The in-flight spectroscopic performance of the Swift XRT CCD camera during 2006-2007
ABSTRACT The Swift X-ray Telescope focal plane camera is a front-illuminated MOS CCD, providing a spectral response kernel of 135 eV FWHM at 5.9 keV as measured before launch. We describe the CCD calibration program based on celestial and on-board calibration sources, relevant in-flight experiences, and developments in the CCD response model. We illustrate how the revised response model describes the calibration sources well. Comparison of observed spectra with models folded through the instrument response produces negative residuals around and below the Oxygen edge. We discuss several possible causes for such residuals. Traps created by proton damage on the CCD increase the charge transfer inefficiency (CTI) over time. We describe the evolution of the CTI since the launch and its effect on the CCD spectral resolution and the gain. Comment: 8 pages, 5 colour figures, submitted to SPIE
The in-flight spectroscopic performance of the Swift XRT
CCD camera during 2006-2007
O. Godeta, A. P. Beardmorea, A. F. Abbeya, J. P. Osbornea, K. L. Pagea, L. Tylera, D. N.
Burrowsc, P. Evansa, R. Starlinga, A. A. Wellsa, L. Angelinib, S. Campanad, G. Chincarinid,e,
O. Citteriod, G. Cusumanof, P. Giommig, J. E. Hillb, J. Kenneac, V. LaParolaf, V. Manganof,
T. Mineof, A. Morettid, J. A. Nousekc, C. Paganic, M. Perrig, M. Capalbig, P. Romanod,e, G.
Tagliaferrid, F. Tamburellig
aUniversity of Leicester, University Road, Leicester, LE1 7RH, UK;
bNASA-GSFC, Greenbelt, MD 20771, USA;
cPennsylvania State University, 525 Davey Lab, University Park, PA 16802, USA;
dINAF-Osservatorio Astronomico di Brera, Via E. Bianchi 46, 23807, Merate, LC, Italy;
eUniversit` a degli Studi di Milano, Bicocca, Piazza delle Scienze 3, I-20126, Milano, Italy;
fINAF-IASF, Via U. La Malfa 153, 90146 Palermo, Italy;
gASI-ASDC, Via G. Galilei, I-00044 Frascati, Italy
The Swift X-ray Telescope focal plane camera is a front-illuminated MOS CCD, providing a spectral response
kernel of 135 eV FWHM at 5.9 keV as measured before launch. We describe the CCD calibration program
based on celestial and on-board calibration sources, relevant in-flight experiences, and developments in the CCD
response model. We illustrate how the revised response model describes the calibration sources well. Comparison
of observed spectra with models folded through the instrument response produces negative residuals around and
below the Oxygen edge. We discuss several possible causes for such residuals. Traps created by proton damage
on the CCD increase the charge transfer inefficiency (CTI) over time. We describe the evolution of the CTI since
the launch and its effect on the CCD spectral resolution and the gain.
Keywords: CCD, X-rays, spectroscopy, Charge Transfer Inefficiency
The Swift gamma-ray burst satellite1was successfully launched on 2004 November 20.
provided observations and positions of GRBs and their afterglows to observers and robotic telescopes typically
within a minute, thanks to its three instruments: the wide-field Burst Alert Telescope2and the two narrow-field
instruments, X-Ray Telescope3and UV/Optical Telescope4. When a GRB is detected by the BAT and a slew
is possible, Swift automatically re-points to bring the burst within the field of view of the XRT and the UVOT.
The XRT uses a grazing incidence Wolter-1 telescope consisting of a thermally controlled carbon fibre tele-
scope tube, an X-ray mirror system of 12 concentric gold-coated electroformed Ni shells with a 3.5m focal length,
and a Focal Plane Camera Assembly (FPCA) housing an e2v CCD-22 with 600 × 602 image pixels, located
behind an optical blocking filter with an optical transmission of about 0.25%. The CCD is mounted on a ther-
moelectric cooler connected via a heat pipe to an external radiator. The FPCA also includes an autonomous Sun
shutter, four55Fe calibration sources and a substantial mass of Al proton shielding (which also reduces thermal
variations). To avoid the effects of pile-up, the XRT is able to autonomously select one of the three following
readout modes5according to the source brightness: Photo-diode (PD) mode at highest count rates with a 0.14
ms time resolution and no spatial information; Windowed Timing (WT) mode at moderate count rates with a
Since then, it has
Further author information: (Send correspondence to O.G.)
