Spitzer IRAC observations of JWST calibration stars
Jessica E. Krick, Patrick Lowrance, Sean Carey, Seppo Laine, Carl Grillmair, Schuyler D. Van Dyk, William J. Glaccum, James G. Ingalls, George Rieke, Joseph L. Hora, Giovanni G. Fazio, Karl D. Gordon, Ralph C. Bohlin
DDraft version February 5, 2021
Typeset using L A TEX preprint2 style in AASTeX63
Spitzer IRAC Photometry of JWST Calibration Stars
Jessica E. Krick, Patrick Lowrance, Sean Carey, Seppo Laine, Carl Grillmair, Schuyler D. Van Dyk, William J. Glaccum, James G. Ingalls, George Rieke, Joseph L. Hora, Giovanni G. Fazio, Karl D. Gordon, and Ralph C. Bohlin IPAC, MC 330-6, Caltech, 1200E. California Blvd. Pasadena, CA 91125 Steward Observatory, The University of Arizona, 933 North Cherry Ave., Tucson AZ 85721 Center for Astrophysics | Harvard & Smithsonian, 60 Garden St., MS-65, Cambridge MA 02138, USA Space Telescope Science Institute, 3700 San Martin Dr., Baltimore, MD 21218
ABSTRACTWe present infrared photometry of all 36 potential JWST calibrators for which thereis archival Spitzer IRAC data. This photometry can then be used to inform stellarmodels necessary to provide absolute calibration for all JWST instruments. We describein detail the steps necessary to measure IRAC photometry from archive retrieval tophotometric corrections. To validate our photometry we examine the distribution ofuncertainties from all detections in all four IRAC channels as well as compare thephotometry and its uncertainties to those from models, ALLWISE, and the literature.75% of our detections have standard deviations per star of all observations within eachchannel of less than three percent. The median standard deviations are 1.2, 1.3, 1.1,and 1.9% in [3.6] - [8.0] respectively. We find less than 8% standard deviations indifferences of our photometry with ALLWISE, and excellent agreement with literaturevalues (less than 3% difference) lending credence to our measured fluxes. JWST ispoised to do ground-breaking science, and accurate calibration and cross-calibrationwith other missions will be part of the underpinnings of that science.
Keywords: infrared:stars INTRODUCTIONCalibration of an instrument proceedsthrough both observations of stars and mod-els of stellar flux density. Calibration stars arechosen to be those where the spectra can bemodeled to the 1% level (e.g., white dwarfs,A, G stars). However, these models need tobe verified and refined through observations.The observations output data in data number
Corresponding author: Jessica [email protected] units which then need to be converted into fluxunits. That conversion comes from comparingthe observed data numbers to predictions of theflux densities in Janskys (Jy) at the effectivewavelength of the observation. Predicted fluxdensities (hereafter we use “flux” for flux den-sity in Jy) are derived by integrating the modelswith the relative spectral response curves for theinstrument.Each telescope’s absolute calibration dependson the accuracy of the models. These mod-els need to be grounded in truth observations a r X i v : . [ a s t r o - ph . S R ] F e b Krick et al. at or near the wavelengths where the modelswill be used. Our goal in this paper is to in-form these models at wavelengths similar to theJWST wavelengths by providing IRAC fluxesfor a set of calibration stars.We need absolute photometry to be able tocompare with either physical models or mea-sured values in different modes or with differentinstruments. Scientific applications for absolutephotometry are many and varied, including ev-erything from solar system objects and zodiacallight to the extragalactic background light andsupernovae.The Infrared Array Camera (IRAC; Fazioet al. 2004) was operational on the Spitzer spacetelescope (Werner et al. 2004) from 2003 - 2020with 4 broad mid-infrared bands with responsecovering 3.15 - 9.25 µ m. The James Webb SpaceTelescope (JWST; Gardner et al. 2006) is cur-rently scheduled to launch in 2021 and will ob-serve from 0.6 - 28.3 microns. This paper pro-vides high precision IRAC fluxes, where avail-able, at 3.6, 4.5, 5.8, and 8.0 µ m (also denotedch1 - ch4 respectively). These filter names arelabels and are not the actual effective wave-lengths (for more detailed information on filtertransmission see Hora et al. 2008). The detectorarrays in IRAC are 256 ×
256 pixels with 1.2 arc-seconds per pixel. Subarray observations use amode where 64 consecutive 32 ×
