The fact that the events related below actually, more or less, happened, is in some ways remarkable--not at all because of the events themselves or anything to do with this author I hasten to add, but because they occurred as a result of the advance in X-ray hardware technology achieved by the designers and builders of the AXAF mirrors and instruments and the maturing of the field of observational astrophysical X-ray spectroscopy that this technological advance represents.
The High Energy Suppression Filter was originally conceived in order to try and address one particular side effect of the high performance of the Low Energy Transmission Grating (LETG) when married with an efficient detector over a broad bandpass, but with limited intrinsic energy resolution. The LETG in combination with the AXAF HRMA and HRC-S detector has a very good response for energies in the range ~0.1 -7 keV--a magnificent achievement in astrophysical instrumentation. This tremendous wavelength coverage though does pose a small problem.
The relatively high efficiency towards high energies means that there is also significant throughput from the LETG in the higher spectral orders. The higher orders overlap with the first order spectrum on the detector causing order confusion. Spectral simulations of late-type stellar coronal sources indicate that the higher order throughput represents a considerable fraction of the total spectrum flux, and at some energies can dominate the first order spectrum. Particularly severe is the contamination caused by the 3rd and 5th orders of the Fe L shell complex. Because the HRC-S microchannel plate detector has very little intrinsic energy resolution, there is no telemetered information to distinguish between first order and higher order photons. This problem has often been referred to as ``order confusion'' and the ``overlapping orders'' problem.
The overlapping orders problem can be thought of as one of signal-to-noise ratio. If one is interested in measuring the flux in a first order line, the higher order flux is an additional background that must be subtracted. As a background problem it is not very pleasant, this pseudo-background being highly inhomogeneous in the case of a line-dominated source and being always commensurate in brightness with the first order spectrum one might be primarily interested in. It should be noted, however, that while analyses aimed at extracting the fluxes of individual lines suffer from this background, methods of spectral analysis that fit the whole spectrum simultaneously using a model of the plasma emission do not necessarily suffer such problems. Since the spectrograph model response will include the efficiencies for higher order transmission (though certainly known at lesser accuracy than the first order response), the model folded through this response is simultaneously fitted to all orders at once.
Why, then, embark on the folly of attempting to filter out these potentially useful photons if one can fit the whole spectrum at once? This issue is still a trifle touchy and perhaps scientifically controversial. Having once attempted to be unbiased (Drake 1996), with a high degree of failure, I will spare the spilling of my, perhaps only slightly polarised, gore on the subject lest the rage thus engendered in half of the readership of this Newsletter induces them to tear it to shreds and destroy the other articles which are in contrast interesting and informative. Under the Sword of Damocles of red editorial pen (but no accompanying pleasurable feast), I would simply add that for global spectral fitting to work successfully for high resolution spectra, the spectral model must include all spectral lines present in the source that give rise to observed counts (whether individually or as an ensemble of otherwise unresolved or undetected individual features) and with wavelengths correct to within a fraction of a resolution element. (Concealing an additional remark in parentheses: except for some relaxation in the requirements of theoretical line wavelengths, these criteria of course remain true in the case of the fitting of low resolution spectra.) It should also be born in mind that in some wavelength ranges AXAF spectra will surpass in quality even solar X-ray spectra; in these ranges, especially in the largely unexplored 30-70 Å range, spectral models remain essentially untested by observation.
Should one sink to the view that filtering higher order photons might sound like a not entirely unreasonable idea, the entirely unreasonable restrictions of having to work within the laws of physics provides further weight to the soul. There is no known design of transmission filter which can attenuate the broad range of higher energies which cause the LETG overlapping orders problem, yet which is transparent to soft X-rays. However, such filtering properties can be found in reflective optics.
The initial work on the reflection filter began around December-January of 1995-1996, during the months of arctic conditions in Cambridge while snowed-in for weeks of unhealthy, forced, contemplative thought. In some of the particularly bleak moments, a very rough design for a relatively simple flat mirror which could be used as a high energy filter was conceived. The flat mirror would be a few cm in height, and would be supported by fixtures attached to the molybdenum frame that sits on top of the HRC-S assembly. It would lie directly above the edge of the HRC-S detectors, so as not to impede the current light path to any large extent.
