the plateau de bure neutron monitor: design, operation and monte carlo simulation

IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 59, NO. 2, APRIL 2012 303

The Plateau de Bure Neutron Monitor: Design,Operation and Monte Carlo Simulation

S. Semikh, S. Serre, J. L. Autran, Senior Member, IEEE, D. Munteanu, Member, IEEE, S. Sauze, E. Yakushev, andS. Rozov

Abstract—This paper describes the Plateau de Bure Neu-tron Monitor (PdBNM), an instrument providing continuousground-level measurements of atmospheric secondary neutronflux resulting from the interaction of primary cosmic rays withthe Earth’s atmosphere. The detector is installed on the Plateaude Bure (Devoluy mountains, south of France, latitude North 4438’ 02’’, longitude East 5 54’ 26’’, altitude 2555 m) as a part ofthe ASTEP Platform (Altitude Single-event effects Test EuropeanPlatform), a permanent installation dedicated to the study ofthe impact of terrestrial natural radiation on microelectronicscircuit reliability. The present paper reports the neutron monitordesign, its operation since August 2008 and its complete numericalsimulation using the Monte Carlo codes GEANT4 and MCNPX.We particularly detail the computation of the neutron monitordetection response function for neutrons, muons, protons andpions, the comparison between GEANT4 and MCNPX numericalresults and the evaluation of the PdBNM counting rate as functionof both the nature and flux of the incident atmospheric particles.

Index Terms—Accelerator factor, atmospheric neutrons,GEANT4, MCNPX, muons, neutron flux, neutron monitor, pro-tons, response functions.

I. INTRODUCTION

W HEN a primary cosmic ray strikes atoms in Earth’s at-mosphere, the collisions may produce one or more new

energetic particles called “secondary” cosmic rays. These sec-ondary particles may strike other atmospheric atoms producingstill more secondary cosmic rays. The whole process is calledan atmospheric cascade or extensive air shower. If the primarycosmic ray has enough energy, the nuclear byproducts of thecascade can reach Earth’s surface. Such extensive air showerswith many particles of different nature (photons, electrons,hadrons, nuclei) arriving on the ground can be detected with awide variety of particle detectors, including different types of

Manuscript received May 02, 2011; revised September 09, 2011; acceptedDecember 07, 2011. Date of publication January 23, 2012; date of current ver-sion April 13, 2012. The neutron monitor construction was supported in part bythe MEDEA + Project #2A704 ROBIN and in part by the French Ministry ofEconomy, Finances and Industry under research conventions #052930215 and#062930210.

S. Semikh, E. Yakushev, and S. Rozov are with the Dzhelepov Labora-tory of Nuclear Problems, JINR, 141980 Dubna, Moscow, Russia (e-mail:[email protected]).

S. Serre, J. L. Autran, D. Munteanu, and S. Sauze are with Aix-Marseille Uni-versity and CNRS, Institute of Materials, Microelectronic, and Nanosciences ofProvence (IM2NP, UMR CNRS 6242), F-13384 Marseille Cedex 13, France(e-mail: [email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TNS.2011.2179945

drift chambers, streamer tube detectors, scintillation counters,optical detectors or Geiger tube detectors [1]. Among them,the so-called “neutron monitors” are especially dedicated tothe counting and detection of cosmic ray induced neutrons [1],[2]. These neutral atmospheric particles are of great importancein microelectronics, a completely different field than cosmicray or elementary particle physics, since their interaction withthe matter (primarily Silicon) has been identified over manyyears as a major production mechanism of Single-Event Effects(SEE) in electronics integrated circuits [3]–[7]. For the mostrecent deca-nanometer technologies, the impact of other atmo-spheric particles produced on circuits has been either clearlydemonstrated (protons) or is still more or less an unexploredquestion (pions and charged muons in particular) [3].

In this context and in order to experimentally study the ef-fects of natural atmospheric radiation on microelectronics cir-cuits, we developed and installed in 2005 a permanent test plat-form at high altitude, the Altitude SEE Test European Platform(ASTEP) [8], [9]. ASTEP is located in the French Alps on thedesert Plateau de Bure (Devoluy Mountains) at 2552 m (Lat-itude North 44 38’ 02’’, Longitude East 5 54’ 26’’), in alow electromagnetic noise environment, and is hosted by the In-stitute for Radio-astronomy at Millimeter Wavelengths (IRAM[10]). It has been fully operational since March 2006. From a ge-omagnetic point-of-view, the ASTEP site is characterized by acutoff rigidity of 5 GV; the natural neutron flux is approximately6 times higher than the reference flux measured at New-YorkCity. This value is called the “acceleration factor” with respectto the gain that we can expect on the duration of the differentexperiments for SEE detection performed in altitude instead ofat sea-level [3], [11].

In 2006 and after suspecting the importance of natural radia-tion (neutrons) background fluctuations in the interpretation andfine analysis of our experiments [12], we launched the construc-tion of a neutron monitor for the ASTEP platform, precisely tosurvey on site and in real-time (typically minute per minute)the time variations of the natural atmospheric neutron flux in-cident on the ASTEP platform. The integration of the instru-ment was finalized in June 2007; its installation on site was per-formed in July 2008 after approximately one year of operationand test in Marseille. The instrument, definitively known underthe appellation “Plateau de Bure Neutron Monitor” (acronymPdBNM), has been fully operational on ASTEP since July 23,2008 [13]. In the present paper, we first describe the design andconstruction of the PdBNM (Section II). In Section III, we re-port its operation since July 2008 and the observation of detectedparticle flux fluctuations due to atmospheric pressure variations

0018-9499/$31.00 © 2012 IEEE

304 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 59, NO. 2, APRIL 2012

Fig. 1. 3D schematics (top) and detailed view (bottom) of the Plateau de BureNeutron Monitor (PdBNM). The instrument is placed on a concrete thick (40cm) floor. Dimensions on the draw are in mm.

and solar events. We also detail the experimental determina-tion of the acceleration factor of the ASTEP location with re-spect to sea-level. In a second part of the paper, we discuss themodeling and numerical simulation of the PdBNM using bothGEANT4 and MCNPX Monte Carlo codes. We successively de-tail in Section IV the physics models involved in simulations,the comparison methodology between GEANT4 and MCNPXresults, the input data for realistic particle sources and finallythe PdBNM detection responses and its sensitivity to other at-mospheric particles (than neutrons). Finally, our conclusions aresummarized in Section V.

