Gas pixel detectors

ARTICLE IN PRESS

Nuclear Instruments and Methods in Physics Research A 572 (2007) 160–167 www.elsevier.com/locate/nima

Gas pixel detectors R. Bellazzinia,, L. Baldinia, A. Breza, F. Cavalcaa, L. Latronicoa, M.M. Massaia, M. Minutia, N. Omodeia, M. Pesce-Rollinsa, C. Sgro´a, G. Spandrea, E. Costab, P. Soffittab a Istituto Nazionale di Fisica Nucleare di Pisa, Largo B. Pontecorvo, 3 I-56127 Pisa, Italy Istituto di Astrofisica Spaziale e Fisica Cosmica, Via del Fosso del Cavalier,100 I-00133 Roma, Italy

b

Available online 15 November 2006

Abstract With the Gas Pixel Detector (GPD), the class of micro-pattern gas detectors has reached a complete integration between the gas amplification structure and the read-out electronics. To obtain this goal, three generations of application-specific integrated circuit of increased complexity and improved functionality has been designed and fabricated in deep sub-micron CMOS technology. This implementation has allowed manufacturing a monolithic device, which realizes, at the same time, the pixelized charge-collecting electrode and the amplifying, shaping and charge measuring front-end electronics of a GPD. A big step forward in terms of size and performances has been obtained in the last version of the 0.18 mm CMOS analog chip, where over a large active area of 15  15 mm2 a very high channel density (470 pixels/mm2) has been reached. On the top metal layer of the chip, 105,600 hexagonal pixels at 50 mm pitch have been patterned. The chip has customable self-trigger capability and includes a signal pre-processing function for the automatic localization of the event coordinates. In this way, by limiting the output signal to only those pixels belonging to the region of interest, it is possible to reduce significantly the read-out time and data volume. In-depth tests performed on a GPD built up by coupling this device to a fine pitch (50 mm) gas electron multiplier are reported. Matching of the gas amplification and read-out pitch has let to obtain optimal results. A possible application of this detector for X-ray polarimetry of astronomical sources is discussed. r 2006 Elsevier B.V. All rights reserved. Keywords: Imaging detectors; Pixels detectors; Proportional counters; X-ray astronomy; Polarimetry

1. Introduction Amongst the various types of micro-pattern gas detectors introduced in the last decade, gas pixel detectors (GPD) represent the last step towards a level of integration and compactness comparable to solid-state detectors. Since the year 2001 we have been actively working at INFN-Pisa on the GPD concept in which a custom CMOS analog chip is simultaneously the pixelized charge collecting electrode and the front-end electronics (amplifying, shaping and charge measuring) of a suitable charge multiplier, a gas electron multiplier (GEM) in our case. Three different generations of application-specific integrated circuit (ASIC) of increased size, reduced pitch and improved functionality have been designed and fabricated (Fig. 1). Performance and test results of the first two VLSI chips, Corresponding author. Tel.: +39 50 2214367; fax: +39 50 2214317.

E-mail address: [email protected] (R. Bellazzini). 0168-9002/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2006.10.171

with 2101 and 22,080 pixels at 80 mm pitch (in 0.35 mm technology) can be found in Refs. [1,2]. Here we report on the third generation, realized in 0.18 mm CMOS technology, with 105,600 pixels at 50 mm pitch organized in a honeycomb array and on the GPD built up by coupling the analog VLSI ASIC to a gas cell charge multiplier. The detector provides unique performances, in terms of imaging, spectral and timing capabilities and it is for these reasons that it has been proposed as a photoelectric polarimeter at the focus of a JET-X like optics in a small satellite mission (POLARIX, [3]). It has also been proposed at the focus of the large area mirror of the XEUS telescope [4], the ESA permanent space borne X-ray observatory planned to be launched in 2015. The sensitivity and efficiency of the instrument onboard of the XEUS observatory will allow to perform energy-resolved polarimetry at the level of few % on many galactic and extragalactic sources with photon fluxes down to milliCrab or fractions of milliCrab. The minimum detectable

ARTICLE IN PRESS R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 160–167

161

Fig. 1. Photo of the three ASIC generations. The large area chip is shown bonded to its ceramic package.

polarization (MDP) for a few representative celestial sources have been measured via a full Monte Carlo simulation of the detector and the results from these simulations are presented in this paper.

