# Description¶

The master background subtraction step subtracts background signal from 2-D spectroscopic data using a 1-D master background spectrum. The 1-D master background spectrum is created from one or more input exposures, or can alternatively be supplied by the user. The 1-D background spectrum - surface brightness versus wavelength - is projected into the 2-D space of source data based on the wavelength of each pixel in the 2-D data. The resulting 2-D background signal is then subtracted directly from the 2-D source data.

Logic built into the step checks to see if the exposure-based background subtraction step in the calwebb_spec2 pipeline has already been performed on the input images, based on the value of the S_BKDSUB keyword. If S_BKGSUB is set to “COMPLETE”, the master background step is skipped. If the calwebb_spec2 background step was not applied, the master background step will proceed. The user can override this logic, if desired, by setting the step argument --force_subtract to True, in which case master background subtraction will be applied regardless of the value of S_BKDSUB (see Step Arguments).

Upon successful completion of the step, the S_MSBSUB keyword is set to “COMPLETE” in the output product. The background-subtracted results are returned as a new data model, leaving the input model unchanged.

## Inputs¶

The primary driver of the master background step is a “spec3” type Association (ASN) file or a ModelContainer data model populated from a “spec3” ASN file. This is the same ASN file used as input to the calwebb_spec3 pipeline, which defines a stage 3 combined product and its input members. The list of input members includes both “science” and “background” exposure types. The master background subtraction step uses the input members designated with "exptype": "background" to create the master background spectrum (see example_asn1). These need to be x1d products created from individual exposures at the end of the calwebb_spec2 pipeline, containing spectra of background regions. The master background signal will be subtracted from all input members designated as "exptype": "science" in the ASN, resulting in a new version of each science input. These inputs need to be cal products created from individual exposures by the calwebb_spec2 pipeline.

There are two main observing scenarios that are supported by this step: nodded exposures of point sources and off-source background exposures of extended targets. A third type of operation is performed for NIRSpec MOS observations that include background slits. The details for each mode are explained below.

### Nodded Point Sources¶

If an observation uses a nodding type dither pattern to move a small or point-like source within the field-of-view, it is assumed that part of the field-of-view in each exposure is also suitable for measuring background. Exposures of this type are identified by the pipeline based on their “PATTTYPE” (primary dither pattern type) keyword value. The value will either contain the substring “NOD” somewhere within the name (e.g. “2-POINT-NOD” or “ALONG-SLIT-NOD”), or will be set to “POINT-SOURCE” (for MIRI MRS). The calwebb_spec2 srctype step recognizes these PATTTYPE values and sets the source type to “POINT.”

This in turn causes the extract_1d step at the end of calwebb_spec2 to extract spectra for both source and background regions. For IFU exposures the background region is typically an annulus that is concentric with a circular source region. For slit-like modes, one or more background regions can be defined in the extract1d reference file, flanking the central source region. In both cases, the extraction regions are centered within the image/cube at the RA/Dec of the target. Hence for nodded exposures, the location of the extraction regions follows the movement of the source in each exposure. The extracted data from the source region are stored in the “FLUX” and “SURF_BRIGHT” (surface brightness) columns of the x1d product, while the background extraction is stored in the “BACKGROUND” column. The master_background step recognizes when it’s working with nodded exposures and in that case uses the data from the “BACKGROUND” column of each background x1d product to create the 1-D master background spectrum.

Below is an example ASN file for a simple 2-point nodded observation consisting of two exposures.

{
"asn_type": "spec3",
"asn_rule": "candidate_Asn_IFU",
"program": "00626",
"asn_id": "c1003",
"target": "t001",
"asn_pool": "jw00626_20190128T194403_pool",
"products": [
{"name": "jw00626-c1003_t001_nrs",
"members": [
{"expname": "jw00626009001_02101_00001_nrs1_cal.fits",
"exptype": "science",
"asn_candidate": "('c1003', 'background')"
},
{"expname": "jw00626009001_02102_00001_nrs1_cal.fits",
"exptype": "science",
"asn_candidate": "('c1003', 'background')"
},
{"expname": "jw00626009001_02101_00001_nrs1_x1d.fits",
"exptype": "background",
"asn_candidate": "('c1003', 'background')"
},
{"expname": "jw00626009001_02102_00001_nrs1_x1d.fits",
"exptype": "background",
"asn_candidate": "('c1003', 'background')"
}
]
}
]
}


As you can see, the same two exposures are defined as being both “science” and “background” members, because they both contain the target of interest and a region of background. The “science” members, which are the cal products created by the calwebb_spec2 pipeline, are the data files that will have the master background subtraction applied, while the “background” members are the x1d spectral products from which the master background spectrum will be created. The combined master background spectrum will be subtracted from each of the two science exposures.

