Deep Herschel PACS point spread functions
AAstronomy & Astrophysics manuscript no. paper˙PSFs˙final © ESO 2018November 5, 2018
Deep
Herschel (cid:63)
PACS point spread functions (cid:63)(cid:63) (Research Note)
M. Bocchio , S. Bianchi A. Abergel Institut dAstrophysique Spatiale (IAS), UMR8617, CNRS, Universit´e Paris Saclay, Universit´e Paris Sud, Orsay F-91405, FranceINAF - Osservatorio Astrofisico di Arcetri, Largo Enrico Fermi 5, 50125 Firenze, ItalyPreprint online version: November 5, 2018
ABSTRACT
The knowledge of the point spread function (PSF) of imaging instruments represents a fundamental requirement for astronomicalobservations. The
Herschel
PACS PSFs delivered by the instrument control centre are obtained from observations of the Vesta asteroid,which provides a characterisation of the central part and, therefore, excludes fainter features. In many cases, however, information onboth the core and wings of the PSFs is needed. With this aim, we combine Vesta and Mars dedicated observations and obtain PACSPSFs with an unprecedented dynamic range ( ∼ ) at slow and fast scan speeds for the three photometric bands. Key words.
Instrumentation: photometers, Techniques: image processing, Techniques: photometric
1. Introduction
The response of a given imaging instrument to a point sourceis known as the point spread function (PSF). In the case ofdi ff raction-limited space telescopes this quantity is dominatedby the configuration of the aperture and it is key to many aspectsof astrophysical observations. First, models are often comparedto observations. This operation is typically carried out by con-volving a given model to the PSF of the observed image. Anerror in the estimate of the PSF would lead to errors in the in-terpretation of the observations. Second, images taken with dif-ferent instruments or at di ff erent wavelength bands intrinsicallyhave distinct resolutions. In order to compare multiple imageson a pixel-by-pixel basis they need to be smoothed to a com-mon (larger) PSF, which is a procedure that involves the use ofconvolution kernels. The construction of a convolution kernel isbased on the knowledge of the PSF of the image that needs tobe processed and the common PSF. Third, the interface betweendi ff erent regions (e.g. photodissociation regions or the outskirtsof galaxies) are often rich in information. A strong gradient in in-tensity usually characterises the interface, however, making thecontrast between these regions very strong. Faint wings of thePSF of the instrument can extend very far from the PSF cen-tre and if the contrast between regions is su ffi ciently high, faintstructures of the PSF can have an intensity that is comparable tothat of the fainter regions. This represents a possible source ofcontamination and needs to be carefully taken into account.The PSF of the photometer of the PACS instrument(Poglitsch et al. 2010) on board Herschel is characterised by(Lutz 2015) a narrow core, a tri-lobe pattern, and knotty struc- (cid:63)
Herschel is an ESA space observatory with science instrumentsprovided by European-led Principal Investigator consortia and with im-portant participation from NASA. (cid:63)(cid:63)
FITS files of our PACS PSFs (Fig. 2) are available in elec-tronic form at the CDS via anonymous ftp to cdsarc.u-strasbg.fr(130.79.128.5) or via http://cdsweb.u-strasbg.fr/cgi-bin/qcat?J/A+A/ tured di ff raction ‘rings’ at sub-percent level. For fast scans instandard and parallel mode, this PSF structure due to the tele-scope is smeared by detector time constants and data averag-ing, resulting in a larger PSF. The PSFs delivered by the PACSinstrument control centre (ICC) have been observed using theVesta asteroid and provide information about the central part ofthe PSFs (until a radius of ∼ (cid:48)(cid:48) ) at di ff erent scan speeds. Marsobservations were used for the characterisation of the encircledenergy fraction (EEF) in the wings of PSFs but were not com-bined with Vesta observations in the PSFs delivered. The goalof this Research Note is to combine Vesta and Mars dedicatedobservations to provide new estimates of the PACS PSFs withan unprecedented dynamic range ( ∼ ) to permit the propercharacterisation of the central part and the wings.This research note is organised as follows: in Section 2 wepresent the PACS observations, give details on the data process-ing followed, and describe the method used to produce the finalPSFs. In Section 3 we analyse the PSFs and compare them toextragalactic images and, finally, in Section 4 we draw our con-clusions.
2. Vesta and Mars PACS data processing
Herschel
PACS dedicated PSF observations are scan maps cen-tred on various objects taken at 70 (blue band), 100 (green band),and 160 (red band) µ m. The core of the PSF is best characterisedobserving faint objects (e.g. the asteroid Vesta), while the wingsof the PSF can only be seen in observations of bright objects(e.g. Mars). Using a combination of images of bright and faintobjects, it is therefore possible to obtain a good characterisationof the PACS PSFs. We consider dual-band observations of Vesta and Mars (see ob-sIDs in Table 1) taken at the three di ff erent PACS wavelength a r X i v : . [ a s t r o - ph . I M ] J un . Bocchio, S. Bianchi A. Abergel: Deep Herschel
PACS point spread functions (RN)
Table 1: ObsIDs of the available observations of Vesta and Mars.
