Improving the calibration of inorganic scintillators comprising the Suzaku Hard X-ray Detector. Shin ya Yamada

Transcription

1 Improving the calibration of inorganic scintillators comprising the Suzaku Hard X-ray Detector Shin ya Yamada April 1, 2008

2 Abstract The Hard X-ray Detector (HXD) on board Suzaku satellite has realized high-quality observations of celestrial hard X-rays since its launch on year By combining the Si-PIN diode (PIN) detector and the GSO scintillator, it covers the wide energy range from 10 kev to 600 kev. To understand the detected signals and estimate accurately the input source spectra, detector response calibration is the key issue. The HXD have been calibrated on the ground and in orbit to a good extent, but still there remains challenging works for improving the performance. Among them, the most important issue is the fine calibration of the pulse-height to energy conversion (PEC) of the GSO signal. There is known discrepancy between the on-ground and in-orbit PEC, which was interpolated as a significant increase in the non-linearity effect below 100 kev. We conducted a consistent and thorough re-analysis of both ground and in-orbit calibration data. trying to confirm whether the PEC has really changed across the launch. Through this analysis, we re-evaluated the light output of the in-orbit calibration lines. As a result, we conclude that the in-orbit change in the PEC is primarily an artifact. We also found that the analog circuit around zero input appears to have changed to some extent. To verify an alternative possibility that some change actually occurred to the HXD hardware, we also performed laboratory experiments using flight-equivalent boards on-ground and in-orbit non-standard operation. Both results are consistent with the re-calibration.

3 Contents 1 INTRODUCTION The Hard X-ray Universe in the Suzaku Era The Aim of this Thesis THE HARD X-RAY DETECTOR (HXD) Concept of the HXD Design policy HXD-S Overall configuration Detail of well-counter units Anti-counter units High voltage supply units HXD-AE Overview ACU HXD-DE Signal processing flow of the GSO/BGO scintillator phoswich detector Scintillator crystals The PMTs and bias distribution (bleeder) circuits The dynode signal preamplifier Pulse shape discrimination (PSD) Trigger Logic Pulse height to energy conversion in GSO

4 CONTENTS Overview Differential non-linearity (DNL) Integrated non linearity (INL) Gain correction GSO Non-linearity correction RESEARCH PLAN In Orbit Performance of the HXD HXD-AE and DE HXD-S Pulse height to energy conversion Temporary modification of the pulse height to energy conversion Other issue with the GSO calibration Objective Methods RE-EXAMINATION OF CALIBRATION DATA Re-examination of Ground Calibration Data Calibration of WPU Calibration of GSO light output Charge vs. ADC channel plot Re-examination of In-orbit Calibration Data Summary of calibration lines Energy vs. ADC relation Charge vs. ADC channel relation for in-orbit data Re-examination of In-orbit Calibration Lines A Working hypothesis Light output from various reactions Re-calculating light output of the calibration lines Application of improved light output to in-orbit data Pedestal ADC channel

5 CONTENTS 3 5 LABORATORY EXPERIMENTS USING FLIGHT-EQUIVALENT HARDWARE Hypotheses Flight equivalent hardware Effects of very large pulses induced by cosmic-rays Simulation of cosmic-ray induced events Experimental setup Effects on PMT and CSA Setup for post-fm WPU Effects on electronics INL and Pedestal examination in the post-fm WPU analog circuits Objective and methods Probing the origin of the INL The waveform of the pedestal and its dependency of voltage IN-ORBIT GAIN-SHIFT OPERATION Gain reduction operation of the HXD Operation Obtained spectra and its analysis Independent method of verifying analog offset Independent calculation Idea Result SUMMARY 78 8 ACKNOWLEDGMENT 79 9 Appendix Software analysis of the GSO response function Object The flow of building the response function Plans Results and discussion

6 List of Figures 1.1 The spectra of Cygnus X-1 obtained with Suzaku (Makishima et al. 2008). The detector response is not removed. Estimated PIN and GSO background is overlaid The view of the Suzaku satellite with the side panels removed Black diagram of the HXD system Appearance of HXD-S: side view (left), top view (right) Numbering rule of the Well and Anti counter units A side view of a Well-counter unit. The signal X-ray comes from right Effective area of 16 Well counters units Schematic view of HXD-AE, seen from the connector side Circuit diagram of the Well bleeder Circuit diagram of the Well preamplifier Signal processing in the phowswich counter processing block Signals in the PSD LSI. Fast and Slow signals before and after the peak hold for GSO (left) and BGO (right) are illustrated. The outputs of Slow LD and PSD out are displayed in the bottom panels Trigger logic diagram A timing chart of the trigger logic The flow from Energy to pulse height Flow of the PEC, and the designation of pulse heights after each correction step The INL table of W31, which was measured on the ground experiments. Abscissa is the output ADC channel, and ordinate is the value which must be subtracted from the ADC channel