Fig. 2. Schematics of the concept of the gas pixel detector.

2. The GPD components: The GEM and the VLSI chip The principle of operation of the GPD is schematically shown in Fig. 2. When an X-ray photon is absorbed in the low Z gas mixture, usually 50% Ne or He and 50% DME, a photoelectron is emitted and its ionization track is drifted by the applied electric field (1 kV/cm) towards the GEM amplification structure. Within the GEM holes, the primary electrons are multiplied by the high electric field (100 kV/cm) and then collected by the pixelized read-out electrode underneath. To match the read-out pitch a 50 mm pitch GEM has been manufactured at CERN. In this configuration, the very high granularity of the device was fully exploited and optimal spatial resolution and imaging capability have been obtained. The technological challenge in the fabrication of this type of GEM has been the precise and uniform etching of the very narrow multiplication holes (33 and 15 mm diameter, respectively, at the top and in the middle of the kapton layer). The GEM has worked very well with the two gas mixtures, He and Ne based, used in the tests. A large effective gain (well above 1000) has been reached at a voltage reduction of 70 V if compared to what was previously obtained with a 90 mm pitch GEM and such results are likely due to the higher electric field line density inside the narrow amplification holes. Typical high voltage settings with gas filling of Ne(50%)–DME(50%) are: VDRIFT ¼ 1800 V, VGEMtop ¼ 750 V, VGEMbottom ¼ 300 V, with the read-out electrode at the charge preamplifier input voltage, i.e. 1.5 V. A photo of the detector during the assembly phase is shown in Fig. 3. The GEM foil glued to the bottom of the gas-tight enclosure

Fig. 3. Assembly phase of the detector. The GEM foil glued to the bottom of the gas-tight enclosure and the large area ASIC mounted on the control motherboard are well visible. Absorption gap ¼ 10 mm, collection gap ¼ 1 mm.

and the large area ASIC bonded to the control motherboard are well visible in the photo. As for the previous two ASIC generations, this last version of VLSI analog chip also has a full electronic chain (low noise charge preamplifier, shaping amplifier, peak and hold and input to a multiplexer stage) implemented for each pixel in the five metal and single poly-silicon layers of the 0.18 mm CMOS technology that lies beneath the top metal layer where 105,600 pixels have been patterned. The pixels are arranged in a 300  352 honeycomb matrix corresponding to an active area of 15  15 mm2 and pixel density of 470 mm2. The chip integrates more than 16.5 million transistors and it is subdivided in 16 identical clusters of 6600 pixels (22 rows of 300 pixels) or alternatively in 8 clusters of 13,200 pixels (44 rows) each one with an independent differential analog output buffer.

ARTICLE IN PRESS 162

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 160–167 Table 1 Main ASIC characteristics Peaking time (externally adjustable) Full scale linear range Pixel noise Read-out mode Trigger mode Read-out clock Self-trigger global threshold Frame rate Parallel analog output buffer Access to pixel counter

Total power dissipation Fill fraction (metal to active area ratio)

Fig. 4. Definition of the self-trigger 4 pixel minicluster.

Each cluster has a customizable internal self-trigger with independently adjustable threshold. Twenty-five thousand predefined groups of 4 contiguous pixels (mini-cluster in Fig. 4) contribute to the local trigger which has a dedicated amplifier whose shaping time is roughly a factor 2 faster (1.5 ms) than the shaping time of the analog signal to allow its peak detection and hold. Each noisy pixel can be masked with respect to its contribution to the trigger signal and/or its analog output. The internally wired OR of all the mini-cluster self-trigger circuits commands the hold of the maximum of the shaped analog signal of each of the 105,600 pixels. The event is then localized in the rectangular area, which contains the triggered mini-clusters plus a user selectable margin of 10–20 pixels. The Xmin, Xmax, Ymin, Ymax coordinates of this region are available as four 9-bit data as soon as the data acquisition process (internally triggered) terminates, flagged by the DataReady signal. The event window coordinates can be copied into a serial–parallel IO interface register (a 36-stage FIFO) by applying an external signal. Subsequently, clock pulses push out the analog data to a serial balanced output buffer compatible with the input stage of a 12-bit ADC. In self-trigger mode the read-out time is largely reduced (at least a factor 100) with respect to the standard sequential read-out of the full matrix (still available) thanks to the relatively small number of pixels (700 on average) within the event region of interest (ROI). Table 1 summarizes the main characteristics of the VLSI chip. To generate and handle command signals to/from the chip a very compact data acquisition system has been implemented on Altera FPGA Cyclone EPIC240 (RISC NIOS-II processor). Analog data area is read, digitally converted by the ADS5270TI Flash ADC and transitorily stored in a static RAM. In self-trigger mode and by using the embedded RISC processor it is possible to read, just after an event acquisition, the pixel pedestals in the