O.G.: E-mail: email@example.com, Telephone: +44 116 223 1039
A.P.B.: E-mail: firstname.lastname@example.org, Telephone: +44 116 252 3583
arXiv:0708.2988v1 [astro-ph] 22 Aug 2007
1.8 ms time resolution and 1-D spatial information; at lower count rates Photon Counting mode (PC) with a 2.5
s time resolution and 2-D spatial information.
During the first year in orbit two major incidents occurred which required modification of the instrument’s
operation, although they have not affected its scientific productivity. First, before the CCD was cooled to its
nominal operating temperature of −100◦C, the XRT thermo-cooler (TEC) power supply system apparently failed,
and therefore the XRT has to rely on passive cooling via the heat pipe and radiator in combination with enhanced
management of the spacecraft orientation to reduce the radiator view of the sunlit earth. In flight the XRT is
nowadays operated with CCD temperatures of -75 to -52◦C (see Kennea et al.6for more details). Secondly, on
2005 May 27 the XRT was hit by a particle (micro-meteoroid) which scattered off the mirror system to hit several
CCD pixels, causing new bright pixels, one bright column and two bright column segments7. Charge leakage
from the top of the bright column affects its immediate neighbours. The evolution of charge leakage depends
on the CCD temperature which can now only be controlled by orientation of the spacecraft. After this event,
the optical filter showed no sign of damage. Similar events were also observed in the XMM-Newton EPIC MOS
CCDs. The bright pixels and columns have been vetoed on-board for the PC and WT modes. This is impossible
in PD mode so that mode is no longer used, which had the beneficial effect of reducing the CCD temperature.
These two incidents had no direct impact on the spectroscopic performance of the XRT, or its ability to image
and locate new GRBs. Indeed, the XRT routinely measures the early X-ray light-curves and spectra of all the
GRB afterglows at which it is promptly pointed (hence 95% of the 183 GRBs detected by the BAT for which a
spacecraft slew was possible within 5 minutes after the BAT trigger were detected by the XRT up until the middle
of July 2007; the BAT has detected a total of 243 GRBs up until the middle of July 2007). These observations
have revealed previously unexpected multiple breaks8,9and flares in the early X-ray light-curves8,10–13suggesting
extended activity of the central engine up to 105s after the trigger for long and short GRBs; which challenges
the current progenitor models. Essential spectral and temporal information was also obtained with the XRT
for the peculiar event GRB 060218 showing for the first time the rise of a supernova14. The XRT has also
discovered the first short burst afterglow15and provided accurate locations of several short GRBs, which are
associated with elliptical galaxies or star-forming galaxies, the burst being located in the outskirts of the galaxy
in the latter case; these results tend to indicate that short GRBs are likely to be due to binary compact object
spiral-in and collision16,17. The XRT also provides essential information for non-GRB targets, for example with
the follow-up of the recurrent nova RS Ophiuchi18or the micro-quasar GRO J1655-4019. The fraction of time
spent on non-GRB targets (excluding the calibration targets) is ∼ 39.4% since the launch, and this is expected
to increase over time.
In this paper, we concentrate on the XRT CCD in-flight spectral calibration, and describe recent improvements
to the response model made available as response matrices through the HEASARC caldb (RMF version 00920).
The results obtained during the pre-flight calibration were covered in Osborne et al.21.
2. IN-FLIGHT CALIBRATIONS
Since the FPCA front door was opened, spectroscopic calibrations have been only performed using a set of well
known celestial objects observed every six months in order to monitor the change in the spectral response (see
Table 1). The fraction of time spent on calibration targets is ∼ 7.5% since the launch. Many of our calibration
targets are also used by other X-ray observatories. This allows us to perform cross-calibration campaigns with
different X-ray instruments (e.g. the XMM-Newton EPIC MOS cameras for the quasar 3C 273 and the blazar
Mkn 421). In addition, we make use of four55Fe calibration sources permanently illuminating the non-imaging
area of the CCD. In order to optimise and facilitate the scheduling of the calibration targets, a new state was
implemented on-board the XRT in 2007 May. This new state allows us, for instance, to control the observation
mode while the XRT is still in auto-state depending on the purpose of the observations.
