How To Design IRS Spectral Maps


The IRS instrument offers a powerful form of spectral imaging known as "Spectral Mapping Mode", in which the telescope moves by small steps through a grid of positions, stopping at each step to collect one or more spectral exposures. Large regions can be mapped to relatively faint surface brightnesses, providing 3D spectral cubes covering extended objects. This document offers some suggestions for designing IRS spectral maps from which accurate flux intensities and spectral images of extended sources can be obtained. Sparse spectral maps, which can be considered equivalent to a collection of non-contiguous spectra, are not covered here. For more general information on the IRS and planning IRS observations, see the Observing Planning Cookbook. An example Short-Low nuclear and Long-Low strip map on the nearby galaxy M51 from the SINGS Legacy program is shown at right.


Basics

The basic set of parameters which specify an IRS spectral map are found in the Instrument Settings of a mapping AOR in Spot. They are:
Ramp Duration:
The duration in seconds of each exposure (BCD) obtained at each mapping position.
Cycles:
The number of exposures (BCDs) to obtain at each position before moving to the next position.
Pointings Parallel:
The number of positions in the mapping grid parallel to the long dimension of the slit.
Step Size Parallel:
The size in arcsec of steps (if any) along the long dimension of the slit.
Pointings Perpendicular:
The number of positions in the mapping grid perpendicular to the long dimension of the slit, (i.e. along the short, wavelength-dispersed dimension).
Step Size Perpendicular:
The size in arcsec of steps along along the short dimension of the slit.
Map Cycles:
The number of times to cycle through the entire map grid.

Note that spectral maps are always performed in a coordinate frame parallel and perpendicular to the slit. It is not possible to map at, e.g., 45 degrees to the slit.


Designing Maps

The mechanics of laying out spectral maps using Spot, and overlaying the slit positions on images, are covered in the Observing Planning Cookbook's mapping section. These details will not be repeated here. Instead, recommendations for selecting appropriate mapping parameters to give useful final spectral cubes will be offered.

Quick Summary (each point is discussed in detail, below)

  1. We recommend that you visit each piece of sky with at least two and preferably four different instrument pixels.
  2. The recommended step size is 1/2 a slit width in either direction.
  3. Planning for robust background subtraction is critical for achieving good performance. We recommend a dedicated background observation for all spectral maps.
  4. Remember that coverage geometry is a function of ecliptic latitude, since the slits rotate as a function of time, by differing amounts.

Map Parameters

The IRS slits are very small, so combining spectra from multiple positions into spectral cubes is limited by the accuracy of the available pointing information. The slits are imaged to two detector pixels, yet the PSF delivered by the telescope to the slit plane is fully sampled only at the long wavelength end of each module order. Since individual IRS modules cover more than a factor of two in wavelength with a single slit, diffraction out of and into the slit can have a strong, wavelength-dependent impact on the extended flux calibration, especially for targets with compact structures at high spatial resolution. The best method to minimize the impact of these issues is to design a spectral map with adequate redundancy. At a minimum, each position should be visited by some portion of the IRS slit at least twice.

A very general rule of thumb, if there are no other constraints, is to construct maps with approximately 1/2 slit length and 1/2 slit width steps.

The recommended step sizes are:

Recommended step sizes
Step Size
ModuleParallelPerpendicular
Short-Low (SL) 26" 1.85"
Short-High (SH) 5" 2.3"
Long-Low (LL) 79" 5.1"
Long-High (LH) 10" 4.5"

Note that the recommended steps are slightly less than 1/2 the slit dimensions in each direction, to provide redundancy and safely avoid the edges of the slits where flat fielding errors are largest. If your mapping observation requires the use larger steps (at the cost of redundancy), it is important to maintain the factor of two redundancy in the perpendicular direction.

