Large—area reflection grating spectrometer for theConstellation—X mission

S. M. Kahn, F. B. Paerels, J. R. Peterson, A. P. Rasmussena,M. L. Schattenburg, G. R. Ricker, Jr., M. W. Bautz, J. P. Doty, G. Y. Prigozhinb,

J. A. Nousek, D. N. Burrows, J. E. Hilic and W. C. Cashd

a Columbia University, 550 West 120th Street, New York, NY, 10027b Center for Space Research, Massachusetts Institute of Technology,

77 Massachusetts Avenue, Cambridge, MA, 02139C The Pennsylvania State University, 525 Davey Lab, University Park, PA, 16802

d Center for Astrophysics and Space Astronomy,University of Colorado, Boulder, CO, 80309-0389

ABSTRACTThe optical chain of the spectroscopic X—Ray telescopes aboard the Constellation—X spacecrafts employs a reflectivegrating spectrometer to provide high resolution (E/zE '400) spectra for multiple spectra as a slitless spectrometerin the spectral feature rich, soft X—Ray band (0.2-2.3 keV). As a part of the spectroscopic readout array, we providea zero—order camera that images the sky in the soft band inaccessible to the microcalorimeters. Technologicalenhancements required for producing the RGS instruments are described, along with prototype development progress,fabrication and testing results.

Keywords: Constellation—X Mission, Gratings, CCDs, X-Ray Spectrometer

1. INTRODUCTIONThe Constellation—X payloads' are equipped with large—area, reflection grating spectrometers (RGS) as part ofthe optical chains of the spectroscopic X—Ray telescopes (SXT). The RGS grating array intercepts a fraction of thecondensing light behind the SXT and, as a slitless grating spectrometer, provides continuous soft X—Ray spectroscopiccoverage for objects in the telescope's field of view.

The soft X—Ray band (0.2—2.3 keV) is very rich in spectral features that are of key interest in X—Ray astron-omy. These spectral features include K—shell edges and transitions of the cosmologically abundant elements andtheir ionized counterparts, as well as the L—shell features for the heavier elements (i.e., Fe and Ni). Unambiguousidentification and measurement of these features offer invaluable astrophysical data, which promise to address manyastrophysically interesting parameters of the X—Ray emissive plasmas. However, in order to acquire these data, aspectral resolving power of A/L;\ > 300 is required.

Microcalorimeters2'3 offer nondispersive spectroscopy with a spectral resolution kernel zE that is approximatelyconstant, independent of energy. The resolving power E/LE drops with decreasing energy, and even if theoreticalperformance (iE r 2 eV) is attained in these instruments, a resolving power of 300 will not be met for energiesbelow 0.6 keV. A dispersive spectrometer, in contrast, features a spectral resolution kernel zA that is approximatelyconstant, and thus yields a rising resolving power with decreasing energy. In addition to superior spectral perfor-mance, the RGS readout does not suffer from poor efficiency at long wavelengths as the microcalorimeter does, sincethere are no requirements for thick optical blocking filters on the CCDs. The science payoff for grating spectrometersis indeed crucial.

The baseline design of the spectrometer is largely derived from the RGS instrument aboard the X—Ray Multi—Mirror (XMM) spacecraft,4'5 but will require a number of technological enhancements to meet mass and power

Further author information: (Send correspondence to A.P.R.)APR.: E-mail: arasmus©astro.columbia.edu

Part of the SPIE Conference on EUV, X-Ray, and Gamma-Ray Instrumentation for Astronomy X94 Denver, Colorado • July 1999 SPIE Vol. 3765. 0277-786X1991$l 0.00

requirements for the Constellation—X spacecrafts. In section 2 we describe the baseline design for the large area RGSspectrometers. In the subsequent sections, we describe requirements of the grating and charge coupled device (CCD)components and progress toward their respective development.

2. BASELINE DESIGNThe baseline design of the Constellation—X's RGS involves an array of thin reflection gratings oriented behind thetelescope and an associated set of CCD detectors which both sample the spectrum and provide an imaging cameracentered at the zero order focus.

