Calorimeter Algorithms

1
Calorimeter Algorithms
Gregorio Bernardi for the CALGO and CAT groups

• A few remarks on the calo. problems
• Overview of the CAT force results
• From CAT to CALGO
• Results of the CALGO workshop
• The road ahead

2
DISCLAIMER
Different more or less important problems affect
the calorimeter However The calorimeter data
are waiting for other subdetector data
improvements in order to go to publication The
Reprocessing is NEEDED mainly to improve the
Tracking The Calorimeter fixing can be done
and redone and redone from the TMBs in a few
days. personal opinion From the analysis point
of view the fake jets problem is, for the moment,
our main problem in the calorimeter.
3
D0 Calorimeter Algorithms Task Force
Members of the Task Force (MAY JULY
2003) Gregorio Bernardi (chair), Jonathan Hays,
Serban Protopopescu, Markus Klute, Jan Stark,
Robert Zitoun, Vishnu Zutshi Emmanuel Busato,
Jean-Roch Vlimant Charge to the Task Force
The task force will optimize in collaboration
with the Physics and ID-groups the calorimeter
algorithms for precision physics. It will start
from the achievements of the Calorimeter Task
Force, continue to improve the algorithms and
propagate them to the physics objects using as
benchmarks the comparison data/simulation of the
W (-gte nu) transverse mass spectrum and Jet
Met resolutions in di-jet events. The task force
is expected to coordinate the efforts in the
different groups related to the calorimeter.
4

• Specifically,
  the task force should
• 1) Make progress on the understanding/correction
  of the calorimeter noise.
• 2) Finalize the development of an optimized
  0-suppression scheme and study
• its consequences for MET, EM, Tau and Jet ID.
• 3) Make progress on the understanding of the EM
  response/resolution, using
• on-line and off-line tools.
• 4) Make progress on the data/mc agreement of the
  calo object
• reconstruction and of the energy flow
  algorithm.
• 5) Use large statistics of W and di-jets events
  to display how the
• progress achieved on points 1-4 propagates in
  the corresponding
• physics distributions.

5
Correcting the DATA

• Many problems are not discovered immediately
  online.
• - BLS electronics problems
• - cable swaps
• - not all hot cells are caught online
• ? Large datasets on tape with some quality
  issues.
• These data sometimes suffer from isolated
  problems. Can fix these
• isolated problems and use the data for
  publication quality analyses.
• But need mechanisms to do this.
• ? cal_corr_dst package
• Takes raw calorimeter data and fixes known
  problems
• ? included in reprocessing
• Can also be included in user analysis jobs
• (because we have the individual cell
  energies in the thumbnail).
• Requires some effort as calorimeter
  reconstruction needs to be rerun.
• CAT MODEL HAS BEEN TAKEN ON BOARD BY COMMON
  SAMPLE GROUP

6
DQ Issues Addressed in Current Version of
cal_corr_dst

• Energy sharing problem
• Draft D0 note
• http//www-clued0.fnal.gov/stark/esp_note.ps
  
• Tower two problem
• Draft D0 note
• http//www-clued0.fnal.gov/stark/tower2/note
  .ps
• BLS cable swap
• next slide
• Short term project
• kill hot cells found by dq_calo

Web page with latest version of cal_corr_dst
http//www-clued0.fnal.gov/stark/cal_corr_on_
tmb.txt
7
Most recent addition to cal_corr_dst
Correction for BLS cable swap The discovery
of the swap was triggered by an analysis
plot ?/? distribution of all EM candidates
pT gt 25 GeV EM fraction gt 0.9 isolation lt
0.15 HMx8 lt 20 pink boxes tower two
problem blue boxes energy sharing problem red
boxes dont know yet Nice example of
constructive feed-back from analysis.
J. Gardner
Donuts ? Cable swaps !
8
Reducing influence of Noise T42T42 reject
ALL isolated cells below 4 sigmas (and ALL
negative energy cells)

• Select high signal cells (4 sigma)
• Keep their significant neighbours (2 sigma)
• ?Thresholds 2? is 2.5? at the moment ? T42.5

Full description is available in DØNotes 4124,
4146
Available in D0Reco in shadow mode Not running
yet We wants to have first a tmbfix
w/o T42