7 LIST OF FIGURES Digital PSD selection criteria (solid trapezoid) applied to on-board GSO data, shown on a twodimensional histogram of fast and slow-shaped pulse heights, while the hard-wired PSD function is temporarily disabled. The dash line indicates the hard-wired PSD cut. The left panel represents the full pulse-height range, while the right panel is an expanded view at lower energies Long-term variation of the pulse height of the 511 kev line in four unit on W Ratio of the background-subtracted HXD spectra of A0535 to those of the Crab Nebula. Black and red represent the PIN and GSO data, respectively Initial in-orbit calibration of the GSO energy scale. Abscissa is the input energy, while ordinate is the estimated energy derived from the GSO pulse heights using the pre-launch PEC. In addition to several background lines (red circles), the cyclotron line of A ( cyan circle), and the pedestal information at 0 kev are also shown. The dashed line represents the proportinal line, energy = estimated energy. The bottom panel represents the ratio between this proportinal line and the estimated energy The spectrum of A using the modified PEC of eq ( 3.1) Reconfirmation of the WPU pre-launch calibration using electronics test pulses, conducted by Kawaharada (2004). Abscissa is ADC channel of the pocket MCA, while ordinate is SLOW-PHA of W Ground calibration of GSO light outputs taken from Kitaguchi (2006). Abscissa is the energy of various calibration isotopes, while ordinate is PMCA channels. The dashed green line represent eq (4.1), while the blue curve eq (4.2), The bottom panel shows the data ratio to the dashed green line The energy vs. PHA SLOW relation of ground calibration data of GSO taken from Kawaharada (2004). Abscissa is energy, while ordinate is PMCA channels. The energies of the utilized isotopes of GSO are also indicated. The bottom panel shows the ratio of the data to the linear function obtained by fitting the data above 100 kev A charge vs. ADC channel plot of the GSO calibration data taken in 2004 June (Kawaharada 2004). Abscissa is charge, converted from the energy using the inverse ralation of eq. (4.2). Ordinate is the same WPU output channel as in Fig 4.1, but the WPU offset by +33 ch has been subtracted. The bottom panel represents the ratio between data and the linear function (dashed green line) pointing from the origin to the 511 kev data point A PHA SLOW spectrum of one well unit acquired from a black sky. The higher data points represent those events which have simultaneous hits with surrounding units

8 LIST OF FIGURES The decay scheme of 153 Eu from 153 Gd The decay scheme of 151 Eu from 151 Gd The energy vs. PHASLOW relation of the background lines observed in the blank-sky spectrum of Fig Abscissa is energy, while ordinate is PHA SLOW. The doted curve indicates the modified PEC relation by Kokubun et al. (2007) Relation between the charge from charge and ADC channel. (Top) Red circles represent he same in-orbit data points as plotted in Fig 4.8, while green triangles are the same pre-launch data as in Fig 4.4. Then relative gains are adjusted using the in-orbit data as a reference. The dashed red line is a proportionality line connection to the 511 kev point. and the dotted magenta curve represents the modified PEC. (Bottom) Ratios of the in-orbit (red circles) and pre-launch (green triangle) data to the common line of proportionality. Magenta circles are ratios of the in-orbit data to the modified PEC The relation between energy gpt various interactions and light output per energy. Abscissa is energy, while the ordinate is light output per energy. L γ (E)/E (red), L e (E)/E (blue), L ek (E)/E (cyan), L el (E)/E(magenta), and L em (E)/E(green) are shown The same as Fig. 4,8, but the charge of the in-orbit lines at 70, 150, and 196 kev are re-calculated ( 4.3.3) SLOW-PHA histogram of events triggered by PSEUD. Blue and green are pre-launch data taken in laboratory and on the spacecraft, respectively. Red and Magenta are in-orbit data acquired during on and off periods of the high voltage The same as Fig 4.10, but the WPU offset for the in-orbit data is changed from the original 33 ch to 25 ch as indicated by the pedestal channel measurements The analog circuit and block diagram of LED The configuration of tests for PMT and CSA is shown Absorption coefficient per cm for typical shielding matter as a function of energy. The dashed curve and dotted curve represent the absorption coefficient of photoelectric absorption and that including all interaction, respectively. Al (black), Cu (cyan), Sn (green), W (yellow), and Pb (red) are shown GSO Spectrum obtained under the setup shown in Fig. 5.2, when irradiating Cd (blue) and Eu (green) and both isotope(red)

9 LIST OF FIGURES A GSO spectrum obtained by the setup in Fig The spectrum before LED radiaton (black), that during LED radiation (red), and that after LED radiation (green) are shown. The left two figures are the spectrum before gain correction, whille the right two are after gain correction The setup of performance test on post-fm WPU under large signals The result of PHA SLOW vs. input test pulse voltage. The top panel represents PHA SLOW ; before LED radiation (black), during it (red), and after it(green). The bottom panel shows the PHA SLOW measured during or after LED radiation subtracted with the PHA SLOW before LED radiation SLOW-PHA and the voltage of peak hold before ADC is shown against input test pulse voltage The oscilloscope waveform of the pedestal. The peak-hold output (cyan), peak hold gate signal (red), sample hold signal (green), and PIN trigger signal (yellow) are shown The oscilloscope waveform of the pedestal. The peak-hold output when power supply is set at 6V (purple), and one at 5.8 V (cyan) The spectra observed in the HV reduction. The red spectrum obtained by gain operation is shown; one is normally selected (red), and the other is coincident event(blue), The normal background spectrum acquired in June 2007 (green) and that selected by coincidence (magenta) is shown for comparison, and in addition, those multiplied to fit the 511 kev line with those obtained by gain operation are also shown The same as Fig. 4.12, but the data obtained by the gain reduction operation (blue rectangle) is added, with a gain factor of relation relative to the nominal in-orbit spectrum(red). The WPU offset is assumed to be 33 ch for the pre-launch data, while 25 ch for the in-orbit data SimHXD flowchart diagram of step SimHXD flowchart diagram of step SimHXD flowchart diagram of step

