X-ray polarimetry with a conventional gas proportional counter through rise-time analysis

Nuclear Instruments and Methods in Physics Research A 421 (1999) 241 248 X-ray polarimetry with a conventional gas proportional counter through rise-time analysis Kiyoshi Hayashida*, Noriyuki Miura, Hiroshi Tsunemi, Ken ichi Torii, Hiroyuki Murakami, Yoshiaki Ohno, Keisuke Tamura Department of Earth and Space Science, Faculty of Science, Osaka University, 1-1, Machikaneyama, Toyonaka, Osaka, Japan Received 20 January 1998; received in revised form 21 July 1998 Abstract We have performed an experiment on the signal rise time of a Xe gas proportional counter using a polarized X-ray beam of synchrotron orbital radiation with energies from 10 to 40 kev. When the counter anode is perpendicular to the electric vector of the incident X-ray photons, the average rise time becomes significantly longer than that for the parallel case. This indicates that the conventional gas proportional counters are useful for X-ray polarimetry. The moderate modulation contrast of this rise-time polarimeter (M"0.1 for 10 kev X-rays and M"0.35 for 40 kev X-rays), with capability of the simultaneous measuring X-ray energies and the timing, would be useful for applications in X-ray astronomy and in other fields. 1999 Elsevier Science B.V. All rights reserved. PACS: 95.55.!n; 95.55.Ka; 29.40.LS Keywords: X-ray polarimeter; X-ray astronomy; Proportional counter; Synchrotron radiation 1. Introduction The technique of polarimetry has been established and used by optical or radio astronomers in their observations. In the field of X-ray astronomy, * Corresponding author. Tel.: #81 6 850 5476; fax: #81 6 850 5539; e-mail: hayasida@ess.sci.osaka-u.ac.jp. Also at: CREST, Japan Science and Technology Corporation (JST). Present address: NASDA TKSC SURP, 2-1-1 Sengen, Tsukuba, Ibaraki 350-8505, Japan. Present address: Department of Astrophysics, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-01, Japan. however, detection of the polarization has been limited to a few sources and is not yet a routine observational technique, as reviewed by Fraser et al. [1]. Difficulties mainly come from the limited sensitivity of the polarimeter so far used in space. For example, a Bragg crystal polarimeter in the OSO-8 experiment, which consisted of hundreds of crystals, detected a significant degree of linear polarization only for a few sources. One problem of the Bragg crystal polarimeter is its limited bandwidth. Although it has a modulation contrast factor M [1] of greater than 0.9, it can make use only of a small part of the photons in the continuous energy spectra of cosmic X-ray sources. 0168-9002/99/$ see front matter 1999 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8-9 0 0 2 ( 9 8 ) 0 1 1 4 5-0

242 K. Hayashida et al./nucl. Instr. and Meth. in Phys. Res. A 421 (1999) 241 248 Alternative techniques have thus been proposed and examined. One of them is a Thomson scattering polarimeter, which makes use of the anisotropic cross section of Thomson scattering. Another example is a series of polarimeters employing the anistropic property of the photo-electron emission, e.g., photo-electrons are preferentially emitted toward a direction parallel to the electric vector of incident photon. Several photo-electron polarimeters are mentioned in Ref. [1]. Recently, Tsunemi et al. [2] indicated that a charge-coupled device (CCD) worked as a polarimeter in experiments at a synchrotron facility. In a CCD chip, an X-ray photon creates a charge cloud elongated along the momentum of the primary photo-electron. Since each pixel in a CCD chip works as a solid state detector, if the charge cloud extends beyond the pixel boundary, we will get signals from the adjacent two pixels. By analyzing those signals, we can measure the degree and the direction of linear polarization in the incident X-rays. Tsunemi et al. [2] used a CCD chip of pixel size 12 μm and obtained modulation amplitude of 3.5% for the 40 kev X- ray beam, which corresponds to the modulation contrast M"0.07. Although the modulation parameter of the CCD polarimeter is small, a great advantage of this detector is that it can work as a high-resolution imaging spectrometer at the same time. In this paper, we describe another type of photoelectron polarimeter employing the same operation principle as the CCD polarimeter. As in the CCD device, an X-ray photon deposited in a gas proportional counter produces a primary charge cloud. The size and the elongation direction of the charge cloud would reflect the rise time of the signal, even