aa r X i v : . [ h e p - e x ] A ug Measuring the NuMI Beam Flux for MINERvA
Melissa T. Jerkins
Department of Physics, The University of Texas at Austin
On behalf of the MINERvA Collaboration Abstract.
MINERvA is employing multiple tools to understand its neutrino beam flux. We utilizeexternal hadron production data, but we also depend heavily on in situ techniques in which wereduce our hadron production uncertainties by tuning our Monte Carlo to both MINERvA detectordata and muon monitor data.
Keywords: neutrino flux
PACS:
INTRODUCTION
MINERvA is performing low energy precision cross-section measurements using theNeutrinos at the Main Injector (NuMI) beamline at Fermi National Laboratory. Figure 1shows the NuMI beamline, which begins with 120 GeV protons hitting a carbon targetand producing secondary pions and kaons. Magnetic focusing horns focus the hadrons,which proceed into a long decay pipe where they decay to muons and neutrinos. Ahadron monitor measures the hadron content of the beam prior to an absorber throughwhich only muons and neutrinos exit. Three muon monitors sit within excavated alcoves,and muons range out in the rock so that the beam contains almost exclusively neutrinoswhen it encounters the MINERvA detector. In order for MINERvA to make absolutecross section measurements, we need to understand our neutrino beam flux, which isnotoriously difficult to measure. Because of the importance of the flux determination,MINERvA utilizes three different tools in order to understand the flux. We use existinghadronic cross section data to predict our yield of pions and kaons off of our target, butwe also emphasize the importance of in situ flux measurements, particularly beam fittingover both MINERvA detector data and muon monitor data.
SOURCES OF FLUX UNCERTAINTIES
The most significant flux uncertainties come from hadron production uncertainties, butsome are due to beam focusing. NuMI utilizes horn focusing to increase its flux, and Fig-ure 2 shows a simulation of the uncertainties that result from inevitable misalignmentsin the focusing elements [1]. The uncertainties are quite small in the focusing peak,and they are largest in the falling edge of the peak. These misalignments are fairly easy http://minerva.fnal.gov IGURE 1.
Diagram of Fermi National Laboratory’s NuMI beamline preceding the MINERvA detec-tor [1]
FIGURE 2.
Relative error introduced into the flux determination by beam focusing uncertainties [1] to understand and model using Monte Carlo techniques, and the effects of correlationsbetween the various sources of uncertainty are being studied.The dominant flux uncertainties derive from uncertainties in hadron production offof the target, and existing Monte Carlo models are not in good agreement with eachother concerning how to model hadron production. For example, Monte Carlo modelsdo not agree on the average transverse momentum of pions leaving the NuMI target, andthose discrepancies result in different neutrino spectra because the pions are focuseddifferently by the horns.
EXTERNAL HADRON PRODUCTION DATA
Given that modern hadron production data sets continue to improve, MINERvA iscertainly utilizing this information to learn about its flux. MINERvA must acknowledge,however, that we cannot depend entirely on experiments like NA49 [2] to simply deliverour flux measurement. Most hadron production data sets are taken on thin targets, but theNuMI target is approximately two interaction lengths, meaning that reinteractions are anon-negligible 20-30% effect. Even if we had a data set that corresponded perfectly to
IGURE 3.
