Nuclear Instruments and Methods in Physics Research B · PDF filecontinue to be the limiting factor in risk assessment for long-duration mission planning purposes for the foreseeable

Nuclear Instruments and Methods in Physics Research B 267 (2009) 125–138

Contents lists available at ScienceDirect

Nuclear Instruments and Methods in Physics Research B

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

Development and evaluation of the Combined Ion and Neutron Spectrometer (CINS)

C. Zeitlin a,*, R. Maurer b, D. Roth b, J. Goldsten b, M. Grey b

a Lawrence Berkeley National Laboratory, Life Sciences Division, 1 Cyclotron Road, MS 74R0197, Berkeley, CA 94720, USAb Johns Hopkins Applied Physics Laboratory, Laurel, MD, USA

a r t i c l e i n f o

Article history:Received 2 June 2008Received in revised form 7 September 2008Available online 17 October 2008

PACS:07.87.+v29.40.�n25.70.Mn98.70.Sa

Keywords:Galactic cosmic raysDosimetryHeavy ionsNeutronsSpectrometry

0168-583X/$ - see front matter � 2008 Elsevier B.V.doi:10.1016/j.nimb.2008.10.006

* Corresponding author. Tel.: +1 510 486 5518; faxE-mail address: [email protected] (C. Zeitlin).

1 We refer to LET in water, in an infinite volume so thincluded, even high-energy delta-rays that escape smalsometimes written LET1.

a b s t r a c t

The Combined Ion and Neutron Spectrometer, CINS, is designed to measure the charged and neutral par-ticles that contribute to the radiation dose and dose equivalent received by humans in spaceflight. As thedepth of shielding increases, either onboard a spacecraft or in a surface habitat, the relative contributionof neutrons increases significantly, so that obtaining accurate neutron spectra becomes a critical part ofany dosimetric measurements. The spectrometer system consists of high- and medium-energy neutrondetectors along with a charged-particle detector telescope based on a standard silicon stack concept.The present version of the design is intended for ground-based use at particle accelerators; future itera-tions of the design can easily be streamlined to reduce volume, mass, and power consumption to createan instrument package suitable for spaceflight. The detector components have been tested separatelyusing high-energy heavy ion beams at the NASA Space Radiation Laboratory at the Brookhaven NationalLaboratory and neutron beams at the Radiological Research Accelerator Facility operated by ColumbiaUniversity. Here, we review the progress made in fabricating the hardware, report the results of severaltest runs, and discuss the remaining steps necessary to combine the separate components into an inte-grated system. A custom data acquisition system built for CINS is described in an accompanying article.

� 2008 Elsevier B.V. All rights reserved.

1. Introduction butions are calculated for the GCR in unshielded space, the impor-

The space radiation environment presents potential health haz-ards to crew and equipment [1–4] owing to the presence of ener-getic charged particles and, in shielded environments, neutrons.In particular, the Galactic Cosmic Rays (GCR) are a source of a moreor less constant flux of energetic nuclei, on the order of 0.2–0.5 par-ticles per cm2 per second per steradian [5]. Many GCR have highenergy (near solar maximum, about half the particles have kineticenergies above 1 GeV/nucleon) and are therefore able to penetratesignificant depths of shielding. In terms of abundance, protonsdominate the flux (85–90%), helium ions make up most of the rest,and highly-charged ions such as silicon (charge Z = 14) and iron(Z = 26) are rare, representing about 1% of the total. However, be-cause the dose delivered by an energetic particle is proportionalto the square of the charge, the heavy ions make a substantial con-tribution to dose. To relate the physical properties of a mixed-fieldradiation environment to biological risk, dose equivalent is calcu-lated by integrating the quality factor Q [6] against the linearenergy transfer (LET or L) spectrum.1 When dose equivalent contri-

All rights reserved.

: +1 510 486 6949.

at all ionization electrons arel volumes such as cells. This is

tance of heavy ions grows even larger, and iron accounts for thesingle largest contribution of any ion species. This is because theLET of high-energy iron ions in water is about 150 keV/lm, closeto the peak of Q(L) at 100 keV/lm.

The other main source of energetic particles in deep space is thesun, when it produces solar particle events (SPEs). SPEs are spo-radic, being more likely near solar maximum, and predominantlyproduce protons at relatively low energies. Typically the flux fallsrapidly with energy and is negligible above 100 MeV; such a spec-trum is said to be ‘‘soft.” In some instances SPEs can be extremelyintense and/or have a ‘‘hard” spectrum with significant fluxesabove 100 MeV [7]. Some events [8] contain significant fluxes ofhigh-energy heavy ions.

Efficient mechanisms exist that allow DNA to correctly repair it-self when there is simple damage along a strand [9]; this sort ofdamage can be caused by, among other things, the low-LET radia-tion from cosmic ray muons on the surface of the Earth. However,high-LET radiation can produce more complex damage sites [10].These sites may be unrepairable, leading to cell death, or, worse,they may be mis-repaired, possibly leading to mutagenic changes.Data on the biological effects of high-LET radiation have histori-cally been sparse, and effectiveness as a function of LET is stronglydependent on the system and endpoint studied. Thus high-LETradiation presents a source of risk that is difficult to quantify withconfidence. Uncertainties in biological response are and will

mailto:[email protected]

http://www.sciencedirect.com/science/journal/0168583X

http://www.elsevier.com/locate/nimb

126 C. Zeitlin et al. / Nuclear Instruments and Methods in Physics Research B 267 (2009) 125–138

continue to be the limiting factor in risk assessment for long-duration mission planning purposes for the foreseeable future[11].

Although energetic, highly-charged ions have high LET, theirfluxes and their contribution to risk can be reduced by shielding,particularly if hydrogenous materials are used, for reasons ex-plained elsewhere [12–14]. The mechanism for the reduction ofdose and risk is nuclear fragmentation, in which the projectile nu-cleus is broken up into lighter, less ionizing particles. Neutrons arealso produced in these interactions and become increasinglyimportant as shielding depth increases [15]. Though neutrons arenot directly ionizing, they deposit energy (and therefore deliver adose) through their interactions with nuclei. In tissue, interactionswith hydrogen are particularly important because the full energyof the neutron can be transferred, giving a proton considerablerange.

Because they do not lose energy through the ionization process,neutrons are more difficult to attenuate than charged particles.There are two principal ways in which neutrons are created in nu-clear interactions. In nucleus–nucleus collisions, neutrons can beremoved from the projectile, emerging from the interaction withvelocities similar to that of the projectile, modified by the Fermimotion of the nucleons. These neutrons are sometimes referredto as ‘‘projectile-like.” The second mechanism is evaporation fromtarget nuclei that receive significant excitation energy in the colli-sion with the projectile. These excited states decay by (isotropic)emission of nucleons. Although these target-fragment neutronshave low energy, they can escape the target or shield and contrib-ute to the dose in tissue.

Measurement of the mixed radiation field in space requires ahighly capable instrument or instrument suite that must be de-signed and implemented within severe mass and power con-straints and many technological advances have been made as aresult of these constraints. An instrument suite with similar capa-bilities but intended for ground-based applications is free of themass and power constraints, but nonetheless many advantagescan be gained from incorporating flight-tested or flight-like com-ponents into ground-based systems. CINS benefits from technologydeveloped for spaceflight, particularly in the area of data acquisi-tion, with the result being a compact, portable system, optimizedfor making measurements at heavy ion accelerators.

2. Detectors and electronics

Given the complexity of the space radiation environment, par-ticularly behind moderate depths of shielding, detailed measure-ments will require multiple types of detectors, each optimized tomeasure a specific component of the environment. CINS thereforecontains three sub-systems: a large plastic scintillator which mea-sures neutrons in the energy range from 1 to 15 MeV, where theirbiological effectiveness is high; a silicon detector system with aveto counter for the measurement of high energy neutrons; anda silicon detector stack augmented with three scintillators, twoof them thin plastic and one fairly thick BGO crystal, for detectionand identification of charged particles from protons to iron. As thesub-systems are not yet integrated, we have performed separatetests on each, which are described below.

