Measurements of the Chandra
ACIS Contaminant using the LETG
This page documents the analysis Ive done using the Chandra Low
Energy Transmission Grating Spectrometer (LETGS) with the ACIS detector.
This combination has been extremely useful in diagnosing the nature of
the contaminant that is building up on ACIS. See the QE degradation web page for public
information on the subject. A
paper will appear in the proceedings of the SPIE, volume 5165.
The presentation is available in either
2. Time Dependence
The ACIS contaminant imparts at least three distinct absorption edges
on the LETGS data due to C-K, O-K, and F-K. The C-K edge is the deepest
and is shown to increase at a rate of about 0.4 optical depths per year.
The other components are readily detected in a recent LETGS observation
of Mk 421 (from F. Nicastro) and marginally detected in a few previous
LETGS calibration observations. The time dependence of the C-K edge is
modeled and the other edges are assumed to have the same scaling with
time, resulting in a time-dependent spectral model of the contaminant
that can be used to correct Chandra grating data for the effect of the
contaminant. This model has been implemented using ISIS for users in a standalone procedure.
There has been a series of CXC internal meetings to analyze the
problem and consider a fix. There were many reports and presentations
at these meetings, so the objective of of this web page is to provide
public access to some of these reports in a coherent way to help users.
While this work concentrates on LETG/ACIS data and the conclusions are
likely to apply to all ACIS observations, I emphasize that there remain
some unresolved issues. The reports are presented in chronological
order within several categories due to the development of our
understanding of the contamination.
At first, only the C-K edge was clearly detected in the LETG/ACIS data
so the first correction file only
included a model of the C-K edge. The O-K edge and an F-K edge were
found in the very long observation of Mk 421 in October, 2002
(from F. Nicastro). These
features are rather weak but Mk 421 was rather bright at the time and it
was observed for 100 ks. A model of the absorption due to the
contaminant was developed in the following procedure.
- The June 2002 LETG/ACIS spectrum of PKS 2155-304 was
extracted using the methods outlined in a previous memo on preliminary LETGS
calibration. This memo describes the correction I employ to account for a
discrepancy between QEs of the BI and FI chips. I have been applying
the correction to FI chips but it could easily be reversed and applied
only to BI chips. The results would be the same except for a
- The PKS 2155 spectrum was fit to a double power law
(2PL) model that smoothly joins a flat PL at low E to a steeper one at
high E, as described the the LETGS cal memo. A pileup correction was applied to the
response functions to account for event loss near the Ir-M edge.
- A model of the C-K absorption was
devised to fit the overall shape of the C-K absorption after heavily
binning the residuals from the 2PL fit.
- Using the October 2002 observation of Mk 421, the O-K
and F-K edges in the contaminant were modeled.
The resulting model of the contaminant is available in a file. There are four columns,
corresponding to wavelength (Å) and the remaining three correspond
to the optical depths due to carbon, oxygen, and fluorine. The fits
gave (Henke equivalent) edge optical depths of 2.09 +/- 0.02, 0.07 +/-
0.03, 0.100 +/- 0.007, and 0.066 +/- 0.005 for the C-K, N-K, O-K and F-K
I obtain the following column densities for C, O, and F: 2.0e+18,
1.75e+17, and 1.45e+17 atoms per sq cm. For N-K, I had a previous upper
limit of the edge optical depth of about 5%, giving a column density of
7.1e+16 atoms per sq cm. Carbon is more abundant than the other
elements by factors of 11.5, 14, and >30 for F, O, and N,
respectively. The uncertainties on these ratios are about +/- 1.
Whatever the compound is, the C atoms are mostly bonded to other C
Back to Summary
2. Time dependence
When it was first reported (originally in July
2000) in the LETGS data, the C-K edge did not show clear variation with
time. Time dependence became clear with observations in 2001 and
Using the contamination model derived above, I fit the LETG/ACIS data
for many sources with smooth spectra, allowing the C-K, O-K, and F-K
edges to vary independently. The results are given in a text file and given in another web page.
Then the C-K edge depth was fit to a simple time dependence model. The
shape of the time dependence was assumed to be the same for each of the
three detected edges and were merely scaled to observed optical depths
(see the previous section) for the date of the Mk 421 observation
(2002.86). Because the optical depth values used to create the
contamination model were slightly different 2.10, 0.09, and 0.06
(for C-K, O-K, and F-K, respectively) the time factors for each
component were merely rescaled. The resulting time dependence file is available
The O-K edge data do not fit the model as
well as expected for reasons that are not completely understood.
Back to Summary
3. Spatial nonuniformity
Two observations of 3C 273 gave us the opportunity to see if the
contaminant is uniform over the surface of ACIS. The analysis shows that there is a
detectable variation in the sense that there is more contaminant at the
edge of the detector than in the center. These results are consistent
with preliminary results from analysis of data from the ACIS external
cal source and with the hypothesis that the thickness depends on the
local filter temperature, which is expected to be warmer in the center
than at the edge.
Back to Summary
Last updated August 12, 2003