[Skip to Content]


General Links
Professional Links
Australian Human Resources Institute Diversity Awards 2013 Finalist

Spectroscopic Observing with IRIS2


Wavelength Formats

With its combination of two different long slits (SLIT_150 and OFF_150), two grisms (SAPPHIRE_240 and SAPPHIRE_316) and a range of order-sorting filters (K, Ks, H, Hspect, J, Jshort, Jlong), IRIS2 offers a variety of wavelength formats. In practice, most observers will be using the grisms in a limited number of formats, which are listed in the following table along with the standard name we use for that format according to the blocking filter used.

Grism Order Slit R Start
Mean Dispersion
K SAPPHIRE_240 m=1 SLIT_150 2250 2.02 2.249 2.37 0.442 Format reversed!
Blue cutoff set by array + filter; red cutoff set by filter.
Ks SAPPHIRE_240 m=1 SLIT_150 2200 2.02 2.249 2.31 0.442 Format reversed!
Blue cutoff set by array edge; red cutoff set by filter.
H SAPPHIRE_316 m=1 OFF_150 2270 1.47 1.637 1.79 0.341 Cutoffs determined by H filter.
Hspect SAPPHIRE_316 m=1 OFF_150 2270 1.46 1.637 1.81 0.341 Blue cutoff set by array edge; red cutoff set by filter + array edge.
J SAPPHIRE_240 m=2 OFF_150 2490 1.17 1.220 1.33 0.225 Format reversed!
Cutoffs determined by J filter.
Jshort SAPPHIRE_240 m=2 SLIT_150 2460 1.04 1.158 1.26 0.232 Format reversed!
Blue cutoff set by array edge; red cutoff set by filter.
Jlong SAPPHIRE_240 m=2 OFF_150 2490 1.10 1.220 1.33 0.225 Format reversed!
Blue cutoff set by array edge; red cutoff set by filter + array edge.

A given grism and slit combination delivers a more-or-less fixed wavelength format onto the detector. This wavelength format is then selected by the particular filter used. So calibrations for the K/Ks, H/Hspect, J/Jlong will each in turn be similar, but with different start and end wavelengths. "Centre" wavelengths in the Table above refer to the wavelength at pixel 513, near the centre of the array.

Note that the format of the J- and K-band spectra is reversed on the detector (i.e., wavelength decreases with pixel number), compared to that for the H-band spectra, because of the way the SAPPHIRE_240 grism has had to be mounted for optimum overall wavelength coverage.