Band Scan speed Vesta Marsblue / red 20 1342195472 -1342195473 -blue / red 60 - 13422311571342195470 13422311581342195471 1342231159- 1342231160green / red 20 1342195476 -1342195477 -green / red 60 - 13422311611342195474 13422311621342195475 1342231163- 1342231164 bands and with a scan speed of 20 (cid:48)(cid:48) / s and 60 (cid:48)(cid:48) / s . The Herschel
Interactive Processing Environment (HIPE; v.12.1.0; Ott 2010)was first used to bring the raw Level-0 data to Level-1 using thePACS calibration tree PACS CAL 65 0 and the pipeline scriptsfor Solar System Objects (SSO). Maps were then produced us-ing Scanamorphos (v.24.0, Roussel 2013) with pixel sizes of 1 (cid:48)(cid:48) and rotated of the roll angle θ RA = . ◦ and 108.6 ◦ clockwise(for Vesta and Mars observations, respectively) so as to have theZ-axis of the spacecraft point up.Unfortunately, there are no su ffi cient dedicated observa-tions in parallel mode that are useful for a correct character-isation of the PSF. We bypassed this problem partly simulat-ing parallel-mode observations at 20 (cid:48)(cid:48) / s and 60 (cid:48)(cid:48) / s from obser-vations in standard mode. This is performed by modifying theHIPE pipeline to average fluxes and coordinates of two consec-utive frames of our data, therefore mimicking the reduced sam-pling frequency of the detectors (for the blue and green bandsonly) in parallel mode (Lutz 2015; private communication).The observations considered have an array-to-map angle(ama) of ± . ◦ , i.e. scan directions form an angle of ± . ◦ with respect to the Z-axis of the spacecraft. This angle corre-sponds to the SPIRE “magic” angle (Valtchanov 2014) and itis always used for observations in parallel mode. Observationsin standard mode can be performed using an ama = ± . ◦ or ± ◦ .The scanning direction does not a ff ect the shape of the PSFfor low scanning speeds (10 and 20 (cid:48)(cid:48) / s ). On the contrary, Vestaimages obtained at fast scanning speed (60 (cid:48)(cid:48) / s ) present a clearelongation along the scanning direction. The central region ofMars images is saturated and no elongation is observed; Marsobservations at lower scan speeds are therefore not required andimages taken at 60 (cid:48)(cid:48) / s are used for the characterisation of thefaint wings at both slow and fast scan speeds. In the Herschel
Science Archive there is no PACS observation (apart from dedi-cated observations for PSFs analysis) at 60 (cid:48)(cid:48) / s in standard modewith an ama = ± ◦ , we therefore consider only PSFs withama = ± . ◦ . For the observations of Mars, the pixels cen-tered on the source are heavily saturated, leading to significanttrails. When using Scanamorphos (Lutz 2015) these artifacts aregreatly reduced, masking the a ff ected regions. Adopting an ama of ± . ◦ provides a good coverage for fully sam-pled maps in the three SPIRE bands. Fig. 1: Average radial profile of Vesta (green) and Mars (blue)observations and of the estimated PSF (black) for 70 µ m imagesat a scan speed of 20 (cid:48)(cid:48) / s . The light green and light blue lines areVesta and Mars profiles for r > c . The shaded region indicatesthe range of profiles along di ff erent directions of the estimatedPSF. Vertical dashed black lines indicate r and r . In order to produce PSFs for images obtained at low scan speeds(10 and 20 (cid:48)(cid:48) / s ), we merge Vesta and Mars observations at 20 (cid:48)(cid:48) / s and 60 (cid:48)(cid:48) / s , respectively, while we use Vesta and Mars observa-tions at 60 (cid:48)(cid:48) / s for PSFs for high scan speeds (60 (cid:48)(cid:48) / s ). The sameis done for the parallel mode, using the corresponding partlysimulated data.First of all, we notice that Mars images have a good signal-to-noise ratios (S / N) up to ∼ (cid:48)(cid:48) from the centre, while Vestaimages are noisy for radii that are larger than 60 (cid:48)(cid:48) . We measurethe background in Mars images and we remove it, while thebackground estimation for Vesta is computed with a more so-phisticated technique (see Sect. 3.1). Images are then normalisedso that the total integrated flux equals unity and the radial profile(averaged over the 2 π angle) for both Vesta and Mars is com-puted (see Fig. 1 for an example at 70 µ m with a scan speed of20 (cid:48)(cid:48) / s ). The central region of the Mars image is saturated andthe ratio between the two profiles is constant over a given radialregion (i.e. between r = (cid:48)(cid:48) and r = (cid:48)(cid:48) for observations inFig. 1). To avoid introducing any artifacts in the PSFs, we thenrescale the Mars images to those of Vesta and compute the PSFas PS F ( r , θ ) = V ( r , θ ) if r ≤ r V ( r , θ )[1 − f ( r )] + M ( r , θ ) f ( r ) if r < r < r M ( r , θ ) if r ≥ r , (1)where V ( r , θ ) and M ( r , θ ) indicate the Vesta and the rescaledMars images, respectively, and f ( r ) = (cid:32) r − r r − r (cid:33) − (cid:32) r − r r − r (cid:33) + (cid:32) r − r r − r (cid:33) , (2)is a smooth Heaviside step function with null first and secondderivatives at the extremes r and r .The average radial profile of the estimated PSF tightly fol-lows that of Vesta at short radii, while it tends to the profile ofMars further from the centre (see Fig. 1). During the operationof merging, the information on the asymmetry of the PSF is notlost and radial profiles measured along di ff erent directions showrather strong variability with respect to the average radial profile(shaded region in Fig. 1).
