Every once in a while, when the Netscape home page pops up in a new window, for those of us who never get round to completing the arduous 5 second task of setting a more sensible local home page, and after the mouse, operating by mysterious forces and surely of its own volition, clicks not on the invitingly dull grey ``Chandra Science'' button on the title bar but on poll: which famous celebrity is your dream date?, one's eye can be caught by intriguing minor headlines--intriguing on the scale of being slightly more interesting than washing the kitchen floor, but somewhat less so than poll: which famous celebrity is your dream date?--that promise fascinating new insights into the human condition. One such study involved hapless participants playing a computer game which psychologists had cunningly programmed to temporarily freeze at certain moments. (No, it was not at one of the CIAO workshops.)
The test is to see whether participants applied only the single mouse click they knew, logically, was necessary, then waited patiently for results, or whether mouse clicks were repeatedly applied in an irrational frenzy of childish impatience. Uncharitably, my wife has no hesitation in pre-assigning me to the latter group, claiming to have watched me perform the test in numerous data analysis scenarios (at one of the CIAO workshops.)
The truth, however, is that my behaviour has been modified from its usual cool rationality by a paranormal psi talent, a power that lowers a veil of negative influence on the goings on of any machinery in my immediate vicinity. Bugs appear in software that had worked perfectly in the hands of others, programs freeze, random numbers aren't, 99% confidence becomes somewhat iffy. It goes without saying that the perfect code I write myself seems to get perturbed into horrible states of error when run. Even, inexplicably, when run by others. So, when called upon every year or so, it is with some considerable relief and gratitude that I cut-and-paste into the Chandra Newsletter draft the usual ``The last twelve months have seen flawless operation of the LETG, and at the time of writing, the LETGS with both HRC-S and ACIS-S detectors is operating nominally, blah blah blah.'' Fortunately, mine is not a long-range force and the data the instrument is returning, happily insulated by thousands of miles of near-vacuum in its high elliptical orbit, are jolly impressive.
Well, most of it is anyway. One much anticipated Chandra event was the first acquisition of a high resolution spectrum of an isolated neutron star. When still X-ray bright, such objects are expected to have temperatures in the range of a few hundred thousand to a million degrees K, and so their spectra are expected to peak nicely in the LETGS bandpass. In 1997, Frits Paerels at Columbia published an important paper pointing out that the gravity-sensitive Stark broadening of X-ray lines of ionised metals in a neutron star atmosphere could be used, together with their observed gravitational redshift, to provide a simultaneous measurement of mass and radius--quantities that could provide useful constraints for the equation of state of ultra-dense matter. A 55 ks LETG+HRC-S observation of one of the more promising candidates in the isolated neutron star pageant, RX J1856.5-3754, was finally obtained in 2000 March. The spectrum was nicely analysed by Vadim Burwitz at MPE and collaborators. Fascinatingly, it exhibited................. absolutely nothing and was perfectly consistent with a blackbody spectrum.
Last year's TAC subsequently declined a deluge of three proposals requesting a significantly longer exposure to draw out those diminutive spectral features lurking bashfully in the grassy noise of the 55ks exposure. It was then left to Our Director to wield the noise-cutting scythe in the form of 450 ks of discretionary time. The data were obtained in 2001 October, were carefully shepherded through processing and, after a brief but intense V&V, the co-added spectrum was finally ready for public release. Fascinatingly, it exhibited.............. absolutely nothing and was perfectly consistent with a blackbody spectrum (Figure 10).
Apparently bereft of any spectroscopic foothold on its glassy, precipitous slopes, if we really knew that RX J1856.5-3754 was a pure blackbody spectrum in the X-ray range it would make an excellent, though rather faint, calibration target, peaking as it does in the 40-50 Å range and with no significant higher order contamination shortward of 80 Å or so. The 25-70 Å range is especially difficult to calibrate in-flight because of a lack of standard candle point source targets. White dwarfs are too cool and show no appreciable flux here, and one is left hoping for featureless AGN continua that, with luck, are still described by simple power laws in this very soft range. Even if luck holds, such spectra inevitably are complicated by higher spectral orders that cannot be separated in the LETG+HRC-S combination.