3. RESPONSE MODEL DEVELOPMENTS FROM JANUARY 2006 TO JUNE 2007
When an X-ray photon interacts within the CCD, it generates a charge cloud which is collected in the depletion
region after spreading in the bulk of the detector. The charge cloud may spread into more than one pixel
depending on its energy and location of interaction. To compute the response matrices we stack simulated
Table 1. Summary of the in-flight calibration targets used up to July 2007.
Low energy response
2E 0102-7217 PC
Gain, energy resolution and
Energy scale offset, gain, shoulder,
CTI and energy resolution
Effective area and cross-calibration
Effective area (cross-calibration
Effective area and cross-calibration
Cas A SNR
3C 273 Quasar
CrabPulsar WT 46
spectra of monochromatic X-rays. We use the grade recognition process used in the analysis software (for PC
mode this is a 3 × 3 pixel matrix centred on the highest pixel). To avoid noise being included in the charge
summation and excessive telemetry usage, a threshold is set on-board below which pixels are not considered.
The X-ray spectrum resulting from monochromatic radiation significantly differs from a simple Gaussian, it
consists of six components: a Gaussian peak with a shoulder on the low energy side of the peak, an escape peak
and a Si Kα fluorescence peak if the photon energy is above the Si K-shell edge, a shelf extending to low energies,
and at the very lowest energies a noise peak (of which only the high energy side may be seen above threshold).
Below 1.5 keV, the shoulder and shelf are mainly produced by the charge losses at the interface between the
SiO2layer and the active silicon volume of the open electrode, possibly due to a local inversion of the electric
field near the detector surface. Above 1.5 keV, the shoulder and the shelf are produced by several processes:
(i) sub-threshold losses; (ii) recombination and trapping in the bulk of the detector; (iii) inhomogeneity of the
electric field in the depletion depth, in addition to the surface losses. Their exact shapes depend on the readout
mode, and hence they are different for the PC and WT modes. Both above and below 1.5 keV, the main peak
broadens with time due to the degradation of the charge transfer efficiency, producing a larger shoulder at lower
energies (see Section 5).
Changes in our spectral response code have been made in order to better reproduce the different components
mentioned above. The recent improvements described below have been released as new spectral response files
3.1 The shelf from photons above ∼ 2 keV
The new response matrice files (v009 RMFs) include an empirical rescaling of the low energy shelf made by
X-rays above 2keV, which significantly improve the quality of spectral fits to the calibration sources (see Fig. 1).
This was previously incompletely modelled for either PC or WT modes, resulting in an underestimation of the
modelled redistributed counts when fitting spectra of heavily absorbed sources (e.g. NH≥ 1022cm−2).
3.2 The shoulder from photons above ∼ 1.5 keV
Before the v009 release of the calibration files, the shoulder, which was modelled by artificially increasing the
event split threshold in order to increase the sub-threshold losses, was not reproduced well (see Fig. 6a in Osborne
et al.21). We showed that modification of the shape of the charge cloud formed in the field-free region using the
Figure 1. Left: XRT PC grade 0-12 spectrum of NGC 7172 using the new v009 (red) and previous v008 (black) response
files. The use of a WABS*POWERLAW model gives a value of NH ∼ 7.3 × 1022cm−2. Right: XRT WT grade 0-2
spectrum of the X-ray binary 4U 1608 using the new v009 (red) and previous v008 (black) response files. The use of a
WABS*(POWERLAW+DISKBB) model gives a value of NH ∼ 1022cm−2.
formalism described in Pavlov & Nousek22can lead to a more physical modelling of the shoulder (see Fig. 6b in
Osborne et al.21), the shape being no longer a 2-D Gaussian (although a 2-D Gaussian remains sufficient in the
depletion region). Fig. 2 shows the relatively good agreement of the model with the very complicated spectrum
of the SNR Cas A (the North knot as noted in Fig. 2).