II. PdBNM DESIGN

The Plateau de Bure neutron monitor consists of three Heproportional counter tubes that detect thermal neutrons via theexothermic He p T nuclear reaction keV .The absorption of a neutron by a He nucleus is generally fol-lowed by the emission of charged particles (a triton and a proton)which then are detected by depositing (part of) their energy inthe gas and creating a charge cloud in the stopping gas. Thegreater the energy deposited in the gas by these particles, thelarger the number of primary ion pairs, the larger the number ofavalanches, and the larger the pulse detected as the output signalby the electronic acquisition chain.

Fig. 2. Top: External view of the ASTEP building showing the metalwalledroom on the first floor which is installed the Plateau de Bure Neutron Monitor.Bottom: ROOT screenshot showing the GEANT4 modeling of the neutron mon-itor and its implantation in the ASTEP first floor room.

The PdBNM, shown in Fig. 1, is very similar to a standard“3-NM64 neutron monitor” as usually labeled in the literature[1], [2]. Its design follows the recommendations published in[14], [15] for the optimization of the apparatus response. Theensemble detector is based on three high pressure (2280 Torr)cylindrical He detectors, model LND 253109 [16]. These de-tectors are long tubes (effective length 1828.8 mm) offering alarge effective detection volume (3558 cm ) and a very highthermal neutron sensitivity of 1267 counts/nv (LND specifica-tions) [16]. Each tube is surrounded by a 25 mm thick coaxialpolyethylene (PE) tube which plays the role of a neutron moder-ator and by 20 coaxial thick (50 mm) lead rings serving as sec-ondary neutron producer. All these elements are placed insidea 80 mm thick PE box to reject low energy (thermal) neutronsproduced in the close vicinity of the instrument. The only differ-ence between a standard 3-NM64 design and this instrument isin the geometrical shape of the lead rings—they have flat bottomside and are directly put on the bottom of the polyethylene box[14]. A Canberra electronic detection chain, composed of threecharge amplifiers model ACHNP97 and a high voltage source3200D, was chosen in complement to a Keithley KUSB3116acquisition module for interfacing the neutron monitor with the

SEMIKH et al.: PLATEAU DE BURE NEUTRON MONITOR: DESIGN, OPERATION AND MONTE CARLO SIMULATION 305

Fig. 3. Plateau de Bure Neutron Monitor response recorded from August 1, 2008 to March 31, 2011. Data are uncorrected from atmospheric pressure and averagedover one hour. �� variations in neutron flux are evidenced, with several peaks �� counts�h� corresponding to the passage of severe atmosphericdepressions (the highest peak correspond to the Klaus storm on January 25, 2009).

control PC. We developed a dedicated software under VisualBasic 2008 to control the PdBNM data acquisition as well asto manage and time stamp data using a GPS time acquisitioncard installed on the same PC. All these operations can be re-motely controlled via a VPN connection on Internet betweenASTEP and the IM2NP laboratory in Marseille. Every minute(real-time), the PdBNM provides the uncorrected counting ratesfor each detection tube, plus temperature, pressure and hygrom-etry values taken at the beginning of the measurement interval.These data are post-processed to provide hourly and monthlyaveraged values posted on the ASTEP website and available fordownload [17].

III. PdBNM INSTALLATION AND OPERATION

Assembled and firstly operated in Marseille during2007–2008, the PdBNM was transported and permanentlyinstalled on the Plateau de Bure in July 2008. Fig. 2 (top) showsthe ASTEP building where the instrument is installed at thefirst floor. This building was specially constructed to host theneutron monitor in 2007–2008; its metallic walls (includinga 10 cm thick sheet of rockwool) are quasi-transparent forhigh energy neutrons. Fig. 2 (bottom) schematically showsthe position of the neutron monitor inside the building. It iscentered with respect to the circular concrete slab (thickness40 cm) forming the floor of the room. We will investigatein Section IV-D-III by GEANT4 simulation the influence ofthe surrounding building on the neutron monitor detectionresponse. Fig. 3 shows the PdBNM averaged response (onepoint per hour) from August 1, 2008 to June 21, 2010. Thisuncorrected response from atmospheric pressure directly givesan image of the neutron flux variation at the ASTEP location;evidencing variations of this averaged flux at groundlevel essentially due to atmospheric pressure variations.

During its installation, the PdBNM was used to experimen-tally determine the acceleration factor (AF) of the ASTEP loca-tion with respect to sea-level [11]. With strictly the same setup,two series of data, shown in Fig. 4, were thus recorded in Mar-seille and on the Plateau de Bure: the difference between the

Fig. 4. Experimental determination of the ASTEP acceleration factor (AF)from the barometric response of the neutron monitor successively installed inMarseille (2007–2008) and on the Plateau de Bure since July 2008. Experi-mental clouds correspond to one month recording (one point per hour).

counting rates and barometric coefficients for the two locationsallowed us to directly evaluate the acceleration factor of ASTEPwith respect to Marseille location, here estimated to 6.7. Takinginto account latitude, longitude and altitude corrections for Mar-seille location with respect to New-York City (the referenceplace in the world for standardization purposes [11]), the finalvalue of the acceleration factor is . Thisvalue is close to 6.2, the average acceleration factor reported inthe Annex A of the JEDEC standard JESD89A [11] and closeto 5.9, the value given by the QinetiQ Atmospheric RadiationModel (QARM) [18], [19] for quiet sun activity.

