Advanced CCD imaging spectrometer (ACIS) instrument on the Chandra X-ray Observatory

TLDR: The ACIS instrument has been operating for three years in orbit, producing high quality scientific data on a wide variety of X-ray emitting astronomical objects, except for a brief period at the very beginning of the mission when the CCDs were exposed to the radiation environment of the Outer van Allen Belts which resulted in substantial radiation damage to the front illuminated CCD.

Abstract: The ACIS instrument has been operating for three years in orbit, producing high quality scientific data on a wide variety of X-ray emitting astronomical objects. Except for a brief period at the very beginning of the mission when the CCDs were exposed to the radiation environment of the Outer van Allen Belts which resulted in substantial radiation damage to the front illuminated CCDs, the instrument has operated nearly flawlessly. The following report presents a description of the instrument, the current status of the instrument calibration and a few highlights of the scientific results obtained from the Guaranteed Observer Time.

Figures

Figure 2.3a Figure 2.3b

Figure 3.4.2

Figure 1 background events. A much more complete description of the

Figure 3.1.1 Figure 3.1.2

Figure 2.1 tilting the CCDs in

Figure 2.2 posts are made from a plastic Torlon to insulate the focal plane

Full text

Advanced CCD imaging spectrometer (ACIS) instrument on the
Chandra X-ray Observatory
G. P. Garmire
*a
,M.W.Bautz
**b
,P.G.Ford
b
, J. A. Nousek
a
, and G. R. Ricker Jr
b
a
The Pennsylvania State University;
b
Massachusetts Institute of Technology
Abstract
The ACIS instrument has been operating for three years in orbit, producing high quality scientific data on a wide variety
of X-ray emitting astronomical objects. Except for a brief period at the very beginning of the mission when the CCDs
were exposed to the radiation environment of the Outer van Allen Belts which resulted in substantial radiation damage to
the front illuminated CCDs, the instrument has operated nearly flawlessly. The following report presents a description
of the instrument, the current status of the instrument calibration and a few highlights of the scientific results obtained
from the Guaranteed Observer Time.
1. Introduction
The Advanced CCD Imaging Spectrometer (ACIS) on board the Chandra X-ray Observatory(CXO) [1] is a powerful
tool for conducting imaging, spectroscopic and temporal studies of celestial X-ray sources. The instrument consists of
ten Charge Coupled Devices (CCDs) especially designed for efficient X-ray detection and spectroscopy [2]. Four of the
front illuminated (FI) CCDs are arranged in a square array with each CCD tipped slightly to better approximate the
curved focal surface of the Chandra Wolter type I mirror assembly. The remaining six CCDs are set in a linear array,
tipped to approximate the Rowland circle of the objective gratings that can be inserted behind the mirrors (see Figure
2.1). One CCD next to the center-line of the grating array is essentially flat and is a back illuminated (BI) CCD that is
useful for imaging soft X-ray objects. Each CCD subtends an 8.4 arc minute by 8.4 arc minute square on the sky. The
individual pixels of the CCDs subtend 0.492 arc seconds on the sky. The on-axis performance of the telescope is better
than 0.5 arc second. The spacecraft normally is commanded to conduct an observation utilizing a dither motion in the
form of a Lissajous pattern over a 16 arc second square area of sky. This motion slowly moves (typically, 0.1 arc second
per CCD exposure) any X-ray object across a 32 by 32 pixel region of the CCDs. It is possible to deconvolve the images
obtained from the CXO/ACIS to better than 0.5 arc second resolution. In Figure 1 the image of SN1987A, the most
recent supernova in the LMC, is resolved even though it subtends
only about 1.3 arc seconds (see Burrows, [3] for more details).
The CCDs utilize a framestore design, such that no shutter is
required, the image being shifted to the framestore area in 41 ms,
very short compared to a typical exposure of 3.24 seconds. The
frame transfer causes streaking for very bright sources, but more
typical exposures have no bright source in the field. Shorter
exposures are possible by using only a portion of the CCD area,
but with a loss of field on the sky and observing efficiency. The
CCDs may also be used in a continuous clocking mode to achieve
a line readout time of 2.85 ms (equivalent to a time resolution of
~6 ms for a point source) but at the loss of one spatial dimension.
The background on orbit is very low (6 x 10
-8
count/pix/sec in the
0.5 – 2.0 keV band and 1.3 x 10
-7
counts/pix/s in the 2.0 – 8.0
keV band) as obtained from the dark moon observations [4 ].
Even in a 2Ms exposure [5] 40% of the pixels are free of
Figure 1 background events. A much more complete description of the
instrument is available at the web site listed at the end of this article. This report is intended to give an over view of the
instrument with sufficient detail to aid the potential observer in understanding how the instrument works and what its
capabilities are for observations.
X-Ray and Gamma-Ray Telescopes and Instruments for Astronomy, Joachim E. Trümper,
Harvey D. Tananbaum, Editors, Proceedings of SPIE Vol. 4851 (2003)
© 2003 SPIE · 0277-786X/03/$15.00
28
Downloaded from SPIE Digital Library on 07 Sep 2010 to 128.103.149.52. Terms of Use: http://spiedl.org/terms

