Using IRAC Images

1
Using IRAC Images

• Sean J. Carey for the IRAC Instrument Support and
  Instrument Teams

2
Outline

• Overview of data
• Products to use
• Mask definitions
• Calibrations applied
• Absolute photometric calibration
• Performing photometry on IRAC data
• Array-dependent correction
• Pixel-phase correction
• PRFs (point response function PRFs are the
  observed PSFs, optical PSF convolved with pixel
  shape)
• Bias correction
• Color correction
• Extended source calibration
• Common data artifacts
• Electronic
• Optical
• Suggested mitigation methods
• Useful resources
• Documentation
• Software

3
Data Products

• Start with the basic calibrated data (BCDs)
• Image data (e.g. SPITZER_I1_6050048_0000_0000_1_bc
  d.fits)
• Image uncertainties (e.g. SPITZER_I1_6050048_0000_
  0000_1_bunc.fits)
• Image masks (e.g. SPITZER_I1_6050048_0000_0000_1_b
  dmsk.fits)
• Calibration files needed
• pmasks (quasi-static bad pixel masks,
  http//ssc.spitzer.caltech.edu/irac/products/pmask
  .html)
• PRFs (http//ssc.spitzer.caltech.edu/irac/psf.html
  )
• Pipeline post-BCD products appropriate for quick
  look (need remosaicing in most cases)
• Data in units of surface brightness (MJy/sr)
• BCDs have astrometry accurate to 0.2 arcsec
• Data well-calibrated
• Linearization applied
• Bias (dark) subtracted (sky dark uncertainties lt
  2)
• Interpixel gain corrected (flat-field
  uncertainties lt 1)
• Absolute calibration using set of calibration
  standards

4
IRAC Data Taking
IRAC takes data using multiple (Fowler) sampling
to minimize read noise. Pairs of signal and
pedestal reads are differenced and then averaged
together for each integration. The exposure time
is the difference between any one pair. Sources
which saturate in the pedestal reads will return
small data numbers and are not flagged as
saturated. These super-saturating sources will
appear to have the peaks carved out.
Fowler sampling
Pi pedestal reads Si signal reads
5
Pointing Refinement
Black pre-refinement IRAC Red post-refinement
IRAC
Current pipeline refinement is good to ? 0.2
arcsec No further refinement should be needed
6
Important Header Keywords

• Astrometry
• CRVAL1 , CRVAL2, CD matrix - positions of BCD
  frame
• Superboresight refined in S13
• RARFND, DECRFND, CDRFND matrix pointing refined
  positions using 2MASS
• USEDBPHF indicates if pointing refinement
  succeeded
• Array distortion coefficients
• A_ORDER, A_0_2, A_1_1, A_2_0, etc.
• Polynomial transformation of distorted x, y to
  undistorted x, y
• Photometry
• FLUXCONV - conversion of DN/sec to MJy/sr
• SKYDRKZB - estimate of Zodiacal background
  subtracted by skydark
• ZODY_EST - estimate of Zodiacal background in the
  BCD
• EXPTIME - exposure time of BCD

7
Mask Definitions

• pmask bit definitions
• 0 not set
• 1 not set
• 2 not set
• 3 not set
• 4 not set
• 5 not set
• 6 not set
• 7 variable dark current
• 8 variable response to light
• 9 response too high (fast saturation)
• 10 dark current too excessive
• 11 not set
• 12 not set
• 13 not set
• 14 dead pixel
• 15 sign bit

dmask bit definitions 0 Muxbleed correction
partial 1 FFcorr not applied 2 optical ghost
contamination 3 electrical ghost contamination 4
wraparound correction applied 5 Latent image
flag 6 EBWC not applied 7 Flat is questionable 8
Flat could not be applied 9 CR detection 10 Data
saturated HDR saturated 11 bad data before
correction (cvi2r4) 12 Linearization not
applied 13 banding flag (not yet implemented) 14
Data missing in downlink 15 sign bit
Use all pmask bits 32767
Use only dmask bits 5, 7, 8, 9, 10, 11, 12, 14,
15 32544
8
imask Files

• Reorder dmask flags, removing less than useful
  bits
• imask definition (tbs to be set)
• Bit 0 reserved for boolean mask of data to mask
• Bit 1 reserved
• Bit 2 optical ghost (post-BCD, tbs)
• Bit 3 stray light (post-BCD stray light masking
  sets this)
• Bit 4 strong saturation (donut) detection
  (post-BCD, tbs)
• Bit 5 muxbleed (post-BCD, set in prototype code)
• Bit 6 banding (post-BCD, set in prototype code)
• Bit 7 column pulldown (post-BCD, set in
  prototype code)
• Bit 8 crosstalk (set in pipeline)
• Bit 9 radhit (set in pipeline)
• Bit 10 latent image (set in pipeline)
• Bit 11 not flat-field corrected (set in
  pipeline)
• Bit 12 data not linear (set in pipeline)
• Bit 13 saturation (set in pipeline)
• Bit 14 bad or missing data (set in pipeline)
• Recommend using bits 3, 8-14 32520 in making
  mosaics, etc.