3–10 ms 30,000 electrons 50 electrons ENC Asynchronous and synchronous Internal, external, self-trigger Up to 10 MHz 2000 electrons Up to 10 KHz (event window self trigger) 1, 8 or 16 Direct (single pixel) or serial (8–16 clusters, full matrix, region of interest) 0.5 W (self-trigger mode) 92%

chip-defined ROI, one or more times, compute the average and transfer the pedestal subtracted data to the offline analysis system. The great advantage of this operation is to allow a real time control of the data quality and to cancel time drift effects of the pedestal values with temperature or any other environmental changes. Disadvantages manifest themselves as a slight pffiffiffi increase of the channel noise (maximum a factor 2 in case of one pedestal reading) and of the event read-out time. For most of the applications, we consider this fact not to be a real problem due to the very high signal to noise ratio (well above 100) and to the very fast operation in window mode. The standard reading of all the 105K pedestals at the beginning or at the end of a data taking is nonetheless still possible. On-line detector control and data acquisition is done with a LabVIEW Graphic Unit Interface, which performs a bidirectional communication via a 100 Mbps TCP connection between the DAQ and a portable PC. 3. Laboratory tests and results Functional tests have been performed on the chip stand alone and on the assembled detector. Due to the very low pixel capacitance an average noise of 50 electrons ENC (3 ADC counts with an input amplifier sensitivity of 350 ADC counts/fC) has been measured. This allows reaching, at a gas gain of 1000 and with a 3 sigma threshold, good single primary electron sensitivity, very important for polarimetric applications. To operate in self-trigger mode a threshold for the whole chip is set and the modality enabled in the configuration register. The lowest applicable global threshold is fixed by the offset variations of the pedestals, 10% of the linear dynamics in CMOS technology, more than the pedestal fluctuations. The fake trigger rate as a function of the global threshold has been measured in self-trigger mode and the results are reported in the plot of Fig. 5. All pixels were working and no mask was applied to suppress noisy channels. In these conditions and at a working threshold of

ARTICLE IN PRESS R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 160–167

163

6

40

4 30

20 Few Hertz @2300 e 10

Y position (mm)

Noise Trigger Rate (counts/s)

50

2 0 -2 -4

0 1200 1400 1600 1800 2000 2200 2400 2600 2800 3000 Threshold (electrons)

-6

Fig. 5. Fake trigger rate as a function of the threshold (in electrons).

-4

2 -2 0 X position (mm)

4

6

-6

-4

-2 0 2 X position (mm)

4

6

6 4 Y position (mm)

2300 electrons the fake trigger rate is reduced to 3 Hz or less. This threshold is much lower than the one which was possible to apply when working with the top GEM signal as external trigger to start the read-out sequence. Many tests have been done on the detector with different gas fillings and radiation sources (55Fe or X-ray generator with Cr anode) in self-trigger mode. On average, the size of the ROI around the triggering event is 700 pixels for 5.9 illumination. The use of a GEM with a very fine pitch that well matches the read-out one has noticeably increased the spatial resolution and the 2D imaging capability of the detector. This fact is very important in X-ray polarimetry, because it allows to reconstruct the photoelectron tracks with a high degree of detail even in the case of low-energy (2–3 keV) photons. It also allows to recognize the initial part of the track through the localization of the absorption point of the photon and then to estimate the photoelectron emission direction. It is well known that in photoelectron absorption, s-photoelectrons are ejected according to a cos2 distribution with respect to the photon electric field (the polarization vector). The angular distribution of the photoelectron emission direction gives, therefore, information on the angle and degree of polarization of the converted X-ray photons. Fig. 6 shows a sort of radiographic image obtained by illuminating a small pendant of few mm in size placed in front of the detector with 5.9 keV photons from a 55Fe source. The two scatter plots of Fig. 6 refer, respectively, to the coordinates of the center of gravity (top panel) and of the reconstructed absorption point (bottom panel). The higher accuracy obtained with the absorption point is due to the choice of a light gas mixture, i.e. based on neon or helium, where the ionization tracks of photoelectrons are long and well sampled by the high detector granularity and the barycentre is far from the absorption point. Nevertheless, the high resolving power reached by using a GEM and a read-out electrode with the same fine pitch, allows