There is considerable commonality in the readout requirementsof the different systems, particularly between the two systemsbased on silicon detectors. These detectors require a charge-sensi-tive preamplifier, located close to the detector itself to minimizecable lengths and the attendant noise pickup, and generally a shap-ing amplifier is used to improve performance. The job of the shaperis to filter and amplify the output of the preamplifier to improvethe signal to noise ratio, and – because fall times from the pream-

plifiers can be long – to restore the signal to baseline soon after thepeak.

2.1. Neutron detectors

2.1.1. Eljen fast neutron detector systemFor the Neutron Energy Spectrometer, JHU/APL has designed,

procured and fabricated a 12 cm � 12 cm Eljen plastic boron-loaded scintillator for detection of medium-energy 1–15 MeV neu-trons that have the maximum dose equivalent weighting factors of10–20 [6]. The large scintillator is mated to a photo-multiplier tubeand the detector system uniquely identifies ‘‘fast” neutrons byobserving a combination of a recoil or scatter pulse and a subse-quent 10B capture pulse that occur within a narrow time window.The average time separation of the pulses is about 2 ls. This detec-tor system was calibrated at Columbia University’s RadiologicalResearch Accelerator Facility (RARAF) in November 2006 and wasused at NSRL for the first time in May 2007.

The medium-energy neutron detector system uses a uniquetrigger technique to positively identify neutrons. The two pulsesdescribed above, with their particular time and amplitude restric-tions, distinguish the waveform produced by a neutron in the en-ergy range of interest from waveforms produced by otherparticle types. The first pulse represents the energy of the incomingneutron which interacts and slows down within the detector; thesecond pulse represents the boron capture of the slowed neutron.The timing of the second pulse with respect to the first pulse is crit-ical. The allowable time window is based on a few criteria, the firstof which is the e-folding time of a capture given the size of thedetector and the amount of 10B present. Dead time and observa-tions during calibration help determine what time restriction touse. The second pulse also has restrictions on both its minimumand maximum height. This trigger technique is less efficient atdetecting neutrons than use of other anti-coincidence techniques,but given the size and volume of the detector it is a reasonable ap-proach. The trigger has been implemented in NIM electronics forour accelerator experiments. It is based on heritage from the Gam-ma-Ray Neutron Spectrometer (GRNS) flying on the MESSENGERspacecraft [16], in which the full trigger logic is implemented inthe FPGA firmware.

2.1.2. High-energy neutron system (HENS)A 5 mm thick lithium-drifted silicon detector is sensitive to

neutrons with energies greater than 10 MeV via elastic andinelastic neutron–silicon interactions that produce charged parti-cles within the active volume of the detector. Charged silicon re-coils (elastic reaction) and secondary particles such as protons,alpha particles and residual nuclei (inelastic reactions) depositcharge in the detector. This charge is subsequently collectedand measured by standard pulse height techniques. Thickerdetectors offer increased efficiency for neutron detection by con-taining a greater number of possible interaction sites for the neu-tron, and consequently presenting a greater fraction of the meanfree path to incident neutrons. This fact is especially important asthe neutron energy increases to tens of MeV. The efficiency of the5 mm thick detector was determined by exposure to mono-ener-getic neutron beams in experiments at the Columbia UniversityRARAF. The measured efficiency is about 5% in the interval from5.89 to 18.5 MeV. The agreement between our data and modelsindicates that we can achieve efficient detection (3% or greater)of neutrons with energy up to about 50 MeV [17]. Since the crosssection for (n + A) reactions falls with energy in this region, theefficiency of the method also falls, to about 1–2% at 100 MeV[18].

To separate the neutrons from the other incident radiation suchas charged particles and gamma rays requires an anti-coincidence

C. Zeitlin et al. / Nuclear Instruments and Methods in Physics Research B 267 (2009) 125–138 127

shield in front of the thick silicon detector. For accelerator experi-ments, where the neutrons and charged particles are predomi-nantly coming from the direction of the target, a plasticscintillator paddle placed between the target and the silicon detec-tor can provide efficient discrimination. In a spaceflight environ-ment, more complete coverage – as close to 4p steradians aspossible – must be implemented.

HENS measures energy deposited in silicon detectors by sec-ondary particles produced in nuclear reactions, and not the inci-dent neutron energy itself. The energy of any single neutroncannot be recovered from the deposited energy. However, the pro-cesses which create secondary particles and the probability ofdepositing a given amount of energy from a reaction are wellknown as a function of neutron energy. Given these probabilities,it is possible to estimate, or unfold, the incident neutron energyspectrum from the energy deposition spectrum in silicon. For theHENS data unfolding, the known response function of the detectoris used. The process involves inverting a matrix equation for theincident neutron spectrum. The first step in the conversion processis to apply the energy calibration. The second step is to use the re-sponse function for the solid-state detector determined throughmodeling and experiment to calculate the maximum likelihoodincident neutron spectrum from the deposited energy spectra data.Details of the deconvolution procedure can be found elsewhere[18].

In earlier versions of HENS dating to 1999, the 5 mm thick sili-con detectors had an active area of 2 cm2; however in 2006 wewere able to obtain refurbished (re-drifted and with guard ringsapplied) detectors of the same thickness but with an active areaof 10.75 cm2 area which increases our count rate for the sameexposure times. An example of the increased efficiency is shownbelow in Section 4.4.

2 Because the detectors present considerable mass, some low-energy particles stopin the stack, allowing for charge, mass, and energy identification. Slightly moreenergetic particles slow down in the stack and leave distinct patterns of energydeposition, allowing for the determination of charge and energy. Further, evenminimum ionizing protons deposit enough energy in a thick detector to bring thesignal well out of the noise.

2.2. Charged-particle stack

Compared to neutrons, the detection and identification ofthe dosimetrically important charged particles in spaceflight isstraightforward – one simply needs accurate measurements ofenergy deposition, with the principal requirement being a widedynamic range in order to accommodate particles from minimum-ionizing protons (dE/dx in silicon of about 0.3 MeV/mm) to slowiron (dE/dx in silicon of 570 MeV/mm for an ion with 200 MeV/nucleon of kinetic energy). Based on these cases, a dynamic rangecapability of 10,000:1 or greater is needed.

Our starting point for the design was the MARIE detector [19]aboard Mars Odyssey [20]. Built by the NASA Johnson Space Center,and similar to instruments flying aboard the International SpaceStation, MARIE is a stack of silicon detectors including two fairlythin detectors (1 mm deep) used for triggering and for the defini-tion of the telescope’s geometry, two strip detectors, and four thick(5 mm deep) lithium-drifted detectors that are the primary particleidentifiers. MARIE also has a Cherenkov detector at the bottom ofthe stack, but due to a design error it did not produce useful data.The MARIE electronics have an excessively high gain, so that manyparticles of interest (ions with dE/dx in silicon above about60 MeV/mm) cause saturation. The pulse-height readout was se-verely rate limited, so that even the nominal GCR trigger rate pre-sented problems. The rate was reduced by raising the triggerthreshold, but this meant that the large majority of high-energyprotons were missed, as their energy depositions were belowthreshold.

The two major problems in the implementation of the MARIEdesign – excessive preamplifier gain and the slow readout – areboth in the electronics; there is nothing intrinsically wrong withthe design of the stack, and in fact the use of thick silicon detectors

for particle identification has some advantages.2 That part of thestack design was therefore retained, and other details were modifiedas follows: the two thin silicon detectors used for triggering were re-placed by thin plastic scintillators, each optically coupled to a com-pact Hamamatsu photosensor module; the silicon strip detectorswere eliminated; and the Cherenkov detector was replaced by aBGO crystal, which is optically coupled to a standard photomultipliertube and base. Details on each of these components, and the ratio-nale behind each choice, are given in the following. The nominal lay-out of the stack is shown in Fig. 1.

The event trigger, described in the accompanying article aboutthe data readout, is flexible. Nominally, it is expected to be a coin-cidence of hits above threshold in the two plastic scintillators. Thecoincidence requirement determines the geometric factor of thetelescope. In the nominal stack configuration (which is also flexi-ble), the order of detectors is, first, a plastic scintillator (P1), thentwo silicon detectors (Si1 and Si2), the second plastic scintillator(P2), followed by the last two silicon detectors (Si3 and Si4), fol-lowed by the BGO crystal. The detectors are spaced on a pitch of4 cm, giving a geometric factor of 0.49 cm2 sr for the viewing angledefined by P1 and P2, for particles entering from the front of thestack. The threshold energy for a proton at normal incidence toreach P2 and fire the coincidence trigger in this configuration isabout 47 MeV. If the nominal positions of the second scintillatorand second silicon detector were exchanged, the geometric factorwould increase by more than a factor of two, to 1.06 cm2 sr, andthe proton energy threshold for reaching P2 would be reduced toabout 33 MeV. This may be a more suitable configuration forspaceflight, however for present purposes (ground-based measure-ments at accelerators), the configuration shown in Fig. 1 isacceptable.

