Comparison of simulated backgrounds with in-orbit observations for HE, ME and LE onboard Insight-HXMT
Juan Zhang, Xiaobo Li, Mingyu Ge, Haisheng Zhao, Youli Tuo, Fei Xie, Gang Li, Shijie Zheng, Jianyin Nie, Liming Song, Aimei Zhang, Yanji Yang, Yong Chen
CComparison of simulated backgrounds with in-orbitobservations for HE, ME and LE onboard
Insight-HXMT
Juan Zhang • Xiaobo Li • Mingyu Ge • Haisheng Zhao • Youli Tuo • Fei Xie • Gang Li • Shijie Zheng • Jianyin Nie • Liming Song • Aimei Zhang • Yanji Yang • Yong Chen Abstract
Insight-HXMT , the first X-ray astronomicalsatellite in China, aims to reveal new sources in theGalaxy and to study fundamental physics of X-ray bi-naries from 1 keV to 250 keV. It has three collimatedtelescopes, the High Energy X-ray telescope (HE), theMedium Energy X-ray telescope (ME) and the Low En-ergy X-ray telescope (LE). Before the launch, in-orbitbackgrounds of these three telescopes had been esti-mated through Geant4 simulation, in order to investi-gate the instrument performance and the achievementof scientific goals. In this work, these simulated back-grounds are compared with in-orbit observations. Goodagreement is shown for all three telescopes. For HE, 1)the deviation of the simulated background rate aftertwo years of operation in space is ∼
5% from the obser-vation; 2) the total background spectrum and the rela-tive abundance of the ∼
67 keV line show long-term in-creases both in simulations and observations. For ME,
Juan ZhangXiaobo LiMingyu GeHaisheng ZhaoYouli TuoFei XieGang LiShijie ZhengJianyin NieLiming SongAimei ZhangYanji YangYong [email protected] Key Laboratory of Particle Astrophysics, Institute of High En-ergy Physics, Chinese Academy of Sciences, Beijing 100049,China University of Chinese Academy of Sciences INAF-IAPS, via del Fosso del Cavaliere 100, I-00133 Roma, Italy
1) the deviation of simulated background rate is within ∼
15% from the observation, and 2) there are no obvi-ous long-term increase features in the background spec-tra of simulations and observations. For LE, the back-ground level given by simulations is also consistent withobservations. The consistencies of these comparisonsvalidate that the
Insight-HXMT mass model, i.e. spaceenvironment components and models adopted, physicsprocesses selected and detector constructions built, isreasonable. However, the line features at ∼ Keywords
Insight-HXMT, Geant4 simulation, back-ground observation
Due to atmospheric absorption, X-ray radiation of as-trophysical sources needs to be detected in space. How-ever, satellite-borne detectors suffer enormous spaceradiation including cosmic rays, diffuse X-rays, so-lar flares, the albedo of the Earth, charged particlestrapped in the radiation belts and so on. Besides caus-ing damage to sensitive detectors, the space radiationalso results in background events during scientific ob-servations of target sources. The space-induced back-ground varies with each instrument, according to thedetector type and operation orbit (e.g. Jahoda et al.2006; Rothschild et al. 1998; Fukazawa et al. 2009; Tawaet al. 2008). The in-orbit background of each spaceinstrument has to be estimated before the launch inorder to optimize the instrument design and to investi-gate how well it can fulfill the scientific goals. Geant4(Agostinelli & Geant4 Collaboration 2003; Allison et al. 2006, 2016) is a general toolkit to simulate the in-teraction of particles with matter. It is widely usedin nuclear physics and accelerator physics, as well asmedical science and space science. Currently, it is alsoa popular tool to predict the in-orbit backgrounds ofspace instruments (e.g. Tenzer et al. 2010; Perinati etal. 2012; Fioretti et al. 2016; Weidenspointner, Pia,& Zoglauer 2008; Campana et al. 2013; Xie & Pearce2018). The accuracy of simulated backgrounds couldbe examined by in-orbit observations obtained afterthe launch. This kind of examination can validate thesimulation methods and the assumed models. For ex-ample, Mizuno et al. (2004) validated the cosmic raybackground flux models based on a
GLAST balloonfight experiment. And Odaka et al. (2018) verifiedthat a simulation process of proton-induced radioac-tivation background could describe background mea-surements of Hitomi/Hard X-ray telescopes very wellthrough comparison between simulations and measure-ments.As the first Chinese X-ray astronomical satellite,the Hard X-ray Modulation Telescope (
HXMT ), alsonamed
Insight-HXMT , was launched into a low-Earthorbit with an altitude of ∼
550 km and an inclination of ∼ ◦ on 15th June 2017 (Li et al. 2018). It aims to scanthe Galactic plane for new sources and to study funda-mental physics of X-ray binaries (Li 2007; Zhang et al.2020). The three scientific payloads, the High EnergyX-ray telescope (HE, 20–250 keV), the Medium EnergyX-ray telescope (ME, 5–30 keV) and the Low EnergyX-ray telescope (LE, 1–15 keV), are slat-collimated in-struments and co-aligned. Each has different sized fieldof view (FOV) and rotation angle, with the aim to sub-tracting in-orbit background using the combined FOVmethod (Jin et al. 2010). A mass model was built inthe framework of Geant4 to estimate the in-orbit back-grounds of these three payloads (Xie et al. 2015; Li etal. 2015). The simulation results showed that the es-timated background flux of HE was comparable to thebackground measurements of RXTE/HEXTE (Xie etal. 2015).
