DATA-DRIVEN EFFICIENCY CORRECTION DUNE-PRISM MEETING FEBRUARY 28 TH , 2019 Cristóvão Vilela
INTRODUCTION • Want to make a far detector prediction from near detector data, relying as little as possible on neutrino interaction model. • Near detector is much smaller than far detector and will have limited efficiency for events with high hadronic energy. • Using MC to estimate near detector hadronic containment efficiency introduces model dependence. • Try to calculate efficiency correction for selected events at the near detector using the event geometry. • Ignoring lepton side of the event for now. DUNE-PRISM meeting February 28 2019 2
GEOMETRIC EFFICIENCY CORRECTION • Start out with a selected event at the near detector. • Non-primary-lepton energy deposits in the LAr shown. DUNE-PRISM meeting February 28 2019 3
GEOMETRIC EFFICIENCY CORRECTION • Randomly translate event within FV (blue) and rotate wrt average neutrino direction. • In this case, keeping x constant. • Not a good idea if we expect efficiency to change ~rapidly within a DUNE-PRISM OA slice and if statistics are limited. • In future randomize x within OA slices and use all MC. DUNE-PRISM meeting February 28 2019 4
GEOMETRIC EFFICIENCY CORRECTION • Repeat N times. • N = 1000 • A little slow, but doable. DUNE-PRISM meeting February 28 2019 5
GEOMETRIC EFFICIENCY CORRECTION • Efficiency for the event is: 𝑢ℎ𝑠𝑝𝑥𝑡 𝑂 𝑞𝑏𝑡𝑡 𝑑𝑝𝑜𝑢𝑏𝑗𝑜𝑛𝑓𝑜𝑢 𝑑𝑣𝑢 𝑂 𝑢ℎ𝑠𝑝𝑥𝑡 DUNE-PRISM meeting February 28 2019 6
GEOMETRIC EFFICIENCY CORRECTION • Test method on on-axis near detector events. • Works reasonably well. • Differences at high energy probably from events that are never seen at the near detector. • For DUNE-PRISM prediction will use far detector MC events to make up for those. DUNE-PRISM meeting February 28 2019 7
GEOMETRIC EFFICIENCY CORRECTION • Efficiency correction seems a little worse at edges of detector. • Need to verify with more MC. • Might need to tighten FV. • Or improve method. DUNE-PRISM meeting February 28 2019 8
GEOMETRIC EFFICIENCY CORRECTION • Efficiency correction seems a little worse at edges of detector. • Need to verify with 𝑢𝑠𝑣𝑓 < 2 𝐻𝑓𝑊 𝐹 𝜉 more MC. • Might need to tighten FV. • Or improve method. DUNE-PRISM meeting February 28 2019 9
WHAT NEXT • To figure out which far detector MC events need to be added to the data-driven prediction, want to use the same method on far detector MC events: • Place far detector event randomly in imaginary near detector FV. • Throw translations and rotations to measure “near detector geometric efficiency” for each far detector event. • Then, when building the prediction loop over, near detector data events and far detector MC events: • If near detector geometric efficiency > threshold, use near detector data event • else use far detector MC event. • Need to flesh out details, but I think basic idea should work. • Found way of accessing true energy deposits in the far detector MC files. • Currently writing an art module to calculate “ near detector geometric efficiency” for far detector events. DUNE-PRISM meeting February 28 2019 10
Recommend
More recommend