Distribution in p z and p T of hadron parents that yield a neutrino that interacts in the detectorfor the low energy, pseudo-medium energy, and pseudo-high energy beam configurations [3] our target, we would still have to account for in-beam temporal variations that changethe flux. We also have to account for downstream interactions where hadrons are createdby interactions in material other than the target. BEAM FITS TO MINERVA DETECTOR DATA
To address these effects MINERvA wants to make an in situ flux measurement thatwill average over real effects in the beam, and specifically we want to determine theflux as a function of the underlying hadron parent p z and p T . The hadron momentum iskinematically related to the resulting neutrino energy, and the transverse momentum ofthe hadron determines how well the hadron is captured by the focusing horns. Figure 3shows a distribution of the hadron parents that yield a neutrino that interacts in thedetector, and it illustrates that different NuMI beam configurations focus particles fromdifferent regions in the parent hadron p z - p T space [3].MINERvA intends to utilize the flexible design of the NuMI beamline. In additionto varying the current in the focusing horns, we can also move our target in and outof the first horn, which changes which hadrons are focused. This technique is notas effective as actually changing the separation between the focusing horns, but it ismuch faster and easier to accomplish. By producing neutrinos of the same energy usingseveral different beam configurations, we are able to deconvolve focusing effects, hadronproduction off of the target, and neutrino cross sections. Because of these flexible beamconfigurations, we can attempt to tune our hadron production yields to match data fromthe MINERvA detector [3, 4]. Each p z - p T bin contributes with a different weight ineach beam configuration. We create a functional form to parameterize our Monte Carloyields off the target in terms of p z and p T , and we then warp the parameters so thatthe Monte Carlo matches the real data. We perform a chi-squared minimization acrossmultiple different beam configurations to learn how to tune our Monte Carlo.The results of such a fit are a set of weights that should then be applied to MonteCarlo pion and kaon yields. MINOS successfully lowered their flux uncertainties usingthis technique, although they used an inclusive charge current data sample, which is notsuitable for deconvolving neutrino cross sections. (GeV) n E0 2 4 6 8 10 12 14 16 f l u x n R e l a t i ve un ce r t a i n t y on FIGURE 4.
Error band applied to a low-energy neutrino flux that represents our uncertainty beforetuning our Monte Carlo. Uncertainties due to kaon production and tertiary production are not included. “Standard Candle” Sample
MINERvA wants to perform its fit on a “standard candle” data sample in which thecross section is approximately constant with neutrino energy. Ultimately we will performour fit over several different standard candle samples, one of which will be quasi-elasticevents of moderate Q . The energy independence of the cross section for this sampledoes not depend on axial mass. Low Q events are excluded because of uncertaintiesdue to nuclear effects, and high Q events are excluded because of reconstruction diffi-culties as well as because of a non-neglible M A dependence. MINERvA will determinethe shape of its flux by performing a global beam fit on our quasi-elastic sample. We willfix our overall normalization by using our inclusive charge current sample above approx-imately 20 GeV and comparing that to data from experiments like CCFR, CCFRR, andCDHSW [5, 6, 7], which measured the cross section in that region to within ∼ Estimate of Flux Error Bands
While MINERvA continues to collect “special run” data using NuMI’s flexible designto map out the parent p z - p T space, we are examining the size of the flux uncertaintiesthat we expect to obtain after we perform a global beam fit. Before we attempt anyhadron production fitting, we want to know what the state of “current knowledge” is.The uncertainties in current knowledge can be assessed by comparing hadron productionmodels of various Monte Carlo simulations and observing the different neutrino fluxesthey yield. Figure 4 shows an error band applied to a low-energy neutrino flux thatrepresents our uncertainty before tuning our Monte Carlo. It includes model differencesin pion production, beam focusing uncertainties, and an overall 5% yield uncertainty for p + production off of the target. It does not include model differences for kaons off ofthe target, which would primarily affect the high energy tail of the spectrum, and it doesnot include uncertainties due to downstream interactions, which are being studied.e would like to know how much those uncertainties could be reduced by perform-ing a hadron production tuning fit, but until MINERvA has taken and analyzed all of itsspecial run data, we cannot actually perform such a fit. A study is underway, however, toestimate what the results of that fit might yield by warping the hadron production modelin Geant4 [8] Monte Carlo until it matches the neutrino fluxes yielded by Fluka [9]Monte Carlo. The assumption is that our underlying parameterization of hadron produc-tion off the target is flexible enough that the uncertainties in the fit will be similar whenwe actually perform it on real data. BEAM FITS TO MUON MONITOR DATA