if the counter has a single anode. Thus measurement of the rise time would enable X-ray polarimetry. The basic idea of this rise-time polarimeter was proposed by Riegler et al. [3] and Sanford et al. [4] in 1970. According to Ref. [1], M parameter of less than 0.08 at 16 kev was reported, while little progress has been done so far. Although gas proportional counters have been conventionally used in the field of X-ray astronomy, rise-time polarimetry has not yet been employed in observations. In order to examine whether the rise-time polarimeter is realizable or not, we performed an experiment in a synchrotron facility. In our experiment, we positively detected modulation of the rise-time against the angle between the polarization direction and the counter anode direction. The modulation contrast M we obtained was 0.1 at 10 kev and 0.35 at 40 kev. The beam energy dependence of M is, however, different from what we expected from the simple charge cloud model in the gas proportional counters. This point will be briefly discussed. We will also discuss the advantages and the disadvantages of the rise-time polarimeter particularly for X-ray polarimetry in astronomy. 2. Polarization degree of the incident beam The experiment was performed at a synchrotron facility of the Photon Factory in the Institute of High Energy Physics, Tsukuba, Japan, in July 1993. We used beam line BL14C, which has a vertical wiggler as its insertion device in order to produce harder X-rays than other beam lines. By employing a double-crystal spectrometer set there, we used a monochromatic X-ray beam whose energy ranged from 10 to 40 kev. The X-ray beam was output into a cage which was open to the air. The X-ray detectors were set in the cage and the electronics for them were set up outside the cage. The same beam line and spectrometer were used in the CCD polarimetry experiments by Tsunemi et al. [2]. The polarization degree of the incident X-ray beam was not calibrated with other devices at that time. We thus constructed a Thomson polarimeter, whose modulation factor is well modeled, in order to measure the polarization degree of the BL14C. Our Thomson polarimeter consisted of a Xe gas proportional counter and a polyethylene target. The proportional counter, which was a backup model of the Transient Source Monitor experiment on board Temma satellite, was set on a rotating stage around the polyethylene target. The target had a cylindrical shape with 1.1 cm diameter and 6.0 cm length. The distance between the target and the window (we used a lead mask to set the window size to be 0.5 cm 1.0 cm) was 10.7 cm. With this configuration, calculated values

K. Hayashida et al./nucl. Instr. and Meth. in Phys. Res. A 421 (1999) 241 248 243 Fig. 1. Counting rate modulation of the gas proportional counter on the Thomson polarimeter for the incident X-ray beam of 34 kev at BL14C. θ is the angle of the counter window direction around the beam measured from the vertical direction (clockwise looking from the beam downstream). Fig. 2. Linear polarization degree of the X-ray beam at BL14C measured with our Thomson polarimeter. The crosses indicate the modulation of the counting rate of the proportional counter. The closed circles indicate the beam polarization, which is derived from the measured modulation by dividing by M. of the modulation constant M is 0.90 at 10 kev and 0.95 at 40 kev. Fig. 1 shows the modulation curve obtained for 34 kev X-rays at the BL14C. The counting rates plotted are corrected for the intensity change of the incident beam, although it is less than 1% during the measurement time for Fig. 1. In Fig. 1, the angle θ indicates the counterwindow direction around the beam direction, where the angle is measured from the vertical upper direction (clockwise looking from the beam downstream). Fig. 1 indicates that the electric vector of the incident beam is vertical and the X-ray beam is scattered principally in the horizontal direction. We measured similar modulation curves for 10 40 kev range. The measured modulation, counting-rate (θ"0 )! counting-rate (θ"90 ) / (counting-rate (θ"0 )# counting-rate (θ"90 )) is summarized in Fig. 2. Dividing the measured modulation by the M value of this Thomson polarimeter (M ) yields the polarization degree of the X-ray beam. As shown in Fig. 2, the polarization degree of the incident beam is about 80%$5% depending on the incident energy. It was also examined whether the offset angle of the first-crystal of the spectrometer affects the polarization degree or not, since we have made use of it to adjust the intensity of the incident beam. However, the dependence was less than 3% for the offset angle range we used in our experiment. 