6 0 0 2 SimulationStudies ofDelta-ray Backgroundsina Compton-Scatter n TransitionRadiationDetector a J 4 JohnF.Krizmanica,b Michael L.Cherryc RobertE. Streitmatterb aUniversitiesSpace Research Association ] t bNASAGoddard Space Flight Center,Greenbelt, MD 20771 USA e cDept.of Physics &Astronomy, Louisiana State University,BatonRouge, LA70803 USA d - s n i . Abstract s c In order to evaluate the response to cosmic-ray nuclei of a Compton-Scatter Transition Radiation Detector in i s the proposed ACCESS space-based mission, a hybrid Monte Carlo simulation using GEANT3 and an external y transition radiation (TR) generator routine was constructed. This simulation was employed to study the effects h ofdelta-rayproductioninducedbyhigh-energynucleiandtomaximizetheratioofTRtoδ-raybackground.The p [ resultsdemonstratetheabilityofaCompton-ScatterTransitionRadiationDetectortomeasurenucleifromboron toiron uptoLorentzfactors γ ∼105 takingintoaccount thesteeply falling power-law cosmic ray spectra. 1 v 0 Keywords: Transitionradiation, Compton scattering, delta rays, cosmicrays,ACCESS 2 PACS: 95.55.Vj,96.50.sb, 96.50.Vg 0 1 0 6 1. Introduction The Compton-scatter technique exploits the 0 facts that 1) the x-ray spectra of emitted TR / s The proposed Advanced Cosmic-ray Composi- can be hardened by a judicious choice of radia- c tion Experiment for Space Science (ACCESS) is tor material, thickness, and spacing and 2) the i s a dedicated space-based mission to perform di- greatest energy dependence of the TR is at high y rectnuclearcompositionmeasurementsuptonear frequencies [2]. For a radiator with thickness l1, h p the ‘knee’ of the cosmic-ray spectrum, E ∼ 106 plasma frequency ω1, and spacing l2, interference : GeV/nucleus [1]. One configuration [2] employs a effects from the superposition of amplitudes from v nearly 6 meter3 Compton-Scatter Transition Ra- eachinterfaceyieldpronouncedmaximaandmin- i X diation Detector (CSTRD) in a cubic geometry imum in the TR spectra. The highest frequency r with a smaller calorimeter underneath. The abso- maximum is at ω = l1ω12(1+ρ) while the sat- a max 2πc lute size of the ACCESS instrument is driven by uration energy is γs ≈ 0.6cω1pl1l2(1+ρ), where the requirementto obtain significantevent statis- ρ=0exceptformaterialswithcomplexdielectric tics in this flux-poor energy regime while being constants (i.e., metals) where ρ = 1. (In order to constrained by the achievable spacecraft size and be conservative, the simulations employed in this mass. study assume ρ = 0). Thus an ensemble of alu- Preprintsubmitted toElsevier Science 2February2008 minumfoilradiatorswith100−200µmthickness Table 1. Response of aCSTRD to normallyincident separated by several mm leads to γ ≈ 105 and carbonnuclei withLorentz factors above TR saturation. s The firstnumber ineach columnrepresents the number ~ωmax ≈ 130 keV (for ρ = 0). As the Compton- ofpixel hits per event with50 keV≤Epixel<500 keV scattering cross-section is significant at these en- whilethe second number (in parentheses) gives the pixel ergies, these photons will scatter away from the hits per event without energy selection. initiating particle’s trajectory and thus spatially CsIlayer TR only δ-rays only Sum TR +δ-rays separate the TR signal from the ionization path. Furthermore, the significant flux of x-rays with Top 6.7(7.9) 0.5(0.7) 7.1(8.6) 7.1(8.6) energies > 50 keV can be detected with an in- organic (e.g., CsI) scintillator. The principle of Sides 23.8 (27.9) 1.5(2.3) 25.3(30.2) 26.3(31.4) Compton-scatterTRDshasbeendemonstratedin Bottom 18.3 (22.3) 1.7(4.5) 20.0(26.8) 19.7(26.3) acceleratortest beams[3]. Total 48.8 (58.1) 3.7(7.6) 52.4(65.7) 53.1(66.3) The thick radiators used to produce such hard x-raysalsoinduceδ-rayproduction.Theimpactof