.
2. Focal Plane Design
The ACIS focal plane (Figure 2.1) is composed of two arrays with one array designed to optimize imaging and the other
array for
spectroscopy
utilizing the
objective gratings
that are part of the
Chandra X-ray
Observatory
(CXO). In the case
of the imaging
array, the CCDs
are tipped to lie
tangent to the
optimum focal
surface of the
Wolter Type I
mirrors employed
for imaging. This
nearly spherical
focal surface lies
10.04 meters from
the joint separating
the paraboloid and
hyperboloid
mirrors and has a
radius of curvature
of 85 mm. By
Figure 2.1 tilting the CCDs in
the imaging array more pixels encompass regions of high angular resolution (see Figure 2.3 for a comparison between a
flat array and a tilted array. The dashed curve in (a) starting at 7 arc seconds is for the S-array portion of the image). The
spectroscopic array is also tilted, but to a lesser degree, since the Rowland circle for the gratings has a radius of
curvature of 4.2 m. A detailed layout of the focal plane is shown in Figure 2.1. The position of the detected X-ray in the
spectroscopic array is directly proportional to the wavelength
of the photon. The intrinsic energy resolution of the CCDs is
used to separate the overlapping orders of the dispersed
spectrum. The spectroscopic array is composed of two
different CCD designs, four (FI) three phase CCDs and two
(BI) three phase CCDs , S1 and S3 in Figure 2.1. The (BI)
device labeled S3 is essentially flat (perpendicular to the
optical axis of the telescope) and provides an additional
imaging capability, especially for sources with mainly low
energy X-ray emission (below 1 keV) where the BI CCDs have
superior quantum efficiency. Fabrication limitations require
small gaps to exist between the CCDs (~11 arcsec for the I-
array and 8.8 arcsec for the S-array). Dithering of the
spacecraft in a Lissajous figure over a 16 arc second square
pattern during an observation fills in these gaps to some extent.
The actual flight focal plane is shown in Figure 2.2. The four
Figure 2.2 posts are made from a plastic Torlon to insulate the focal plane
from the rest of the camera body. Gold-coated aluminum bars cover all of the framestore areas to shield them from
focused X-rays. The curved straps attached to each CCD are flexprints which carry the electrical signals to and from
each CCD.
Spacecraft coordinates are used with
outer edges of the array.
Dimensions marked * apply to the
Datum dimensions apply to the active area,
at operating temperature.
Dimensions are in inches [mm]
ACIS Focal Plane Array
other dimensions apply to the silicon.
the origin of the on-axis focal point
of the HRMA when the SIM is at IP#1.
.000 [0.00]
IP#1
.057 [1.46]
.078 [1.99]
Proc. of SPIE Vol. 4851 29
Downloaded from SPIE Digital Library on 07 Sep 2010 to 128.103.149.52. Terms of Use: http://spiedl.org/terms

