Table Of Content

Astronomy&Astrophysicsmanuscriptno.DarioFadda (cid:13)cESO2016 August19,2016 Transient effects in Herschel/PACS spectroscopy DarioFadda1,(cid:63),JefferyD.Jacobson2,andPhilipN.Appleton2 1 InstitutodeAstrofisicadeCanarias,E-38205,LaLaguna,Tenerife,Spain UniversidaddeLaLaguna,Dpto.deAstrofísica,E-38206LaLaguna,Tenerife,Spain 2 NASAHerschelScienceCenter—CaliforniaInstituteofTechnology,MC100-22,Pasadena,CA91125,USA e-mail:[email protected],[email protected] August19,2016 6 1 ABSTRACT 0 2 Context.TheGe:GadetectorsusedinthePACSspectrographonboardtheHerschelspacetelescopereacttochangesoftheincident fluxwithacertaindelay.Thisgeneratestransienteffectsontheresultingsignalwhichcanbeimportantandlastforuptoanhour. g Aims.Thepaperpresentsastudyoftheeffectsoftransientsonthedetectedsignalandproposesmethodstomitigatethemespecially u inthecaseofthe“unchopped”mode. A Methods.Sincetransientscanarisefromavarietyofcauses,weclassifiedtheminthreemaincategories:transientscausedbysudden variationsofthecontinuumduetotheobservationalmodeused;transientscausedbycosmicrayimpactsonthedetectors;transients 8 causedbyacontinuoussmoothvariationofthecontinuumduringawavelengthscan.Weproposeamethodtodisentangletheseeffects 1 andtreatthemseparately.Inparticular,weshowthatalinearcombinationofthreeexponentialfunctionsisneededtofittheresponse variationofthedetectorsduringatransient.Analgorithmtodetect,fit,andcorrecttransienteffectsispresented. M] Results.Thesolutionproposedtocorrectthesignalfortheeffectsoftransientssubstantiallyimprovesthequalityofthefinalreduction withrespecttothestandardmethodsusedforarchivalreductioninthecaseswheretransienteffectsaremostpronounced. I Conclusions.Theprogramsdevelopedtoimplementthecorrectionsareofferedthroughtwonewinteractivedatareductionpipelines h. inthelatestreleasesoftheHerschelInteractiveProcessingEnvironment. p Keywords. methods:dataanalysis—techniques:spectroscopic—infrared:general - o r t s1. Introduction In this paper we briefly review the observational modes of a the PACS spectrograph and the data reduction techniques. We [TheHerschelspaceobservatory(Pilbrattetal.2010)completed describehowthe“unchopped”modeisparticularlysensitiveto itsmissiononApril29,2013afterperformingatotalof23,400 3 suddenchangesinincidentflux,andcosmicrayglitches,which vhours of scientific observations during almost 4 years of activ- cause memory effects in the detector responses. Finally, we in- 9ity ( 1446 operational days). One of the most used instruments troducethetechniqueusedtocorrectmostoftheseeffects,and 2aboard Herschel was the PACS spectrometer (Poglitsch et al. show a few selected examples. The methods and software de- 72010).Approximatelyonequarterofthetotalscientifictimewas scribedinthepaperarenowimplementedintheHerschelInter- 7devoted to PACS spectroscopy. Although most of the observa- activeProcessingEnvironment(HIPE1)(Ott2010).Thecompar- 0tions were performed in the standard “chop-nod” mode, a sub- isonspresentedinthepaperarebetweenthearchivedSPG(stan- .stantialfraction(30%oftheobservations,correspondingtothe 1 dard product generation) products, generated with HIPE 14.0 025%ofthetotalspectroscopytime)usedtwoalternativemodes. andcalibrationdataversion72,andourpipelineinHIPE15. 6The“wavelengthswitching”modewasreleasedtousersafterthe 1startofthemissiontoallowPACSspectrometerobservationsto v:bemadeincrowdedfieldswherechoppingwasnotpossible.A 2. ThePACSspectrometer yearlater,thismodewasreplacedbytheso-called“unchopped” i Xmode.BytheendoftheHerscheloperationmission,thismode WerecommendreadingPoglitschetal.(2010)andthePACSob- rwasusedforapproximatelyone-thirdofallPACSspectroscopic servationalmanualforadetaileddescriptionofthePACSspec- aobservations. The primary focus of this paper is to describe an trometer. As a way to introduce some terminology used in the optimalwaytoreducedatatakeninthe“unchopped”mode. paper,wegivehereaconcisedescriptionoftheinstrumentand The PACS spectrometer was able to observe spectroscopi- howitworks. callybetween60µmand210µmusingGe:Gaphoto-conductor ThePACSspectrometerwasanIFU(integralfieldunit)com- arrays. This type of detectors suffers from systematic memory posedbyamatrixof5×5spatialpixels(spaxelsorspacemod- effectsoftheresponsewhichcanbiasthephotometryofsources ules)coveringafieldof47(cid:48)×47(cid:48) squarearcminutes.The5×5 andincreasethenoiseinthesignal.Sucheffectshavebeendoc- pixel image passed into an image slicer, which rearranged the umentedandstudiedforsimilardetectorsonpreviousspaceob- 5×5two-dimensionalimageintoone-dimension(1×25),which servatories such as ISOCAM (Coulais & Abergel 2000; Lari thenfedaLittrow-mounteddiffractiongratingwhereitoperated etal.2001)andMIPS(Faddaetal.2006,Fig.17). at1st,2nd,and3rdorder.Thefirstorder(red)andsecondorthird (cid:63) MovingtotheSOFIAScienceCenter(USRA) 1 www.cosmos.esa.int/web/herschel/hipe-download Articlenumber,page1of14 A&Aproofs:manuscriptno.DarioFadda order(blue)wereseparatedwithadichroicbeam-splitter,where the spectra were re-imaged onto separate detectors. The wave- lengthrangecoveredis51-105µmfortheblue(choiceof51-73 and 71-105 µm for the 3rd and 2nd order) and 102-220 µm for thered,respectively. The dispersed light was detected by two (low and high- stressed) Ge:Ga photo-conductors arrays with 25 × 16 pixels. In the following we call spectral pixels the individual pixels of thetwoarrays.Wecall“module”or“spaxel”(fromspatialpixel) thesetof16spectralpixelscorrespondingtothedispersedlight fromasinglepatchof9.4×9.4squarearcsecondsonthesky. Since the instantaneous wavelength range covered by 16 spectral pixels is small (∼1500 km/s for many observations), the grating was typically stepped through a range of grating angles during an observation. This operation is referred as a “wavelength scan”. During a standard observation, the desired Nod B Chop 2 −10.58 wavelength range is covered by moving the grating back and Off forth. These movements are called up- and down-scans, since s] Slew thewavelengthseenbyasinglespectralpixelfirstincreasesand g e−10.62 d