Peak-Up

Peak-Up is typically not required for spectral mapping observations, unless you don't know the position of your desired target center, and wish to peak-up directly on the target. The spacecraft pointing is good enough, and the reconstructed star-tracker and gyroscope positions accurate enough, that simply giving accurate coordinates for the map center is sufficient.

Redundancy

Redundancy in a spectral map helps mitigate the under-sampling and diffraction issues mentioned. It also serves to sample the same portion of the sky on different parts of the IRS detector arrays, which can be of tremendous help for maps of faint, low surface intensity sources. This is particularly true for redundancy along the slit. Stepping 1/2 or even 1/3 of the slit length, even if spending 1/2 or 1/3 of the time at each position, will dramatically improve the quality of the map. Redundant maps offer the ability of detecting and masking out warm (often called "rogue") pixels directly from the maps themselves.

When designing maps, prioritize redundancy where possible. At least a factor of two perpendicular stepping redundancy is essential for the highest quality maps, which is why the maximum recommended step size is 1/2 of the slit width. Although very useful maps can be made without any parallel (along the slit) redundancy (as is done, for instance, in the SINGS Short-Low maps), this type of redundancy can improve the quality of faint object maps considerably.

For slit-parallel steps, the redundancy at each position will not be uniform over the entire map. Some useful formulae for calculating map sizes and areas of redundant coverage are given here. These apply to a single map, specifically for either Short-Low or Long-Low, with "n" parallel steps of size "d" arcseconds, a sub-slit length of "s" arcseconds, and a gap between sub-slits of "g" arcsecs. The dimension of the map in the slit-parallel direction has the following properties:

Length of maximum coverage in either sub-slit (in arcsec):
(n-1) d + 2s + g
Length over which there is coverage in both sub-slits (for d < s):
(n-1) d - g
Length over which there are at least two visits per position per slit (for d < s/2):
(n-3) d - g
This layout is illustrated below:

Full spectral coverage

Often it is desirable to map the same area with more than one sub-slit or module. The Long-Low slit is roughly 10x the area of the Short-Low slit, and more than 100x the area of the smallest slit, Short-High. Obviously, mapping similar areas with all four modules can be prohibitive. Mapping LL+SL or LH+SH(+SL) is more common, and can be achieved in at least two different ways:
  1. The LL slit is roughly perpendicular to the SL slit in the Spitzer focal plane. For similar areal coverage between the modules, you can wait several months for the slits to rotate to match orientation, and conduct two similar maps to cover the same area. A similar method can be used for SH and LH (though these slits are much closer to square, making it much less useful). Note that this solution is most effective only for high ecliptic latitude targets.
  2. Two or more slits can be mapped in the same AOR (or grouped AORs) in such a way as to maximize the usable area of overlap.

The second approach, which takes advantage of the orientation of the slits in the focal plane to construct two or more perpendicular, yet overlapping maps, is usually preferred. An example of this for the Long-Low and Short-Low slits is illustrated above in the case of M51. Although the Short-Low map is small in this case, the general idea is the same. Here, the "Low Long Both" option was chosen with enough perpendicular steps to cover the length of the Short-Low slit. This map produces full Long-Low and Short-Low coverage around the nucleus. A larger Short-Low map (more perpendicular steps) would increase the area covered by both slits. Obviously, since the Short-Low and Long-Low slits are not precisely perpendicular, the maps will not have full overlap at large radii.

For an 1xn Short-Low map, 11 Long-Low (perpendicular) steps of five arsec each will suffice to cover one of the sub-slits. The "Both" mapping method treats low-resolution IRS slits as a single full slit. The center of the full slit is placed at each mapping grid position. Since the central ~25" of each slit is obscured, this method requires enough slit-parallel steps to achieve uniform coverage in both sub-slits, and at least one slit-parallel step to obtain any coverage on the targeted position. This type of map leads to along-the-slit extensions on either side of the main mapped area with coverage in a single sub-slit only. Rather than being wasted, these areas are often very useful for background estimation (see below).