-6

2 0 -2 -4 -6

Fig. 6. Radiographic image of a small pendant obtained with 5.9 keV photons from a 55Fe source, (a) scatter-plot of barycenters and (b) scatterplot of reconstructed absorption points.

getting a better 2D reconstruction with the absorption point with respect to the barycentre, even with high Z Arbased mixtures as shown in Figs. 7(a) and (b). The phantom of Fig. 7 was machined with a set of holes of 0.5 mm of diameter and spaced, edge-to-edge, 0.5 mm. With a standard 90 mm pitch GEM, even with a read-out at 50 mm, the difference between barycentre and absorption point reconstruction would have been negligible. The GPD described in this paper is not only an excellent imager but also a good proportional counter with energy resolution of about 14–15% at 6 keV, depending on the gas filling. This feature is suitable for crucial energy resolved measurements, for example in polarimetry to separate the

ARTICLE IN PRESS R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 160–167

164

1200

6

25

mean2 = 494.38 +- 0.15 sigma2 = 45.10 +- 0.13

1000 800

4

ΔE/E (FWHM) = 15%

Y position (mm)

20 2

600 15

0

400

-2

10

mean1 = 151.99 +- 0.77 sigma1 = 53.93 +- 0.91

200

-4

0

5

0

-6 0 -6

-4

-2

0

2

4

6

100

200

300

400

500

600

Fig. 8. Pulse height distribution of top GEM signals. Gas filling: 50% Ar–50% DME. Resulting energy resolution (@5.9 keV) ¼ 15%, FWHM.

X position (mm)

600 35 6

500 30

Y position (mm)

4

400 25

2

PHeight Entries 9567 χ2 / ndf 48.12 / 36 Prob 0.08536 103.2+- 4.4 P0 1984+- 16.6 P1 385+- 16.7 P2 582.6+- 8.3 P3 4 4317+- 6.9 P 573.9+- 6.0 P5

300 20

0

200 15

-2

100 -4

10

-6

5

0 1000

2000

3000

4000

5000

6000

7000

8000

Pulse Height (ADC counts) 0 -6

-4

-2

0

2

4

6

X position (mm)

Fig. 9. Energy resolution (at 5.9 keV) obtained summing the charge of the ionization track collected on the read-out pixels in 50% Ar–50%DME gas mixture.

Fig. 7. 2D Image reconstruction of a multihole phantom. Hole diameter ¼ 0.5 mm, pitch ¼ 1 mm, (a) scatter-plot of barycentres and (b) scatter-plot of reconstructed absorption points.

4. An application in X-ray polarimetry

unpolarized fluorescence from the partially polarized continuum in reflection spectra. This property will allow for example to test general relativity effects in the matter around a black hole through the measurement of the polarization angle as a function of energy [5,6]. Fig. 8 illustrates the pulse height distribution of the signals from the top GEM at 5.9 keV in 50%Ar–50% DME. At this energy DE/E15%, FWHM, is measured. A good energy resolution is still obtained by collecting the charge of the ionization track on the read-out pixels (Fig. 9). It is important to remark that, after several months of intensive operation, no die nor a single pixel has been lost for electrostatics or GEM discharges, or for any other reason.

A long-term theoretical analysis has foreseen the possibility to test models of X-ray sources and to derive relevant parameters of the models by measuring the linear polarization of X-rays. Unfortunately, traditional polarimeters based on Bragg diffraction or Compton scattering are characterized by a poor sensitivity and have given to date positive results only for a very few bright sources, for example the first and only generally accepted measurement of the Crab nebula made with a Bragg crystal polarimeter more than 30 years ago [7,8]. Moreover, scattering polarimeters are practically insensitive below 5 keV and are background limited, while Bragg crystal polarimeters are efficient only around a narrow band fulfilling the Bragg condition. The detector described in this paper can detect