The spectrometer's zero order focus images the sky in the low energy band inaccessible to the microcalorimetersand also anchors the dispersion solution for each source illuminating the spectroscopic readout. Spectral orderseparation is performed using the CCD's moderate spectral resolving power and thus qualitatively resembles theRGS instruments aboard XMM.6'7

Figure 1 . Left: Geometry of the RGS grating array, simplified (with only two gratings) for clarity. The telescope,at top, condenses light from an on—axis source toward point F. The grating array sits directly behind the telescopeand intercepts part of the converging beam. The gratings lie on a toroidal surface and here only the tangential radiusis shown as the Rowland circle. The diffraction angle is the same for all gratings, and the rays meet at a spectroscopicfocus, labelled here as F . Right: The diffraction geometry for the blaze condition, showing the various angles andtheir definitions.

2.1. GeometryThe grating array is arranged in an inverted Rowland circle configuration as pictured in Figure 1. The gratings havea variation in ruling density from one end to the other and are identical. They are positioned along a toroidal surfaceand oriented such that the light has a single angle of incidence on all gratings. This configuration is analogous of theRowland circle design and features first order aberration—free focussing (in the dispersion plane) for points F alongthe circle.

2.2. OptimizationDesign parameters are derived through optimization. The prescription starts with the dispersion equation linkingthe angle of incidence a and the diffraction angle fifor a grating rule spacing d (see Fig. 1), diffraction order m andwavelength A (Eq. 1) and blaze condition relating the graze angle 'y and blaze angle to the blaze diffraction angle9B (Eq. 2):

mAcos$=cosa+—ã--, (1)

a+ö=flB—6E-y. (2)

95

96

Using these equations, the blaze wavelength AB is expressed in terms of y and ö (Eq. 3), and the resolving power ofa grating spectrometer RB at AB may be cast into the form of Equation 4 where O is the angular resolution of thetelescope.

mAB 2d sin 'y sin ö, (3)

(4)

The first step in deriving the design parameters listed in Table 1 is setting the spectrometer's resolving powerat blaze and also the blaze wavelength (RB = 400 and AB = 20A). Using the baseline telescope angular resolution(z9 = 15 ), Equation 4 provides a relation between the graze angle 'y and the chromatic magnification i. Thesecond equation relating these two quantities is a simplified expression for the efficiency for the spectrometer systemat blaze:

EfficB 7)2Refi('Y, AB), (5)

where one factor of i comes from scalar diffraction theory (groove efficiency) and the other from the maximum(non—vignetting) grating filling factor for the grating array. The reflectivity used is that for gold, which depends on"B and 'y. The two variables are found simultaneously by numerically optimizing the efficiency (Eq. 5) where thesame two variables are constrained by Equation 4. The rest of the design parameters are readily derived from theprevious equations, and are provided in Table 1.

Table 1. Optimized grating parameters

RB = 400AB = 20 Ad = 24547 A6 = 0.605 degrees'y = 2.21 degreesa = 1.61 degrees11 = 0.57Amin = 5 AAmax50 A

2.3. GratingsThe gratings to be used in the RGS require substantial technological advances over the RGS aboard XMM, sincea factor of three weight reduction per unit area is required for Constellation—X. For XMM, the grating elementsconsisted of epoxy replicated gratings on flat, 200x100 mm silicon carbide substrates which were milled down to1 mm thickness except for five stiffening ribs extending the length of the substrate. Grating flatness was provided bythe mounting scheme, which provided contact between the four substrate corners with four co—planar contact points,or bosses, which are in turn attached to the integrating structure. Substrate fabrication was time intensive and wasthe process that limited grating production.

In the case of Constellation—X, the number of gratings required per module will indeed be large, and the weightrequirement stringent. Substrate production may not be the throughput limiting process simply because thereare very many gratings to fabricate. We therefore turned to another approach for grating fabrication, which usesphotolithography and microfabrication to produce very thin gratings "membranes" which also have flat, smoothgroove profiles. The smooth, optimally shaped grooves are fabricated using anisotropic etching techniques onto Si(111) planes which are graze—cut to the correct blaze angle (see Fig. 2). Accurate thinning of the substrates is madepossible by performing grating fabrication on buried oxide (BOX) Si wafers, which are processed in a simple sequencefollowing grating patterning (see Fig. 3).

1i(a) Interferometric

Lithography1I I1._siiiconj(b) Reactive ion Etch Sb2,

ARC, and S13N4

(c) RCA Clean

(d) Anisotropicaily Etchwith KOIl

(e) Evaporate Cr &Wet Etch S13N4

(f) Reactive Ion Etch SI &Wet Etch Cr

Figure 2. Process detail for grating patterning. The graze cut substrates are prepared such that there is a smallangle (the blaze angle 6) between the (111) planes and the wafer surface. The ruling density variation is achievedby interference lithography (left) where a stablized standing wave is set up to expose and develop the photoresist.Next, through a multistep process (right), various etchants are used to form groove surfaces that coincide with theatomically smooth (111) crystal planes.