9
Absolute Calibration

• Absolute calibration factors
• Derived from set of 4 stars (A dwarfs) observed
  twice each observing campaign
• Cal factor for ?F? constant (color corrections
  necessary!)
• Photometry performed using 10 pixel radius source
  aperture, 12-20 pixel reference annulus
• Predicted fluxes from Kurucz models
• Photometry stable to 1
• Absolute photometry accurate to 3
• Irreducible error in truth models 1.5
• See paper Reach et al. (2005), PASP accepted
  (astro-ph/0507139)

10
Photometric Accuracy

• With effort (careful planning and data
  reduction), relative photometry can be done to
  better than 1. Absolute photometry to 3,
  limited by underlying stellar calibration.
• Out-of-the-box calibration (i.e. just measuring
  stuff) is considerably worse, typically on the
  order of 5-10.
• Several effects contribute to the error budget at
  the few level. These effects (in order of
  decreasing importance) must be compensated for by
  the user.
• Array-location-dependent correction
• Pixel size (distortion) correction
• Pixel-phase correction
• Color correction

11
BCD Images are in Surface Brightness Units
After flattening (gain correction), images are in
units of surface brightness (flux per unit solid
angle). The pixel solid angle is not uniform in
the BCDs due to distortion. This means that when
measuring fluxes you need to account for the
pixel solid angle to get to Jy. Due to
distortion, the pixels in IRAC do not all have
the same solid angle. For best solution,
reproject images (e.g. with MOPEX) onto a
projection with equal (or nearly so) fixed pixel
solid angles. For BCDs, maps of the pixel solid
angle are available at http//ssc.spitzer.caltech.
edu/irac/solid_angles/.
12
IRAC Is Not an Absolute Photometer

• Absolute bias level unknown for any given BCD
• Shutterless calibration
• Temporal bias variation (first-frame effect)
• Depends on previous readout history of array
• Currently not well calibrated for first frame
  (12s or longer) in AOR
• Effect strongest in 5.8 ?m data
• Most significant for data using repeats
• Mitigate first-frame effect by matching
  overlapping frames
• overlap.pl does a good job
• Check for introduction of large scale gradients
• Correct for first-frame effect in repeats using
  repeat0 frames to create delta-bias images

13
Additional Array-Dependent Correction for Compact
Blue Sources
Filter bandpass shifts as a function of angle of
incidence and thus position on array. This
results in a color correction that is position
dependent. There are also effects depending on
extended nature of zodiacal background. The zody
is extended and red. Most IRAC sources are
compact and blue. While the IRAC flat-field
corrects the zody perfectly, it is wrong for many
other types of objects. This effect may be as
large as 10. It is the dominant source of
uncertainty in IRAC.
14
Additional Array-Dependent Correction for Compact
Blue Sources
Correction images are available from the SSC.
These images provide correct factors for
photometry of a point source with a stellar-like
SED.
http//ssc.spitzer.caltech.edu/irac/locationcolor/
15
Additional Array-Dependent Correction for Compact
Blue Sources

• For well-dithered data, this effect substantially
  averages down, and on average may be lt1. But for
  some locations in the dither map, may be up to
  10.
• Several approaches to application exist
• Correction of BCDs directly.
• Mosaics of correction factors for post
  application to extracted photometry (shown at
  right).

