Astroparticle Physics 30 (2008) 72–84

Contents lists available at ScienceDirect

Astroparticle Physics

journal homepage: www.elsevier .com/locate /astropart

PoGOLite – A high sensitivity balloon-borne soft gamma-ray polarimeter

Tuneyoshi Kamae a,*, Viktor Andersson a,b, Makoto Arimoto c, Magnus Axelsson d, Cecilia Marini Bettolo b,Claes-Ingvar Björnsson d, Gilles Bogaert e, Per Carlson b, William Craig a,1, Tomas Ekeberg a,b,Olle Engdegård b, Yasushi Fukazawa f, Shuichi Gunji g, Linnea Hjalmarsdotter d, Bianca Iwan a,b,Yoshikazu Kanai c, Jun Kataoka c, Nobuyuki Kawai c, Jaroslav Kazejev b, Mózsi Kiss a,b, Wlodzimierz Klamra b,Stefan Larsson b,d, Grzegorz Madejski a, Tsunefumi Mizuno f, Johnny Ng a, Mark Pearce b, Felix Ryde b,Markus Suhonen a,b, Hiroyasu Tajima a, Hiromitsu Takahashi f, Tadayuki Takahashi h, Takuya Tanaka f,Timothy Thurston i, Masaru Ueno c, Gary Varner j, Kazuhide Yamamoto f, Yuichiro Yamashita g,Tomi Ylinen a,b, Hiroaki Yoshida f

a Stanford Linear Accelerator Center (SLAC) and Kavli Institute for Particle Astrophysics and Cosmology (KIPAC), Menlo Park, CA 94025, USAb Royal Institute of Technology, Physics Department, SE-106 91 Stockholm, Swedenc Tokyo Institute of Technology, Physics Department, Meguro-ku, Tokyo 152-8550, Japand Stockholm University, Astronomy Department, SE–106 91 Stockholm, Swedene Ecole Polytechnique, Laboratoire Leprince-Rinquet, 91128 Palaiseau Cedex, Francef Hiroshima University, Physics Department, Higashi-Hiroshima 739-8526, Japang Yamagata University, Physics Department, Yamagata 990-8560, Japanh JAXA, Institute of Space and Astronautical Science, Sagamihara 229-8510, Japani The Thurston Co., 11336 30th Avenue, NE, Seattle, WA 98125, USAj University of Hawaii, Department of Physics and Astronomy, Honolulu, Hawaii 96822, USA

a r t i c l e i n f o

Article history:Received 14 April 2008Received in revised form 8 July 2008Accepted 8 July 2008Available online 25 July 2008

PACS:95.30.Gv95.55.�n95.55.Ka95.55.Qf95.75.Hi95.85.Pw97.60.Gv97.60.Lf97.80.Jp98.54.Cm

Keywords:Instrumentation, detectorsTechniques, polarimetricPulsars, generalX-ray, binariesStars, neutronGalaxies, active

0927-6505/$ - see front matter � 2008 Elsevier B.V. Adoi:10.1016/j.astropartphys.2008.07.004

* Corresponding author. Tel.: +1 650 926 4525; faxE-mail address: kamae@slac.stanford.edu (T. Kama

1 Present address: Lawrence Livermore National Lab

a b s t r a c t

We describe a new balloon-borne instrument (PoGOLite) capable of detecting 10% polarisation from200 mCrab point-like sources between 25 and 80 keV in one 6-h flight. Polarisation measurements in thesoft gamma-ray band are expected to provide a powerful probe into high energy emission mechanisms aswell as the distribution of magnetic fields, radiation fields and interstellar matter. Synchrotron radiation,inverse Compton scattering and propagation through high magnetic fields are likely to produce high degreesof polarisation in the energy band of the instrument. We demonstrate, through tests at accelerators, withradioactive sources and through computer simulations, that PoGOLite will be able to detect degrees of polar-isation as predicted by models for several classes of high energy sources. At present, only exploratory polar-isation measurements have been carried out in the soft gamma-ray band. Reduction of the large backgroundproduced by cosmic-ray particles while securing a large effective area has been the greatest challenge.PoGOLite uses Compton scattering and photo-absorption in an array of 217 well-type phoswich detectorcells made of plastic and BGO scintillators surrounded by a BGO anticoincidence shield and a thick polyeth-ylene neutron shield. The narrow field of view (FWHM = 1.25 msr, 2.0 deg� 2.0 deg) obtained with detectorcells and the use of thick background shields warrant a large effective area for polarisation measurements(�228 cm2 at E = 40 keV) without sacrificing the signal-to-noise ratio. Simulation studies for an atmosphericoverburden of 3–4 g/cm2 indicate that neutrons and gamma-rays entering the PDC assembly through theshields are dominant backgrounds. Off-line event selection based on recorded phototube waveforms andCompton kinematics reduce the background to that expected for a �100 mCrab source between 25 and50 keV. A 6-h observation of the Crab pulsar will differentiate between the Polar Cap/Slot Gap, Outer Gap,and Caustic models with greater than 5r significance; and also cleanly identify the Compton reflection com-ponent in the Cygnus X-1 hard state. Long-duration flights will measure the dependence of the polarisationacross the cyclotron absorption line in Hercules X-1. A scaled-down instrument will be flown as a pathfindermission from the north of Sweden in 2010. The first science flight is planned to take place shortly thereafter.

� 2008 Elsevier B.V. All rights reserved.

ll rights reserved.

: +1 650 926 5566.e).oratory, Livermore, CA 94550, USA.

T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84 73

1. Introduction

Celestial X-ray and gamma-ray sources have been studied usingtheir spectrum, time variability and projected image since the early1960s (see, for example, a review by Fabian et al. [1]). For manysources, such observations alone do not identify the dominantemission mechanism and polarisation measurements are expectedto add decisive information. Polarimetry will be particularlyimportant for studies of pulsars, accreting black holes and jet-dom-inated active galaxies. Strong X-ray and gamma-ray polarisationcan arise from synchrotron emission in ordered magnetic fields,photon propagation in extremely strong magnetic fields (>1012 G)and anisotropic Compton scattering, as has been discussed in [2]and reviewed by Lei et al. [3]. The orientation of the polarisationplane probes the intensity and direction of the magnetic and radi-ation fields, as well as the matter distribution around sources.

Despite the potential importance of polarisation measurements,the Crab nebula is the only source outside the solar system fromwhich polarisation has been significantly detected in this energyrange. The first clear detection of polarisation was at 2.6 keV and5.2 keV, with an instrument on board the OSO-8 satellite [4,5].Due to the limited effective area, the instrument could not detectpolarisation from the Crab pulsar or other sources [6,7].

A polarisation detection has been reported from a gamma-rayburst observed with an instrument on the RHESSI satellite in2002 [8]. This claim has been seriously challenged and hence re-mains controversial [9,10]. A new detection of polarisation intwo gamma-ray bursts archived in the BATSE catalog has beenpublished recently [11]. The authors fitted the observed time-bin-ned counting rates including polarisation-dependent Comptonreflection off the atmosphere. The fit indicates a polarisation de-gree exceeding 35–50%.

Two techniques have been developed to measure hard X-ray(soft gamma-ray) polarisation from astrophysical sources. Below10 keV, polarisation can be deduced by tracking the electron fromphoto-absorption in an imaging gas detector. From about 25 keV to1 MeV, polarisation can be determined by measuring the azi-muthal angle distribution of Compton scattered photons as hasbeen exemplified in papers contained in the proceedings editedby Turner and Hasinger [12]. Due to atmospheric opacity, instru-ments have to be launched into a satellite orbit if the first tech-nique is to be used. Instruments relying on the latter techniquecan be flown on high altitude balloons since photons with energiesgreater than 20–25 keV reach the instrument through typicalatmospheric overburdens of 3–4 g/cm2. In this energy band, how-ever, the large background produced by cosmic-ray particles posessignificant detector design challenges. Several instruments are un-der development to detect X-ray or gamma-ray polarisation fromastronomical sources using the Compton scattering technique inballoon and satellite experiments. An important figure of meritfor these instruments is the lowest degree of polarisation detect-able at the 3r level, referred to as the minimum detectable polar-isation. Most instruments have gamma-ray bursts as their primarytarget. Examples include POLAR [13] (10–300 keV, satellite) andGRAPE [14] (50–300 keV, long-duration balloon or satellite). PHE-NEX [15] is designed to detect a 10% polarisation from the Crabnebula in the 40–300 keV range in a 3-h balloon observation. CI-PHER [16] will study steady-state objects between 10 keV and1 MeV with the capability of detecting 5% polarisation from theCrab nebula in a 3-h satellite orbit observation.