Chop2 then decreases as the grating executes a scan first in one direc- c [ On tion,andthenback. e Chop 1 D Theothermobilepartofthesystemisthechopperwhichlies −10.66 at the entrance to the whole PACS instrument. This is a mirror which allows the IFU to point at different parts of the sky. In Chop1 Nod A Off standardchop-nodmode,thechoppingmirrorisusedtoalterna- −10.70 tivelypointrapidlytothesourceandthenabackgroundposition, 310 320 330 340 350 whilemaintainingthesametelescopepointing.Thelargestchop- Time [s] perthrowavailableis6arcminutes,whichisalimitationofthe chop-nodmode.Ifthechoppermirrorismovedtolargerangles, Fig.1.Exampleofchop-nodobservationfootprints(top),andobserved itcanaccesstwointernalcalibrators(BB1andBB2)whichare position projected on the sky (bottom). In the first nod, the detector usedforcalibrationduringtheslewtothetargetsource. alternatesbetweentwoopticalpathstoobservetheobjectandtheback- Sincethemirrorispassivelycooled,theemissionofthetele- ground. These paths have slightly different telescope background sig- scopeissignificant,andtypicallydominatesthetotalsignal.For nals.Inthesecondnod,theopticalpathsareinvertedafterslewingthe instance,inthered,theemissionseenbyagivenpixelat120µm telescope.Byaddingthetwoobject-backgrounddifferences,itispossi- isaround300Jy.Thishightelescopebackgroundlevel,although bletoremovetheeffectsofthedifferenttelescopebackgroundlevels. not generally desirable in IR astronomy, has the advantage that ithelpstomitigatetransienteffectsinthedetectorsbyhelpingto effects. This paper is directed at ways of solving some of these keepthesignallevelconstantatthedetectors.Asecondproperty problemswithintheHIPEsoftware. ofthedominanttelescopesignalisthatitisalwayspresent,and canbeusedasarelativelyconstantreferenceflux.Thisproperty canbepositivelyexploitedintheanalysisofthesePACSdata. 3.1. Chop-nodmode The standard way to observe with the PACS spectrometer was the “chop-nod” mode. During a wavelength scan, the chopper 3. Observationalmodes modulatesbetweenan“on-source”and“off-source”position.In Spectroscopic observations consist of a calibration block fol- thisway,anyvariationoftheresponseslowerthanthechopping lowedbyaseriesofscienceblocks.Thecalibrationblockisper- timecanbesubtractedusingthe“off-source”constantreference. formedwhenthetelescopeisslewingtoreachthetargetandin- However,sincetheon-andoff-sourcechopsignalsfollowdiffer- volvesrapidchoppingbetweenthetwointernalreferenceblack- ent optical paths through the telescope, the observed telescope bodies.Thesciencetargetisthenobservedinoneormorebands background emission is slightly different in the two chop posi- using either the “chop-nod” or the “unchopped” observational tions. Toremovethis effect,the telescopeisnodded. Inthe left mode. The “chop-nod” mode is used for isolated sources and panelofFigure1weshowthefirstnodposition(nodB)whenthe involves the continuous chopping between target and an “off” sourceisobservedinthechop2position.Theright-handpanel position. The “unchopped mode” is used to observe crowded showsthesituationafterasmallslewwhichplacesthesourcein fields or extended sources. It consists of a staring observation chopposition1wheretheobservationisrepeated(nodA).The of the science target followed by an observation of the refer- optical paths used to observe source and off-field are inverted ence“off”positionbymovingthetelescope.Thismodereplaced so that the average of the “source minus off” signals is largely the“wavelength-switching”modeusedforearlierobservations. unaffectedbythedifferenttelescopebackgrounds. Instead of chopping, the “wavelength switching” used a wave- Two types of submodes are defined in the chop-nod mode. length modulation to move the line on the detector array by a Inthecasetheobservationinvolvesanunresolvedline,the“line wavelength equivalent to the FWHM of the line. This allowed chop-nod”modedefinesanoptimalwaytoobservethelineand thedifferentialspectrumtobemeasured,butitwasfoundtobe the surrounding continuum. If a larger wavelength range is re- inefficient during the verification phase and deprecated. Obser- quired,the“rangechop-nod”modecanbeusedtomanuallyen- vations in “unchopped mode” suffered from detector transient tertherangeofwavelengthstobescannedbythegrating. Articlenumber,page2of14 DarioFadda etal.:TransienteffectsinHerschel/PACSspectroscopy Thedatareductionisdoneoneachindividualpixel.Thesim- This methodfrees theresult fromthe effects ofthe variablere- plestreductiontechniqueusesthecalibrationblocktocompute sponse although it requires the knowledge of the dark. Since theaverageresponseduringtheobservation.Theresponseisde- PACS has no shutter, the dark was measured in the laboratory fined for each pixel as the ratio between the measured and ex- beforethelaunchofHerschel.Itisnotknownwithgreataccu- pectedsignalfromtheinternalcalibratorswhichwasmeasured racyalthoughitsvalueisnegligiblewithrespecttothetelescope inthelaboratory.Themeasuredsignaliscomposedofthetele- background.Thetelescopebackgroundwasderivedinflightby scope background (T and T in the two chopping positions), calibrating the emission of the mirror with observed spectra of A B sourceflux(s),anddark(d)signal.Thetelescopebackgroundis brightasteroids. dominated by the emission of the primary mirror, and depends on the temperature of the spot which is seen by the detector. Becauseofdifferentialvariationsinthemirrortemperature,and 3.2. Unchoppedmode its emissivity with position on the mirror, the different chopper At the end of the verification phase, the “wavelength switch- positionstendtoseedifferingdegreesoftelescopebackground. ing” mode (see Poglitsch et al. 2010) was released3 and used Although, during a given observation, the overall temperature in a few key programs. Several disadvantages became appar- of the mirror changed slowly, it was measured to vary due to ent with this mode during its early use, leading to its replace- changes in illumination of the spacecraft by the Sun on longer mentwiththe“unchopped”mode.Forexample,byconstruction, timescalesthanasingleobservation. the