ar lines blend together.  A spectral arc atlas for representative setups is given at https://www.lco.cl/?epkb_post_type_1=imacs-spectral-atlas.

B. Getting Acquainted with the Control Software

The observing station in the Baade control room has a computer dedicated to controlling the instrument functions of IMACS (presently Llama). Llama is Mac minl running Mac OSX. (See our mini-tutorial if you are unfamiliar with the Mac OSX environment.) Data can be burned onto 4GB DVDs (see the tutorial), transferred to a personal laptop, or onto a personal USB or FireWire external hard drive. Llama has two widescreen flat-panel displays and supports multiple workspaces, so there is room to display the GUIs and the Quick Look tools and a window to run a data reduction package such as IRAF.  The IMACS operation software runs on the Llama computer only. 

The Startup Window is started by a script called 'imacs'. (It is in the observer's path and no parameters need to be given.)  The user learns from the Startup Window the parameters of the configuration. The observer can change the Observer name but is unlikely to modify any of the other entries.  When satisfied, click 'OK'.

How to use of the IMACS-GUI, including details about the fields and pull-down menu choices control the instrument functions, can be found at http://instrumentation.obs.carnegiescience.edu/Software/IMACS/mechanics.html#user. The IMACS-GUI controls all the mechanical functions of IMACS but does not control the outboard guiders (the province of the TO) or the CCD mosaic cameras (separate GUIs -- see below). The IMACS-GUI displays, in a simple format, the present complement of changeable elements: mask, filters, gratings, grisms, and calibration lamps. Each of these has a pull down menu for selecting a new element. 

NOTE: Configuration changes of the Mosaic Camera are locked out while observations are in progress, but not configuration changes of the instrument itself.

Those who are interested in viewing the technical data from IMACS -- temperatures, pressures, guider positions, etc. -- may choose "Hardhat" in the Options pull-down menu.

C. Running the Mosaic CCD Cameras
The 8K x 8K Mosaic CCD cameras, Mosaic2 and Mosaic3 (and the guide cameras as well) run from Fedora LINUX PCs; these computers reside in the 'Equipment Room' rather than the Observing Room. The Mosaic Camera GUI will look somewhat familiar to those who have used other LCO instruments, although CamGUI has some added features. The GUI for the f/4 camera has a blue tint, and the f/2 GUI a red tint; when the Quick Look tool displays a picture, the right frame area changes color to match the camera that is displayed displayed.  The data paths for the two CamGUI can be set to tsame or to different directories.  Mosaic2 and Mosaic3 are completely independent, so they can be run at any time.

Filenames are specific to Mosaic2 and Mosaic3.  Mosaic3 writes files iffxxxx.fits, where iff stands for IMACS file "four" (f/4), and Mosaic2 writes iftxxxx.fits for IMACS file two.  The difference in the filenames means that a single data directory can be used to store both f/2 and f/4 data, although separate directories are a perfectly good alternative.

A small, bright-green icon of the 8 CCDs is used to select the specific data that will be stored, that is, the observer can 'switch off' CCDs (although, since the CCDs are read out in parallel, this will not shorten the readout time.)

The IRAF program mosaic will pack this information into single frame that includes all 8 CCDs and a coordinate system that takes into account the gaps between CCDs.

The f/2 and f/4 CamGUIs can be launched from the pull down menu "Modules" in the Mechanical GUI, or from each other, also from the "Modules" menu.

Submenus of the Camera GUIs 

ExpTime is the exposure time in seconds for the next observation: the user can change this time during the exposure by entering a new number and hitting return. Setting up a loop can be a convenient way to take flat fields or make a sequence of timed observations.  The bar to the right of ExpTimes shows the progress of the exposure in graph form --- the barfills up as the exposure is taken and empties as the exposure is read out.

From the pull-down menu 'Options,' ExpProgress opens a window where the time remaining on the exposure, and a running average of the seeing from the Principal Guider (when available) are displayed, large enough to be seen across the room.

ExpType includes bias and dark frames. Observations and calibrations are done by choosing Object or Nod&Shuffle (N&S), which brings up another GUI to set the nod-and-shuffle parameters.

The "Disk" bar at the bottom of the GUI shows the available disk space.  The bar changes color from green to red as it moves from right to left and a warning is given if (ever) disk space is dangerously small. 