  1. To use the special slit-finding and marking plugins provided by Chris Tinney, you must enter cgt_skycat before starting skycat.
  2. Depending on whether you plan to do Js, Jl, H, or K-band spectroscopy, you must first ask the Night Assistant to load the correct aperture definitions into the CCS for the SLIT_150 (for Js or K) or OFFSET_150 (for Jl or H) slit.
  3. Slew to the target, but leave the telescope in the Reference Axis.
  4. If you are working at a similar hour angle to the last observation with this slit, you should not need to re-check the slit position, and can skip to Step 8.
  5. Configure IRIS2 for imaging in the band you plan to get spectra in, but also with the appropriate slit in the slit wheel.
  6. Take a dummy image of the sky background through the slit, say 5s in K or H, or 10s in J. Occasionally the image WCS isn't picked up by Skycat. In this case, reload the frame manually from disk (e.g. /data_vme14/aatobs/iris2_dummy/110619/ccd/a.fits)
  7. When the image appears in Skycat, select Telescope -> Find Slit in Current. After a few seconds, you should get a dialog box with the parameters of the fit to the slit curvature. Click "Yes", and you will then be shown a new overlay of the slit fit in green on the slit image.
  8. You can turn the latest slit trace overlays on or off at any time using the Telescope -> Show Slits / Hide Slits options.
  9. Move the slit wheel back to the OPEN position, and take a dummy sky image at the Reference Axis.
  10. Switch to Axis A in the PTCS, and take another image. In the File -> Bias Image... dialog box, load the dummy image at the reference position, and turn on subtraction. Your target should now be visible somewhere near the cross marked on the slit at Y=580 corresponding to Axis A. Position the cursor as close as possible to the centroid of the target, then hold down Shift while clicking the left mouse button. This brings up a menu of options, including "Move object to CENTRE slit (Y=580.0)" and "Move object to OFFSET slit (Y=580.0)". Select the appropriate one, wait for the move to complete, then take another dummy image.
  11. Once the target is within 10 pixels or so of the Axis A position, relative telescope offsets are too coarse to get it lined up exactly. Instead, ask the Night Assistant to find a nearby guide star, and begin guiding in Axis A. Take a new image, in case the start of guiding has displaced the image at all. Shift + left-click again on the target, and select the appropriate "Move to..." option. This time however, instead of accepting the suggested telescope offsets, take note of the recommended Guide Probe offsets, and call them out to the Night Assistant, who will offset the Guide Probe accordingly. Click "Cancel" in the suggested offsets pop-up window, and take a new image, which should then have the target directly beneath the green cross for Axis A. If not, re-determine and apply new Guide Probe offsets from the image. If the offset is only a pixel or so, then take another image to make sure the offset is real, and not just an artefact of the seeing.
  12. Notify the Night Assistant that you will be switching to Axis B. Select Axis B in the PTCS, then wait for the move to complete, and for the Night Assistant to resume guiding at Axis B. Take another image, which should show the target still directly underneath the slit overlay, but in row Y=640. If not, you should ask the Night Assistant to re-start the guiding at Axis B, in case guiding commenced at a bad time.
  13. Switch back to Axis A, and take one more image to ensure the target comes back to exactly where it started, under the cross.
  14. After checking the object name, exposure parameters, number of quads, and DR recipe, you can now save and execute the observing sequence.
Additional notes for blind acquisition (courtesy of Karl Glazebrook)
    Assume we have coordinates for a reference star and a blind target (on the same coordinate system) and we have worked out telescope offsets star->target
  1. Slew to reference star. [Image slit if changing fields/large RA slew]
  2. Image it (remember to take a sky in the reference axis and turn on bias sub in skycat) in Axis A.
  3. Shift-left click to bring it to the right slit (we are using the offset slit for H) and image again to check
  4. If OK now start guiding
  5. Image and do the shift-left click again, this time don't DO the offset just give the GP offsets to the Night Assistant
  6. Iterate until happy that the star is exactly on the slit.
  7. [up to here it is the same as the current manual for non-blind acquisition]
  8. Offset the telescope to the target using the offsets previously worked out (IRIS2 PTCS screen, do 'OFFSET relative to current position' then 'set base here')
  9. Offset the guide probes so they are still on the guide star. The Night Assistant can do this from the GPOFF tab in his interface. (Note the SIGNS are the SAME as the PTCS offsets)
  10. Start guiding again. When the guider locks on to the star and drags it back in to the center any telescope offset error is taken out.
  11. You are now guiding at the target position. Image again as a sanity check (right direction? East/North not mixed up?)
  12. Start spectroscopy sequence.

Spectroscopic flatfielding

There are pixel-to-pixel sensitivity variations of up to +/-20% in IRIS2, so flatfielding of your data is essential. The best procedure for this is to take 'lamp on'-'lamp off' sequences with the telescope pointed at the white patch on the dome. The dome flat patch on the windscreen can be illuminated using lamps mounted at the Prime Focus access area. There are two lamps, which can be plugged into a socket, the voltage for which can be controlled from the control room. The two lamps are a desk lamp, and a much brighter flood lamp - you will need the latter for spectroscopic flatfields, turned all the way up to maximum brightness. If the telescope is not positioned for dome flats, ask the night assistant, support astronomer or afternoon technician to put it there (Dome=0o, Windscreen=21o). It is recommended that dome flats should be taken in the same read mode as your object observations (i.e. MRM observations would need MRM dome flats, and DRM observations would need DRM dome flats). For MRM observations/flats you may probably want to have roughly the same number of reads in both dome flats and on-sky observations. If you've just taken arcs, take a 1.5s x 30 dummy dark to flush any residual images.

The simplest way to obtain spectroscopic flatfields is to edit a copy of the AAO_Arc_Flat.tcl sequence to select only the bands of interest, and follow the instructions as you run it.

It should only be necessary to obtain one set of flats and arcs per IRIS2 run, and the same set of calibrations can be applied to different nights by ORAC-DR.

Wavelength Calibration

Here we include arc line lists, sample arcs, and crude calibrations you can apply to your data. Because of the relatively fixed spectral format with each grism and slit, ORAC-DR does not require an arc to be taken before reducing object data. Instead, it applies an approximate linear wavelength scale to each reduced image (usually good to better than 0.002 µm), and also flips J- and K-band images to have wavelength increasing with pixel number. If you need better wavelength accuracy, you will need to obtain arcs and calibrate these yourself.