With the coming of some phantom signs of spring, this rough design was rather optimistically advertised as a ``simple bolt-on fix for the overlapping orders problem''. The first rough sketch of the filter was grafted onto an existing HRC-S figure and is illustrated for entertainment purposes in Figure 10. In this sketch, the author had thought it might be somewhat amusing to add the tongue-in-cheek labeling; it is this label that subsequently hatched the ``Drake Flat'' nickname.
Figure 10: The first ``rough sketch" of the proposed location with respect to the
HRC-S detector assembly of a flat mirror which might, at a pinch,
be useful for reducing higher order flux from the LETG.
This first filter ``design'' (using the word somewhat loosely) had a vanadium coating. The incoming spectrum would reflect off this coating at grazing angles of a few degrees (~5 or so), at which angle V absorbs X-ray photons at energies higher than the V L-edge. The residual high energy flux reflecting off the filter would be scrubbed down to a minimum by surface roughness. Furthermore, the range in cone angles ( ± 3.5 ° ) of the AXAF telescope would be dealt with by added surface roughness which varied over the mirror according to the particular graze angle at each location.
Though quite reasonable-sounding to a stellar astronomer (whose instrument building talents peaked when he was still working with Lego), even ingenious after a few stiff drinks, such concepts are apparently quite hilarious to those with real talents for instrument building. Purely hypothetically, one can conceive of a grossly embarrassing situation of an ignorant stellar astronomer blundering onwards and even consulting the soft X-ray astronomical optics guru, Dr. Eberhard Spiller, on the matter. Dr. Spiller might, in such a circumstance, had it ever occurred, have suggested, probably through clenched teeth (though it is difficult to tell on the telephone, in this entirely hypothetical situation of course) that the variable roughness approach was ``theoretically valid'', and pointed in the direction of the MIT ion-beam etching laboratory as a means for roughening things up a bit. Falsely buoyed by this encouraging euphemism, a true naïve imbecile might, had such a hypothetical situation arisen, even have contacted the MIT ion beam laboratory, world leaders in the field and working on multi-million dollar contracts with multi-national silicon chip manufacturers and other cutting-etch, if you forgive the pun, technologies, asking whether or not they might be able to roughen up his (or her) mirror for him (or her) a bit please. What would they have replied? I imagine, hypothetically, something like an aloof and humiliating ``...making a polished mirror rough does not sound like a project that would interest our laboratory". How fortunate that feckless endeavour does not result in such humiliation in real life.
The variable roughness design, though useful for providing good joke material for instrument builders, was not a very practical solution. One alternative that was examined was to use a reflection grating as a cross-dispersion element. However, such an optic would have been too inefficient when combined with the rest of the instrument response. The only other solution appeared to be to make the filter with multiple facets, each designed to cope with a particular range of the incoming light cone. With the sensible help of Paul Gorenstein and the very tolerant and sage advice of Steve Murray and Martin Zombeck (the HRC Principal Investigator and Project Scientist, respectively), and of Bert Brinkman (the LETG Principal Investigator) the final two-facet design was produced. Paul Gorenstein's invaluable expertise in X-ray optics also fashioned my ill-conceived V coating into a practical design consisting of two separate coatings of C and Cr, each one to handle a different wavelength range. The cut-off energies of each of these coatings are again dictated mainly by the change in reflectivity over the ionization thresholds of their L (in the case of Cr) and K (in the case of C) edges. The C K edge at 284 eV (~44 Å) provides a sharper cut-off than the L-edge of the Fe-group elements, while providing a higher reflectivity at lower energies. However, the C coating does not provide any throughput for the wavelength range shortward of the C edge, and so the Cr coating served to extend the filter to higher energies. Shortward of the Cr filter cutoff, corresponding roughly to the wavelength range 0-37 Å, the central portion of the filter was removed in order to allow the higher energy spectrum to be detected with no reflection.
Figure 11: The two-faceted reflector as seen face-on from the
cross-dispersion direction. The upper and lower facets (``F1'' and
``F2'') are shown, together with the extents of the C and Cr coatings
in physical and first order wavelength spaces.