As it is, based on the blackbody hypothesis, RX J1856.5-3754 does illustrate nicely the current state of the LETG+HRC-S effective area calibration. The effective area of this combination depends on the effective area of the HRMA, the diffraction efficiency of the LETG, the transmission of the HRC-S UV/Ion shield, and on the microchannel plate quantum efficiency. As outlined in an earlier Newsletter article, this last quantity completely dominates the calibration uncertainties in 1st order spectra. The comparison of the blackbody fit to the observed data in Figure 10 shows areas of some systematic deviation beyond the counting statistics amounting to about 10% or so, as well as areas of smaller discrepancy. In this range, both the HRMA area and LETG efficiencies are relatively flat with wavelength. We note, in particular, the apparent edge at 60 Å. This feature occurs at the HRC-S plate gaps and at, in this plot of summed + and - orders, the transition from seeing only the +1 order in the approximate range of the -1 order gap (-58 Å) to seeing only the -1 order in the approximate range of the +1 order gap (-66 Å). This effect results from a residual discrepancy in the +1 vs -1 quantum efficiency calibration.
More globally, absolute calibration uncertainties are determined by the normalizations and power law slopes adopted for AGN calibrators--with most weight placed on our primary calibrator PKS2155-304--and at longer wavelengths on knowledge of the absolute flux of our primary white dwarf calibrators Sirius B and HZ 43. At shorter wavelengths ( 20 Å) cross-calibration with HETGS suggests the LETG+HRC-S effective area is good to about 10 % or so, absolute. Toward longer wavelengths, and in higher spectral orders, uncertainties are likely to be larger--we estimate about 15 % or so for the former. Higher order calibration uncertainties result from both the HRC-S QE and the LETG diffraction efficiencies which have proved both difficult to model and to calibrate empirically. Uncertainties range from about 20% in third order to factors of two or more in still higher orders.
The LETG+ACIS-S calibration is currently more certain than LETG+HRC-S by virtue of the QE calibration of ACIS-S, and again, for 1st order, is presently at a level of about 10 % in absolute terms. Progress in improving this calibration will be presented in a future Newsletter.
A major milestone in the LETG+HRC-S calibration was recently achieved with the inclusion of a spatially non-uniform HRC-S QE map. This map was based on data obtained by the HRC team in flat field tests immediately prior to the instrument delivery, and now enables a fairly accurate HRC-S effective area, with or without the LETG, to be constructed for general off-axis angles using the CIAO tools fullgarf (previously mkgarf) and mkarf.
The LETG group has also recently made available response matrices that we have constructed for LETG+HRC-S combined orders 1-6. We emphasise response matrices here because these are crucially different from the RMF/ARF approach adopted by CIAO, in which effective area and spectral response functions are separated into two separate entities. As will be well-known to the majority of readers, a response matrix is comprised of an instrument spectral response function normalised to its effective area. The difference is important because the CIAO fitting engine Sherpa cannot at this time model the overlapping spectral orders in the LETG+HRC-S response that requires combining multiple RMF's/ARF's, whereas utilising a response matrix to describe multiple orders is straightforward (as it is in XSPEC). The drawback to the response matrix approach is the number of elements required in the matrix to describe accurately the response in the higher spectral orders that have higher resolving powers. The current matrices include orders 1-6. The importance of orders above 6 will depend on the source spectrum, but for most sources their omission should be of little consequence for wavelengths 50 Å. At longer wavelengths, omission of higher orders could be a bit more important; if in doubt, the best approach is to compute a trial model for increasingly higher orders until the incremental change by adding further orders is small compared to the other uncertainties. Such an operation can be done within CIAO, for example by using fakeit for each spectral order in turn, then summing the resulting spectral orders using SLANG. An analysis thread describing more precisely the exact steps is in preparation. The effective areas for individual spectral orders are of course available to perform the same type of computation in custom-written software.
