We compare the AXAF CCD Imaging Spectrometer (ACIS) to ASCA's Solid-state Imaging Spectrometer. Both instruments use charge-coupled devices developed at MIT Lincoln Laboratories. Progress in detector development, together with experience gained with ASCA have led to a number of improvements in instrument performance. These include:

• Better detection efficiency at very high and very low energies, by means of deeper depletion and back-illumination, respectively;
• Better spectral resolution by means of lower noise detectors;
• Significantly greater tolerance for charged-particle radiation, by means of greater shielding, detector improvements, lower detector operating temperature, and pixel-by-pixel bias correction.

The ACIS team has conducted an extensive calibration program (see, for example, the poster by Pivovaroff et al). To assess the impact of residual calibration errors on ACIS measurements, we have simulated several representative observations. We report the expected magnitude calibration-related errors in source model parameters.

Figure 1
Figure 1: The ACIS flight focal plane, which consists of 8 front-illuminated and 2 back-illuminated MIT Lincoln Laboratory CCID17 detectors. Each detector contains 1024 x 1026 pixels, and each pixel measures 24 x 24 m. Thus each detector measures about 1 inch square. The four imaging detectors, arranged in a square, are tilted to approximate the AXAF High-Resolution Mirror Assembly's focal surface. The field of view of the imaging detectors is 17 arcmin on each side. The six detectors in the linear array conform to the Rowland circle of the AXAF High-Energy Transmission Gratings, and can be used for grating spectroscopy or imaging. The back-illuminated detectors are the second and fourth from the left in the spectroscopy array.

 
Figure 2: An ASCA SIS focal plane detector. Each of the two SIS sensors contains a focal plane like the one shown here, comprised of 4 front-illuminated MIT Lincoln Laboratory CCID7 detectors. Each detector contains 420 x 422 pixels, and each pixel measures 27 x 27 m. Thus each CCD detector measures about 1/2 inch square. The field of view at the ASCA focal plane is about 22 arcmin square.

 
Figure 3: Comparsion of ACIS and ASCA SIS Detector Quantum Efficiencies. All curves are derived from ground calibration data. The ASCA data were obtained by Gendreau (MIT PhD Thesis, 1995). ACIS detectors shown in red (Back-illuminated detector; BI) and green (Front-illuminated detector; FI) and the ASCA SIS detector in blue. The ACIS curves are for ASCA grades 0,2,3,4 and 6, with split-event threshold (spt) of 15 e; the ASCA SIS curve is for grades 0,2,3 and 4, for spt = 40 e. Both curves represent pre-launch performance.

Why do ACIS detectors have higher quantum efficiency?

• The ACIS BI device has much greater efficiency at low energy because its dead layer is very thin.
• The efficiency improvement at high energies is attributable to the larger depletion depth of the ACIS devices: 65-75 m ACIS FI; 40-45 m ACIS BI; 30-35 m for ASCA SIS.
• The ACIS FI device has somewhat higher efficiency just below the oxygen edge, owing to thinner silicon dioxide in the dead layer ( m for ACIS FI; m for the ASCA SIS.
• ACIS Grade 6 events can be accepted with negligible penalty in spectral resolution because of the lower readout noise. This further improves ACIS detection efficiency.

 
Figure 4: Comparsion of spectral resolution of ACIS front-illuminated (FI) and ASCA SIS detectors. Both curves are derived from ground calibration data. The ASCA data were obtained by Gendreau (MIT PhD Thesis, 1995). ACIS FI detectors shown green; ASCA SIS detector in blue. The ACIS curves are for ASCA grades 0,2,3,4 and 6, with split-event threshold (spt) of 15 e; the ASCA SIS curve is for grades 0,2,3 and 4, for spt = 40 e. Both curves represent pre-launch performance. Above 2 keV the FWHM curves of ACIS and ASCA converge.

Why do ACIS Front-illuminated detectors have better spectral resolution?

• The ACIS readout noise is only half that of the ASCA SIS (2-3 e- RMS for ACIS; 4-6 e- for ASCA SIS).
• The lower ACIS readout noise permits a lower split event threshold: ACIS uses a split threshold of 15 e; ASCA uses 40 e. Thus, ACIS measures smaller quantities of ``split event''charge in the vicinity of events. This improves spectral resolution.

NB: The spectral resolution of ACIS BI devices is typically FWHM 90 - 120 eV for E < 1.5 keV.