On Fig. 4 is also reported the value of the PdBNM barometriccoefficient used to correct the neutron monitor counting ratefrom the effect of atmospheric pressure [1]. Data for baro-metric coefficient calculation have been selected during theperiod August 2008–December 2010 for which no disturbanceof the interplanetary magnetic field and magnetosphere wasreported. The least-squares method was used for the data of


Fig. 5. Comparison between pressure-corrected signals from five neutron mon-itors during the Forbush effect observed on February 15–17, 2011. Data fromJungfraujoch, Roma, Athens and Kerguelen neutron monitors are online avail-able [22]. Data courtesy from the Neutron Monitor DataBase (NMDB).

Fig. 4 to obtain the regression coefficient in the semi-loga-rithmic representation Ln(N) versus whereN is the hourly neutron monitor counting rate at atmosphericpressure P, is the so-called barometric coefficient andis the reference atmospheric pressure. Averaged values of

hPa has been obtained for thePdBNM with a reference pressure equal to hPa atASTEP location. Under this reference pressure, a counting rateof counts h is measured, corresponding to thePdBNM reference counting rate. From these values, PdBNMdata can be easily corrected from atmospheric pressure usingthe following well-known transformation:

(1)

To conclude this section, we would like to illustrate the timedependency of the PdBNM signal and the comparison with theresponse of other ground neutron monitors for a recent “cosmicray Forbush decrease” (after the scientist who first related thissolar event effect [20]), occurring on February 15–17, 2011. Itis the first Forbush decrease in the new solar cycle #24, it meansafter a very long solar minimum from December 2006. The Sunwas active with a X-class flare and many M-class flares duringthis period [21]. Fig. 5 shows the responses of these differentneutron monitors located at Jungfraujoch (Switzerland), Roma(Italy), Athens (Greece), Kerguelen (French islands in southhemisphere) and Plateau de Bure (PdBNM) [22]. The tempo-rary narrow dips in each respective neutron monitor signals aredue to increases in the intensity of the magnetosphere shieldingstrength caused by the sudden solar flare event. The coincidencebetween the different monitor signals is spectacular during theidentified Forbush-effect, as well as the correlation between thesignals in terms of amplitude variations before and after thistransient magnetic disturbance.

IV. PdBNM MONTE CARLO SIMULATION

In this section, we report in detail the modeling and MonteCarlo (MC) simulation of the PdBNM detector to study the re-sponse of the monitor to the total flux of particles in atmosphericshowers, coming from the primary cosmic rays. Our initial ob-jective was to estimate the overall counting rate (and the de-tection efficiency) of the PdBNM with respect to the total in-coming flux and to extract partial contributions from each typeof particle (in particular, neutrons) to the overall counting rate.For this purpose, the atmospheric fluxes of all the basic primaryparticles should be taken into account with corresponding en-ergy spectra and angular dependencies. Following the referencepaper by Clem and Dorman in this field [15], we performed sim-ulations for neutrons, muons , protons, charged pions

, and, in addition, photons, which are considered furtheras primary particles for PdBNM. Besides, the impact of sur-rounding to the functioning of PdBNM (concrete floor, buildingetc.) should also be estimated. As an important intermediate stepthe evaluation of the PdBNM detection response functions foreach type of the incoming particle is required.

Like any other detector system, PdBNM requires an explicitcalibration procedure, either with neutron source or with another(calibrated) detector. MC simulation is an important step com-pleting the experimental calibration. In particular, the calibra-tion procedure (measurements of the counters spectra) servesalso for the correct definition of the detection event in termsof ROI (Region Of Interest in the energy range). For PdBNMthe explicit calibration is planned for the future, but presentedserious practical difficulties due to its specific location, whichincreases the role of MC simulation.

As neutron detection event every single instance of neutroncapture on the He nucleus in the neutron counter is considered[15]. Generally, the neutron detection event should be definedin terms of response signal level from the proportional counter,belonging to certain ROI, because not all of the events of neu-tron capture produce the same response due to the wall effectsetc. It also allows one to reject possible background and noise.Our experience shows that if the ROI is correctly defined thenthe difference between the adopted definition of detection eventand the ROI definition is not huge in terms of the total countingrate, at least it is enough for the estimation of the relative con-tributions from each type of particle.

Besides, due to the processes of multiple secondary neutronproduction in the elements of PdBNM (basically, in the leadproducer), in this study we have to distinguish between the fol-lowing quantities: overall counting rate of PdBNM, its totalcounts and detection efficiencies for given particle species. Bydefinition, the efficiency of detecting the given primary particlecannot exceed 1; for the total counts (which is always referredto given particle type) it is not so, because for any single pri-mary incoming particle having high enough kinetic energy thereis a certain probability of multiple neutron production and de-tection (see the exact definition of the “total counts” quantitybelow). By overall counting rate we always mean the sum oftotal counting rates, corresponding to all the components of nat-ural background. The multiplicity of secondary neutron produc-


TABLE ILIST OF THE CONSIDERED GEANT4 CLASSES IN THE SIMULATION FLOW FOR THE DESCRIPTION OF NEUTRON INTERACTIONS

tion in the parts of PdBNM and their detection in counters willbe discussed below.

A. GEANT4 Physics Models Involved in Simulations

To perform MC simulations of the PdBNM detector withinthe GEANT4.9.1 toolkit, we composed the list of physical pro-cesses on the base of the reference physics list QGSP_BIC_HP[23], also including the additional interactions of neutrons withpolyethylene (PE). First of all, concerning the hadronic interac-tions in general, in QGSP group of physics lists the Quark GluonString Precompound model is applied for high energy interac-tions of protons, neutrons, pions, kaons and nuclei. The high en-ergy interaction creates an exited nucleus, which is passed to theprecompound model describing the nuclear de-excitation. Nu-clear capture of negative particles is simulated within the ChiralInvariant Phase Space (CHIPS) model. QGSP_BIC_HP list in-cludes binary cascade for primary protons and neutrons withenergies below GeV, and also uses the binary light ioncascade for inelastic interaction of ions up to few GeV/nucleonwith matter. In addition this package includes the data drivenHigh Precision neutron package (NeutronHP) to transport neu-trons below 20 MeV down to thermal energies [23].

