VIEWS: 0 PAGES: 26 POSTED ON: 12/24/2012
Controlling Systematics in a Future Reactor q13 Experiment Jonathan Link Columbia University Workshop on Future Low-Energy Neutrino Experiments April 30 − May 2, 2003 A Simple Counting Experiment Study Look for disappearance in the ratio R, defined as N far L2far R 2 ε N L near near Where: • The N’s are the number of observed events • The L’s are the baselines and • e is the relative efficiency of the near and far detectors. Disappearance is measured as a deviation of R from 1 and the sensitivity to sin2q13 at 90% CL is just 1.64 R S m13 2 L far ( E ) ( E ) sin(1.27m 2 13 )dE E E Counting vs. Shape • Huber, Lindner, Schwetz and Winter have shown that a pure shape analysis works well with large statistics. • A combined shape and rate analysis improves sensitivity over a pure rate analysis only slightly at the scale of current proposals. 50 tons, 6 GW, 3 years and 1200 meters Counting Experiment Shape & Rate • Therefore, the counting experiment is sufficient to study/compare these scenarios. Significant Contributions to the Error 1. Statistics in the far detector N far N bg stat N far 2. Uncertainty in the relative efficiency of the near and far detector 2 e (with movable detectors) N near f where f is the fraction of run time used for cross calibration 3. Uncertainty in the background rate in the far detector bg rate N bg bg N far Kr2Det Proposal • This elegant proposal can be simply stated as 2 detectors and one reactor • Identical near and far detectors target the dominate source of error in CHOOZ and Palo Verde − flux uncertainty • It explicitly address the background error by doubling the depth compared to CHOOZ and has 65 reactor off days a year • The reactor power (~2 GW) is low by modern standards • The 1000 metes far baseline may not be ideal Few Words on Methodology This analysis starts with the assumptions in the Kr2Det proposal (Mikaelyan et al.): • Two identical, 46 ton (fiducial) detectors at 115 and 1000 meters • 55 events/day in far detector, 4200 near • Reactor is on for 300 days in a year • Relative efficiency of near and far detectors know to 0.8% • 600 mwe shielding Background of 0.1 events/ton/day • The background rate is measured during reactor off days Spreadsheet Study Allowing the variation of: • reactor power • near and far baselines • number of far detectors • run time • background rate • fraction time for cross calibration • detector size • background sensitivity • one or two reactor scenarios • reactor capacity factor den 0.85 g/cm^3 L near 150 m coverage 0.2 depth 300 mwe flux 2.00E+20 nu/s/GWth L far 1200 m diam_pmt 8 in veto_ineff 0.05 power_th 6.1 GWth near flux 4.315E+11 /cm^2/s area_pmt 0.032429 m^2 eff depth 6000 mwe flux0 1.22E+21 nu/s far flux 6.742E+09 /cm^2/s near_bg 0.01 /ton/day years 3 Hden 7.85E+22 H/cm^3 far_bg 0.01 /ton/day upfrac 0.89 xsec*eff 5.59E-44 cm^2 near_bg_err 50 % uptime 3.25E+02 day/year far_bg_err 25 % tons/unit fid far units active frac near bg sub far bg sub near events near error far tot/unit far err/unit far total eff error R error 50 1 1 547.5 547.5 7977144 2837.72 124642 353.82128 124642 0.008 0.0086 50 1 0.9 492.75 492.75 7179429 2690.84 112178 335.66464 112178 0.001669 0.003621 25 2 0.9 246.375 246.375 3589714 1898.72 56089 237.35074 112178 0.00236 0.00364 16.6 3 0.9 163.593 163.593 2383570 1546.1 37243 193.40784 111729 0.002897 0.003666 10 5 0.85 93.075 93.075 1356114 1165.49 21189 145.88377 105946 0.003136 0.003662 5 10 0.9 49.275 49.275 717942 847.702 11217 106.14271 112178 0.005278 0.00379 far units = The number of identical detectors at the far location (n). active frac = The fraction of time that each far unit spends at the far location (time spent at the near location is 1-active frac). bg sub = The total number of background events subtracted from each unit. near events = The number of signal events (after BG subtraction) seen by the near detector during the active fraction (N_n). far events = The number of signal events seen by each far detector during the active fraction (N_f). eff error = The error on the reletive efficiency (eff) of a far detector wrt the near detector as measured side-by-side. R = far^2/near^2 * 1/n * sum(N_f*eff)/N_n R error = The error on R. Ways of Improving the Statistics at the Far Detector There are three ways… 1. More target volume at the far detector site 2. More reactor power 3. More running time Twice Volume = Twice Power = Twice Run Time (Statistical errors only) More Target Volume at the Far Detector Site Small near detector and bigger far detector: Important errors may not cancel if the detectors are not identical Bigger detectors near and far: Error cancellation intact Possible attenuation problems in large Gd loaded detectors Detectors are impossible to move More same size far detectors: The errors scale like one big detector Could phase in the experiment or improve sensitivity by adding more detectors $$$$ Add More Reactor Rower See earlier talk: We can get ~9 GW with French reactor sites ~8 GW in Germany, ~7 GW in the U.S. and Less elsewhere. I’ll show later in this talk that no reactor off running is not needed. More Running Time I think that it is a bad idea to plan on an extra long run (more than 3 years) • More time for efficiency to drift (i.e. degradation of Gd loaded scintillator) • Hard on young scientists • Could get beat by off-axis Extra running time could be useful if we get to the end of our run and we have a marginal (≤3) effect, but we must not be systematics limited. Controlling the Relative Efficiency Systematic • Bugey (the only near/far reactor exp.) had e = 2% • 1.8% if you ignore the solid angle error • Kr2Det assumes 0.8% What value should we be using? How will we determine/measure e? One possibility is movable detectors Movable Detectors This idea originated with Giorgio Gratta and Stan Wojcicki • Our idea is to have a far detector(s) that can be moved