Coherent pion analysis with garsoft: status report December 10, 2019 Leo Bellantoni MPD meeting
From DESY workshop • 135 CC coh π + in 11 shifts of “data” • It’s a pretty plot • Reverse the |t| cut & get 1 π + ~60% pure for free • Reverse dE/dx cut & get CCQE ~60% pure for free • Shows we can kill converted pairs & ECAL background • Shown to LBNC last week But! No overlaid inspill activity 2
Adding overlays GarSoft does not yet handle 10 μ s spill well – MC vertex can • be up to 10 μ s × 3cm/ μ s = 30cm closer to the endcap than the reconstructed vertex I can match in the transverse plane fairly well • Now must loop over reco vertex & quiz matching MC • vertex; earlier I started at the MC vertex and found matching reco vertex Therefore, have to change the definitions of my • backgrounds What about tracks that cross the membrane & are late? • Then there are disconnected tracks and they might form vertices Tom’s been stitching these, and I’ll talk about them later • 3
Adding overlays Back-to-back verts killable with cos( Θ OPEN ) > − 0.90 • Narrow angle cut more critical as there is signal there • I did a crude optimization on ~40% of the data cos( Θ OPEN ) < +0.98 • 4
“Mystery” background ~56 days of “data” (with downtime) • What’s that white “Mystery” background? It is • disconcertingly similar to the signal It was in the Hamburg analysis, just not visible • 5
“Mystery” background 1 st look at events • shows that many of these are reco verts with no matching MC vertex Only track quality cut • available is(basically) N º TPCClusters – no χ 2 available yet Cut at ≥ 55 Clusters • 6
“Mystery” background Helps but does not cure • A slightly prettier version of this plot might be • possible – this would be our “money” plot 7
“Mystery” Background • Total 1113 events in the “Mystery” background • 209 actually can be matched to other kinds of MC vertices than {RES, DIS, CCQE}: p – π − ; I’ve seen Λ s on smaller samples 106 p bar – π + 2 π + – π − , presumably from K S 74 e + – e − . Remember that this is after the 26 P TRACK > 150 MeV cut π + with some other thing 1 • In addition to good vertices from bad tracks we can also have bad vertices from good tracks. • Only vertex quality cut available is distance between trackends in the vertex – no χ 2 available yet 8
“Mystery” Background • Red line is reco verts that match to MC vertices, black dots are ”Mystery” backgrounds (including those 209 that did match) • Pretty clear bias towards bad reco verts in the unmatched reco verts 9
A Bump in the Middle After the opening angle cut the membrane bump is still there, but smaller. Left, the before/after ratio when I do not require that there be an MC vertex matched back to the reco vertex; right, when I do. 10
A Bump in the Middle After the opening angle cut the membrane bump is still there, but smaller. Left, the before/after ratio when I do not require that there be an MC vertex matched back to the reco vertex; right, when I do. We thought the shoulders here were delayed particles produced by GEANT. So why are they cos(Θ OPEN ) ≅ ± 1? ∃ some background of bogus reco vertices with cos(Θ OPEN ) far from ± 1 and it is uniform in x , These bumps have σ = 23.5 cm. Cut off distance between trackends to even be vertexed in the 1 st place is 12 cm, half that. 11
A Bump in the Middle After the opening angle cut the membrane bump is still there, but smaller. Left, the before/after ratio when I do not require that there be an MC vertex matched back to the reco vertex; right, when I do. It makes sense that “real” MC matchable vertices generally pass the -0.90 < cos(Θ OPEN ) < +0.98 cut But why is there a bump still in the middle? There is supposed to be only 50 μm of material there, right? And why the dip before the bump? (These plots before the Θ FWRD , dE/dx, Nº TPC Cluster cuts) 12
Conclusions / Discussion • I believe we know what these events are caused by • If it were a real dataset, it wouldn’t be ready for PRD • It will be hard to advance this analysis further without sinking a lot of time into GArSoft, which would be a good thing to do actually • How to wrap this up into something for the CDR & call it done? How long a section in the CDR? • Could show the with-overlays plot or show the Hamburg era • plot and then talk about how overlays impact the analysis 13
Backup Hamburg analysis (no overlays) with new background • definitions 14
Recommend
More recommend