Si

(a) Prepare Burled Oxide(BOX) Substrates

(b) Pattern Grating

(C) Etch BOX Parting LayerandThm

(d) Bond to SupportStructure and Metalize

Figure 3. Use of buried oxide (BOX) layer wafers to provide thin grating membrane.

97

Figure 4. Si (111) faceted test ruling grating fabricated at MIT. X—Ray diffraction eflicieiicy measurement resultsfor this grating are displayed in Figure 6.

98

Figure 5. Micrograph comparison of anisotropically etched, microfabricated test ruling and the mechanically ruledmaster grating used on for the XMM RGS grating replicas. Left: atomic force microscope (AFM) image of the testruling. Right: scanning tunnelling micrograph (STM) image of the mechanically ruled master. The test ruling showssuperior smoothness, groove shape and efficiency (see Fig. 6).

Si (111) planes,<0.4 nm rms roughness.

ft4eDy cm•s o* •pxottp 1 vUng I$4&1 •rneflDy uvetV:POtQtPe EXi3:34iAoLALq4p

r . wflpt tdth- '\ : :::' a rnj.-2 r P34OC.

T--* I •. * m,-2rNGO-A-

\J '\:T:kii i 2

nc4eAe 1gk [dgt.esj

Figure 6. Sample test results of the anisotropically etched test ruling fabricated at MIT and comparisons tomeasurements of other gratings. The two plots correspond to measurements at two different wavelengths: A1K/8.34A(left), and CuL/13.34A (right). On both plots there are curves for m = —1, m = —2 and m = 0 efficiencies as afunction of incidence angle c. Grating measurements included for comparison are for a holographically ruled gratingprovided by Tayside Optical (TOT) and the mechanically ruled master grating by Perkin—Elmer (PMG). As thesefigures show, the microfabricated grating is equivalent to or better than the other master gratings.

Several prototype gratings have already been fabricated at MIT (see Fig. 4) and tested at Columbia to explorethe anisotropic etching techniques applied to blazed gratings. Figure 5 provides a comparison of groove profilesbetween the microfabricated grating and the mechanically ruled master grating which was used for the XMM RGSreplicas. The replication proces yielded progressively worse micro—roughness at each stage, and final flight partshad roughness of order '' 13A rms. In contrast, the microfabricated grating grooves are atomically smooth, i. e.,difficult to measure but less than 4 A rms.

The reflectivity of the sample ruling is also very good. Figure 6 is a comparison of grating efficiencies as a functionof incidence angle for two X—Ray energies, and shows equivalent or better performance to other master gratings inthe comparison. With refinements to the process these efficiencies will most likely improve.

Fixturing the gratings within the array will be done in a modular fashion. While many details regarding thefixturing have yet to be determined, the general approach is pictured in Figure 7. This figure shows an integratingstructure that supports the reflection grating array at the exit plane of the telescope. To reduce weight and improvestrength, the mirror support spider and attachment ring might also be integral to the monolithic RGA supportstructure.

2.4. CCDsAs mentioned above, the spectroscopic readout and zero order camera form an optical bench with tight weight andpower requirements, and is schematically pictured in Figure 8. In the spectroscopic readout, the first order, 5—60Aband is imaged by six, 50x26 mm chips employing rectangular pixels that approximately match the aspect ratio ofthe spectrometer's PSF.

The RGS CCDs must have high quantum efficiency (QE) at low energies and significant power and weightreductions must be made relative to advanced CCD cameras such as ACIS aboard Chandra. Backside illuminatedCCDs are an obvious choice for their QE, however their production acceptance yield is currently too small to bea practical choice. Our baseline choice for the detector array elements is a resistive gate CCD (RGCCD) whichpromises to combine the low energy quantum efficiency properties of a backside illuminated chip with the superiorproduction acceptance yield, charge collection and charge transport properties of frontside illuminated CCDs.