32 pixel imagesare taken at a higher readout rate without mov-ing the telescope.We attempt to include in this work all cur-rently available, potential JWST calibrationstars. These are listed in Table 1. The sourceof this list is the James Webb Space TelescopeUser Documentation on Absolute Flux Calibra-tion (Institute 2016). Stars are chosen to havea range of flux levels to be observable with asmany modes of the four JWST instruments aspossible. This paper works towards a high levelgoal of providing an accurate cross-calibrationof Hubble-Spitzer-Webb. Having observations of the same stars taken with all three observato-ries will provide the basis for a strong combinedcalibration of these three NASA observatories.We use all possible archival IRAC data for thiswork including both the cryogenic mission (Aug2003 - May 2009) and the warm mission (July2009 - Jan 2020). When the cryogen on Spitzerwas depleted in May 2009, the IRAC instru-ment warmed up, which rendered two channelsinoperable ([5.8] & [8.0]), and changed the cal-ibration of the remaining two channels ([3.6] &[4.5]). We are careful to use the appropriatecalibrations for the appropriate mission.In Section 2 we describe the archival data usedfor this project. Section 3 covers our meth-ods for reducing the data, measuring photome-try, and applying photometric corrections. Sec-tion 4 discusses how we validate our photom-etry by studying the uncertainties, comparingthe photometry with model fluxes, Wide-fieldInfrared Survey Explorer (WISE; Wright et al.2010) fluxes, and the literature. We make con-cluding remarks in Section 5. THE DATAWe use publicly available data from theSpitzer Heritage Archive (SHA) availableon the NASA/IPAC Infrared Science Archive(IRSA). Since the Spitzer mission is now com-plete, this work includes all available IRAC dataon these targets. We include all data availablein the archive regardless of actual observed tar-get, which means we include data where the tar-get was not the calibration star, but the cali-bration star was observed serendipitously. Thisadds complication to our work because not allobservations were optimally designed for abso-lute photometry and/or do not have optimal ex-posure times to maximize signal to noise ratioon the calibration star. Specifically, in many https://sha.ipac.caltech.edu/applications/Spitzer/SHA/ pitzer IRAC observations of JWST calibration stars § µ m MIPS band (Zellem et al. 2014;Crossfield et al. 2012). Furthermore the flux ofthe star changes by 0.15% at secondary eclipse,and the same amount throughout its orbit asthe phase of the planet changes. While this issmaller than the desired 1% photometry for acalibration star, we feel that the planetary com-panion makes this star not a suitable choice forcalibration.There is a time frame early in the warm mis-sion (BMJD 54966.5 to 55093.5) when the tem-peratures and bias levels were changing signifi-cantly. As the instrument calibration varied byseveral percent during that period we do notinclude any data taken during that time.We define here the Spitzer IRAC nomencla-ture for observations and file naming. An in-dividual Spitzer observing sequence is an AOR(Astronomical Observation Request). A singleAOR can be any reasonable number of individ-ual data images (frames) and can include allfour channels. Templates for the AORs wereprovided by the Spitzer Science Center, butthe AORs themselves were designed by the ob- servers. The individual data frames generatedby the Spitzer IRAC pipeline are called BasicCalibrated Data (BCD) and are formatted asFITS files. They have had instrumental sig-natures removed (darks, flats, etc.), and havebeen calibrated using an absolute calibrationinto physical units. The uncertainty image cor-responding to an individual frame is a BCD un-certainty (BUNC) FITS file (see the IRAC In-strument Handbook for a detailed descriptionof the pipeline and data products). PHOTOMETRYFor an overview of the absolute photometriccalibration of Spitzer IRAC, see Reach et al.(2005) and Carey et al. (2012) We include thespecifics of our photometry pipeline especiallywhere it differs from these previous works.The following is an overview of our data pro-cessing steps: 1) download data, 2) removezeroth frames (the first image taken in a se-quence), 3) convert images to units of electrons,4) measure centroids for our target stars, 5) per-form aperture photometry, 6) apply photomet-ric corrections.From the SHA, we download the level 1 prod-ucts (BCD & BUNC) FITS files as well as theraw data. The raw data are the definitive way ofdetermining if a target star is saturated. BCDframes have a saturation correction applied tothem so it is not always clear when a star hassaturated; however, the raw data counts willshow it. High accuracy photometry on mosaicsis not recommended (see Hora et al. 2008, for athorough discussion of this).We remove the zeroth frames in each set ofobservations. The zeroth frames of every obser-vation request (AOR) have a different per-pixelbias than other frames. This is caused at leastin part by a significantly different delay between https://irsa.ipac.caltech.edu/data/SPITZER/docs/irac/iracinstrumenthandbook/ Krick et al.
Table 1.