The high-resolution slits are small enough and close enough to square that it is not difficult to create grouped maps in Short-Hight and Long-High which cover roughly the same area. Typically 4 times as many Short-High steps are needed to cover the length of the Long-High slit.

Cycle and exposure time

The number of cycles obtained at each map position is not critical. Whether, e.g., you use four, 30sec Long Low exposures, or a single 120 sec exposure is a matter of taste, data volume, and increased BCD failure tolerance (at the expense of some overhead). While spacecraft drift was a concern for some early Cycle-1 maps, recent modifications to the pointing system have virtually eliminated this as an issue, and, with well-sampled maps, any combination of ramp duration, cycles, and map cycles which achieves sufficient depth per pixel should be acceptable.

In terms of mapping efficiency, overhead can be a concern when mapping bright sources. Since it takes roughly 6-8secs for Spitzer to offset position, finish the conditioning frames, and settle before another map exposure can begin, using the 6 sec ramp time, unless necessary to avoid saturation, can be very wasteful. For a Short-Low map, with 20% or less additional total AOR time, you can increase the BCD exposure time per position by a factor of 2.3 simply by choosing the 14sec ramp time, instead of the 6sec ramp time. Use Spot to try various combinations and pay close attention to the exposure vs. total observing time estimates.

In order to calculate the required exposure per pointing for sources of a given surface brightness, observers should use SPECPET and assume 3x3 detector pixels correspond to a point source, in order to compute expected S/N for a given flux intensity in MJy/sr.

Very Large Maps

The maximum IRS AOR duration is six hours. Mapping large regions (e.g., larger than 30 square arcmin) typically requires more mapping steps than this AOR limit will allow. To achieve large maps in this case, two or more mapping AORs can be tiled to cover a large region. The IRS slits can rotate up to 1 degree per day, and since Spitzer cannot roll, you are forced to live with this rotation. For a 6hr map, the differential roll is small enough that no gaps open up in a properly designed map.

However, scheduling a tiled group of AORs is complicated by the fact that all mapping AORs rotate about their geometric center. Unfortunately, a group of spectral mapping AORs cannot be made to rotate about a common center. Therefore, even if all AORs are grouped within a time period which limits their relative slit position angles to a degree or less, large gaps can open up within the tiled AORs, depending on the absolute roll angle of the slit at the date of observation. There is no simple solution to this problem, but timing constraints may help.

One option is to place absolute constraints on the date(s) of observation, consulting the BIC to pick a date within an IRS campaign. For coordinate positions for which the slit rotates quickly (near the ecliptic poles) a one day absolute constraint is required. For lower ecliptic latitude observations, this can be relaxed somewhat. Visualize the group of AORs in Spot, and vary them about the chosen day by 1-3 days to see how tight of a constraint is required to avoid gaps in the map. Remember that the slit footprint visualized in Spot is slightly larger than the usable slit area.

If you are planning a very large map requiring two or more grouped AORs, you should consult Observer Support through the Help Desk for more information on scheduling your observations.

An example large map created with 4 grouped SL mapping AOR, comprised of over 1200 total individual pointings is shown below. Sample AORs which produce a square area of uniform coverage if executed on 5 Jan 2005 can be found here.


Backgrounds

The best scientific results are obtained from IRS spectra when a proper background is subtracted at the BCD-level from each position in the map. The background exposures should be taken as close in time and position to the primary target as is possible. Not only does this remove the astrophysical foreground emission (zodiacal and cirrus), it eliminates many systematic artifacts which exist at or even above the level of the background in unsubtracted frames, and mitigates the effects of time-varying warm pixels (an especially significant factor in the LL and LH modules). While staring mode observations of point sources with the Short-Low and Long-Low slits have a built in background observation, since each science target is placed at two nod positions for every AOR, you must explicitly obtain your own background exposures for all mapping mode observations. High-res staring observations also use two nods per slit, but the nods are very close together and there is significant PSF overlap. Care must be taken to avoid the outer regions of extended science targets when selecting regions for the background. The best method for deciding which type of background to use or where to place background observations is overlaying AORs directly on an available Spitzer IRAC or MIPS image within Spot.