16
Pixel Solid Angle
In the BCD, failure to account for the pixel
solid angle while performing photometry will lead
to an error of up to 1.5 near the array corners.
SSC post-bcd software accounts for variations in
pixel solid angle. Variations are of order 1.
17
Pixel Phase Dependence
Particularly at 3.6 ?m, it appears that the
filling factor of the pixel active region is less
than 1. As a result, combined with the
undersampling of the beam, the flux of an object
varies slightly with position of the beam
relative to the pixel center. This effect is as
large as 4 peak-to-peak at 3.6 ?m. The 3.6 ?m
correction factor (divide measured flux by) is
, p is distance from pixel center in pixels
18
Aperture Corrections
IRAC flux calibration tied to a 12 (10 pixel)
radius aperture. Generally, this is very large
compared to an optimal aperture for point sources
(2-4). Aperture corrections must be applied for
smaller or larger apertures. The IRAC data
handbook provides aperture corrections, but
observers are encouraged to check with their own
curves-of-growth.
Encircled energy curves normalized to 1 at 10
pixel radius.
19
Point Response Function
SSC provides a set of oversampled PRFs derived
from in-flight measurements. They can be used to
derive aperture corrections, etc. Due to the
undersampling of IRAC, the exact details of the
PRF depend somewhat on the dithering and
mosaicing used. It is desirable to derive your
own PRFs from your actual data (e.g. with
DAOPHOT, Starfinder, etc.). Beware of using the
faint galaxies in the frame (which are slightly
extended), stick with stars.
http//ssc.spitzer.caltech.edu/irac/psf.html
20
Extended Point Response Function and Internally
Scattered Light
The PRFs of 5.8 and 8.0 ?m arrays are more
extended than predicted by optics. In these
channels, the wings of the PRF are enhanced by
light scattered internally in the arrays. Red
3.6 ?m Yellow 4.5 ?m Green 5.8 ?m Cyan 8.0
?m
21
Color Corrections

• Absolute calibration assumes source spectrum F? ?
  ?-1
• For any other source function, a color correction
  must be applied
• Corrections are typically a few percent for
  stellar and blackbody sources
• Corrections can be more significant for sources
  with ISM-like source functions (gt 20 depending
  on spectrum and passband)
• Measured flux is flux at effective wavelength of
  array 3.550, 4.493, 5.731 and 7.872 ?m, for
  channels 1-4 respectively

22
Extended vs. Point Source Calibration

• The 5.8 and 8.0 ?m arrays have significant
  scattered light issues
• Internal scattering in array
• PRFs have extended diffuse components (still
  being characterized)
• IRAC is calibrated for point sources
• 22 of flux is scattered out of nominal PRF at
  5.8 ?m
• 26 of flux is scattered out of nominal PRF at
  8.0 ?m
• Extended source calibration
• Need to account for light that is scattered back
  onto extended source aperture
• Apply ratio of point source to extended source
  throughput to account for this light
• Extended source correction multiply by 0.63, 0.69
  for 5.8 and 8.0 ?m
• Ratio derived for truly diffuse source (Zodiacal
  light)
• There is a color dependency of the ratio
• Ratio will depend on aperture size and angular
  size and surface brightness distribution of
  source

23
Test of extended source calibration ?

• Large aperture photometry of NGC 777
• Assume no dust emission from galaxy
• Extrapolate 2MASS photometry to IRAC using color
  information
• Resulting extended source correction is smaller
  than correction derived from observations of
  Zodiacal light
• Suggest that scattering is source color dependent
• 0.86 for 5.8 ?m
• 0.75 for 8.0 ?m
• These factors may be more appropriate for
  extended sources dominated by stellar-like SEDs
• More work on correction versus aperture size
  needs to be done

24
Test of extended source correction ??

• Observations of 3 extended HII regions
• Sh2-152, Sh2-138 and IRAS 201784046
• Truth data at 8.3 ?m from MSX (5 absolute
  calibration)
• Measured fluxes in various aperture sizes
• MSX (8.3 ?m) and IRAC (8.0 ?m) data with same
  astrometry
• Aperture sizes from 20 - 350 arcsec
• As MSX resolution is 20 arcsec, comparison
  contaminated by point sources for lt 50 arcsec
• Use representative off-source regions to subtract
  background
• Compare ratio of IRAC 4 / MSX flux as a function
  of aperture
• Apply extended source correction for Zodiacal
  light
• Color correct both data sets for ?eff 8.0 ?m
  (used ISO SWS spectrum of Sh2-138)
• Extended source correction from Zodiacal
  observations appears to be appropriate for
  apertures gt 50 arcsec

25
Red extended source calibration data
26
Red extended source calibration dataapertures lt
100 arcsec
27
Extended Source Calibration Summary

• 5.8 and 8.0 ?m photometry of extended sources
  must be corrected
• Scaling factor applied to account for diffuse
  scattered light scattered back into extended
  source aperture
• The scaling factor appears to be color dependent
• For Zodiacal light, HII regions (red diffuse
  emission) the best current scaling factors are
• 0.63 at 5.8 ?m
• 0.69 at 8.0 ?m
• Scaling appears to apply to sources gt 50 arcsec
  in extent
• No information exists on smaller apertures
• Observation of a stellar dominated extended
  source suggest the correction factors
• 0.86 at 5.8 ?m
• 0.75 at 8.0 ?m
• No data on scaling as a function of aperture size
  exists
• The above scaling factors are provisional, will
  depend on the exact color of the source, and are
  not well defined for intermediate sized sources