RXTE/HEXTE was selected for comparisonbecause it was also a slat-collimated instrument, withthe same kind of scintillators (NaI/CsI) and a simi-lar operation orbit. A big amount of data on
Insight-HXMT background observations has been accumulatedso far. In this work, we analyze the background mea-surements of HE, ME and LE, and make comparisonswith previously simulated results, with the motivationto examine the mass model. It is worth noting that thebackgrounds given by simulations are an average esti-mate after a long-term operation in space and it cannotbe used in data analysis for any practical observations.The paper is structured as follows. In Section 2, themass models of the
Insight-HXMT
HE, ME and LE are
Fig. 1
The whole structure of Insight-HXMT. HE is sur-rounded by ACDs. ME consists of three boxes which are inblue color. LE consists of three boxes in green. The graycubic at the bottom is the service module. briefly introduced. This is followed by Section 3, wherebackground observations are analyzed and results be-tween simulations and observations are compared. Thediscussion and conclusion are given in Section 4.
The mass model of
Insight-HXMT has been built underthe framework of Geant4 Version 9.4.p04 (Xie et al.2015) and is updated to Version 10.05.p01 currently.Three objects that needed to be specified are detectorconstructions, primary particles and physics processes,and they are described as follows.2.1 Detector ConstructionsThe whole satellite and detailed structures of HE, MEand LE are shown in Fig. 1, which is generated byGeant4 visualization. HE is in the middle of the plat-form. It is surrounded by 18 anti-coincident detectors(ACDs). ME and LE are at the two edges of the plat-form. Both ME and LE have three boxes, which areillustrated in blue and green colors respectively.HE is equipped with NaI(3.5 mm)/CsI(4 cm) scintil-lators. The collimators of HE are made of aluminumalloy and tantalum. ACDs above and around HE are
Fig. 2
Top-down view of the collimators of 18 HE modules.The white grids are the collimator slats. The top-right one(a white empty circle) is the blind FOV collimator. Thetwo with sparse grids are large FOV collimators. The other15 are small FOV collimators. This figure is generated byGeant4 visualization directly. used to veto charged particles in space. With pulseshape discriminator (PSD), signals in the CsI could bedistinguished from those in the NaI. This can be usedto reject high energy gamma rays and charged particlebackgrounds. HE consists of 18 detector modules, oneof which has a blind FOV collimator, two have largeFOV (5 . ◦ × . ◦ ) collimators, and the other 15 havesmall FOV (1 . ◦ × . ◦ ) collimators. Arrangements ofthese collimators are shown in Fig. 2. The cut-awayview of one small FOV detector module is presentedin Fig. 3. These figures are generated by Geant4 andillustrate how the mass model of HE is built in detail.Both ME and LE are silicon detectors (Zhang etal. 2020). Three boxes of ME are placed with a ro-tation angle of 120 ◦ relative to each other. The sensi-tive detectors, which aim to detect X-rays from 5 keVto 15 keV, are Si-PINs with thickness of 1 mm and cov-ered by 50 µ m beryllium window. Each box of ME has576 detector pixels. A group of 32 pixels is read outthrough one ASIC. There are 18 ASICs in each MEbox. Silver glue with a thickness of ∼ µ m is used tofix the Si-PIN pixels on the bottom ceramics. The MEcollimators are made of aluminum alloy and tantalum.There are three different FOVs in each box. The largeFOV has a size of 4 ◦ × ◦ , the small FOV is 1 ◦ × ◦ ,and the covered FOV is sheltered by 0.6 mm tantalum.One of the ME boxes is shown in Fig. 4. All the pixelsin one box are divided into 18 modules according to thecollimators and readout ASICs. Each vertical stripe inFig. 4 corresponds to one module. In summary there Fig. 3