JWST calibrators observed with Spitzer IRACStarname RA Dec Spec. V K A V Alternate(h m s) ( ◦ (cid:48) (cid:48)(cid:48) ) Type (mag) (mag) NameHD2811 00 31 18.49 −
43 36 23.0 A3V 7.50 7.06 0.26HD14943 02 22 54.67 −
51 05 31.7 A5V 5.90 5.44 0.032MJ03323287 03 32 32.88 −
27 51 48.0 F6V 16.64 14.82 0.21 C26202G191B2B 05 05 30.62 +52 49 54.0 DA0 11.78 12.76 WD 0501+527lam lep 05 19 34.52 −
13 10 36.4 B0.5V 4.29 5.09 HD34816HD37962 05 40 51.97 −
31 21 04.0 G2V 7.85 6.27 0.04HD37725 05 41 54.37 +29 17 50.9 A3V 8.31 7.90 0.13mu col 05 45 59.89 −
32 18 23.2 O9.5V 5.18 5.99 HR1996, HD38666HD38949 05 48 20.06 −
24 27 49.9 G1V 7.80 6.44 0.00GD71 05 52 27.51 +15 53 16.6 DA1 13.03 14.12 WD 0549+158eta01Dor 06 06 09.38 −
66 02 22.6 A0V 5.69 5.75 0.00 HR2194, HD42525HD55677 07 14 31.29 +13 51 36.8 A4V 9.41 9.16 BD+14 1605WD1057+719 11 00 34.31 +71 38 03.3 DA1.2 14.80 15.47HD101452 11 40 13.65 −
39 08 47.7 A9mIV 8.20 6.82 0.12HD106252 12 13 29.51 +10 02 29.9 G0 7.36 5.93 0.00GD153 12 57 02.37 +22 01 56.0 DA1 13.35 14.31 WD1254+223HD116405 13 22 45.12 +44 42 53.9 A0V 8.34 8.48 0.00HR5467 14 38 15.22 +54 01 24.0 A1V 5.83 5.76 0.03 HD128998HD146233 16 15 37.27 −
08 22 10.0 G2V 5.50 4.19 HR 6060, 18 Sco2MJ16313382 16 31 33.85 +30 08 47.1 G2V 12.92 11.37 0.11 P330EBD+60 1753 17 24 52.27 +60 25 50.7 A1V 9.64 9.64 NPM1+60.0581HD158485 17 26 04.84 +58 39 06.8 A4V 6.50 6.14 0.14 HR6514HD159222 17 32 00.99 +34 16 16.1 G1V 6.56 5.00 0.00 HR6538HD166205 17 32 13.00 +86 35 11.3 A1V 4.34 4.26 0.02 del Umi2MJ17325264 17 32 52.64 +71 04 43.1 A4V 12.21 12.25 0.11 TYC 4424-1286-1HD163466 17 52 25.37 +60 23 46.9 A2 E 6.85 6.34 0.102MJ17571324 17 57 13.25 +67 03 40.9 A3V 12.00 11.16 0.11 TYC 4212-455-12MJ18022716 18 02 27.17 +60 43 35.6 A2V 11.98 11.83 0.02HD165459 18 02 30.74 +58 37 38.2 A1V 6.86 6.582MJ18083474 18 08 34.75 +69 27 28.7 A3V 11.69 11.53 TYC 4433-1800-12MJ18120957 18 12 9.56 +63 29 42.3 A3V 12.01 11.29 0.02 TYC 4205-1677-1HD180609 19 12 47.20 +64 10 37.2 A2V 9.42 9.12 0.11 NPM1+64.0581HD186427 19 41 51.97 +50 31 03.1 G3V 6.20 4.65 16 Cyg BLDS749B 21 32 16.01 +00 15 14.3 DBQ4 14.67 15.22 WD 2129+00HD205905 21 39 10.15 −
27 18 23.7 G2V 6.74 5.32 0.01HD214680 22 39 15.68 +39 03 01.0 O9V 4.88 5.50 10 Lac pitzer IRAC observations of JWST calibration stars Table 2.
JWST calibratorswithout Spitzer IRAC dataStarnameHD15318HD27836HD60753HD115169HD1423312MASS J15591357+47364192MASS J16181422+00000862MASS J16194609+5534178WD1657+3432MASS J17430448+66550152MASS J18052927+6427520HD167060 the zeroth frame and the previous one at the endof previous AOR as opposed to the delays be-tween subsequent frames within the AOR dueto the larger slew time to acquire the new tar-get. This larger delay time introduces a largeper-pixel bias offset in the frame (sometimesreferred to as the‘first frame effect”) which isnot as well characterized nor removed by thebias subtraction part of the pipeline producingthe BCD. Consequently, attaining the highestpossible accuracy requires that we ignore theseframes.The zeroth frame is not the only frame af-fected by different delay times. In particularstaring mode data (non-dithering) is processedin the pipeline in the same manner as dithereddata, including using a dark image which wasmade by dithering. That dithered dark datawill therefore have different (longer) delay timesthan the staring mode data. To correct for this,we use calibration observations in PID 1345 tomake our own staring mode dark suite fromnon-dithered data. We then use this staringmode dark on the staring mode data by first removing the dithered dark and then applyingthe staring dark.We use coordinates for the target stars fromSIMBAD to identify the calibration star in eachframe. For all observations, we calculate thecentroids of the target stars using the first mo-ment box centroider available in IDL on theIRAC contributed software website . We usea box width of 7 pixels with a background boxwidth of 6 pixels separated by 3 pixels from thecentroiding box.We convert the BCD images into units of elec-trons to enable estimation of the Poisson noisefor each BCD . The conversion from MJy/sr(units used in the BCD images) to electrons is e = M J y/sr ∗ gain ∗ exptime/f luxconv where gain, exptime and fluxconv are keywordsin the BCD headers and depend on exposuretime, channel, and mission phase (warm orcryo). The pixel gain is subject to approx-imately 10% uncertainties, but is a constantterm that affects all observations equally.We then perform circular aperture photome-try on the images in units of electrons at thereturned centroid locations. We use a 3 pixelaperture radius with 3 - 7 pixel aperture back-ground. We choose this smaller aperture sizeto improve signal to noise ratio of the detec-tions. Specifically we are using IDL aper withthe keywords /exact and /flux set which doesa better job of calculating the intersection of acircular aperture with square pixels and keepsmeasurements in flux units instead of convert-ing to magnitudes. The background is calcu-lated using a three sigma clipped mean. Wetake read noise values from the image headers. https://irsa.ipac.caltech.edu/data/SPITZER/docs/irac/calibrationfiles/pixelphase/box centroider.pro https://idlastro.gsfc.nasa.gov/ftp/pro/idlphot/aper.pro Krick et al.
