1. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 1 ) Andy Beardmore and the Leicester Swift-XRT calibration team CCD22 Simulator Mk II

2. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 2 ) Why Rewrite ? Existing code is written in FORTRAN-77(!) – Difficult to follow let alone maintain Been through many hands (~6)! Often find hard-coded parameters Over reliance on COMMON blocks – Only approximates the XRT onboard and ground s/w event recognition schemes split event threshold is constrained to be same as event threshold Grading schemes not identical

3. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 3 ) Code base diverged between PC and WT – WT uses gaussian approx to line spreading – PC uses more rigorous “charge cloud spreading” model in field-free region Never implemented in WT mode - as existing WT code was deemed “OK” – Hard-coded energy dependent shelf and line profile “fixes” Ended up with two different WT codes with different shelf “fixes” for grade 0-2 and 0 Most importantly - need to fix WT mode – Grade 1,2 redistribution clearly inaccurately 10 row binning an afterthought

4. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 4 ) CCD22 Simulator MkII Written in C++ with the aim to be modular. ReadoutMode base class – implements the image, store, and serial-register arrays; pixel charge accumulation WTRawFrame, PCRawFrame derived classes implement the specific readout modes, CTE and readnoise application – Image → Store transfer for PC – 10 row binning → serial register for WT WTMode, PCMode are further derived classes implement the event recognition and grading

5. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 5 ) Generic source input. Photon obtained from – Photon/s/bin spectrum at top of CCD – Spatial distribution Uniform source (circle or rectangle) Drawn from the XRT PSF Full charge-cloud spreading Si-Kα flourescence Outputs an “event list” ( bin, rawx, rawy, pha, grade) OpenMP parallelisation over the photon loop – 1 Million photon PC mode door source simulation takes 10 minutes using 10 cores. – Not so effective for WT mode

6. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 6 ) The difficult stuff – i.e. Physics! X-ray detection is determined by a photon's ability to pass through the electrode structure, while interacting in the active layer of the device – accurate Si, SiO2, Si3N4 absorption coefficients required → using ACIS values. Charge cloud spreading, followed by the pixel- mapping, event-recognition and grading determine how well the original photon energy can be recovered

7. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 7 ) CCD Operation Overview Electrode Structure

8. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 8 ) Open Electrode Structure

9. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 9 ) Charge cloud spreading Initial Size (Pavlov & Nousek 1999) – 3D Gaussian, with 2D cylindrical radius – Less than 1 micron for energies of interest Depletion region (Hopkinson, 1984; PN 1999) – Expansion radius

10. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 10 ) Charge cloud spreading (ii) Field-Free region (Hopkinson, 1984; PN 1999) – Radial charge distribution at depletion/FF boundary Substrate (Hopkinson 1984) – Radial charge distribution at substrate/FF boundary

11. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 11 ) 2D Charge Distribution

12. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 12 ) Surface Loss Function Loss to surface states occur for events detected at low z (< buried channel depth of a few microns) Model with a loss function, Q o = Q i x f(z) where f(x) = S + B (z/l) ^c (z<l) = 1.0 – A exp(-(z-l)/tau)

13. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 13 ) Operation controlled by an input file – Sets input source (spectrum, spatial model), pixel geometry, mode, thresholds, CTI, read noise, charge spreading params, etc Pixel geometry file

14. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 14 ) The Importance of Getting Good Grades Make image from PHAS array for each frame. Find next max PHA value in image above threshold Extract 3x3 Grade 3x3 Remove 3x3 from image Repeat until no pixels above threshold Or testing simulator PC event recognition algorithm on door source data Original algorithm: Mismatch in the redistribution tail

15. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 15 ) The Importance of Getting Good Grades (II) Loop through each pixel (x, y) Test if PHA > thresh Test if pixel is the local max in 3x3 Grade 3x3 Leave 3x3 in image array Repeat over all pixels Final algorithm:

16. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 16 ) The Importance of Getting Good Grades (iii) Find next max PHA above threshold in 200 pixel readout array Extract 7 pixel array centred on the max PHA pixel Grade 7 pixel array, taking care with “don't care” pixels Mask out pixels in grade Repeat until no pixels remain above threshold WT mode test Algorithm:

17. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 17 ) Simulation Results Best door source modelling in terms of grade distribution and spectral redistribution for both PC and WT achieved for – Open electrode depletion depth = 27 micron, FF depth = 53 micron – Closed electrode DD = 37 micron, FF depth = 43 micron – Depletion Na = 2e13 cm^-3, S=1 – Field-free diffusion length, Lff = 250 micron. – Substrate diffusion length, Ls = 2 micron

18. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 18 ) Results - PC Si Kα mean free path X 1.7 to produce line flux

19. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 19 ) Results - WT

20. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 20 ) PC spectra – by grade Black solid – door data grade 1-4 Black dashed – door data grade 5- 12 Blue – grade 1-4 simulation Green – grade 5-12 simulation

21. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 21 ) Further refinement ? Drop depletion depth near vertical edges → variation in grades 1,3 & 2,4 events No visible effect on spectrum, however.

22. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 22 ) WT door & point source comparison Point source simulation (red 170c/s, yellow 105c/s) – grade 1-2 bump is more prominent than for the uniform, door source case

23. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 23 ) WT - absorbed source GX301-2 data taken in 2010-10 Simulation Model: phabs*power. (NH=1.68e23cm^-3) – NB event threshold is ~300 eV !

24. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 24 ) WT simulation – power spectrum Fall-off in power at high frequencies, as seen in real WT data from hard sources Origin is event splitting at the 10 row boundary of multipixel events. Converted PSF Mn-Kα source event file to light curve and made a powerspectrum

25. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 25 ) WT simulation – power spectrum (cont'd) No fall-off in power at high frequencies, as seen in real WT data from soft sources. Powerspectrum of a soft source (O-Kα line) simulation – i.e. nearly all single pixel events

26. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 26 ) Pile-up Simulator naturally reproduces pile-up

27. Swift-XRT Team Meeting – Clemson, October 23 rd, 2011 ( Page 27 ) ToDo Presently have an event list generator, rather than an RMF generator Check the redistribution as a fn of energy – Create energy dependent loss fn, based on existing RMF redistribution, if necessary Check QE - have a reasonable idea what it should be. Make RMFs !