10 List of Tables 2.1 Electric power consumption of HXD-AE Properties of the GSO and BGO scintillators Counting rate of PIN and GSO at each stage of event selection Activation lines in the GSO spectrum The basic parameters of Eu Energy and internal conversion coefficient on Eu Summary of the light output calculation Parameters of the HV reduction operation photon, EVS value map

11 Chapter 1 INTRODUCTION 1.1 The Hard X-ray Universe in the Suzaku Era The fifth Japanese X-ray satellite, Suzaku (Mitsuda et al. 2007), was launched on 2005 July 10. Utilizing the X-ray CCD camera (XIS) and the Hard X-ray Detector (HXD), we have achieved unprecedented high sensitivity over a wide energy range from 0.2 kev to 600 kev. This satellite has observed more than 700 objects so far, and unveiled new aspects of the high energy universe; these include a more accurate measurement of relativistic effects around super massive black holes, a discovery of highly obscured Active Galactic Nuclei, and reliable evidence of contribution of supernova remnants to the production of cosmic rays. In many of these findings, hard X-ray information has played an important role. Figure 1.1 shows the spectrum of the most famous black hole binary, Cygnus X-1, observed by Suzaku. Thus, an accreting black hole radiates soft X-rays ( 1 kev) from its accretion disk via thermal process, and hard X-rays ( 100 kev) from high temperature corona via thermal Comptonization process. Therefore, the wide band Suzaku spectra covering three orders of magnitude provide a powerful diagnostic tool. Indeed, the Suzaku data have given an important clue as to the 30-year unsolved mystery of the rapid X-ray variability, which is seen only in black hole binaries at low accretion rate. 9

12 CHAPTER 1. INTRODUCTION 10 Counts/sec/keV 10 XIS HXD-PIN HXD-GSO GSO_bkgd PIN_bkgd 5% 3% Energy (kev) Figure 1.1: The spectra of Cygnus X-1 obtained with Suzaku (Makishima et al. 2008). The detector response is not removed. Estimated PIN and GSO background is overlaid. 1.2 The Aim of this Thesis The HXD (Takahashi et al. 2007; Kokubun et al. 2007) is a novel cosmic hard X-ray experiment, developed under a collaboration of the University of Tokyo, JAXA, Hiroshima University, Saitama University, Stanford University, RIKEN, Osaka University, Aoyama Gakuin University, Kanazawa University, and a few other domestic groups. Figure 1.2 shows the Suzaku satellite, where the HXD sensor is located on the spacecraft base panel. The HXD is a hybrid detector, utilizing silicon PIN diodes (HXD-PIN) for lower energies (10-70 kev), and GSO scintillators (HXD-GSO) placed behind the PIN diodes for higher energies ( kev). The HXD-GSO energy range is extremely important, because it addresses such issues as the physics of accreting black holes, and formation of electron-positron annihilation lines. In spite of its importance, the performance of HXD-GSO is significantly more complicated that of HXD-PIN. This is partly because the X-ray interaction above 100 kev is dominated by Compton scattering, and partially because the response of its components (scintillators, phototubes etc.) is subject to various changes. Actually there has been a suspect that the energy vs. pulse height relation of HXD-GSO has changed in low energies ( 70 kev) across the satellite launch, and its cause is not yet understood well. Given these, to improve the lower energy response of HXD-GSO is the aim of the present thesis. We first describe the HXD in 2, and the objectives and methods of the research in 3. In 4, we

13 CHAPTER 1. INTRODUCTION 11 re-analyze pre-launch calibration data and actual in-orbit data, focusing on the suggested changes of the low-energy GSO response. In 5 we conduct follow-up laboratory experiments trying to get deeper understanding of the issue, followed by a non-standard operation of the HXD for a diagnostic purpose described in 6. The results are discussed in 7, and concluded in 8. Figure 1.2: The view of the Suzaku satellite with the side panels removed.

14 Chapter 2 THE HARD X-RAY DETECTOR (HXD) 2.1 Concept of the HXD Design policy The HXD consists of five components: Suzaku HXD Block Diagram Overall Structure HXD-S HXD-AE Well Counter 16units Signal WPU Command Data Control DP Anti Counter 20units Signal TPU Command Data HXD-DE Data HV T Sensor Control Monitor ACU Command Time Latch AE,HV-Status Data Command Bi-level Status HXD-PIM HK HXD-PSU Power Line Figure 2.1: Black diagram of the HXD system. HXD-S (Sensor) Detection part with front end electronic (pre-amplifier etc) and high-voltage bias generator. 12