It is a well known fact that the moderation of neutrons withkinetic energies below 4 eV in PE should be considered in a spe-cial way (see e.g., [24] and references therein). In this low-en-ergy region the scattering of neutrons on the hydrogen nucleiin PE cannot be treated as scattering on free protons due to thepossible excitation of vibrational modes in PE molecules. Suchcollective motion of molecules significantly change the thermalneutron scattering characteristics in PE, so dedicated thermalscattering dataset and model should be included for neutron en-ergies less than 4 eV to allow the correct treatment of neutronmoderation and capture processes in the elements of PdBNM.As QGSP_BIC_HP list in GEANT4.9.1 doesn’t contain thermalscattering, this process was introduced and the list of neutroninteractions was re-composed as it is shown in Table I (modelsand datasets are available since GEANT4.8.2). Another mod-ification of the QGSP_BIC_HP list in the current study con-sists in adding (enabling) the process of muon-nuclear interac-tion for high energy muons from atmospheric showers, which isdisabled by default. However, it was found that mostly the con-tribution of such processes is not very important here. No otherchanges in QGSP_BIC_HP list were made.

Fig. 6. Front (top) and side (bottom) view of the test device for compara-tive modeling between GEANT4 and MCNPX codes. The He neutron countermodel LND 253109 with a 2 inches diameter is inserted into a polyethylene tubewith variable external diameter.

B. Comparison With MCNPX Simulation Results

To check the adequacy of the adopted physics in GEANT4 itis very informative to compare the results of modeling some testdevice within different MC codes, e.g., to simulate some simpleneutron detector within GEANT4 and MCNPX. Such an inde-pendent cross check could reveal possible mistakes in the list ofphysical interactions. As a test system we consider the employedin PdBNM He neutron counter LND 253109 [16] inserted intoPE tube with variable external diameter, as shown in Fig. 6. Let’sstudy the dependence of this detector’s response from the en-ergy of mono-energetic neutron flux, taking the neutron sourcefor the simulation as a homogeneous and parallel neutron beamfully illuminating the lateral surface of the system perpendicularto the counter’s axis. In Fig. 7, the results obtained are shownfor the bare counter and different thicknesses of surroundingPE. Symbols correspond to MCNPX 2.6.0 evaluation [25], solidlines—to GEANT4.9.1. Statistical uncertainty for MCNPX re-sults is less than 1% and is not shown. For GEANT4 the numberof primary neutrons for each energy value is , so for mostof points, the statistical uncertainty is much less than 1% and isalso not shown. Although the divergence between GEANT4 andMCNPX at the lowest and highest energies is still significant,we clearly observe a general agreement in the shape of curvesand reasonable numerical agreement, which confirms the cor-rectness of the adopted physics list in GEANT4.


Fig. 7. Response functions of the system defined in Fig. 6 for a bare counterand different external polyethylene tube diameters.

The evaluation of these response functions in GEANT4 simu-lation practically consists in directly counting the neutron detec-tion events, i.e., the events of neutron capture on He nuclei inactive volume of the counter. During the simulation such eventsare identified by the reaction products—proton and triton in thefinal state of the reaction. But it is also very useful to calcu-late within GEANT4 the energy-binned neutron fluence in theactive volume of the counter [26] in order to reproduce the pro-cedure of calculating the “Tally F4” in MCNPX (see [26], [27]and Appendix for detailed explanation). Convolution of this flu-ence with neutron capture cross section on He is proportionalto the value of response function. Performing this operation inthe current study it was explicitly checked that both methods ofthe test system response evaluation lead to the same results.

C. Input Data for Realistic Particle Sources

To study the response of PdBNM to natural radiation, oneneeds realistic energy spectra for each component of the nat-ural background. Energy spectra of particles , , , , ,

, in cosmic-ray induced atmospheric showers, which areemployed in corresponding GEANT4 particle sources to sim-ulate real atmospheric fluxes, are available in the literature oron the web as functions of latitude, longitude and altitude. Theyare obtained from direct measurements and/or from MC sim-ulations. For the neutron flux we use the well-known scalablespectra from [28], which are actually the part of JEDEC Stan-dard JESD89A [11]. Another possibility to obtain all the re-quired spectra is, for example, to refer to the QinetiQ Atmo-spheric Radiation Model (QARM) [18], [19] or to the PARMAmodel [29], [30], which are specifically developed for predictionof the radiation in the atmosphere for a given location and date.In our calculations we mostly employ the PARMA model; thecomparison between its predictions in the neutron part with thespectrum from JESD89A is shown in Fig. 8. MC simulations ofPdBNM are needed both for Marseille and ASTEP locations. Asan example, in Fig. 9 the PARMA differential fluxes for muonsand protons are shown at ASTEP.

Another important issue in MC simulation is the strong zenithangular dependence of atmospheric showers. To make GEANT4

Fig. 8. Comparison between the JEDEC and PARMA atmospheric neutronspectra given for the reference location corresponding to New York City. After[28] and [29].

Fig. 9. Differential fluxes of muons and protons given by the PARMA modelfor ASTEP conditions.

primary particle sources more realistic, we introduce in simula-tions the angular dependence of the primary flux intensity in theform:

(2)

where is the zenith angle.Equation (2) was employed in the GEANT4 built-in General

Particle Source [31], allowing us to generate the primary par-ticles with such a given angular distribution. For neutrons weadopt [32], and for muons [33].