9
T42 on high energy electrons
Estimators for EM candidates in data, when pTgt13
GeV
OK !
10
T42 on Missing and Scalar ET
WZ skim , p13.06 , W? e nu selection (ET gt 20
GeV, MET gt 25 GeV) Compared to Pythia MC
No T42
After T42
11
Jets properties before and after T42
Top groups alljets skim (reco version
p13.06.01) Passes the 4JT10 trigger.
At least 4 jets (JCCB). HT gt 100 GeV
(just plain sum of uncorrected JCCB jets).
? ? 20000 events
Jets are JES corrected. No quality cuts applied
because jet id distributions change
significantly.
t42 no t42

12
New Jet Seeding in p14
Remove CH and MG cell energies from the seed
tower energy before starting preclustering
(i.e., a p13 seed might have an energy lt 500
MeV in p14)
Jets not found with new
seeding
Jets found in
both cases

• about 3-4 less jets
• with new seeding

Jets not found with new seeding - mostly low pT
- in ICR - large number of
proto-jets merging
13
JET ID Plans for the future

• High priority task remains reduction of fake jets
• Are T42, new jet seeding, hardware fixes
  enough ?
• Jet ID cuts ? new certification
• Use L1 confirmation ?
• Tuning of Jet ID criteria for ICD region.
• Tuning of merging algorithm ?
• Consider lowering jet pt threshold
• 6 GeV (to make analysis at 10 - 12 GeV)

14
MET subgroup Goals

• MET Resolution related to Jes, T42
• Treatment of non reconstructed jets
• Overall correction strategy of MET
• ? Decides how to correct EM/Jets etc..
• Understanding of unclustered energy
• - in QCD processes
• - in EW events

15
CALGO Activities Structure
CAT
CALOP - calo hardware operations R. Zitoun
CALGO - calorimeter algorithms
objects U.Bassler/ G. Bernardi
cal-simulation Leslie Groer, Michel Jaffre
em-id Harald Fox, Jan Stark
calib online
?-id Yurii Maravin Drew Alton
data taking
cal dq Slava Shary, Jan Stark
jet-id Slava Kulik Alexander Kupco
hardware
met Patrice Verdier, Sophie Trincaz
cal/icd-softw. Jan Stark/Lee Sawyer
slow control
eflow Anna Goussiou Jon Hays
cps-software A. Magerkurth/D.Alton
trigger L1/L2
jes Ia Iashvili/ Nirmalya Parua
fps-software A. Patwa/A. Turcot
icd
l3cal-software
tau-id Dhiman Chakraborty / Serban Protopopescu
16
Welcome/Goals of the Workshop (15')
U. Bassler/G. BernardiShutdown
Status/Operation Organization R.
ZitounData Quality V. Shary/J.
Stark Overview J. Stark zero-bias
monitoring bad cell correction V. SharyLevel
3 (20') V. BuescherSimulation
Status and Prospects L. GroerTau-id
D. Chakraborty/S. Protopopescu Jet-id
S. Kulik/A. Kupco
jet-id status and plans S. Kulik
jet reconstruction and simulation V.
Zutshi Jet and Met while studying Wbb
production G. BernardiMissing ET (30')
S. Trincaz/P. Verdier MET status (15')
P. Verdier cal_t42 status (15')
J.R. Vlimant
CALORIMETER WORKSHOP contributions
17
Jet Energy Scale I. Iashvili/N.
Parua JES status and plans
I.Iashili Response measurement for b-jets
T.Kurca Showering measurement with MC
method J.Rani Energy Flow (30')
A. Goussiou/J. Hays electron-id/calibration
H. Fox/J. Stark Introduction
H. Fox/J. Stark EM
reconstruction packages S.
Crépé-Renaudin H-matrix, data/MC agreement
M. Jaffré/T. Vu Anh Electron
likelihood (2) J. Kozminski, S.-J.
Park EM scale corrections (10')
S. Kermiche photon-id
D. Alton/Y. Maravin FPS software
J.
Lazaflores/A.Patwa CPS Status
D. Alton
18
Jet Energy Scale
Offset (p13.06 data, p13.08 MC) Respons
e (p13.06 data, p13.08 MC) New ICR
correction (p13.06 data, p13.08 MC) same as
previous correction in data (25), smaller in MC
(11 ? 6.5)
Special Min Bias runs
Pythia UE 0.8 minbias
19
JES Smaller Uncertainty

• Offset
• Luminosity dependence OK
• Response
• Increased statistics 15 ? 55 pb-1
• EM scale included
• Systematics background, topology, vertex
• MC used for extrapolation at high energies
• Total
• in central region 9.5 ? 5.5

?0.0
20
Response b-jets vs q-jets (MPF method)
21
Jet Resolutions

• Resolution at ET50 GeV
• Run I 11.6
• Run II 14 (data/MC)
• More to Understand !
• Z ? bb resolution
• various algorithms (p13.08, uncor.)
• Rcone 0.5 18.1
• Rcone 0.7 17.5
• 0.4 kT 18.3
• 1.0 kT 17.7
• Rcone0.5
• before correction 18.1
• after JES correction 17.0 Use
  Tracksclusters?