Start begins a science exposure. The exposure can be paused (e.g., for clouds) and resumed by toggling with the "pause" button, but observers should remember that cosmic rays are accumulating while the exposure is suspended. The progress of the exposure can be seen in the graph-bar that fills to the right; as the CCDs read out the bar empties to the right. Test pictures will typically be taken rebinned by 4 x 4 or 2 x 2 to save readout time. The "Snap" button automatically sets the 4 x 4 mode, takes an exposure, and -- unlike the "start" command -- does not update the frame number at the end of the exposure (will be overwritten).  The binning is returned to its former value automatically after "Snap" finishes.

Exposures can be stopped or aborted. At any time a user can stop the exposure and readout will immediately begin. During read out -- at the intended exposure time or as the result of hitting "Stop" -- the user can abort the readout, which results in the data being dumped and the file number not updated. A wipe of the chips follows, but if the user has an itchy trigger finger, it is recommended that a "Snap" of minimal exposure be taken to insure a clean chip, before any science exposure is begun.

The Mosaic2 and Mosaic3 CCD cameras have two readout speeds, "fast," and  "slow"  The default "fast" readout speed takes 81 sec with read noise of about 4-5 e- per per pixel, varying from chip-to-chip. The "slow" readout takes 144 sec with a read noise of about 3 e-.  For 2x2 binning these times are 29, 37, and 46 sec, respectively.  (Chip-by-chip read-noise figures are tabulated in Section 1 -- Description of IMACS.)  Binning does not change the readout noise, but it does increase the signal, so it should be considered whenever S/N per exposure will be low, especially at f/4, where the data are usually oversampled for a reasonable slit width or for median seeing.

Recommendation: To avoid confusion, observers should in genera leave the speed in the "fast" readout mode, except in the case of very thinly exposed spectroscopic exposures.  Good to know: As with the binning, the read speed can be changed any time before the readout begins, e.g., during an exposure. --- AD

The 8 CCDs are read out in parallel. A QuickLook (QL) Tool -- launched from the 'Options' menu of the CamGUI -- will display and image of the 8 CCDs in real-time as the readout proceeds and as the data are stored to disk. The default display fills the screen, but the user can select --- from the "options" menu --- a quarter-sized QL-tool.  Due to the limited number of resolution elements on the display, the image of the mosaic is undersampled, a lot or a little, depending on the binning. The panel on the right is tinted red for an f/2 camera image, blue for an f/4 image.  The QL can be used to call back a previous frame for inspection.

The Quick Look Tool: The circle with the green arrows shows the orientation of the screen, North (long) and East (short).

A magnifier window allows the user to zoom in at 1X, 2X, 4X, and 8X to see the data full-scale (at whatever was the rebinning the data was obtained).  The magnifier window can be stretched to a two-times-larger box size.  Placing the cursor in the main window (if the cursor is in the magnifier window it will be automatically returned to the main display) and hitting the left mouse button will sample the exposure level at a given pixel position. Statistical information is given by hitting "return" or the space bar when the cursor is positioned appropriately. The radius feature controls the area of the fit for that exposure.  This rms diameter of an image is not equivalent to what is calculated in IRAF imexamine, which is more representative of the FWHM, but is still useful for relative measurements.

There are several autoscale options and the scaling can be set manually -- the choices are listed as a pull-down menu on the side of the QuickLook screen.  Note that you can force the chips to be scaled together, or separately, with the two choices below the autoscale menu.

 The magnifier window of the Quick Look Tool. 

D. Reading and Saving Data: Rebinning and Subrasters

It is not necessary to save the data from the whole array for each observation. Clicking on the small green moniker on the Camera GUI brings up a menu, which allows the user to specify any subset of CCD chips that are to be recorded as disk file.

The user can also select to rebin the array (or the selected CCDs) separately in X and Y -- this can be particularly useful in spectroscopic observations where it may be desirable in either the spatial or dispersion dimensions to rebin the data (lower read noise; compacted data).

The software also provides for a very convenient way to set up subrasters. As many as 8 subrasters can be selected for each CCD; the user fills in a table of X0 and Y0 coordinates and the size of each sub-array. The cursor on the QuickLook display can be used to select the center of the subraster: type 'a' to transfer the position and set the subraster size. Taking a picture with the subraster feature will save time, since the CCDs are clocked rapidly in between the readout of subrasters (at whatever binning).  However, as the number of subrasters increases, this advantage will diminish. Data for subrasters can be stored separately by choosing "save mode = minimal" or embedded in a full frame, with zeros filling the non-subrastered areas, by choosing "save mode = full" in the Subraster dialog box.  The default is Save mode  = full -- it is essential to use the "full" mode if you are using the mask alignment software, ialign and ifalign.