In RGCCDs, the capacitance between charge clocking electrodes (gates) is substantially reduced and effectivelyreplaced by a distributed resistance (see Fig. 9). The reduced capacitance allows for very low clocking power,

99

:$4*Aee Je a (dgra*a]

Figure 7. A Sketch of a possible approach for the integrating structure holding the grating array (left) whichcontains sub—module groups of gratings (right). The thin silicon gratings are bonded to precision ground sideribswith the correct taper to stack the gratings at the appropriate angles and separations. The ribs are notched forprecision locating against reference bosses in the full-up assembly.

26

1dator [>

Power Drtal

[stemEctronic

Po*erBs

Figure 8. Baseline design schematic for the spectroscopic readout and zero order camera for the RGS. The twochips to the right form the zero order camera and will be spaced farther from the spectroscopic CCD array. The5—60A wavelength band (in first order) are intercepted by the six chips shown.

100

Grating Sets(indtvtdual gratingsnot shown)

Inner Mirror Shell

Outer Mirror Shell(remaining shellsremoved for darity)

Mirror Support Spider

Attachment Ring

RGA Support Column(attaches to spder)

RGA MonolithicSupport Structure

Si Grattng

Notch for PrecisionLocating Bosses(4 bosses per grating set)

Bonded Set of Gratings

Siderib,Ground withAppropriateTaper

Inset View ol CCD

JIIIIIIIIIlIIIIIIIII

RESISTIVE-GATE VS. CONVENTIONAL CCD

Figure 9. A schematic comparison of a conventional frontside illuminated CCD (left) to a resistive gate CCD (right)showing the lateral potential wells set up by the clock pattern generator.

Figure 10. RGCCD designs. Left: a cartoon diagram showing the structure of the RGCCD clocking gates. Right:an electron micrograph image of a prototype RGCCD fabricated at MIT Lincoln Laboratories. A single pixel is threetimes longer than the squares seen in the upper left. The meshlike pattern is the resistive polysilicon gate.

permitting 10 to 20 times lower than the conventional MOS devices. They provide very good low energy quantumefficiency, since the overlapping gates in a frontside illuminated device are replaced by a single, thin mesh of polysiliconelectrode overlapping the photosensitive silicon bulk. Production of RGCCDs involve a factor of two fewer processingsteps than conventional CCDs and thus promise a higher yield. They are intrinsically radiation—hard, since the onlysignificant trapping losses occur under doped gate regions which occupy less than 5% of the device area. Furthermore,a rectangular pixel geometry which is optimally suited for the RGS PSF is a natural choce for the RGCCD structure.

Prototype RGCCDs have been fabricated at MIT Lincoln Laboratories and X—Ray testing is in progress. Aschematic diagram for the RGCCD architecture as an electron micrograph are given in Figure 10. Initially somecharge transport problems were observed, which required adjustment to doping levels within the resistive gate. Suchdopant tuning may be performed on existing RGCCD chips, which is another beneficial property of these devices,ensuring a yield substantially higher than backside illuminated CCDs, for instance.

101

I I I

CONVENTIONAL CCD RESISTIVE-GATE CCD

TYPICALLY THREE POLY LEVELS

GATES MUST BE CLOSELY SPACED:— HIGH INTERELECTRODE CAPACITANCE- SUSCEPTIBLETO GATE-GATE SHORTS

HIGH CLOCK RATE TO MOVE CHARGELONGDISTANCES

SINGLE POLY LEVEL

GATES CAN BE WIDELY SPACED:- LOW INTER ELECTRODE CAPACITANCE— VERY LOW SUSCEPTIBIUTY TO SHORTS

LOW CLOCK RATE CAN MOVE CHARGELONG STANCES

Layout of Prototype RGCCD

10 20 30 40 50wavelength (A)

Figure 1 1 . Baseline scientific performance for the Constellation-X RGS instrument. Optimized parameters, nominalmirror module effective area functions (multiplied by six), nominal RGCCD quantum efficiency functions and a vectordiffraction efficiency calculation were combined to produce the effective area curves (left). The right two plots show,for m = — 1 , the resolution element LA and the resolving power, plotted as a function of wavelength.

3. BASELINE SCIENTIFIC PERFORMANCEUsing the baseline telescope effective area, nominal RGCCD quantum efficiency, grating filling factor iand a vectordiffraction theory calculation for a triangular groove profile for the design incidence angle in Table 1, preliminaryeffective areas are calculated and given in Figure 11. The figure also provides FWHM resolution elements £\ andresolving power A/LA for the first spectral order (m =—1) as a function of wavelength over the first order bandpassof the spectrometer. As mentioned above, the SXT zero—order camera effective area curve given in this plot, whichcomplements the SXT/microcalorimeter effective area curve, which is heavily absorbed at long wavelengths. Theresolving power for the spectrometer is limited by the telescope's angular FWHM; an alignment budget characteristicof the XMM RGAs was used in these calculations, and are provided in the plot. Any technological improvements inthe ultimate mirror performance will translate directly to an improvement in the spectrometer's resolving power.