.
Figure 2.3a Figure 2.3b
3. CCD Characteristics
The ACIS CCDs were developed at the MIT Lincoln
Laboratory on high-purity p-type float-zone wafers of silicon
with resistivities of about 7000 ohm-cm. A complete
description can be found in [2]. Each CCD is a
1024x1026-pixel frame-transfer imager which is divided into
four sectors. As shown in Figure 3.1, the framestore is split,
and the framestore pixels are smaller than those in the imaging
area. This arrangement allows for four independent output
amplifiers, and facilitates 3-side abutment of the detectors.
Each amplifier is a floating-diffusion output circuit with
responsivity of 20 microvolts/e
-
and a noise of about 2 e
-
RMS
at the operating rate of 100 kpix/s All four amplifiers are
operated in parallel to minimize the frame readout time, which
is typically 3.24 s. The room temperature dark currents were
usually found to be about 500 pA/cm
2
with about 50 pA/cm
2
coming from the bulk emission and the remainder from
surface-state dark current. By operating the device below -100
C and with frame times less than 10 s, the dark current per pixel
is less than 1 e
-
/pixel/frame. Each pixel is 24 by 24 microns
which at the focal surface of the CXO corresponds to 0.492 arc
seconds. A cross section of the FI and BI CCDs is shown in
Figure 3.2. The FI CCD design results in the charge from
cosmic rays being generated in a nearly field free region of the
CCD which spreads over a large number of pixels by the time it
reaches the buried channel where it is stored for readout.
Figure 3.1 Also, high energy X-rays interact deeper in the silicon on
average thus giving them more opportunity to diffuse into a large charge cloud as illustrated in the Figure 3.2. The BI
CCDs, on the other hand, are just the reverse for X-rays, where the low energy X-rays interact far from the buried
channel and diffuse into a larger charge cloud. Cosmic rays generate much smaller charge clouds in the BI architecture,
since there is a much thinner field free region near the back surface. Nevertheless, the background rejection efficiency is
much higher in the FI CCDs, with the result that the background is higher in the BI CCDs by a factor of 2 –3, depending
on the energy.
The CCD design used for the ACIS includes a trough implant in the buried channel that increases the ability of the
device to tolerate charge particle radiation encountered in the space environment. Two types of radiation damage are
Image Array
Frame Store
Shift Registers
Output Nodes
Row 0
Row 1
Row 2
Row 1025
Row 1
Row 1025
Row 0
Row 2
Column 0
Column 1
Column 511
Column 512
Column 1023
Dummy 0
Dummy 3
Node A Node B
(1026 rows by
1024 columns)
(1026 rows by
1024 columns)
(260 pixels per Node)
Node C Node D
30 Proc. of SPIE Vol. 4851
Downloaded from SPIE Digital Library on 07 Sep 2010 to 128.103.149.52. Terms of Use: http://spiedl.org/terms