When viewing an image on the Quick-look tool, typing the number 1 to 8 on the keyboard will move the cursor to that chip image.

Recommendation:Use the QL to check that the frames are what you expected, but use IRAF and DS9 to measure seeing, look for faint objects or features, look for spectral features, etc.-- AD

E.  Let the Observations Begin!

OPEN THE HATCH! From long experience, the IMACS team strongly recommends opening the hatch before attempting observations -- use the popup-menu at the top of the IMACS-GUI.  

Please remember to CLOSE THE HATCH when you are finished observing.

Direct Imaging
The simplest way to use IMACS is as a direct camera, in either the f/4 (~15 arcmin square) or f/2 (~27 arcmin diameter) modes. All that is required of the observer is to select a filter and provide coordinates of the field to the TO, who will set find stars for both the Principal and Shack-Hartmann guiders.

No focus required: A main feature of IMACS is that, in normal operation, focus is automatic. IMACS is internally focused on the surface defined by the slit mask (matching the telescope's focal surface). The optical design, in combination with the focusing stage of the CCD mosaic array, automatically adjusts for temperature to maintain both focus and scale. The Shack-Hartmann guiders (both centerfield and outboard) are set up to maintain the telescope focus at the same surface. The observer is strongly advised to use this autofocus feature of IMACS -- attempts to focus "manually" will have unpredictable results.  The standard filters provided in IMACS have offsets that have already been determined, so part of the autofocus procedure is to automatically compensate when they are inserted. In the case of user-provided filters, a focus zero point at a given IMACS temperature needs to be determined by the Instrument Specialist and entered in the offset table

As mentioned above, and flat fields are best taken with the pupil screen of the telescope or the twilight sky.  Even the internal flat field lamp will be sufficient to measure pixel-to-pixel variations, but for overall illumination patterns the first two options are much preferred. The "loop" feature provided in the camera control may be convenient in taking flatfield exposures. Generally the object is to expose the flatfield to a level several times the count per pixel level of the deepest science exposure. This is sufficient for all but the most critical applications.

Fringing with the Mosaic2 and Mosaic3 CCD cameras is only a few percent in in the CTIO-I-band, and ~5% in the Sloan 'z'-band.  Fringing can generally be removed from direct images by making a "fringe frame" out of a dozen or more exposures of relatively empty sky fields and making a median image to eliminate stars and galaxies. The fringe frame, appropriately scaled for the exposure time and/or sky brightness, is subtracted from the science image.  

Flat field exposures for imaging (see below) need not be taken at the same telescope (IMACS) position as the observation.

Making an IMACS f/2 or f/4 mosaic

As described in "Imaging with IMACS" in the first section of this manual, a script feature has been added that allows the observer to make mosaics, by taking in succession 2 to 5 exposures that are offset in order to cover the chip gaps. A default offset is stored in the system, so the observer needs only to set the exposure time and the number of exposures.  Two exposures are adequate to cover all but a small square in the middle, but 4 or 5 exposures should be taken if the goal is to obtain a relatively uniform S/N ratio across the entire mosaic.

Parameters for the script are set by clicking on the 'create' button on the 'script' line in the appropriate camera GUI.  Choosing the 'default' option will bring up a window where the number of positions and the integration time can be entered.  The user can name and save this file if the configuration is to be used frequently. 

In order to use the Mosaic script feature, you must open the 'Telegui,' one of the modules in the f/2 or f/4 camera (see above).

When after setting on the field center, have the telescope operator tweak the field center so that the chosen guide star is in the middle of its 105 arcsec radial range.  This ensures that the guide star will not be lost during the offsets.  The Shack-Hartmann guider is not used during this process of making a mosaic, due to its constraint of sending light down a small aperture. The SH test should be done before the mosaic exposures are started. If the total exposure time is under 20-30 minutes, the corrections made at the start should be adequate to hold focus and mirror shape. If focus is changing more rapidly than this, a break in the execution of the mosaic script (in order to run a SH test) can be accomplished using the 'pause' feature.