2. Bocchio, S. Bianchi A. Abergel: Deep
Herschel
PACS point spread functions (RN)
Fig. 2: Our estimates of PACS PSFs (in logscale) as a function of the filter and scanspeed. All images are 300 (cid:48)(cid:48) × (cid:48)(cid:48) . Redband PSFs in standard and parallel modesare exactly the same. The spacecraft Y- andZ-axis are to the left and to the top, respec-tively.We also tested for the impact of the finite size of Mars on thePSF determination. At the time of observations, the apparent di-ameter of the planet as seen from the L2 point was d M ∼ . (cid:48)(cid:48) r > r .When a circle of diameter d M ∼ . (cid:48)(cid:48)
55 is convolved with our com-posite PSFs, the profile of the resulting images for r > r remainsunchanged with respect to the PSFs. This demonstrates that thefinite size of Mars does not a ff ect the shape of the derived PSFs.
3. PSF analysis
The EEF is computed using the same notation and following themethod by the ICC (Lutz 2015) as follows:
EEF obs ( r ) = (cid:82) r (cid:82) π [ PS F ( r , θ ) − c ] rdr d θ − c (cid:82) c (cid:82) π [ PS F ( r , θ ) − c ] rdr d θ − c , (3)where c represents the background value to be removed fromthe observed image, c is the maximum radius out to which wecompute the EEF, and c is the flux missing in the PSF core dueto saturation.We compute the EEF for the Vesta and Mars images (seeFig. 3) assuming c = c = (cid:48)(cid:48) and 1000 (cid:48)(cid:48) for Vesta andMars, respectively. We deduced the value of c for Mars imagesby comparing the integrated observed flux to the nominal fluxgiven by the ICC (i.e. S [70] = S [100] = S [160] = ∼ . , . .
8% of the nominal values at 70, 100, and 160 µ m, respec-tively.However, the flux of Vesta for radii larger than c is non-negligible and must be taken into account for the calculation of Ephemerides can be accessed at: http://ssd.jpl.nasa.gov/?horizons the EEF. Furthermore, since the S / N in Vesta maps is very lowat r > c , the value of c cannot be estimated directly from ob-servations. From a visual comparison of the slope of the Vestaand Mars EEF, we find c (cid:39) − (normalised to the peak of theVesta image) for all pairs of observations. The flux outside ra-dius c in Vesta observations is 8.3%, 9.8%, and 12.3% for blue,green, and red bands. We then correct the Vesta EEF for the fluxlost due to the non-zero c and the flux outside radius c (dottedline in Fig. 3). The corrected EEF curve tightly follows that ofVesta at short radii and tends to that of Mars for r (cid:38) (cid:48)(cid:48) . We fi-nally compute the EEF for our estimate of the PACS PSF (solidline in Fig. 3) and note that it agrees very well with the correctedVesta EEF, therefore supporting the methodology used. The EEFcurve that we obtain is comparable to that presented by the PACSICC (red dashed line in Fig. 3, Lutz 2015), with a little di ff erencefor r (cid:46) (cid:48)(cid:48) due to the adopted centering technique.Fig. 3: Encircled energy fraction (blue band) for Vesta (greenline), Mars (blue line), for the estimated PSF (solid black line),and as obtained by the ICC (red dashed line). The dotted blackline indicates the corrected Vesta EEF (see text for details).
