1. Application of a Charge Transfer Model to Space Telescope Data Paul Bristow Dec’03 www.stecf.org/poa

2. The STIS Calibration Enhancement Project  A comprehensive empirical calibration pipeline already exists for STIS. We aim to improve those components which benefit from physically motivated corrections.  Current work includes:  Wavelength Calibration Calibration lamp line list - new lab measurements at NIST Optical Model  Optimal Spectral Extraction  Detector Model

3. STIS CTE trails grow with time:

4. CCD Model: Concept Forward Simulation:  Start with 2D Charge distribution on chip Distribution of bulk traps  Shift the charge under each electrode as during readout  Calculate capture and emission of charge in each shift Take into Account:  Status of bulk traps  Dark current  Trap capture and emission timescales  Chip clocking frequency, architecture, gain etc.

5. Correction: Further iterations (usually not necessary): Corrected image Difference image Simulation output Raw data =- = - Difference image Raw data >> SIMULATION Corrected image >> SIMULATION

6. Cleaning CTE Trails

7. CCD Model: Implementation Follows a prescription outlined in Philbrick 2001 (Thanks to Rob for correspondence)

8. CCD Model: Implementation Types: P-V (Si-E), O-V (Si-A), V-V Timescales:  Emission - essentially from lab data  Capture - always smaller than clock period (for STIS at least) Trap density:  NIEL appropriate to HST orbit  Linear scaling with time

9. Application to STIS BIAS offset and pattern must be removed from raw data before input to CTI model Divide by GAIN Negative value pixels corrected (unphysical) Reference dark files also suffer from CTE and must be corrected Reference bias files contain hot columns which are actually corrected by the CTI model => bias files must be modified

12. Comparison to Empirical Corrections for Imaging Data Empirical calibration used as a test of physical model  On average we should expect agreement  This comparison can also help to calibrate the physical model Good general agreement (especially for objects to which the empirical calibration applies)

13. Special Cases: Images

14. Trap Distribution: Mean (over chip) density fixed by NIEL and time on orbit. Uniform versus arbitrary clustering Mapping traps  Ground testing no good, traps only begin to appear on orbit.  Pocket pumping  On board Fe 55 (Chandra) Dark current 

15. “Initial” Trap Status Possibilities:  All empty/All full  Function of charge collected during exposure (collecting phase electrodes)   Long timescale traps Turns out not to be critical

16. Notch/Mini-channel The STIS CCD was designed to have a mini- channel There seems to be no pre- launch data which characterises it. Represent channel with 2 parameters: n mc (<n t ) and N sat-mc (<<N sat ) CCD readout model produces better results with N sat-mc =0 

17. Quantisation and Noise There’s no such thing as half an electron! Original implementation of readout model used integer, i.e. quantised, representation of charge.  e.g. n ex-INT =Poisson(n ex-FLOAT ) i.e. random Current version uses floating point representation (fractional e - ), i.e. deterministic  Similar results (for CTI effect on fluxes, charge trails etc), However….

18. Quantisation and Noise (cont.) Quantised model predicts that the readout process increases the background  background as a function of row  More transfers, more random events => seems reasonable!  Not observed! For this reason the deterministic version has been adopted

19. Exporting to other Detectors (WFPC2) Initial attempts to apply to WFPC2 require large, arbitrary changes to trap normalisation to get reasonable results. Difficult to find a normalisation that fits a range of data. Differences (STIS to WFPC2):  WFPC2 is front illuminated Charge/volume relation may be different? Different mix of trap timescales?  Clocking, chip size, gain, time on orbit etc. should be easily accounted for.

20. Conclusions We have developed a model of the STIS CCD readout and used it to correct real data The result for STIS provide a substantial calibration enhancement for CTI effected data The model is generic and portable to other CCD detectors There remain aspects which, while not critical to STIS calibration, could be better understood and implemented. www.stecf.org/poa

21. Comparison to Empirical Corrections for Spectroscopic Data Same principle as for imaging data Once again, good general agreement

22. Special Cases: Spectroscopy