2.2.1. Silicon detectorsSilicon detectors have a long and successful history in space-

flight applications. For example, the cosmic ray detectors aboardthe Voyagers 1 and 2 spacecraft [21] have operated for 30 yearsin deep space and continue to function as of this writing. This out-standing record, along with the well-understood properties of sili-con detectors, make them a low-risk, high-reliability choice for thisapplication. Silicon detectors have also been used successfully inaccelerator-based experiments to measure GCR-like heavy ions(see [22] and references therein).

The choice of method to be used for obtaining the charged-par-ticle dose using silicon detectors depends on the environment, butin all cases depends on the energy loss recorded. This in turn isgoverned by the Bethe–Bloch equation.

dEdx¼ �k

Zmat

Amat

� �Zb

� �2

ln2mec2b2c2

I0

!� b2 � d

2

" #

where k is a constant, me is the electron mass, c is the speed of lightin vacuum, Zmat, Amat and I0 refer to the atomic number, mass num-ber, and ionization potential of the medium being traversed, andthe remaining variables, Z, b and c, refer to the projectile. The quan-tity d/2 represents the density effect [23]. In practice, one does notmeasure dE/dx, but rather the energy deposited in a finite detectordepth, DE/Dx. These can differ as some of the energy lost by the pro-jectile goes into the creation of energetic delta rays that can escapethe detector volume.

Fig. 1. Schematic drawing of the charged-particle stack as used in accelerator experiments. Beam is incident from the left. Silicon detectors are drawn as open rectangles,plastic scintillators as filled rectangles. Not to scale.


It is crucial for our purposes to understand the detailed behav-ior of the Bethe–Bloch equation. First, it is not a single valuedquantity for a given value of dE/dx, that is, for a given medium, dif-ferent combinations of projectile Z and b can produce the sameenergy deposition, at least over a small depth interval. (For thatmatter, even a single value of Z does not yield a single-valuedfunction, owing to the relativistic rise.) The lack of uniquenessleads to ambiguities in identification of charge and energy, whichcan be mitigated to some degree by increasing the detector depth.Fig. 2 demonstrates these points; the black bars are the calculatedenergy depositions, DE’s, for protons with 100 MeV of kineticenergy incident on a series of four silicon detectors, each 5 mmthick, followed by a BGO detector 3.8 cm deep. The gray bars arefor 4He ions traversing the same detectors with an incident energyof 2 GeV/nucleon. The energies are chosen so that, over the first5 mm of silicon traversed, the deposited energies are equal. Withonly the first detector, it would be impossible to distinguish thetwo cases. The additional detectors cleanly resolve the ambiguity.

Fig. 2. Calculations of deposited energy for protons and 4He in a series of 5 mmthick silicon detectors (numbers 1 through 4) and 3.8 cm of BGO (5). The energiesare chosen to illustrate the divergent behavior of the two particles in detectorsdownstream of the first detector, although they deposit equal energies in detector1. The proton stops in the BGO, but the 4He ion penetrates and exits the stack.

The clearest distinction is seen in the BGO, where the proton stopsand deposits much less energy than the 4He ion, which punchesthrough.

The second important aspect of the equation for present pur-poses is that as b approaches 1, dE/dx becomes quite insensitiveto velocity. Thus, while a series of energy deposition measure-ments at relatively low energies can in many cases be used to dis-tinguish the charge, energy, and sometimes mass of the incidentparticle, only the charge can be determined by dE/dx for high en-ergy particles. Additional information must be obtained by othermeans, e.g. with a Cherenkov counter as in the HEAO-C experiment[24], or with a magnetic spectrometer, but such approaches are ona different mass scale than we are considering here. For practicallow-mass experiments based solely on energy deposition sam-pling, there will always be some maximum energy above whichonly charge measurements can be made; one can therefore only re-port integral fluxes above that maximum.

The design of the charged particle stack is shown schematicallyin Fig. 1 above. The detectors ahead of the BGO are capable of stop-ping protons and 4He ions with energies up to about 70 MeV/nu-cleon. The full stack including the BGO stops these same particleswith energies up to 165 MeV/nucleon. This is intended to capturethe full energy of protons in the range of energies seen in typicalsolar particle events, but as will be shown below, the BGO alsofunctions as a useful DE detector for higher-energy particles thatpass through it.

In heavy-ion fragmentation experiments with sufficiently high-energy beams and sufficiently thin targets, primary beam ions thatsurvive the target and the majority of charged projectile fragmentsthat are produced emerge from the target with roughly equalvelocities. More precisely, there is a small downshift effect (frag-ments lose a small amount of energy in the collision process com-pared to the incoming primary), compensated by the smallerenergy loss of fragments (compared to the primary) in traversingthe target from their point of creation to the target exit. The narrowvelocity distributions result in energy deposition spectra in which(integrating the Bethe–Bloch equation over modest detectordepths, with velocities remaining approximately constant)DE / Z2.

The proper approach to measuring the charged-particle dose ina mixed radiation field containing (among other components)high-energy heavy ions and protons using silicon detectors de-pends on the details of the detector stack and the ranges of parti-cles in question. At energies low enough for the stack to providecharge and energy identification of the particle as it entered thestack, that information can be fed into a straightforward calcula-tion of LET in water and therefore the dose from those particles.At higher energies, where detailed identification is not possibleusing dE/dx alone, one can simply relate the measured DE of agiven particle in a known depth of silicon to its LET in water via


a single constant of proportionality. This is an energy-dependentapproximation, but it is good to about ±5% above about 30 MeV/nucleon.

2.2.2. Plastic scintillatorsPlastic scintillators possess two particularly useful characteris-

tics for present purposes. First, for triggering purposes, their rise-times are a few nanoseconds, allowing for fast trigger decisions.Second, unlike silicon detectors, their output signals (as read outthrough a photomultiplier tube, pin diode, or photosensor module)are not linear with dE/dx. Instead, the response begins to roll off asthe ion charge increases due to quenching, as shown in Fig. 3,where one can also observe that the scintillator/photosensor com-bination has much poorer resolution than the silicon detectors. Thecompression of the response caused by quenching and the lack ofresolution mean that the plastic scintillators do a poor job ofresolving the higher charges, but those are adequately resolvedin the silicon detectors. However, the relatively large signals fromlow-LET particles allows the plastic scintillators to cleanly resolvethe lightest fragments, charges 1 and 2. This is also helpful in trig-gering, since the silicon detectors (which could in theory be usedfor triggering) must have low gain in their readout electronics tokeep the largest energy depositions below the saturation level,and consequently have comparatively little separation betweenminimum-ionizing particles and their noise level.

In typical applications, plastic scintillators are optically coupledvia a light guide to a photomultiplier tube and base, where highvoltage is applied and the signal (anode and/or dynode) is avail-able. However, photosensor modules that do not require an exter-nal high voltage bias supply are available. The Hamamatsu H5773Pis used here. As in more standard designs, the scintillators – 3.3 cmdiameter circular pieces of Pilot B each about 1 mm thick – areembedded in light guides, which are epoxied to the photosensors.The photosensor requires only a single +12 V power supply. Gaincontrol is achieved by an input on pin 3, which must be in the

1 GeV/nucleon 28Si + 20 g cm -2 CH2

Silicon Detector 2 Pulse Height (ADC Counts)0 100 200 300 400 500

Plas

tic S

cint

illato

r 2 P

ulse

Hei

ght (

ADC

Cou

nts)

50

100

150

200

Fig. 3. Scatter plot of pulse-height obtained with a plastic scintillator (coupled to aphotosensor) versus pulse-height obtained with a silicon detector, in an experimentwith 1 GeV/nucleon 28Si ions incident on a 10 g cm�2 polyethylene target. The lineapproximately indicates what a linear response in DE would be, in contrast to thequenching of the scintillator output.

range of 0.25–0.9 V. We provide this via an onboard programmableDAC (digital to analog converter) that is part of the high voltagedistribution board described below. The control voltage deter-mines the voltage applied across the photomultiplier amplificationchain. The anode output of the photosensor is available on pin 7. Inexperiments with 56Fe, the control voltage was varied between0.4 V and 0.6 V; with 28Si, the voltage was set to 0.84 V. In all cases,the anode outputs were used as inputs to a conventional NIM shap-ing amplifier, which slowed, inverted, and amplified the signals.