3. Bocchio, S. Bianchi A. Abergel: Deep
Herschel
PACS point spread functions (RN)
Fig. 4: PACS 70 µ m observations of NGC253 with overplotted contours of our PSF,rotated of an angle θ RA ∼ ◦ clockwise(see text). Right panel, zoom to the centralregion. The computed PACS PSFs are presented in Fig. 2. They are notaxisymmetric and are characterised by a large width variabil-ity depending on the direction. Using a two-dimensional (2D)Gaussian fit of the PSFs, we measure the FWHM along the Yand Z directions. The resulting values are reported in Table 2and are comparable to those obtained by the PACS ICC (Lutz2015). As expected, the FWHM increases in both Y and Z direc-tions from blue to red filters. PSF features are observed along theZ direction (see Fig.2) and the width is therefore systematicallylarger (up to 20%) along this axis with respect to the Y direc-tion. The scan speed contributes to the enlargement of the PSFand a clear elongation is visible towards the scanning direction( ± . ◦ with respect to the Z axis). Astronomical objects presenting a strong contrast between dif-ferent regions can present evident PSF features. NGC 253 is aintermediate spiral galaxy currently undergoing an intense starformation. This galaxy has been observed by PACS at 70 and160 µ m at a scan speed of 20 (cid:48)(cid:48) / s and with the + Z direction of thetelescope rotated by an angle θ RA ∼ ◦ west of north (clock-wise). At these wavelengths, the central region of the galaxy isvery bright compared to the rest of the galaxy, which is thencontaminated by the lobes and faint structures of the PSFs.Fig. 4 illustrates the PACS 70 µ m image of this galaxy withthe contours of our estimated PSF overplotted. On the rightpanel, we show a zoom to the central region and overplot con-tours of the PSF core. Both the faint structures and core of thePSF dominate over the extended emission of the galaxy andTable 2: FWHM of the estimated PSFs in arcsec along the Y andZ directions. Scan speeds are indicated in arcsec / s, 20p and 60pare in parallel mode. Band Dir. Scan speed20 20p 60 60pblue Y 5.77 6.28 7.75 9.71Z 6.35 6.94 8.72 10.77green Y 6.90 7.31 8.68 10.67Z 7.26 7.75 9.51 11.72red Y 10.59 11.80Z 12.29 13.70 match very well with the PSF . This observation represents anexample of a case where a good characterisation of the PSF ofthe instrument is needed to correctly interpret astrophysical data.Similarly, our computed PACS PSFs were used in a recentstudy of the scale height of the dust distribution in a nearby edge-on galaxy, NGC 891 (Bocchio et al. 2016). The larger width ofthe PSFs compared to that of modelled PSFs for the ‘as built’telescope (Geis & Lutz 2009) and their radial asymmetry lead toa narrower dust scale height by up to a factor of ∼
4. Conclusions
Using dedicated observations of Vesta and Mars, we provide newestimates of the PACS PSFs for scan speeds of 20 (cid:48)(cid:48) / s and 60 (cid:48)(cid:48) / s both in standard and parallel mode. The obtained PSFs have awide dynamic range ( ∼ ) enabling a proper characterisationof both the core and faint structures of the PSFs.As an example we consider NGC 253, a galaxy with a strongcontrast between the central and peripheral regions. From a com-parison between our estimated PSFs and PACS observations ofNGC 253, we obtain an excellent matching, therefore supportingthe reliability of the method used. Acknowledgements.
We would like to acknowledge Prof. D. Lutz for a use-ful discussion and for giving us part of the code needed to simulate parallel-mode observations. We thank the PACS ICC for their insightful comments onthe paper. We acknowledge K. Dassas for making the PSFs available online onthe IDOC website. Part of this work has received funding from the EuropeanUnions Seventh Framework Programme (FP7 / ◦ FP7-SPACE-606847).
References
Bocchio, M., Bianchi, S., Hunt, L. K., & Schneider, R. 2016,A&A, 586, A8Geis, N. & Lutz, D. 2009, Herschel-PACS document PICC-ME-TN-029, v2.2, http://pacs.ster.kuleuven.ac.be/pubtool/PSF/PACSPSF_PICC-ME-TN-029_v1.0.pdf
Lutz, D. 2015, private communicationLutz, D. 2015, Herschel-PACS document PICC-ME-TN-033,v2.2, http://herschel.esac.esa.int/twiki/pub/Public/PacsCalibrationWeb/bolopsf_22.pdf
Okumura, K. 2010, Herschel-PACS document SAp-PACS-KO-0716-10 Faint linear artefact in PACS photometerOtt, S. 2010, in Astronomical Society of the Pacific ConferenceSeries, Vol. 434, Astronomical Data Analysis Software andSystems XIX, ed. Y. Mizumoto, K.-I. Morita, & M. Ohishi,139 The faint stripes visible in the left panel are due to crosstalk(Okumura 2010) and are not directly related to the shape of the PSF.4. Bocchio, S. Bianchi A. Abergel: Deep
Herschel
PACS point spread functions (RN)