28
Data Artifacts
Electronic
Optical

• Induced by bright sources
• Artifacts property of detector and readout
  multiplexers (muxes)
• InSb (3.6 and 4.5 ?m arrays, channel 1 and 2)
• SiAs (5.8 and 8.0 ?m arrays, channels 3 and 4)
• Artifacts function of fluence
• Saturations
• Multiple artifacts produced by single source
• Artifact nomenclature
• Muxbleed
• Muxstripes
• Column pulldown
• Banding
• Latent images
• Electronic crosstalk

• Scattering of light from bright sources
• Source brightness
• Source position relative to Spitzer focal plane
• Optical path and detector components
• Artifact nomenclature
• Ghost images
• Filter
• Beamsplitter
• Pupil
• Stray light
• Optical banding
• Extended scattered light

29
Muxbleed

• Cause Relaxation of multiplexer after observing
  bright source
• Present only in 3.6 and 4.5 ?m arrays
• Symptom Trail of bright pixels in read direction
  (every 4th column)

3.6 ?m 30 Doradus
Muxbleed
30
Muxstripes

• Related to muxbleed
• Symptom Variation in level of column segments
  due to very bright source
• Bright stars trigger and stop stripes

Muxstripe starts
Muxstripe ends
31
Column Pulldown

• Symptom Bias change in column containing bright
  source
• Strongest at 3.6 ?m, also present at 4.5 and 5.8
  ?m
• Different above and below source - Fowler
  sampling dependent

32
Banding

• Horizontal and vertical stripes produced by
  bright sources at 5.8 and 8.0 ?m
• Partially electronic, also optical
• Strength of banding can be different on opposite
  sides of triggering source
• Weak banding may also be present in 3.6 and 4.5
  ?m images

33
Residual (Latent) Images

• Short term latents
• Significant for longest frame times
• 1 of latent producer in subsequent 200s
  integration
• Latents are much weaker for shorter frame times
• Flagged but not currently modeled
• Long term latents
• 3.6 and 8.0 ?m arrays
• Persistent for hours
• Mitigated through annealing and scheduling of
  bright sources
• Not flagged in data
• Slew latents
• Avoid slewing across bright sources in an AOR
• Constrain AOR sequences to avoid slewing across
  bright source
• Redundancy is key to mitigation
• Examine median stack image provided

34
Latent Flagging
35
Long Term and Intra-AOR Latents
36
Ghost Images

• Caused by internal reflections in filters and/or
  beamsplitters
• Can be flagged by using post-BCD software
• No good mitigation method
• Ghost images exist as part of provided PRFs
• Filter ghosts
• Position of ghost dependent on position of source
  on array for 3.6 and 4.5 ?m arrays
• Multiple orders possible for 3.6 and 4.5 ?m
• Less severe for 5.8 and 8.0 ?m
• Beamsplitter ghosts
• 5.8 and 8.0 ?m arrays only
• Much fainter than filter ghosts

37
Filter and Beamsplitter Ghost Examples
3.6 ?m
4.5 ?m
5.8 ?m
8.0 ?m
38
Stray Light

• Scattered light from sources off array onto array
• Stray light producing regions identified for all
  4 arrays
• 2-3 scattering regions for 3.6 and 4.5 ?m
• Scattering regions on edges of 5.8 and 8.0 ?m
  arrays
• Diffuse scattered light accounted for in
  flat-field pipeline processing
• Stray light can be flagged in post-BCD processing
• Dithering with sufficient step size is key to
  mitigation
• Use large or medium dither
• Or small dither if mapping with 1/3-1/2 array
  offsets

39
Stray Light Example
Stray light producing star
40
Optical Banding
5.8 ?m

• Light scattered in array for SiAs arrays
• Row and column bands
• Diffuse component
• Verified by ground tests on spare arrays
• 10-20 of incident flux scattered into bands
• Diffuse scattering is 1.3? band flux
• Banding has curvature