The polarised gamma-ray observer – light-weight version (PoG-OLite) is a balloon-borne astronomical soft gamma-ray polarimeteroptimised for point-like sources. It measures polarisation in the en-ergy range 25–80 keV from sources as low as 200 mCrab by usingthe azimuthal angle anisotropy of Compton-scattered photons.

Several features distinguish PoGOLite from other balloon-borneinstruments proposed to observe astronomical sources using theCompton scattering technique such as PHENEX [12,15]:

� the narrowest field of view: 1.25 msr (FWHM);� the largest effective area between 25 and 60 keV: �228 cm2 for

polarisation measurements at 40 keV;� sensitivity extends as low as to 25 keV to address the sources

and processes mentioned above;� the lowest background rate: �100 mCrab for 25–50 keV;� a lower modulation factor, �33%.

The instrument is currently under construction and an engi-neering flight of a 61 unit ‘‘pathfinder” instrument is planned for2010 from the Esrange facility in the North of Sweden. The instru-ment was originally designed to record Compton scattering andphoto-absorption in an array of 397 phoswich detector cells madeof plastic and BGO scintillators, surrounded by active BGO antico-incidence shields (called PoGO). Through trade-off studies includ-ing detector simulation and design, cost estimation andprototype testing, we converged on a lighter version of the originalPoGO instrument [17]. This new design (PoGOLite) will be able toreach a higher altitude (41–42 km) at balloon facilities where theballoon size is limited to a 1 million m3 balloon and extend thelower energy limit down to 25 keV. The lighter design also simpli-fies implementation of the mechanism which permits the polarim-eter to rotate around its longitudinal axis, which is essential inreducing systematic errors in polarisation measurements. Thepointing system and gondola design are inspired by the flight-pro-ven design of high energy focusing telescope (HEFT, [18]) and bal-loon-borne large-aperture sub-millimeter telescope (BLAST, [19]).The overall design of the PoGOLite polarimeter and gondola hasincorporated features needed to accomplish long-duration balloonflights from Sweden to North America in the future.

In this paper we describe the key PoGOLite polarimeter designfeatures, summarise the results obtained during tests with proto-type instruments, and present the scientific potential. Relevanttheoretical models for high energy emission from selected targetswill be discussed together with simulated measurements thereoffor one 6-h balloon flight. We show that PoGOLite will open anew observational window on high energy astrophysics, with thepromise of clarifying the emission mechanism of many sources.

2. PoGOLite detector and data processing

2.1. PoGOLite detector

The PoGOLite detector consists of a hexagonal close-packed ar-ray of 217 well-type phoswich detector cells (PDCs) and 54 sideanticoincidence shield (SAS) detectors made of bismuth germanate(BGO) scintillators. This detector assembly is housed in a rotatingcylindrical structure (the inner cylinder), which is placed inside an-other cylindrical structure reinforced with ribs and filled withpolyethylene (the outer cylinder) as shown in Fig. 1. The arrange-ment of the PDCs is detailed in Fig. 2.

Each PDC, Fig. 3, is composed of a thin-walled tube (well) ofslow plastic scintillator at the top (Eljen Technology EJ–240, fluo-rescence decay time 285 ns), a solid rod of fast plastic scintillator(Eljen Technology EJ–204, decay time �2 ns), and a bottom BGOcrystal (Nikolaev Institute of Inorganic Chemistry, decay time�300 ns), all viewed by one photo-multiplier tube (HamamatsuPhotonics R7899EGKNP). The thin-walled well is wrapped with areflective layer (3M VM2000) and 50 lm thick foils of lead andtin, and the bottom BGO is coated with a reflective BaSO4 layer.In the PDCs, the wells serve as active collimators, the fast scintilla-

Fig. 1. Axial cross-section of the PoGOLite polarimeter. The inner cylinder holds 54 BGO crystal modules, which comprise the Side Anticoincidence Shield (SAS), 217 PhoswichDetector Cells (PDCs), 271 photo-multiplier tubes (PMTs), a data acquisition system and the lower polyethylene neutron shield. The outer cylinder houses the mechanismwhich allows the polarimeter to rotate around the longitudinal axis, the pivot for elevation pointing, and the lateral polyethylene neutron shield. The assembly is �140 cm inlength and �100 cm in diameter and is estimated to weigh about 800 kg.

Fig. 2. An overview of the detector arrangement in the PoGOLite polarimeter. Theoutermost units are the SAS BGO modules. For clarity, not all modules are shown, sothe 217 PDCs are partially exposed, showing the slow and fast scintillators, thebottom BGO crystals, and the PMTs. The mechanical structures and the neutronshield are not shown.

Fig. 3. One Phoswich Detector Cell: (from top to bottom) 60 cm long slow plasticscintillator well, 20 cm long fast scintillator rod, 4 cm long BGO crystal and 19 cmlong phototube assembly. All scintillators are covered with reflective materials,VM2000 or BaSO4. In addition, the well portion of the PDC is wrapped with 50 lmthick lead and tin foils for passive collimation.

74 T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84

tor rods as active photon detectors, and the bottom BGOs as a low-er active shield. The overall length of the active collimator,600 mm, together with the 24 mm mean diameter well opening,sets the solid angle event acceptance, which is 1.25 msr(2.0 deg � 2.0 deg). Signals from each PMT are continuously sam-pled, digitized and recorded at 36 MHz. Through on board exami-nation of the recorded waveforms, signals consistent with beingfast scintillator light are selected from those mixed with slow/BGO scintillator light. This well-type phoswich detector technologywas developed and used to reduce the cosmic-ray-induced back-

grounds by more than one order of magnitude for the WELCOMEballoon experiments [20–27]). Based on this success, the technol-

T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84 75

ogy has been applied to the Suzaku Hard X-ray Detector [28–30].The hard X-ray detector is operating in orbit with the lowest back-ground achieved in its energy band, 12–600 keV [31,32].

The side anticoincidence system [33] consists of 54 modules ofBGO crystals which cover two thirds of the height of the PDC ele-ments. Each module is built from three crystals glued together,as shown in Fig. 4. The crystals have a pentagonal cross-sectionand are tightly packed around the PDC assembly. A module-to-module gap of �100 lm is foreseen, mainly due to the reflectiveBaSO4 layer applied to each crystal. The BGO crystals are suppliedby the Nikolaev Institute of Inorganic Chemistry and are read outwith the same type of phototube that is used for the PDC units.

With an estimated total side anticoincidence rate of about100 kHz at float altitude, a simple veto would reject about 6% of de-tected events by random coincidence. The segmentation allows theside anticounter and the PDC hits to be correlated, which will fur-ther reduce the number of valid events rejected. Segmentation alsoallows possible asymmetries in the backgrounds to be studied andcorrections to be applied in off-line analysis. The anticoincidencethreshold is around 75 keV.

In the PoGOLite design, the PDCs serve to detect and measureboth Compton scattering(s) and photo-absorption of astronomicalsoft gamma-rays. Hence, the design is scalable in size and facili-tates end-to-end tests from developmental stage. If the two func-

Fig. 4. The three types of BGO modules and crystals used in the PoGOLiteanticoincidence system. A corner module (center) and an edge module (right) of theSAS are shown, together with a bottom BGO crystal of the PDC (left).

Fig. 5. The flight model Flash ADC board,

tions had been assigned to different components, it would nothave been easy to attain a large effective area nor to performend-to-end tests.