Flat Fields and Spectral Arc-Line Calibrations

When not in use, IMACS is sealed from the dome air by a roll-up hatch, actually a small movie projection screen riding in channels. In the closed position the screen reflects a series of calibration lamps mounted within the front section of the instrument. There are 5 arc-line (Penray) lamps: neon, neon + mercury, krypton, argon, and xenon. There is a halogen lamp is used for flat fields; it has 7 levels of illumination -- faint levels for direct flats, bright levels for spectroscopic flats. All lamps are operated from the IMACS MechGUI.

These internal lamps are useful for daytime checks, but for science flats and arcs the observer should use the lamps that illuminate the screen at the pupil of the telescope itself, just in front of the secondary mirror.  (The flat-field screen and the lamps are operated from a program 'ffs' that is initiated from any terminal window on Llama.) The illumination of the instrument through the pupil screen is greatly preferred because it closely matches the light coming from the sky itself. This is generally the reason for doing sky flats (instead of more traditional "dome flats" done with a screen on the inside of the telescope enclosure), so the pupil screen flats should make sky flats unnecessary, except perhaps when the bluer color of the sky is better suited than relatively cool halogen light sources.  Using the pupil screen does not involve the primary mirror, but the mirror cover must be open for the tertiary to see the pupil screen and reflect the light to IMACS.

Imaging flat fields and spectral flats (with the grating or grism in the beam) are both taken this way, as well as arc-line calibrations, which use the arc-line sources available in the ffs menu. Example arc spectra for the main dispersers (long slit mode) are available in the IMACS spectral atlas, and also in the telescope dome under "Calibrations_and_Tools" on each of the datadisks on the observer workstations.

Spectroscopy

Long-Slit Spectroscopy:IMACS has a set of long-slits that have been cut into multislit-mask blanks, in a range of widths.  One or more of these, installed in the slit-mask server, can be inserted in order to do single object (point or extended) spectroscopy, or spectroscopy for for 2-3 objects along a single line.  The procedure for placing the object(s) at the proper position is described in the IMACS cookbooks. It involves taking a picture with IMACS to identify the target(s), a first move to get close to the target position using the IRAF command 'toslit', and then a fine adjustment  that uses the CCD pixels of the guider -- 'icobject' -- that does not require the guider to be moved, so the operation is accurate to a small fraction of an arcsecond. 

After a longslit mask itself is inserted, it is a good idea to check that the target position is still where it was on the setup frame -- sometimes the IMACS mask does not return to exactly the same position. Once the observer is satisfied that the object or object are in the right place, the disperser can be inserted (grating or grism), the proper filter inserted (usually Spectroscopic but could be a blocking filter), the binning and exposure time chosen, and the spectroscopic observation can begin.  The procedure is described in careful detail in the IMACS Cookbooks.

After the spectroscopic observation, or a set to a new sky position, one or more spectral flats and arc-line calibrations should be made before the grating or grism is uninstalled.

Setting up Multislit Spectroscopy:  Multislit spectroscopy requires considerable preparation, as described above, but observations are comparatively simple to perform.  With the information copied from the ObjName.SMF file produced by the 'maskgen' program 'intgui' (the line that starts by '!.OC' (Observing Catalog)

!.OC R0221M1 02:21:56.000 -03:47:10.00 2000.0 0.0 0.0 -226.15 OFF  02:20:52.9 -03:46:28 2000.0 02:22:56.3 -03:44:48 2000.0 

the TO has all the information (Object name and coordinates, rotator angle, PG and SH guide star coordinates) required to accomplish the initial setup, using the GMAP display that communicates with the TCS (Telescope Control System).  The observer should prepare a catalog file with these lines to pass to the TO at the beginning of the night (or run). As the telescope slews to the target coordinates, the IMACS guiders are preset to the positions required to acquire the guide stars.  In general, when the target is reached, the two guide stars will be displace by some number of arcseconds (typical 2-10) because of pointing errors in the telescope and small errors in the rotator angle.  In order to take advantage of the precision metrology of the IMACS guiders, it is critically important to move the stars to the guiders rather than the other way around, through a 'coordinated offset'  of the telescope that adjust azimuth, altitude, and rotator angle to place the stars on the guider crosshairs.  This process is called 'centering' -- make sure it is being used, as this is not general practice for Magellan instruments. In this way, the IMACS-Magellan system corrects the pointing and rotator errors.  The procure is sufficiently accurate to bring the alignment stars of the slitmask within their target boxes at the first setting, if the field center and guidestar positions are accurate to ~1 arcsecond.