### Extended Source with Dedicated Background Exposures¶

Observations of extended sources must obtain exposures of a separate background target/field in order to measure the background. Exposures of a background target are identified by the keyword “BKGDTARG” set to True in the header. During calwebb_spec2 processing, the srctype step recognizes these and sets their source type to “EXTENDED”, because all dedicated background exposures are to be processed as extended sources.

This in turn causes the extract_1d step at the end of calwebb_spec2 to extract a spectrum in extended source mode, which uses the entire field-of-view (whether it be a slit image or an IFU cube) as the extraction region. The extracted spectral data are stored in the “FLUX” and “SURF_BRIGHT” columns of the resulting x1d product, with the “BACKGROUND” column left blank. The master_background step recognizes when it’s working with a background exposure, which is always treated as an extended source, and in that case uses the data from the “SURF_BRIGHT” column of each background x1d product to construct the master background spectrum.

Below is an example ASN file for an extended source observation that includes background target exposures, using a 2-point dither for both the science and background targets.

{
"asn_type": "spec3",
"asn_rule": "candidate_Asn_IFU",
"program": "00626",
"asn_id": "c1004",
"target": "t002",
"asn_pool": "jw00626_20190128T194403_pool",
"products": [
{"name": "jw00626-c1004_t002_nrs",
"members": [
{"expname": "jw00626009001_02101_00001_nrs1_cal.fits",
"exptype": "science",
"asn_candidate": "('c1004', 'background')"
},
{"expname": "jw00626009001_02102_00001_nrs1_cal.fits",
"exptype": "science",
"asn_candidate": "('c1004', 'background')"
},
{"expname": "jw00626009001_02103_00001_nrs1_x1d.fits",
"exptype": "background",
"asn_candidate": "('c1004', 'background')"
},
{"expname": "jw00626009001_02104_00001_nrs1_x1d.fits",
"exptype": "background",
"asn_candidate": "('c1004', 'background')"
}
]
}
]
}


In this example there are two exposures of the science target, labeled as “science” members, and two exposures of the background target, labeled as “background” members. As before, the science members use cal products as input and the background members use x1d products as input. The master background step will first combine the data from the two background members into a master background spectrum and then subtract it from each of the two science exposures.

### NIRSpec MOS with Background Slits¶

NIRSpec MOS exposures that have one or more slits defined as background require unique processing. The background slits take the place of both nodded and off-target background exposures, because the background can be measured directly from the spectra contained in those slits. Because the background spectra come from the same exposure as the source spectra, the master background processing is applied within the calwebb_spec2 pipeline as each individual exposure is calibrated, rather than during calwebb_spec3 processing. In this scenario, all source and background slits are first partially calibrated up through the extract_2d and srctype steps of calwebb_spec2 to produce 2D cutouts for each slit. At this point the master_background step is applied, which completes the remaining calibration steps for all of the background slits, resulting in multiple 1D extracted background spectra. The background spectra are combined (as with other observing modes) to create a 1D master background spectrum, and then the master background spectrum is interpolated back into the 2D space of each source slit and subtracted. The background-subtracted source slits then have all of their remaining calwebb_spec2 calibration steps applied.

Note that special corrections are applied to the 2D master background data before being subtracted from each source slit, as explained in detail below.

## Creating the 1-D Master Background Spectrum¶

The 1-D master background spectrum is created by combining data contained in the x1d products listed in the input ASN as "exptype": "background" members. As noted above, the background members can be exposures of dedicated background targets or can be a collection of exposures of a point-like source observed in a nod pattern.

For the case of dedicated background target exposures, the spectrum contained in the “SURF_BRIGHT” column of the background x1d products will be used for creating the master background spectrum. For the case of nodded exposures, the spectrum contained in the “BACKGROUND” column of the x1d products will be used. The data in both columns are in units of surface brightness, which is appropriate for eventually computing the 2-D background signal.

When all the input background spectra have been collected, they are combined using the combine_1d step to produce the 1-D master background spectrum. See the combine_1d step for more details on the processes used to create the combined spectrum.

## Subtracting the Master Background¶

The 1-D master background spectrum is interpolated by wavelength at each pixel of a 2-D source spectrum and subtracted from it. The source data instances can be, for example, a set of NIRSpec or MIRI IFU exposures, a set of NIRSpec MOS or fixed-slit 2-D extractions, or a set of nodded MIRI LRS fixed-slit exposures. The subtraction is performed on all data instances within all input science exposures. For example, if there are 3 NIRSpec fixed-slit exposures, each containing data from multiple slits, the subtraction is applied one-by-one to all slit instances in all exposures. For each data instance to be subtracted the following steps are performed:

• Compute a 2-D wavelength grid corresponding to the 2-D source data. For some observing modes, such as NIRSpec MOS and fixed-slit, a 2-D wavelength array is already computed and attached to the data in the calwebb_spec2 pipeline extract_2d step. If such a wavelength array is present, it is used. For modes that don’t have a 2-D wavelength array contained in the data, it is computed on the fly using the WCS object for each source data instance.