The mass model of one HE small FOV detectormodule. The yellow layer in the middle is 4 cm CsI. Thelayer above is 3.5 mm NaI, which is in magenta. The greenlayer above NaI is the 1.5 mm beryllium window. All ofthese are installed inside the crystal holder, which is alu-minum alloy and is shown in orange color. Above the crys-tal holder are collimated slats, which are installed inside analuminum alloy cylinder. The signals in NaI/CsI crystalare read out by PMT, which is illustrated below the crystalholder. are 15 small FOV modules, 2 large FOV modules, and1 blind FOV module in each box of ME.LE uses the swept charge device (SCD) CCD236(Zhao et al. 2019). It has 96 SCDs, with a sensitivethickness of 50 µ m. LE aims to detect X-rays from1 keV to 15 keV. There are eight collimators in eachbox. Every 4 SCDs share one collimator. The grids ofcollimators are made of aluminum alloy, Al7075, whichcontains about 90.3% pure aluminum and 5.5% zinc.A layer of 0.2 mm tantalum is pasted around the col-limator, to reduce cosmic gamma-ray background andcharged particles. On the top of each collimator, to re-duce stray lights, there is a very thin layer of shadingfilm ( C H O N ), with 100 nm aluminum below andabove. Between the shading film and the collimator,there is a layer of nickel mesh to fix the film. The colli-mators of LE have three different FOVs. The large FOVis 4 ◦ × ◦ , the small FOV is 1 . ◦ × ◦ , and the coveredFOV has 1 mm aluminum alloy Al7075 on top. Besidesthese three different FOVs, there is also a very shortcollimator above 4 SCDs inside each LE box, which ex-tends the FOV as large as 50 ∼ ◦ × ∼ ◦ . Giventhe extended FOV, these 12 SCDs in the three boxesare used as sky monitors. In total, there are 20 smallFOV detectors, 6 large FOV detectors, 2 covered FOVdetectors and 4 wide FOV detectors in each box of LE. Fig. 4
The mass model of one ME box, generated byGeant4 visualization. The blocks in green are Si-PIN de-tectors. Each green block shown in this figure contains twominimum Si-PIN pixels. The white grids above these pixelsare the collimator grids. The strip in gray is the shelter thatcovers the FOV of the blind module. The two strips on theleft of this shelter are large FOV modules. The remaining15 strips are small FOV modules.
All these collimators and detectors are surrounded byan aluminum box, which can shield X-ray and cosmicray radiation. The whole structure of LE is shown inFig. 5.
Fig. 5
The mass model of one LE box, generated byGeant4 visualization. The pixels in red are the SCDs. Thegrids in white are the collimator grids. The two green blocksare the shelters used to cover the FOV of blind detectors.The top left unit collimator is used as sky monitor.
The service module is located at the bottom of thesatellite and is used to hold the fuel and other equip-ment like attitude control assemblies, etc. It is made ofhoneycomb boards of aluminum alloy. During the sim-ulations, an increase of the cosmic-ray proton inducedbackgrounds on HE by dozens of counts is noted if thismodule is in the whole satellite construction. The raiseis mainly from the secondary events generated by theenergetic cosmic protons interacted with the massivemodule. Therefore, the service module is necessary and important in the mass model, as shown to be a hollowgray cubic in Fig. 1. It is constructed according to thereal size of its outside dimension with the inner simpli-fied into box structures, and is assigned by the averagedensity derived from the true mass of this module.2.2 Space EnvironmentFor space environment, various components were in-cluded in the simulation, including cosmic X-ray back-ground (CXB), cosmic ray protons and electrons,albedo gamma rays&neutrons, and the trapped protonsin South Atlantic Anomaly (SAA) (For more details ofeach component, please refer to Xie et al. 2015; Li et al.2008, and references therein). According to the pre-vious simulation, CXB, cosmic ray protons and SAAprotons have dominating contributions to the in-orbitbackgrounds of
Insight-HXMT . These components arepresented in more details here. An attitude of zenith-pointing is assumed, therefore the incident CXB andprimary cosmic rays are obscured by the Earth.