41
Radiation Hits (Radhits)

• Affect 3-6 pixels / second of exposure time
• Radhits more significant at 5.8 and 8.0 ?m
• Redundancy is key to mitigation
• Flagged in pipeline processing (single frame
  radhit detection)
• Need to be flagged in post-BCD processing
• SSC provided MOPEX software does this
• Suggest RM_THRES 0.05 (in mosaic.nl,
  particularly for 5.8 and 8.0 ?m)
• For high redundancy data (gt4 samples/position) ?
  temporal (multi-frame) outlier rejection
• For low redundancy data (2-4 samples/position) ?
  dual outlier
• Temporal outlier rejection works much better with
  5 frames than it does 4.
• Can use dual outlier rejection to prevent
  rejection of PRF wings
• REFINE_OUTLIER 1 (in mosaic.nl)
• REFINE_OUTLIER_THRESH gt10 (in mosaic.nl)

42
Radhit Example
8.0 ?m mosaic w/outlier rejection
8.0 ?m BCD
43
Electronic Artifact Mitigation Method

• Artifact flagging
• Cleaning up of mask files (dmask ?imask)
• Identification of triggering pixels
• Masking of affected data
• Estimation of true sky under artifacts
• Interpolation using nearby data
• Gaussian kernel which mimics PRF
• Modeling of artifacts
• Fit to difference between data and estimation
• Fit functional forms for each artifact
• Correction of BCDs
• Replace data with data - model
• Update uncertainties

44
Artifact Flagging

• Find triggering pixels based on unsaturated
  fluence (number of e-)
• e- MJy/sr GAIN EXPTIME / FLUXCONV
• HDR saturations
• Use short frame data to estimate fluence
• Replace pixels if short gt 100000 e- and long lt
  0.5 ? short
• Full frametime saturations
• Need additional data
• Can estimate from 2MASS for most sources
• Tool to PRF fit in testing
• Muxbleed (3.6 and 4.5 ?m)
• 30000, 60000, 150000 e- at 3.6 ?m to mask 32, 64,
  192 samples
• 55000, 160000, 250000 e- at 4.5 ?m to mask 32,
  64, 128 samples
• Column pulldown (3.6 and 4.5 ?m)
• 30000 e- at 3.6 ?m
• 110000 e- at 4.5 ?m

45
3.6 ?m Flagging
Mask
BCD
46
5.8 and 8.0 ?m Flagging

• Horizontal banding 200000 e- (5.8 ?m), 55000 e-
  (8.0 ?m)
• Vertical banding 80000 e- (5.8 ?m), 200000 e-
  (8.0 ?m)

5.8 ?m BCD
5.8 ?m mask
47
Sky Estimation

• Interpolate over masked pixels
• Use Gaussian kernel with FWHM of 1.44, 1.43,
  1.49, 1.71 arcsec for 3.6, 4.5, 5.8, 8.0 ?m

3.6 ?m BCD
3.6 ?m sky estimate
48
Model Fit

• Fit corrective functions to BCD-sky_estimate
  difference
• Banding, pulldown DC offset on either side of
  triggering source
• Linear fit may be appropriate for some data
• Muxbleed model derived from data in AOR
• Fowler sampling dependent
• Scaled and filtered muxbleed instances
• Optionally, smooth with a
• polynomial fit

49
3.6 ?m Results
BCD
Corrected BCD
50
8.0 ?m Results
BCD
Corrected BCD
51
Stray Light Masking

• Update masks when bright sources are in stray
  light producing region
• Procedure uses 2MASS source table provided as
  post-BCD product
• Finds instances of bright enough 2MASS sources
  off array and updated masks

52
Documentation Resources

• Spitzer Space Telescope Observers Manual -
  Chapter 6
• Infrared Array Camera Data Handbook
• IRAC Pipeline Description Document
• Calibration paper by Reach et al. (2005)
• IRAC webpage http//ssc.spitzer.caltech.edu/irac/
• IRAC interest group archive https//lists.ipac.cal
  tech.edu/mailman/listinfo/irac-ig

53
Software Resources

• IRAC contributed software
• http//ssc.spitzer.caltech.edu/archanaly/contribut
  ed/browse.html
• IDL procedures
• Straylight and ghost masking (mask_straylight_v5p1
  .pro)
• Pulldown corrector (extragalactic observations)
• Muxbleed corrector (extragalactic observations)
• dmask to imask procedure (make_imask_from_dmask.pr
  o)
• Alternate overlap corrector (proto_overlap.pro)
• General muxbleed, banding and pulldown corrector
  (irac_preprocess.tar) saturated source estimator
• Coming soon
• Saturated source fitter
• PRFs as a function of fluence
• IST recommends not using cosmetic.pl for
  photometric data (cosmetic.pl is good for making
  press release images)