6.2. Spectroscopy

The three science cases we consider here are: point sources, point-like/semi-extended sources, and extended sources. The three types of observation, i.e. AOTs, that we consider are: pointed, mapping, and tiling. The differences between mapping and tiling in terms of the raster design are mapped out in Table 2.4, how to identify a pointed versus a mapping versus a tiling observation is given in the associated text and associated Table 2.5, and the full list of products provided for the various AOTs is given in Table 2.3.

The cubes or tables to use for inspection or for science for the combination of the three science cases and the three observing modes mentioned above are listed in Table 6.1. The product to use for "inspection" is that which gives the easiest access to immediate visualiation of the target, the product to use for science is either the same or that on which the scientist can run recommended post-pipeline processing tasks. Note that it is a fine line whether one or another product is the best to use, and sometimes it really is a matter of personal preference.

Please do read Section 3.2.5 for general comments/warnings applicable to all the science cubes discussed in this chapter.

Table 6.1. The recommended products for different science case and AOT type, for science and/or just for inspection.

  Pointed AOTs Mapping AOTs Tiling AOTs
Point sources Science spectrum tables or rebinned cubes see Section 6.2.1 see Section 6.2.1
Point sources Inspection interpolated cubes projected or drizzled cubes interpolated cubes
Semi-extended sources Science rebinned cubes see Section 6.2.2 see Section 6.2.2
Semi-extended sources Inspection interpolated cubes projected or drizzled cubes interpolated cubes
Extended sources Science interpolated cubes projected or drizzled cubes interpolated cubes
Extended sources Inspection interpolated cubes projected or drizzled cubes interpolated cubes

The spectrum tables to be used for point sources can be found in the Level 2/2.5/3 and as standalone browse products. The cubes at Level 2 and 2.5 have the non-equidistant wavelength grid and those provided as standalone browse product have an equidistant grid (this is explained in Section 5.3.2), but otherwise they are the same. For easy viewing of a cube outside of HIPE, the standalone cubes are the ones to use, as they are easier to load into other software. For use in HIPE, for line scans it does not really matter but for range scans the non-equidistant cubes from Level 2 or 2.5 are to be preferred (for the reason why, see Section 5.3.2).

Where a Level 2.5 is available, these cubes or tables are the ones to use: but this level is present only for unchopped range scan observations. For the spectrum tables, those of Level 3, if present, are the most convenient to use.

Table 6.2. Product name key for Table 6.1.

Product names @L2 @L2.5 @L3 Standalone Browse Product
Spectrum table HPSSPEC[R|B] HPSSPECBS[R|B] HPSSPEC HPSSPEC if present, else HPSSPECBS[R|B] if present, else HPSSPEC[R|B]
Rebinned cubes HPS3DR[R|B] HPS3DRBS[R|B]    
Interpolated cubes HPS3DI[R|B] HPS3DIBS[R|B]   HPS3DEQIBS[R|B] if present, else HPS3DEQI[R|B]
Projected cubes HPS3DP[R|B] HPS3DPBS[R|B]   HPS3DEQPBS[R|B] if present, else HPS3DEQP[R|B]
Drizzled cubes HPS3DD[R|B] HPS3DDBS[R|B]   HPS3DEQDBS[R|B] if present, else HPS3DEQD[R|B]

6.2.1. Point sources

The task extractCentralSpectrum can be used to extract a fully-calibrated point-source spectrum. The requirements are:

  • All of the flux must come from the source itself, not e.g. also from any background.

  • The task needs to be run on a rebinned cube, which normally means you are working with a pointed observation; for a mapping observations, it is necessary to locate the rebinned cube in the raster in which the source is the most centrally located.

  • The source should be centred somewhere within the central spaxel.

If you are sure these conditions are met for your target, the point-source spectrum table (HPSSPECR/B at Level 2, HPSSPECBSR/B at Level 2.5, or HPSSPEC at Level 3), which contain the outputs of this task, can be used to get the spectrum of your point source. There are up to four spectra in the spectrum table: the first is the spectrum of the central spaxel and is not point-source corrected. The other two to three are the spectra to use.

ExtractCentralSpectrum works on the rebinned cubes. It removes the extended source correction (ESC: which is part of the pipeline flux calibration) and produces three point-source corrected spectra, called c1, c9, and c129 in the language of the pipeline scripts. This task and its output are explained in chp. 8 of the PDRG, and the key between the "c1/9/129" name and the column titles in the spectrum tables can be found in Section 5.3.4. For pointed chop-nod observations, any of the three spectra provided in this table can be used—c1, c9, c129—with c129 being the recommended as it normally has the best SNR and where there was slight mispointing or pointing jitter during the observation, the flux levels will be more correct. For pointed unchopped observations two spectra are provided—c1 and c9—with c9 being the recommended unless its SNR is the lower.

If you wish to rerun extractCentralSpectrum—for example to see the diagnostic information it produces, or to use for semi-extended sources (see below)—this has to be done on the rebinned cubes.

To obtain a fully-calibrated point-source spectrum from mapping or tiling observations, it is necessary to run the task extractCentralSpectrum on the rebinned cube in the raster in which the point source is centrally located. To only add up the flux in an aperture will not produce the correct spectrum, since (i) the ESC will not have been removed and (ii) the uneven illumination of the PACS FoV will not have been corrected for.

For non centrally-located point sources and/or to run an interactive pipeline script which applies a more advanced flux correction for point sources (the "pointing offset correction script" which is for chop-nod pointed observations only), read the advice in chp. 8 and chp. 2 of the PDRG.