CXB is thought to result from unresolved sources out-side the Galaxy. The broken power-law spectrum(Gehrels 1992) is adopted in our simulation, F = . × E − . , E < .
02 MeV0 . × E − . , .
02 MeV < E < . . × E − . , E > . E is in MeV and F is photons cm − s − MeV − sr − . The ∼
7% normalization fluctuation (Revnivt-sev et al. 2003) on angular scales of ∼ In a low-Earth orbit, cosmic ray protons outside theSAA region consist of the primary cosmic ray protons,which are modulated by the solar activity and geomag-netic field, and the secondary protons below the geo-magnetic cutoff rigidity (Gehrels 1992). To obtain aconservative background estimation, a spectrum corre-sponding to the minimum solar modulation (Mizuno etal. 2004) and a high geomagnetic latitude is adopted,as shown in Fig. 6. The orbital modulation of geomag-netic fields and the east-west effect are ignored. Similarto CXB, an isotropic distribution except for the Earthdirection is used.
E, GeV
10 110 - M e V - s r - s - F l u x , m AMS(0.0,0.2)AMS(0.2,0.3)AMS(0.3,0.4)AMS(0.4,0.5)AMS(0.5,0.6)AMS(0.6,0.7)AMS(0.7,0.8)AMS(0.8,0.9)AMS(0.9,1.0)AMS(>1.0)primary
Fig. 6
The primary cosmic ray proton spectrum usedfor the
Insight-HXMT orbit. Also shown are the unfoldeddownward spectra reported by AMS-01 for different geo-magnetic latitudes (Gehrels 1992).
Due to the low-Earth orbit with an altitude of 550 kmand an inclination of 43 ◦ , Insight-HXMT spends ∼ . The orbit-averaged differentialspectrum generated from the AP-8 model for the solarminimum is used, as shown in Fig. 7. The spectrum forsolar maximum is also presented for comparison. Thedirection distribution of SAA protons is assumed to beisotropic and is not blocked by the Earth due to theirlocal origin. E, MeV
10 110 - M e V - s - F l u x , c m AP8 minAP8 max
Fig. 7
The differential spectra of SAA protons for
Insight-HXMT orbit. Insight-HXMT , 1–250 keV, low energy electromagneticphysics is preferred. Under Geant4 Version 9.4.p04, theShielding Physics List was chosen in our mass model.As mentioned in Xie et al. (2015), we added radioactivedecay process and replaced the standard electromag-netic physics with low energy electromagnetic physics.In Geant4 Version 10.5.p01, the radioactive decay pro-cess is already included in the Shielding Physics List,so we only make a minor modification by replacing thestandard electromagnetic physics with the livermoreelectromagnetic physics.2.4 Simulation OutputSensitive detectors are defined for NaI/CsI of HE, Si-PIN of ME and SCD of LE in the
Insight-HXMT mass model. The information on these sensitive de-tectors, such as the deposited energy and its responsetime, is recorded. These signals are classified intoprompt backgrounds and delayed (or radioactive) back-grounds by their response time relative to the incidence.Prompt backgrounds trigger the sensitive detector im-mediately after the incidence. Radioactive backgroundscould trigger signals on sensitive detector hours, daysor even months and years later after the incidence.For these radioactive backgrounds, the deposited en-ergy and the corresponding response time are firstlyrecorded in Geant4 simulation, then integrated alongthe radiation history to obtain the final backgroundlevel. and ROSAT All-Sky Survey Bright Source catalog (Vo-ges et al. 1999) to calculate the significance of eachsource on HE, ME and LE. The significance of celes-tial sources is calculated by using signal to noise ratio(SNR), S/ √ B , where S denotes the source counts and http://isdc.unige.ch/integral/catalog/39/catalog.html B the background counts. The simulated backgroundrates are used in this SNR calculation and an expo-sure time of 10 seconds is assumed. After searchingaround the medium galactic latitudes, twenty-one di-rections were chosen, where no sources exceed a signif-icance of 5 within the FOVs of the three instruments.These regions are illustrated in Fig. 8. We define theseregions as blank sky directions of Insight-HXMT . Posi-tions of these blank sky directions are given in Table 1. -150 -100 -50 0 50 100 150 longitude , degree -80-60-40-20020406080 l a t i t ud e , d e g r ee LEMEHE
Fig. 8
Blank sky directions of
Insight-HXMT , used forbackground observations. The cross points are the sourcesvisible on LE. The green dots are sources visible for ME,and the magenta dots are sources visible for HE. The 21blue circles with a cross in the center are the blank sky di-rections we choose for
Insight-HXMT . And the backgroundobservation directions used for
RXTE/PCA (Jahoda et al.2006) are also shown as the red open circles.