2.2. Data processing

Signals from all 217 PDCs and 54 SAS PMTs are fed to individualflash ADCs on 38 Flash ADC (FADC) Boards (see Fig. 5) and digitizedto 12-bit accuracy at 36 MHz. Field programmable gate arrays(FPGAs) check for a transient signal compatible with an energydeposition in the fast scintillator (‘‘clean fast signal”) between aminimum (�15 keV) and a maximum (�200 keV). A veto signalis issued when the FPGAs detect a transient signal exceeding theupper discrimination level to suppress cosmic-ray backgroundsor signals compatible with the rise time characteristic of the slowscintillator or the bottom BGO. Digital I/O (DIO) boards (see Fig. 5)collect the trigger signals from eight FADC boards and process localtrigger signals (both clean fast signals and veto signals). A globalDIO board collects local trigger signals from DIO boards and pro-cesses a global trigger signal. If a clean fast signal is found withouta veto signal in a time-window, a global trigger is issued and dig-itized waveforms are stored for a period of �1.6 ls starting �0.4 lsearlier than the trigger. Only PDC data with a transient signalgreater than a value corresponding to around 0.3 photo-electron(0.5 keV) will be stored in the buffer of the FADC boards for dataacquisition. Note that we conservatively assume the threshold tobe 1.0 photo-electron in all computer simulations. The FADCboards also save information on what channels have stored wave-forms for all triggered events (we call it the hit pattern). In order tominimize the dead time, the DIO boards start accepting triggers assoon as it has received signals from all FADC boards that all wave-forms are buffered. The trigger rate is expected to be about 0.5 kHzresulting in a dead time fraction of about 0.6% and a data-rate ofabout 240 kB/s or 0.9 GB/h with zero suppression. An onboardcomputer, the Space Cube (Developed by JAXA and Shimafuji Elec-tronic Inc. [34], see Fig. 5), collects hit patterns and waveformstagged by trigger identification numbers from FADC boards asyn-chronously for dead-time-less data acquisition and records theminto flash memory drives. The Space Cube also collects the hit pat-tern from FADC boards to confirm that all waveforms are collectedfor every event.

The recorded data is processed in the following steps:

Step 1: Select the PDC where a photo-absorption took place bychoosing the highest energy deposition compatible with beinga clean fast signal in the fast scintillator. Fig. 6 shows how cleanfast signals in the fast scintillator of gamma-rays from 241Amare selected in a high background environment created by elec-trons from 90Sr irradiating the slow scintillator. Each dot in thefigure represents one recorded event: the horizontal and verti-

digital I/O board, and Space Cube TM.

Fig. 6. Selection of clean gamma-ray hits from 241Am on the fast scintillator rod while the slow scintillator well is irradiated with electrons from 90Sr. Each point in the lowerpanel corresponds to one triggered event: the charge collected in a short interval (120 ns) gives the abscissa and that in a long interval (1 ls) the ordinate. Signals in the fastscintillator rod are gamma-rays from 241Am (a line at 60 keV and several lines between 15 and 36 keV) lying between the two dashed lines. The pulse-height distribution ofthe selected fast scintillator signals is shown in the upper panel. Background electron signals form the thick band to the left of the dashed lines in the lower panel. A crudeenergy scale for gamma-rays detected in the fast scintillator has been added.

Fig. 7. Selection of Compton-scattered coincidence events for data obtained with a25 keV polarised beam at the KEK Photon Factory. The beam is arranged to scatterin the central PDC and be photo-absorbed in a surrounding PDC (cf. Fig. 8). Thedashed box shows the valid kinematic area for Compton scattering events where asmall amplitude fast signal from the recoil electron and a larger amplitude fastsignal from the photo-absorption are registered.

76 T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84

cal coordinates are the charges integrated over the fast(�120 ns) and slow (�1 ls) intervals respectively. The diagonalconcentration of dots between the two dashed lines corre-sponds to clean fast signals in the fast scintillator.Data inFig. 6 indicate strong concentrations around the dominant lineat 60 keV and the several weaker lines between 15 and36 keV in the diagonal slice, all produced by 241Am gamma-raysin the fast scintillator. Included in the lower-energy concentra-tion are Compton scattered events where only a fraction of thegamma-ray energy is deposited in the fast scintillator. The sitewith the highest clean fast signal pulse-height is selected as thephoto-absorption site.Step 2: Waveforms from the nearest neighbors plus the nextnearest neighbors (total 18 PDCs) around the identified photo-absorption site are searched for a Compton scattering signal.We estimate the chance of finding one Compton scattering hitto be about 20.5%, two hits about 10.5% and more than three hitsabout 15.5% when averaged between 20 and 80 keV. We plan toaccept the events with one and two Compton scattering sites. Inthe beam tests, the Compton site threshold was set to �0.3photo-electrons where 0.65 photo-electron corresponds to1 keV in the fast scintillator. If more than one Compton site isfound, the site with highest pulse-height is chosen. This strategyis justified by the fact that a low energy Compton recoil electronmeans that there is less effect on the azimuthal scattering angle.We note that the single photo-electron peak was clearly identi-fied during the beam tests described in this paper and isexpected to be so during balloon observations. Our computersimulation shows that the modulation factor of two Comptonsite events is �23% while it is �35% for one Compton site event.We conservatively assumed a 1.0 photo-electron threshold fordetecting Compton scattering in all simulations.Step 3: Distributions of the sum of the two energy depositions(the photo-absorption site and the Compton scattering site)versus the energy deposition at the Compton scattering siteare studied for the selection of valid events. Fig. 7 shows sucha distribution obtained with a polarised photon beam at the

KEK Photon Factory, Japan, in 2007. Seven flight-model PDCswere arranged as shown in Fig. 8. The area enclosed by the threedashed-line segments in Fig. 7 denotes the allowed kinematicalregion for Compton scattered and photo-absorbed events.

For these selected events, we define the azimuthal angle of scat-tering by a line connecting the centers of the two PDCs where

Fig. 8. Arrangement of the seven PDCs in the KEK 2007 beam test. The polarised25 keV pencil beam hits the center of PDC 0. Photo-absorption was recorded in oneof the six peripheral PDCs. The set-up was rotated in azimuth in 15-deg steps.

T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84 77

Compton scattering and photo-absorption are detected. Since thepolarimeter orientation drifts constantly in the celestial coordinatesystem during observations and since the polarimeter will be ro-tated around its axis, the azimuth angle fixed thereon has to bealigned against the position angle around the target object. Thedistribution of scattering position angles aligned in this way showsa sinusoidal modulation when the incident gamma-rays are polar-ised. This measured distribution is fitted with a sinusoidal curvewith a constant offset, the modulation curve. The modulation fac-tor is defined as the ratio between the amplitude of the sinusoidalpart and the offset.

Fig. 9. Modulation measured in a polarised 25 keV pencil beam at the KEK PhotonFactory. The beam polarisation was (90 ± 1)% and the measured modulation factoris (33.8 ± 0.7)%.

Table 1Expected performance characteristics of PoGOLite at an atmospheric overburden of 4 g/cm

25 keV 30 keV

Minimum detectable polarisation in 6 hfor a 100 mCrab source 10.5%for a 200 mCrab source 6.5%with the Crab spectrum

Field of view 1.25 msr (2.0 deg � 2.0Time resolution 1.0 lsGeometric area 994 cm2

Effective area for polarisation measurement(attenuation by air not included) (cm2)

93 167

Signal rate for a 100 mCrab source 0.039/s/keV 0.044/s/Background rate 0.043/s/keV 0.041/s/Modulation for 100% polarised beam with Crab spectrum (%) 33 29

Fig. 9 shows three sets of modulation curves obtained for threecoplanar pairs of PDCs in the KEK 2007 prototype. The abscissa isthe rotation angle of the prototype relative to the plane of thebeam polarisation. The PDC arrangement is shown in Fig. 8. Thebeam had an energy and polarisation of 25 keV and (90 ± 1)%,respectively. It was aimed at the center of PDC 0 and the six possi-ble coincidence pairs (PDC 0–PDC 1, PDC 0–PDC 2, PDC 0–PDC 3,PDC 0–PDC 4, PDC 0–PDC 5 and PDC 0–PDC 6) were combined intothree coplanar sets in the figure. The azimuthal angle of Comptonscattering is offset by 0, 60 and 12 deg for the three coplanar pairsPDC 1–PDC 4, PDC 2–PDC 5 and PDC 3–PDC 6, respectively. Themeasured modulation factor is (33.8 ± 0.7)% for the beam polarisa-tion of 90 ± 1%. We note that this modulation factor is for the 7-unit prototype model and for a pencil beam hitting at the centerof the array. In astronomical observations with a larger PDC array(either 61-unit or 217-unit), gamma-rays will illuminate uniformlyover the Fast Scintillator surface and a significant fraction of scat-tered gamma-rays will be absorbed at next nearest neighbors.Modulation factor will decrease by the former factor but increaseby the latter. The combination of the two effects brings the ex-pected modulation factor for 100% polarised astronomical sourceto the values tabulated in Table 1.

2.3. Background suppression

Potential backgrounds affecting the polarisation measurementcan arise from extraneous gamma-ray sources within the field-of-view, gamma-rays and charged particles that leak through theside and bottom BGO anticoincidence systems, and neutrons pro-duced in the atmosphere and the gondola structure. For high lati-tude flights, auroral X-ray emission due to bremsstrahlungemitted at altitudes of �100 km is also a potential background.