thisbackgroundisexacerbatedincosmic-raymea- with an areal size of 160×160 cm2. The radiator surements by the vast number of sub-TR events stackissurroundedbysix160×160cm2 planesof resultingfromthepower-lawnatureofthecosmic- CsI scintillator in a cubic geometry. Each scintil- ray spectrum. The study presented in this paper lator module contains a layer of 2 mm thick CsI, optimizes the CSTRD signal in relation to this 250 µm silicon PIN diode readout, and 5 mm of background. G10 encapsulated in a 1 mm Al housing. The CsI is segmented into 2.5×2.5 cm2 pixels with each pixel recording the deposited energy on an event- by-eventbasiswith a50keVthresholdenergy. 2. SimulationStudies The pixel TR energy distribution peaks at an energyperpixelnear100keVwithnearly100%of A hybrid Monte Carlo simulation has been the Compton-scattered TR signal being recorded constructed using GEANT3 [4] interfaced to an below500keV.Theδ-raypixelenergydistribution external TR generator routine based on the for- showsalowenergypeakbelow100keVwithapro- malismofTer-Mikaelian[5].GEANT3modelsthe nouncedtailextendingwellabove1MeV.Apixel dominantelectromagnetic processesin the energy energy selection of 50 ≤ E < 500 keV retains pixel range 10 keV to 10 TeV with the cross-sections virtually100%oftheTRhitpixelswhilerejecting smoothly extending to higher energies. For heavy > 40% of the δ-ray hits. The pixels recording the ions, ionization losses are simulated along with ionizationoftheprimariesandtheirnearestneigh- thediscreteprocessofδ-rayproduction.Hadronic borsareexcludedinthis pixelcounting. interactions are not simulated for heavy ions Table 1 details the response for normally inci- in GEANT3, and the ACCESS calorimeter and dentcarbonnucleiwithγ >105 forsimulatedTR ancillary charge measurement detectors are not only, δ-ray background only, and signal + back- modeledinthis study. ground.Notethatthepixeloccupancyisatsucha Although multiple CSTRD configurations were low level that summing the individual signal and investigated, the results employing a radiator ar- backgroundsamples equals the resultfor simulta- rangementwiththehighestTRsaturationLorentz neouslysimulatingthetwoprocesses;i.e.,thereare factorarepresented.Theconfigurationconsistsof very few events in which a δ-ray and TR photon alargeradiatorstacksurroundedonallsixsidesby hit the same pixel. The results indicate that the highlysegmentedscintillatorswhichmeasureboth signal is 48.79/3.65 = 13.4 times larger than the the Compton-scattered TR and ionization losses, background. By varying the scintillator thickness thus providing the primary particle charge deter- from 1 to 8 mm, the signal/backgroundratio was mination.Theradiatorconsistsof150Alfoilswith maximized at 2 mm CsI thickness. Furthermore, 150µmthickness separatedby1 cmvacuumgaps a substantial reduction in the δ-ray background 2 s100 s Hit Hit el el 1000 x x Pi Pi er of 80 er of mb mb800 u u N N 60 600 40 400 20 200 0 0 4 5 4 5 10 10 10 10 Lorentz Factor g Lorentz Factor g (cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:0) (cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:0) Fig. 1. The simulated total number of pixels hit for normally incident carbon (left) and iron (right) nuclei. The response to the background δ-ray process aloneis illustratedbythe lower,horizontal distribution. was obtained by placing the scintillators on the no more than βN , then the probability of a signal outside of the radiator stack versus interspersing low-energy event with δ-rays fluctuating upward them within the stack. No significant gain in the to mimic the signal of a valid event above E thr signal/backgroundratiowasobtainedbyrejecting mustbenomorethanP =β(E0/Ethr)α−1.Using pixels further