15 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 13 HXD-AE (Analog Electronics) An box which performs analog waveform shaping analog-to-digital conversion (ADC) and simple hard ware screening of signals from HXD-S. HXD-DE (Digital Electronics) A signal processing part using CPUs. Receiving digital data packets from HXD-AE, performing software processing such as data selections and sending the final screened packets to satellite data processing unit. HXD-PSU (Power Supply Unit) Convert the satellite power bus voltage (40-50 V) to a regulated power lines for HXD-DE, HXD-AE, and HXD-S. HXD-PIM (Peripheral Interface Module). Command (and HK-data) handling interface module. The detail of each component is described in the following section. In 2.1.1, we show the block diagram of the HXD system. 2.2 HXD-S Overall configuration The HXD-S is mainly consists of 16 HXD-Well counter units and 20 HXD-ANTI units, arranged in 6x6 compound-eye configuration. A well unit consists of 4 PIN Si detectors and 4 GSO scintillator, surrounded by large well-shaped BGO scintillator acting as an active shield. The GSO, is placed beneath the PIN Si diodes, and photons with energy below 50 kev is mainly absorbed by the PIN detector; while those with energy above 50 kev are absorbed by the GSOs. The orientation and unit IDs are presented in 2.3. The 16 HXD-Wells are surrounded by 20 HXD-ANTI units, made of 4-cm BGO scintillators. HXD-ANTI units are also used to detect gamma-ray burst, transient X-ray burst, and solar flares. HXD-S is operated at temperatures between 12 and 20 degree. Using the thermometers installed in the HXD-S, we monitor the temperature variance.

16 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 14 Figure 2.2: Appearance of HXD-S: side view (left), top view (right). Configuration of Sensor Units (Top View) T00 T01 T02 T03 T04 T10 T34 W00 W01 W10 W11 T11 Y T33 W03 W02 W13 W12 T12 X T32 W30 W31 W20 W21 T13 T31 T30 W33 W32 W23 W22 T14 P0 P1 P3 P2 T24 T23 T22 T21 T20 PIN in Well 1 unit Figure 2.3: Numbering rule of the Well and Anti counter units

17 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) Detail of well-counter units Figure 2.4 shows the details of the well unit. The main detection part (PINs and GSOs) are surrounded by BGO crystals with 4 well-type holes, restricting the field of view to 4 deg x 4 deg. In addition, a fine collimator made of Phospler-Bronz sheet is installed in each holes (wells) restricting the FOV to 35 x 35 below 100 kev. Combining 64 Si-PIN diodes and 64 GSO crystals, HXD-S obtains an effective area of 150 cm 2 at 15 kev and 260 cm 2 at 150 kev, respectively. Calculated effective area is presented in Figure 2.5. In a Well unit, four PIN detectors are read by 4 independent charge sensitive amplifiers mentioned below. A signal from GSO crystals and BGO shield is read by a single photo-multiplier tube (PMT), coupled with a bias bleeder specially designed to handle high counting rate of 1 khz and very large signal, as large as 1 GeV, the BGO shield detector. The anode signals are directly fed into HXD-AE, while the final dynode signal goes through the preamplifier mounted below the PMT, and then into HXD-AE. Signals for the GSO and the BGO separated afterward in the HXD-AE Plastic ring Groove for cables 5 2 Thickness BGO well Fine collimater of BGO 3.0 width 25.5 Reny Cap1 Mu-metal with aluminum ring Reny Skirt GSO 24x24x5 PIN diode assembly BGO Figure 2.4: A side view of a Well-counter unit. The signal X-ray comes from right Anti-counter units Anti-counter units are utilized to provide their passive and active shield for the HXD Wellcounters. Due to the cosmic ray bonberbent and cosmic X-rays from all around, each unit shows a high counting rate of 1 khz. BGO signal is read by a PMT and goes through a preamplifier. A total of 5 counters are fed into a single board in HXD-AE. While each counter provides hit timing signals for Well-units, analogy signals from 4 counters (excluding that on the corner) are summed and thus, spectral information is obtained. The main detector is BGO scintillators. HXD-Anti consists of 20 Anti-counters units. Five units have one common board, the rule of which is the same as in the case of HXD-Well. It also has detected more than 100 Gamma ray bursts, transient

18 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 16 Figure 2.5: Effective area of 16 Well counters units outbursts, and solar flares. In this context, HXD-Anti is also called HXD-WAM. The transient data is archived to all over the world via the internet as early as possible High voltage supply units A total of 12 high voltage supply boxes are mounted on HXD-S, four each for the PIN detectors, PMTs of well counters, the HV-PIN, and those of anti-counters. The bias voltage is controlled via operation command and taken any value below its highest limit, which is 1200V for the PMT-HVs and 600V for the PIN-HVs. In normal operation, PMT-HVs are operated at 900V, and set at 0 V at several times in a day to secure the PMT from high partial background regions. PIN-HVs are nominally operated at 500 V, but noise increase in a few PIN causes 2 out of 4 HVs to be lowered to 400V. See Kaukauna et al. for details. 2.3 HXD-AE Overview The HXD-AE is composed of four Well-type Detector Processing Unit (WPU), four Transient detector Processing Unit (TPU), and an Analog-electronics Control Unit (ACU).

