Readme file for the deliverable 4.3 of the H2020 EWC project ============================================================ 1. Introduction The goal of this deliverable is to provide validated data products of HST observations that can be used to emulate Euclid images and calibrate shape measurement biases related to galaxy color gradients in WP5. This deliverable builds up on the earlier deliverable D4.1, which provided an initial reduction of CANDELS data. 2. Methodological approach This deliverable D4.3 provides a significantly refined reduction of the earlier tile-wise reduction of CANDELS HST/ACS observations in filters F606W and F814W which provided the basis for the earlier deliverable D4.1. The refined reduction D4.3 provides two important major updates: Firstly, it employs our recalibrated model for the correction of charge-transfer inefficiency (Massey et al. in prep.). Secondly, we now employ the improved astrometric modelling provided by astrodrizzle (D4.1 was based on the older multidrizzle tool), for which we carefully validated the astrometric matching. The deliverable D4.3 includes new science frame stacks and corresponding r.m.s. noise model images for a total of 149 HST/ACS pointings observed in the filters F606W and F814W (the selected exposures are detailed in Schrabback et al. 2018, MNRAS, 474, p.2635-2678). Importantly, the world coordinate system of the new image stacks is consistent with the corresponding stacks in the D4.1 reduction. This enables a fast reprocessing for the analysis conducted within WP5, which employs the HST reductions from WP4. Accordingly, D4.1 data products such as masks for bright stars, ACS PSF models, and matched catalogs including photometry from Skelton et al. (2014 ApJS, 214, 24) can directly be used in combination with the new D4.3 data. In addition to these complementary data products, which were already provided in D4.1, Deliverable D4.3 is also supported by our extensive tests conducted using simulated data (see Sect. 3). We include validated data products that cover the full CANDELS-Wide fields AEGIS, UDS, and COSMOS, which have a homogeneous coverage in the HST/ACS filters F606W and F814W. These are the best-suited bands for the Euclid shape calibration, since they cover the Euclid/VIS band-pass almost perfectly together. 3 Summary of activities and research findings An important component of the deliverables related to WP5 are our refined models of the HST/ACS point spread function (PSF). These have been described in more detail in deliverable D4.2a. Here we only summarize the most important points relevant for deliverable D4.3: An earlier PSF analysis of CANDELS F606W+F814W data was conducted by Schrabback et al. (2018), employing the principal component analysis (PCA) approach from Schrabback et al. (2010, A&A 516, A63), which operates on spatial interpolation coefficients for PSF characteristics such as half-light radius and PSF polarisation. Schrabback et al. (2018) calibrated their PCA models on a set of stellar field exposures observed in the F606W and F814W filters. Gillis et al. (2020, MNRAS 496, 5017) tested and recalibrated ACS PSF models computed using tinytim (Krist & Hook 2011, SPIE, 8127) in the filter F606W. Using this refined version of tinytim we then estimated the best-fit telescope focus values for each of the star fields from Schrabback et al. (2018). This allowed us to fit a relation between the focus and the first PCA coefficient from Schrabback et al. (2018), showing that linear models provide good interpolations in most cases except for very negative focus offsets, where larger discrepancies were detected. We then used these relations to compute an exposure-averaged best-fit focus value for each of our CANDELS galaxy field stacks, employing the PCA coefficients from Schrabback et al. (2018). These can be used along with the provided position- and focus-dependent tinytim models as PSF models in the colour gradient analysis of WP5. A key complement to our data deliverables are the tests conducted using simulated images. Parts of the simulation-based findings were already described in Deliverable 