LUKE PICKERING, JAKE CALCUTT DUNEPRISM PRODUCTION UPDATE
2017-11-30 2 BEAMLINE AND ND GEOMETRY SETUP 575m Beam sim z 5.8 o 58.1m TGT ND HALL (cm) (cm) ▸ Flux simulation fires hadrons along +z. ▸ Needs to be rotated to expose the near detector to the correct neutrino flux. ▸ Rotate beam coordinates by dip angle (5.8 degrees) and then shift beam sim origin to (0,58.1m,-575m) Luke Pickering
2017-11-30 3 SANITY CHECK OF GENIE EVENTS Beam is definitely heading downward as expected Neutrino momentum y:x Neutrino momentum y:z Spectrum looks sensible enough (N.B. averaged over entire off-axis range) Interaction pos y:x Interaction pos y:z Neutrino energy Flux window was too short for the dip angle so the back of the detector was clipped. Re-running this now, but didn’t feel that it needed to delay as it doesn’t really affect any conclusions Luke Pickering
2017-11-30 4 PRODUCTION STATS: PER FILE GENIE RooTracker edepsim Total argon_box.py Stage g4lbnf GENIE (5000) (10000) (10000) (10000) (10000) (5000x2) Time ~2 hours 80m 180m ~10s 20m ~4.75 hours 94m POT 1E+05 1.04E+16 2.09E+16 — — — — Entries 101000 5000 10000 — — — — Neutrinos 101000 e.g. 370000 e.g. 750000 — — — — Used File Size 80 Mb 9.7 Mb 20 Mb 7 Mb 2.4 Gb 2.5 Gb 7.4 Gb ▸ Using a scaled up example of the argon_box.py results from Jake for comparison. ▸ Might be worth moving to edep-sim. ▸ Current aim is to push 20k events through from the start and get that to Guang ASAP (i.e. Friday Dec 1st). ▸ Can work on intermediate steps while that output is being assessed by Guang. Luke Pickering
2017-11-30 5 ARE OUR PRODUCTION GOALS FEASIBLE ▸ Scaling the numbers on the previous slide up: ▸ This simulates at all off-axis positions simultaneously so ‘collect data’ at NStops times faster. ▸ 4m wide FV for 32m off-axis span we need 8 stops. ▸ Expected yearly POT is 1.4E21 POT ▸ So for 11x 7year POT/8 stops, we need to simulate 13.475E21 POT on this 36m box ~= 650,000 files @ 2.1E16 POT per 10k events: ▸ ~3 Million CPU hours ▸ ~1.6 PB of disk space if we save intermediate stages ▸ ~6.5 Trillion events ▸ I think we need to re-assess the production size, can easily scale predictions to such event rates and force poisson errors on these. ▸ I think 1k files would be a good start — Even at this level cannot feasibly save G4 output @ 2+ TB. (10 mil events for nu-mode) Luke Pickering
2017-11-30 6 SANITY CHECK OF PREDICTED EVENT RATES ▸ Possible that I am screwing something up in the geometry and thus the NEvents <=> POT correspondence is not correct. ▸ Check against event rates that I previously calculated. ▸ Use 5x4x2.5 ‘FV’ as for the ND workshop study ▸ Ask for one cycle’s worth of events at the on axis position ▸ I think this shows good correspondence. O ff set ν µ CCInc NCInc 0 m 5.1 1.5 5 m 3.9 1.2 10 m 1.8 0.53 15 m 0.85 0.24 20 m 0.47 0.13 Running GENIE + Flux driver 25 m 0.28 0.081 30 m 0.18 0.053 ND Workshop Luke Pickering
2017-11-30 7 A NOTE ON NEUTRINO RE-USE ▸ When throwing actual events in the detector, decay parents are re-used a huge number of times: Ray tracing rejection ▸ ~100k decay parents used as input (equiv to 1E6 POT). (flux de-weighting) Interaction rejection (xsec de-weighting) ▸ The problems that Cris saw before were exaggerated because that flux generation tool forces neutrinos through (XOffset,0,0) in detector coordinates. ▸ My tools here: https://github.com/luketpickering/DUNEPrismTools are built on top of those scripts but are hopefully much more configurable and allow you to specify a flux window (in a same way as the GENIE flux driver). ▸ If there is interest in using/building on these rather than just the scripts that Laura sent, I will add Cris as a contributor to the repo. ▸ I propose to re-use each decay parent O(10) times but randomly spraying them over the 36m wide ‘detector’ — this should stop the ‘line’ problems that Cris saw where one high energy neutrino is dragged off axis and leaves a line in the plot. ▸ It does mean that the parents are re-used, I’m not sure we can get around this… ▸ This isn’t implemented in the code, but should be a five minute job. ▸ Ideas, Concerns, Expectations? Luke Pickering
2017-11-30 8 NEXT STEPS ▸ Running a larger nominal beam simulation sample today (1k files). ▸ Will get those pulled to MSU so that Cris can start playing with them. ▸ Build FD Fluxes (and FD/ND flux ratios). ▸ Will run in a few other configurations and determine OAA vs. ENu reweighting histograms (similar to Tejin’s talk at the DUNE ND meeting 2017-11-30): ▸ Horn shift, Current shift, Water layer thickness ▸ Want to also get PPFX working for hadronisation parameter reweights. Luke Pickering
WE APOLOGISE FOR ANY INCONVENIENCE THANK YOU
Recommend
More recommend
