Energy Flow Technique and where I am

1
Energy Flow Technique and where I am

• Lily

Have been looking at the technique developed by
Mark Hodgkinson, Rob Duxfield of Sheffield. Here
is a summary. Last slide is RE hadronic W
reconstruction
2
Current jet calibration schemes

• There are currently 2 calibration schemes use by
  the jet group
• The standard H1-style calibrates cells in jets
  to the jet energy scale
• Local hadron style calibrates cells to the
  hadron scale- no correction in place yet.
• In rel 12, local hadron weights are applied for
  eflow. This results in an improvement in
  linearity and a comparable sigma compared with
  local hadron jets.
• Improvement is expected to be less in r13 (better
  local hadron jets)
• H1 style calibration is currently better than
  local hadron

Running eflow current situation

• In rel 12
• Need to run eflow on ESDs- cells in hadronic topo
  clusters
• Standard Jet finder is run on eflow objects
  instead of caloclusters or towers
• Can run on ESD in RecExCommon with
  eflowJobOptions to create AOD containing eflow
  jet collections, I.e
• Eflow4ConeParticleJets - Cone
  0.4 with eflow objects as input
• In rel 13, should be able to just run
  eflowJoboptions on AOD as part of analysis job

3
What is it good for?
At low Pt (lt140GeV) the Inner Detector has
significantly better resolution than the
calorimeters, so tracking can be used to improve
jet resolution

Can also use E(calo)/P(track) to check
calorimeter calibration.
How does it work?

• Cluster-level method (simple way)
• Match tracks to clusters
• For lowish Pt use energy measurement from track
  to get expected energy deposit in calorimeter
  Eexp and sigma (Eexp)
• Ecalo - Eexp Eunexpected
• Compare Eunexpected with sigma(Eexp) which is
  sigma(Ecalo) and possibly away cluster

ATLAS TDR SINGLE PIONS, ETA0
4
How does it work?

• Cell-level method (complex way)
• Devise an ordering method to remove cells using
  Monte Carlo
• Use this method to remove cells until Ecalo -
  Eexp
• Use cell-based weights to calibrate remaining
  cluster energy

Weighted mean energy density for all cells in
layer
Notes

• Tracks must have gt1 pixel hit
• Separate electrons from pions using the ratio of
  TRT hits (number of high threshold hits)/(number
  of total hits). Electrons have mean ratio 2.5,
  pions 0.7

5
More detail- Cell based method

• Cells are classified by (a) calorimeter region
  and (b) distance from track axis
• Compare energy distribution relative to track
  axis with that of Monte Carlo (single pions)
• Extrapolate track axis to form shower axis in
  calorimeter
• Look at a very small region around track axis
  (comparable with error on track extrapolation)
• Weight cells around track axis according to their
  distance from the axis
• Descriptive plots courtesy of Rob Duxfield on
  following slides

6
More detail- Cell based method
The following plots show a cluster formed by a
neutron and pion overlapping, and the pion has
been identified as having first interaction in
the 2nd layer of the LAr calorimeter
Plots are Energy density in eta-phi space
7
More detail- Cell based method
The energy of the cluster is subtracted,
beginning with the EM core and continuing ordered
in energy density, until the subtracted energy
expected energy from tracking.
Plots are Energy density in eta-phi space
8
More detail- Cell based method
The remaining cluster, with its heart removed, is
discarded. .
SUM-UP Calo Energy is replaced with Track Energy
9
Does it work?
Sigma0.1775/-0.0072
10
Does it work?
Sigma0.1775/-0.0072 Sigma0.1625/-0.0062
11
Does it work?
Sigma 0.1775/-0.0072 Sigma
0.1625/-0.0062 Sigma 0.1382/-0.0051
YES
this is all for rel 11 with local hadron cell
weights
12
Latest results
Barcelona workshop J1 sample okay, but J2 sample
had worse sigma with eflow jets than with H1
topocluster jets..

• Problem found to be due to
• Double counting effect-gt too much energy in
  reconstructed jets.
• Introduced new parameter
• If Ecluster lt(Eexpected-(k2 sigmaexpected)) the
  use cluster energy, NOT track energy
• E/P must used cluster which is matched contains
  95 of Energy in 0.4 cone around track

13
More Information

• All the information in these slides was gathered
  from talks given by Mark Hodgkinson and Rob
  Duxfield from Sheffield, and from very helpful
  emails from Mark Hodgkinson.
• The talks are attached to the UCLttH wiki page
• Cell based method RD_eflow_dec05
• Latest Results MH_14-03-07
• Everything else go to Indico jetEtMiss page and
  search for Mark Hodgkinson

Jobs for Lily

• Practise eflow using any old ESD sample at UCL
• Make 12.0.6 HG5 0570,0571 ESDs using panda as
  they dont currently exist. (As far as I can
  tell)
• Orhow long until we get rel 13 standard
  production datsets?

14
More Stuff Im doing now

• Our reconstruction of the hadronic W
• Currently wjjTool defines correctly reconstructed
  W if
• Two combined light jets have mass within 30GeV of
  W mass
• Both these jets have a W as their parent
• My problems and worries
• Found that matching these reco correct Ws to
  gen Ws (as I did for plots shown last week) was
  producing a lot of events where the reco W
  clearly does not correspond to the gen W

Have put in WdR match, indexing and playing with
GenParticle iterator . Can we check for certain
that the two light jets come from the same W ?