22
JCCA cone jets have no energy correction
applied CellNN calibrated using linear fit to
response measured using single charged pion
Monte-Carlo. Potential significant improvement
need to calibrate tracks/objects on data. Here
results z?qq MC with p13
23
Electron identification
calorimeter cells
For low-pT non-isolated electrons road
method. Start from tracks, look at energy
deposits in a narrow road around the track
extrapolated through the calorimeter see if
these deposits are EM-like.
cone, cellNN
clusters
C/FPS clusters
tracks possibly with dE/dx
electron candidates
calibration - geometric corrections - EM
scale corrections - phi cracks
electron candidates for analysis
sophisticated discriminants H-matrix, likelihood
The whole chain needs to be certified. Quite
some Monte Carlo dependence geometric
corrections, training of H-matrix, analysis,
24
Low-pT di-EM triggers in trigger list v12.x
Trigger list v12 contains a new low pT di-EM
trigger. ? useful data samples for EM
validation studies Summary of the selection
criteria used by this trigger L1 CEM(2,3.)
CEM(1,6.) TTK(2,3.) TTK(1,5.) L2 ?
restriction to central region L3 two road
electrons ( gt 3 GeV and gt 5 GeV, one tight one
loose)
7.6 pb-1
Reconstructed using road method.

• Fit result (signal parameters)
• N (one tight) 614 /- 113
• Mass 3.040 /- 0.005
• 0.071 /- 0.005
• a 0.61 /- 0.12
• n 1.7 /- 1.1

25
Data J/?, ? ? e e- using EM clusters
HMx8 lt 50
HMx8 lt 20
Fit results (signal parameters)
N (J/Psi) 152 /- 15 Mass (J/Psi)
3.166 /- 0.040 GeV Resolution (J/Psi) 0.382
/- 0.043 GeV N (Upsilon) 70 /- 18 Mass
(Upsilon) 10.31 /- 0.18 GeV Resolution (Ups)
0.584 /- 0.144 GeV
N (J/Psi) 82 /- 9 Mass (J/Psi) 3.166
/- 0.040 GeV Resolution (J/Psi) 0.305 /-
0.037 GeV N (Upsilon) 55 /- 13 Mass
(Upsilon) 10.12 /- 0.17 GeV Resolution (Ups)
0.689 /- 0.190 GeV
comparison of fitted yields gives information on
HMx cut efficiency.
26
Calibration at low Energy

• To investigate the high ? mass, take the
• HMx8 lt 20 sample and compare the mass
• distributions obtained from clusters and tracks
  for the same events.
• mass obtained from tracks appears to be
• more reasonable, but need to check resolutions
  with ? Monte Carlo.

yield ? Mass (GeV) resolution (GeV)
clusters 55 / 13 10.12 /- 0.17 0.69 /- 0.19
tracks 68 /- 11 9.43 /- 0.11 0.62 /- 0.10
Potential calibration Improvement
27
Electron Resolution

• From Monte-Carlo with corrections
• S (0.199 0.008) N (0.42 0.08) GeV
  C (0.0076 0.0014)
• MC momentum resolution 3 _at_ 50 GeV
• expect Z mass resolution 2Z MC 2.2
• Data 4.0 ? 3.1

28
Various Corrections
MZ/GeV sZ/GeV sZ/MZ()
raw ADC 81.5 3.6 4.4
non linearity 86.7 3.7 4.3
gain correction 86.6 3.6 4.2
crate equalization 86.5 3.6 4.1
geometric correction 90.2 3.6 4.0 0.1

• Corrections in D0reco emcandidate
• Further
• calibration
• corrections
• Other trials

calibration timing 90.2 3.6 4.0
physics timing 90.2 3.6 3.9
calib/phys amplitude 90.0 3.4 3.8
after tuning weights 90.4 3.3 3.7
use 2 calweights 90.4 3.3 3.7
pulser amplitudes 90.2 3.4 3.8
10 h slices 90.4 3.3 3.7
29
Module Boundaries (Phi) ?