Besides these 21 blank sky directions, pointing ob-servations of some certain sources could be also con-sidered as background observations, such as Cas A andPSR B0540–69 for HE. The continuum spectrum of CasA from 3 to 500 keV could be fitted with a thermalbremsstrahlung plus a power-law components with anindex of 3 . ± .
03 (Wang & Li 2016). Based on thisspectrum, the expected contribution of Cas A to HE at20–300 keV is ∼ . ± .
01 (Ge et al. 2019).A rate of ∼ Table 1
Insight-HXMT blank sky directions. blank sky R.A. Dec
Insight-HXMT software (Zhao et al. 2016). The recommendedselection criteria are used, which are also listed in Ta-ble 2.3.2 HE BackgroundsFig. 9 plots the simulated background spectrum of HEand its constituent components after one year oper-ation in orbit. It is clearly seen that the SAA in-duced background is the most prominent component.The peak structures of this component at ∼
31 keV,56 keV, 67 keV and 191 keV, result from the radioac-tive decay of iodine isotopes caused by SAA trappedprotons. The same structures are presented in thedelay background caused by cosmic ray protons. Forprompt components induced by cosmic ray protons,CXB, and albedo gamma rays, the fluorescence lines http://hxmt.org/index.php/usersp/dataan Table 2
Blank sky observations used in this work. sky regions a observation period b label c exposure d selection criteriaYYYYMMDD (seconds)HE PSR B0540–69 20170719-20170722 20170719 45605 For HE, ME and LE,ELV > ◦ COR > >
300 sTN SAA >
300 sANG DIST < = 0 . ◦ ;for LE, plusDYE ELV > ◦ PSR B0540–69 20170913-20170918 20170913 457481-3,10,12,14,15,19,21 20171204-20171231 20171216 59063PSR B0540–69 20180902-20180914 20180902 83506Cas A 20190713-20190715 20190714 65464Cas A 20190814-20190816 20190815 42146ME 1-4,6,8,10-12,14,15,19-21 20171102-20171231 20171103 1033202-6,11,14 20180611-20181005 20180621 486583-6,10,11,14-16,19,20 20190429-20190626 20190625 53286LE 1-4,6,8,10-12,14,15,19-21 20171102-20171231 20171103 671882-6,11,14 20180611-20181005 20180621 321563-6,10,11,14-16,19,20 20190429-20190626 20190625 29633 a Numbers in this column indicate the corresponding blank sky directions in Table 1. b The start and end dates of each period. Background observations during these periods are chosen. c The labels in this column are used in Figs. 10, 11 and 12 to indicate that the corresponding background spectra are obtained fromthe corresponding observation data here. d For the same observation data, HE, ME and LE have different effective exposure time. This is because the data processing has someprocedures related to the intrinsic instrument performance itself. around 57 keV, which originates from HE tantalum col-limators, are clearly seen. All of these line structuresin background spectrum could be used in the in-orbitcalibration of HE (Li et al. 2020).
20 30 40 50 60 70 100 200 300
E, keV - k e V - c oun t s s cxbcrpalbSAAcrp delaytotal simulation Fig. 9
Simulated spectrum of the averaged background ofHE after 1 year operation in space. This background spec-trum consists of prompt background induced by cosmic X-ray background (cxb), cosmic ray protons (crp), the albedogamma rays of the earth atmosphere (alb), and radioactivebackground caused by SAA trapped protons (SAA) and cos-mic ray protons (crp delay).
After a long term operation in space, the back-ground rate of HE increases, especially at the energyof ∼
67 keV, as presented in Fig. 10. Note that the first peak at ∼
25 keV in the observation spectra is not a realpeak. It is caused by the electronic noise and thresh-old cutoff (Zhao et al. 2020). These effects are notincluded in the simulation process, so this peak is notshown in the simulated spectra. Six post-launch periodswere chosen to compare the simulation and the observa-tion of long-term variation. For observation, spectra areshown at the periods corresponding to about 1 month(20170719), 3 months (20170913), 6 months (20171216),1 year (20180902) and 2 years (20190714&20190815) af-ter the launch. Observation data used in this figureare listed in Table 2. For comparison, the simulatedHE background spectra after an operation time of 1,3, 6 months and 1, 2, 4 years are given. The relativeincrease at ∼
67 keV is obvious, which is due to the longdecay periods of some iodine isotopes. The relativeabundance of ∼
67 keV increases quickly especially inthe first year of operation, and gradually approachesstable.The total observed and simulated background ratesfrom 30 to 300 keV on the seventeen large and smallFOV modules of HE are shown in Table 3. Theaverage background rate after two year operation is(538 . ± .