2.2.3. BGOThough most commonly used to detect gamma rays, BGO scin-

tillators can also be used for the detection of energetic heavy ions.Shielding studies [13] show that, the higher the Z of a target mate-rial, the less effective it is as a shield against the high-energy ionsin the GCR. The reason for this is that, per unit mass, the interactioncross sections for high-A target materials are relatively small andstopping power is relatively large. These are useful properties fora heavy ion detector, since one would like to stop as many particlesas possible, and for both stopping and penetrating particles, tohave as few interactions in the detector as possible, since thoseadd considerable complication to particle identification. For thesereasons, we decided to augment the silicon stack with a BGO crys-tal, read out with a standard photomultiplier tube. An assembledunit with a diameter of 3.8 cm and a depth of 3.8 cm was obtainedfrom Scionix, Inc.

The energy deposited by a heavy ion is much larger than that ofa photon, and hence the light output of BGO in this application ismuch larger than in more typical use. The PMT voltage must be re-duced accordingly in order to avoid saturation. The optimal settingdepends on measurement priorities. The gain can be reduced bylowering the voltage to the point where even very heavy and rela-tively slow ions can be detected, at the cost of losing resolution oflighter particles. Alternatively, if the priority is to measure low-LETparticles, the PMT voltage can be kept relatively high, with a loss ofresolution for the higher-LET particles. As with the plastic scintilla-tors, the PMT’s anode outputs was input to a conventional NIMshaping amplifier for signal processing ahead of the ADC. In thenext iteration of CINS, the BGO will be handled in a manner similarto the other detectors in the stack. It will be mounted on a detectorboard similar to those used for the silicon detectors, with a photo-sensor module replacing the photomultiplier tube, and a shapingamplifier attached to the photosensor output to bring the timingof the signal into line with those from the silicon detectors.

2.3. Electronics

For this application, we decided that a motherboard/daughter-board approach was acceptable, since there is no particular needto share lines across boards. Each silicon detector and plastic scin-tillator/photosensor is mounted on its own board, which connectsto a motherboard which has 10 slots. Six of the slots are used bythe detectors, another slot is used by a high voltage supply board,another slot is used by a signal driver board, and two additionalslots are available for expansion. It was initially planned that oneof the slots on the motherboard would be used by a board contain-ing analog-to-digital converters (ADCs) for each of the channels,and another slot would be occupied by a communications andCPU board. However, for the time being, these functions have beenrelegated to an external readout system, and it is only necessary tobe able to ship out the signals over long coaxial cables. This is thefunction of the driver board.

2.3.1. Silicon detector readout – DE channelsThe detector boards are designed for use with silicon detectors

equipped with guard rings. The silicon detectors in the HENS also


have guard rings for better definition of the geometric factor. Theguard ring is a distinct annulus at the edge of the detector withits own connection on the n side, which is also where the centralpart of the detector is connected; the p side is held at ground po-tential. The guard ring is not electrically isolated from the centraldisk, but the connection has an impedance of about 1.5 kX. Thebias voltage for both parts of the detector is shared by the centerand guard ring, with a 100 kX resistor in between the two filter cir-cuits. This approach keeps both portions at the same potential rel-ative to ground. If the guard ring is not biased, it floats to apotential near that of the (biased) central part of the detector.

A conventional approach to processing the signals from the sil-icon detectors, using commercially available chip preamplifiersand shaping amplifiers, was adopted initially and is implementedin the board design. The Amptek A250, a low-noise, high-perfor-mance, high-reliability preamplifier was selected for the first stage.Commercial chip shaping amplifiers from Cremat were selected forthe second stage; these have a fixed gain of 10 and a choice ofshaping times ranging from 100 ns to 4 ls. The 1 ls version waschosen for CINS. The Cremat gain is affected by the pole-zero com-pensation. The Cremat circuit is designed for a pulse with a longfall time, but we have substantially reduced the fall time of the pre-amplifier output from the nominal 300 ls, to 4.5 ls, which has theeffect of slightly reducing the gain of the shaper.

Because the system must handle large energy depositions with-out saturation, up to 3000 MeV in a single detector, it is necessaryto add a large external feedback capacitor to the A250 circuit. Thisvalue was chosen to be 56 pF, so that the maximum signal at thepreamplifier output is Vmax = 0.0444 pC/MeV � 3000 MeV/56 pF = 2.38 V. The maximum swing specified in the A250 litera-ture is 2.8 V with ±6 V power supplies. A voltage this large willdrive the Cremat shaper, with its nominal �10 gain, into satura-tion, so it must be reduced. However, the signal from a minimum-ionizing particle (which will deposit about 2 MeV in 5 mm of sili-con) will give an output on the order of 1.5 mV at the A250 output,which should not be significantly reduced. The A250 output istherefore split into two branches, one going to a resistive dividercircuit with output equal to 1/6 of the input (the low gain path),and the other to a similar divider with output 5/6 of input (highgain). The low gain channel should then produce an output of peakamplitude just under 4 V for 3000 MeV deposited in 5 mm of sili-con, corresponding to LET in water of about 300 keV/lm.

The choice to reduce the low-gain channel by a factor of 6, andto therefore obtain a maximum output pulse of 4 V, is driven inpart by the fixed sensitivity of the CAMAC ADCs used in early tests.These units have 12 bits of resolution and accept pulses withamplitudes up to 4 V, so that the scale is very close to 1 ADC countper mV of pulse-height on the input. On the high-gain path, a min-imum-ionizing particle gives an amplitude of about 12.5 mV, suffi-ciently above the pedestal to be seen.

The large feedback capacitance on the A250 requires the addi-tion of a compensation capacitor between pins 6 and 7 to avoidringing. (The need for a compensation capacitor eliminates thepossibility of using the more compact A250F model.) A value of5 pF was found to work well. The fall time of the preamplifier out-put is governed by the product of the feedback capacitance andresistance, and by default is long; using the internal elements inthe A250 yields a time constant of 300 ls. This has been reducedto 4.5 ls in our design by adding a 80 kX external feedback resistorin parallel with the 56 pF external feedback capacitor.

2.3.2. Silicon detector readout – guard ringsThe geometric factor of the telescope is defined by the active

areas of the silicon detectors and plastic scintillators. An importantadvantage of the guard ring design is that the central area is welldefined, which is not the case for similar detectors that lack a guard

ring. In those detectors, the area over which the silicon is fully de-pleted is not equal to the full area; instead, a large area is fully de-pleted, but the depletion depth decreases near the edge. Thiscreates a region in which charge liberated by a particle traversalis not fully collected, an artifact that has the potential to compli-cate the data analysis. With the guard ring, the area of varyingdepletion depth is entirely outside the central region, resulting inimproved resolution and removing edge hits from the data. Theguard ring channel, if read out, can provide useful informationabout particles outside the cone that defines the telescope’s geo-metric factor. Since the pulse-height information from the guardring is, almost by definition, not reliable, it is sufficient to use theguard ring signal as input to a preamplifier and discriminator,and to simply count the number of hits per channel. This is accom-plished on the detector board with an Amptek A111F chip, whichcontains has an internal discriminator. The default threshold sensi-tivity is 5 � 104 electrons; by shorting pins 3 and 4, as done in ourdesign, the threshold is increased to 5 � 105 electrons, correspond-ing to 0.08 pC of charge, or <2 MeV deposited energy. This is veryclose to the mean energy deposited in 5 mm of silicon by a mini-mum-ionizing particle.

2.3.3. Plastic scintillator readoutThe Hamamatsu photosensor modules are described above. A

dedicated circuit board will be built to accommodate these units,but in the tests described here, the photosensors were simply at-tached to unloaded spare silicon detector boards and point-to-point wiring was used to provide the needed voltages (+12 V andthe sensitivity control) and to route the output onto the appropri-ate line on the motherboard.