.
typical of silicon based CCDs; displacement damage and
ionizing damage. The former is induced by protons and other
heavy particles in the cosmic ray flux, while the latter is
induced by electrons or UV or X-ray photons that can deposit a
charge in the insulating layer separating the gates from the
buried channel. This has the effect of changing the potential
well created by the clock voltages which in tern can affect the
charge transfer efficiency. Over a five year baseline life of the
Chandra Mission the accumulated dose of the displacement
damage was estimated to be 1-3 kRad, while the ionizing dose
is computed to be less than 10 k Rad (however, see section 3.1
concerning On-Orbit anomaly). These levels of displacement
damage will produce a measurable increase in Charge
Transfer Inefficiency (CTI) over a five-year period.
The CCDs were carefully calibrated in the laboratory at MIT
using ten X-ray lines on 32 by 32 pixel regions of theCCDs to
match the dither pattern size. This process smoothes out pixel
to pixel variations and reduces the calibration time by a factor
of one thousand. The calibrations were compared to a reference
CCD that was calibrated at the synchrotron light source at
BESSY in Berlin, Germany. The overall calibration was
deemed to be accurate to a few percent ([6], ACIS Calibration.
Report). An Optical Blocking Filter (OBF) was placed about 2
centimeters above the CCDs to limit the
Figure 3.2 amount of light that could reach the CCDs from stars and from
scattered light in the spacecraft. The filter was supplied by the Luxel Corporation using a 200nm substrate of free
standing polyimide and coated with 160 nm of aluminum over the imaging portion of the array and 130 nm of aluminum
over the spectroscopy array where less visible light was anticipated since much of it would be dispersed by the gratings.
The X-ray transmission was calibrated over the same 32 by 32 pixel equivalent areas at the synchrotron light source at
the University of Wisconsin. The transmission over the range 200 to 1200nm range was measured at Penn State, at the
Denton Vacuum Corp using a Perkin Elmer UV/VIS/Near IR spectrometer and at Brookhaven National Laboratory. The
results of these calibrations are included in the ACIS Calibration
Report referenced above. The observational consequences of
the transmission of visible light are discussed below.
The overall, average quantum detection efficiency of the two
types of CCDs used on ACIS is shown in Figure 3.3. The fine
structure near the absorption edges (EXAFS) have been omitted
in this figure. The range of variations from CCD to CCD is
about 10% and within a FI CCD the variations are less than 3%
from pixel to pixel. The variations within the BI CCDs are
greater and reach about 15% for extreme cases. The QE of the
CCDs is dependent upon what combination of pixels is used to
create each X-ray event. It is frequently the case, depending
upon X-ray energy, that some of the charge of the X-ray induced
ionization reaches more than one pixel. The onboard processing
Figure 3.3 electronics can be programmed to send several different pixel
combinations to the ground through the telemetry of the spacecraft. Normally, a 3x3 pixel island (faint mode) is
transmitted, with the central pixel containing the peak charge. An event threshold is set in the front end processor to
only select events above the threshold. A second threshold is used by the processor to detect the charge from the
adjacent pixels, the “split event threshold’ which is usually lower than the event threshold. Each event is generally
“graded” as determined by the pattern of split events. The QE will depend upon which grades are selected. For
example, if only unsplit events are selected, the QE will be less than events where splits are allowed. See the above
referenced ACIS Calibration Report for details.
Gate Structure
~0.5 µm
(deadlayer)
Depletion
Region:
50-75 µm
Field-free
Region:
40-200 µm
Front-Illuminated X-ray CCD Structure
( not to scale)
Phase 2Phase 1
Phase 3
Phase 2Phase 1
Phase 3
Buried Channel (charge collection & transfer)
Depletion Region:
•E≠0
•Little lateral diffusion
•Good charge collection;
•Accurate spectroscopy.
Field-Free Region:
•E=0
•Significant lateral diffusion
•Poor charge collection
•Poor spectroscopy
•Good "anti-coincidence"
layer for charged-particle
rejection
1 pixel (24 µm)
Incident X-rays
pixel boundary Ne = Ex/3.65 eV
Phase 2Phase 1
Phase 3
Phase 2Phase 1
Phase 3
Buried Channel (charge collection & transfer)
Depletion Region:
•E ≠0
•Little lateral diffusion
•Good charge collection;
•Accurate spectroscopy.
Backside Surface & Implant:
• ~100 A SiO2 deadlayer
• 300-600 A B implant partially depleted
• Surface recombination significant
• Lateral diffusion could be significant
• Largest effects on lowest energies
1 pixel (24 µm)
Incident X-rays
pixel boundary
Ne = Ex/3.65 eV
High-energy
events typically
suffer little
diffusion
Low-energy
events typically
suffer larger
lateral diffusion, require multi-pixel summation
Low-energy
events typically
suffer little
diffusion
High-energy
events typically
suffer larger
lateral diffusion, require
mulit-pixel summation
Back-Illuminated X-ray CCD Structure
( not to scale)
Depletion
Region:
45 µm
Gate Structure
~0.5 µm
(deadlayer)
Backside Surface
& Implant
Proc. of SPIE Vol. 4851 31
Downloaded from SPIE Digital Library on 07 Sep 2010 to 128.103.149.52. Terms of Use: http://spiedl.org/terms