The bottom BGO crystals integrated into the 217 PDCs, in com-bination with the 54 BGO crystal assemblies in the side anticoinci-dence system (SAS), efficiently suppress the gamma-raybackground together with the slow plastic scintillator tubes andthe lead/tin foils around the tubes. Polyethylene walls surroundthe aluminum structures. The averaged thickness of the polyethyl-ene is about 10 cm in the upper part and about 15 cm in the lowerportion, as shown in Fig. 1. These walls, along with the slow plasticscintillator tubes, slow down fast (MeV) neutrons to keV energies,thereby making it less likely that a trigger is generated and theselection criteria described in the previous section satisfied.

The PoGOLite design adopts five new background suppressionschemes compared to traditional phoswich technology: (1) an ac-tive collimator limits the field of view to 1.25 msr (2.0 deg �2.0 deg); (2) a thick polyethylene neutron shield is included; (3)all non-zero PDC waveforms are sampled at 36 MHz for about1.6 ls starting 0.4 ls before the trigger to eliminate fake eventsproduced by pulse pile-up; (4) additional suppression of neu-

2

40 keV 50 keV 60 keV 80 keV Total/Average

deg)

228 198 172 158

keV 0.039/s/keV 0.025/s/keV 0.015/s/keV 0.0056/s/keV 1.52/skeV 0.030/s/keV 0.029/s/keV 0.022/s/keV 0.018/s/keV 1.77/s

26 27 32 40 32.5

78 T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84

tron-induced background is possible using recorded PDC wave-forms as shown in Fig. 10; (5) the counting rate and energy spec-trum are recorded for all 54 BGO assemblies of the SAS tofacilitate correction for possible azimuthal modulation in back-ground events, as discussed further below.

2.3.1. Background simulationBackground rates have been calculated using simulation pro-

grams based on the Geant4 package [35,36]. The program incorpo-rates background flux models, the PoGOLite detector geometry, thetrigger logic, energy-loss due to scintillation light conversion, andthe on board and off-line event filtering algorithm.

Background gamma-rays can reach the fast scintillators bycrossing the BGO crystals unconverted as well as through gaps be-tween the BGO crystals. The gamma-ray background flux modelhas been developed based on observational data [37] and gener-ates diffuse cosmic and atmospheric gamma-rays representativeof a balloon environment with 4 g/cm2 atmospheric overburden.

Neutrons can fake valid Compton scattering events throughelastic scattering off protons in the fast scintillators. Two or moreclean fast signals have to be produced with kinematics compatiblewith Compton scattering. Since the neutron mean-free-path in theplastic scintillator is less than 1 cm for kinetic energies below3 MeV, fake events cannot come from neutrons with energies lessthan �1 MeV. Neutrons must also escape detection in the BGOscintillator. In this regard, we note that the gamma-ray emittinginelastic cross-sections exceed 1 barn in bismuth for neutrons withenergies greater than �5 MeV. A combination of these two condi-tions limits background neutrons to the 1–100 MeV range. Thescintillation light yield in one fast scintillator must be consistentwith that for the photo-absorption site in the PoGOLite energyband, and the other one with that for the Compton scattering site.Scintillation light emitted by the low energy recoil proton in n-pelastic scattering is suppressed as described by Verbinski et al.[38] and Uwamino et al. [39]. Our simulation program estimatesthese processes with dedicated programs within the Geant4 frame-work using the QGSP_BERT_HP physics list [40]. We note that 96%of the neutron flux is attenuated by the combination of a 10 cmthick polyethylene wall and 6 cm (in average) slow scintillator ar-ray. More details can be found in [41].

The neutron flux used in the simulation is from a model pro-posed by Armstrong et al. [42] for an atmospheric depth of 5 g/cm2. The angular distribution is assumed to be isotropic, but wenote that according to Armstrong et al. [43], the upward and down-

Fig. 10. Distributions of the ratio of the pulse-heights integrated with the fast(120 ns) and slow (1 ls) time constants for neutron-induced (---) and gamma-ray(—) events in the energy band between 45 and 75 keV. The neutrons are from 252Cfand the gamma-rays are from 241Am. About 51% of neutron-induced backgroundevents can be rejected while keeping about 72% of the gamma-ray signals.

ward fluxes are typically 80% and 20% of the isotropic flux, respec-tively. This anisotropy in background neutron distributiontherefore affects the background rate at less than �20% level.

The neutron flux produced by cosmic-rays in the passive com-ponents of the payload has been calculated to be negligible basedupon an empirical formula developed by Cugnon et al. [44].

Background due to charged particle interactions has been deter-mined to be negligible. This background is eliminated by requiringthat the pulse-height lies between lower and upper discriminationlevels, and by the anticoincidence function of the BGO shieldembedded in the PDC assembly and the SAS.

The background presented by auroral X-rays in high latitudeflights (e.g., from the Esrange facility in Northern Sweden) has beenstudied by Larsson et al. [45]. In a conservative scenario, significantauroral X-ray emission is predicted for less than 10% of the flighttime during a long duration flight from Esrange. The resulting X-ray fluxes in the PoGOLite energy range are expected to be of afew tens of mCrab. Short duration spikes exceeding 100 mCrabare also possible. In order to identify periods with significant auro-ral activity, PoGOLite will be equipped with auroral monitoringinstruments (narrow passband photometers and magnetometers).It is noted that although a background for PoGOLite, polarisationmeasurements of auroral emission will provide unique results,allowing the pitch angle distribution of the incident electrons tobe determined, thereby giving essential clues to the accelerationprocesses at play.

We also expect higher charged particle, neutron and gamma-ray backgrounds during flights from Northern Sweden than fromTexas or New Mexico, USA. We plan to assess backgrounds duringan engineering flight from Sweden and increase the thickness ofthe neutron shield before making long-duration flights if needed.

The total background rate is shown in Fig. 11 with filled circles,the neutron background rate with open circles, and the gammabackground rate with filled squares. These rates can be comparedwith signal coincidence rates for a 1 Crab source (thick solid histo-gram) and a 100 mCrab source (thin solid histogram). The expectedtotal background coincidence rate is equivalent to that of a�100 mCrab source between 25 and 50 keV. Our simulation pro-gram predicts about 60–70% of the total background to be pro-duced by albedo neutrons and the rest by albedo gamma-rays.The contribution from charged cosmic-rays will be less than 10%.

2.3.2. Minimizing possible systematic biasBackgrounds can affect polarisation measurements in two dif-

ferent ways: First, the minimum detectable polarisation deterio-

Fig. 11. Expected background rates at an atmospheric overburden of 4 g/cm2

compared with signal rates expected for a Crab source (thick solid histogram) and a100 mCrab source (thin solid histogram). (d) total background, (s) neutronbackground and (j) gamma-ray background.

Fig. 12. The detector geometry used in the neutron test. The width of the array isabout 10 cm. A polyethylene shield with a thickness of 10 cm (not shown here)

T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84 79

rates due to statistical fluctuations in the background counts. Formost polarised astronomical sources, 10–20% polarisation is ex-pected. This means an instrument must be able to detect a 3–6%modulation factor. Instrumental response must be axially symmet-ric to better than a few percent. Various sources of asymmetry areexpected: anisotropy in background events, systematic offset inthe pointing and asymmetry in the instrument response.

To reduce systematic bias, the PDC assembly will be rotated axi-ally in 15-deg steps. Such a rotation mechanism has been imple-mented in the beam tests and the measured modulation factorsfor six combinations of photo-absorption and Compton scatteringsites have agreed to within �2% in absolute value.

Second, the background due to albedo neutrons will be aniso-tropic when the polarimeter axis is tilted off the zenith. Such ananisotropy will be monitored by recording counting rates for all54 BGO crystal assemblies of the SAS and for the outermost 48PDC units. The measured anisotropy will be used to remove mod-ulation artifacts introduced by background events.

surrounds the entire detector array. The SAS BGO shields are facing the neutrongenerator (not shown in this picture) to resemble the situation with atmosphericneutrons incident on the side anticoincidence shield surrounding the instrument.

Fig. 13. Neutron spectra from simulation (thick solid line) and from measurements(thin solid line) for a plastic scintillator in a simple detector array surrounded by a10 cm thick polyethylene shield. The two peaks in the measured spectrum are fromsaturation in the electronics and are therefore not expected in the simulation. Thedistributions are normalized to the value at 10 keV.