19 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 17 The board s orientation is shown in Figure 2.6. HXD-AE is designed to operate in lower power, as shown in Table 2.1. A WPU handles signals from 4 Well-counters, and generates and sends the data packed to HXD-DE event-by-event. A WPU sends up to 1 khz photon to HXD-DE. Counter values from lower and upper discriminator and other are also accumulated each second. It also produces hit-timing signals used for anti-coinsidence of separate counters. Processing chain of WPU is shown in detail in the next sub-section. As presented in 2.2.3, a TPU handles 5 Anticounter units. It distributes hit timing signals to all WPU and also acquires 1 second accumulated spectra covering 60 kev up to 5 MeV. Detail is described in Yamaoka et al. in preparation. HXD-AE TPU0 TPU0 DE TPU1 DE TPU1 WPU0 W02 W00 W03 W01 DE W12 W10 W13 W11 DE WPU1 TPU3 TPU3 DE TPU2 DE TPU2 WPU3 W32 W30 W33 W31 DE W22 W20 W23 W21 DE WPU2 ACU HV,TempMPX PSU DE from connector side Figure 2.6: Schematic view of HXD-AE, seen from the connector side Table 2.1: Electric power consumption of HXD-AE Voltage +5VD +5VA -5V +12V -12V ACU (1 module) 30 (ma) WPU (4 module) TPU (4 module) ACU The final board ACU mainly controls the power lines of individual WPU and TPU board, as well as the HVs. It also monitors a total of thermometers in HXD-S. Level controls of HVs are also

20 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 18 handled. 2.4 HXD-DE HXD-DE composes of Interfaces (I/Fs) to the nine HXD-AE boards, HXD-PIM, satellite data processor (DP), and two redundant identical CPU boards. On a space qualified CPU operated in 12 MHz clock, data handing and command controlling software is running. A specific faction made by astrophysical scientists (PI-program) is also running, and used to further screen the photon event data. Filtering detail is presented in Kokubun et al. HXD-DE can send up to 1 khz event data to the DP. However, in daily operation, allowed band-width is further restricted typically to 1 khz. Thus, onboard data screening of the PI program is playing a key role. It processes the data sent from HXD-AE and then sends it to Data Processing Part of HXD (hereafter DP), or decodes the command sent from DP and then sends it to HXD-AE. 2.5 Signal processing flow of the GSO/BGO scintillator phoswich detector Scintillator crystals Table 2.2: Properties of the GSO and BGO scintillators GSO(Ce): Gd 2 SiO 5 (Ce) BGO: Bi 4 Ge 3 O 12 effective atomic number density (g/cm 3 ) light yield ( photon/mev ) effective decay time (ns) ;20 degree effective decay time (ns) ;-20 degree Basic characteristics of the GSO and BGO crystals are presented in table 2.2. Both crystals have relatively high density and atomic number and hence higher photon cross section to gamma-rays, as compared to classical inorganic scintillators used in orbit, such as NaI(Tl) and CsI(Tl). Since

21 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 19 the GSO is relatively new scintillator, we checked its basic performance, such as input energy to light yield relation, by ourselves in the HXD development (e.g. Uchiyama et al. IEEE TNS 2001). Effective decay time of scintillation light of the GSO and the BGO are relatively short, 100 ns and 500 ns, respectively. In a phoswich configuration, we utilize this difference to distinguish the signals from each crystal. Details are described later in this section. As already presented, GSOs and BGO are optically coupled together. The 4 GSO blocks are glued on a BGO block, called the bottom part, using a silicon glue. A well-shaped active collimator part made of BGO (BGO well part) is also glued on the bottom part. The BGO bottom part is coupled to the PMT (R ). Thus, both the scintillation light from the GSOs and the BGO well part are accumulated though the BGO bottom part and their light-yield is almost halved. In a typical operation condition, the relative light-yield detected by the PMT is 1.0, 0.5 and 0.25, for the GSOs, BGO bottom part. In this process, and the BGO well part, respectively. The HXD-Well counter is characterized by the large volume of the BGO. Thus the counting rate in orbit is as high as 1 khz because of the cosmic X-ray signals, direct hit of high-energy particles, and signals from the activated radio isotopes synthesized by them. In addition, a single hit from very high energy ions can deposit large amount of energy and causes saturation of signal processing circuits The PMTs and bias distribution (bleeder) circuits Photomultiplier tubes must satisfy general requirements; a high energy resolution, a large dynamic range, and a low dark current. In addition, for the use on spacecraft, it must be quakeresistant and have prolonged stability of gain and resolution under the large signals for more than five years. Considering performance assessment study of PMT [5][3], 2-inch PMT, Hamamatsu R , for Well and 1 1/8 inch PMT, Hamamatsu R MOD for ANTI have been selected [5]. Hereafter, we focus on the Well units and their PMTs. In the case of Well units, anode signal is used for trigger, while dynode signal is for spectroscopy. In orbit, the voltage of PMT is off to avoid high counting rate when the satellite is passing through South Atlantic Anomary, in which the rate of secondary cosmic ray becomes two orders of magnitude higher than in the other place, The gain of these PMTs are proprotional to V 7. The bleeder circuit is shown in Fig There are three characteristic as follows. One is small

22 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 20 resistor (1M) for increase of bleeder current so as to stabilize PMT gain. The Zener diodes is implemented in order to stabilize the last inter-stage voltage, with the capacitor (0.01 µ F) and a series resister (100 kω) suppressing the noise from the Zener (that is, RC filter). The clump diodes are installed in both dynode and anode readouts in order to release excess charge into the bleeder ladder, when the very large sinal is detected. Note that we also use final dynode signals for spectroscopy. Figure 2.7: Circuit diagram of the Well bleeder The dynode signal preamplifier The dynode signal is fed into a preamplifier. It is optimized to discriminate the GSO ( 100 ns) and the BGO ( 700 ns) signal and recover fast enough not to increase dead time against large signals at the rate of 1 khz. The circuit is shown in Fig 2.8. This charge-sensitive preamplifier (CSA) receives the dynode charge signal of the PMT. Its time constant is set at 4 µ s using a feedback capacitor of 100 pf and a feedback resistor of 39 kω. The linearity is within 1 % (Tanihata (1998)).