• Compute the background signal at each pixel in the 2-D wavelength grid by interpolating within the 1-D master background spectrum as a function of wavelength. Pixels in the 2-D source data with an undefined wavelength (e.g. wavelength array value of NaN) or a wavelength that is beyond the limits of the master background spectrum receive special handling. The interpolated background value is set to zero and a DQ flag of “DO_NOT_USE” is set.

• Subtract the resulting 2-D background image from the 2-D source data. DQ values from the 2-D background image are propagated into the DQ array of the subtracted science data.

## NIRSpec Background Corrections¶

Once the 1-D master background spectrum has been interpolated to the 2-D space of the science data, NIRSpec data can sometimes need additional corrections to make the computed background match the science data. This is due to two primary effects of NIRSpec calibration:

• Point sources in MOS and fixed-slit mode receive wavelength offset corrections if the source is not centered (in the dispersion direction) within the slit. Hence the wavelength grid assigned to the 2-D slit cutout is shifted slightly relative to the wavelengths of the background signal contained in the same cutout. And because the flat-field, pathloss, and photom corrections/calibrations are wavelength-dependent, the pixel-level calibrations for the source signal are slightly different than the background.

• Point sources and uniform sources receive different pathloss and bar shadow corrections (in fact point sources don’t receive any bar shadow correction). So the background signal contained within a calibrated point source cutout has received the wrong pathloss correction and hasn’t received any bar shadow correction. Meanwhile, the master background is created from data that had corrections for a uniform source applied to it and hence there’s a mismatch relative to the point source data.

The 2-D background that’s initially created from the 1-D master background is essentially a perfectly calibrated background signal. However, due to the effects mentioned above, the actual background signal contained within a calibrated point source slit (or IFU image) is not perfect (e.g. it still has the bar shadow effects in it). So all of these effects need to be accounted for in the computed 2-D background before subtracting from the source data.

### NIRSpec IFU Mode¶

For the NIRSpec IFU mode, the only effect that needs to be accounted for is the difference between point source and uniform source pathloss corrections, because no wavelength or bar shadow corrections are applied to IFU data. The 2-D background generated from the master background spectrum must have the uniform source pathloss correction removed from it and the point source pathloss correction applied to it, so the operation performed on the IFU 2-D background is:

$bkg(corr) = bkg * pathloss(uniform) / pathloss(point)$

### NIRSpec Fixed-Slit Mode¶

NIRSpec fixed slit data receive flat-field, pathloss, and photometric calibrations, all of which are wavelength-dependent, and the pathloss correction is also source type dependent. Fixed slit data do not receive a bar shadow correction. Only slits containing a point source can have a wavelength correction applied, to account for source centering within the slit, hence slits containing uniform sources receive the same flat-field and photometric calibrations as background spectra and therefore don’t require corrections for those two calibrations. Furthermore, the source position in the slit is only known for the primary slit in an exposure, so even if the secondary slits contain point sources, no wavelength correction can be applied, and therefore again the flat-field and photometric calibrations are the same as for background spectra. This means only the pathloss correction difference between uniform and point sources needs to be accounted for in the secondary slits.

So if the primary slit (as given by the FXD_SLIT keyword) contains a point source (as given by the SRCTYPE keyword) the corrections applied to the 2-D master background for that slit are:

$\begin{split}bkg(corr) = bkg &* [flatfield(uniform) / flatfield(point)]\\ &* [pathloss(uniform) / pathloss(point)]\\ &* [photom(point) / photom(uniform)]\end{split}$

For secondary slits that contain a point source, the corrections applied to the 2-D master background are simply:

$bkg(corr) = bkg * pathloss(uniform) / pathloss(point)$

### NIRSpec MOS Mode¶

Because the master background is subtracted from only partially calibrated slit data, the number of correction terms is reduced relative to that for fixed slits. At the time the 2D master background is subtracted from each source slit, those source slits have not yet received any calibrations that are specific to point sources and hence the background signal in the source slits has effects in it characteristic of uniform source data. Therefore the fully-calibrated 2D master background signal needs to have those uniform source effects imposed on it (e.g. impose the uniform source flat-field pattern or the bar shadow pattern into the background data), so that it matches the actual background signal in each slit before being subtracted. The corrections applied to the 2D master background for MOS slits are:

$\begin{split}bkg(corr) = bkg &* flatfield(uniform) * pathloss(uniform)\\ &* barshadow(uniform) / photom(uniform)\end{split}$