continuum of the source could not be measured. Moreover, ForthemeasuredsignalC,inV/s,inachop-nodobservation only observations of unresolved lines could be measured cor- wehave: rectly. Asymmetric lines or lines broader than the wavelength switchingintervalwerefoundtobedifficulttorecover.Forlarge C =(T +d+s)·R(t) C = (T +d)·R(t) 1 A 2 B wavelength ranges, SED shape variations can lead to peculiar (1) C(cid:48) = (T +d)·R(t) C(cid:48) =(T +d+s)·R(t) baselines in the final differential spectrum. This made observa- 1 A 2 B tionsoverlargewavelengthrangeschallenging.Finally,therapid for the two chop positions in the first (C and C ) and second switching of the grating between two distant wavelength posi- 1 2 nods(C(cid:48) andC(cid:48)),respectively.HereR(t) = ρ ·r(t)istheprod- tionscreatedmechanicaloscillationswhichrequiredalongtime uctofth1erelativ2espectralresponsefunction(λRSRF)ρ andthe todamp.Thiseffectwasmoresevereinspacethanduringground λ response function r(t) (V s−1 Jy−1) which can vary with time. testing,leadingtolongertimeintervalsbetweenusefulobserva- T andT correspondtothesignaldetectedfromthetelescope tionalsamplesandthuspreventinganefficientremovalofrapid A B backgroundatthechopperpositionsAandB,respectively. responsevariations. The simplest technique consists in estimating the signal It appeared clear that performing a slow grating scan while sourcebycomputingthedifferentialsignal,sˆ,betweenthechop- staring at an object, followed by a similar observation pointed pingpositionsandcombiningthenods: to an “off” position, yielded superior results compared with “wavelength-switching”. This “unchopped” mode had the ad- sˆ=[(C −C )+(C(cid:48) −C(cid:48))]/(2·ρ ·r ), [Jy] (2) vantage of allowing large wavelength range scans, such as far- 1 2 1 2 λ CB IR SEDs of confused or extended regions. Unlike the “chop- where rCB = (cid:104)r(t)(cid:105) is the average response estimated from the nod” mode, where the two chopper positions sample different calibrationblock.Aswecansee,themethodworksifr(t)isap- opticalpathsandthereforedifferentmirrortemperatures,inthe proximatelyconstantduringachoppingperiod(1sec)sothatthe “unchopped”modetheopticalpathsusedforon-andoff-source subtraction cancels the effect of a transient. This approach was observationsarethesame.Therefore,themirrortemperaturein usedbythesocalledSPG(standardproductgeneration)pipeline the on- and off-source are exactly the same. Moreover, in the to populate the Herschel Science Archive (HSA) for chop-nod caseofchop-nod,thefootprintofthedetectorontheskyrotates observations2untilrecently. slightlybetweenthetwochoppositions(seethePACSobserver SinceSPG13,analternativeapproachexploitingtheknowl- manual, Figure 4.7). The effect, though small, is worse for the edge of the telescope background is used. This technique does largerchopperthrows.Thismeansthatinthetwonodpositions, notrequiretheapplicationoftheRSRFandresponsecorrections theonlyspaxelseeingpreciselythesamepartofthesourceisthe sinceitusestheratiosbetweentheobservationsinthetwochop- centralone.Spaxelsfurtherfromthecenterbecomesuccessively pingdirections.Nevertheless,theaccuracyoftheresultsdepends mismatchedinthetwonods.So,thereductionworksbestforthe on the knowledge of the mean telescope background. Also, ra- centralspaxel.Thisproblemdoesnotexistforthe“unchopped” tios introduce more noise in the final signal with respect to the mode because the observation of the on- and off-positions are standardreduction. madewithafixedcentralchopperposition. Informulae,ifwedefinenormalizationas: The “unchopped” line and range modes were released on September2010,supersedingtheprevious“wavelengthswitch- C −C C(cid:48) −C(cid:48) 2s N = 1 2 + 1 2 = , (3) ing” mode4. A specific “unchopped” mode for bright lines was C1+C2 C1(cid:48) +C2(cid:48) TA+TB+2d+s releasedonApril20115. In “unchopped” mode, a complete wavelength scan is done the signal normalized to the average telescope background can first on the “on-source” (hereafter ON) position and then later beexpressedas: 3 herschel.esac.esa.int/Docs/AOTsReleaseStatus/ s = s = N . (4) PACS_WaveSwitching_ReleaseNote_20Jan2010.pdf (cid:104)T(cid:105)+d (T +T )/2+d 1−N/2 4 herschel.esac.esa.int/Docs/AOTsReleaseStatus/ A B PACS_Unchopped_ReleaseNote_20Sep2010.pdf The source flux can be therefore derived by multiplying the 5 herschel.esac.esa.int/twiki/pub/Public/PacsAotReleaseNotes/ normalizedsignalbythetelescopebackground(darkincluded). PACS_UnchoppedReleaseNote_BrightLines_15Apr2011.pdf Note that all the AOT release notes will be provided on the Herschel 2 http://www.cosmos.esa.int/web/herschel/science-archive LegacyLibrarywebsitestartingfrom2017. Articlenumber,page3of14 A&Aproofs:manuscriptno.DarioFadda 10 BB1 9 2.5 BB1 8 BB2 2.0 ] 7 s V/ 6 [ Calibration block 1.5 x 5 BB2 u Fl 4 Previous Transient 1.0 Calibration block 3 values Observation Previous 2 0.5 values Obs 1 Red array Blue array 0.0 0 10 20 30 40 50 60 70 80 90 100 110 120 10 20 30 40 50 60 70 80 90 100 110 120 1.10 1.10 1.05 1.05 > x u Fl1.00 1.00 < ux / 0.95 BB1 0.95 BB1 Fl BB2 BB2 0.90 0.90 Red array Blue array 0.85 0.85 20 30 40 50 60 70 80 90 100 110 20 30 40 50 60 70 80 90 100 110 Time [s] Time [s] Fig.2.Transientsaffectingthecalibrationblockforthered(left)andblue(right)arrays.Duringthecalibrationblock,thedetectorsseealternatively thetwoblackbodies(toppanels).Sincetheinitialfluxissignificantlydifferentfromthefluxesoftheinternalblackbodies,astrongtransientoccurs duringthefirstminuteoftheobservation.Thesignalsnormalizedtothemedianvaluesofeachblackbody(bottompanels)showhowthesame transientaffectsthefluxesfromthetwoblackbodies(blueandredlines).Thetransienteffectismorepronouncedfortheredarray. ona“off-source”(hereafterOFF)positionclearofsourceemis- evident after sudden variations of flux on detector arrays (see, sion. Unfortunately, variations of the response during the ON e.g., Coulais&Abergel2000).InFigure2weshowthesignal scan cannot be corrected using the OFF scan because the two detected during the calibration