4.2a, and a more in-depth description will be provided in Deliverable 4.2b (Euclid Collaboration: Scognamiglio et al. in prep.). In the following we summarize the two most important findings that are relevant for the use of the D4.3 data in weak lensing calibration image simulations: As already explained in Deliverable 4.2a, ACS PSF model uncertainties can lead to significant biases in weak lensing shear calibrations. In D4.2a we showed that resulting additive shape measurement biases can be mitigated via the introduction of an extra random image rotation in the Euclidization procedure following the deconvolution for the ACS PSF. In further tests we identified a second source of bias in the Euclidization procedure, which occurs if Euclid VIS images are emulated using galsim at the native VIS pixel scale (0.1”). As we found out recently, this bias can be mitigated if emulated Euclid-like data are initially created using the Euclidization procedure at a finer pixel scale (0.05”). This mitigation strategy also works if the images are subsequently re-binned to the VIS native pixel scale. This suggests that the biases are likely the result of aliasing effects that occur at the native VIS pixel scale given the relatively poor sampling of the VIS PSF. A more detailed analysis of this effect will be presented in Deliverable D4.2b (Euclid Collaboration: Scognamiglio et al. in prep.). 4. Detailed description of files: The latest version of reduced images (based on astrodrizzle and including our latest CTI calibration, Deliverable D4.3) are found in subdirectories of: https://marvinweb.astro.uni-bonn.de/data_products/euclid-hst/HST/CANDELS/v1.0/ Additionally, we provide our earlier reduction (Deliverable D4.1, based on multidrizzle), which is available in https://marvinweb.astro.uni-bonn.de/data_products/euclid-hst/HST/CANDELS/v0.9/ There are subdirectories for each of the three included CANDELS fields (aegis_*, cosmos_*, uds_*) for both filters (606, 814), including science frame stacks (*sci.fits) and rms noise maps (*rms.fits). Both reduction versions are matched to the same output WCS, so that additional data products, which have been created for the v0.9 stacks (and can be found in the corresponding subdirectories) can also directly be used for v1.0. The subdirectories *_flagmasks include flag masks, that mark regions affected by bright stars, galaxies, image boundaries, etc. (good areas have a flag mask value of 0). Extracted catalogs (available for v0.9 in the same sub-directories as the image stacks) are named: catforer_${canfield}_${filter}.txt where ${canfield} indicates the CANDELS field (aeg, uds, cos) and ${filter} the filter (606, 814). These are KSB+ weak lensing shape catalogs described in Appendix A of Schrabback et al. (2018) matched to the photometric redshift catalogs from Skelton et al. (2014). Columns: # 1 Ra R.A. (deg) # 2 Dec Dec. (deg) # 3 rh half light radius (analyse) [pixel] # 4 FLUX_RADIUS Fraction-of-light radii [pixel] # 5 MAG_AUTO_F606W Kron-like elliptical aperture magnitude [mag] # 6 MAGERR_AUTO_F606W RMS error for AUTO magnitude [mag] # 7 IMAFLAGS_ISO FLAG-image flags OR'ed over the iso. profile # 8 Flag Extraction flags # 9 FLUX_AUTO Flux within a Kron-like elliptical aperture [count] # 10 FLUXERR_AUTO RMS error for AUTO flux [count] # 11 goodgal # 12 goodregion mask value (addmask) # 13 e1iso_snCal_rot e1iso_snCal for North=up # 14 e2iso_snCal_rot e2iso_snCal for North=up # 15 e1iso_snCal -0.078*(SN/2)^-0.38 [""] # 16 e2iso_snCal -0.078*(SN/2)^-0.38 [""] # 17 f_F606Wcan F606W aperture magnitude from Skelton et al. 2014 # 18 e_F606Wcan F606W flux error from Skelton et al. 2014 [in counts for a zeropoint of 25.0] # 19 f_F814Wcan F814W aperture magnitude from Skelton et al. 2014 # 20 e_F814Wcan F814W flux error from Skelton et al. 2014 [in counts for a zeropoint of 25.0] # 21 z_p Peak photometric redshift from Skelton et al. 2014 # 22 A semi major axis (analyse) # 23 B semi minor axis (analyse) # 24 Theta position angle (analyse) [degree] # 25 Xpos good centroid position x (analyse) [pixel] # 26 Ypos good centroid position y (analyse) [pixel] # 27 fieldID