We apply four photometric corrections; thearray location-dependent correction , the pixelphase correction , an aperture correction, anda time-dependent correction (for [3.6] & [4.5]only) as described below. Array location andpixel phase effects are both largely mitigatedwith a good dithering strategy. However, sincemany of the archival AORs were not designed ascalibration observations and have non-optimalmapping and dither strategies, it is necessary tomake these corrections. We take care to use thedesignated corrections for cryogenic and warmdata as appropriate.The array location-dependent correction takesinto account the variation in system response ofthe instrument across the field of view primar-ily due to the change in angle of incidence oflight through the bandpass filter as a functionof position on the array. This variation mani-fests itself mainly as a shift in the location of thefilter edges as a function of wavelength. The useof zodiacal light as a source of flux for determin-ing the flat fields and normalizing the per-pixelresponse maximized this variation for stars likethe calibrators used in this paper.The pixel phase correction accounts for chang-ing gain as a function of position within apixel coupled with the undersampling of a pointsource by IRAC especially at the shortest wave-lengths. There is no pixel phase correction for[5.8] & [8.0].Both the array location-dependent correctionand pixel phase correction are corrected withthe SSC provided code irac aphot corr.pro .The IRAC calibration is referenced to a 10pixel aperture with 12 - 20 pixel background https://irsa.ipac.caltech.edu/data/SPITZER/docs/irac/calibrationfiles/locationcolor/ https://irsa.ipac.caltech.edu/data/SPITZER/docs/irac/calibrationfiles/pixelphase/ https://irsa.ipac.caltech.edu/data/SPITZER/docs/dataanalysistools/tools/contributed/irac/iracaphotcorr/ annulus. We apply an aperture correction toour photometry done with smaller aperture andbackground annulus. Values for this correctionare taken from the instrument handbook andare [1.125, 1.120, 1.135, 1.221] for the cryogenicmission in [3.6] - [8.0] respectively, and [1.1233,1.1336] in the warm mission for [3.6] and [4.5].There is a known degradation in flux sensi-tivity with time over the mission for channels1 and 2 (Krick et al. 2016). This result comesfrom studying the observed signal as a functionof time of seven primary calibration stars binnedin two week intervals from the start of the mis-sion through 2016. This type of analysis hasnot been done for the [5.8] and [8.0] bandpassesso we do not make a correction to those ob-servations. However, since [5.8] and [8.0] wereonly operational for the roughly 5.5 year cryo-genic mission, the maximum correction wouldbe smaller than for the 16 year [3.6] and [4.5]data. We apply this 0.1% per year ([3.6]) and0.05% per year ([4.5]) correction to our photom-etry. While this correction is small, we do havephotometry for some stars which spans manyyears or even in a few cases (IRAC calibrators)the entire 16 year mission.We do not apply a color correction so ourfluxes are actually F*K* in the nomenclatureof the IRAC data handbook.To test that our photometry pipeline works,we have run our pipeline on a sample of fivestars used to do the absolute calibration ofIRAC. We choose these stars as our test datasetbecause they have well-studied, published ab-solute fluxes. The five stars are: ’KF09T1’,’NPM1p67’, ’KF06T2’, ’KF06T1’, ‘NPM1p68’.Our measured fluxes are within two percent offlux values in the absolute calibration papers(Reach et al. 2005; Carey et al. 2012).3.1. Rejections
We seek out and remove the following typesof higher uncertainty photometry in this order: pitzer IRAC observations of JWST calibration stars • BCD flagged data. We use the imask dataprovided by the SSC which to reject allpoints where the central pixel was foundin the BCD processing to be errant ineither having optical ghosts, stray light,saturation, muxbleed/bandwidth effect,banding, column pulldown, crosstalk, rad-hits, or latents. Additionally, we use theraw data frames and the table of maxi-mum unsaturated point sources per fram-etime from the IRAC Instrument hand-book to confirm any observations notflagged which were saturated. • centroids within 5 pixels of the edge of thearray. • signal-to-noise ratio of the individual pho-tometry points is less than 6.0. • outliers within each AOR that are greaterthan three sigma from the mean. Wedo this iteratively, where we reject threesigma sources, then re-calculate the meanand re-reject, until there are no more re-jections. • outliers within each star’s measurements(per channel) that are greater than threesigma from the mean. This is also doneiteratively.Lastly, there is one [3.6] AOR (40643584)found to have a relatively high flux persistentimage in it from a previous observation, sowe rejected the entire AOR from consideration.While it may have been possible to subtract outan average latent image, the star (HD180609)has 1005 other observations in [3.6], so usingthose to calculate