D. Simulation Results

1) Detection Response Functions: In [15] the detection re-sponse functions of neutron monitor NM-64 are evaluated fordifferent particle species as “total counts” versus the energy ofmono-energetic particle fluxes, arriving at the upper edge ofmonitor. For this purpose the FLUKA toolkit was employed.To perform the same analysis for PdBNM in GEANT4.9.1, letthe incident mono-energetic particles be uniformly distributedupon the upper edge of the PdBNM and arriving in the vertical


Fig. 10. Plateau de Bure Neutron Monitor detection responses for neutrons,protons, muons and pions. Lines correspond to GEANT4, large empty symbolscorrespond to MCNPX (CEM and LAQGSM models). For any given kineticenergy, the simulation run consists of �� primary particles (events).

direction, as it is done in [15]. The obtained dependences of thetotal counts from the energy for different particle species areshown in the Fig. 10. Note, that for the given simulation run bydefinition one has:

(3)

and for the primary particles with high energy this value exceeds1 due to the massive capture of neutrons coming from the pro-cesses of secondary neutron production with high multiplicity(see detailed analysis below). For the case of primary neutrons,if there are no secondary neutrons produced or captured thentotal counts is exactly equal to the detection efficiency.

The construction of PdBNM monitor is very similar to theconsidered in [15] (the difference is basically in geometrical di-mensions, but they are still close to each other), so the detectionresponse functions of these monitors can be qualitatively com-pared. Despite the corresponding curves from [15] cannot besuperimposed to Fig. 10, their behavior is in convincing agree-ment. In particular, all the relative variations are very similar,although for NM-64 the functions are not normalized to the sur-face area of monitor.

Additionally, the same modeling of PdBNM was performedwithin MCNPX 2.6.0 [25] for several values of initial ener-gies of neutrons, negative muons and protons (these particletypes give dominant contributions to the overall countingrate). MCNPX data points are presented by symbols as ex-plained in Fig. 10. The models employed are CEM03.01 andLAQGSM03.01 [34]–[37]. The results obtained demonstrategood agreement, and we can also conclude again that theadopted physical picture of the neutron interactions with matterin GEANT4.9.1 is adequate and close enough to that employedin FLUKA and in MCNPX 2.6.0.

2) Secondary Neutrons Multiplicity: The monitor PdBNMcontains a great amount of lead (about 2 tons), so the contribu-tion of the secondary neutron production to the total countingrate deserves separate investigation. First of all, to verify MC

Fig. 11. 2-dimensional histograms corresponding to the projections of PdBNMto coordinate planes (front and side views) and describing distribution of thesecondary neutrons production vertices from incoming muon flux.

TABLE IIDEPENDENCE OF THE EFFICIENCY, TOTAL COUNTS AND SECONDARY NEUTRON

PRODUCTION RATIO ON THE PRIMARY NEUTRON ENERGY

simulation it is quite informative to visualize the distributionof the secondary neutron production vertices within the volumeof the monitor. The example of such a distribution is given atFig. 11 for the case of primary muons having energy spectrumfrom PARMA model [29], [30] for the ASTEP conditions andarriving in the vertical direction, which is axis (for simplicitywe do not reproduce the realistic angular distribution of atmo-spheric muons in this example). Naturally, the lead tubes pro-duce most of such vertices, although PE walls also contribute.To analyze the role of secondary neutrons in the detection of pri-mary particles, GEANT4 simulations were performed for bothmono-energetic primary particles and realistic spectrum. For ex-ample, let’s consider in detail the neutron curve at Fig. 10. Foreach numerical point of the curve the columns of Table II con-tain the primary neutron kinetic energy , detection effi-ciency, total counts and the ratio of the total number of sec-ondary neutrons produced in the run to the number of pri-mary neutrons . Any primary particle (neutron,muon, etc.) is considered as detected in PdBNM if at least oneneutron (either primary or secondary) is captured in He tubein the current event. Neutrons with energies below 1 MeV arealmost completely reflected by 80 mm PE walls of monitor.Starting from a few MeV, increases very rapidly, finallyexceeding for the energies of few tens MeV. The ob-served difference between efficiency (which is proportional tothe number of detected primary particles) and total counts is ex-plained by the capture of these additional neutrons in He tubeswithin given event.

In principle, to estimate the total counting rate of PdBNMfrom the atmospheric neutron flux it is sufficient to convolutenow the neutron curve from Fig. 10 with the spectrum fromFig. 8. But we prefer to perform direct MC simulation of


TABLE IIICONTRIBUTIONS TO THE TOTAL COUNTING RATE FOR THE GIVEN PARTICLE TYPE FROM THE EVENTS

WITH DIFFERENT MULTIPLICITIES OF NEUTRON CAPTURES IN He COUNTERS

TABLE IVDEPENDENCE OF THE EFFICIENCY, TOTAL COUNTS AND SECONDARY NEUTRON PRODUCTION RATIO ON THE PRIMARY PROTON AND NEGATIVE MUON ENERGY

Fig. 12. Set of the histograms connecting the number of secondary neutronsper event and the number of neutron captures in He counters for the JEDECprimary neutron spectrum. Each curve corresponds to the fixed number of neu-tron captures in event.

PdBNM for the realistic neutron source with JEDEC spectrum,thus avoiding, for instance, the errors of numerical interpolationand integration of strongly varying functions over the hugeinterval of energy.

In Fig. 12 the histograms obtained for the JEDEC neutronspectrum are presented, which connect the number of secondaryneutrons per event and the number of neutron captures in Hecounters. The neutron production multiplicity here can achievevery high values (up to 120 secondary neutrons per event), andthe same for the neutron capture multiplicity (up to 14 capturedneutrons per event). From this data the contribution of the sec-ondary neutrons to the total counting rate can be analyzed. Infact, the histogram in Fig. 12 labeled by digit 1 corresponds tothe events characterized by a single captured neutron, primary

or secondary. Its first bin contains the events with no secondaryneutrons produced, thus describing the capture of only (mod-erated) primary neutrons, which corresponds to a (limited) per-centage of 9.7% of all the captured neutrons in the simulationrun. Totally this histogram accumulates 51.1% of the capturedneutrons. The other 48.9% of neutrons are involved in multiplecaptures (double, triple, etc.) which distribution is summarizedin Table III in the following way: the integral from the histogram1 in Fig. 12 divided by the total number of captured neutronsin the run is equal to 0.511; the integral from the histogram 2(events with double neutron captures) multiplied by 2 and di-vided by the total number of captured neutrons is 0.26 etc.