block when the chopper points scansareperformedatdifferenttimes.So,althoughunchopped alternativelybetweenthetwointernalblackbodies(whichhave observationsoffersomeadvantagesoverthechop-nodmode,the differenttemperatures).Passingfromthepreviousobservationto effects of transients require mitigation. In this paper, we show theinternalblackbodiescausesacleartransienteffectwhichis thetypicaltransientsfoundinthesignal,andsometechniquesto more important in the case of the red array (left panels). When modelandsubtractthem. thesignalisnormalizedtotheasymptoticflux(bottompanels), Threedifferentsub-modesexistfortheunchoppedmode: theresponsevariationbecomesevident.Itisinterestingtonote (leftbottompanel)thattheshorttransientsduetocosmichitson – unchoppedline,forsinglelineobservations. thearrayhaveatimescalelongerthanthechoppingtime.Thisal- – unchoppedbrightline,forsinglebrightlines.Itis30%more time efficient than the standard line mode since bright lines lowsthecorrectionoftheseeffectsusingthe“chop-node”mode. Wecanclassifythreeseparatetypesoftransients: requirelesscontinuumtocomputethelineintensity. – unchopped range, for observations of the continuum or a – continuum–jump transients: transients due to a sudden complexoflines. change of the incident flux from one almost constant level A further difference between these sub-modes is that the refer- toanotheralmostconstantlevel; – cosmicraytransients:transientsinducedbycosmicrayhits enceOFFisobservedduringthesameAOR(AstronomicalOb- whichproduceglitches,followedbyaresponsevariation; servation Request) in thecase of the “unchopped line”, but has – scan dependent transients: transients along a wavelength tobeprovidedasaseparateAORinthe“unchoppedrange”.In scan produced by rapid variations of the (dominant) tele- ordertoensuretheobservationsareperformedsequentially,the scope background and continuum (in the case of a bright OFFobservationintheunchoppedrangescanisconcatenatedto target). themainobservationwithintheobservationsequence. ItisworthmentioningthattheStandardProductGeneration (SPG)pipeline,whichpopulatestheHSA,doesnotuseanytype 4. Transients of transient correction. So the products available in the archive The term “transient” refers to a delayed response of a detector for the “frequency switched” and “unchopped-mode” observa- to the variation of the incident flux. Transients are particularly tions contain many transient effects that could adversely affect Articlenumber,page4of14 DarioFadda etal.:TransienteffectsinHerschel/PACSspectroscopy science goals. However, from HIPE 14 onward, users have the 5 option of processing their data with special scripts designed 1 1. specifically to correct transients. The current paper describes 0 1 how these corrections are made. In the following we describe 1. indetailthedifferenttypesoftransients. 5 0 1. 0 0 4.1. Continuum-jumptransients 1. 5 9 Asuddenchangeintheilluminationofadetectorpixelpassing 0. from one flux level to another, can induce a major transient in 0 2 thesignal.Thisoccurstypicallyatthebeginningofeachobser- 1. 0.00 0.05 0.10 0.15 vationjustafterthecompletionofthecalibrationblock.Because 5 of the difference between the flux from internal calibrators and 1.1 g k the telescope background, a transient is usually visible during B 0 theentireobservation. 1.1 Tel cahsaonAugrnecoeot.hfFebroarncdoex,moammrcophnlaenf,ogtrhemeoufoswfearthvmieslaekynignhtdahvooefcrcteruqarunseswisethenidtlsetwiosbowsedhriveffnienrga- 1.05 gnal / entlinesoccurringatdifferentwavelengths.Whenthegratingis 00 Si commandedtoaccessadifferentpartofthespectrum,orchange 1. to a completely different region of the spectrum, a jump in the 95 0. Time [hour] signalusuallyoccursbecauseofthechangingemissionspectrum of the telescope background at different wavelengths (see Sec- tion 5.2). If the source is very bright, differences in continuum 0.0 0.3 0.6 0.9 1.2 1.5 1.8 levelfromthesourcecanalsoinduceatransient. Ifthesourcecontinuumisnegligiblewithrespecttothetele- Fig. 3. When the signals from a spaxel are normalized to the ex- scope background, it is possible to normalize the signal during pectedtelescopebackgroundalong-termtransientafterthecalibration the whole observation to the telescope background which has block becomes clear. In this example (an observation of M81, ObsID beenpreviouslycalibrated.Thetransientduetothejumpincon- 1342269535)anOFFpositionisobservedatthebeginningandatthe tinuumbetweenthecalibrationblockandthetargetobservation endoftheobservation(bluedots),whilea2x2rasterobservationisper- appearsclearly,seeFigure3.Itispossible,inthiscase,tofitthe formedinthemiddle(greendots).Sincethecontinuumemissionofthe behaviourwithamodelandsubtractitfromtheobservation.Itis objectisnegligiblewithrespecttothetelescopebackground,itispossi- bletofitthegenerallong-termtransient(redline)andcorrectthesignal. interestingtonotethat,withoutthiscorrection,anOFFobserva- Notethatmanyhugetransientsappearinindividualspectralpixelsbe- tioncanhaveameasuredfluxgreaterthananONobservationif causeofcosmicrayhitsonthedetectors.Ifthiscorrectionisnotmade, ithappenstobeperformedduringaperiodofresponsestabiliza- andanaveragebackgroundissubtractedfromthesignal,thefinalim- tion. This effect is not taken into account by the SPG pipeline, age will present an artificial gradient and the flux will be negative in so that many observations in the archive have an artificial and someregions.Topinsetpanel:aclose-upoftheinitialpartofthesignal confusingnegativebackground. justafterthecalibrationblock,showingthestrongtransient.Theperi- Someobservationsthatusedtheunchoppedmodecontained odicvariationinthesignalisaleft-overfromtheimperfecttelescope requestsformorethanonebandinasingleAOR.Forextended backgroundestimateusedfornormalization.Theemptyspacesbetween sources, a typical raster observation allows one to cover an ex- pointsoccurduringtelescopeslewing. tended region by moving the telescope to cover a grid of positions. However, for efficiency reasons, the AOR was designed