.
A detailed calculation was made to determine the sensitivity of the CXO with ACIS to stray light. First, it was important
to protect the CCDs from receiving light scattered in the telescope structure and from the background sky, since the
CCDs are extremely sensitive to visible light. Secondly, light from the object under study must also be rejected. In the
case of a bright star or planet this is a challenging requirement because the thickness of the filter for visible and near IR
light also affects the transmission of the filter for low energy X-rays. The filter choices are given above. The aluminum
was coated on both sides of the polyimide for electrical grounding, with 30nm on one side and the remainder on the
other. The measured transmission curves are given in the ACIS Calibration Report. The limiting magnitudes for various
configurations of ACIS and are given in Table 3.1. It is worth noting that the effects of light from bright objects can be
circumvented, somewhat, by using a 5x5 island (very faint mode) for each event and then using the outer 16 pixels of the
island to recompute a bias for the event. In addition, the event threshold must be increased so that any bias offset created
by the visible light does not trigger the event recognition algorithm. As an example, for Jupiter at opposition, the event
threshold should be increase from 20 ADU to 53 ADU and the split threshold from 13 ADU to 46 ADU to prevent every
pixel over the disk of the planet from triggering the threshold for the BI CCD. In the initial observations of Jupiter, the
event threshold was not raised, and every pixel illuminated by the disk of Jupiter produced an event. Since the on board
processor was set to reject all events in which all of the pixels in the 3x3 island was above either the event or split
thresholds, most of the events, including the X-ray events, were discarded on orbit. By inserting the MEG/HEG an
optical attenuation of about 2.6 magnitudes is introduced.
Table 3.1
The magnitude for which one electron is generated per 3.24 s exposure from a point source. It takes 13 electrons to
exceed the split event threshold and 20 to exceed the event threshold for the BI CCD.
Stellar Temperature (K) BI Chip in S-Array
(V-magnitude)
FI Chip in I-Array
(V-magnitude)
3000
4000
5000
6500
10000
20000
8.5
8.3
7.7
7.3
6.8
6.7
3.5
3.2
2.6
2.2
1.6
1.4
3.1 On-Orbit anomaly
The ACIS was launched with the S3 CCD at the aimpoint of the telescope in the launch-lock position. Initial
performance of the detectors determined by the Fe
55
in the ACIS door, was found to be consistent with the ground
calibration. After the ACIS door was opened on day 220 of 1999, and the mirror covers were opened on day 225, the
ACIS S3 CCD was used to obtain the focus of the SIM and boresight information on the relative alignment of the SIM
and the star tracker. Then a number of observations were conducted using both the ACIS and the HRC until day 252,
when the ACIS was placed under the calibration source to record calibration data. The data showed that there had been
a very large increase in the FI CCD charge transfer inefficiency (CTI)! Such a change was totally unexpected, based on
all earlier analyses of radiation damage. A number of emergency meetings were held to determine what the cause of
such a large increase might be. At a meeting held at Lincoln Laboratory on 09/23/99, a nuclear physicist suggested that
the most likely cause of the damage was 100 keV protons from the outer trapped radiation belt forward scattering
(Rutherford scattering) off of the very smooth iridium coating on the X-ray mirrors. These protons would not be
deflected significantly by the magnets and would reach the CCDs with nearly their original energies. Given the high
flux of these protons, a significant dose to each CCD could be expected. By using the Science Instrument Module (SIM)
translation to place the ACIS behind a shield for each radiation belt passage, the CCDs could be protected. Subsequent
analyses have confirmed this suggestion and the protection procedure has worked very well since it was implemented.
32 Proc. of SPIE Vol. 4851
Downloaded from SPIE Digital Library on 07 Sep 2010 to 128.103.149.52. Terms of Use: http://spiedl.org/terms