The same analysis can be performed for primary particles ofany type. It is clear that in detection of the particles other thenneutron the role of secondary neutrons production is principal(they are detected as far as they are able to produce neutrons).Considering primary protons and negatively charged muons, forthe corresponding curves in Fig. 10 we obtain data shown inTable IV. It is clear, that for protons the secondary neutron pro-duction multiplicity is even higher, then for primary neutrons,and the corresponding multiple neutron capture processes inHe are more intensive, as it follows from Table III. For negative

muons, there is a primary energy interval with high secondaryneutron production ratio, where the processes of muon modera-tion and capture in the producer are effective. But for any energythe difference between the detection efficiency and total countsdoes not become as large as for primary neutrons and protons.

3) Impact of the Surrounding to PdBNM Counting Rate:Another important issue for simulation is the impact of sur-roundings on the PdBNM overall counting rate. As describedin Section II, the monitor is placed inside the first floor of theASTEP building, which possibly distorts the original radiation


Fig. 13. ROOT screenshots of GEANT4 simulations showing the tracks ofprimary and secondary particles (green lines—neutrons, yellow—gammas,red—electrons) for different views of the instrument: a) top view including theASTEP surrounding building (first floor); b) face view of the instrument placedon the concrete floor of the building first floor; c) detailed view at the level ofthe polyethylene box of the neutron monitor.

background. Although the building is relatively light, it containssignificant amounts of steel and concrete, which could lead tothe secondary neutron production (in steel walls) and neutronreflection (on concrete floor), so the effect of it to the countingrate should be estimated.

The building was included into the total geometry of mod-eling as shown in Fig. 13(a). In this simulation, we found thatthe number of secondary neutrons produced in the system“building monitor” for the primary protons and neutrons is

times higher than for standalone monitor, and it is timeshigher for the primary negative muons. But the resulting impactof this additional neutron flux to the total counts of PdBNMis not so significant: it increases the total counts only in 4.1%for primary neutrons, 4.5% for protons and 1.7% for negativemuons. Thus, these extra neutrons are mostly not detected byPdBNM.

To conclude this part, Fig. 13(b) and (c) show for illustrationdifferent views of a simulated events for five incoming atmo-spheric neutrons (with energy 100 MeV) interacting with thematter of the PdBNM.

4) Contributions of Different Particle Species to PdBNMCounting Rate: It is clear that until the experimental cali-bration of PdBNM is done, only the relative contributionsfrom the different particle species of the natural radiationenvironment into the overall counting rate of PdBNM monitorcan be reliably estimated from MC simulation. Such contri-butions are directly proportional to the following factors: i)the partial detection efficiency (or total counts, see Fig. 10)for the given particle type and ii) its partial flux intensity inthe natural conditions depending on the latitude, longitude,altitude, atmospheric pressure, solar cycle etc. To obtain therealistic partial fluxes for the given conditions, we employ the

TABLE VRELATIVE CONTRIBUTIONS TO THE PDBNM OVERALL COUNTING RATE OF

THE DIFFERENT PARTICLE SPECIES. FOR THE SIMULATION OF THE INSTRUMENT

RESPONSE AT ASTEP LOCATION, THE IMPACT OF THE BUILDING IS TAKEN

INTO ACCOUNT

PARMA model and construct a GEANT4 particle source as it isdescribed in Section IV-D-III. Calculated relative contributionsto the PdBNM overall counting rate from different types ofparticles are listed in Table V for two completely differentlocations—Marseille and Plateau de Bure. Higher and lowerlimits for the percentages arise from the difference between thedetection species of efficiency and total counts which, in itsturn, is caused by the detection of secondary neutrons, as it isdescribed above. It is also worth noticing, that to evaluate totalcounts for primary neutrons it is necessary to consider not onlyJEDEC part of the neutron flux (from 1 MeV and higher) butalso lower part from 1 eV to 1 MeV (available both in PARMAand QARM models), which gives significant contribution dueto the very intensive flux in this low energy region.

Obviously, PdBNM is quite sensitive to atmospheric protonsand negatively charged muons. The latter, being moderated in itsmaterials and captured in nuclei produce a lot of secondary neu-trons (it is also shown in Table IV). The sensitivity of PdBNM toother particle species (like electrons and pions) is estimated asvery low, either due to the small corresponding value of partialdetection efficiency (as for electrons) or low atmospheric partialflux intensity (charged pions).

V. CONCLUSION

In conclusion, this paper summarizes a five years effort tocompletely develop, install and characterize by simulation a newneutron monitor permanently installed on the Plateau de Bure inFrench south Alps. The primary interest of the instrument wasmotivated by the real-time measurement of the atmospheric neu-tron flux impacting microelectronics experiments deployed inaltitude to precisely investigate the impact of natural radiationon electronics. Nevertheless, data obtained with this instrumentcan be also considered for cosmic ray investigations equally toother monitors installed around the world. Almost three years ofcontinuous operation demonstrated high stability and reliabilityof the instrument which will soon be integrated into the neutronmonitor database (NMDB) network for real-time data accessi-bility on the web.