a weighted mean flux is abetter option than adding the uncertainty frompersistent image removal.After these rejections we have 62361 photom-etry points on 36 stars.Finally, we do not report fluxes for stars thathave fewer than three observations, or for which the final flux is within three standard deviationsof zero (non-detections). RESULTS & DISCUSSIONWe report weighted averages, standard devi-ations as percentiles of the weighted averages,and the final number of observations after allrejections in Table 3. Measurements leadingto the averages are weighted by one over thesquared uncertainty on each individual photom-etry point (Bevington & Robinson 2002). Thisassumes that the observed population is the par-ent population. Errors in the means can becalculated by dividing the standard deviationby the square root of the number of observa-tions. Blank entries in the table indicate ei-ther no IRAC data or non-detections. Note thatthere is a large disparity in the number of obser-vations per star as some of these stars were in-tentionally observed for this type of calibrationprogram, some were serendipitously observed inother programs, and some are also used as rou-tine IRAC calibrators (meaning they have a lotof observations).Median values of the distributions are verysimilar to weighted averages. We choose to re-port weighted averages because they account forthe known uncertainties when deriving the av-erage flux.To determine the quality of the photometry,we examine 1) the uncertainties in the mea-sured fluxes and do a comparison to 2) predictedfluxes in each band, 3) WISE fluxes in each bandand 4) comparison with literature values.The plots below were the main diagnostic forfinding cases where the IRAC photometry wasin some way amiss. Outliers in these plots wereexamined individually. In cases where the rootcause was determined to be one of the rejectioncriteria listed in Section 3.1, those observationswere removed from consideration. Some out-liers have reasonable distributions that are ex-pected based on signal to noise ratio or number
Krick et al.
Table 3.
Measured IRAC Fluxes for JWST calibratorsStar 3.6 µ m 4.5 µ m 5.8 µ m 8.0 µ mflux std. N flux std N flux std N flux std NmJy % mJy % mJy % mJy %HD2811 424.54 0.74 4 276.91 1.20 5 178.23 0.99 5 99.77 1.26 5HD14943 1888.33 1.66 2262 1207.59 2.05 22612MJ03323287 0.31 3.66 91G191B2B 2.02 1.51 113 1.28 1.80 63 0.80 8.75 56 0.45 13.88 34lam lep 2550.29 1.38 2318 1589.92 1.71 2324 969.77 5.76 63 540.18 3.38 63HD37962 878.91 2.41 252 555.71 4.55 252 211.11 6.86 209HD37725 192.51 0.92 534 125.63 0.88 629 80.22 0.51 51 45.25 0.92 57mu col 1043.31 0.71 2252 656.35 0.92 2261HD38949 771.74 2.82 251 490.68 5.18 252 198.68 4.29 82GD71 0.68 2.76 60 0.44 3.67 68eta01Dor 1394.50 1.76 752 886.70 2.47 501 557.91 0.91 252 314.91 0.63 252HD55677 61.03 1.01 739 39.79 0.99 719 25.24 0.75 86 14.03 1.21 74HD101452 338.31 1.34 5 218.33 0.38 5 122.02 1.05 5WD1057+719 0.14 3.96 48 0.11 5.13 45HD106252 1239.36 2.29 4755 749.77 3.52 251 278.21 6.87 251GD153 0.49 2.71 98 0.31 3.84 81HD116405 113.31 0.67 8 73.42 0.75 10 46.40 0.84 5 25.60 0.42 4HR5467 1335.03 0.76 866 865.59 0.81 755 537.85 0.89 564 304.84 0.62 566HD146233 7386.14 0.73 2512MJ16313382 7.74 1.08 56 5.01 0.90 10 3.17 0.68 5 1.90 13.68 4BD+60 1753 39.56 0.92 438 25.72 1.12 433 16.28 0.93 96 8.94 0.87 81HD158485 991.71 2.90 684 644.60 0.74 442 399.91 1.53 424 228.39 1.04 433HD159222 2777.04 1.27 4752 1719.22 1.87 252 626.51 3.79 252HD166205 5678.95 0.73 250 3658.35 0.97 2512MJ17325264 3.56 1.23 42HD163466 815.65 0.86 1049 529.82 1.08 1049 332.82 1.30 725 187.54 1.14 7382MJ17571324 9.65 0.56 14 6.28 0.74 15 4.02 1.39 24 2.22 1.86 242MJ18022716 5.11 0.79 130 3.31 1.13 138 2.13 2.71 117 1.17 3.11 164HD165459 652.32 0.72 7082 425.27 0.70 7516 268.67 0.58 30 150.07 1.17 302MJ18083474 6.55 1.73 1776 4.36 1.09 242MJ18120957 8.69 0.89 146 5.69 1.09 135 3.63 1.47 35 2.00 2.20 26HD186427 3913.80 0.92 252HD180609 64.38 2.98 1131 41.94 1.47 601 26.69 0.67 10 15.09 1.16 399LDS749B 0.25 5.26 71 0.16 7.59 67HD205905 2164.74 1.21 250 1369.37 2.44 252 500.80 4.01 250HD214680 1621.04 2.97 120 1024.87 2.50 63 640.01 7.74 60 343.79 5.08 63 pitzer IRAC observations of JWST calibration stars Uncertainties
We examine the standard deviations of all ourflux distributions as a function of flux. Theseare shown in Figure 1, color coded by channelon the left and by signal to noise ratio (SNR)on the right.The source of the multiple observed turn-offsto higher percentage uncertainty seen here (atthe lowest fluxes and again at around 500 -1000 mJy) is lower signal to noise observationsfrom different exposure times. Observationswith uncertainties in the 3 - 5% range are due toexposure times too short to accurately measurethe flux of the star. Those observations havesignal to noise ratios ranging from just aboveour threshold of 6 to a few tens. These are stillsolid detections so we include them in our table.The large majority of our standard devia-tions are less than 3% (76% of the detections).The median standard