In contrast to RX J1856.5-3754, and back to the jolly impressive, is the LETG+HRC-S spectrum of the dwarf nova WZ Sge that undergoes an outburst every 30 years or so. Caught in its latest at the Director's Discretion following multiple TOO requests (PI P. Wheatley, Leicester University, pressed Send first), its spectrum (Figure 11) is replete with a feast of crinkly bits Slartibartfast could have won an award for had he turned to astrophysics, and also bears more than a passing resemblance to a Stegosaurus' rib cage. When time filtered and split into different orbital phases, the ribs also wiggle about. While the identifications of some of the more prominent ribs is straightforward by comparison with appropriate spectral line lists, there is still much in there to discover for the spectroscopist/paleontologist, let alone those studying the nova outburst itself.
To what accuracy could you measure the wiggling of the ribs? The answer depends on the accuracy and stability of the dispersion relation. While now perhaps the most difficult part of the LETG+HRC-S combination to understand and calibrate, the dispersion relation is already pinned down to fairly good accuracy for the central HRC-S plate. Geometrically, it is simple enough. However, it was clear soon after launch that there were complications when one or two narrow lines in the spectrum of Capella did not quite land exactly where they should. Differences were small ( Å or less, compared to an instrument resolution of Å), but measurable. Further study using different targets that probe different wavelength regions has indicated that the problem lies in detector spatial non-linearities localised to certain regions with a millimetre scale size. One particular ``glitch'' occurs in the negative order between about 17 and 22 Å. Work is ongoing to map out these spatial non-linearities and to determine their exact cause and how to correct for them. Outside of these problem areas, the dispersion relation for the HRC-S centre plate is good to an accuracy of Å. The outer plates also appear to show signs of spatial non-linearity and exhibit larger departures from the dispersion relation of the central plate. In particular, a primary departure appears to be that their underlying dispersion relations are slightly different from that of the central plate, with deviations reaching up to 0.05 Å. Again, we are working to map out these effects, though analysis of the dispersion relation in the 70-170 Å range is hampered by a lower instrument effective area and consequent paucity of very bright spectral lines.
There is evidence that the LETG+ACIS-S dispersion relation could be systematically different from that of LETG+HRC-S, based on the current sets of parameters that describe the two instruments. This is impossible physically, since both instruments share the same Rowland geometry; the discrepancy lies elsewhere and work is ongoing to pin down the cause (likely due to a residual small error in the pixel sizes of one, or both, of the instruments, or in the ACIS-S plate gaps). The difference in the implied dispersion relations amounts to a maximum of about 0.04% .
Regarding stability, we have never seen any significant shifts of spectral lines from one observation to the next of the primary wavelength calibration source Capella with either HRC-S or ACIS-S, confirming that the instruments are very stable.
LETG operations were once again back to normal at the end of July 2001 when the spacecraft systems group succeeded in implementing a software work-around that enabled a return to commanded insertion and retraction. The original problem was due to a HETG (rather than LETG) limit switch failure that onboard software safety procedures used to prevent an undesirable clanging together of LETG and HETG support structures should either be accidentally activated at the wrong time. The software patch replaced the temporary procedure of inserting the LETG ``by hand'' through commands that nudged it gradually into place.
The entire set of LETG+HRC-S data has recently been reprocessed in order to correct for two problems found with earlier automatic processing. The first problem pertained to filtering on status bit 21, which we found to be appropriate for HRC-I but which was found to remove a significant number of valid events in a spatially non-uniform way for HRC-S. This filter has now been switched off for the S detector. The second regards the area of the detector passed on to level 1.5 processing, which turned out to be slightly too small to accommodate all of the region required for estimating the background spectrum. The AP problems were corrected in the DS6.1.0 software release.
Observer and proposer information and news on the performance of the Chandra LETGS can be found on the instruments and calibration page:
- Jeremy Drake
Paerels, F., 1997, ApJ, 476, L47