• CC has 32 modules with Dj0.2
• loss of clusters (?)
• loss of energy
• Broader effect on mass

Broader effect on mass Origin ?
Statistics ?
MZ/GeV sZ/GeV sZ/MZ
All 90.4 3.3 3.7
djgt0.05 91.3 2.8 3.1
djlt0.05 89.6 3.2 3.6
30
Dead Material Simulation

• Modeling of the phi cracks (electric fields and
  charge collection)
• Modeling of the dead-material
• Modeling of preamp charge collection signal
• Improvement done for solenoid
• Improving the material map for the SMT cables,
  infrastructure, cooling

31
EM ID estimators data vs MC
EM CC Cluster widths in r-phi

• p14.03.00 data and MC
• EM ID v4.1
• 2 good electrons, pT gt 20 GeV, ?_det lt 2.5, not
  within cracks, Mee gt 50 GeV
• At least 1 track match in data
• Many distributions are ok
• pT, isolation, transverse shower shapes in r-phi,
  EMfractions, HMx41 CC
• Discrepancies in a few (MC normally
  narrower)
• HMx8 components
• 2 ( logE and ?z) and 4 (?z)
• Number cells in lower layers CC
• CC profile/shower shape in ? and R
• HMx6 ?

EM CC Cluster widths in z
32
CPS

• Spatial resolution
• Resolution in f of 1.58 mr (MC 1.5 mr)
• Resolution in z of 3.2 mm(MC 2.5 mm)

MC study energy resolution Green shows the
improvement using the CPS.
sE/E
33
Progress on Photon ID

• Currently, a photon is an electron with no
  matched central track, Why is this not good?
• HMatrix does not work well for electrons, not
  easy to make it work for photons
• NEW TOOLS UNDER STUDY
• Tracks
• Matching tracks to CPS-EM objects
• Isolation
• Hit counting
• Use both CFT and SMT
• PS
• Matching to EM/TRK
• Shapes of clusters, p0-g separation?
• HMx
• Check if electron HMx works well enough on
  photons
• Train it on a sample of g (?)

34
TRK-CPS-EM matching

• From a sample of good Z?ee candidates
• Require exactly two good EM objects
• Match them with CPS clusters
• 3d floor method
• Match CPS and EM (3d floor) in (h, f)
  space
• c2 method
• Use all floors
• Fit to a straight line

Wrong cluster
Good cluster
35
Priority Tasks

• Em-energy resolutions (electrons, photons)
• Data/simulation agreement shower shapes, cracks
• Data quality strategy (on-line/off-line)
• Noise understanding / Fake Jets / T42
  integration
• J.E.S. / hadronic energy resolution
• Calorimeter compensation studies (E-Flow)
• ?-id

36
Future dates

• CALGO meetings devoted to specific projects
• October 21st met corrections, DQ October
  28th em-resolution / simulation
• p14 certification of calorimeter software and
  objects November 21st (JES Nov. 28th)
• god-parenting by CAT
• Next calorimeter workshop December 2nd/3rd
  or 3rd/4th

37
Backup slides
38
MET Run Selection
Only good runs

• Using METB (no CH, except in good JCCB jets)
• Pretty stable with time (METxy most sensitive)
• Good run selection based on
• rms(METxy)
• ltSETgt
• 91 good runs (feb-jun 2003)
• could be lower with tighter cuts
• could be higher with software corrections
• MUST be update for p14
• for TMBfixed data

39
TRK-CPS / TRK-CPS-EM matching

• Require exactly two CPS-EM objects
• Propagate track to the middle of CPS
• Calibrate track propagation
• Check TRK-CPS relative alignment
• Select the closest track in (z, f) space
• In the Z?ee sample
• TRK-EM (traditional) matching probability is 84
  2 (AA)
• TRK-CPS-EM matching probability is
• floor method 95.8 1.1
• c2 method 92.8 1.3
• Fake rate estimation is in progress

40
JES b-quarks parton-level

• New semi-leptonic correction for b-jets (p13.08
  MC)
• Added dependence on parent quark (in addition to
  Emu, ptrel, Ebjet
• Parton-level
• Correction
• light q
• b-q

b ? µ? X
b ? c ? µ? X