Improvements in ACIS Detector Radiation Tolerance

As we expected before the launch of ASCA, the SIS detector performance has been degraded by high-energy charged-particle radiation. A variety of measures have been taken to improve ACIS radiation tolerance. These include:

Lower detector operating temperature
ACIS detectors will operate at -120C, some 60C colder than ASCA SIS detectors. For a given level of radiation exposure, this temperature reduction reduces radiation-induced charge transfer inefficiency (CTI) by about one order of magnitude, and reduces radiation-induced dark current by a factor greater than 200.

Faster detector readout
Although an ACIS detector contains six times as many pixels as an ASCA SIS CCD, the ACIS detector can be read out in only 3.3 seconds. This is less than the readout time required in the SIS's 1-CCD mode. For the same reasons that ASCA's 1-CCD mode performs better than 4-CCD mode, the faster ACIS readout time improves radiation tolerance.

Pixel-by-pixel bias correction
ACIS maintains a bias estimate for each active pixel in the focal plane. This estimate can be revised at will. The ASCA SIS has sufficient memory for only a coarse bias map ( 1 bias value for every 40 pixels). ASCA experience has shown that radiation-induced dark current varies significantly from pixel to pixel. Uncorrected, this dark current degrades ASCA's spectral resolution. We believe the ACIS bias map will substantially reduce the effects of radiation-induced dark current on ACIS spectral resolution.

Bad-pixel rejection
The ACIS flight software maintains a programmable list of up to 10,000 ``bad'' or defective pixels. Events originating in pixels designated as ``bad''can be rejected onboard, so that they do not consume scarce telemetry bandwidth. The ASCA SIS has insufficient memory to provide this function.

Improved detector hardening
The charge transfer channels in ACIS CCD's are only 2 m wide; those in ASCA devices are 3 m wide. The narrower channels reduce the chance that radiation-induced traps will impede the charge transfer process.

Better shielding
The mean thickness of shield surrounding the ACIS detectors is roughly equivalent to 10 gm cm of aluminum. This is about twice the mean shielding carried by ASCA.

Onboard calibration source
ACIS's onboard calibration source will allow us to monitor gain, charge transfer efficiency and spectral resolution at discrete energies from 0.7 to 5.9 keV. This will allow us to calibrate the effects of radiation damage on ACIS detectors more directly.

The bottom line: After 5 years on orbit, (given reasonable assumptions about solar activity), we expect ACIS detector performance to be better than ASCA SIS performance was 1 year after ASCA launch.

Summary of ACIS Calibration Error Analysis

Approach:

• We have analyzed the effects of three types of calibration error in ACIS front-illuminated CCDs on observational parameters derived from model fitting in XSPEC.

• The three types of calibration error studied were depletion depth, gain, and gate structure thickness.

• The effects of these errors on the best-fit parameters for some typical astrophysical spectra were determined using thermal bremsstrahlung and power-law models.

Results:

• Depletion depth errors of caused the best-fit column density to change by in the power-law model with low column density. All other parameters in the thermal and power-law models were insensitive to errors of this magnitude in the depletion depth.

• Gain errors of produced peak-to-peak changes of 7--11% in the best-fit temperature for the thermal plasma models, and had little effect on the photon index in the power law models. The gain errors caused a change of in the best-fit column density in the power-law model with low column density, and smaller changes () in the thermal models.

• The observational model parameters studied were insensitive to errors of in the thickness of the CCD gate structure (Si and ).

Conclusions:

• Accurate parameter fitting of typical astrophysical spectra at the 5--10% level can be achieved with the ACIS front-illuminated CCDs with calibration errors of in the depletion depth and gate structure, and errors in the gain.

• Sub-assembly calibrations performed at the MIT Center for Space Research and flight assembly calibrations performed at XRCF indicate that the calibration accuracy of the ACIS front-illuminated CCDs will meet or exceed these levels.

 
Table:

Thermal Plasma (kT = 2 keV, , abund. = 0.1, Z = 0).

 
Table:

Thermal Plasma (kT = 2 keV, , abund. = 1.0, Z = 0).

 
Table:

Thermal Plasma (kT = 7 keV, , abund. = 0.3, Z = 0).

 
Table:

Thermal Plasma (kT = 7 keV, , abund. = 0.3, Z = 0.5).

 
Table:

Absorbed Power Law (, ).

 
Table:

Absorbed Power Law (, ).