The observer has already run the IRAF-IMACS program icbox (part of the IMACS package -- type 'imacs' to load) on the slitmask image (see the Multi-Slitmask Cookbook), which has produced an alignment file and a subraster file.  After the guide stars have been acquired, the next step is to load the subraster file ('full' or 'subraster' -- on the CamGUI) and take a picture.  (Take a full subraster frame, not minimal! -- see the Subraster GUI). Both this picture and the picture of the slitmask taken in preparation for observing should be binned 1x1 for f/2 or 2x2  for f/4.  This should be an R-band exposure, regardless of the wavelength range of the observations. (If the difference between the R-band and wavelength region of interest is an issue, make sure the mask is designed with the slits at the paralactic angle.)  With the recommended R=18-19 mag for alignment stars, 15-25 sec exposures should be sufficient, (but longer in very poor seeing).  As the picture is read out, the observer checks to see if the alignment  stars are visible in the subraster images -- if not, there is a problem!  A full frame could be taken to look further afield for the stars (in other words, for an RA-DEC shift), but it is likely that the problem is more serious, for example, the wrong rotation angle or problems with the astrometry.

With this subraster picture, the observer then runs the IRAF-IMACS program ilaign, which will -- 'uno por uno' -- display the subraster images with the alignment stars, which the observer confirms or re-marks -- see the Cookbook for details.  The result of this process is to provide offsets in RA, DEC, and rotator angle to bring the stars to the centers of the boxes, accomplished through a coordinated offset in which these offsets are sent to the Telscope Control System.  The aligment procedure would be finished at this point if the slitmask inserted in exactly the same position each time it is inserted.  Since this not always the case, the next step is to locate the multislit mask for the observation and take a second picture.  When this subraster exposure through the mask is finished, the IRAF-IMACS program ifalign is used to calculate a coordinated offset, this time taking account of the actual position of the alignment boxes for this insertion of the mask.  The residuals at this point should be less than an arcsecond, or two at the most -- throw out any bad stars!   The result of this final alignment should be a calculated offset of < 0.2 arcsec in both RA and DEC and angular misalignment < 0.010 degrees.  A second iteration with ifalign may be required, but generally should not be.

Recommendation:Don't spend additional time tweaking up with ialign, once you get a reasonable result.  The slitmask may not insert at exactly same position as it did when you took the previous mask picture, and you need to move on to ifalign to take to correct for this. --- AD 

Once completed, the alignment will hold as the field is tracted, within the limits of the flexure between focal plane and guiders, as described above.  The telescope must be set for rotation guiding mode, which means that it is tracking in altitude, azimuth, and Nasmyth rotator, to keep the PG star and the SH star centered.

Making and calibrating the mutlislit observations: After alignment is complete, insert a grating (f/4) or grism (f/2). When operating properly, latchup of gratings and grisms results in only small shifts of wavelength.  The grisms in particular repeat well (within 1-2 pixelx at the detector), but at certain gravity orientations the gratings (which are mounted in heavy grating-tilt-mechanisms) appear to latch up differently from one insertion to another, resulting in a (x,y) zero point shift on the detector of as much as 5 pixels. The reason this is an issue is the degree to which spectral flat fields taken in the afternoon, for example, can be used for the subsequent observations. The safest procedure is to finish each multislit spectral observation with 1-2 spectral-flat-field exposures and an arc-line exposure using the pupil screen in the telescope. The guiding needs to be turned off when the puil screen is in, but the guide stars will be re-acquired afterwars with no loss of precision in the mask alignment. 

If one grating setup is to be used throughout the night, another way to deal with the f/4 latchup uncertainty is to use the grating in 0^th-order (angle = 0 degree) to do the imaging, that is, leave the grating installed and simply move back and forth between 0^th-order and the angle used for observation, usually 1^st or 2^nd order. This procedure also has a built-in uncertainty, but it is thought to be a pixel or less and, at any rate, an internal arc can be taken to reset the tilt exactly, should the program require this unusual precision.The COSMOS software package (see below) provides a quick-look spectral extraction tool that will subtract the sky and apply a wavelength scale so that the observer can gauge the quality of selected multislit spectra as the data are gathered.   It is perfectly possible to use COSMOS to fully reduce data during the observing, but it generally requires a person dedicated to the task.

