IRS Data in the Spitzer Archive (17 June 2004) A number of Guaranteed Time, Legacy, and Early Release observations have now been processed with the S10 software and deposited in the Spitzer Data Archive. While the data are generally of very high quality, there are a number of features which are apparent, and which require some explanation for observers not familiar with IRS data. Order Mis-matches between SL and LL ----------------------------------- In some cases observers may see "jumps" in flux between the SL and LL spectral orders for a source observed with both modules. These jumps are typically less than 10%, once background emission from zodiacal dust and cirrus have been removed. The mismatch can be higher before the background correction, depending on the local background levels, due to the different slit sizes. We are currently investigating the cause of mismatches remaining after background corrections, which may have contributions from both pointing offsets in SL, and a known software error in how signal ramps are fit, that lead to errors in the computed total flux incident on the array. The errors are present in data from all of the arrays, greatest at the shortest integration times where the charge current may be overestimated by ~25%. The errors will be corrected together with the affected flux calibrations over the next few weeks. Horizontal Stripes in SL data ----------------------------- Horizontal stripes (of reduced signal) which stretch from the peakup areas through the SL spectra may be present in some SL BCD products. When present, the stripes can be seen in the "bcd_fp.fits" product, and manifest themselves as unexpected absorption features in the spectral extractions (1D data). The stripes are caused by an over-subtraction of the peakup stray light (see ch. 7 of the SOM) at the locations of the spectral orders. Observers should examine the 2D BCDs in all cases, but especially when an unexpected absorption feature is seen in the 1D spectra. The SL straylight removal algorithm uses a mask file and bad pixel map which we are continuously updating at the SSC. However, in cases of severe stray light or new bad pixels, the correction can go awry. In all cases, we provide both corrected ("bcd_fp.fits") and un-corrected 2D ("f2ap.fits") data, and provide extractions from both products ("spect.tbl" and "spect2.tbl"). Observers may refer to http://ssc.spitzer.caltech.edu/irs/products/sample.html for more information. The upcoming release of SPICE (Spitzer IRS Custom Extractor) will allow observers to extract 1D spectra from BCDs if the extractions provided by the SSC at this stage do not meet the observers' science needs. Absolute Flux Calibration ------------------------- The IRS uses narrow slits and is spatially undersampled (except at the longest wavelengths of each order). In addition, the IRS pipeline is designed to work on point source data. All spectral calibrations are based upon matching point source science data to point source calibration data. For a complete description of the calibrations, please see ch. 7 of the SOM. Absolute spectrophotometry with the IRS is therefore very difficult, even in the case of point source science targets. However, the IRS detectors are very stable, showing deviations of less than 1% from campaign to campaign. Standard star flux calibrations are generally better than 5% from campaign to campaign, but there are some exceptions. (These measurements are affected by the same issues outlined above under low-resolution order mis-matches.) At this stage, the overall photometric uncertainties of the IRS are 30% in SH and LH data, and 20% in SL and LL data. We expect the absolute photometry of the IRS 1D spectral data to improve with future pipeline releases, but we recommend all observers carefully check their spectra against known photometric values and scale where appropriate, as they would for all narrow-slit spectra taken with any other ground- or space-based instrument. Background Subtraction ---------------------- The mid-infrared background, predominantly from zodiacal light at IRS wavelengths, can contribute significantly to observations of faint targets with the IRS. The arrays are sufficiently sensitive to detect levels of 20 MJy/sr or more at any wavelength. Therefore, we recommend that observers examine their BCD data and use the background estimation tool in SPOT to gauge the background levels in order to weigh these against the science goals. Background subtraction should be performed on IRS SL and LL data before making any absolute photometric scaling (see above). In Staring mode, all science targets are placed at two nod positions along the IRS slits, so this background subtraction can be accomplished by differencing the nod positions at the BCD level. Alternatively, the accompanying subslit (e.g., SL2 for SL1 spectra) can also be used for background subtraction at the BCD level. This cannot be done with SH or LH data, due to the sizes of the slits and PSFs. It is also important to realize that the local background may not be suited to using the nods or alternate slit data for correction of zodiacal light if, for example, the area is crowded or contains extended emission around the science target. In addition, the AOR itself may have been designed by the original observer in such a way that data suited to background correction were not acquired. These are the principal reasons why automated background subtraction cannot be reliably performed by the SSC pipeline. If background corrections can be made by the observer with the SL or LL BCDs, then the observer will also have to perform the 1D spectral extractions. See the upcoming release of SPICE for details on how to extract your own 1D spectrum. SPICE will also allow observers to specify a region for an extraction of a background 1D spectrum, which can then be subtracted from the 1D pipeline products without doing any subtraction at the 2D level. Warm/bad Pixels and NaNs in IRS data ----------------------------------- All four IRS arrays were subject to the damaging effects of solar flares during the In Orbit Checkout phase, which deposited in only two days an equivalent dose of protons expected over 2.5 years. This damaged 1% of the SL and SH pixels, and 4% of the LL and LH pixels (mainly the LH pixels). These pixels are now masked from data processing, by assigning "Not a Number" (NaN) status to them. This prevents numerical operations from being affected by their presence. Assigning any other real number would lead to improper processing of the surrounding signal information; for example, during the extraction process. Since signal information has been lost from these pixels as a result of the solar activity, an extraction element containing a damaged (NaN-ed) pixel will obviously contain lower signal than expected, and will appear as an outlier below the continuum (if a continuum is present). These outliers will be identified by a certain range of values in the 5th column of the extraction (1D spectra) tables. The SSC is investigating means of approximating (via interpolation) the signals of the damaged pixels through use of adjacent, unaffected pixels.