lost jet update

12
Lost Jet Update John Krane Iowa State University jet/MET Meeting 8/29/02 reminder of problem performance metric parameter choice

Upload: damia

Post on 06-Jan-2016

18 views

Category:

Documents


0 download

DESCRIPTION

Lost Jet Update. John Krane Iowa State University jet/MET Meeting 8/29/02. reminder of problem performance metric parameter choice. CDF Physics groups are not using Run II algorithm. Problem: unclustered event energy. CDF: Matthais Toennesmann DØ: me, Vishnu Z., Bob H. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Lost Jet Update

Lost Jet Update

John Krane

Iowa State University

jet/MET Meeting 8/29/02

• reminder of problem• performance metric• parameter choice

Page 2: Lost Jet Update

John Krane -- Iowa State University 2

Problem: unclustered event energy

CDF: Matthais Toennesmann

DØ: me, Vishnu Z., Bob H.

Cones can iteratate

away from “small”

Energy clusters

The Run I algorithm did this too...

CDF Physics groups are not using Run II algorithm

Page 3: Lost Jet Update

John Krane -- Iowa State University 3

Modified Cone Algorithm

CDF’s “search cones”

Use a small cone to find jets and iterate locations

Expand cone size to full 0.7 and save

Find midpoints found a phi-wrap bug; using pT-weight not 4vec?

Iterate 0.7 size midpoint jets

Page 4: Lost Jet Update

John Krane -- Iowa State University 4

Results on selected sample (45 evts)

Seed tracking

Plot shown in

OK workshop

Symmetric in y-,

so just use R...

Abs y drift

Abs

d

rift Each point was a seed

p11’

I never found a lost jet, but they must exist...

Page 5: Lost Jet Update

John Krane -- Iowa State University 5

Drift distance for 0.7 cones, pT>15 GeV

If a seed is too close (R/2) to existing jet, ignore it

Standard cones can drift very long distances!

Search cone R/2 limits drift to R...

Page 6: Lost Jet Update

John Krane -- Iowa State University 6

R=0.5 Cones

Same comments apply...

Page 7: Lost Jet Update

John Krane -- Iowa State University 7

R=0.3 cones

Again...

Page 8: Lost Jet Update

John Krane -- Iowa State University 8

Normalize drift distances by R

R=0.5 cones,scaled distance

Page 9: Lost Jet Update

John Krane -- Iowa State University 9

x-axes have suppressed zero

Page 10: Lost Jet Update

John Krane -- Iowa State University 10

Conclusions

Cones can drift quite far from the seed, even for reasonably high-pT Jets >15 GeV

This doesn’t mean a jet is “lost” every time this happens (I have yet to find a lost jet)

Search cones can limit drift as much as we like

R/2 works well (almost perfectly) R

Page 11: Lost Jet Update

John Krane -- Iowa State University 11

Suggestions for Future Work

Implement my rcp-driven code (but not necessarily a search cone setting just yet)– Search_Factor=1.0 means no search cone– Search_Factor=0.5 eventually...

Run full Reco tests for CPU time and consistency

Consult CDF and try to converge on a parameter – Informally, Joey Huston thinks R/2 works well– Would like permission to show this talk externally

Page 12: Lost Jet Update

John Krane -- Iowa State University 12

Files altered for Search Cone mod

energycluster/ILConeAlgorithm.hpp(Constructor allows search_cone argument,midpoint calculation fixed for phi-wrap,is_stable() need not iterate,min jet pT reduced during search phase)

calreco/CalClusterReco.cpp(Read and pass Search_Factor from rcp)

calreco/rcp/CalILCone07.rcp

calreco/rcp/CalPreSCILCone07.rcp

calreco/rcp/CalPreSCILCone05.rcp

calreco/rcp/CalPreSCILCone03.rcp