Frequently asked questions

Q1. How is the BEP commanded to ignore subsequent frames?

With the downlink bandwidth limited to 24 kbps, the BEP’s output buffers are soon saturated and the FEPs are commanded to ignore subsequent frames until space becomes available. 

Q2. How many analog processing chains can be used in the ACIS system?

There are a total of 40 analog processing chains within the DEA, one for each of the 4 output nodes of each CCD, but no more than 24 chains can be in use during an observing run. 

Q3. How many kByte RAM instruction caches does the BEP have?

The BEP has a 512 kByte RAM instruction cache and 256 kByte data cache, each with 35 ns access time to support the Mongoose running at 10 MHz. 

Q4. How long does it take to read the entire framestore?

Since it takes approximately 3.2 seconds to read the entire framestore, typical full-frame exposures last 3.2 4seconds and the frame-to-frame time is 3.241 seconds. 

Q5. How long has the CXO been making observations?

The CXO has been making observations for the past three years, interrupted only by ~8 hour periods once per orbit when the observatory comes within ~60,000 km of Earth, and during occasional high radiation events emitted by the active Sun, which has just gone through solar maximum. 

Q6. What is the main purpose of the sub-array readout mode?

A sub-array readout mode has been used extensively to reduce exposure time and pileup, but which also reduces observing efficiency, as explained above. 

Q7. What is the spectral resolution of the ACIS/CXO?

Although clusters of galaxies have been studied extensively by previous X-ray observatories, the high angular resolution with moderate spectral resolution of ACIS/CXO permits the examination of the cores of the clusters in sufficient detail to place limits on the interaction cross section of dark matter that is thought to constitute 90% of the cluster mass [30]. 

Q8. Why was there a slow accumulation of condensation on the optical blocking filter?

Because the focal plane is operated at -120C and the optical blocking filter is at about -60C it was anticipated that there might be a slow accumulation of condensation on either or both components, although there is only a very small path from the main cavity of the telescope to the focal plane. 

Q9. What is the purpose of the heater?

A heater is provided to protect the camera and electronics from cooling excessively in eclipse conditions, or to heat the camera and focal plane to bake out any condensations that might build up over time. 

Q10. How do you reduce the exposure time?

Shortened exposure times may be achieved by discarding part of the framestore, i.e., processing only a sub-frame, or by intentionally pre-flushing and discarding the contents of the image area after reading the framestore. 

Q11. What is the dramatic new field of study opened up by the CXO?

Perhaps one of the most dramatic new fields of study opened up by the CXO is the study of the X-ray source populations in external galaxies. 

Q12. What is the temperature of the plasma surrounding SgrA?

The entire 17 x 17 arc minute field surrounding SgrA* is covered with emission from a plasma with a temperature of nearly 100 MK. 

Q13. How long does it take to create a new bias map?

a new bias must be created for each CCD after a change of observing mode, perigee passage and radiation induced shutdown. 

Q14. What is the way to study the X-ray sources in external galaxies?

The high angular resolution of the X-ray mirrors combined with the small pixel size of ACIS and the moderate energy resolution of the CCDs creates a powerful tool to reveal the distribution and types of X-ray sources found in external galaxies. 

Q15. How long does it take to read the framestore?

This transfer takes 41 ms, after which the framestore is read out to determine where each X-ray has interacted in the CCD, while the next exposure is being recorded on the imaging portion of the CCD.