from the trajectory of the primary the conservative values of α = 2.75, β = 0.01, thanjustthe hitpixelplus nearestneighbors. E0 =1GeV,andEthr =105GeV,theprobability Thesesimulationswereperformedusingnuclear is evaluated as approximately 2×10−11 or ∼ 7 σ primariesfromborontoironandincidentenergies assuming a single-sided Gaussian probability dis- chosen from a flat logE distribution to enhance tribution function. Thus the low end of the TR themeasuredhighenergyresponse.Figure1shows dynamic range is defined as that energy where the results for carbon and iron for TR + back- the TR responseis 7 σ abovethe averagevalue of groundandforδ-raybackgroundalone.Thisradi- the sub-TR response. On a power-law decreasing atorconfigurationyieldeda meansaturatedvalue spectrum, the flux of particles at the high end of of∼960TRphotonsforcarbonand∼18,000TR theenergyrangeisdepletedascomparedtolower photons for incident iron nuclei with, on average, energies. Thus, the high end of the TR dynamic ∼27%(20%)oftheTRphotonsatenergiesgreater range can have a more modest 2 σ separation than50(100)keV. betweenthe signalandthe saturatedTRvalue. The effective dynamic range of the TR mea- Table 2 details the effective dynamic ranges for surement is obtained from a statistical analysis. various nuclear species based upon the results of Assuming a power law spectrum ϕ(E) ∼ E−α this analysis. As the data demonstrate, the mod- and a TR threshold energy of E , the num- eledCSTRDcanmeasuretheLorentzfactorsofin- thr ber of events above the TR threshold goes as cident nuclei from boronto iron. Furthemore,the N ∼ E1−α. Some number of sub-threshold rangeofLorentzfactorsthatguaranteesignalsep- signal thr events accompanied by δ-rays will masquerade as arationfromthe sub-TRbackgroundandthe sat- high-energyeventswithTR.Ifthenumberofthese urated value significantly increases as the atomic misidentified low-energy events is required to be numberofthe incidentnucleiincreases. 3 Table2. Dynamicrange of TRmeasurements for various incident nuclei. References Nuclei Effective γlowEffective γhighDynamic Range [1] T.L. Wilson & J.P. Wefel, NASA report TP-1999- 11Boron 2.4×104 4.5×104 1.9 209202 (1999); M.H.Israel etal.,NASA NP-2000-05- 12Carbon 2.2×104 5.2×104 2.4 056-GSFC (2000) 16Oxygen 1.6×104 5.8×104 3.6 [2] M.L. Cherry & G.L. Case, Astropart.Phys. 18, 629 (2003) 28Silicon 1.1×104 7.2×104 6.5 [3] G.L. Caseet al.,NIM A524,257 (2004) 56Iron 7×103 8.3×104 11.9 [4] CERNProgramLibraryLongWriteupW5013 (1993) [5] M.L.Cherry,Phys. Rev. D10,2245 (1978) [6] L. Sihver et al.,Phys RevC47, 1225 (1993) Although hadronic interactions of the incident nuclei are notmodeled for this study, anestimate oftheattenuatingeffectsoftherelativelythickra- diatorsandCsIscintillatorsintheCSTRDcanbe obtained by consideration of the nuclear inelastic cross-sections.Forironprimariesincidentuponan aluminumtarget,thenuclearinteractionlengthis givenas20.3g/cm2 [6]whilethegrammageofthe 2 CSTRDmodeledinthisstudyis11.5g/cm .Thus, approximately 43% of an incident iron flux will be lostdue to hadronicinelastic collisions.Events with charge-changing interactions will be recog- nizedbyobservingadifferenceinthe chargemea- surement of the particles entering versus exiting the detector. 3. Conclusions AMonteCarlosimulationhasbeenemployedto simulate the response of a Compton-scatter tran- sition radiation detector and to optimize the de- sign with respect to the background process of δ- rayproduction.The results indicate thata space- based CSTRD can perform nuclear composition measurementsuptoaLorentzfactorof105.How- ever,whilethedynamicrangeismorethanafactor of10forcosmic-rayironmeasurements,the range is more limited for lower-Z nuclei due to the rela- tive reductioninTRgeneration. 4