3. Performance verification with accelerator beams andcomputer simulations

Various prototype PDCs have been tested in polarised gamma-ray beams to verify the simulation program – including physicsprocesses implemented in Geant4 – and subsequently to guideoptimisation of trigger thresholds and the event selection algo-rithm. In 2003, a beam test was conducted at the Argonne NationalLaboratory, in which the analysis on the azimuthal modulationdemonstrated errors in the treatment of photon polarisation inthe Compton/Rayleigh scattering processes implemented inGeant4 [46]. The next test, conducted at KEK in Japan, was aimedat studying the low energy response of the PMT assemblies [47].One flight PDC and six prototype PDCs were then tested at KEKin 2004 and the revised Geant4 model was verified to �3% (abso-lute) in modulation factor [48]. The most recent test at KEK wascompleted in March 2007 with seven flight-version PDCs andfront-end electronics. The measured modulation for (90 ± 1)%polarised 25 keV gamma-rays is presented in Fig. 9 and the detailswill be described in a separate publication [49].

The response of the PDC fast plastic scintillator to neutrons inthe MeV energy range has been studied in great detail. Tests withneutrons produced in the decay of 252Cf confirm that the neutron-induced background can be reduced by about 49% while sacrificingabout 28% of gamma-ray events between 45 and 75 keV (seeFig. 10). This filtering can be applied during off-line analysis tothe outer-most layer of PDCs where neutron background eventsconcentrate. The PDC response in the 10 MeV range and the valid-ity of the Geant4 simulations of the in-flight neutron backgrounddescribed in Section 2.3.1 were tested with 14 MeV neutrons froma D–T reaction. In these tests, a simplified detector geometry, con-sisting of four fast plastic scintillators and three SAS BGO crystalsas shown in Fig. 12, was used. This detector array was surroundedby a 10 cm thick polyethylene shield, mimicking the PoGOLitedetector construction.

The neutron count rate in the central scintillator (Fast0) wasmeasured both with and without active vetoing in the surroundingscintillators. In both cases, the recorded spectrum was well repro-duced in simulations. An example is given in Fig. 13, which showsthe measured and the simulated spectrum obtained with the veto-ing system in use. The neutron count rate was reduced by (58 ± 4)%after active vetoing, and the corresponding result from simulationis (57.0 ± 1.4)%. This agreement indicates that Geant4 and theQGSP_BERT_HP physics list provide a reliable reconstruction ofneutron interactions in the PoGOLite instrument. Further detailsof these measurements and simulations can be found in [50].

Simulation studies show that charged cosmic-ray interactionsin the PDC assembly can be filtered out to a negligible level be-cause of their higher energy deposition. However, saturation ef-fects due to large signal depositions could mimic low level fastscintillator signals in subsequent events. Moreover, in the SAS, cos-mic-ray crossing can introduce dead time and lower off-line eventfiltering efficiency. Various tests were conducted with a protonbeam (392 MeV) at the Research Center for Nuclear Physics in Osa-ka to study the behavior of the front-end electronics under cosmic-ray bombardment. In a test run, a gamma-ray spectrum from241Am irradiating the fast scintillator was recorded while protonsbombarded either the slow scintillator tube or the bottom BGOof the PDC. We found that the spectrum is little affected down toabout 20 keV for a proton rate of 5 kHz, which is more than an or-der of magnitude higher than the expected cosmic-ray rate. Themeasurement described in Section 2.2 and Fig. 6 also show thatclean fast signals (gamma-rays from 241Am) are recovered even un-der an intense electron bombardment (�10 kHz) of the slowscintillators.

The expected PoGOLite performance as given in Table 1 hasbeen calculated based upon beam test results and simulations

Table 2Expected PoGOLite northern targets

Target Coincidencerate (s�1)

Minimum detectablepolarisation (%)

Crab (total) 15.2 2.4Cyg X-1 (Hard state) 14.9 2.4Cyg X-1 (Soft state) 5.3 4.6Hercules X-1 2.7 8Mkn 501 (Flare) 0.82 17V0332+53 (burst) �4 5.34U0115+63 (burst) �4 5.3GRS 1915 (burst) �4 5.3

80 T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84

(Table 2). By reducing backgrounds and having a large effectivearea (about 228 cm2 at 40 keV), PoGOLite is able to detect a 10%polarised signal from a 200 mCrab source in a 6-h flight.

4. Science with PoGOLite

4.1. Crab nebula

The Crab nebula has been the prime target of polarisation mea-surements in several electromagnetic wave bands. The first detec-tion of large polarisation in the optical band, by Vashakidze [51],Oort and Walraven [52] and Woltjer [53], led to the establishmentof synchrotron radiation as the dominant emission mechanism inthe optical band, similar to the process suggested as an explanationfor the polarised radio emission by Shklovsky [54]. Stimulated bythe implication that electrons are accelerated to high energies inthe nebula, many X-ray observations followed, including high spa-tial resolution (�1500) mapping of the nebula in the X-ray band (1–6 keV) by Oda et al. [55]. The authors concluded that the X-raysource is extended and that its intensity profile is consistent withits optical image. Several high spatial resolution maps were ob-tained in the soft gamma-ray band using lunar occultation of theCrab nebula (see references given in [56]).

Aschenbach and Brinkmann [56] modeled the X-ray emittingstructure of the Crab nebula based on observations available thenand concluded that high energy electrons are trapped in a mag-netic torus around the Crab pulsar and emitting synchrotron radi-ation. We note that Rees and Gunn [57] demonstrated theoreticallythat a magnetic torus will be built up by the relativistic outflow ofthe Crab pulsar. The plane defined by the magnetic torus was pre-dicted to be perpendicular to the spin axis of the Crab pulsar. Theradiation region can extend well beyond the torus and its size isdetermined by the magnetic field and involves both electron diffu-sion and bulk motion. The diameter was calculated to be around 10

at 20 keV and 2–40 in the optical band by Aschenbach and Brink-mann [56]. The authors noted that the observed elongation inthe optical emission region was orthogonal to that in the X-rayemission region, probably due to propagation conditions of theelectrons. For higher energy X-rays, the emission region was pre-dicted to shrink to an extended elliptical configuration with a ma-jor axis of about 4000 and a minor axis of 2200.

The entire Crab nebula was imaged in the soft gamma-ray band(22–64 keV) to about 1500 resolution by Makishima et al. [58] andPelling et al. [59]. The observed images were consistent with theemission being synchrotron radiation from high energy electronstrapped around a magnetic torus similar to that described above.The radiation region in the 43–64 keV band was significantly smal-ler than in the 22–43 keV band, consistent with the prediction byAschenbach and Brinkmann [56].

Fine X-ray images taken with the Chandra X-ray Observatoryhave brought a renaissance to the structural study of the Crab neb-ula and other pulsar wind nebulae (see [60] and references given ina review by Arons [61]). The discovery of two concentric torii is one

of the many key observations brought forward by Chandra. Ng andRomani [62] used the Chandra image to determine the orientationof the magnetic torii accurately: the position angle of the Crab pul-sar spin axis was determined to be 124.0 ± 0.1 deg and126.31 ± 0.03 deg on the inner and outer torii, respectively. The an-gle relative to the line of sight was determined to be 61.3 ± 0.1 degand 63:03þ0:02

�0:03 deg for the inner and outer torii, respectively.If high energy electrons are trapped in a macroscopic toroidal

magnetic field, we expect the polarisation plane to be parallel tothe spin axis. Nakamura and Shibata [63] built a 3D model of theCrab nebula and predicted the polarisation position angle depen-dence. A 3D relativistic MHD simulation of pulsar wind nebulaewas presented by Zanna et al. [64]. Both models are axisymmetricand predict the polarisation plane to be parallel to the spin axis.Zanna et al. [64] predict the surface brightness distribution be sig-nificantly different in the optical and X-ray bands.

The X-ray polarisation position angle measured by Weisskopfet al. [4] is 161 ± 2.8 deg at 2.6 keV and 155.5 ± 6.6 deg at5.2 keV, i.e., about 30 deg from the spin axis determined by Ngand Romani [62]. The polarisation angle is consistent with162.0 ± 0.8 deg measured in the optical band of the central region(diameter 3000) of the Crab nebula [52]. A spatially resolved polar-isation measurement of the entire Crab nebula by Woltjer [53] hasrevealed that the emission is extended to 40 � 30 and highly polar-ised (up to �50%) in the outer part of the nebula. In the centralpart, the measured degree of polarisation fluctuates highly at 1000

scale, possibly associated with filamentary structures. Woltjer[53] has estimated the error in his position angle measurementto be �6 deg.