2.4. Driver board

All detector signals travel along the motherboard to a driverboard mounted in the last slot. The driver board contains twelveLMH6321 chips from National Semiconductor, each capable ofdriving 480 milliamps of current as configured in our system.These are needed since the NSRL provides connections from thebeam cave to the counting room via long 50X coaxial cables. TheCAMAC ADCs provide 50X termination. The drivers must be ableto source a current of at least 80 mA, so the specification is wellabove the requirement. However the observed slew rate of the dri-ver chips is barely adequate for this application, and this choice ofparts will be revisited.

2.5. HVPS board

Bias voltage for the silicon detectors was provided by a dedi-cated board mounted just ahead of the driver board in the stack.This board contains four Model CA10P high voltage supplies man-ufactured by Emco, and a Cygnal C8051F120 microprocessor tocontrol them. The microprocessor communicates via a serial lineto a laptop PC running software that provides a GUI to adjust targetvoltages and ramp rates for the bias supplies. These are set viaDACs on the microprocessor. Additional DACs in the microproces-sor control the sensitivity of the photosensor modules used for theplastic scintillators.

2.6. Data acquisition

A custom data acquisition system has been designed for CINS byJHU-APL engineers. The system makes use of off-the-shelf compo-nents to digitize the waveforms produced by the charge-sensitivepreamplifiers. The custom system is described in detail in anaccompanying article [25]. In the accelerator-based testing of com-ponents performed to date, either CAMAC or VME systems were


used to record data from the charged particle stack. The VME sys-tem is still under development and has been used with mixed re-sults. More success has been had using the same CAMAC systemused in other experiments (see, e.g. [22]). Since, at least for the timebeing, we consider the CAMAC system to be the reference standard,it has been used in parallel with the custom system for the data de-scribed here. Data taken simultaneously in both systems can becompared to make sure that the results are as they should be, thatis, completely independent of the data acquisition hardware.

3. Neutron detector collision experiments

3.1. NASA Space Radiation Laboratory (NSRL)

Two identical high-energy neutron systems were employed fordetecting energy spectra from the collision of high-energy iron and

Fig. 5. Integral neutron energy spectra for neutrons per incident ion versus neutron enerpolyethylene.

Fig. 4. Integral neutron energy spectra for neutrons per incident ion versus neutron enertargets.

silicon beams with aluminum and polyethylene targets at theNASA Space Radiation Laboratory in 2006 and 2007. The 5 mmthick lithium-drifted silicon detectors were enclosed in a light-tight aluminum box with a scintillator paddle for charged particlerejection placed immediately in front of the silicon detector’s aper-ture position. Polyethylene targets of depth 10 and 20 g cm�2 wereused; aluminum targets were 5, 10, 20 and 34.4 g cm�2. The detec-tor positions were usually at angles of 30� and 90� from the centerof the exit face of the target at distances of about 35 cm.

Due to the low detection efficiency for high energy neutrons,the experimental runs were extended until approximately1 � 106–4 � 106 ions were incident on the targets. We show com-parisons of neutron production for: 1 GeV/nucleon 56Fe and 28Sibeams on 20 g cm�2 Al and polyethylene targets (Fig. 4); 1 GeV/nu-cleon 56Fe on 20 and 34.4 g cm�2 Al targets and 20 g cm�2 polyeth-ylene targets (Fig. 5); and 1 GeV/nucleon 28Si on 10 and 20 g cm�2

gy for 1 GeV/nucleon iron ions on two different thicknesses of aluminum and one of

gy for 1 GeV/nucleon 28Si and 56Fe beams on 20 g cm�2 aluminum and polyethylene

Fig. 6. Integral neutron energy spectra for neutrons per incident ion versus neutron energy for 1 GeV/nucleon silicon ions on two different thicknesses of polyethylene targetsnear and perpendicular to the beam-target axis.


polyethylene targets at two detection angles (Fig. 6). The yield datapresented in the following show some interesting features and willbe compared to Monte Carlo simulations in a future article.

3.1.1. Effects of projectile and target typeFig. 4 shows the normalized (per incident ion) neutron integral

energy spectra for 1 GeV/nucleon 28Si and 56Fe beams on 20 g cm�2

aluminum and polyethylene targets at an angle of 30� from thebeam-target axis (basically in the forward direction). The ±40% er-ror produced by the statistical deconvolution of the energy deposi-tion spectra into the incident neutron spectra means that there isnot a significant difference between individual data points forthe aluminum and polyethylene targets for either the silicon oriron projectiles. The shapes of the Al target curves are steeply fall-ing, with sharp corners around 40 MeV. In contrast, the curves seenfor polyethylene targets are comparatively smooth below about50 MeV, showing moderating effects that arise primarily fromthe hydrogen content. The iron beam neutron production is signif-icantly greater than the silicon beam production at all but a veryfew individual points at high energy. The integrated neutron yields(neutrons per ion) above 10.75 MeV are given in Table 1.

The large uncertainties notwithstanding, the 30� data show lar-ger neutron yields for the 56Fe beam than for the 28Si beam. Themean neutron yield per incident ion at forward angles is expected,for a given target material and depth, to be significantly higher forthe iron beam than for the silicon beam. This is due both to the lar-ger cross section for iron projectiles to interact in the target (whichaffects both projectile- and target-like yields), and to the greaternumber of neutrons available to be removed from the projectile.Even though Fig. 4 shows that, at a few individual energies, thetwo beams produce error bars that touch, the mean values over lar-ger energy intervals are significantly different due to the reduced

Table 1Neutron yields per incident beam ion at 1 GeV/nucleon with 20 g cm�2 Targets.

Ion Target material Integrated yield

56Fe CH2 0.3756Fe Al 0.2228Si Al 0.1628Si CH2 0.077

standard error of the mean. Experience allows one to concludefrom visual inspection that the means are significantly differentat a high confidence level when the nominal value of one set ofdata is consistently greater than that of a second set for 20 pointsor more with the binning shown here.

3.2. Effects of target thickness or areal density

Having seen that projectile type is more dominant than targettype in high-energy nucleus-nucleus collisions, we examine the ef-fects of target depth on neutron yield. Fig. 5 shows the neutronyields from collisions of 1 GeV/nucleon 56Fe ions on two thick-nesses of aluminum with a comparison to a polyethylene targetfor the smaller areal mass density. All data are again at 30�.

The neutron yields for the 34.4 g cm�2 Al target are significantlygreater than those for the 20 g cm�2 Al target for almost all datapoints (there are a very few points where the error bars just touchat the higher energies) across the 10.75–500 MeV energy range.The curve for the 20 g cm�2 polyethylene target mostly falls be-tween the two aluminum spectra, with error bars that overlapboth. The 34.4 g cm�2 Al energy spectrum is significantly higherthan the 20 g cm�2 polyethylene spectrum in the very lowest en-ergy bins. Also, the 34.4 g cm�2 Al target does not have the sharpcorner seen with 10 and 20 g cm�2 Al target depths; the curve isqualitatively more similar to that seen with the 20 g cm�2 polyeth-ylene target. This data set again shows the better moderating capa-bility of polyethylene. More data with aluminum targets havingdepths between 20 and 34 g cm�2 would be useful, as we wouldexpect them to show the transition in the spectral shape as a func-tion of target depth.

The integrated neutron yields (neutrons per incident beamion) above 10.75 MeV are given in Table 2. The comparatively

Table 2Neutron yields per incident 56Fe ion at 1 GeV/nucleon with varying target depths.

Target material Depth (g cm�2) Integrated yield

Al 34.4 0.92CH2 20 0.37Al 20 0.22

Table 3Neutron yields per incident 28Si ion at 1 GeV/nucleon with detector off axis.