3. Experiment for rise-time polarimeter The experimental configuration for the rise-time polarimeter is illustrated in Fig. 3. The experiment was performed on adjacent days to the Thomson polarimeter measurement of the beam line. We shone the polarized X-ray beam on a conventional Xe gas proportional counter and digitally sampled the signal pulses. The beam energy selected by the spectrometer ranged from 10 to 40 kev. The Xe gas proportional counter we used for the signal pulse sampling was a backup model of the All Sky Monitor experiment on board the Japanese X-ray astronomy satellite Ginga [5]. The counter measured about 30 cm 5cm 5 cm and was filled with Xe gas of 736 mm Hg and CO of 25 mm Hg. The high voltage was fixed tobe 1950 V in the experiment. Although it comprised three gas cells at the front and one at the back, we used a signal from the front-center cell. The counter was set on a rotating stage. We define the rotation angle Θ to be the counter anode direction measured from the vertical upper direction. The angle is measured clockwise looking from the beam downstream, as for θ of the Thomson polarimeter. The counter window was covered with a lead sheet except for the beam entrance. In front of the counter stage, we fixed a lead collimator with a diameter of 2 mm. In addition, when we measured the fluorescence X-ray

244 K. Hayashida et al./nucl. Instr. and Meth. in Phys. Res. A 421 (1999) 241 248 Fig. 3. (a) Setup of the experiment at BL14C. (b) Sampling system for the output pulses from the Xe gas proportional counter. Fig. 4. Signal from the pre-amplifier of the Xe gas proportional counter sampled with the digital oscilloscope. Incident X-ray energy is 40 kev. We reduce the data of each signal pulse into PH (pulse height) and RT (rise time). lines as a calibration, a filter made of Sn was put on the beam exit to the experiment cage. The output signals from the built-in pre-amplifier of the counter were sampled by a digital oscilloscope (Hitachi VC-6165), which had a maximum sampling frequency of 100M samples/s and a voltage resolution of 8 bit in the storage mode. An example of the sampled pulse is shown in Fig. 4. Owing to a pre-trigger function in the oscilloscope, we could obtain data before the pulse rises. The data sampled by the oscilloscope were transferred to a personal computer via GP-IB interface and stored in its hard disk unit. We set the sampling frequency as 80 MHz and 2000 points of data were transferred to the computer. This assembly, mainly its GPIB interface, limited the processing frequency to one pulse per 1 s. The typical number of pulses we sampled for one setup was a few thousand. The signal from the built-in pre-amplifier was also processed with a shaping amplifier and monitored with a scalar and a pulse-height analyzer. We adjusted the offset angle of the first crystal in the spectrometer so that the X-ray beam intensity did not exceed 3000 c/s in order to avoid a pile up problem. Note also that the incident beam contained higher-harmonic components besides

K. Hayashida et al./nucl. Instr. and Meth. in Phys. Res. A 421 (1999) 241 248 245 the fundamental component we wanted. Those higher-harmonic components, however, could be rejected in the course of the data analysis. 4. Analysis and results The pulse-shape data stored are accessible from a personal computer, and thus we can take flexible ways for reducing those. As a first step, however, two quantities are extracted from each pulse: one is pulse height (PH), which is defined as the maximum voltage the minimum voltage, the other is the rise time (RT), which is the interval between the time when the signal exceeds 20% of the pulse height and the time when it exceeds 80% of the pulse height (see Fig. 4). We select the data whose pulse height corresponding to the incident energy we selected; escape events, higher harmonic photons and piled up events were rejected. Distributions of the rise time for the incident X-ray energy of 34 kev are shown in Fig. 5. The distribution of the rise time has a skewed form, e.g., a sharp cut-off at the shorter end and a tail toward the longer end (we cut the events whose rise time exceeds 1 μs). However, it is Fig. 5. Distribution of the signal rise time for Θ"0, 90, 180, and 270. The incident X-ray beam energy is 34 kev. When the counter anode direction is perpendicular to the electric vector of the incident photons (Θ"90 or 270 ) the rise-time distribution spreads toward the longer side.

