nm Samples and Sensitivities Kirk Bays Caltech Oct
nm Samples and Sensitivities Kirk Bays (Caltech) Oct 19 2012 NOn. A Collaboration Meeting
Procedure • • S 12. 06. 17 FD MC Genie files (unswapped) 720, 000 events (2 E 23 POT) No rock events Oscillation assumptions: – 810 km – Dm 223 = 2. 35 E-3 – sin 2(2 q 13) = 0. 1 – sin 2(2 q 23) = 1 • Make DSTs (CAFs later) with many variables • Tune cuts, develop samples • Use Chris’s contour making framework 2
Events types • “Preselection”: Require > 0 visible E, oscillate – 720 k 400 k events equivalent after osc – 41% NC/7% QE/52% non. QE CC • Containment cut: Every hit must be within the FV (within 50 cm from detector edge) – Contained: 191, 000 events • 57% NC/ 7% QE /36% non. QE CC – Uncontained: 209, 000 events • 26% NC/ 7% QE /67% non. QE CC
Handles to separate NC from CC • Likelihood of most muon like track (Nick’s PID) • Track length of that best track • These handles sufficient for decent separation muon pion proton gamma electron non QE CC QE NC
Sample 1: Contained QE • subdivide QE: – events with one (Kalman) track (51% contained QE) – events with two (Kalman) tracks: • where the 2 nd track is a 2 D track (22% of contained QE) • where the 2 nd track is a 3 D track (7% of contained QE) non. QE CC QE NC # Kalman Tracks
contained events, before any cuts non. QE CC QE NC true calo Ge. V 1 track events after NC separation: best PID > 0. 4 && best. Track. Length>220 cm true calo
Separate QE from other CC non. QE CC QE QE used (pass cut) NC Final QE Sample (0 -5 Ge. V cal. E): 90% QE (~43 events 18 e 20) 8% non. QE CC 2% NC Efficiency of QE: ~43% of 1 -track QE events ~23% of contained QE events ~12 % of all QE events
true E – after NC cuts, 1 track true E – all contained non. QE CC QE true. E – QE sample Ge. V NC Visible E – QE sample
QE energy estimation • Compare 3 methods: – Total visible E (calorimetric) – Susan’s energy estimator – Using QE formula with reconstructed muon angle use this! Susan’s E QE formula E res (reco-true)/true QE Formula
QE sample contours true sin 2(2 q 23)=0. 99 90% C. L. QE final sample 70% E res +30% efficiency Both improvements
Sample 2: Contained non. QE CC Final sample (0 -5 Ge. V Cal. E): Kill NC: PID of best track > 0. 5 best. Track. Length > 250 76% non. QE CC 19% QE 5% NC Exclude events in QE sample ~450 CC events (18 e 20) CC eff: 65% final true E non. QE CC QE NC Ge. V final cal E
CC sample contours Using Susan’s energy estimator (~12% E-res) 90% C. L. CC final sample 70% E res +30% efficiency Both improvements
Uncontained Samples • Split into 2 samples: – UC 1: the muon exits detector (2/3) • 32% NC/ 20% QE/ 48% non. QE CC (0 -5 Ge. V true E) – UC 2: the muon is fully contained (1/3) • 66% NC/ 6% QE/ 28% non. QE CC (0 -5 Ge. V true E) UC 2 true E UC 1 true E Ge. V non. QE CC QE NC Ge. V
Uncontained: sample 1 (m exits) E resolution: non. QE CC QE NC • E res very poor (RMS = 29% all events using cal E) • Separate NC: – best Track Length > 220 cm – PID > 0. 44 • • After NC cuts E res 26% (RMS) Can improve further by restricting vertex to not be near edge Can get E res to 22%, but loss of statistics not worth it Need E estimator for uncontained events; for now just use vis E/ Susan’s
Uncontained: sample 1 (m exits) Final sample: non. QE CC: 73% QE: 25% NC: 2% true E (final) ~77 events (18 e 20) (0 -5 Ge. V) 2/3 of non QE CC events not contained; most are in this uncontained sample. But, most are > 5 Ge. V and contribute no oscillation information cal E (final)
Uncontained: sample 1 contours 90% C. L. UC 1 final sample 70% E res +30% efficiency Both improvements
Uncontained: sample 2 (m FC) E resolution: non. QE CC QE NC non NC events better E res than UC S 1 NC events very poor resolution Use standard NC cuts E Res before NC cut: 35%, after: 19% Final sample: 79% non. QE CC / 17% QE / 4% NC ~86 events (18 e 20)
Uncontained: sample 2 contour 90% C. L. UC 1 final sample 70% E res +30% efficiency Both improvements
Contours 18 e 20 POT event counts 0 -5 Ge. V Using 0 -10 Ge. V makes no difference in contours 90% C. L. CC: ~450 events ~12% E res QE ~43 events ~9% E res Uncon 1 ~73 events ~26% E res Uncon 2 ~86 events ~19% E res Combined contour
CAF Files • Chris’s contour frameworks with CAFs • Some variables have problems / need changes / additions • Mostly I can reproduce DST analysis 90% C. L. PID is done in a vector (doesn’t work with TTree Draws used by code) PID has non-vector value for longest track, NOT for best track I can mostly do the DST analysis on CAFs except everything is longest track, not best track I would also request some minor changes regarding containment