ARTICLE IN PRESS R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 160–167

and convert photons in a large range of energies, from a few keV up to tens of keV, by choosing the appropriate gas mixture filling and image with high resolving power photoelectron tracks, making possible focal plane photoelectric polarimetry. An example of reconstructed tracks obtained with 5.4 keV photons in two different gas mixtures (50% Ne–50% DME and 40% He–60% DME) is reported in Fig. 10. The black cross refers to the barycentre while the white cross indicates the reconstructed absorption point. The direction of the principal axis of the charge distribution (passing through the center of gravity) and of the reconstructed emission direction of the photoelectron are shown. The requirement of measuring the degree of polarization at a few % level for a large variety of galactic and

4.8

extragalactic sources sets a strong constraint to the amount of systematic effects that can be accepted without spoiling the polarimetric capability of the device. A residual modulation of 0.7070.32% (statistical error) has been measured with a totally unpolarized 55 Fe source (Fig. 11). This value represents the limit on the MDP, but it could be reduced further by identifying possible nonuniformities in the system, which can affect the data. To check the sensitivity of the detector to polarized radiation and to measure the modulation factor of the angular distribution, 99% linearly polarized photons have been produced by 901 Thomson scattering of a pencil beam from Cr X-ray tube (5.4 keV, 20 kV, 35 mA). The scatterer was a lithium target.

-3

50% Ne - 50% DME

40% He -60% DME

-3.2 Y cocordinate (mm)

4.6 Y cocordinate (mm)

165

4.4 4.2 4.0 3.8 3.6

-3.4 -3.6 -3.8 4 -4.2 -4.4

3.4

-4.6 -4.8

3.2 4.6 4.8

5

5.2 5.4 5.6 5.8

6

2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4

X cordinate (mm)

X cordinate (mm)

Fig. 10. Real photoelectron tracks obtained by irradiating the detector with 5.4 keV X-rays from a Cr tube, for two gas mixtures. The black cross refers to the barycenter while the white cross indicates the reconstructed absorption point. The major principal axis (through the barycenter) and the reconstructed emission direction of the photoelectron (through the absorption poit) are also shown.

2500

2000

Theta1 Entries χ2 / ndf Flat term Modulated term Angle

1500

192311 68.59 / 77 2386+- 9.5 33.75 +- 15.54 0.3535+- 0.2290

1000

Residual modulation: 0.70+-0.32

500

0 -3

-2

-1

0

1

2

Phi (rad) Fig. 11. Residual modulation from an unpolarized X-ray source (55Fe).

3

ARTICLE IN PRESS 166

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 160–167 Theta1 Entries 10993 62.56 / 77 χ2 / ndf 62.39+- 1.73 Flat term 147+- 3.5 Modulated term -0.02402+- 0.01205 Angle

Theta1 Entries 23519 2 / ndf 77.31 / 77 χ 143.3+- 2.5 Flat term 299.5+- 5.2 Modulated term -0.00748+- 0.00872 Angle

500

220 200 180 160 140 120 100 80 60 40 20 0

μ (5) = 51.11+- 0.89

400 300 200 100

50% Ne - 50% DME

0 -3

-2

-1

0 1 Phi (rad)

2

3

μ (5) = 54.26+- 1.24

40% He - 60% DME

-3

-2

-1

0 1 Phi (rad)

2

3

Fig. 12. Modulation factor measured at 5.4 keV with two different gas mixtures: 50% Ne–50% DME and 40% He–60% DME at 5.4 keV.

μ (%) = 57.01+- 1.38

200 180 160 140 120 100 80 60 40 20 0

μ (%) = 64.33+- 1.54

160 140 120 100 80 60 40 20

50% Ne - 50% DME

40% He - 60% DME

0 -3

-2

-1

0 1 Phi (rad)

2

3

-3

-2

-1

0 1 Phi (rad)

2

3

Fig. 13. Modulation factor obtained at 6.4 keV with same gas mixtures of Fig. 12.