1) counts s − from the observation, and(563 . ± .
6) counts s − from the simulation. The ob-served background levels correspond to 3 . ∼ . × − counts s − keV − cm − , and the simulated ones3 . ∼ . × − counts s − keV − cm − . For the simu-lated backgrounds of HE, deviations from observations
50 100 150 200 250 300
E, keV - k e V - c oun t s s observation - c m - k e V - c oun t s s - E, keV
50 100 150 200 250 300 - k e V - c oun t s s simulation - c m - k e V - c oun t s s - Fig. 10
The background spectra of HE at different operation periods. Left: the observed spectra from July 19, 2017 toAugust 15, 2019. Right: the simulated spectra at about 1 month ∼ Table 3
The observed and simulated background rates ofHE from 30 to 300 keV. period observation simulation deviation(counts/s) (counts/s) (%)1 month 395.2 469.2 18.73 months 399.9 512.2 28.16 months 461.3 536.0 16.2 ∼ ∼ ∼
30% in the first three months afterthe launch. This deviation reduces to ∼
16% in the sixthmonth, ∼
10% after 1 year operation and ∼
5% after 2year operation. The large deviations in the first threemonths could be attributed to the fact that
Insight-HXMT was still in the performance verification phase.On the other hand, as discussed in Xie et al. (2015),the input cosmic proton spectrum we used in simula-tion corresponded to a high geomagnetic latitude re-gion, which will cause more prompt and delayed back-grounds than that of low geomagnetic latitude regions.While considering that the dominated background com-ponent of HE is induced by SAA protons, the deviationfinally approaches into a small value after a long termoperation in space. In addition, the simulated SAAbackground contributions are obtained by folding de-layed background events output by Geant4 with theaveraged SAA irradiation history for
Insight-HXMT or-bit, not the reality history. This will cause some differ-ence in the beginning, but a stable value is approachedafter a long time accumulation. 3.3 ME BackgroundsDue to electronic noises presented at low energies, theeffective energy range of ME shifts into a bit higher en-ergy band, ∼ ∼ ∼ . × .
10 15 20 25 30 35 40
E, keV - k e V - c oun t s s observation - c m - k e V - c oun t s s -
10 15 20 25 30 35 40
E, keV - k e V - c oun t s s totalcxbcrpalbSAA simulation - c m - k e V - c oun t s s - Fig. 11
Left: The observed background spectra of ME at different operation periods. The black line(20171103) isthe spectrum of ∼ by ∼
50% after veto between Si-PIN pixels. This vetomethod hardly has effects on the component of X-rays.The contributions from CXB before and after veto arealmost the same. This means the veto method betweenpixels could reduce the charged particle background ef-ficiently while keeping the same detection efficiency ofX-rays.The simulated ME background rate after veto from10 to 40 keV on small FOV detectors is (35 . ± . − after 1 year operation, which correspondsto a background level of 1 . × − counts s − keV − cm − . The observed rate is (37 . ∼ . ± . − , i.e. 1 . ∼ . × − counts s − keV − cm − . Thesimulated background is about 15% lower than the ob-served value. In addition, the relative abundance ofsilver lines given by simulation is slightly different fromthose given by observations. As for the silver line, theexact thickness of the silver glue layer is unknown andthe thickness of this layer below different pixels are notthe same. We used an uniform thickness of 14 µ m inthe simulation. For the background level, note thatwe do not consider any electronic and readout pro-cesses in simulation and that the simulation result isobtained after veto between pixels for each simulatedbackground event. This veto method is an ideal andmuch more strict selection criterion than the conditionof “Grade==0”, which is used by default in the ob-servation data analysis procedure. This will cause alower simulated proton background estimation. In ad-dition, the number of pixels that were selected duringobservation data analysis and processing could also in-troduce some difference between simulation and obser-vation as well. Si-PIN pixels with high electronic noise are screened out during observation data analysis. Thebackground counts on the remaining pixels were scaledto all pixels of the 45 small FOV detector modules toobtain the observed rate of ME background. While thesimulated rate are obtained from the counts on smallFOV detectors directly.3.4 LE BackgroundsThe observed and simulated background spectra of LEare shown in Fig. 12. Only the small FOV detectorsare used to obtain these spectra. Similar to ME, theobserved spectra are from different periods after thelaunch, i.e. ∼ ∼ ∼ ∼ (12 . ± .