23 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 21 Figure 2.8: Circuit diagram of the Well preamplifier Pulse shape discrimination (PSD) The Anode signal are used to produce Anode-LD (lower-discriminator) trigger and Anode-UD (upper-discriminator) flag. Anode-LD is used to initiate the data aquisition in the WPU. The Anode-UD flag is used to veto high energy deposit event like cosmic ray. The energy level of Anode LD is typically set at 30 kev for the GSO, while that of Anode UD is typically 1 MeV. Because of the smaller light yeild and longer decay constant, the Anode-LD value is 200 kev for the BGO signals. The Dynode signals from the PMT, after being amplified by the Well preamplifier, is fed into the analog circuits for pulse shape descrimination. For the pulse shape discrimination, we applied two different shaping time to the preamplifier, with τ 150 ns optimized for only GSO and another with τ 1000 ns accumulation before the GSO and the BGO signals. The difference of signal amplification between the GSO and BSO signal is shown in Fig The Fast and Slow-shaped pulses are fed into peak-hold circuit and then into ADC. Here, we call the Fast pulse height PHA-FAST, and Slow pulse height SLOW-PHA. By comparing the two signals, the PSDOUT flag is initiated. By the flag, we can distinguish between the GSO and the BGO signals. For the detection of large signal, SUD (super-upper-discriminator)-trailing trigger stands when pulse height of large signal is overshot.

24 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 22 Figure 2.9: Signal processing in the phowswich counter processing block. Figure 2.10: Signals in the PSD LSI. Fast and Slow signals before and after the peak hold for GSO (left) and BGO (right) are illustrated. The outputs of Slow LD and PSD out are displayed in the bottom panels.

25 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) Trigger Logic The steps in trigger process is executed in the following order: 1. Wait trigger signal: ANODE LD, PIN LD, SUD, and PIN LD 2. Accept an inhibit signal 3. Judgment of trigger accept or reject. If reject, go to A/D conversions of Slow, Fast, PIN signals, and transmission of an event data to FIFO. 5. Cancellation of the inhibit signal after ANALOG BUSY signals become inactive. 6. Return to 1 We show the trigger logic timing chart Fig When a trigger is activated, the subsequent triggers are inhibited by the time when the event sequence ends, And a peak-hold gate is created immediately. Even for a simultaneous event, the fastest one starts the sequence. Timing chart of the trigger logic is shown in Fig Pulse height to energy conversion in GSO Overview In any radiation detector, the relation between the incident radiation energy and the output pulse height is an important element of its performance. Since this is an essenctial issue in the presenct thesis, we describe, in some detail, this relation in HXD-GSO referring to Once a photon is photo-absorbed in the GSO scintillator, a photo electron is created, which immediately produce secondary electrons typically per 10 ev of the incident radiation. These electrons deposit their energies in the scintillator, creating optical photons of which the number is typically given as N 0 1 (E/MeV), where E is the incident X-ray photon energy. These optical photons enter the photocathode of the photomultiplier, and get multiplied into electrons of which the number is given as N 1 N 0 η10 6 (V/1 kv) 7, where η is quantum efficiency of the photocathode and V means high voltage of the photo-multiplier. The output charge is converted into a voltage pulse via the capacitor of CSA. Finally, in WPU, the pulse is further amplified, then

26 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 24 Figure 2.11: Trigger logic diagram.

27 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 25 Figure 2.12: A timing chart of the trigger logic

28 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 26 peak-holded in PSD-LSI, and digitalized by ADC. This digitalized raw pulse height is called PHA-SLOW(FAST). Through this process, the energy of incident X-ray is converted to the pulse height. Since the CSA utilizes a capacitor of C = 100 pf, we obtain a typial pulse height of en 1 /C = η(v/1kv) 7 (E/MeV) 1 volts for a 1 MeV photo. Thus, the digitized pulse height we obtain is approximately proportioanl to the incident photon energy, as long as the photon deposits its full energy. However, this strict proportionality is broken by many factors, including the non-linearity in the GSO light yeild, that in the phototube multiplication, zero-point offsets in electronics, differential and integrated bib-linearities is the electronics, and so on. Since what we can measure is a pulse height of each event, we have to convert it back to the incidenct energy through a reverse process, with knowledge on all these non-ideal effects. Here, let us call all this reverse process Pulse-height to Energy Conversion, or PEC. Figure 2.13: The flow from Energy to pulse height. The steps of the PEC is executed in the following order: correction for differential non-linearity (DNL), that for integrated non-linearity (INL) of electronics, phototube gain correction, and GSO non-linearity correction. We emphasize on the difference between the non-linearity correction of analog circuit and that of scintillators: the former must be considered in voltage space, while the latter in energy. Accordingly, in reference to the flow chart of Fig. 2.14, we define five pulse height: PHA-SLOW(FAST), ADCDNL-PI-SLOW(FAST),ADCINL-PI-SLOW(FAST), which is equivalent to charge output from the phototube, GAIN-PI-SLOW(FAST), which corresponds to the GSO light output, and PI-SLOW(FAST). Thus, the quantity PI-SLOW is our best estimate of the incident X-ray energy, and should coincide with the energy within the detection energy resolution (statistical effect) and calibration uncertainties in the PEC (systematic effects). Below, we describe each step of the PEC.

