XDF Data Release 1.0

Page Navigation

• Introduction
• Information on Observations
• Explore the Images with Interactive Displays
• Data Products, Zero Points, and Download Options
• Citation Information

Introduction

This summary describes the Hubble eXtreme Deep Field (XDF) project data release. The XDF includes all optical and infrared data taken by Hubble covering the original Hubble Ultra-Deep Field (HUDF) program, with just a few exceptions. Small amounts of data taken with unusual or rarely used filters and spectroscopic data are not included. The XDF also includes images which only partially cover the HUDF, as shown in the figure below. The weight maps are appropriately adjusted for the partial coverage.

The data used to generate the XDF were taken over the decade from mid-2002 to the end of 2012. The total exposure on the XDF is about 1.8 million seconds using data from Hubble's two main workhorse cameras, the Advanced Camera for Surveys Wide Field Channel (ACS/WFC) and the Wide Field Camera 3 InfraRed Channel (WFC3/IR). Substantial gains accrue from using all the ACS data taken over the decade since 2002, as well as from the improved processing techniques since the HUDF was released. The gains in the XDF are equivalent to adding about 180 orbits of data to the original HUDF dataset — split roughly equally between gains from improved processing techniques and from new data.

Further background on the XDF, including how and why this project was initiated, can be found at http://xdf.ucolick.org/xdf.html. The generation of the processed, combined and aligned images for the XDF was carried out over the last year by XDF team members Dan Magee, Pascal Oesch, Rychard Bouwens and Garth Illingworth (see Illingworth et al 2013). The XDF team can be found at http://xdf.ucolick.org/team.html.

The XDF Field

Observations

Description

Observations included in this version of the XDF were taken from July 2002 to December 2012 from 19 different HST programs. The table below lists the median exposure time (in kiloseconds) in the deepest parts of the ACS and WFC3/IR XDF images along with the total number of exposures combined in each filter.

Programs Used

Data Processing

The process begins with individual flt files requested from the MAST HST archive. These images have already been subject to bias correction, dark subtraction, flat-field correction, and cosmic ray rejection. The flt files obtained from the MAST are reasonably well calibrated, given that the on-the-fly pipeline processing by STScI already takes advantage of darks and flat fields constructed from latest on-orbit calibration data. However, in some cases further instrument specific processing is required as described below.

All images used to produce the XDF were first visually inspected to identify any data quality issues (loss of guiding, excessive background, pointing inaccuracies). Any image which could not be corrected was rejected and not processed. During this visual inspection we also identify images affect by satellite trials and optical ghosts from filter reflections generated by bright stars (ACS) and updated the data quality array to ensure that these artifacts were masked during final processing.

For ACS/WFC data taken after HST Servicing Mission 4 (SM4) we also applied corrections for charge transfer efficiency (CTE) degradation and removed bias striping noise.

ACS/WFC Data

The XDF ACS/WFC dataset were reduced by the ACS GTO team data processing pipeline APSIS. For a detailed description of the reduction process please refer to "An Automatic Image Reduction Pipeline for the Advanced Camera for Surveys" Blakeslee, et.al. 2003. In order to improve the reductions of the XDF ACS/WFC dataset we applied one additional correction not performed by APSIS.

To optimize the pixel-by-pixel S/N (and correct for any imperfections in the flats or darks), we median stacked all of the observations in the F435W, F606W, F775W, F814W & F850LP bands to create supermedian frames. These are subsequently subtracted from the individual images. To further optimize the pixel-by-pixel S/N in the reductions, we repeated the median stacks of exposures in each band — but now masking out the sources apparent in the final stacks (rather than just those apparent in the individual exposures). We then subtracted these supermedian images from of the each individual exposures and drizzled the data together to generate the final reductions.

WFC3/IR Data

The XDF WFC3/IR data were reduced by the HUDF09 team's WFC3/IR data reduction pipeline WFC3RED. The reduction process is described in "WFC3RED: A HST Wide Field Camera 3 Image Processing Pipeline" Magee, Bouwens & Illingworth 2007.

The WFC3RED processing of the WFC3/IR observations in the XDF follow the procedures in (Magee, Bouwens & Illingworth 2007), but with a few additional corrections in order to improve the reductions.

While most WFC3/IR flt files requested from MAST are ready for use (and stacking), in some cases CALWF3 can falsely flag nearly all pixels as cosmic rays. For WFC3/IR MultiAccum mode observations, CALWF3 assumes that accumulating background counts over the entire observation is a linear function. This assumption may not be the true for all observations. Depending on the circumstances for an observation the background count rate may vary over the duration of the observation. In order to determine if non-linearities in the background count rate are a problem, we compared the average exposure time in an flt file's TIME array extension with the exposure time listed in the header EXPTIME keyword. If we find the exposure time varies by more than 2% we introduce one additional step and sky subtract each of the individual readouts in the MultiAccum observation. This additional step ensures the background count rate is linear before reprocessing the image with CALWF3.

We also mask out pixels on WFC3/IR images that are affected by source persistence. This masking was performed by utilizing persistence models generated by the STScI WFC3 Persistence Project. A persistence model is created for each WFC3/IR exposure which incorporates an internal persistence model (persistence within a visit) and an external persistence model (persistence from earlier visits). Using the total (internal plus external) persistence model we flag all pixels above a threshold of 0.2 electrons/s to ensure that the final reductions are not significantly affected by source persistence.

To further refine the registration of images, we use MultiDrizzle to create image stacks for each visit and then cross-correlated the sources in each image stack with the HUDF F850LP-band images. Minor adjustments are made to the coordinates of each visit to obtain a more optimal alignment.

Finally, in order to improve the pixel-by-pixel S/N (and correct for any imperfections in the flats or darks), we median stacked all the observations in the F105W, F125W, F140W and F160W bands to create supermedian frames. These are subsequently subtracted from the individual images. To further optimize the pixel-by-pixel S/N in the reductions, we repeated the median stacks of exposures in each band — but now masking out the sources apparent in the final stacks (rather than just those apparent in the individual exposures). We then subtracted these supermedian images from each of the individual exposures and drizzled the data together to generate the final reductions.

Back to top.

[ (53.187414,-27.779152), (53.159507,-27.759633), (53.134517,-27.787144), (53.161906,-27.807208) ]

FLAG_IMAGE       hlsp_xdf_hst_deepest_flag_v1.fits      # filename for an input FLAG-image

hlsp_xdf_hst_wfc3ir_F160W_60mas_v1_sci.fits