2) counts s − and the observed rate is(12 . ∼ . ± . − , which corresponds to abackground level of 4 . ∼ . × − counts s − keV − cm − . The long term variation is not seen due to theless contributions from radioactive components. Theline features at 7.5 keV, 8.0 keV and 8.6 keV in the ob-served spectra are assumed to be fluorescence lines ofnickel, copper and zinc, respectively. As mentioned inSection 2, zinc is the main composition of the LE alu-minum alloy collimators. The 8.6 keV line of zinc is also E, keV - k e V - c oun t s s observation - c m - k e V - c oun t s s - E, keV - k e V - c oun t s s totalcxbcrp simulation - c m - k e V - c oun t s s - Fig. 12
LE background spectra. Left: observed spectra. Lines are the same as in Fig. 11, which corresponds to anoperation time of ∼ presented in the simulated spectrum, but line structuresat 7.4 keV and 8.0 keV are not evident. These discrep-ancies may result from the uncertainties of detector con-structions in simulation. For instance, there is a gluelayer between the SCD detectors and the ceramics, butthe composition of this glue is unknown, therefore it isabsent in the mass model. While it has high probabilitythat this layer contains some elements that might showline structures in the background spectrum, as the gluedoes on ME background. The real-time background situation in space is muchmore complex than the simulation. Our simulation,with averaged space environment models and zenithsatellite pointing attitude, could only give a generalbackground estimation. On the one hand, for the spaceenvironments, the real time charged particles are notonly modulated by solar activity and geomagnetic cut-off rigidity, but also affected by some short term turbu-lence. Take the LE background as an example, there areseveral hundred seconds flares below 7 keV presentedin the light curve, which may result from the real-timelow-energy charged particles (Liao et al. 2020). The ob-served ME background rates could vary by up to 50%even in the same COR region (Guo et al. 2020). In ad-dition, the relative direction of the Earth in the FOVhas some effects on the background. The Earth blocksCXB and primary cosmic rays, but it is also a radiator,especially for LE. That is why there is an additional se-lection condition that the bright earth angle is greaterthan 30 ◦ , as given in Table 2. On the other hand, thedetector responses are also affected by the charge trans- fer and readout processes, especially for silicon detec-tors. There are already some available softwares thatmimic these processes as realistically as possible, e.gthe SIXTE software (Dauser et al. 2019). These pro-cesses are beyond the scope of Geant4, and we have notcombined these processes currently.In summary, we compare the simulation and the ob-servation of Insight-HXMT backgrounds in this work.The simulations are based on Geant4 framework. Theobservation results are obtained from the analysis of theblank sky measurements. Deviations of the simulatedbackground rates are within ∼
20% from observationsafter the performance verification phase. For HE, thisdeviation approaches ∼
5% after an operation time oftwo years, and the simulation well depicts the observedlong-term increases of the background spectrum andthe relative abundance of the ∼
67 keV line. For MEand LE, long-term increases are not shown neither insimulation nor in observation, and the background lev-els given by simulations are also consistent with the ob-servations. The predicted line structures in simulatedbackground spectra, i.e. ∼
31 keV, 56 keV, 67 keV and191 keV lines of HE, the silver line of ME and the zincline of LE, are observed after the launch. These lines arevital for the in-orbit calibration. For LE, the absenceof fluorescence lines of nickel and copper in simulatedspectra may result from the uncertainties of mass mod-eling. The agreement of these comparisons indicatesthat the space environment we adopted, physics pro-cesses selected and the detector constructions we builtare reasonable in the