29 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 27 PHA-SLOW(FAST) DNL Corrected ADCDNL-PI-SLOW(FAST) INL Corrected ADCINL-PI-SLOW(FAST) charge Gain Corrected GAIN-PI-SLOW(FAST) light output - GSO Non-Linearity Corrected PI-SLOW(FAST) Figure 2.14: Flow of the PEC, and the designation of pulse heights after each correction step Differential non-linearity (DNL) As already confirmed in the ground experiment, the 12-bit A/D converter (ADC) in WPU, ADC 12062, has relatively large differential non-linearity (DNL). The DNL in the present case appears as excess/deficient counts (up to 50%) in particular ADC channels, which recur every 64 channels. The cause has been unknown, though unstable reference voltage seems likely. This DNL effect is roughly reproducible, and has been tabulated as variations of pulse-height width from ADC channel to channel. Over every 64-channel interval, we adjust the center of each ADC channel so that its effective width is appropriately represented. Through this correction, the original integer ADC channel turns into a real floating number, called ADCDNL-PI-SLOW (FAST) Integrated non linearity (INL) Even correcting for the DNL, the ADC output is not generally perfectly proportional to the input voltage; this is called Integrated non linearity (INL). This effect is often prominent at the lowest input voltage. Therefore, the next step is the correction for this INL. In the HXD case, the INL causes typically 10 % higher channels especially below 100 ch than expected when the ADC performance is ideal. It is confirmed that this deviation does not depend on the analog gain [4]. The INL effect of each ADC channel was measured also in the ground experiments, and tab-

30 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) 28 AE_PI_FAST:ADC_PI_FAST {UNIT_ID==13} AE_PI_FAST ADC_PI_FAST Figure 2.15: The INL table of W31, which was measured on the ground experiments. Abscissa is the output ADC channel, and ordinate is the value which must be subtracted from the ADC channel. ulated as exemplified in Figure By subtracting this deviation from ADCDNL-PISLOW (FAST), we obtain ADCINL-PISLOW(FAST), which is considered to represent the charge output from the phototube; this is the second step Gain correction Although the quantity ADCILN-PI-SLOW (FAST) has been made roughly proportional to the incident energy, the coefficient of proportionality is rather arbitrary, depending on the GSO light yield, the CSA capacitance, the PMT gain, and electronics settings. In particular, the first two factors are temperature sensitive. Therefore, this coefficient, or simply gain, must be determined with a great care, and as frequently as possible. The gain can be determined if we can find at least two (hopefully more) spectral line of which the energy is well known. The reason why we need at least two calibration points is that we generally allow to vary both the slope and intercept of the energy vs. output voltage relation. In other words, we assume a linear relation, instead of an exact proportionality, between the two quantities. This procedure is applied separately to all the Well units, and the obtained result is named GAIN-PI-SLOW (FAST).

31 CHAPTER 2. THE HARD X-RAY DETECTOR (HXD) GSO Non-linearity correction The last step is to correct GAIN-PI-SLOW (FAST) for non-linearity of the GSO scintillator. Since the K shell energy of Gd is present at 50 kev, L shell effects are not negligible below 100 kev, which causes non-linearity [2]. This effect was measured under the condition where INL is negligible [4]. According to [4], the input energy E and the light output E LO of W31 is described as E = E LO [1 exp( E LO 0.534)]. (2.1) By equating Gain-PI-SLOW (FAST) with E LO, we finally obtain via this relation our best estimate on E, to be denoted as PI-Slow (FAST). Here PI stands for pulse invariant. The parameters of this equation were evaluated both for Fast and Slow pulse height of each Unit.

32 Chapter 3 RESEARCH PLAN 3.1 In Orbit Performance of the HXD This Chapter describes the basic plan of the present thesis. Since our objective is to improve the HXD-GSO calibration in low energies, we start with a brief description in 3.1 of the in-orbit performance of the HXD. Then in 3.2, we introduce one particular issue, namely unexpected changes in the pulse-height to energy conversion in HXD-GSO. Our methods to tackle the issue are presented in HXD-AE and DE Since the launch of Suzaku, the performance of HXD-AE and HXD-DE was verified relatively soon ( 2 months). As an overall measure of the AE/DE performance, we summarize in Fig. 3.1 counting rates of PIN and GSO at each stage of event selection. After the analog event selection using UD veto and analog PSD cut ( 2.5.5), the total event rate summed over the four WPU modules typically reached a few kilo count/s, which is significantly higher than the nominal telemetry limit ( 300 count/s). However, as designed, an efficient reduction in the event rate has been achieved by the onboard software in HXD-DE [1] in the following manner. Figure 3.1 shows the software DE cut, namely a digital PSD cut, which we applied to the GSO events from HXD-AE. In this way, we effectively lowered the UD (left panel), and excluded low pulse-height BGO events that survived the hardwired PSD cut in HXD-AE. As a result, the average GSO count rate was dramatically reduced to 5 10 counts/s. 30