tocyclethroughalltherequestedbandsateachrasterposition, bandchangesappeartolimitthequalityoftheobservation,even beforemovingthetelescopetothenextpositionintherasterse- withthemostadvanceddatareduction.Indeed,weareabletosee quence.Anunexpectedsideeffectofthisstrategywasthegen- jump-transients even in the blue channel, which has the fastest eration of transients at each band change. It is not possible to responsestabilizationcomparedwiththered(seeFigure4). remove these jump–transients in a clean way in case of such multiple-bandobservations. There is an important lesson to be learned from this expe- 4.2. Cosmicraytransients rience with PACS. In retrospect, it would have been better to observe the complete raster in a single band, before cycling to ItiswellknownthatforGe:Gadetectors,energeticcosmicrays the next band. This would have minimized the effects of band- usuallyproduceglitchesinthesignal,followedbyresponsevari- inducedjump-transient.Weadvisethatanyfuturemissionwhere ations. Depending on the energy of the cosmic ray, the glitch transient could be an issue, should avoid as much as possible can be followed by a tail or the variation can be more compli- suddenchangesinfluxduringtheobservation.Thisisprobably cated (lowering temporarily the response). A similar behavior thecaseofFIFI-LSonboardSOFIA,whichisessentiallyaclone wasnotedinthepastforpixelsintheISOCAMarrayontheIn- of the PACS spectrograph. Unfortunately, when the observa- frared Space Observatory (see, for example, Lari et al. 2001). tionalmodewasintroducedforHerschel,thereductionpipeline Wewillshowthattheresponsevariationcanbedescribedwitha wasnotfullydeveloped,anditwasverydifficulttoevaluateall combinationofexponentialfunctions(seeSection5).Themain theeffects.Sincetheresponsewasknowntostabilizerelatively challenge to correcting and masking the damage in the signal fast, it was assumed that the effect of a band change was mini- fromcosmicrayhitsistoselectthemostsignificantevents,and mal,comparedwiththeadvantageofbeingmoretimeefficient. thenfindthestartingtimewhichmarksthebeginningofthedis- Experiencelatershowedthatjump-transienttransientcausedby continuityinthesignal. Articlenumber,page5of14 A&Aproofs:manuscriptno.DarioFadda 300 CR transients ] Transients y J 250 [ l Bkg gna 200 el si T d nal / lize 150 Band 3 Band 3 Sig Drift ma 100 Band 1 Band 2 Band 1 Band 2 r o N 50 On−source Off−source 0 −0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Time [hours] Time [hours] Fig.4. Long-termtransientforthepixelsofmodule16inthebluechan- Fig. 5. The central spaxel of an observation of NGC6543 (ObsID neloftheobsid1342246963.Oneofthepixelsshowsalsoatransient 1342215704) with three lines observed. For each observation block, duetoacosmicrayhit.Thereisa30%variationbetweenthetwoends thesignalineachwavelengthscanisdividedbythelastscanandnor- oftheobservation.Blueandgreendotsrefertooff-targetandon-target malized to its asymptotic median flux. The red line shows the tran- partsoftheobservation,respectively. sientmodelfitted.Thefirstblockcomesafterthecalibrationblock(not shown)whosefluxishigherthanthecontinuumduringthispartofthe observation.Thethreelinesareobservedfirstonthetarget(greendots) 4.3. Scandependenttransients andthenontheOFFposition(bluedots).Thegapsinthesesciencedata correspond to the time spent moving the grating, and slewing to the Finally,inthecaseofobservationsspanninganextendedwave- OFFposition.Alsointhiscasesomepixelsappeartobehavedifferently lengthrange,rapidvariationofthecontinuumduringthewave- becauseoftheimpactofcosmicrays. length scan produced transients in the signal (see Section 5.4 foranexample).Insomeobservations,thislasteffectisrespon- revisittheindividualpixelresponsesbynormalizingtheirsignal sible for different apparent fluxes in the source spectrum for to this “first guess” spectrum. In this way, each individual de- upward and downward scans over the same wavelength range. tectorpixelresponsetocanbestudiedfortheeffectsofcosmic Luckily, many PACS observations using the unchopped mode raytransients,andtheseCRtransientscanbeeithercorrectedor targeted only one line with a short range scan. For these cases, masked.Inthefollowing,wedescribethemodelusedtofitthe scan-dependent transients had negligible effects. It was found responseasafunctionoftimeandthealgorithmsappliedtothe to be important only for observations which scanned over ex- signaltomitigatetheeffectofthetransients. tended wavelength ranges. As a result, two different interac- tive pipeline scripts (accessible within HIPE as so–called ipipe scripts)havebeenwrittentotreatthe“unchoppedline”and“un- 5.1. Model choppedrange”casesseparately. Theresponsevariationscanbedescribedingeneralusingacom- binationofexponentialfunctions.Wefoundthatacombination ofthreeexponentialfunctionscloselydescribetheseveraldiffer- 5. Detectingandmodelingtransients entcasesoftransientsobserved. To detect, model, and correct transients in the signal, one has to decouple the signal and the effect of the transients. The best waytoproceedistoobtainanestimateoftheresponsebynor- f(t)=a+ashorte−t/τshort +amediume−t/τmedium +alonge−t/τlong (5) malizingthesignaltotheexpectedspectrum.Aswewillseein Theparameteracorrespondstothemedianlevelofthenor- thefollowingsections,ifthesourceemissionisnegligiblewith malizedsignal.Inthecasethattheresponseisobtainedbynor- respecttothetelescopebackground,theidealwaytoproceedis malizing the signal to the last scans, a is fixed to unity by con- tonormalizethesignaltotheexpectedtelescopebackground.If struction. The three exponential functions have very different the source emission is not negligible, one has to have a better timescales.Afirstfunctionaccountsforthefastvariationofthe guessofthesignal.Onepossibilityistoassumethatthedetector signal just after the flux change or a cosmic hit. The timescale eventually stabilizes after a jump in the continuum, so that one is between 1 and 10 seconds. The second term has an interme- canusethelastscansneartheendoftheobservingsequenceto diate timescale (between 10 and 80 seconds). Finally, the third estimate the “true” continuum