Insight-HXMT mass model. Thedetailed differences between simulation and observationhelp us to study the low-Earth orbit space environmentand the effect of detection process, which will be usefulfor the following EP and eXTP missions on instrumentdesign, background rejection and estimation, etc. Acknowledgements
We thank the referee for help-ful suggestions and comments. J. Zhang thanks YingTan for her help with the ME instrument performanceand Helen Poon for improving the English of the wholetext. This work is supported by the National Nat-ural Science Foundation of China under the grantNos. 11403026, U1838201, and U1938201. The observa-tion data used in this work is from the
Insight-HXMT mission, a project funded by the China National SpaceAdministration (CNSA) and the Chinese Academy ofSciences (CAS). We gratefully acknowledge the supportfrom the National Program on Key Research and Devel-opment Project (grant No. 2016YFA0400801) from theMinister of Science and Technology of China (MOST),and the Strategic Priority Research Program of the Chi-nese Academy of Sciences (grant No. XDB23040400). References
Agostinelli, S., & Geant4 Collaboration 2003, Nuclear In-struments and Methods in Physics Research A, 506, 250Allison, J., Amako K., Apostolakis J., et al. 2006, IEEETransactions on Nuclear Science, 53, 270Allison, J., Amako, K., Apostolakis, J., et al. 2016, NuclearInstruments and Methods in Physics Research A, 835, 186Campana, R., Feroci, M., Del Monte, E., et al. 2013, Ex-perimental Astronomy, 36, 451Dauser, T., Falkner, S., Lorenz, M., et al. 2019, Astron.Astrophys., 630, A66Dean, A. J., Bird, A. J., Diallo, N., et al. 2003, Space Sci.Rev., 105, 285Fioretti, V., Bulgarelli, A., Malaguti, G., et al. 2016, Proc.SPIE, 99056WFukazawa, Y., Mizuno, T., Watanabe, S., et al. 2009, Publ.Astron. Soc. Jpn., 61, S17Ge, M. Y., Lu, F. J., Yan, L. L., et al. 2019, Nature Astron-omy, 3, 1122Gehrels, N. 1992, Nuclear Instruments and Methods inPhysics Research A, 313, 513Guo, C.-C., Liao, J.-Y., Zhang, S., et al. 2020, arXiv e-prints, arXiv:2003.06260Jahoda, K., Markwardt, C. B., Radeva, Y., et al. 2006,Astrophys. J. Suppl. Ser., 163, 401Jin, J., Chen, Y., Zhang, S.-N., et al. 2010, Chinese PhysicsC, 34, 66Li, G., Wu, M., Zhang, S., et al. 2008, Chin. J. Space Sci.28(6), 531Li, G., Xie, F., Zhang, J., et al. 2015, Progress in AstronomyVol. 33, No. 2Li, T., Xiong, S., Zhang, S., et al. 2018, Science ChinaPhysics, Mechanics, and Astronomy, 61, 31011Li, X., Li, X., Tan, Y., et al. 2020, arXiv e-prints,arXiv:2003.06998Li, T.-P. 2007, Nuclear Physics B Proceedings Supplements,166, 131Liao, J.-Y., Zhang, S., Chen, Y., et al. 2020, Journal of HighEnergy Astrophysics, 27, 24Mizuno, T., Kamae, T., Godfrey, G., et al. 2004, Astro-phys. J., 614, 1113Odaka, H., Asai, M., Hagino, K., et al. 2018, Nuclear In-struments and Methods in Physics Research A, 891, 92Perinati, E., Tenzer, C., Santangelo, A., et al. 2012, Exper-imental Astronomy, 33, 39Revnivtsev, M., Gilfanov, M., Sunyaev, R., et al. 2003, As-tron. Astrophys., 411, 329Rothschild, R. E., Blanco, P. R., Gruber, D. E., et al. 1998,Astrophys. J., 496, 538Tawa, N., Hayashida, K., Nagai, M., et al. 2008, Publ. As-tron. Soc. Jpn., 60, S11Tenzer, C., Warth, G., Kendziorra, E., et al. 2010, Proc.SPIE, 77420YVoges, W., Aschenbach, B., Boller, T., et al. 1999, Astron.Astrophys., 349, 389Wang, W., & Li, Z. 2016, Astrophys. J., 825, 102Weidenspointner, G., Pia, M., & Zoglauer, A. 2008, in“IEEE Nuclear Science Symposium Conference Record”,Dresden, Germany, p.2877-2884.Xie, F., & Pearce, M. 2018, Galaxies, 6, 50 Xie, F., Zhang, J., Song, L.-M., et al. 2015, Astrophys.Space Sci., 360, 13Zhang, S.-N., Li, T., Lu, F., et al. 2020, Science ChinaPhysics, Mechanics, and Astronomy, 63, 249502Zhao, H. , Li, C. , Li, X. , et al. 2016, Chinese Journal ofSpace Science, 36, 6Zhao, H. , Li, C. , Li, X. , et al. 2020, Astronomical Research& Technology, in press, chinaxiv.org/abs/202008.00071Zhao, X.-F., Zhu, Y.-X., Han, D.-W., et al. 2019, Journalof High Energy Astrophysics, 23, 23