33 CHAPTER 3. RESEARCH PLAN 31 As to the HXD-PIN signal, we employed the DE software to remove common-mode noise, produced by electric inference in the spacecraft. That is, any PIN-triggered event was removed if it has a simultaneous hit in at least one of the other 3 PIN detectors in the same Well unit. This has successfully reduced the PIN count rate to Hz. (Table 3.1) Slow Fast Figure 3.1: Digital PSD selection criteria (solid trapezoid) applied to on-board GSO data, shown on a twodimensional histogram of fast and slow-shaped pulse heights, while the hard-wired PSD function is temporarily disabled. The dash line indicates the hard-wired PSD cut. The left panel represents the full pulse-height range, while the right panel is an expanded view at lower energies HXD-S After verifying the basic performance of HXD-AE and HXD-DE, the performance of HXD-S was examined using various test observations. Since the energy resolution of GSO remained the same as that measured on ground, we became confident that the launch vibration did not cause any crack or mechanical damages to the GSO/PMT assembly. The gain of PMT, in contrast, has been changing since the launch mainly due to two causes. One is short-term ( hour) gain fluctuations due to temperature changes (by several degrees), because the GSO light yield and the PMT gain both depend inversely on the temperature, with typical coefficients of 0.5%C 1 and 0.2%C 1, respectively. The other is long-term gain degradation of the PMT; over half a year since the launch, all the Well units have exhibited gradual gain decreases, by 5% (minimum) to 20 % (maximum).

34 CHAPTER 3. RESEARCH PLAN 32 Table 3.1: Counting rate of PIN and GSO at each stage of event selection. Selection pattern PIN (count/unit/s) GSO (count/unit/sec) HXD-AE Analog LD rate Analog UD rate After analog PSD cut < 100 HXD-DE After PIN UD cut After PIN trigger cut 1-10 After digital LD After digital PSD 5-10 Ground analysis After anti-coincidence applied These PMT gain variations, however, are what we have already expected, and can be corrected by referring to various background lines ( 2.6.4) seen in kev of the GSO spectrum Pulse height to energy conversion After the confirmation of basic operation of HXD-GSO, more detailed calibrations of its performance were carried out by Kitaguchi (2006) and Kokubun et al. (2007). One important item here is the pulse-height to energy conversion (PEC), described in 2.6. Using the background lines, it was then confirmed that the PEC determined in pre-launch measurements remains correct in energies above 70 kev, when we apply appropriate gain corrections to the data. However, this was not clear in lower energies, since there is no appropriate calibration lines in this energy range. In order to calibrate the PEC in energies below 70 kev (down to 35 kev), a celestial object named A was utilized (Terada et al 2006), because it is known to exhibit cyclotron resonance absorption at 50 kev of its spectrum. Therefore, this object was observed on 2005 September 14. The PIN and GSO spectra obtained from A , after subtracting background, are shown in Fig.3.3. Here, we converted the GSO pulse heights back to energy, using the PEC relation determined in pre-launch measurements ( 2.6). In order to appropriately remove the detector response, we divided the spectra of A by those of the Crab Nebula, which is known to

35 CHAPTER 3. RESEARCH PLAN 33 Figure 3.2: Long-term variation of the pulse height of the 511 kev line in four unit on W1. exhibit a featureless power-law spectrum with photon index of 2.1. A hollow is seen at about 50 kev in the PIN spectrum, which must be the cyclotron absorption feature. On the other hand, the same feature is seen at about 60 kev in the GSO spectrum, thus implying a 20% discrepancy between the PECs of PIN and GSO. The energy calibration of the PIN detector is more reliable than that of GSO, because its PEC is mostly determined by the preamplifier, and is hence much simpler than that of GSO. In addition, it was confirmed that the Gd escape peak, expected at 43 kev, correctly appeared at 43 kev within 1 kev in the in-orbit background spectrum of PIN. Therefore, we have come to conclude that the PEC of GSO has changed across the launch Temporary modification of the pulse height to energy conversion To confirm further the discrepancy suggested by A data, we used GSO background lines, seen at 70 kev, 150 kev, and 350 kev. Figure 3.4 shows these lines, together with other higherenergy lines and the A cyclotron feature, on the plane of incident energy vs. estimated energy. Here, the estimated energy means PI-SLOW (FAST) in Figure 2.6.2, namely our best estimate of the incident energy based on the PEC procedure determined in the pre-launch calibration. Thus, the data points exhibit systematic deviations, toward lower energies, from a strict proportionality. We decided to replace equation (2.1) with a more suitable one. Since we do not have

36 CHAPTER 3. RESEARCH PLAN 34 Figure 3.3: Ratio of the background-subtracted HXD spectra of A0535 to those of the Crab Nebula. Black and red represent the PIN and GSO data, respectively. suitable calibration lines below 50 kev, we temporarily regarded the pulse height when no signal is in GSO, hereafter pedestal, as 0 kev. As a result, instead of equaiton (2,1), the data points in figure 3.4 have been explained by assuming a relation between true energy E and the modified energy E MOD as, E = E MOD 55.2 exp( E MOD /38.0 kev), (3.1) where the parameters is for slow-shaped pulse height in W31. These parameters have been determined for fast and slow shaped pulse height of all units. Figure 3.5 shows the same spectrum of A as in figure 3.5, but with the GSO pulse height converted to energy using the modified PEC of eq (3.1). Thus, the cyclotron lines appear to better agree between PIN and GSO Other issue with the GSO calibration In addition to the change in the PEC, the in-orbit GSO data showed several unexpected properties, which cannot be explained by the pre-launch calibration. These are listed below: 1. The GSO Branch in the two dimensional spectrum below 100 kev became wider than seen in grand experiments.