REFERENCES1. H. Tananbaum, N. E. White, J. A. Bookbinder, F. E. Marshall, and F. Cordova, "Constellation x—ray mission:

implementation concept and science overview," in EUV, X—Ray, and Gamma—Ray In3tiiimentation for Astronomyx, 0. H. Siegmund and K. A. Flannagan, eds., Proc. SPIE 3765 (this volume), 1999.

2. R. L. Kelly, K. R. Boyce, S. R. Breon, S. S. Holt, J. S. H. Moseley, D. B. Mott, F. S. Porter, C. M. Stahie,A. E. Szymkowiak, M. D. Audley, K. C. Gendreau, R. Fujimoto, K. Mitsuda, Y. Ishisaki, D. McCammon, andT. Mihara, "Astro—e high resolution x—ray spectrometer," in EUV, X—Ray, and Gamma—Ray Instrumentationfor Astronomy X, 0. H. Siegmund and K. A. Flannagan, eds., Proc. SPIE 3'765 (this volume), 1999.

3. C. K. Stahie, "Toward a 2—ev microcalorimeter x—ray spectrometer for the constellation—x mission," in EUV,X—Ray, and Gamma—Ray Instrumentation for Astronomy X, 0. H. Siegmund and K. A. Flannagan, eds., Proc.SPIE 3765 (this volume), 1999.

4. A. C. Brinkman, H. J. M. Aarts, A. J. F. den Boggende, T. Bootsma, L. Dubbeldam, J. W. den Herder,J. S. Kaastra, P. A. J. de Korte, B. J. van Leeuwen, R. Mewe, E. van Zwet, T. Decker, C. J. Hailey, S. M.Kahn, F. Paerels, S. Pratuch, A. Rasmussen, G. Branduardi-Raymont, P. Guttridge, J. Bixier, K. Thomsen,A. Zehnder, and C. Erd, "The reflection grating spectrometer on—board of xmm," in EUV, X—Ray, Gamma—RayInstrumentation for Astronomy VII, 0. Siegmund and M. Gummin, eds., Proc. SPIE 2808, pp. 463—480, 1996.

5. S. M. Kahn, J. Cottam, T. A. Decker, F. B. S. Paerels, S. M. Pratuch, A. Rasmussen, J. Spodek, J. V. Bixler, A. C.Brinkman, J. W. den Herder, and C. Erd, "The reflection grating arrays for the reflection grating spectrometer on—board xmm," in EUV, X-Ray, Gamma-Ray Instrumentation for Astronomy VII, 0. Siegmund and M. Gummin,eds., Proc. SPIE 2808, pp. 450—462, 1996.

4000

3000

2000

1000

010

0.08

0.06C0

0.040)

total

telescope

0)

00.

>0U,0)

30wavelength (A)

alignment0.02 .. .

0.00 ______________

10 20 30 40 50wavelength (A)

102

6. A. Rasmussen, J. Cottam, T. A. Decker, S. M. Kahn, J. Spodek, M. Stern, C. Erd, A. den Boggende, A. C.Brinkman, J. den Herder, F. B. Paerels, and C. de Vries, "Performance characterization of the reflection gratingarrays (rga) for the rgs experiment aboard xmm," in X—Ray Optics, Instruments, and Missions, R. B. Hooverand A. B. C. W. II, eds., Proc. SPIE 3444, pp. 327—337, 1998.

7. A. Rasmussen, J. Cottam, T. A. Decker, S. M. Kahn, M. Sako, J. Spodek, M. Stern, A. C.Brinkman, J. den Herder, F. B. Paerels, C. de Vries, A. den Boggende, G. Branduardi-Raymont,C. Erd, G. Vacanti, H. Brauninger, and W. Burkert, "Analysis of real data acquired using the rgs," inhttp : //astro . estec .esa.nl/XMM/news/wsl/procs/rasmussena.ps.gz, M. Dahiem, ed., Proc. First XMM

Workshop: Science with XMM, Sept. 30—Oct. 2 1998.

103