Standard-Star Spectral Calibrations: Slit-viewing mode (described in the next section) is the preferred way to make observations of standard stars for flux or radial velocity calibrations.  For radial velocity standards, ask the TO to set to the star and then move it to one of the sections of the multi-width slit -- it is a good idea to choose the slit that best matches the seeing at the time.  For flux calibrations, give the TO the coordinates of, for example, one of the Hamuy standards stars, and ask the TO to move the star into the 7-arcsecond diameter hole.  This assures that nearly all the light passes through to the spectrograph.  Arcs through the hole will be low-resolution, but the ajoining arc from the multi-width slit are applied when using the CFslit_f2.SMF or CFslit_f4.SMF descriptors with the COSMOS software.  If relative flux calibrations are all that is required, the widest slit position is a good choice.

As described below, It is not necessary to save the whole Mosaic array for spectroscopy in Slit-Viewing mode, as the spectrum will only appear in a relatively narrow band, on chips 1-4 in N&S orientation and on chips 2 and 5 in normal mode.  However, if these are the only observations made with the slit-viewing mode (as opposed to a full program of spectroscopy), it is simpler in terms of data reduction to take full-frame exposures (though 2 x 2 binning is sufficient).

One of the slits or alignment boxes in a multislit mask may also be used for spectrophotometric calibrations.  This is the preferred method when the observations are in N&S mode with the f/2 camera, where -- because of the rotation of the grism --- the multi-width slit of Slit-Viewing mode runs in the wrong direction relative to the dispersion (and the slits may disappear in the CCD gap). The 7" hole can still be used, of course, and an approximate wavelength calibration can be done with an arc taken through the hole, but for this case it is probably easier to use one of the alignment boxes on the multislit mask,  and then to use the "spectral-map" solution of the COSMOS reduction software, with the SMF file appropriate to the multislit mask, to determine a wavelength scale appropriate for the alignment box.

No setup-form request required for this mode, since it is always available. 

In addition to its two outboard guiders that swing around the ~27 arcmin square field, IMACS has a centerfield guider (CFG) that travels radially into the centerfield (optical axis) of the instrument. By switching between the CFG's two sets of optics (housed in 2 barrels), the observer can view centerfield - a square field 105 arcsec on a side - or perform an on-axis Shack-Hartmann test.

From its fully-inserted position on the optical axis the CFG is pulled back approximately 2 arcmin to turn IMACS into a slit-viewing spectrograph. The utility of SlitView mode is that sufficiently bright objects can be acquired in real time with the TV camera and placed on the desired slit -- time is saved because the CCD Mosaic Camera is not needed for target acquisition. The 'pickoff' (viewing) mirror incorporates a multi-width slit at the focal surface of the telescope. The layout is 5 side-by-side slits each about 17 arcseconds long, with widths of 0.25, 0.50, 0.75, 1.0, 1.3 arcseconds, and a displaced 7.0 arcsec diameter hole for photometric calibrations. When inserting the CFG any installed slit mask will be removed. In its place, a baffle will be automatically inserted behind the CFG to mask out the entire field except for the slit and hole (to eliminate surrounding sky light). The Principal Guider (PG) can also interfere with the CFG. If this conflict exists, the user will also be prompted for permission to move the P, which will result in the loss of any fine adjustment such as might have been accomplished in a multi-slit spectroscopy setup. The PG will be moved to near the end of its range, and is at any rate mostly occulted by the CFG, so its use as a guider is very limited.

Observing in the slit-viewing mode is straightforward -- the same as for typical non-imaging spectrographs. The observer provides target coordinates to the TO, who will acquire for the observer to identify) the object at the centerfield position. The TO will then apply an offset of 115 arcseconds west or east (for the two standard settings of the rotator, north up, or north down, respectively).  At this point the obsever should see the object reflected off the slit, close to the small cross that defines the center of the field. (If a number of objects are to be observed in this way, the observer can ask the TO to skip the Centerfield acquisition step and offset directly to the SlitView position.)  A second procedure is to use the toCFslit script in the IRAF-IMACS package, which sends the telescope by a coordinated offset to the center crosshairs of the CF guider.