running field id # 28 orientat position angle of image y axis (deg. e of n) The image pixel scale is 0.05". To select "good" galaxies (well-PSF corrected, not in masked area, stars excluded) one has to select: goodgal=1 goodregion=1 IMAFLAGS_ISO=0 F606W-F814W color can be computed as f_F606Wcan-f_F814Wcan These are PSF matched magnitudes from 3D-HST. For total magnitudes the SExtractor auto magnitudes can be used: MAG_AUTO_F606W (MAG_AUTO_F814W for the F814W based catalogs) As size estimate FLUX_RADIUS from Sextractor or the half-light radius from analyseldac rh can be used. The fully corrected KSB+ ellipticity/shear estimates are provided in both image coordinates (e1iso_snCal, e2iso_snCal) and sky coordinates (e1iso_snCal_rot, e2iso_snCal_rot, rotated to North=up, East=left). Xpos, Ypos are the image coordinates from the original image. z_p is the peak photometric redshift estimate from 3D-HST. fieldsID identifies the original image in which the measurement was done (see below). =============================================================================== The information to translate fieldsID to the tile names is provided in files fieldID_list_${canfield}.F{filter}_withPSF fieldsID is in column 3, the field name in column 1 These files contain further important information: Column 4 lists the image position angle on the sky in deg. Column 5 lists the mean first principal component coefficient from the PCA PSF analysis, averaged over all exposures contributing to the stack. Column 6 lists the corresponding tinytim PSF focus offset for chip 1 (based on a linear fit between these quantities in star fields). Column 7 lists the corresponding tinytim PSF focus offset for chip 2 (based on a linear fit between these quantities in star fields, excluding star fields with a PCA coefficient >1 for F606W due to outliers). =============================================================================== PSF models The tinytim PSF model fits files are stored in the sub directory tinytim. These were created using tinytim with updated parameters from Gillis et al. (2020). They have a pixel scale of 0.0155 arcs/pix, and no camera distortion has been applied to the models. File names are: tt2_${x_psf}_${y_psf}_${focus}_F${filter}_psf_chip${chip}.fits Where x_psf and y_psf are the positions in the detector frame, chip indicates the chip number (1 2) as explained below, focus indicates the focus offset in microns, and filter indicates the band pass (606 814). In order to identfy the correct focus file for a particular galaxy first identify the correct focus offset from the file fieldID_list_${canfield}.F{filter}_withPSF via the "fieldsID" index in the catforer_${canfield}_${filter}.txt catalog. The galaxy positions Xpos and Ypos listed in catforer_${canfield}_${filter}.txt correspond to the distortion corrected stacks. In order to approximately transform them to detector coordinates (for typical small dither patterns) first the chip into which the galaxy falls at the average dither position must be determined as: Chip 2 if (Ypos>(Xpos*0.0392927+2053.91)) Chip 1 if (Ypos<(Xpos*0.0392927+2053.91)) Then, the detector position can approximately be determined as For Chip 1: x_psf ~ Xpos-0.0539895*Ypos y_psf ~ Ypos-0.0262454*Xpos For Chip 2: x_psf ~ Xpos-0.0554415*Ypos+3.0 y_psf ~ Ypos-0.0350099*Xpos-2089.0 Based on chip, focus, x_psf, and y_psf the nearest tinytim PSF model can be identified and used. =============================================================================== Known issues: The applied KSB+ implementation fails for very large galaxies given a limited postage stamp size, which is why these galaxies are excluded from the catalogs. =============================================================================== Contact: In case of questions please contact Dr. Tim Schrabback (schrabba@astro.uni-bonn.de) =============================================================================== Acknowledgements: We acknowledge support from the EU Horizon 2020 EWC project, as well as support provided by the German Federal Ministry of Economics and Technology (BMWi) via grants 50QE1103 and 50QE2002.