As the emission region shrinks for higher energies, the polarisa-tion position angle is expected to approach that of the spin axis(�124 to 126 deg). PoGOLite can isolate the Crab nebula from theCrab pulsar by removing gamma-rays detected within 1.65 msand 3.3 ms of the two peaks, P1 and P2 in Fig. 14, respectively. PoG-OLite can detect 2.7% polarisation (3r) and determine the positionangle with a precision of �3 to 4 deg for the nebula component be-tween 25 and 80 keV. Energy dependence of the polarisation posi-tion angle will either confirm the magnetic torus around the Crabpulsar or seriously challenge the standard model of high energyemission. Either way, it will make a critical contribution to themodeling of the Crab nebula.

4.2. Emission mechanism in isolated pulsars

Rapidly-rotating neutron stars are a prime target for polarimet-ric studies. Their emission region is known to be different in differ-ent wavelength bands. Historically, phase-resolved polarimetryhas had enormous diagnostic capability at radio and optical wave-lengths. The expected signature of emission near the poles of a di-pole field, an ‘‘S”-shaped swing of the polarisation position anglethrough the pulse profile [65], has been seen in many radio pulsars.This is generally accepted as proof that the radio emission origi-nates from the open field lines of a magnetic dipole. In the X-rayand gamma-ray regimes, each pulse period reveals two peaks,called P1 and P2. A phase-resolved polarisation measurement ofthe corresponding soft gamma-ray flux will indicate where in themagnetosphere the emission occurs.

Models for the high energy emission fall into two general clas-ses. In the polar cap model and its various modifications [66,67],particle acceleration occurs near the neutron star surface and thehigh energy emission results from curvature radiation and in-verse-Compton induced pair cascades in the presence of a strongmagnetic field. Outer gap models [68–70] assume that accelerationoccurs in vacuum gaps that develop in the outer magnetospherealong the last open field line between the null-charge surfacesand the light cylinder, and that high energy emission results from

Fig. 14. Model predictions versus phase: intensity (top panels), polarisation position angle (middle panels) and polarisation degree (bottom panels) for (left) the polar capmodel [67], (center) the outer gap model [69] and (right) the caustic model [71]. The areas between a pair of vertical lines in the figure correspond to the first pulse (P1) andthe second pulse (P2).

T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84 81

electron-positron cascades induced by photon–photon pair-pro-duction. These mechanisms intrinsically produce highly polarisedradiation (up to 70%, depending on the particle spectrum) beamedalong the magnetic field lines, with electric vectors parallel or per-pendicular to the local field direction.

The Crab pulsar is the primary target for PoGOLite. Fig. 14shows the theoretical polarisation position angle and polarisationdegree as a function of pulse phase for the polar cap, outer gapand slot gap or ‘caustic’ [71,72] models. In the caustic model, theCrab pulse profile is a combination of emission from both poles,whereas in both the polar cap and outer gap models, radiation isseen from only one pole or region. The signature of caustic emis-sion is a dip in the polarisation degree and a rapid swing of the po-sition angle at the pulse peak. The expected azimuthal modulationfor P1, as defined in Fig. 14, has been calculated for a 6-h observa-tion and is shown in Fig. 15. In this calculation, we assumed thatthe nebula component is polarised with a polarisation degreeand angle as measured by OSO-8 [4].

Fig. 15. Modulation in the azimuthal Compton scattering angle predicted for P1 ofthe Crab pulsar by three pulsar models: (—) polar cap model, (- --) outer gap model,(-�-) caustic model. A 6-h simulated observation by PoGOLite at 4 g/cm2 atmo-spheric overburden is shown.

The PoGOLite instrument will distinguish between these threemodels at the 5r level using P1 alone. Predictions of the polarisa-tion characteristics at P2 for the three models are rather similar.Measurements of the region between the two peaks will also beimportant in understanding the pulsar emission mechanism inthe gamma-ray band. Even if none of these models turns out tobe correct, the PoGOLite data will provide strong constraints uponany future models.

4.3. Accreting black holes

Another class of intense galactic X-ray sources where soft gam-ma-ray polarisation is expected are X-ray binary (XRB) systems,such as Cygnus X-1, where a compact object accretes matter froma companion star (for reviews, see [73,74]. Dramatic spectral dif-ferences suggest that the two major states seen in such sourcesare related to changes in the mass accretion rate, resulting in dif-ferent structures of the accretion flow (see, e.g. [75]). At high accre-tion rates, accretion takes place through a geometrically thin,optically thick accretion disc, extending into the innermost stableorbit around the black hole [76]. At lower accretion rates, the innerpart of the disc is replaced by a geometrically thick, optically thin,hot inner flow, possibly advection-dominated [77]. Most XRBs ex-hibit transitions between these two spectral states, commonly re-ferred to as the soft and the hard state, respectively.

In the hard state, the primary X-ray emission arises via repeatedCompton upscattering of thermal X-ray photons from the trun-cated outer disc by hot electrons in the inner flow. The resultinghard X-ray spectrum also shows signatures of a reprocessed com-ponent, arising from the primary hard X-ray photons reflecting offthe cooler accretion disc. Evidence for this reflection component isprovided by a fluorescent Fe-line at �6.4 keV, an Fe absorptionedge at 7 keV, and a broad ‘‘hump” at �10 to 200 keV [78]. Whilethe primary emission is not expected to be significantly polarisedas it arises via multiple Compton scatterings, the reflection compo-nent is expected to display significant intrinsic polarisation [79,80].The observed degree of polarisation is also dependent on the incli-nation of the system, giving polarimetry the potential to deriveinclinations of XRB systems, which is often difficult by othermeans.

82 T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84

Some viable, but not widely accepted, models suggest that apart of the high energy emission arises via synchrotron processesin a collimated outflow or jet ([81]; see also Section 4.5). Thesemodels predict a much higher degree of polarisation.

In the soft spectral state, hard X-ray emission arises from Comp-ton upscattering of photons by energetic non-thermal electronsdistributed in active regions above the accretion disk. As both thegeometry and electron distribution differ from that in the hardstate, a different polarisation signature is expected. Apart fromgeometrical constraints, polarisation measurements in the softstate can provide information about such a non-thermal electrondistribution. Measuring the polarisation at energies above 10 keVwill thus provide independent constraints on the inferred accretiongeometry, and constitutes a test for the models inferred from spec-tral and temporal studies.

A natural candidate for polarisation measurements is the per-sistent source Cygnus X-1, one of the brightest and best studiedXRBs, where the compact object is a stellar-mass black hole. Sincethe albedo of the reprocessed component contributes �30%, a netpolarisation of �10% is expected in the hard state, and a slightlylower degree in the soft state. Simulations show that even a lowdegree of polarisation (3–5%) is detectable with PoGOLite in eitherspectral state. In a long-duration flight, PoGOLite will also be ableto measure energy dependence of the polarisation (see [82,83]).

The degree of polarisation may be even higher in other XRBsources, e.g., Cygnus X-3. Recent results show that it may exhibita state totally dominated by Compton reflection [84].

The spectral formation process in active galactic nuclei – mainlyin Seyfert nuclei – is believed to be similar to that inferred forGalactic XRBs, although the mass of the accreting black hole is atleast 106 times higher. While a short PoGOLite flight is unlikelyto be sufficiently sensitive to measure soft gamma-ray polarisationin such sources (e.g., for NGC 4151, the emission is much fainter,reaching 10–80 keV fluxes of �10 mCrab), they certainly are excel-lent candidates for future polarisation studies in the soft gamma-ray band.

4.4. Accreting neutron stars

Accretion onto highly magnetized neutron stars, such as Hercu-les X-1 and 4U0115+63, provides a unique opportunity to studyphysical processes under extreme conditions. The observed X-rayand gamma-ray radiation originates from accreted material flow-ing along magnetic field lines onto distinct regions near the polarcaps (for a review, see [85]). The localized emission region andneutron star rotation lead to a modulation of the observed flux.The timing of these pulsations can be used to determine the orbitalparameters for the neutron star in the binary system. A harmonicabsorption feature has been detected first from Hercules X-1 byTrümper et al. [86] and later from several other objects in the hardX-ray spectrum [87]. These absorption features are interpreted ascyclotron resonances in magnetic fields of order 1012 G.