To inspect the cubes, to see e.g. where your point source is, use the interpolated cubes for pointed or tiling observations, and for mapping observations use the projected or drizzled cubes—which, is a matter of presonal preference.

Which level to take the table or cube from?

  • If there is a Level 3 spectrum table, take that (it is the combined table of all obsids that were taken to cover the full SED of the source). If there is a level 2.5 present, take that (these are provided only for unchopped range scan AOTs). Otherwise take the Level 2.

  • Alternatively, take the same tables from the browseProduct (Chapter 5) "directory" in the observation (either from within HIPE or from disc: Section 2.3.5)

  • If using the rebinned cube, take it from Level 2.5 if present, otherwise from Level 2.

6.2.2. Semi-extended sources

For semi-extended sources—those located fully within the central 3x3 spaxels of the rebinned cubes (with a practical limit of a diameter of about 15")— and which are mostly centred in the FoV, to recover the total flux from the source from pointed observations you can use a set of tasks in HIPE which: undo the extended source correction, extract a point-source corrected spectrum, and then apply a correction based on the difference between the model of the extended source's surface bright distribution and the flux of a point source. The recommended tasks to use are extractCentralSpectrum followed by specExtendedToPointCorrection; for cases of off-centred semi-extended sources (i.e. which are not centred in the centre of the FoV) an alternative to extractCentralSpectrum is possible. See chp. 8 of the PDRG for more information on this set of tasks: which outputs of extractCentralSpectrum to use, and in particular for advice on using it for sources which are either not fully or not at all contained within the central spaxel. The main requirements are the following:

  • All of the flux must come from the source itself, not e.g. also from any background.

  • The task needs to be run on a rebinned cube, which normally means you are working with a pointed observation; for a mapping observations, it is necessary to locate the rebinned cube in the raster in which the source is the most centrally located.

  • A surface brightness distribution model or an image of the source must be provided.

  • The source should be located in the central 3x3 region and that the surface brightness model also takes into account where, in this central part of the FoV, the source is placed; if it is not centred within the central spaxel, it is recommended to use the "3x3" output of the task extractCentralSpectrum, rather than that from the central spaxel-spectrum only.

To only add up the flux from the source will not produce the correct source spectrum, since (i) the ESC will not have been removed and (ii) the uneven illumination of the PACS FoV will not have been corrected for.

To inspect the cubes use the interpolated cubes for pointed or tiling observations, and for mapping observations use the projected or drizzled cubes.

6.2.3. Flat extended sources

For sources that are extended such that the slope from spaxel to spaxel is gradual (no more than 20% within a single 47" square FoV) and the source is larger than (and so fully encompassed in) a single PACS footprint (47" square), the total flux from the source (included in the total FoV) can be obtained by integrating over the cube spatially. Use the interpolated cubes for tiling observations, and the projected or drizzled for mapping observations. For pointed observations either the interpolated or rebinned cubes can be used, however note that by definition a fully-extended source is not fully captured by a pointed observation and the flux you measure is limited spatially by the field covered.

For extended sources that have structure within a single footprint, e.g. which are "clumpy" or have a strong gradient, the same cubes can be used, but if you simply sum up over the field the measured flux will be incorrect, since it is not corrected for the uneven illumination of the PACS IFU (see Section 6.2.4 for an explanation of this).

Which level to take the table or cube from?

  • If there is a level 2.5 present, take that (these are provided only for unchopped range scan AOTs). Otherwise take the Level 2.

  • For the projected, interpolated, or drizzled equidistant cubes, you can also use the equidistant version provided in the the standalone browse product (Chapter 5) "directory" of the observation, either from within HIPE or from disc (Section 2.3.5).

6.2.4. Other extended sources, crowded fields, off-centred point sources, and semi-extended sources: correcting for the uneven illumination of the PACS IFU

The spaxels of the PACS IFU are not evenly illuminated, with the result that there is effectively some flux loss between the spaxels over most of the PACS wavelength range. To correct for this an extended source correction was created, and this is applied to all observations by the pipeline. This fully corrects the integrated fluxes for all fully-extended sources. For point sources sources, this correction is taken out before the point source corrections are applied, and the fluxes will still be correct. This also applies to semi-extended sources if using the extended-to-point correction in HIPE. However, the uneven illumination of the PACS IFU FoV affects the flux distribution in the rebinned cubes and the spectral calibration in surface brightness. This may result in an incorrect integrated spectrum extracted from an aperture in spectral cubes that contain:

  • crowded fields

  • off-centred point sources

  • semi-extended structures

  • sources which are extended but with steep flux gradients

The inaccuracy in the flux will depend on the morphology of the source at the wavelength of your observation and its coupling to the detector's beam efficiencies on every pointing (mapping) pattern. If you do aperture photometry (with any sized aperture) you will not recover the correct flux, and how incorrect the result is depends on the morphology of the source and its coupling to the spaxels' pointing pattern on the sky.

To estimate the inaccuracy in the flux, if you know (or can model/estimate) the surface brightness distribution of the source (i.e. its morphology at the wavelength of your observation), you can apply a "forward projection" in HIPE: this takes in your input surface brightness distribution and a model spectrum, and working with the pointing/mapping pattern of your observation, it produces result which folds in the uneven illumination. You can then compare the modelled result to your observed result. Scripts and explanation for this tool will be provided on the PACS documentation and the Herschel Explanatory Legacy Library pages, see herschel.esac.esa.int/twiki/bin/view/Public/PacsCalibrationWeb. These tasks are also explained in chp. 8 of the PDRG and chp. 9 of the PDRG, however the script and the release notes for the FMT can only be found on the HSC web-pages.