Dedicated Backgrounds

The recommended way of obtaining background frames is to create an additional dedicated background AOR which maps adjacent dark sky. Either mapping or staring mode can be used for this. A typical format might be a 1x2 map with 2 cycles at each position. Remember that, for the low-resolution modules, background is obtained in both subslits (e.g. SL1 and SL2) at once.

The total time to spend on the background depends on the signal quality requirements of your program. Typically, spending 10% or less of the total map time on the background is sufficient: this is true for high-resolution as well as low-resolution maps. Even a single background frame can radically improve the data quality, but five or more frames are recommended. In all cases, use the same ramp time for the background exposures as for the main map. If there is a large gradient in the background expected at your position, consider using cluster mode to group backgrounds in an AOR on either side of the target. Be sure to keep the full slit clear of any source emission at all permitted rotations.

Outrigger Backgrounds

For sources which are small enough, and which will be mapped sequentially with both orders of a low-resolution module, you can use the non-primary, or "outrigger" order to obtain an accurate background spectrum without spending any time on dedicated observations. Since, while, e.g. data is being collected in SL order 1, SL order 2 is also integrating 70" away, this is often enough separation to provide clean background data. Assuming both orders of the module are being used to map the primary target, often no additional observations will be required.

When using the "Both" method of mapping, in which both subslits are treated as a single larger slit, to obtain matched coverage in both sub-slits over a given area of interest, one sub-slit will extend much further in one direction along the slit, and the other will extend a similar amount in the opposite direction. Often 1, 2, or even 3 "columns" of a map designed this way can be useful for background frames. Note that the background for the two spectral orders will come from opposite regions centered on the mapping target. An example AOR with marginally usable Short-Low outrigger background frames (purple and yellow) is shown here:

The outrigger method does not work for high-resolution modules (which have only a single sub-slit). If in doubt, include a grouped, dedicated background. This can never hurt, and is typically quite inexpensive for large maps. If outrigger background data is, in the end, usable, it can be combined with the dedicated exposures to improve the background signal even further, or look for nonuniformity in the background across the map. In all cases, it is very important to visualize your AOR within SPOT to make sure you are getting well off your science target with all IRS slits of interest.

Archive Backgrounds

When all else fails, e.g. for working with older spectral maps from the archive which offer no in situ backgrounds, you may be able to obtain useful background data directly from the Spitzer archive. Often this means waiting up to a year after the data of interest were first placed in the archive, so that data from the same instrument campaign are freely available.

The basic approach is to search for staring mode data (typically) which were obtained within a few days of the observation, and within a small range of ecliptic latitude of the mapped target, using the same instrument module and (ideally) the same ramp exposure time. Because of the time-varying nature of the warm pixels in the Long-Low and Long-High modules, background observations taken very nearby in time to the science data are best. This method should be considered as a last resort for dealing with poorly designed maps, and we strongly recommend obtaining your own background data as part of your IRS mapping AOR via one or both of the methods outlined above.


Data Reduction and Flux Calibration

The CUBISM tool, which is a SINGS Legacy team deliverable and will be made available and supported through the SSC in Spring 2006, is designed to reconstruct and extract full spectral cubes from well-designed spectral mapping data sets. It relies on several extended-source specific calibrations and corrections provided by the SSC.


Acknowledgements

The advice presented on this page was made possible through the work of a number of Spitzer observers associated with the SINGS Legacy project (R. Kennicutt, PI). The primary author of this document was J.D. Smith (University of Arizona). Other major contributors include D.A. Dale (University of Wyoming), L. Armus, G. Helou, K. Sheth (SSC), H. Roussel (MPIA), T.H. Jarrett, and B.A. Buckalew (SSC).