deviations per channel are1.2%, 1.3%, 1.1%, and 1.9% for [3.6] - [8.0], re-spectively.Some bright stars have larger than expecteduncertainties. This is caused by a combinationof residual effects that will bin down with moreobservations.One example of this is shown in Figure 2. Foreta 01 Dor, channel 2 measured fluxes (880 mJy)are plotted as a function of frame number. Thex-axis shows a sequential view of the photome-try and is not actually time as the AORs werenot taken consecutively. The first four chunksof frames are the 64 subarray frames at 0.1 sfor each of four BCDs in a single AOR with asignal to noise ratio on the individual points ofaround 160. It is not possible to dither withina set of 64 subarray frames. The measurementswith the large error bars in the right half of theplot are the 0.02 s frame times with signal tonoise ratio of around 20. The 0.02 s data aretaken in subarray mode with likewise relatively few dithers but appear to not have this 3-4%photometry difference. Other stars observed inthe subarray exhibit similar behavior at the 1 -4% level.The list of potential sources of statistical un-certainty that will bin down with more dataincludes inaccuracies in the flat field, residualpixel phase effect, latents, photon noise, lowlevel cosmic rays, and bias level offsets. Thepixel phase effect was derived as a correctionfor the average pixel, but each individual pixelmay have a different response. Bias level offsetsare caused by imperfect bias correction due touse of darks made from sky images.The standard deviation in Table 3 accuratelydenotes that we have not removed all sources ofuncertainty.4.1.1.
Systematic Uncertainties
Some sources of uncertainty can not be re-duced by adding more data (or not as quicklyreduced by adding more data). We are awareof three of these systematic uncertainties thataffect IRAC photometry. First, photometry indifferent observing modes (full vs. subarray) ortaken with different exposure times has a dif-ferent absolute level. From an analysis of starsobserved in the warm mission, we find this tobe at most a 3% effect in [3.6] and a 1% effectin [4.5] (Krick et al., future work). Proper datato quantify this effect does not exist at [5.8] &[8.0].Second, positive and negative persistent im-ages in either the photometry aperture or back-ground annulus in either the dark frame or thescience frames can lead to apparent flux in-creases or decreases for several hours after thebright star was observed. Low level persistentimages that we would not have detected in ourrejection work (see Section 3.1) can be on the0.1% level (Krick et al. 2016). This numbercomes from an analysis of all of the warm mis-sion frames used in creating a dark calibration.Note that uncertainty from persistent images0
Krick et al.
Figure 1.
Standard deviation in the photometry as a function of measured flux for all detections. Leftplot is color coded by channel. 76% of detections have standard deviations less than three percent. Rightplot is color coded by average signal to noise ratio of the photometry per star, per channel. This shows thatthose stars with higher standard deviations are derived from noisier observations
Figure 2. [4.5] flux as a function of frame num-ber for eta 01 Dor. These images were not takenconsecutively and so the x-axis is not time. Thefirst three sets of 63 images (with the zeroth frameremoved) are individual subarray images from oneAOR taken at 0.1 s. The last 200+ points are 0.02 ssubarray observations taken at 0.02 s. The tele-scope was dithered from one set of 64 images tothe next, but not within the 64 subarray images. can be reduced by making observations at mul-tiple times and not by taking more observationsat the same time.Third, at the demanding accuracy level of ourphotometry, there can be residual effects fromimperfect linearity corrections that cause small offsets in the results for the brighter stars rela-tive to those for the fainter ones.We know that photometric stability as a func-tion of time for IRAC is excellent, and the smallamount of degradation noted in Section 3 for[3.6] & [4.5] is accounted for in the photometrypipeline, so this will not be a source of system-atic uncertainty for those channels.4.2.
Comparison to Model Fluxes
To get a rough idea of whether or notour measured fluxes are in the correct range,we compare our measured fluxes to predictedfluxes from Kurucz–Lejeune atmospheric mod-els (Lejeune et al. 1997) as provided bythe Spitzer Flux Estimator For Stellar PointSources . Figure 3 shows the percentage differ-ence from the same model value as a function offlux using the same color scheme as in previousplots.All but a couple of stars have measured fluxeswithin about 10% of their predicted values. Be-cause this is just an estimate of the model fluxes,this level of agreement with the models is per-fectly acceptable. The original version of this https://irsa.ipac.caltech.edu/data/SPITZER/docs/dataanalysistools/tools/pet/starpet/index.html pitzer IRAC observations of JWST calibration stars Figure 3.