Conclusions • It is possible to do a full analysis based only on reconstructed variables using DSTs or CAFs • I have an example analysis ready now • Nick’s muon PID works well • MDC CAF files usable now (contours within minutes) • QE events are hard to isolate; for 1 track events doable, for more very hard; better isolation the priority for this sample • QE formula works well as an E estimator for these • Other CC events easier to work with (NC can be eliminated effectively), and currently provide best sensitivity by far • CC sensitivity equally improved by better E res, efficiency • Uncontained events mostly high energy (not useful); some are useful (not very yet), mostly needs better E estimation
Future • Need to add rock events to simulation • Try different ways of separating QE/non. QE CC; optimize contained samples • Need better energy estimators for uncontained events (more important with rock) • Continue to optimize PID • Continue to work on better E estimators for contained CC events • Optimize containment cut • Look at contours for various situations (half detector, different POT, etc)
Backup
Calorimetric E of 1 track contained events after NC cuts Ge. V
90% C. L. combined contours, where all samples have: standard final sample 70% E res +30% efficiency Both improvements
PID plots
PID plots pid vs particle type best. Track. Length>600 cm (almost all muons) muon pion proton gamma pid vs particle type best. Track. Length<600 cm (almost all non-muons here) muon pion proton gamma
Event Displays
NC in CC final sample 1 n. Tracks=2, best. Track. PID = 0. 91, best. Track. Length = 472 cm neutrino. E = 2 Ge. V , total vis E = 1. 9 Ge. V best. Track = pion
NC in CC final sample 2 n. Tracks=9, best. Track. PID = 0. 69, best. Track. Length = 269 cm neutrino. E = 5. 2 Ge. V , total vis E = 4. 2 Ge. V, best. Track = muon
NC in CC final sample 5 n. Tracks=1, best. Track. PID = 0. 85, best. Track. Length = 263 cm neutrino. E = 2 Ge. V , total vis E = 0. 9 Ge. V, best. Track = pion
NC in CC final sample 4 n. Tracks=17, best. Track. PID = 0. 74, best. Track. Length = 308 cm neutrino. E = 16. 6 Ge. V , visible E = 14. 4 Ge. V, best. Track = gamma
NC in CC final sample 5 n. Tracks=10, best. Track. PID = 0. 83, best. Track. Length = 503 cm neutrino. E = 18. 4 Ge. V , visible E = 9 Ge. V, best. Track = proton
true QE contained, fail CC separation cut 2 total. Ge. V-track. Ge. V = 0. 043 (~86 Me. V), cut at 0. 035
true QE contained, fail NC separation cut 2 total. Ge. V-track. Ge. V = 0. 1 (~200 Me. V), cut at 0. 035
Real CC in final sample 1 n. Tracks=13, best. Track. PID = 1, best. Track. Length = 788 cm neutrino. E = 9 Ge. V , visible E = 9 Ge. V, total vis E = 9 Ge. V
Real CC in final sample 2 n. Tracks=3, best. Track. PID = 0. 96, best. Track. Length = 612 cm neutrino. E = 2. 2 Ge. V , visible E = 3. 1 Ge. V
Real CC in final sample 3 n. Tracks=5, best. Track. PID = 0. 75, best. Track. Length = 2094 cm neutrino. E = 6 Ge. V , visible E = 6. 7 Ge. V
Real CC in final sample 4 n. Tracks=4, best. Track. PID = 0. 89, best. Track. Length = 424 cm neutrino. E = 2. 3 Ge. V , visible E = 2 Ge. V
Real CC in final sample 5 n. Tracks=6, best. Track. PID = 0. 85, best. Track. Length = 283 cm neutrino. E = 2 Ge. V , visible E = 2 Ge. V
UC final sample 1 event 1 n. Tracks=4, best. Track. PID = 0. 98, best. Track. Length = 545 cm neutrino. E = 2. 4 Ge. V , visible E = 2 Ge. V
UC final sample 1 event 2 n. Tracks=3, best. Track. PID = 0. 98, best. Track. Length = 670 cm neutrino. E = 2. 1 Ge. V , visible E = 2. 2 Ge. V
UC final sample 1 event 3 n. Tracks=2, best. Track. PID = 0. 99, best. Track. Length = 835 cm neutrino. E = 2. 2 Ge. V , visible E = 2. 3 Ge. V
UC final sample 1 event 4 n. Tracks=1, best. Track. PID = 0. 84, best. Track. Length = 511 cm neutrino. E = 2 Ge. V , visible E = 1. 3 Ge. V
UC final sample 2 event 1 n. Tracks=3, best. Track. PID = 0. 94, best. Track. Length = 850 cm neutrino. E = 2. 5 Ge. V , visible E = 3. 1 Ge. V
UC final sample 2 event 2 n. Tracks=3, best. Track. PID = 0. 9, best. Track. Length = 502 cm neutrino. E = 2. 2 Ge. V , visible E = 2 Ge. V
UC final sample 2 event 3 n. Tracks=6, best. Track. PID = 0. 9, best. Track. Length = 386 cm neutrino. E = 2. 1 Ge. V , visible E = 2. 2 Ge. V
UC final sample 2 event 4 n. Tracks=3, best. Track. PID = 0. 53, best. Track. Length = 904 cm neutrino. E = 2. 6 Ge. V , visible E = 3. 1 Ge. V
- Slides: 48