Like all grism systems in focal reducers, there is signficant curvature of the IRIS2 spectral format. If your data is taken in just one or two locations along the slit, you can derive wavelength calibrations for just these locations. If your objects are scattered along the slit, you'll need to either do a full 2-D wavelength calibration, or do a calibration for the location along the slit of each observation.

Obtaining calibration data

There are two ways to obtain line spectra for your data:

The recommended procedure then is

  1. Obtain arcs at the start or end of your night, for the purposes of dispersion-correcting your data (i.e. putting all the rows of the detector onto a linear wavelength scale, and removing the curvature along the slit). If wavelengths shift during the night they are most likely to do so in small shifts - not changes to the overall dispersion curve. So data at the start or end of the night should be adequate to dispersion-correct and undistort all your data.
  2. Use the night sky lines in your data to correct for small shifts in wavelength zero-point during the night.

Calibrating with Arcs

Plots of the arc lamp spectra obtained with the sapphire grisms in the J,H,K passbands are available. We also provide a couple of arc line lists:

The following arc atlases will only be appropriate for spectra extracted near the centre (i.e. Y~500-600) of the array. Experience indicates you must use at least a third-order (or higher) polynomial to get a good model of the dispersion curve.

Format Arc lamp Plots
Js + Sap240 + SLIT_150 Xe GIF Postscript
Jl + Sap240 + OFF_150um Xe GIF Postscript
Hs + Sap316 + OFF_150um Xe GIF Postscript
K + Sap240 + SLIT_150 Xe GIF Postscript

If you already have a Figaro-calibrated arc spectrum, you can quickly (and very crudely) calibrate your spectra by doing the following:

  1. pair subtract your image (i.e., subtract a sky image).
  2. straighten the image (not entirely necessary for quick-and-dirty reduction).
  3. extract your spectrum from the raw image. Make sure you use data from roughly pixels Y=500-600. Because the spectra are curved, these quick-and-dirty calibrations only work in this range of pixels. For other areas of the array you'll have to re-do the wavelength calibration, with arc spectra from that Y-pixel range.
  4. flip the spectrum in X for J- and K-band spectra, but don't flip it in X for H-band spectra.
  5. delete the existing (meaningless) .AXIS[1] information.
  6. copy the .AXIS[1] information from the spectrum in the 'calib' column with the xcopy command.

So the Figaro cammands will look something like the following (N.B. quotes are used to protect [ and ] which are special in Unix). If you want to do the straightening step, you will need the appropriate .sdist file and rename it to sdist.dat in the directory where you want to do the processing.

    isub raw_data sky_data tmp
    cdist tmp 1 1024 tmp1 2
    extract tmp1 550 560 tmp1
    irevx tmp1 tmp1            (for J and K spectra only!)
    delobj tmp1.axis'[1]'
    xcopy tmp1 Jl_Xe_cal_oct02 tmp1

Calibrating with night-sky OH emission

Plots of the night sky OH emission spectra obtained with the sapphire grisms in the J,H,K passbands are available (Note: these are from Oct 2002, so will not match precisely the current grism + slit set). The line list used to obtain these fits is available as two Figaro arc files oh_eso.arc and oh_eso_all.arc. These files are in plain text, and should be readily convertible into wavelength files for use in other data reduction systems. These files are derived from the work of Rousselot et al. 2000, A&A, 354, 1134 (ESO preprint) for OH calibrating VLT ISAAC data). The oh_eso_all.arc file is the list of all wavelengths for OH lines, including hyperfine splittings. The oh_eso.arc file is more useful - it is the list of lines indicated in the figures of the paper as being useful for calibrating spectra.

Experience indicates you must use at least a third-order (or higher) polynomial to get a good model of the dispersion curve.

Format Plots Comments Calibration
Js + Sap240 + SLIT_150um GIF Postscript Reverse raw images in X Js_OH_cal_oct02.sdf
J/Jl + Sap240 + OFF_150um GIF Postscript Reverse raw images in X Jl_OH_cal_oct02.sdf
Hs/H + Sap316 + OFF_150um GIF Postscript Do not reverse raw images in X Hs_OH_cal_oct02.sdf
K + Sap240 + SLIT_150um GIF Postscript Subset raw images from X=200-1024, and reverse in X Ksky_per_pixel_calib.sdf

The link in the last column contains a Figaro-calibrated OH spectrum. You can use the same procedure as above to apply this calibration crudely to your data. But the arc calibrations are better, typically having RMS for the calibration fits that are a factors of 2-3 lower. You may wish to use the OH sky spectra to perform fine adjustment of the zero-point of your wavelength scale after dispersion correction with the arc spectra, since then you really only want to know one coefficient - the wavelength zero-point, and the OH lines can do that quite well.