level. This asymptotic level can term accounts for the long-term variation of the signal and has then be used to normalize the entire signal to reveal the jump– the longest timescale (between 200 and 1200 seconds). In the transient.Oncethetransientisrevealedinthismanner,itcanbe caseofcosmicrayinducedtransients,thea coefficientisal- short fittedandremovedfromthesignal. wayspositive,whereasforthecontinuum-jumptransientsa short Once the transients due to expected continuum-jumps are canbeeitherpositiveornegativedependingonthedirectionof corrected, it is possible to obtain a first estimate of the final the jump. The parameter space considered for the fitting in the spectrum,combiningalldatafromallindependentupanddown twocasesissimilar. scans,andmeasurementsmadewithall16pixelscombinedto- In the case of cosmic rays, the time constant τ is kept short gether. Oncethisaverage spectrumis created,itcan beused to lower than one second. For the continuum-jump transient, the Articlenumber,page6of14 DarioFadda etal.:TransienteffectsinHerschel/PACSspectroscopy parameter space for τ considered is between 1 and 10 sec- short onds.Fortheothertwotimeconstants,theparameterboundaries are10<τ <80secondsandτ >100seconds. medium long 5.2. Correctingcontinuum-jumptransients If the continuum of the source is negligible with respect to the telescope background, it is possible to see the long-term transient caused by the calibration block all along the observation bynormalizingthesignaltotheexpectedtelescopebackground (seeFigure3).Ifthiseffectisnotcorrectedandanaverageback- ground is subtracted from the signal, there will be parts of the image with a signal lower than the average background which will have negative fluxesat the end of the datareduction. Even worse, in mapping observations, each raster position taken at a different phase of the transient will have a different continuum levelresultinginanartificialgradientinthefinalimage(seelater inFigure15). Unfortunately, this correction depends on the knowledge of thetelescopebackgroundemission.Thecurrentmodelavailable throughHIPEisfairlyaccuratefortheredarray,whilethesitu- ationisslightlyworseforthebluearray. There are several situations where the simple correction above will not work. In the case of a very bright continuum, thenormalizationofthesignalbythetelescopebackgroundwill not be appropriate, and this will make any correction difficult. Anothercaseisthatofamulti-bandobservationtakenconsecu- tively.InFigure5weshowanexampleofanobservationtakenin 3differentbands.Atthebeginningofeachobservationalblock (observationofoneband)thereisacleartransientintroducedby thesuddenvariationofthecontinuumlevel. Fig.6.Anexampleofhowweestimatethe“Guess”spectrumforthe Tocorrectthetransientsinbothofthesecases,wecannor- C+ line of Arp 220 (obsID 1342202119). We compute the histogram malizethesignaltothefluxdetectedduringthelastwavelength of the normalized χ2 of each individual spectral scan with respect to themedianspectrum(toppanel).Thebluelineshowsthewellbehaved scanofeachobservationalblock.Wemaketheassumptionthat scansandtheredlineshowsthosetoberejected.TheGuessspectrum the detector response has stabilized by this time. This assump- (bottompanel;greenline)istheresultofco-addingthewellbehaved tionworkswellinthecaseoflongobservations(∼1hr).Inthe scans(bluelines).Thescanswithχ2 deviatingmorethan3σfromthe caseofshorterobservations,alittlebiaswillremaininthedata medianvaluearediscarded(redlines). sincetheasymptoticlevelhasnotyetbeenreached. weobtainanapproximationtothefinalspectrumcladdingallthe 5.3. Correctingtransientscausedbycosmicrays individual pixels that contribute to the scan (the “Guess” spec- Themostdisruptivetransientsarecausedbytheimpactofcos- trum). Next, the signal for each individual pixel is normalized mic rays on the detectors. Since they happen at random times by this spectrum to obtain an estimate of the response. At this withdifferentenergies,theycauseunpredictablechangesinthe point,thediscontinuitiesintheresponseareidentified,andonly baseline of the data which can last from several minutes up to the most significant are selected. We fit the model (equation 5) onehour.Themaindifficultyinfittingthesetransientswithour totheresponseintheintervalbetweentwoconsecutivediscon- model is to disentangle the real signal from the variation in re- tinuities. A correction is then applied to the original signal to sponsecausedbythecosmicrayimpact,andtodetectthetime attempttoremovetheeffectofthecosmicrays.Inthefollowing, whentheimpactoccurs.Cosmicrayeventsaresofrequentthat, wedescribeindetaileachoneofthesesteps. inpractice,wemustsetathresholdabovewhichweattemptto correctthesignal. 5.3.1. Estimatingthe“Guess”spectrum Lebouteiller et al. (2012) developed a procedure to identify features associated with transients using a non-parametric We assume that the continuum jump transients have been cor- method(multi-resolutiontransformofthesignalandidentifica- rected to this point. We need to identify the scans affected by tionofpatternsonatypicalscale).Thefeaturesfoundwerethen severe transients, and discard them to obtain a good first guess fitted with an exponential, and subtracted from the signal. This of the spectrum. This step is particularly important in the case algorithm is identical to the one developed for the treatment of oflowredundancy,sinceevenafewbadscanscancompromise ISOCAMdata(Starcketal.1999). thecoaddedspectrum.Theproceduretoobtainaguessspectrum Theapproachwedescribedinourpaperisdifferentinboth isiterative.Thefirststepconsistsofcomputingamedianspec- the way we identify the cosmic ray events, and how remove trumbycoaddingallthewavelengthscansforthespectralpixels them.OurmethodsaresimilartothosedevelopedbyLarietal. whichcontributetoagivenspaxel.Thespectrumisobtainedby (2001) for the treatment of ISOCAM data, although the algo- computing the median of all the valid fluxes for each bin of a rithms presented here differs in several significant details. First wavelength grid: the median spectrum. Scans with very