246 K. Hayashida et al./nucl. Instr. and Meth. in Phys. Res. A 421 (1999) 241 248 Fig. 6. Average rise time for 34 kev X-ray incidence as a function of Θ. shown that the distribution of Θ"90 or 270 spreads toward the longer side more than that of Θ"0 or 180. We then calculate the average rise time for each Θ (Fig. 6). It illustrates that the average rise time is longer when Θ is 90 or 270, i.e., when the counter anode direction is perpendicular to the electric vector of the incident photons, than the case when Θ is 0 or 180. This result is what we expected from the operation principle of the risetime polarimeter mentioned in the introduction. For other energies, 10, 20, 30, 36 and 40 kev, we took the data for Θ"0, 30, 60 and 90. Fig. 7 shows the summary of the average rise time for each set up. The average rise time becomes longer for the higher-energy incident X-rays. This might be expected because the primary electron cloud becomes larger for higher-energy photons. For Fig. 7. Average rise time for incident X-ray energies of 10, 20, 30, 34, 36, 40 kev and for Θ"0, 30 60, 90. Error bar denotes the statistical error.

K. Hayashida et al./nucl. Instr. and Meth. in Phys. Res. A 421 (1999) 241 248 247 each energy, the average rise time for Θ"90 is longer than that for Θ"0, as for the 34 kev case. RT(Θ"90 )!RT(Θ"0 ) is 0.02 μs for 10 kev X-rays and 0.07 μs for 40 kev X-rays. We define the modulation contrast of the rise-time polarimeter as M" RT(Θ"90 )!RT(Θ"0 ) /2σN /P, Fig. 8. Modulation contrast for the rise-time polarimeter derived from our experiment. M" RT(Θ"90 )!RT(Θ"0 ) / 2σN /P, where σn " σ (Θ"0 )#σ (Θ"90 ) /2, σ denotes the standard deviation of the rise time distribution, RT is the average rise time, and P is the polarization degree of the incident beam (Fig. 2). where σn " σ (Θ"0 )#σ (Θ"90 ) /2, σ denotes the standard deviation of the rise-time distribution, RT is the average rise time, and P is the polarization degree of the incident beam (Fig. 2). This definition is appropriate for the comparison with other types of polarimeters. Fig. 8 shows the modulation contrast of the rise-time polarimeter measured by us, M"0.1 for 10 kev X-rays and M"0.35 for 40 kev X-rays. In order to verify that the rise-time modulation is produced by the beam polarization and not by other unknown factors, e.g., geometrical set up of the counter, we performed an extra experiment employing Sn filter at the beam exit to the experiment cage. The energy of the incident polarized beam was set to be 36 kev. The Sn filter attached to the beam exit produces fluorescent X-rays of Sn Kα (25.2 kev) and Sn Kβ (28.5 kev), which are not polarized. The 36 kev polarized X-rays, Sn Kα and Sn Kβ fluorescence detected with the Xe gas proportional counter are distinguished by looking at the pulse height of each event, and their average rise time is derived separately. As shown in Fig. 9, no significant difference is found in the rise time between Θ"0 and Θ"90 for the Sn fluorescent X-rays, while the rise time for the incident polarized beam events indicate the modulation as for Fig. 7. We thus conclude that the Θ-dependent rise time is owing to the beam polarization. Fig. 9. Average rise time against polarized beam and non-polarized beam. The left panel shows the rise time for 36 kev polarized beam at Θ"0 and 90. The center and the right panel is that for non polarized Sn Kα and Sn Kβ X-ray incidence. This result indicates that the measured rise time modulation against Θ really reflects the polarization of the incident beam.