Due to the strong magnetic field, the radiation from the accre-tion column is expected to be linearly polarised. As the neutronstar rotates, the orientation of the magnetic field with respect toour line of sight changes and so does the direction and strengthof the polarisation. By observing these modulations, one can deter-mine the orientation of both the rotation and the magnetic axis ofthe star. With observations separated in time, one can also searchfor neutron star precession, which should show up most clearly inpolarimetry.

The beaming pattern of the emitted radiation depends stronglyon how the accreting matter is decelerated as it hits the neutronstar. If a stand-off shock is formed, the radiation will come froma vertically extended accretion column with the strongest emissionnormal to the magnetic field (fan beam). On the other hand, decel-

eration by Coulomb collisions will result in a thin hot plasma slabwith strongest emission in the vertical direction (pencil beam),which is parallel to the magnetic field [88]. The two geometrieswill have opposite correlations between flux and polarisation, sowith polarisation measurements, it will finally be possible to dis-tinguish between these two alternatives.

In both models above, the polarisation is expected to vary withenergy. PoGOLite will not be able to study polarisation variationswithin a cyclotron feature but it has sufficient energy resolutionto detect a strong variation of polarisation across the full spectralband of the instrument [82].

Hercules X-1 is the highest priority target among the persis-tently bright sources in this category. In addition to the persistentsources, the galaxy also contains a population of similar transientaccreting X-ray pulsars. During outbursts, which may last forweeks or months, some of these sources are among the brightestX-ray sources on the sky. One example is V0332+53, which atthe end of 2004 had its fourth outburst since the early 1970s. Atmaximum, the source flux exceeded one Crab and showed threestrong cyclotron features in its hard X-ray spectrum (see, e.g.[89]). The high flux and strong cyclotron features would make thisan extremely interesting target for hard X-ray polarimetry.

4.5. Jets in active galaxies and galactic binaries

Imaging and variability observations in many wavelength bandsindicate that a fraction of all active galactic nuclei are associatedwith relativistic jets pointing close to our line of sight. Emissionis highly variable, and extends over all accessible bands. EGREThas revealed that many active galaxies possessing jets are powerfulgamma-ray emitters, and that the observed gamma-ray flux origi-nating in the jet by far dominates that measured in other bands(see, e.g. [90]); such jet-dominated active galaxies are commonlyknown as blazars. Jets are thus common and energetically impor-tant ingredients for such classes of active galaxies. However, theformation, acceleration, collimation, and contents of a jet arepoorly understood, although the most promising models invokemagneto-hydrodynamic processes as the mechanism for the jetproduction (see a review by Sikora et al. [91]).

From an observational standpoint, the broad-band emissionspectra of blazars generally show two pronounced humps: one inthe radio to soft gamma-ray range, most likely produced by thesynchrotron process; and another, peaking in the MeV/GeV bandand extending in some objects up to TeV energies, most likelydue to inverse-Compton processes via the same electrons that pro-duce the synchrotron emission. The inference that emission fromthe low energy hump in blazars is produced by synchrotron radia-tion is based on the observed high degree of polarisation in theradio through UV bands. However, we know nothing about the le-vel of polarisation in the X-ray/soft gamma-ray band, although thepolarisation properties are a powerful tool to discriminate betweenemission models (see, e.g. [92–94]). In one sub-class of blazars, theso-called High frequency-peaked BL Lac (HBL) objects, the synchro-tron hump definitely extends to the soft gamma-ray band, and thelevel of polarisation as a function of the energy can be a powerfultool to study the details of the distribution of particle energy andthe intensity of magnetic field in the jet. PoGOLite is expected todetect high polarisation (>10%) in these objects; non-detection ofpolarisation, on the other hand, would put many current modelsof blazars into question.

Blazars are variable, with high states (occurring every few years)lasting for months. There are two primary Northern targets forPoGOLite balloon flights, Mkn 501 (see [95]), and 1E1959+65 (see[96]). The brighter one will be selected, depending on the currentflare state. In the Southern hemisphere, the most promising targetis PKS 2155-304 (see, e.g. [97]).

T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84 83

The galactic analogues to quasars, X-ray binary systems, alsosometimes show powerful jets, and are then referred to as micro-quasars, as reviewed by Fender et al. [98]. The jets in microquasarsare studied in the radio band, as the emission region is too compactto be resolved in X-rays. A detection of a high degree of polarisa-tion in the soft gamma-ray band would support the jet origin ofgamma-rays, a model that is viable but not at all widely accepted(see [98]). Since the radiation is from high energy electronstrapped in jets, a measurement of the plane of polarisation will re-veal the direction of the magnetic field. The prime microquasar tar-get is GRS 1915+105, which is the most spectacular of suchsources, often reaching soft gamma-ray fluxes twice that of theCrab system (for a recent overview, see [99]).

5. Summary and conclusions

PoGOLite is a balloon-borne soft gamma-ray polarimeter de-signed to enhance the signal-to-noise ratio and secure a largeeffective area for polarisation measurement (�228 cm2 at 40 keV)by limiting the field of view of individual pixels to 1.25 msr(FWHM) based on well-type phoswich detector technology. Thickneutron and gamma-ray shields have been added to reduce back-ground to a level equivalent to �100 mCrab between 25 and50 keV. Through extensive detector characterisation at polarisedsynchrotron beams and tests with neutrons, radioactive sourcesand accelerator protons, we have confirmed the predicted instru-ment to perform in accordance with Geant4-based simulation pro-grams. PoGOLite can detect a 10% polarisation from a 200 mCrabobject in a 6-h balloon flight and will open a new observationalwindow in high energy astrophysics.

Acknowledgements

We wish to thank Jonathan Arons, Makoto Asai, Roger Blandford,Blas Cabrera, Pisin Chen, Persis Drell, Steven Kahn, Tatsumi Koi, Ka-zuo Makishima, Lennart Nordh, Takashi Ohsugi, Masashi Takata,and Marek Sikora for their continuing support and encouragement.We are grateful to Alice Harding for providing pulsar model predic-tions in numerical form. We acknowledge Charles Hurlbut and EljenTechnology for developing the slow scintillator tube, HamamatsuPhotonics for improving the performance of R7899EGKNP, and theNikolaev Institute of Inorganic Chemistry for supplying BGO crys-tals of excellent quality. The Space Wire based I/O and ADC boardswere developed in JAXA’s program ”Research and Development forFuture Innovative Satellite.” T.K. has benefited greatly from discus-sion with Jonathan Arons on emission mechanisms in pulsar windnebulae. During the early developmental phase, the PoGOLite Col-laboration benefited from discussion with John Mitchell, RobertStreitmatter and Daniel Marlow.

We gratefully acknowledge support from the Knut and AliceWallenberg Foundation, the Swedish National Space Board, theSwedish Research Council, the Göran Gustafsson Foundation, theUS Department of Energy contract to SLAC no. DE-AC3-76SF00515,the Kavli Institute for Particle Astrophysics and Cosmology (KIPAC)at Stanford University through an Enterprise Fund, and the Ministryof Education, Science, Sports and Culture (Japan) Grant-in-Aid in Sci-ence No. 18340052. J.K. and N.K. acknowledge support by JSPS Ka-kenhi No. 16340055. J.K. was also supported by a grant for theInternational Mission Research provided by the Institute for Spaceand Astronautical Science (ISAS/JAXA). T.M. acknowledges supportby Grants-in-Aid for Young Scientists (B) from JSPS (No. 18740154).

References

[1] A.C. Fabian, K.A. Pounds, R.D. Blandford, Frontiers of X-ray Astronomy,Cambridge University Press, Cambridge, 2004.

[2] Production and analysis of polarized X-rays, in: D.P. Siddons (Ed.), SPIE Proc.1548, 1991.

[3] F. Lei et al., Space Sci. Rev. 82 (1997) 309.[4] M.C. Weisskopf et al., Ap. J. 208 (1976) L125.[5] M.C. Weisskopf et al., Ap. J. 220 (1978) L117.[6] E.H. Silver et al., Ap. J. 225 (1978) 221.[7] K.S. Long et al., Ap. J. 238 (1980) 710.[8] W. Coburn, S. Boggs, Nature 423 (2003) 415.[9] R.E. Rutledge, D.B. Fox, MNRAS 350 (2004) 1272.