A comparison of this work with modelphotometry for all four channels. Model fluxes arederived from Lejeune et al. (1997). A rough agree-ment is expected and confirmed. plot was how we found many of the saturatedAORs since those stars had predicted fluxesthat were much higher than measured. ThoseAORs with saturated data are not included inthis analysis, as described in Section 3.1It is beyond the scope of this paper to workwith high resolution models for these stars suchas those found on CALSPEC (Bohlin et al.2014). Instead we hope our photometry willinform the next phase of this project whichis to pin down those models at these infraredwavelengths in preparation for the calibrationof JWST.4.3.
Comparison to WISE Fluxes
We use VizieR to collect ALLWISE magni-tudes (Cutri et al. 2013) for these targets. Acomparison to the WISE fluxes is shown in Fig-ure 4. Because the WISE and IRAC bandpassesare not the same, we expect to find differencesin the reported fluxes. We find an average dif-ference of about 10% between W1 & [3.6] aswell as W2 & [4.5]. The [5.8] and W3 bands aretoo different to warrant comparison. A detailedcomparison of the IRAC and WISE photometryincluding the instrument response functions isbeyond the scope of this paper. Instead we use this rough comparison to make sure there are nounwarranted trends or outliers in the relationbetween IRAC and WISE that would signal wehad a problem with our IRAC photometry.We find good agreement between the two tele-scopes, with standard deviations of the distri-butions of their differences at 2.5% for [3.6] -W1 and 7.8% for [4.5] - W2 The larger value at[4.5] is caused by saturation in W2. The causeof the dip in the [4.5] points and rise in the[3.6] points at higher fluxes is saturation in theWISE bands. WISE W1 & W2 saturate at 200- 300 mJy. While WISE employs a correctionfor saturation, points at fluxes larger than theseshould be interpreted with some degree of skep-ticism.There are two stars that have [3.6] & [4.5]measured fluxes quite discrepant from WISEphotometry: HD159222 and 2MASSJ18120957.HD159222 is saturated in W1 & W2. Wecheck how these individual stars compare to themodels and previously published photometry.HD159222 has a 5% and 7% difference from the[3.6] & [4.5] models and has no published IRACphotometry. 2MASSJ18120957 has a 0.9% &0.01% difference from the [3.6] & [4.5] modelsand was published by Bohlin et al. (2011) witha 1.8% and 1.0% difference from our measuredphotometry (see Section 4.4). Visual inspectionof some of the images from these stars revealsnothing out of the ordinary. We interpret thisdiscrepancy from WISE, with no other evidencefor photometry problems, to mean that thereis nothing suspicious about our photometry ofthese stars, but potentially future work fittingmodels to photometry should consider the dis-crepancy.4.4.
Comparison to Literature Values
We compare IRAC calibrations from Reachet al. (2005), Bohlin et al. (2011), and Careyet al. (2012) in Table 4. The values from Reachet al. (2005) are as tabulated in their paper andare based only on their A-type calibration stars.2
Krick et al.
Figure 4.
A comparison of this work with All-WISE photometry for the two channels where thebandpasses are most similar as a function of flux.Data points show the difference of [3.6] to W1 and[4.5] to W2. Note the relatively small scatter inthe relations between IRAC and WISE. The abso-lute level of the differences is caused by differingbandpasses between the two instruments.
Although they had also observed potential K-giant calibration stars, the results from the tworanges of spectral type disagreed and they de-cided that the A-star values were more reliable.The goal of Bohlin et al. (2011) was to testthe consistency of calibrations based on whitedwarfs, A-type, and G-type stars. The valuesin Table 4 are weighted averages from their Ta-ble 4. We assigned 2% errors to those values.The calibration of Carey et al. (2012) benefitedfrom improvements in reduction approaches forthe IRAC data and also made use of improvedK-giant spectral models (Engelke et al. 2006);the tabulated values are averages from the A-type and K-type stars. If only the A-stars areused, the results agree very closely with thoseof Reach et al. (2005), albeit with the caveatthat different array location dependent correc-tions and flat fields were used in the differentcalibrations. Our values in Table 3 are basedon the Carey et al. (2012) calibration; adjust-ing them to either of the others (or to any future
Figure 5.
A ratio of this work with Bohlin et al.(2011) photometry as a function of flux for all starsand channels where available. Error bars are errorson the mean. calibration) can be made according to the ratiosof the flux conversion factors.Table 4 shows excellent agreement among thethree calibrations. We now compare with thephotometry of individual stars in Bohlin et al.(2011). They published IRAC photometry ofnine of the same targets as in our sample withthe goal of cross calibration between the HubbleSpace Telescope and IRAC. Figure 5 shows ex-cellent agreement of our photometry with thatwork, except for small offsets that are reflectedconsistently for all the stars. These offsets justreflect the systematic differences in calibrations,as also shown in Table 4. The small scatter inthe comparison shows that the sets of photom-etry are consistent with each other to signifi-cantly better than 1%. The most discrepant[8.0] data point at 0.4 mJy is G191B2B whichhas a larger standard deviation (14% at [8.0]).The only star common to both datasets above100mJy is HD165459.4.5.
