Update on primary vertex reconstruction E Crescio A

Update on primary vertex reconstruction E. Crescio, A. Dainese, M. Masera, F. Prino ALICE offline week - 02. 10. 06 Primary Vertex 1

Barrel Tracking & Vertex Recons. Initial vertex: (xvtx, yvtx, zvtx from SPD) 1 st pass high efficiency needed Tracking (TPC ITS TPC …) using SPD vertex info Vertex from tracks with 5 -6 points in ITS (s 1/ d. Nch/dy) 2 nd pass Compute diamond transverse profile, as mean and RMS of vertices in high-multiplicity events of each “LHC fill” Repeat vertex with SPD starting from mean (x, y) important, to have zvtx for all events (even with one track) Repeat tracking Repeat vertex with tracks, using info on diamond profile final (x, y) resolution better than diamond RMS for ALL events ALICE offline week - 02. 10. 06 Primary Vertex 2

Vertex reconstruction using tracks ALICE offline week - 02. 10. 06 Primary Vertex 3

Selection on tracks Idea: use “primary” tracks for vertex reconstruction Select primary tracks on the basis of their d 0 (transverse) Cut |d 0| < n sd 0(pt) = svtx strack(pt) y track vertex x need an estimate of the vertex and its cov. matrix ALICE offline week - 02. 10. 06 parametrized resolution (TO DO: use track cov. matrix) Primary Vertex 4

New procedure Ali. Vertexer. Tracks: : Find. Primary. Vertex(Ali. ESD* esd) Vertex reconstruction in 2 iterations: 1 st iteration: Ø TRACKS SELECTION: |d 0(0, 0)| < 3 cm Ø VERTEX FINDING Ø VERTEX FITTING (xvtx, yvtx, zvtx) 2 nd iteration: Ø TRACKS SELECTION: |d 0(xvtx, yvtx)| < n s(pt) Ø VERTEX FINDING Ø VERTEX FITTING (xvtx, yvtx, zvtx) n = 3 provides optimal resolution and “good” cov. matrix (pulls ~ 1) ALICE offline week - 02. 10. 06 Primary Vertex 5

1 st pass: resolution ALICE offline week - 02. 10. 06 Primary Vertex 6

Computing diamond transverse profile Interaction diamond: stable position and spread in a LHC fill: few hours ~103 s 100 Hz ~ 105 events Strategy: select only events with “many” tracks used for the vertex diamond centre from weighted mean of vertices: (0 1)mm diamond spread from RMS of vertices: estimate MC RMS Ntrks>5 ALICE offline week - 02. 10. 06 Ntrks>10 Primary Vertex Ntrks>20 7

Computing diamond transverse profile Use Ntrks > 20 RMS overestimated due to ~30 mm E-by-E resolution ALICE offline week - 02. 10. 06 Primary Vertex 8

2 nd pass: resolution improvement ALICE offline week - 02. 10. 06 Primary Vertex 9

Efficiency 1 st pass (no info on diamond) 2 nd pass (use info on diamond, RMS = 50 mm) Most failures due to secondary tracks, not well rejected in 1 st pass, that messup initial estimate and cause primary tracks to be rejected Most failures due to tracks with <5 points & secondary tracks that are rejected too few tracks left (see next slide) ALICE offline week - 02. 10. 06 Primary Vertex 10

Efficiency 2 nd pass eff VS good tracks 100% efficiency with >3 tracks with 5 or 6 points ALICE offline week - 02. 10. 06 Primary Vertex 2 nd pass eff VS good “primaries” almost 100% efficiency with >2 tracks with 5 or 6 points and “primary” (not from strangeness decay) 11

Comments on Efficiency (1) In 1 st pass (no info on diamond) there can be failures even in events with many tracks: due to secondary tracks (from s) that are not well rejected in 1 st pass and mess-up the initial vertex estimate In 2 nd pass (using diamond constraint) efficiency is 100% if >5 tracks reco in ITS&TPC residual ineff. due to: tracks with less than 5 points in ITS and secondary tracks (from s) which are rejected, and too few (<2) “good” tracks are left New feature: if Vertexer. Tracks fails in 2 nd pass, x and y are taken from diamond info, z is taken from SPD vertex All events will have a vertex, except those where SPD vertexer fails (no clusters in ITS) ALICE offline week - 02. 10. 06 Primary Vertex 12

Comments on Efficiency (2) Char_t title = Ali. ESD Get. Primary. Vertex() Get. Title() title = “Vertexer. Tracks. No. Constraint”: 1 st pass (not diamond constr. ) title = “Vertexer. Tracks. With. Constraint”: 2 nd pass (with diamond constr. ) Int_t N = Ali. ESD Get. Primary. Vertex() Get. NContributors() N > 2: Vertexer. Tracks found vertex using N tracks N < 0: Vertexer. Tracks failed Ø N = -1: (x=0, y=0, z=0) Ø N = -2: (x=0, y=0, z=z. SPD) Ø N = -3: (x=xmean, y=ymean, z=0) [only after 2 nd pass] Ø N = -4: (x=xmean, y=ymean, z=z. SPD) [only after 2 nd pass] After 2 nd pass, all events will have x and y, and all events with clusters in ITS will have also z Events with N = -4 should not be rejected in any analysis, because they may contain strangeness!!! ALICE offline week - 02. 10. 06 Primary Vertex 13

pull 2 nd pass: Resolutions and Pulls 3 s cut Residuals distributions now gaussian (no tails as in 1 st pass) Errors for x/y slightly overestimated at low mult. (pull<1), to be checked/tuned on PDC events ALICE offline week - 02. 10. 06 Primary Vertex 14

Conclusions: Ali. Vertexer. Tracks More transparent and flexible tracks’ selection procedure implemented Possibility to use average vertex information to constrain vertex fit during second reco pass All events have vertex resolution equal to diamond spread in the worst case Open points: technical procedure to compute average vertex using subsample of high-Ntracks events (possible biases due to pile-up? ) adapt class to be usable also for tracks reconstructed only in the TPC ALICE offline week - 02. 10. 06 Primary Vertex 15