Figure 2. Left: PC grade 0 image of the SNR Cas A. Right: PC grade 0 spectrum of the North knot of the SNR Cas A
(see the image on the left). The data were taken in 2005 February 17. The spectrum is fit using a PHABS(VNEI+VNEI)
model (see Willingale et al.24).
4. THE RESIDUALS AROUND AND BELOW THE OXYGEN EDGE
The fits of several continuum sources reveal negative (less than 20%) residuals around the Oxygen edge (0.54
eV) using v008 and v009 RMFs.
Recently, observational evidence has shown that the CCD bias level can significantly vary during the timescale
of an orbital snapshot on an astrophysical target. The bias level is mode-dependent and is subtracted on-
board during the XRT observations. Bias variations during a snap-shot can occur due to changes in the CCD
temperature and/or scattered optical light from the sunlit Earth23. This can result in energy scale offsets, which
give rise to residuals when fitting spectra, especially at low energies (e.g. around the Oxygen edge). Energy
scale offsets can be seen in both PC and WT modes. A new command option was implemented in the 2.6 XRT
software (WTBIASDIFF) in order to correct the WT data for this effect. A tool XRTPCBIAS included in the 2.7
XRT software∗has been released in July 2007 by the XRT software team, which corrects the PC data. The
use of these contemporary time-dependent bias estimators can significantly improve the energies of low energy
events (see Fig. 3).
Charge transfer inefficiency due to accumulating proton damage also results in a slow change in the energy
scale. This is discussed in Section 5.
Figure 3. Left: Comparison of the energy centroid of the Si and S lines in the North (N) and South East (SE) knots of the
SNR Cas A as observed by the XMM MOS cameras (N: black; SE: red) and the XRT (N: green, cyan; SE: blue, magenta).
The green and blue crosses correspond to XRT/PC grade 0 data for which the bias was contaminated by optical light
from the sunlit Earth. The data were processed with the CALDB2.6 software which does not allow to bias corrections. In
this case, an energy scale offset is observed when compared with the XMM MOS curves. The cyan and magenta crosses
correspond to the same data processed with the new CALDB2.7 software including the tool XRTPCBIAS, which corrects
the data. Right: XRT WT grade 0-2 spectrum of RS Ophiuchi: (black) the data not corrected to the bias problem and
(red and blue) the data corrected using the new command option WTBIASDIFF.
Cross-calibration performed with other X-ray instruments in orbit such as XMM-Newton and Suzaku on the
supernova remnant 2E 0102-7217 reveals that at least these CCD cameras seem to suffer from an overestimation
of the model with respect to the data around the Oxygen edge. Recently, the MOS calibration team has decided
to apply an ad hoc correction to their quantum efficiency (QE) by decreasing the QE below the Oxygen edge
by 10-15%, in order to correct the residuals. We are investigating whether a similar approach could work in the
case of the XRT, once the effects of energy scale offset have been corrected.
5. CHARGE TRANSFER INEFFICIENCY
CCD detectors provide good X-ray imaging and spectroscopic performance. However, the CCD energy resolution
and gain degrade with time due to the increase of the CTI. The main origin of CTI is the increase of charge
traps, which are mainly due to the irradiation of high-energy protons on the CCD passing through the shielding.
Although the low-Earth orbit of Swift and the thick Al shielding around the CCD detector reduce the proton
flux, the frequent passages of the spacecraft through the South Atlantic Anomaly (SAA) can cause formation
of charge traps, and hence an increase of CTI. Since the launch, the FWHM measured using the four55Fe
calibration sources (located in each corner of the CCD; the area of the detector covered by these corner sources
being small and outside the imaging area) increased from 146 eV at 5.9 keV in Feb 2005 to 210 eV in March
2007 when using the bad and good columns (i.e. columns with and without significant traps), respectively. The
broadening of the line is due to the energy scale shifting effect of traps in the pixels through which the charge
has to be transported. The increase of traps in the CCD imaging area, the serial register and the frame-store
area also causes energy scale offset as shown in the left panel in Fig. 4. In this Figure, the data were processed
with a gain file not corrected from the CTI increase.
∗see the following URL: http://heasarc.gsfc.nasa.gov/docs/software/lheasoft/release−notes.html