In this simple mode of operation a grating or grism is left in place (using either the f/4 or f/2 cameras) and the observer proceeds from object to object without reconfiguring IMACS, except for optional grating or grism change. Make sure a filter is inserted when spectra are being taken. This will normally be the "Spectroscopic" filter, but other filters can be used for limiting the bandpass of the observation.  Traditionally,  guiding during these observations has been with reflected light from outside the slit of the object itself, however, at Magellan we use the regular IMACS guiders as "offset guiders." The TO chooses suitable guide stars for the Principle and Shack-Hartment guiders (PG and SHG). The SH test provides continual adjustment of telescope focus and alignment and corrects the mirror figure -- this can be skipped if the observation will be a few minutes or less, but after a series of such short observations with the Slit-viewing feature the SH test should be run to restore the image quality.  The Centerfield Guider occults almost half of the field of the PG, so it could be that there is no suitable guide star in the field. (Remember, however, that most single-object observations have the option of rotating the field, which will almost certainly bring a guide star into the available range.  However, if paragalactic angle constraints, or some angular constaint (for example, two objects on the slit separated by 10") results in no suitable star for the PG, the TO can implement what is called "tilt guiding," in which the SH Guider both guides and performs the mirror adjustments. Although the rotator works open loop in this case, and the guiding updates are done only every 30 sec or more, neither of these present a problem. (Note: Tilt guiding can also be used as a backup for other kinds of IMACS observations, should the PG be unavailable, even for direct imaging.

Data acquired in the slit viewing mode will only appear on four CCD chips, #1, #2, #3 and #4, of the f/4 Mosaic1 camera in its Nod&Shuffle orientation where dispersion follows the short axis of the 2K x 4K detectors. The user can save only the data from these 4 chips, or choose only a range of columns or rows, using the controls on the Camera GUI to create subraster files.  A link to a subraster file specifying sections of chips 1-4 can be found in the IMACS Cookbooks, "Observing with the Centerfield Slit-Viewing Mode" -- paragraph 3.  In "normal" orientation mode, such as is used in the at f/2 with the Mosaic2 camera (and with the grisms also in "normal" mode rather than rotated for N&S)  the data will appear only on CCD chips #3 and #8.

The repeatability of the slit-viewing position has not been measured, but it is expected to be less than a few pixels.  If very accurate flat-field spectra are required, it is advisable to use the pupil screen and take flat fields and arcs before returning the CFG to its stowed position -- the CFG may not return to the exact same position if inserted later for the flat field exposure.  Another approach is to use a single flat or set of flat and to account for any mis-registration by creating a separate 'slit flat' -- the one-dimensional sensitivity function that accounts for the structure along the slit itself.  The  shift between the standard flat and the actual observation can be determined by collapsing the sky continuum in the dispersion direction and cross correlating with the standard flat to determine the shift.

A final warning regarding scattered light: the baffling of this mode is imperfect, so about 1% of the sky light intensity will be added to the star signal, at very low spectral spectral resolution because it is leaking through a broad aperture.  For moderately bright targets, this will be removed through normal sky-subtraction. However, targets that are themselves near or below the sky brightness level should not be attempted with this mode: use the multislits or one of the very long slits in this case.F.  A Checklist for IMACS Observations
Essentially, there are nine items to double-check before beginning an IMACS exposure, especially a long one:  CamGUIs

1. Exposure time
2. ExpType: Obj or N&S
3. Binning
4. Full readout (or selected chips) or subraster
5. Object name
6. Readout speed (and read noise)

             IMACS-GUI: (MechGUI)

1. Calibration lamps off?
2. Slitmask -- proper mask?
3. Disperser -- proper grating or grism? Correct tilt of grating?
4. Correct Filter

If you are new to IMACS, it's a good idea to keep this checklist handy.  There's a lot to remember!

F. Shutting Down 

Observers should close the hatch and turn off all calibration lamps at the end of an observing night. It is not recommended to shut down the GUI displays. As with other Magellan instrumentation, data will normally remain on disk for a substantial time, but not necessarily forever -- its still a good idea to archive data before the end of the run.

G. Data Reduction Pipeline

A complete data reduction package for multislit observations -- COSMOS (http://obs.carnegiescience.edu/Code/cosmos) -- has been developed by Gus Oemler. It includes the sky-subtraction algorithm perfected by Dan Kelson (PASP, 115, 688, 2003). COSMOS includes programs that deal with the complicated data format produced by the Durham-IMACS IFU. In addition to the sophistication of its algorithms, COSMOS processes data fast enough, and produces of high enough quality that extracted, flux calibrated spectra can be provided contemporaneously during observing. Although some programs will benefit from storing raw data and additional reduction procedures,  observers have the opportunity to leave Las Campanas with data that are ready for scientific analysis. 

Information about installing and running the COSMOS software can be found at http://code.obs.carnegiescience.edu/cosmos/Running.html.