In addition to fitting MINERvA detector data to better understand our flux, MINERvAalso utilizes in situ muon monitoring, which is our second key tool for measuring ourflux. The NuMI beamline preceding the MINERvA detector contains four excavatedrock alcoves, three of which are instrumented with arrays of ionization chambers, andwe plan to instrument the fourth alcove soon to increase our sensitivity. Each of ourthree arrays of ionization chambers are filled with helium gas. The muons ionize thegas, and we detect the resulting electrons. Significant amounts of rock separate themonitors, meaning that each monitor is sensitive to a different muon energy threshold,which translates into a different neutrino energy threshold. This type of muon monitoringis a proven technique that has been used for decades in both flux measurements andbeam diagnostics, and NuMI’s flexible beam design only increases the utility of thesemeasurements [10].We can tune our Monte Carlo hadron production model to match muon monitor datain the same way we tune it to match MINERvA detector data. We acquire data in themuon monitors from various beam configurations, and we perform a global fit to warpour underlying hadron production model in our Monte Carlo so that it agrees with thedata.The data points over which we perform the fit are integrals of muon monitor fluxesfrom different beam configurations, and we correct the data for changes in variables likeambient temperature and pressure. In order to tune our Monte Carlo using this data,we also have to apply some corrections to our Monte Carlo data. Most importantly,we have to apply an overall scale factor to account for the charge we expect in themonitor per muon. We obtain this factor from beam tests done on prototype chambers.Unfortunately this factor is very sensitive to gas impurities and can vary 5-10% even dueto a contamination as small as 20 ppm. Once we have made the necessary correctionsto both data and Monte Carlo, then we can tune our Monte Carlo to match the muonmonitor data taken across multiple beam configurations. MINOS performed such afit in which they floated the p + parameters and fixed the p + / p − and p + / K + ratiosusing existing hadron production data, and they successfully obtained a flux shapemeasurement using muon monitor data [11]. Many backgrounds had to be estimatedin that analysis, which contributed a significant amount of uncertainty to the fit andunfortunatley forced MINOS to normalize their flux measurement using near detectordata. MINERvA intends to repeat this analysis, but we hope to reduce some of thedominant uncertainties, particularly those related to gas impurities and backgroundsrom d -rays. We are attempting to improve the purity monitoring of the gas, as wellas our Monte Carlo modeling of d -ray production. Modeling d -Ray Production d -rays are knock-on electrons created by muons as they travel through rock, air, oreven the monitors themselves. These electrons can penetrate the monitor, ionize thehelium gas, and result in a background signal. We expect the energy deposited by d -rays to increase with muon momentum and decrease with the amount of air in front ofthe monitors. Summing over all of the muons that reach each alcove, the Monte Carlopredicts that d -rays can contribute as much as 30% of the muon monitor signal, meaningthat accurate Monte Carlo modeling of this background is essential.To better understand d -ray production, we placed aluminum absorber plates in frontof the monitors to deliberately introduce more d -rays into the data in a controlled way.By measuring the change in d -ray production with and without these various curtains ofabsorbers, we hope to constrain Monte Carlo d -ray production. A full analysis of thisdata is underway and will greatly improve the utility of the muon monitor fit. CONCLUSIONS
MINERvA is employing multiple tools to measure, check, and cross-check its flux. Weare gleaning what information we can from hadron production data, and we are depend-ing heavily on in situ techniques where we reduce our hadron production uncertaintiesby tuning our Monte Carlo to both MINERvA detector data and muon monitor data. Bydiversifying our measurement tools, we hope to lower our flux uncertainties from ∼ REFERENCES
1. Z. Pavlovic, “A Measurement of Muon Neutrino Disappearance in the NuMI Beam,” PhD Thesis, UTAustin (2008).2. C. Alt et al , Eur. Phys. J. C , 897 (2007).3. S. Kopp, Phys. Rep. , 101 (2007).4. M. Kostin et al , “Proposal for continuously-variable neutrino beam energy for the NuMI facility,”Fermilab-TM-2353 (2001).5. P. S. Auchincloss et al , Z. Phys. C , 411 (1990).6. R. Blair et at , Phys. Rev. Lett. , 343 (1983).7. P. Berge et al , Z. Phys. C , 443 (1987).8. S. Agostinelli et al , Nucl. Instrum. Methods Phys. Res. A , 250 (2003).9. A. Ferrari et al , “FLUKA: a multi-particle transport code,” CERN 2005-10 (2005), INFN/TC_05/11,SLAC-R-773.10. S. Kopp et al , Nucl. Instr. Meth. ,568