Our modeling and numerical simulation work, performedwith two different Monte Carlo codes, GEANT4 and MCNPX,allowed us to obtain the neutron monitor detection responsefunctions for neutrons, muons, protons and pions and theirrespective contribution into the overall counting rate of theinstrument considering realistic atmospheric particle sources.We highlighted a relative importance of proton and negative


muon contribution in the monitor response (for muons—es-pecially at sea-level), and a negligible impact of the ASTEPsurrounding building on the counting rate. Finally, we carefullycharacterized the secondary neutron multiplicity processes,essential to understand the physics of this neutron monitor. Fu-ture work should include modifying the electronic acquisitionchain of the instrument to experimentally characterize such aneutron multiplicity on the electrical signals delivered by thedetection tubes and to perform an experimental calibration ofthe PdBNM.

APPENDIX

MCNPX SIMULATION OF NEUTRON FLUX AND REACTION

RATE IN THE He COUNTER WITH TALLY F4: FLUENCE

RESPONSE CALCULATION

In order to determine the fluence response by MCNPX cal-culation, it is assumed that the number of He p T reactionsin the sensitive cell of the proportional counter is corre-lated to the reading of the counter (detector signal). The fluenceresponse is determined using an estimation, or tally, of the neu-tron average flux (integrated over time) on the cell containingthe sensitive gaseous volume of detection. This tally is definedby the “F4” physics card of the MCNPX code and consists in ob-taining the so-called track length estimate of fluence throughthe sensitive cell. It is proportional to the sum of those K pathlengths of neutrons having the energy that pass throughthe sensitive counter volume,

(A1)

Finally, is determined by the tally F4 as the area densityof the number of neutrons (in cm ) normalized to one sourceparticle passing through the sensitive cell.

The fluence response in counts per neutron per cmfor the cylinder of diameter d, uniformly irradiated by a “planeparallel” and mono-energetic neutron beam with incident en-ergy is then given by

(A2)

with is the area of neutron source (in cm ), is the atomicdensity of the He (in cm ), is the volume of the countersensitive cell (in cm ), is the cross section of the reac-tion He T for the neutron energy (in barn).

ACKNOWLEDGMENT

The idea to install a Neutron Monitor on the ASTEP plat-form came after the RADECS Conference organized in Glyfada(Athens) in September 2006 and after the visit of the AthensNeutron Monitor Data Processing Center (ANMODAP) or-ganized in the framework of the IEEE-RADECS SECSTANWorking Group (Susceptibility of Electronic Componentand Systems To Atmospheric Neutrons). The authors wouldlike to particularly thank our fellow colleagues Dr. Jean-LucLeray (CEA) who initiated this legendary visit and Dr. HelenMavromichalaki (Athens University), principal investigator

and scientific leader of the ANMODAP, for her support andcontinuous encouragement. The logistical support of the Insti-tute for Radio-astronomy at Millimeter Wavelengths (IRAM) isalso gratefully acknowledged. Special thanks are due to PierreCox, Bertrand Gautier and Cécile Di Léone for their help.

The first author, S. Semikh, would like to specially acknowl-edge the kind hospitality of the IM2NP laboratory and thegroup headed by J. L. Autran in conducting this research. Theneutron monitor construction was supported in part by theMEDEA Project #2A704 ROBIN and by the French Ministryof Economy, Finances and Industry under research conventions#052930215 and #062930210.

Finally, the authors would like to sincerely thank the Editor,the Senior Editor and the referees for their very valuable com-ments and suggestions on our submitted manuscript that greatlycontribute to improve the clarity of this paper.

REFERENCES

[1] L. I. Dorman, Cosmic Rays in the Earth’s Atmosphere and Under-ground. Norwell, MA: Kluwer, 2004.

[2] J. A. Simpson, “The cosmic ray nucleonic component: The inventionand scientific uses of the neutron monitor,” Space Sci. Rev., vol. 93, pp.12–32, 2000.

[3] J. F. Ziegler and H. Puchner, SER—History, Trends and Challenges,. San Jose, CA: Cypress Semiconductor, 2004.

[4] E. Normand, “Single event upset at ground level,” IEEE Trans. Nucl.Sci., vol. NS-43, no. 6, pp. 2742–2750, Dec. 1996.

[5] R. D. Schrimpf and D. M. Fleetwood, Radiation Effects and Soft Er-rors in Integrated Circuits and Electronic Devices. Singapore: WorldScientific Publishing, 2004.

[6] T. Nakamura, M. Baba, E. Ibe, Y. Yahagi, and H. Kameyama, Terres-trial Neutron-Induced Soft Errors in Advanced Memory Devices. Sin-gapore: World Scientific Publishing, 2007.

[7] M. Nicolaidis, Soft Errors in Modern Electronic Systems. New York:Springer, 2010.

[8] J. L. Autran, P. Roche, J. Borel, C. Sudre, K. Castellani-Coulié, D.Munteanu, T. Parrassin, G. Gasiot, and J. P. Schoellkopf, “AltitudeSEE test European platform (ASTEP) and first results in CMOS 130nm SRAM,” IEEE Trans. Nucl. Sci., vol. 54, no. 4, pp. 1002–1009,Aug. 2007.

[9] J. L. Autran, P. Roche, G. Gasiot, T. Parrassin, J. P. Schoellkopf, andJ. Borel, “Real-time soft-error rate testing of semiconductor memorieson the European test platform ASTEP,” in Proc. 2nd Int. Conf. MemoryTechnol. Design (ICMTD 2007), Giens, France, May 7–10, 2007, pp.161–164.

[10] [Online]. Available: http://www.iram-institute.org/[11] JEDEC Standard Measurement and Reporting of Alpha Particles

and Terrestrial Cosmic Ray-Induced Soft Errors in SemiconductorDevices, JESD89 JEDEC Solid State Technology Association,Arlington, VA [Online]. Available: http://www.jedec.org/down-load/search/JESD89A.pdf

[12] J. L. Autran, G. Gasiot, D. Munteanu, P. Roche, and S. Sauze, “Real-Time soft-error rate characterization of advanced SRAMs,” in Radi-ation Effects in Semiconductors, K. Iniewski, Ed. Boca Raton, FL:CRC Press, 2010, pp. 225–247.