0.20 Ne 40% + DME 60% 1atm 1cm (solid line) DME 60% + HE 40% 1atm 1cm (dotted line) DME 80% + HE 20% 1atm 1cm (dashed line)

Mu∗SQRT (Efficiency)

Different combinations of gas filling and energies have been tested and high modulation values, spanning from 51.1170.89% (5.4 keV, 50% Ne–50%DME) to 64.337 1.54% (6.4 keV, 40% He–60%DME), have been measured. The correspondent angular distributions are shown in Figs. 12 and 13. The optimal performance obtained with the detector filled with a He–DME mixture, is well tuned with the baseline design of a polarimeter for the XEUS telescope [4]. Being a focal plane-imaging instrument, the proposed polarimeter has very low background [9] and sensitivity essentially dominated by the source flux, down to very low values. The choice of the main parameters of the detector (i.e. gas mixture and pressure, absorption gap, window thickness,) determine the quality factor of the polarimeter, pffiffi defined as m , where m is the modulation factor and e the detection efficiency. The plot of the quality factor as a function of energy, obtained with a full Monte Carlo simulation [10,11] of the polarimeter at the focus of XEUS optics, is shown in Fig. 14. From the obtained results, it is well evident that mixtures based on helium can better exploit the high effective area of the XEUS mirror (5 m2

0.15

0.10

0.05

0.00 0

2

4

6 Energy (keV)

8

10

12

Fig. 14. Quality factor obtained with Monte Carlo simulations of the detector at the focus of the XEUS optics for three different gas mixtures. He-based mixtures provide better response at lower energies.

and 35 m focal length) at low energy. From the plot of the MDP as a function of the source flux reported in Fig. 15, it results that polarization of several AGNs with milliCrab

ARTICLE IN PRESS

Minimum Detectable Polarization (%)

R. Bellazzini et al. / Nuclear Instruments and Methods in Physics Research A 572 (2007) 160–167

102 He (40%) DME (60%) 1 cm 1 atm Dotted line=observation time 1 hour Solid line=observation time 24 hour

10 FSRQ

1

Circinus Circinus MGC-6-30-15 Centaurus A

BL Lac

10-1

Hercules X-1

From top to bottom Spectral index = 0.5 Spectral index = 2 Spectral index = 3.5

10-2 10-2

10-1

Vela X-1

Crab

1

10

10-2

10-3

Flux 2-10keV (mCrab) Fig. 15. Minimum detectable polarization as a function of the photon flux for the proposed detector at the focus of the XEUS optics. A few representative milliCrab sources are shown.

fluxes, can be measured down to a few % levels with observation time of 1 d. 5. Conclusions With devices like the one described in this paper, the class of gas pixel detectors has reached a level of integration, compactness and resolving power so far considered typical of solid-state detectors. Depending on the type of electron multiplier, pixel and die size, electronics shaping time, analog or digital read-out,

167

counting or integrating functionality, many different applications can be envisaged. In particular, the possibility to use the presented detector as a high-sensitivity polarimeter has been extensively studied. The obtained performance, in terms of very low residual modulation and high modulation factor, well above 50%, makes it a high throughput tool for X-ray astronomy. It is for these reasons that it has been proposed as focal plane instrument for the XEUS telescope. It is worth noticing that, with a similar approach, a digital counting chip developed for medical applications (Medipix2) has been shown to work coupled to GEM or micromegas gas amplifiers for TPC applications at the next generation of particle accelerators [12,13]. References [1] R. Bellazzini, et al., Nucl. Instr. and Meth. A 560/2 (2006) 425. [2] R. Bellazzini, et al., Nucl. Instr. and Meth. A 535 (1–2) (2004) 477. [3] R. Bellazzini, et al., in: Proceedings of SPIE 6266-29, Orlando, 24–31 May 2006. [4] R. Bellazzini, et al, in: Proceedings of SPIE 6266-149, Orlando, 24–31 May 2006. [5] P.A. Connors, R.F. Stark, Nature 269 (1977) 128. [6] P.A. Connors, R.F. Stark, T. Piran, Astrophys. J. 235 (1980) 224. [7] R. Novick, M.C. Weisskopf, et al., Astrophys. J. 174 (1972) L1. [8] M.C. Weisskopf, et al., Astrophys. J. 220 (1978) L117. [9] A.N. Bunner, Astrophys. J. 220 (1978) 261. [10] L. Pacciani, et al., in: Proceedings of SPIE 4843, pp. 394–405. [11] R. Bellazzini, et al., in: Proceedings of SPIE 4843, pp. 383–393. [12] M. Campbell, et al., Nucl. Instr. and Meth. A 535 (1–2) (2004) 11. [13] A. Bamberger, et al., Nucl. Instr. and Meth. A, to be submitted for publication.