CH2 depth (g cm�2) Detector angle Integrated yield

20 30� 0.07710 30� 0.06420 90� 0.02410 90� 0.023


large yield seen with the 34.4 g cm�2 Al target may in part be ex-plained by energy dependence of the interaction cross sections.The range of the 56Fe beam in Al is about 34 g cm�2, so thatprimaries that survive without interaction stop in this target. Sec-ondaries and higher-generation fragments produced deep in thetarget have much lower energies than those produced nearerthe target entrance, and because the cross sections for p-nucleusand n-nucleus reactions rise significantly with decreasing projec-tile energy below about 20 MeV, it is reasonable to expect thatthis target will produce a greater yield of low-energy neutronsthan does the 20 g cm�2 Al target. The data appear consistentwith this hypothesis: taking the ratios of the yields per incidentFe ion at 34.4 and 20 g cm�2, we find that as energy decreases,the ratio increases steadily from a value of 2.48 at 416 MeV to3.24 at 42 MeV, stays somewhat constant in the range from42 MeV down to 23.7 MeV, and then increases rapidly to a valueof 4.19 going from 23.7 MeV to 10.75 MeV. Thus, the significantlylarger yield from the thicker target is largely due to an abundanceof lower-energy neutrons, which is at least qualitatively consis-tent with the known increase of the interaction cross sectionsat low energy.

3.3. Effects of detection angle

Fig. 6 shows the effect of the angle of neutron detection for col-lisions of 1 GeV/nucleon 28Si projectiles with 10 and 20 g cm�2

polyethylene targets. There is considerable overlap of the errorbars for the two depths of polyethylene at both 30� and 90�. Thepath lengths of neutrons at 90� can potentially intersect more tar-get atoms or areal density than the normally incident 10–20 g cm�2 for our 38 cm wide targets (the pencil like beam collidesat the center of the target). However, this effect is more pro-nounced for higher mass density materials than for polyethylene.The result of Fig. 6 is that the neutron yield at 30� near normal inci-dence is significantly greater than at 90� for the ranges of targetthickness and neutron energies studied. At 90� the yield data arealmost identical for 10 and 20 g cm�2 targets for neutron energiesbelow about 50 MeV. The integrated neutron yields (neutrons perion) above 10.75 MeV are given in Table 3.

Fig. 7. Calibration curve of voltage versus energy for Eljen 12 cm � 12 cm boron-loadedV = 0.113E1.58.

3.4. RARAF

The Columbia University Radiological Research AcceleratorFacility (RARAF) uses a 5 MV Singletron Cockroft–Walton linearaccelerator (installed at the end of 2005) to generate fast neutronsusing nuclear reactions in thin targets. These essentially mono-energetic neutrons in the 220 keV–15 MeV range are producedby protons and deuterons bombarding hydrogen isotope targets.As an example T(p,n)3He, D(d,n)3He and T(d,n)3He reactions areused to create 2.8 MeV, 6.0 MeV and 13–15 MeV neutrons, respec-tively. The neutron flux from the third reaction is nearly indepen-dent of angle while that from the other two varies sharply withangle.

3.4.1. EljenFig. 7 shows the energy calibration curve from the RARAF inves-

tigations with mono-energetic neutrons. The voltage versus energycurve obeys a best least-squares fit relationship of V = (0.113 ±0.003) � E(1.58 ± 0.01), in good agreement with Byrd and Urban[26], who predict an E1.6 response for organic scintillators.

3.4.2. HENSThe neutron detection efficiency of HENS is governed by the

detector thickness, or number of target nuclei that intersect thepath of a neutron. Fig. 8 shows the comparison of the responses(counts versus channel number) of the two detector types (smallarea and large area) used in the 14 MeV neutron beam at RARAF.By 2006 we had improved the data acquisition electronics forHENS and RARAF had a new accelerator. The results for the twodetector areas are very similar in structure, with the larger areadetector as expected having a considerably higher count rate.

plastic scintillator for six neutron energies. The straight line is a least-squares fit,

Fig. 8. Response of large and small area 5 mm thick Si(Li) detectors to 14 MeV neutrons.


4. Charged-particle experiments and calibration

The beam tests performed with the charged-particle stack areuseful for several purposes, including debugging the hardware,understanding the calibration of the system, and – when the sys-tem performs to specification – acquiring data of scientific interest.Also, the photosensor modules’ control voltage and the BGO PMTvoltage have been varied in these tests, in order to find the best set-tings for specific beams. These give us an idea of what settingswould be optimal in the spaceflight environment.

Good charged-particle data have been obtained with beams of1 GeV/nucleon 12C and 28Si at the NSRL. The data are of sufficientquality that they can be analyzed to extract the fragmentationcross sections. The 28Si data analysis has been completed and theresults will be compared to the cross sections obtained with 800and 1200 MeV/nucleon beams [27]. The thick targets used here al-low for improved precision in the determination of the charge-changing cross sections. Other data sets demonstrate the success-ful implementation (and limitations) of the dynamic range require-ments, the utility of the scintillators as trigger detectors, and thehigh resolution of the BGO. The data also allow for a comparisonof spectra obtained with the standard CAMAC acquisition systemand with the custom-designed system described above, thoughthat analysis is not yet complete. One good 28Si data set was ob-tained with an aluminum target, two with polyethylene targets,and one with no target, those data being needed to properly nor-malize the target-in results.

To facilitate the comparison of data acquisition systems, thehigh-gain signals from the silicon detector boards were not used.Instead, the preamplifier output on each high-gain branch was sentdirectly into the new system, with a jumper wire temporarilyreplacing the CREMAT shaping amplifier on that branch. The CA-MAC system received the shaped low-gain signals. The scintillatorsignals were amplified in NIM units with dual outputs, so for thosedetectors, each data acquisition system received the same signal.

4.1. Spectra on the beam axis

As the development of the stack has proceeded, the stack hasbeen gradually populated, starting with three (of four) silicondetectors and one (of two) plastic scintillators. As hardware prob-

lems are overcome, the emphasis during the experiments shiftsfrom debugging hardware to obtaining scientifically useful data.For the data described here, all elements of the charged-particlestack worked acceptably except for a failure on the readout boardof Si4. The BGO photomultiplier tube was set to a potential of�400 V and the photosensor modules for the plastic scintillatorswere set to 0.9 V. If the beam had higher LET, these settings wouldresult in saturation for the primary beam ions and heavy frag-ments. However, in the 28Si beam, the signals stayed on-scale,though they do show quenching (discussed more below). Further,these settings might be appropriate for flight, as they provide rel-atively high gain so that signals from minimum-ionizing particlesare easily observable above the noise.

Because the low gain paths in the silicon detector readouts areset to accommodate relatively slow iron particles, as would beencountered in space, the signals from 1 GeV/nucleon 28Si are rel-atively small and the resolution is not optimal. Even so, as Fig. 9shows, the resolution was sufficient to cleanly identify the primarybeam ions and several fragment species. Fig. 9 contains two panels;on the left is a scatter plot of the summed deposited energy in sil-icon detectors 2 and 3 versus the energy deposited in Si1. A selec-tion cut has been made to require good correlation. Each island ofevents corresponds to a particular species of particle hitting thedetectors, either primaries that survived traversal of the target,or fragments. Only four distinct islands are seen, correspondingto (from highest to lowest DE, Si, Al, Mg and Na), but this is dueto the excessive density of events in the plot. The peaks clarifywhen plotted as a histogram, as in the right panel. Here, the sameevents as in the scatter plot are used to create a charge histogram,by plotting the square root of the sum of the DE’s and applying ascale factor so that the peak for the primary falls at exactly its nom-inal value. Clear peaks are seen for fragments with charge as low as5. This is better resolution than was obtained in a comparableexperiment at 1200 MeV/nucleon [27] with different detectorsand older front-end electronics. The fragment peaks do not lineup exactly with the nominal (integer) values, since they tend toemerge from the target with higher velocities than the survivingprimaries. This effect manifests itself as a slightly lower energydeposition than would occur if the velocities matched the prima-ries, since DE / (Z/b)2, and hence the charge distributions arepeaked slightly below the integer values.

Fig. 9. Left: scatter plot of deposited energy in Si2 and Si3 (summed) versus energy deposited in Si1, with correlation cut imposed. Right: charge histogram for the eventsshown in the panel on the left.

Table 5Fragment yields (�102) Per incident 1 GeV/nucleon 28Si ion in Al and CH2.