[10] C. Wigger et al., Ap. J. 613 (2004) 1088.[11] D.R. Willis et al., A& A 439 (2005) 245.[12] M.J.L. Turner, G. Hasinger (Eds.), SPIE Proc. 6266, 2006.[13] N. Produit et al., Nucl. Instr. Meth. A 550 (2005) 616.[14] J. Legere et al., SPIE Proc. 5898 (2005) 413.[15] Y. Kishimoto et al., IEEE Trans. Nucl. Sci. 54 (2007) 561.[16] R. Silva et al., SPIE Proc. 4843 (2003) 543.[17] V. Andersson et al., Proceedings of the 22nd Texas Symposium at Stanford

University, <http://www.slac.stanford.edu/econf/C041213/proc/>.[18] K. Gunderson et al., SPIE 5165 (2004) 159.[19] E. Pascale et al., 2008, <arXiv:0711.3465v2>.[20] T. Kamae et al., SPIE Proc. 1734 (1992) 2.[21] T. Kamae et al., IEEE Trans. Nucl. Sci. 40 (1993) 204.[22] T. Takahashi et al., SPIE Proc. 1734 (1992) 44.[23] T. Takahashi et al., IEEE Trans. Nucl. Sci. 40 (1993) 890.[24] S. Gunji et al., Ap. J. 397 (1992) L83.[25] S. Gunji et al., Ap. J. 428 (1994) 284.[26] S. Miyazaki et al., PASJ 48 (1996) 801.[27] N. Yamasaki et al., Ap. J. 481 (1997) 821.[28] T. Kamae et al., SPIE Proc. 2806 (1996) 314.[29] K. Makishima et al., in: H. Inoue, H. Kunieda (Eds.), ASP Conference

Proceedings, vol. 251, 2001, p. 564.[30] M. Kokubun et al., IEEE Trans. Nucl. Sci. 51 (2004) 1991.[31] T. Takahashi et al., PASJ 59 (2007) 35.[32] M. Kokubun et al., PASJ 59 (2007) 53.[33] C. Marini Bettolo et al., in: Proceedings of the International Cosmic Ray

Conference, Merida, Mexico, 2007.[34] T. Takahashi et al., Proceedings of the 1st International Space Wire Conference

at Dundee University, 2007, <http://spacewire.computing.dundee.ac.uk/proceedings>.

[35] R. Brun et al., CERN Data handling Division Report DD/EE/84-1, 1987.[36] K. Amako et al., in: Proceedings of CHEP94, LBL-35822 CONF-94049, 1994.[37] T. Mizuno et al., Ap. J. 614 (2004) 1113.[38] V.V. Verbinski et al., Nucl. Instr. Meth. 65 (1968) 8.[39] Y. Uwamino et al., Nucl. Instr. Meth. 204 (1982) 179.[40] T. Koi et al., in: Proceedings of Hadronic Shower Simulations Workshop,

September 2006, Batavia, USA, published as AIP Conference Proceedings, 2007,vol. 89, p. 216.

[41] J. Kazejev, Master’s thesis, Royal Institute of Technology (KTH), 2007.[42] T.W. Armstrong, K.C. Chandler, J. Barish, J. Geophys. Res. 78 (1973)

2715.[43] T.W. Armstrong et al., in: 1997 Conference on High Energy Radiation

Background in Space, Snowmass, USA, available from IEEE Xplore, 1997.[44] J. Cugnon, C. Volant, S. Vuillier, Nucl. Phys. A 625 (1997) 729.[45] S. Larsson et al., in: Proceedings of the 18th ESA Symposium on European

Rocket and Balloon Programmes and Related Research, Visby, Sweden, 3–7June 2007, ESA SP-647, 2007.

[46] T. Mizuno et al., Nucl. Instr. Meth. A 540 (2005) 158.[47] J. Kataoka et al., SPIE Proc. 5898 (2005) 133.[48] Y. Kanai et al., Nucl. Instr. Meth. A 570 (2007) 61.[49] H. Takahashi et al., 2008, in preparation.[50] M. Kiss et al., Licentiate Thesis, Royal Institute of Technology (KTH), 2008.[51] I.S. Vashakidze, Doklady Ak. Nauk. USSR 94 (1954) 1021 (cited in Oort–

Walraven, 1956).[52] J.H. Oort, T. Walraven, Bull. Astron. Inst. Netherlands 12 (1956) 285.[53] L. Woltjer, Bull. Astron. Inst. Netherlands 13 (1957) 301.[54] I. Shklovsky, Dokl. Akad. Nauk USSR 90 (1953) 983.[55] M. Oda et al., Ap. J. 148 (1967) L5.[56] B. Aschenbach, W. Brinkmann, A& A 41 (1975) 147.[57] M.J. Rees, J.E. Gunn, MNRAS 167 (1974) 1.[58] K. Makishima et al., Space Sci. Rev. 30 (1981) 259.[59] R.M. Pelling et al., Ap. J. 319 (1987) 416.[60] M.C. Weisskopf et al., Ap. J. 536 (2000) L81.[61] J. Arons, Adv. Space Res. 33 (2004) 466.[62] C.-Y. Ng, R.W. Romani, Ap. J. 601 (2004) 479.[63] Y. Nakamura, S. Shibata, MNRAS 381 (2007) 1489.[64] D.L. Zanna et al., A& A 453 (2006) 621.[65] V. Radhakrishnan, D.J. Cooke, Astrophys. Lett. 3 (1969) 225.[66] S. Sturner, C. Dermer, C. Michel, Ap. J. 445 (1995) 736.[67] J. Daugherty, A. Harding, Ap. J. 458 (1996) 278.[68] K.S. Cheng, C. Ho, M.A. Ruderman, Ap. J. 300 (1986) 500.[69] R. Romani, I. Yadigaroglu, Ap. J. 438 (1995) 314.[70] R. Romani, Ap. J. 470 (1996) 469.[71] J. Dyks, B. Rudak, Ap. J. 598 (2003) 1201.[72] J. Dyks, A.K. Harding, B. Rudak, Ap. J. 606 (2004) 1125.[73] Y. Tanaka, W. Lewin, in: W. Lewin et al. (Eds.), X-ray Binaries, Cambridge

University Press, Cambridge, 1995, p. 126.

84 T. Kamae et al. / Astroparticle Physics 30 (2008) 72–84

[74] P. Charles, in: M. Abramowicz et al. (Eds.), Theory of Black Hole AccretionDisks, Cambridge University Press, Cambridge, 1998, p. 1.

[75] A.A. Zdziarski, M. Gierlinski, Prog. Theor. Phys. 155 (Suppl.) (2004) 99.[76] N.I. Shakura, R.A. Sunyaev, A& A 24 (1973) 337.[77] A.A. Esin, J.E. McClintock, R. Narayan, Ap. J. 489 (1997) 865.[78] A.P. Lightman, T.R. White, Ap. J. 335 (1988) 57.[79] G. Matt, MNRAS 260 (1993) 663.[80] J. Poutanen, K. Nagendra, R. Svensson, MNRAS 283 (1996) 892.[81] S. Markoff, M. Nowak, J. Wilms, Ap. J. 635 (2005) 1203.[82] M. Axelsson et al., Astropart. Phys. 28 (2007) 327. Available from:

<arXiv:0704.1603>.[83] O. Engdegård, Master’s thesis, Royal Institute of Technology (KTH), 2006.[84] L. Hjalmarsdotter et al., MNRAS 384 (2008) 278.[85] A. Harding, in: S.S. Holt, C.S. Day (Eds.), Proceedings of Maryland Conference

on Evolution of X-Ray Binaries, AIP, New York, 1994, p. 429.

[86] J. Trümper et al., Ap. J. 219 (1978) L105.[87] W. Coburn et al., Ap. J. 580 (2002) 394.[88] P. Meszaros et al., Ap. J. 324 (1988) 1056.[89] N. Mowlavi et al., A& A 451 (2006) 187.[90] C. von Montigny et al., Ap. J. 440 (1995) 525.[91] M. Sikora et al., Ap. J. 625 (2005) 72.[92] J. Poutanen, Ap. J. 92 (1994) 607.[93] M. Sikora, M. Begelman, M. Rees, Ap. J. 421 (1994) 153.[94] D. Lazzati, in: T. Bulik, B. Rudak, G. Madejski (Eds.), Astrophysical Sources of

High Energy Particles and Radiation, AIP, New York, 2005, p. 179.[95] J. Kataoka et al., Ap. J. 514 (1999) 138.[96] B. Giebels et al., Ap. J. 571 (2002) 763.[97] J. Kataoka et al., Ap. J. 528 (2000) 243.[98] R. Fender, T. Belloni, E. Gallo, MNRAS 355 (2004) 1105.[99] A.A. Zdziarski, MNRAS 360 (2005) 825.