Smooth Spectrum (Atmospheric / Telluric) Standard stars

Finding standard (i.e. "smooth spectrum" stars of your favourite type) can be most easily done with the Gemini Telluric Standard Search Tool. Recommendations on how to go about choosing the most appropriate standards are also available.

You can also look at the ESO VLT ISAAC Spectroscopic Standard information.

Spectrophotometric Standards

There aren't any!

In the optical, one uses spectrophotometric standards that have tabulated fluxes and wavelengths. In the IR, there are no such standards. There are some "pseudo-standards", in the sense that tables of flux versus wavelength do exist for some stars. However the fluxes are either derived from models (for DA white dwarfs) or from a scaled version of the solar spectrum (for solar analogs). These standards are for space-based missions and are not particularly useful in calibrating ground based data.

So the usual procedure is to use 'smooth spectrum' standards (see above) to remove the effects of terrestrial absorption on your spectra. Then you can use the same standard (if its effective temperature and magnitude are known), or another standard (with known effective temperature and magnitude) to create an absolute flux scale.


A range of TCL sequences have been defined which can be used, with minor modifications, for most spectroscopic observing applications. These sequences are described in full on the IRIS2 Observing Sequences page.

Sky Subtraction

Sky subtraction is the arguably the single most important aspect of infrared data processing. In the optical, you can usually just flat field your data, extract a spectrum from regions of the slit adjacent to your object, and then subtract that as your sky (call this "subtraction along slit"). In the infrared, the night sky is so bright that residual flatfielding errors (i.e., incorrect match of slit illumination function; variation in spectrograph sensitivity along slit; spatial variation in OH emission; etc.) at the 0.1-1% level can mean that residual errors from this process can be significant.

So the procedure used in the infrared is to nod your target along the spectrograph slit (or for extended objects, to nod the telescope to blank sky) so that you can perform sky subtraction directly (call this "direct sky subtraction"). This adds sqrt(2) more photon noise to your data, but usually results in greatly reduced systematic errors, because you observe with exactly the same instrumental set-up and through exactly the same optics and pixels for both object and sky. Unfortunately, because you don't observe the sky and object simultaneously, variations in the OH emission as a function of time may mean that you still get residual errors in the sky subtraction. However these can be readily removed by applying a subtraction along slit subtraction step after doing the direct sky subtraction step.

Straightening spectra

Like all grism systems in focal reducers, there is signficant curvature of the IRIS2 spectral format. The format is also slightly tilted with respect to the detector rows. It will usually be necessary to 'straighten' your spectra (i.e. shift the columns of the image up or down slightly so that the dispersed light is aligned along detector rows) before extracting (this has the advantage that you need to extract from fewer rows along the detector, and so suffer from less sky noise). Otherwise, you can do a careful tracing of the aperture profile, and separately wavelength-calibrate your object and sky spectra before subtraction.

For data taken prior to September 2003, we provide the following images, which have been obtained by sliding a bright star up and down the slit, and taking observations. You can use your favourite data reduction package to track these spectra and derive a straightening function (if you are using the Figaro reduction package, you can use the 'sdist.dat' files below directly to straighten your images).

J/Jlong/Jshort Jdist_1_020727.fits Jdist_1_020727.sdf Jdistortion_020727.sdist
K/Ks Kdist_1_020727.fits Kdist_1_020727.sdf Kdistortion_020727.sdist

In September 2003 the detector and grisms were removed and remounted, so the spectra are now tilted at a different angle. Similar distortion data have yet to be acquired.

Object Extraction

Object extraction works almost identically to the techniques typically used in the optical. There are lots of ways to do this.
  • You can assume your object is straight on the detector and just extract a fixed number of rows (generally not a good assumption).
  • You can 'straighten' your spectra (recommended for IRIS2 - see above), and then extract a fixed number of rows.
  • You can 'straighten', then apply any number of 'optimal extraction' algorithms which will weight the flux extracted from each row by the brightness of the object in that row.

Which of these you choose to do, and how much effort you spend on it, will depend somewhat on your application.

Return to the main IRIS2 page.

These pages contain information on the functionality of the IRIS2 Infrared Imager and Spectrograph. Pages maintained by Angel Lopez-Sanchez (alopez -@- aao.gov.au) and Stuart Ryder (sdr -@- aao.gov.au). Page last modified by Stuart Ryder.