devi- Articlenumber,page7of14 A&Aproofs:manuscriptno.DarioFadda 240 235 0.12 230 225 se0.08 n o 220 p es0.04 215 r d e 210 er 0.00 205 Filt −0.04 200 195 −0.08 190 1.3 1.20 1.15 1.2 1.10 e nse1.05 ons o p1.1 esp1.00 Res R 0.95 1.0 0.90 0.85 0.9 4200 4400 4600 4800 5000 5200 4500 4600 4700 4800 4900 5000 Frame No. Frame No. Fig.7.Toppanel:Acomparisonoftheobservedsignalofasinglespec- Fig.8.Onthetoppanel,partoftheresponseconvolvedwiththefirst tralpixel(cyan)withthe“Guess”signal(blue;seetext).Thesignalis derivative of the Gaussian with dispersion equal to two frame units. plottedagainsttheFrameNumber,whichisafunctionoftimeduring Thiswidthgivesthebestsensitivitytosuddenchangesoftheresponse the scan. We show three scans in the figure (two up and one down) caused by cosmic ray hits on the detectors. The green lines mark the whichcapturestheC+lineshowninthepreviousfigure.Justafterthe region of ±5 times the local dispersion and the red circles marks the framenumber4720astrongcosmicrayhitsthedetectorproducinga startingpointsoftransients.Inthebottompanel,theresponsivityforthe suddenchangeoftheresponseofthedetector.Bottompanel:Theratio sameblockofframeswitheachintervalbetweentwoconsecutivedis- ofthetwocurves,i.e.theresponseofthepixelasafunctionofframe continuitiesintheresponsefittedwithourtransientmodel(darkgreen). number.Manytransientsareclearlyvisible,inparticularthehugetran- Theredpointsmarktheframemaskedbecausetheintervalbetweentwo sientafterframenumber4720. consecutivediscontinuitiesistooshort(lessthan10frames). ating values compared with the median spectrum are rejected. Thescanswithχ2 valueswhichdeviatemorethan3σfromthe naliscomparedwiththesignalfromasinglepixel,allowingthe median of the χ2 distribution are discarded and a new median discontinuitiestobeeasilyseen. spectrum is computed (see top panel of Figure 6). To obtain a Tocorrectlyfitthetransients,onehastoidentifythestarting robust estimate of the dispersion of the distribution, we make point,i.e.themomentatwhichthediscontinuityinfluxappears useofthemedianabsolutedeviation(orMAD).Theprocedure inthesignal.Thisisaclassicalproblemofsignaltheoryandit isiteratedthreetimestoobtainacleanmedianspectrumwhich hasbeenshown(Canny1986)thattheoptimalfiltertofinddis- serves as “Guess” spectrum. Figure 6 shows visually how the continuities in a mono-dimensional signal is the first derivative process works in the case of an observation of the C+ line of ofaGaussian.ThedispersionoftheGaussianwasfoundempiri- Arp220.Theprocedureworksbecauseofthehighredundancy callytobetwoframeunits(equivalentto0.25s).Ifweconvolve of the PACS data, which has a minimum of four up-and-down theresponsewiththisfilter,spikesappearjustaftercosmicrays scansforeachofthe16spectralpixels. impactthedetector.AsshowninFigure8,wecanselectthemost importantonesbycomparingtheirintensitytothelocalnoisein ourresponseestimate.Inouralgorithmwemakeacutof5σto 5.3.2. Findingsignificantdiscontinuities selectthesignificantevents. Bynormalizingthesignalofeachpixelbytheexpectedfluxof the Guess spectrum, we obtain an estimate of the response as 5.3.3. Maskingandcomputingtheresponse a function of time for each pixel. This function can be used to study the effect of cosmic rays on the signal by selecting the Atthispointwe breakthesignalinintervalsbetweentwo con- mostimportantones,fittingthetransientsafterthem,andeven- secutive discontinuities, and fit the model (equation 5) in each tuallymaskingpartofthesignalexcessivelydamagedbytheef- oneoftheseintervals.Ifthecosmicraysoccurtooclosetogether fectofcosmicrayimpactsonthedetector.InFigure7weshow (less than 10 frame units) we simply mask that part of the sig- an example using the M82 C+ spectrum. Here the Guess sig- naltimeline.Theresultingmodelresponseandmaskedregions Articlenumber,page8of14 DarioFadda etal.:TransienteffectsinHerschel/PACSspectroscopy 200 190 180 170 160 x u l 150 F 140 130 120 SPG 110 190 180 170 160 x u 150 l F 140 130 120 TC Fig.10. Signalvariationsasafunctionofthewavelengthscan.Tosee 110 the effect of the flux variations on the measured signal, we can com- parethesignalmeasuredduringtheup-scans(blueparts)withtheone 100 1800 2000 2200 2400 2600 2800 measuredduringthedown-scans(redparts). Frame No. lengthscanthereisalmostnovariationintheincidentfluxwhich Fig. 9. Comparison between masking in the SPG pipeline (top) and couldcausetransienteffectsonthesignal.Thisisnotthecasein transient-correction pipeline (bottom). Signal is blue and masked observations of extended wavelength ranges, where the contin- frames are represented as red circles. The SPG pipeline masks a few deviant frames as GLITCHES, while the transient-correction pipeline uum can vary substantially during the scan. This leads to tran- masksframeseitherasGLITCHESorasUNCORRECTED.Asaresult, sient behaviour which becomes apparent when comparing the thesignalfromtheSPGpipelinewillcontainunmaskedresidualsofbig scansintwooppositedirections.InFigure10,weshowthatdur- transients,whilethetransient-correctionpipelinewilleithercorrectthe ing up and down scans, the incident flux can change by, in this transientsormaskedthemcompletely.Thisleadstoabigreductionof example,afactorof7.Thisleadstoalargetransient. noise in the final reduction with respect to the product from the SPG In Figure 11 (top panel) the ratio of the two fluxes shows pipeline. clearly the effect. Here we consider only one scan for a given spaxel.Fortheup-scan(bluepoints),weseethefluxdecreases arethenappliedtotheoriginalsignaltoremovetheeffectofthe aswescantolongerwavelengths.However,whenwescanback with a down-scan, (red points) there is a systematically lower transients.Thisautomaticallycorrectsthesignalforvariationsin signal compared with the up scan. This is due to transient be- theresponseaftercosmicrayimpacts,aswellascorrectingfor largeoffsetsinindividualscansrelativetothemediansignal. haviorcausedbytherapidlychangingfluxduringthescans.The effect is clearly shown in the ratio of the up and down scans Onemaindifferencewiththereductionpipelineforchopped (bottom panel of Figure 