Fragmentcharge

CH2, 10 g cm�2 CH2, 20 g cm�2 Al, 20 g cm�2 Ratio (Al:CH2)at 20 g cm�2

13 3.35 ± 0.05 2.76 ± 0.06 2.08 ± 0.06 0.7512 4.28 ± 0.05 3.98 ± 0.07 2.53 ± 0.07 0.6411 2.83 ± 0.04 2.87 ± 0.06 1.64 ± 0.05 0.5710 3.25 ± 0.04 3.37 ± 0.06 1.73 ± 0.05 0.519 2.07 ± 0.03 2.15 ± 0.05 1.11 ± 0.04 0.528 3.90 ± 0.05 4.22 ± 0.07 2.20 ± 0.05 0.527 3.26 ± 0.04 3.78 ± 0.07 1.83 ± 0.05 0.486 4.21 ± 0.04 5.09 ± 0.08 2.68 ± 0.05 0.535 2.39 ± 0.03 2.75 ± 0.06 1.51 ± 0.04 0.55<5 28.9 ± 0.1 50.8 ± 0.2 33.7 ± 0.2 0.66


4.2. Fragmentation of 1 GeV/nucleon 28Si

Five data sets were taken with the 1 GeV/nucleon 28Si beam andthe detector stack placed on the beam axis. The beam was incidenton two Al targets (10 and 20 g cm�2) and two CH2 targets (10 and20 g cm�2). A blank target run was also taken. The data analysismethods used here parallel those used in our other cross sectionmeasurements, although the target depths here are substantiallygreater. The increased depths cause the beam to lose substantialenergy in traversing the targets, so if there were any energy depen-dence of the cross sections, the measurements could suffer from aloss of accuracy. However, we have shown in [27] that the charge-changing cross sections slowly increase with energy for hydrogentargets, and are essentially energy independent for other targets.Therefore the charge-changing cross sections obtained here are di-rectly comparable. The results are shown in Table 4, with the poly-ethylene target result being the weighted average of the crosssections obtained at two target depths. The beam energies listedin the table are calculated values at the target center. We also showthe mean free paths for the 28Si beam in the two materials.

The newly obtained charge-changing cross sections agree wellwith the older results. The values shown in Table 4 are from ourpreviously published thin-target data with 28Si beams at800 MeV/nucleon.

Because the target depths are so great, it would be necessary toapply large, model-dependent corrections to extract the cross sec-tions for production of the various fragment species. These correc-tions are estimated to approach 100% for the thickest targets usedin this experiment. Rather than present results with such large cor-rections (and hence large uncertainties), we present instead in Ta-ble 5 the probability per incident ion for detecting a leadingfragment of a particular species for each of the three targets. Thesecan be thought of as approximating the fragment fluences per inci-dent primary ion, though that is not precisely true since our detec-tors are larger than the nominal 1 cm2 area in the definition.

The data in Table 5 clearly show two effects that have beennoted in earlier work, a strong enhancement in the production ofeven-Z fragments relative to odd-Z, and a significant suppression

Table 4Charge-changing cross sections of 1 GeV/nucleon 28Si in Al and CH2.

Targetmaterial

Avg. Ebeam

(MeV/nucleon)MFP (g cm�2) rcc (mb) Previous

result (mb)

CH2 877 11.7 ± 0.2 1995 ± 21 1958 ± 28Al 876 28.7 ± 0.5 1562 ± 31 1546 ± 27

of F (charge 9) production. Further, in comparing the yields behindpolyethylene targets, it is interesting to note that the fragment flu-ences do not simply increase with target depth. In fact, for the spe-cies closest to the beam, there is a net loss of fragments due tosecondary interactions; for charges 9–11, there is little differencein yield between the two target depths; and it is only for charges8 and below that the thicker target produces an increase in thefragment yields. In the far-right column of Table 5, the aluminumtarget yields are divided by the polyethylene target yields at thesame 20 g cm�2 depth. Fragment yields in aluminum should belower simply on the basis of the charge-changing cross sectionsand the fractions of surviving primaries, 18% for the CH2 targetand 49% for the Al target. The overall fragment yield (not account-ing for differences in multiplicities produced in the targets) in alu-minum is therefore 0.62 the yield in CH2. The data deviate fromthis in an interesting pattern. In the range of fragment charges6–10, the ratios show little variation and are all close to 0.5, belowthe nominal ratio of 0.62. For Na, Mg and Al fragments (charges11–13), the ratios are somewhat larger than nominal, due to thesignificantly higher losses of these fragments in the CH2 targetdue to secondary interactions, and the ratio is also larger thannominal in the summed category below charge 5. The latter effectis due to the fact that the hydrogen targets found within the poly-ethylene are unable to completely break up the projectile into verylight fragments, even in central collisions. As the mass number ofthe target increases, complete break-up of the projectile into lightfragments becomes increasingly probable.

In Fig. 10, the fluences from Table 5 are plotted along with pre-dictions of the PHITS model [28]. For calculation of hydrogen-tar-get cross sections, used in the simulation of CH2 target data,PHITS offers the user a choice of models. We used the default Ber-tini model.

Fig. 10. Fragment fluences per incident beam ion, as measured (circles) and aspredicted by PHITS (histogram).


We have recently published several comparisons to PHITS usinga variety of beam ions: 12C at 290 and 400 MeV/nucleon [29]; 35Cl,40Ar and 48Ti at energies from 290 to 1000 MeV/nucleon [22] and56Fe at 1000 MeV/nucleon [30]. For the 12C beams, it was foundwhile PHITS was reasonably accurate in its prediction of thecharge-changing cross section, the fragment cross sections weresignificantly off. A similar trend was seen in the Cl, Ar and Ti data,although there it was observed that PHITS was superior to oldermodels in its prediction of the cross sections for lighter fragments(below charge 9). The 28Si beam presents an intermediate case, andthe thick targets used here present an additional challenge to themodel, in that secondary and tertiary interactions are relativelycommon. These tend to deplete the observed fluences of the hea-vier fragments and increase the fluences of the lighter fragments.Thus, if a model were to, for example, over-predict the light frag-ment yields even for thin targets, the effect would be amplifiedin thick-target data.

Fig. 11. Histograms of effective charge (bin width of 0.05 charge unit)

4.3. Off-axis spectra

In the left panel of Fig. 11, we show the charge spectrum ob-tained with the 1 GeV/nucleon 28Si beam incident on the 20 g cm�2

aluminum target, with the detector approximately 10� off thebeamline. Deposited energy signals from the first three silicondetectors were summed and scaled to make the histogram. Unlikethe spectra obtained on-axis, which are typically dominated bysurviving primary beam ions, very few primaries are seen. Thefragment distributions are also completely changed compared toon-axis spectra. Off-axis, the probability of observing a fragmentof a particular charge generally increases with decreasing charge(with the exception of charge 9) over the range from charge 13to 5. Note that, because only the low-gain channels were used, res-olution is lacking below charge 5, and some 85% of the events arerecorded as charge 0, even though we know a charged-particle hitthe stack (to produce the trigger). In the right panel, we show thedeposited energy spectrum in the first silicon detector for the highgain channel for the same event sample as in the left panel. For thisoff-axis data set only, this channel was fully implemented as de-signed (rather than being routed, without shaping, to the customdata acquisition system). There is clearly a substantial improve-ment in the resolution at the low end of the spectrum, as expected.The peak centered just above 200 MeV is due to F fragments (Z = 9),so one can count the additional peaks, down to Z = 1.

4.4. BGO as a DE detector

The same off-axis data set used to produce the histograms inFig. 11 above were used to make Fig. 12, a histogram of the BGOpulse-height. (The pedestal has been subtracted and events withpulse-height consistent with the pedestal have been suppressed.)Again, the peak at the far right of the histogram is due to charge9 fragments. The spectrum is quite clean, in particular the charge4 and 5 peaks are better resolved in the BGO than in the high-gainsilicon channel shown in Fig. 11. In between the peaks, there ap-pear to be subsidiary peaks. These are likely due to the coincidentaldetection of a second fragment, for instance, the small peak cen-tered around channel 900 is probably due to Be and He fragmentsin coincidence. So, although the primary purpose of the BGO is tostop relatively low-energy particles (in particular, solar protons

, left, and of deposited energy in Si1 (bin width of 1.5 MeV), right.

Fig. 12. Histograms of the BGO pulse-height spectrum for the 20 g cm�2 aluminum target, with the stack 10� off axis.


in the spaceflight environment), the data show that it works wellas a DE detector to augment the silicon detectors.

