ACS G800L grism pointings in the Archive were first grouped in data associations, following pre-defined rules on
sky position, roll angles, and availability of direct images, resulting in 153 fields (see Figure and Table).
The PHLAG pipeline [3] was used for the end-to-end data processing of all associations. The core component of this pipeline is the aXe software,
developed by ECF [4] to reduce the slitless mode data of HST, which was used for the NICMOS grism spectra and then further developed to
optimally deal with ACS spectra. The data reduction process includes the following steps:
- data preparation
- data retrieval from the archive
- detection of cosmic rays on direct imaging using the LACOS algorithm
- re-alignment of the direct images
- multi-drizzle the direct images in the available filters
- co-add of the filter images to the deep detection image
- object detection with SExtractor on the detection image
- spectral extraction with aXe
- metadata assembly
- generation of the various products
- production of the quality control plots
As part of the quality control of the input association, the following strategy was adopted:
- a minimum of 2 grism images in an association is required;
- a minimum of 2 direct images in one filter is required for an association;
- direct images for which the alignment could not be improved were discarded;
- associations with an object density higher than 2000 per square arcminute are discarded;
- the accuracy of multi-drizzling and the alignment process for direct and grism images is assessed;
- the accuracy of the background subtraction for the grism data is monitored. Defective grism images are discarded;
- the wavelength calibration accuracy is verified.
As a result of this first QC process, 15 associations were not included in this release. These are indicated in
the Table of failed associations, along with the main reason for the exclusion.
Quality control of individual objects within associations:
- internal duplicates, removal of fake objects with too close neighbours
- parameter check in multi-dimensional space, identification and removal of fake objects
Global correction to the object positions:
- re-centering of the spectra traces
- improvements to the wavelength calibration
Assessment of direct imaging fidelity:
- marking the stellar locus in parameter space