11). When the flux is changing very spectroscopyisthat,atthispoint,thereisnoneedfortheappli- rapidly (between 145–190 µm), the up-scan flux is 10% higher cationofaspectralflatfieldcorrection.Inthestandardpipeline, thanthedown-scanflux.However,from190–210µm,wherethe this is a separate task that takes into account small variations fluxchangeismuchsmaller(onlyafactoroftwo),theeffectof oftheresponseinthedifferentspectralpixels.Ourmodelingis thetransientisalsosmaller(lessthan2%). able to automatically rescale all the pixels relative to the same Before we describe in detail how we apply a correction for spatial module in each observational block without any further thiskindoftransient,wefirstneedtodemonstratethatthetran- operation. sient caused by a sudden increase in continuum has a similar The net effect of all our corrections is to significantly de- formtothatofasuddendecreaseincontinuum.Wewillusethis crease the noise in the final coadded spectrum compared with symmetricalbehaviouroftransientstoprovideasolutiontothe the SPG pipeline. Although the SPG pipeline attempts to mask problem. To prove this symmetry, we consider an observation largertransients,importantresidualsremain(seeFigure9). whichwasespeciallydesignedtoinducetransientbehaviourof thedetectorsduringobservationalday27(obsID1342178054). In this observation, the grating was moved to several different 5.4. Correctingscandependenttransients positions to have different levels of flux on the detectors. After For many of the PACS spectra, only a small wavelength range each move, the grating was halted for 4 minutes to allow sta- is observed. In this case we can consider that the continuum is bilization. We show in Figure 12 a part of this observation to essentially constant during the observation. So, during a wave- demonstratetheresponsetosuddenjumpsinrecordedflux.The Articlenumber,page9of14 A&Aproofs:manuscriptno.DarioFadda 1.1 malized Flux01..90 Nor 0.8 1.2 0 400 800 Frame No. 0.9 x u Fl 0.6 0.3 2500 3500 4500 Frame No. Fig. 12. Transient effects on the measured signal when changing be- tweentwogratingpositionsperformedduringanengineeringobserva- tion during OD 27 (see text). In the inset, comparison of the upward (blue)anddownward(red)transients.Thedownwardtransienthasbeen flipped to compare the two behaviours. For a regime of flux changes similar to the one found in the range observations, the two transients followessentiallythesamebehaviour.Thisjustifiesthechoiceofget- Fig.11. Toppanel:fluxintheup-scan(blue)comparedtothefluxin tingthemediansignalbetweenup-anddown-scansignals. thedown-scan(red).Whentheincidentfluxisdecreasing(bluelinebe- tween140and195µm)themeasuredfluxishigherthanthatmeasured when the flux is increasing (red line in the same wavelength region). of the symmetry, we can assume that the differences between Thesituationisinvertedbetween195and205µm,whilethetwofluxes thefluxesinFigure11isentirelyduetotransientbehavior.This areverysimilarinthewavelengthregionbeyond205µmwherethereis means that a reasonable solution is to scale the up- and down- nobigvariationinflux.Bottompanel:ratiobetweenup-anddown-scan scan values of the flux to the average of the two values at each measuredfluxesforallthe16pixelinaspacemodule.Thebehaviouris wavelengthinterval.Withthissimplesolution,accordingtothe verysimilar. observationinFigure12,thefluxrecoverediswithinafewper- centoftheasymptoticincidentflux.Wenotethatachangeofa factor of three in the continuum in one minute (as in the above figureincludesthesignalreadoutjustbefore(blackline)asud- example)isanextremecase.Inmostobservations,thechanges denchangeinthepositionofthegrating,whichtheninducesan influxthroughoutascanarelikelytobemuchless. upward transient and new stabilization (blue line). After a suit- ableinterval,thegratingischangedbacktotheoriginalposition leading to a downward jump and transition (red line), followed 6. Examples againbyaperiodofstabilization.Toinvestigatetheshapeofthe upwardanddownwardtransientsinducedbythischangeincon- Inthefollowingsectionweprovideafewexampleswhichshow tinuum, we show in the inset figure, the normalized signal for the improvements introduced by our transient-correction algo- the upward– and downward–going signal, but with the down- rithms, as compared with the standard unchopped-mode HIPE ward signal flipped in sign to allow a close comparison. If we analysis(SPGpipeline,vers.14).Weremindthatitisimportant ignorethecosmicraytransientswhicharepresent,thebehavior toavoidexecutingpipelinescriptsblindlywithoutexaminingthe oftheupward–going(blue)anddownward–going(red)transient effects of the corrections in the different steps of the reduction responsesarealmostidentical. process. In the above example, the largest part of the transient re- In this section we also show a few comparisons of obser- sponse occurs within a time interval of approximately one vations made with unchopped and chop-nod mode of the same minute,andtheexamplewechosecorrespondedtoachangein targets,toshowhowthereductiondescribedinthispapergives flux of about a factor of three. We recall that in Figure 11, the consistentfluxcalibrationforthesameobservedlines. up-and-down scan induced changes in continuum which were largerthanthisoverthewholescan(whichlastedtwominutes), 6.1. Line but are about the same order (roughly a factor of three) in one minute. Therefore, the test of the hypothesis that the symmetry Asanexampleoftheimprovementinthereductionofalineun- intheshapeoftheupwardanddownwardtransientisvalid. chopped observation, we show here an observation of Arp 220 If we make the assumption that indeed the upward– and (obsID1342202119).WeshowinFigure13(topleftpanel)the downward–going transients have similar behaviors (except for cloud of flux–points for the central spaxel of the spectrometer the sign), this suggests a workable way to correct for the tran- arrayafterpassingtherawdatathroughtheSPGpipeline.Inthe sientsinalongscanoverarapidlychangingcontinuum.Because toprightpanelofthesamefigure,thesameobservationwaspro- Articlenumber,page10of14