to sit at the same baseline as the near detector • The two detectors record events in the same flux at the same time (head-to-head calibration) • Relative efficiency error: 2 e N near f • Near running fraction of 10 to 15% optimizes the total error • A movable detector experiment is best achieved by connecting the two detector sites by a tunnel • Such a tunnel might cost $10 to $20 million depending on the site geology, topology and hydrology. Sensitivity of Kr2Det Kr2Det is ultimately limited by the 0.8% error on the relative efficiency of their two detectors. Physics Reach of Kr2Det Proposal 0.045 Kr2Det Proposal Sensitivity at 90% CL 0.04 Systematics Limit 0.035 0.03 0.025 0.02 0 5 10 15 20 Years The limit in sensitivity imposed by the 0.8% error. It is possible to overcome this limit with a shape analysis and high statistics (à la Huber, et al.) but only after about 65 years of running (~6000 GW ton yrs)! One can do better with a movable far detector… Sensitivity of Kr2Det with Movable Detectors 10% of the running time is spent doing the cross calibration. Physics Reach of Kr2Det with Movable Detector 0.04 Modified Kr2Det Sensitivity at 90% CL 0.03 0.02 0.01 0 0 5 10 15 20 Years 12 years With this modification you get to a sensitivity of 0.01 at m2 of 2.5×10-3 eV2 by adding fiducial mass (138 tons) or time (12 years). The effect is even more dramatic when considering reactor sites with higher power, where the systematic limit is reached sooner. Moving Detectors at a 6 GW Site Consider 50 ton target detectors at 150 0.03 6 GW Reactor, 1200 meter Baseline meters and 1200 meters and a 3 year run. 0.025 Fixed Detectors Movable Detectors Sensitivity at 90% CL 0.02 The far detector spends 10% of the run 0.015 time at the near site for cross calibration. 0.01 Or the relative efficiency is measure to 0.005 0.8% with fixed detectors 0 0 5 10 15 20 Years Controlling Uncertainty in the Background Rate 1. Measure background with reactor off time 2. Put detectors very far underground so that the background is insignificant (The KamLAND solution) 3. Create a large effective depth with an external veto/shielding system (The KARMEN solution) 4. Measure the heck out of it Combining 3 and 4 seems to work well Measure Background with Reactor Off Time This works best at single reactor sites • Commercial reactors can have as little as 3 weeks of down time every 18 months. • For 3 GW, 300 mwe, 1200 BL bg ≈ 2×far • Need 2 months a year to bg ≈ far CHOOZ ran the detector before their reactors were Extrapolation to zero commissioned power from CHOOZ Over time the Gd loading degraded their attenuation length. When they were forced to lower their trigger threshold their background rate changed When extrapolating to zero power at two reactor sites the error scale as N f far half so there is no advantage to greater depth. This is not a reliable plan for future experiments. The KamLAND Solution • KamLAND is so far underground that they estimate only one background event in their entire dataset. • Neglecting this event does not significantly affect their result. • Finding a site with an acceptable reactor and the ability to get far underground at the optimal baseline would be very hard. Perhaps Dave Reyna has a solution? The KARMEN Solution KARMEN was a surface level neutrino detector that achieved an effective depth of about 3000 mwe by using an active veto shield. Saw background reduction of 97% 3 meter thick steel shield with embedded muon detectors at 2 meters. • Spallation neutrons created outside the veto are stopped • Muons penetrating the veto are detected. The KARMEN Solution (Continued) For a reactor experiment it might look something like this: In my studies I assumed a 95% efficient veto. Then 0.2 bg/ton/day at 300 mwe becomes 0.01 bg/ton/day. The difference between 150 mwe and 300 mwe becomes less important. So we might save money with a shallower site. Measure the Heck Out of It Even with a 95% efficient veto we still need to estimate the surviving background to within about 25% to make this error significantly smaller than the statistical error. We can achieve this precision by using vetoed events to study distributions of various parameters and use them to extrapolate into the signal region for non-veto events. Measure the Heck Out of It (Continued) Various Distributions from CHOOZ Distributions of • Positron energy • Neutron capture energy • Spatial separation • Temporal separation as determined from vetoed events, could be used to estimate correlated backgrounds. These distributions also contain uncorrelated background events. Measure the Heck Out of It (Continued) Matching these vetoed distributions outside the signal range to the data could easily result in a background uncertainty in the signal region of ≥ 25%. interactions Proton recoils Neutron transport simulation Detector resolution not included ? From CHOOZ Can we expect distributions from vetoed events and events that evade the veto to be the same? Detailed simulations will tell. Conclusions By controlling the dominant sources of systematic error and maximizing reactor power a next generation reactor experiment can be sensitive to sin2q13 down to 0.01 at 90% CL in 3 years or less. The dominate sources of systematic 9 GW, 50 tons, error 1200 m, 3 years 15% cross calib. • Relative efficiency & 95% eff. veto • Background Rate can be controlled by designing an experiment with movable detectors and an active external veto shield. Systematics are tied to measurements, they go down as stats go up Optimal Baseline 6 GW and 3 Years 0.05 dm^2=5.0e-3 0.04 dm^2=2.5e-3 Sensitivity 90% CL dm^2=1.0e-3 0.03 0.02 0.01 0 500 1000 1500 2000 Baseline (meters) With m2 = 2.5×10-3 the optimal region is quite wide. In a configuration with tunnel connecting the two detector sites, choose a far baseline that gives you the shortest tunnel.
Pages to are hidden for
"ppt"Please download to view full document