248 K. Hayashida et al./nucl. Instr. and Meth. in Phys. Res. A 421 (1999) 241 248 5. Discussion 5.1. Is the rise-time modulation owing to the anisotropic emission of photo-electrons? Our experiment revealed that signal rise-time in a gas proportional counter surely modulates as a function of the angle between the anode direction and the electric vector direction of the incident X-rays. The rise-time is longer when these two directions are perpendicular, which is qualitatively consistent with what we expected from the concept of a rise-time polarimeter. However, some of the results could not be explained, if we consider the observed rise-time modulation to be fully due to the anisotropic photo-electron emission and the elongation directionality of the first electron cloud following it. One of the remarkable contradictions we found is the energy dependence of the rise-time modulation. Since K-edge energy of Xe is 34.6 kev, 34 kev X-rays mainly suffer L-shell photo-absorption, while 40 kev X-rays undergo K-shell photoabsorption. If 34 kev X-rays are photo-absorbed by L-shell electrons, the kinetic energy of the photo-electrons are about 28.5 29.2 kev. On the other hand, when 40 kev X-rays are absorbed by K-shell electrons, the emitted photo-electrons have kinetic energy of 5.4 kev. The size of the primary electron cloud, or equivalently range of electrons, in Xe-gas is described in 1.7(E/30 kev) mm [6]. It is 0.08 mm for 5.4 kev electrons and 1.6 mm for 29 kev electrons. Therefore, if the rise-time modulation simply reflects the elongation of this cloud size, the modulation for 40 kev incidence should be much smaller than the 34 kev case. Therefore, we have to consider other explanations at least for the observed modulation for 40 kev incidence. Further discussions, concerning K-fluorescence, Auger electrons, Thomson scattering will be given in a separate paper. 5.2. Future prospect for the rise-time polarimeter If we have two sets of gas proportional counters each having an effective area of 2000 cm and set orthogonally, 10 s exposure will enable us to detect about 1% polarization degree for 0.1 Crab sources or 10% polarization for 1 m Crab sources with 99% confidence. Although the rise-time polarimeter we examined has lower modulation contrast than either the crystal polarimeter or the Thomson polarimeter, its capability of simultaneous spectral and temporal measurement would be a great advantage for astronomical use. Acknowledgements We acknowledge the nice support by the staff in the Photon Factory, KEK, Japan. References [1] G.W. Fraser, J.E. Lees, J.F. Pearson, Nucl. Instr. and Meth. A 284 (1989) 483. [2] H. Tsunemi, K. Hayashida, K. Tamura, S. Nomoto, M. Wada, A. Hirano, E. Miyata, Nucl. Instr. and Meth. A 321 (1992) 629. [3] G.R. Riegler, G. Garmire, W. Moore, J. Stevens, Bull. Amer Phys. Soc. 15 (1970) 635. [4] P.W. Sanford, A.M. Cruise, J.L. Culhane, in: L. Gratton (Ed.), Non-solar X- and Gamma-ray astronomy, D. Reidel, Dordrecht, 1970, p. 35. [5] H. Tsunemi, S. Kitamoto, M. Manabe, S. Miyamoto, K. Yamashita, M. Nakagawa, Publ. Astron. Soc. Japan 41 (1989) 391. [6] Valkealahi et al., J. Appl. Phys. 65 (1989) 6.