(Plot is from the SAO HRC team.)
The final mirror design is described in more detail in the Observatory Guide. The design was dictated largely by the geometry of the problem of intercepting the light cone while fitting into the existing HRC instrument without obstructing fiducial light paths or compromising the conventional mode of operation of the HRC-S with the LETG. The angles of the mirror facets to the incoming light cone were also very tightly constrained by the width of the active area of the HRC-S detector, because this limits the distance in the cross-dispersion (z) direction between the straight through (unreflected) spectrum and the deflected spectrum bouncing off the mirror. Consequently, the two-faceted mirror could not easily intercept the whole 7° cone, and so a design which intercepts 5° , and leaves the remaining 2° of cone to fall onto the detector unreflected and unvignetted, was arrived at. Furthermore, the tolerances in the instrument alignments and in the pointing accuracy of the spacecraft, together with the practical engineering problems of manufacturing the mirror out of a light, rigid and polishable material in a very short space of time, added additional stringent constraints.
A schematic face-on view of the filter showing the respective wavelength ranges of operation of the different coatings is illustrated in Figure 11. The side view of the reflection geometry, in which the dispersion direction is into the plane of the page, is illustrated in Figure 12. From this latter figure, it can be seen that in the range of operation of the filter, a typical pointing would produce three spectral traces on the detector, corresponding to the two reflected spectra, and the ``straight through'' spectrum corresponding to the part of the cone not intercepted by the flat. In the central gap, of course, the three traces reduce to one. Side effects of the reflections are that the two reflected traces come to two different foci slightly above the detector at its nominal focus. The detector can be re-focussed, though not perfectly for both reflected traces simultaneously. Consequently, use of the filter will result in a small loss in spectral resolution.
Figure 12: A sketch of the reflection geometry for the two-faceted
reflector. This figure essentially represents a cross-cut through the
HRC-S dispersion direction (the dispersion or y axis is
perpendicular to
the plane of the page). Each facet is designed to intercept 
of the
incoming light cone, with the remaining 
of cone falling
directly onto the detector.
(Images are from the MPE LETG team.)
Once the final mirror concept was arrived at, the HRC engineering team lead by Randy Moore carried out the impressive feat of producing a highly sophisticated practical engineering design and procuring the flight optics, machined from radioactive isotope-free beryllium, on a timescale of only a few weeks. Paul Gorenstein organised the C and Cr coatings, and, with the help of Ahsen Hussain and Eric Gullikson (the latter of the Lawrence Berkeley Laboratory), analysed data taken at the LBL synchrotron facility to verify the reflectivity of the witness samples. Regarding these measurements, Paul Gorenstein reported ``For both coating materials the reflectivity is below the theoretical reflectivity of the bulk material with handbook values of density. Very likely the density of the coatings is lower. However, the reflectivity deficit is acceptable in the wavelength ranges where the filters are actually used." Consequently, the filter was placed into the HRC instrument by Steve Murray and Jon Chappell in what was described as ``the most nerve-wracking part of the assembly of the HRC instrument"--one slip could have torn flight UV/ion shields or damaged the microchannel plates themselves.
At the time of writing this article, the massive calibration effort undertaken at the X-ray Calibration Facility of the Marshall Space Flight Center had just been completed. The reflection filter did see its first X-rays during the testing: to this author, at least, it came as a pleasant surprise that low energy photons were reflected and high energy photons were attenuated. However, the data analysed to-date are suggestive of possible alignment problems which, though very small, might compromise the flight operation of the filter. The overall reflectivity of the flight optic remains to be calibrated and it is not clear whether the data gathered at the Calibration Facility can provide this to sufficient accuracy at the low energies the filter is designed to operate at. Hence, it remains to be proven whether or not the overall efficiency of the system is of practical value for astrophysical sources, and moreover, whether the resulting spectra (all three traces!) can be analysed in a scientifically useful fashion.
If only such feckless endeavour did not result in such humiliation in real life.
Drake, J. Unbiased, 1996, in Cool Stars, Stellar Systems and the Sun, 9th Cambridge Workshop, eds. R. Pallavicini and A. Dupree, PASP Conf.Ser., San Francisco, p 203.
Jeremy Drake