[13] J. L. Autran, D. Munteanu, P. Roche, G. Gasiot, S. Martinie, S.Uznanski, S. Sauze, S. Semikh, E. Yakushev, S. Rozov, P. Loaiza, G.Warot, and M. Zampaolo, “Soft-errors induced by terrestrial neutronsand natural alpha-particle emitters in advanced memory circuits atground level,” Microelectron. Rel., vol. 50, pp. 1822–1831, 2010.

[14] P. H. Stoker, L. I. Dorman, and J. M. Clem, “Neutron monitor designimprovements,” Space Sci. Rev., vol. 93, pp. 361–380, 2000.

[15] J. M. Clem and L. I. Dorman, “Neutron monitor response functions,”Space Sci. Rev., vol. 93, pp. 335–359, 2000.

[16] [Online]. Available: http://www.lndinc.com/products/pdf/568/[17] [Online]. Available: http://www.astep.eu[18] F. Lei, S. Clucas, C. Dyer, and P. Truscott, “An atmospheric radiation

model based on response matrices generated by detailed Monte CarloSimulations of cosmic ray interactions,” IEEE Trans. Nucl. Sci., vol.51, no. 6, pp. 3442–3451, Dec. 2004.


[19] F. Lei, A. Hands, S. Clucas, C. Dyer, and P. Truscott, “Improvement toand validations of the QinetiQ atmospheric radiation model (QARM),”IEEE Trans. Nucl. Sci., vol. 53, pp. 1851–1858, 2006.

[20] S. E. Forbush, “World-Wide cosmic-ray variations 1937–1952,” J.Geophys. Res., vol. 59, no. 4, pp. 525–542, 1954.

[21] [Online]. Available: http://www.swpc.noaa.gov[22] Data of the Neutron Monitors Located at Jungfraujoch, Roma, Athens

and Kerguelen are Online Available on the Website of the MonitorDataBase (NMDB) Project, a Real-Time Database for High Resolu-tion Neutron Monitor Measurements [Online]. Available: http://www.nmdb.eu/

[23] [Online]. Available: http://cern.ch/geant4/UserDocumentation/Users-Guides/PhysicsReferenceManual/fo/PhysicsReferenceManual.pdf

[24] S. Garny, G. Leuthold, V. Mares, H. G. Paretzke, and W. Ruhm,“GEANT4 transport calculations for neutrons and photons below 15MeV,” IEEE Trans. Nucl. Sci., vol. 56, no. 4, pp. 2392–2396, Aug.2009.

[25] , D. B. Pelowitz, Ed., MCNPX™ User’s Manual Version 2.6.0. LosAlamos, NM: Los Alamos National Laboratory, 2008.

[26] V. Mares, G. Schraube, and H. Schraube, “Calculated neutron responseof a Bonner sphere spectrometer with He counter,” Nucl. Instrum.Methods in Phys. Res., vol. A307, pp. 398–412, 1991.

[27] B. Wiegel, A. V. Alevra, and B. R. L. Siebert, Calculations of the Re-sponse Functions of Bonner Spheres With a Spherical He ProportionalCounter Using a Realistic Detector Model, PTB-Bericht N-21, 1994,ISBN 3-89429-563-5.

[28] M. S. Gordon, P. Goldhagen, K. P. Rodbell, T. H. Zabel, H. H. K.Tang, J. M. Clem, and P. Bailey, “Measurement of the flux and energyspectrum of cosmicray induced neutrons on the ground,” IEEE Trans.Nucl. Sci., vol. 51, no. 6, pp. 3427–3434, Dec. 2004.

[29] T. Sato, H. Yasuda, K. Niita, A. Endo, and L. Sihver, “Development ofPARMA: PHITS-based analytical radiation model in the atmosphere,”Radiation Res., vol. 170, pp. 244–259, 2008.

[30] EXcel-Based Program for Calculating Atmospheric Cosmic-Ray Spec-trum, [Online]. Available: http://phits.jaea.go.jp/expacs/index.html

[31] Geant4 General Particle Source, [Online]. Available: http://reat.space.qinetiq.com/gps/

[32] U. Bravar, E. O. Flückiger, K. Godin, Z. C. Hansen, J. R. Macri, M. L.McConnell, R. S. Miller, M. R. Moser, and J. M. Ryan, “Atmosphericneutron measurements with the SONTRAC science model,” in Proc.IEEE Nucl. Sci. Symp. Conf. Rec., 2005, N14-126.

[33] J. Kempa and I. M. Brancus, “Zenith angle distributions of cosmicray muons,” Nucl. Phys. B—Proc. Supplements, vol. 122, pp. 279–281,2003.

[34] S. G. Mashnik, K. K. Gudima, A. J. Sierk, M. I. Baznat, and N. V.Mokhov, CEM03.01 User Manual. Los Alamos, NM: Los AlamosNational Laboratory Rep., 2005.

[35] K. K. Gudima, S. G. Mashnik, and A. J. Sierk, User Manual for theCode LAQGSM. Los Alamos, NM: Los Alamos National LaboratoryReport, 2001.

[36] S. G. Mashnik, A. J. Sierk, K. K. Gudima, and M. I. Baznat, “CEM03and LAQGSM03-new modeling tools for nuclear applications,” in J.Phys.: Conf. Series—EPS Euroconf. XIX Nucl. Phys. Div. Conf., 2006,vol. 41, pp. 340–351.

[37] S. G. Mashnik, K. K. Gudima, R. E. Prael, A. J. Sierk, M. I. Baznat, andN. V. Mokhov, “CEM03.03 and LAQGSM03.03 event generators forthe MCNP6, MCNPX, and MARS15 transport codes,” in Joint ICTP-IAEA Advanced Workshop on Model Codes for Spallation Reactions,Trieste, Italy, Feb. 4–8, 2008, LANL Report LA-UR-08-2931.

the plateau de bure neutron monitor: design, operation and monte carlo simulation

Documents