Mu Girl a muon identification package S Tarem

Mu. Girl – a muon identification package S. Tarem 1, 2, Z. Tarem 2, N. Panikashvili 1, 3 , D. Primor 4 1 Technion, 2 CERN, 3 U. of Michigan, 4 Tel Aviv U.

Algorithm flow n Initialize Muon candidate from ID track parameters Extrapolate track to Muon Spectrometer chambers Look for hits in a road around the track extrapolation Make segments from hits Improve extrapolation by using segment information Collect hit & segment information to identify muon ANN to create one figure of merit Select “muon like” candidates n Cut values on lowest p. T & ANN output are job options n n n n

Replacing the Muid. Low. Pt algorithm n A new package is under development ¡ ¡ ¡ ¡ Run independently of MOORE/Muid Improved extrapolation TGC/RPC segments MDT segments using Hough transform – D. Primor CSC segments using Hough transform – D. Primor & N. Amram ANN used to combine program outputs into one optimal discriminator Radically improved SW design n ¡ ¡ Component based, modular, reusable, documented…. Uses new Muon EDM Will replace program Muid. Low. Pt from release 12

Where to look for hits Region. Se n io Reg lector r Extrapolation from ID, chamber selection, extrapolate to chamber ct o e l e S n

Segment making (TGC) n TGC has ¡ 2 layers in the inner station – can measure a point n ¡ ¡ n n Only partial geometric coverage 7/6 layers in the middle station – can measure a segment Nothing in the outer station TGC segments made using hits with 2<10 Interpolation from TGC segments to MDT ¡ ¡ If too few points for segment, the measured position is used with direction taken from extrapolated ID track Improvement in road width

Hits and extrapolations

Roads from ID vs roads from TGC Δη between extrapolated position and hit MDT tube Inner station Middle station Outer station extrapolated from middle n n 6 20 20

Barrel ID vs MDT segment n n Δη between extrapolated position and hit MDT tube Inner station Middle station 20

Segment making (MDT) n n Use Didi’s Hough transform segment maker Interface sends RIO_On. Track in 2 vectors (chambers) ¡ n n Will return lower quality segment if just one multilayer Modified interface to send also road parameters Get back Muon. Segment – quality factor more than just 2

Preliminary performance n We ran on DC 3 validation files of ¡ ¡ n H 4 Some single muons Bs J/Ψ(μ 3μ 6) φ bb μ 6 X A few selections attempted efficiency/fake-rate calculated p. T Fake rate ¡ efficiency η

Number of MDT segments per candidate

Performance for J/Ψ(μ 3μ 6) n n Right – Muid. Low. Pt Bottom – Mu. Girl ¡ ¡ Background bb μ 6 X Low background statistics J/Ψ μ+ μμ-fake

Plans n Complete unfinished features from slide 3 ¡ ¡ ¡ n n RPC segments CSC segments using Hough transform ANN used to combine program outputs into one optimal discriminator Make a thorough test of performance with cavern background etc. Try it on more Physics samples RTT New package structure

Backup

More on fakes n n Do we have fakes from k/pi decays? How can we eliminate them We looked at the change of direction (in eta for now) from the ID track to the TGC segment for muons and fakes Muons fakes

Recall Muid. Low. Pt – our old program n Running after MOORE & Muid, using Muon hits MDT + ++ + RPC/TGC + + +++ + ++ μ ++ ++