5. Systems integration

The CINS subsystems have all been successfully fabricated,tested, and calibrated with beams, as described above. The remain-ing challenges are in the areas of integration and automation ofdata processing. The best approach depends on the intended appli-cation of the system.

5.1. Charged stack/HENS integration

Since the charged-particle stack and HENS use the same 5 mmthick silicon detectors, with similar readouts, we envision the pos-sibility of transferring the HENS functionality to the stack. Toaccomplish this, stack detectors Si1 and Si4 would be fitted withanticoincidence shields made from plastic scintillator or CsI. Thesewould not replace P1 and P2, which would continue to serve as thecoincidence triggers. In addition to the P1*P2 coincidence triggerfor charged particles, we would also implement a neutron triggerin the data acquisition, which would fire when a hit above thresh-old is seen in Si1 or Si4 and no hit is recorded in the correspondinganticoincidence shield. The readout is already set up to accommo-date this.

This approach makes sense in a heavily shielded spaceflightenvironment, or for accelerator measurements made with thedetector placed well away from the beam axis. In both of thesesituations, the high-energy neutron flux is large enough relativeto the charged-particle flux that one would see a reasonableshare of neutron triggers. However, the approach would not beuseful in an accelerator environment when the detector is placedon, or at a small angle with respect to, the beam axis, and thetarget is not thick enough to completely stop the beam and mostsecondaries. In that situation, the rate of charged particles com-pletely dominates the observed neutron rate (in part due to thesmall efficiency for detection), and it makes no sense to attemptto perform an integrated neutron and charged particle dosemeasurement.

5.2. Charged stack/Eljen integration

In this concept, the BGO is either replaced by the Eljen, or iscoupled to a photosensor to make a transmission type mounting,with the Eljen placed at the downstream end. In either case, theEljen would be the last element in the stack, and would serve asa stopping detector for charged particles with modest energies.The areal density of the Eljen is about 13 g cm�2, so it is not aseffective at stopping particles as the BGO is. If we replaced theBGO with the Eljen, the maximum proton stopping energy of thestack would be decreased to 155 MeV. If the BGO were left in thestack as a transmission detector and the Eljen added to the stack,the maximum stopping energy would increase to about 230 MeV.

The same considerations apply to this integration as to the pro-posed integration of the charged stack and HENS. It would not bepractical to expect a good measurement of the dose from medium-energy neutrons in situations where the charged particles aredominant. However, in other situations where the contributionsare more nearly equal, such an approach would likely prove useful.

Acknowledgements

We extend great thanks to Drs. A. Rusek and M. Sivertz for mak-ing the NSRL a congenial place to do experiments. We also thankthe accelerator operators at BNL for their efforts on our behalf. Thiswork is supported at Johns Hopkins University Applied PhysicsLaboratory under NASA Cooperative Agreement 9-58 with the Na-tional Space Biomedical Research Institute.

References

[1] F.A. Cucinotta, W. Schimmerling, J.W. Wilson, L.E. Peterson, G.D. Badhwar, P.B.Saganti, J.F. Dicello, Radiat. Res. 156 (2001) 682.

[2] E.R. Benton, E.V. Benton, Nucl. Instr. and Meth. B 184 (2001) 255.[3] S.J. Hoffman, D.I. Kaplan, Human exploration of mars: the reference mission of

the NASA mars exploration study team, NASA Johnson Space Center, 1997.http://exploration.jsc.nasa.gov/marsref/contents.html.

[4] F.A. Cucinotta, M. Durante, Lancet. Oncol. 7 (2006) 431.[5] J.A. Simpson, in: M.M. Shapiro (Ed.), Composition and Origin of Cosmic Rays,

Reidel, Dordrect, 1983, p. 1.[6] International Commission on Radiation Protection, Publication 92, Vol. 33,

2003, 80p. ISBN-13:978-0-08-044311-9; ISBN-10:0-08-044311-7.

http://exploration.jsc.nasa.gov/marsref/contents.html


[7] J.M. Ryan, J.A. Lockwood, H. Debrunner, Space Sci. Rev. 93 (2000) 35.[8] A.J. Tylka, W.F. Dietrich, Radiat. Meas. 30 (1999) 345.[9] A. Sancar, L.A. Lindsey-Boltz, K. Unsal-Kacmaz, S. Linn, Annu. Rev. Biochem. 73

(2004) 39.[10] W.R. Holley, A. Chatterjee, Radiat. Res. 145 (2) (1996) 188.[11] National Aeronautics and Space Administration, Bioastronautics Critical Path

Roadmap (BCPR), An approach to risk reduction and management for humanspace flight: extending the boundaries, Johnson Space Center Publication62577, 2004.

[12] J.W. Wilson, F.F. Badavi, F.A. Cucinotta, J.L. Shinn, G.D. Badhwar, R. Silberberg,C.H. Tsao, L.W. Townsend, R.K. Tripathi, NASA Technical Paper 3495, 1995.

[13] C. Zeitlin, S.B. Guetersloh, L. Heilbronn, J. Miller, Nucl. Instr. and Meth. B 252(2006) 308.

[14] N. Vana et al., Radiat. Prot Dosimetry 120 (2006) 405.[15] L.C. Simonsen, J.E. Nealy, NASA Technical Paper 3079, 1991.[16] J. Goldsten et al., Space Sci. Rev. 131 (2007) 339.[17] R.H. Maurer, D.R. Roth, J.D. Kinnison, J.O. Goldsten, R.E. Gold, R. Fainchtein,

Acta Astronaut. 52 (2003) 405.[18] J.D. Kinnison, R.H. Maurer, D.R. Roth, R.C. Haight, Radiat. Res. 159 (2003) 154.[19] C. Zeitlin, T. Cleghorn, F. Cucinotta, P. Saganti, V. Andersen, K. Lee, L. Pinsky, W.

Atwell, R. Turner, G. Badhwar, Adv. Space Res. 33 (2004) 2204.

[20] R.S. Saunders et al., Space Sci. Rev. 110 (2004) 1.[21] D.E. Stilwell, W.D. Davis, R.M. Joyce, F.B. McDonald, J.H. Trainor, W.E. Althouse,

A.C. Cummings, T.L. Garrard, E.C. Stone, R.E. Vogt, IEEE Trans. Nucl. Sci. 26(1979) 513.

[22] C. Zeitlin, S. Guetersloh, L. Heilbronn, J. Miller, A. Fukumura, Y. Iwata, T.Murakami, D. Mancusi, L. Sihver, Phys. Rev. C 77 (2008) 034605.

[23] R.M. Sternheimer, Phys. Rev. B 24 (1981) 6288.[24] J.J. Engelmann, P. Ferrando, A. Soutoul, P. Goret, E. Juliusoon, L. Koch-

Miramond, N. Lund, P. Masse, B. Peters, N. Petrou, I.L. Rasmussen, Astron.Astrophys. 233 (1990) 96.

[25] M. Grey, J. Goldsten, R. Maurer, D. Roth, C. Zeitlin, Nucl. Instr. and Meth. B 267(2009) 139.

[26] R.C. Byrd, W.T. Urban, Los Alamos National Laboratory Report LA-12833-MS,1994.

[27] C. Zeitlin, A. Fukumura, S.B. Guetersloh, L.H. Heilbronn, Y. Iwata, J. Miller, T.Murakami, Nucl. Phys. A 784 (2007) 341.

[28] H. Iwase, K. Niita, T. Nakamura, J. Nucl. Sci. Technol. 39 (2002) 1142.[29] C. Zeitlin, S. Guetersloh, L. Heilbronn, J. Miller, A. Fukumura, Y. Iwata, T.

Murakami, Phys. Rev. C 76 (2007) 014911.[30] C. Zeitlin, L. Sihver, C. La Tessa, D. Mancusi, L. Heilbronn, J. Miller, S.B.

Guetersloh, Radiat. Meas., in press, doi:10.1016/j.radmeas.2008.02.013.

http://dx.doi.org/10.1016/j.radmeas.2008.02.013

Documents

Nuclear Instruments and Methods in Physics Research B · PDF filecontinue to be the limiting factor in risk assessment for long-duration mission planning purposes for the foreseeable