Relative Spectral Response Curves
The photometry points we present would needto be combined with spectral response curves tobe useful for careful modeling of the stellar flux.For the stars that were used as IRAC calibrators pitzer IRAC observations of JWST calibration stars Table 4.
Flux Conversion Factors for Different CalibrationsConversion values (MJy sr − DN − s)channel Carey et al. Reach et al. Bohlin et al.[3.6] 0.1069 ± ± ± ± ± ± ± ± ± ± ± ± and other stars with high N values in Table 3,observations were obtained at many different lo-cations on the array, and so an array-averagedspectral response curve can be used. We includethose here as Tables 5 - 8 in machine-readabletable format. There are four tables, one for eachchannel, with 2 columns; wavelength in micronsand spectral response in electrons per incom-ing photon. These response curves reflect ourcurrent knowledge of the telescope throughputand detector quantum efficiency. The responsecurves use measurements of filter and beam-splitter transmissions (Quijada et al. 2004) overthe range of angles of incidence corresponding toa distribution of incident angles across the fieldsof view of the IRAC detectors. For stars with alimited number of samples or sources that wereobserved at one or a few array locations, arraylocation-dependent curves would need to be ap-plied. These are available on the IRAC SpectralResponse page at IRSA . For detailed informa-tion on how these curves were derived and howto apply them, see Hora et al. (2008). CONCLUSIONSWe have described our methodology for de-riving fluxes from IRAC four band data on 36JWST calibration stars. We report weightedaverages and standard deviations for all detec-tions. https://irsa.ipac.caltech.edu/data/SPITZER/docs/irac/calibrationfiles/spectralresponse/ Table 5.
IRAC Channel 1 Full Ar-ray Average Instrument ResponseCurveWavelength Response( µ m) (electrons/photon)3.08106 0.0006473.08289 0.0006933.08473 0.0005873.08656 0.0006023.08840 0.0005623.09024 0.0006443.09208 0.0007703.09393 0.0005363.09578 0.0006043.09762 0.000754 · · · (This table is available in its en-tirety in machine-readable form.) This project was enabled by a rich archiveand includes data from programs designed todo absolute photometry as well as observationswhere the JWST calibrators were ancillary tar-gets. This non-uniform dataset created somedifficulties and in cases added to uncertaintiesin the fluxes where for example a single targetwas observed in different modes (subarray or fullarray) or exposure times, or strategies (dither-ing or staring). This rich archive also providedus enough data in some cases to reject somenon-ideal data (saturated or extra noisy due to4
Krick et al.
Table 6.
IRAC Channel 2 Full Ar-ray Average Instrument ResponseCurveWavelength Response( µ m) (electrons/photon)3.72249 0.0012413.72516 0.0011323.72784 0.0011593.73052 0.0012343.73321 0.0012413.73590 0.0012793.73859 0.0012573.74129 0.0012333.74399 0.0012933.74669 0.001263 · · · (This table is available in its en-tirety in machine-readable form.) Table 7.
IRAC Channel 3 Full Ar-ray Average Instrument ResponseCurveWavelength Response( µ m) (electrons/photon)4.74421 0.0000884.74856 0.0000904.75291 0.0000864.75727 0.0000774.76164 0.0000644.76601 0.0000624.77040 0.0000734.77479 0.0000714.77919 0.0000674.78360 0.000074 · · · (This table is available in its en-tirety in machine-readable form.) Table 8.
IRAC Channel 4 Full Ar-ray Average Instrument ResponseCurveWavelength Response( µ m) (electrons/photon)6.15115 0.0000806.15846 0.0001236.16578 0.0001756.17312 0.0001866.18048 0.0001596.18786 0.0001416.19525 0.0001056.20266 0.0000556.21009 0.0000866.21753 0.0001316.22500 0.000165 · · · (This table is available in its en-tirety in machine-readable form.) latents) which lead to cleaner, more accuratephotometry.In order to validate our IRAC photometry, weperformed the following steps:First we validate our pipeline by applying itto archival photometry of five IRAC calibrationstars and confirmed that we measure the samedistribution of fluxes as those provided in theliterature.Second, we validate our actual photometryby examining the distribution of uncertaintiesand confirm that we find expected levels in thestandard deviations. Approximately 75% of ourdetections have standard deviations less thanthree percent (the quoted IRAC precision level).The median standard deviations are 1.2, 1.3,1.1, and 1.9% in [3.6] - [8.0] respectively. Thosefluxes with uncertainties higher than three per-cent are mostly due to exposure times that weretoo short to achieve ideal signal to noise ratiosor caused by multiple different observing strate-gies for a single target. Both of these are likely pitzer IRAC observations of JWST calibration stars Facility:
Spitzer (IRAC)REFERENCES