Tevatron ttbar cross section combination

Theoretical ttbar cross section

For the ttbar cross sections CDF and D0 agreed to use Moch and Uwer, http://arxiv.org/abs/0804.1476
This is an approximate NNLO with CTEQ6.6, which for a mass of Mt=172.4+/-1.2 GeV the cross section is sigma_ttbar = 7.48 +0.56-0.72 pb (+7.5%-9.7%)
The theory error (scale+PDF) is calculated using Eq. (4) in the paper:
sigma(2*mt) - Delta PDF(2*mt) <= sigma <= sigma(mt/2) + Delta PDF(mt/2)
The uncertainty due to the top mass error (+/-1.2 GeV) is added in quadrature to the theory error.


Note 1: We will be combining the top cross sections at 172.5 GeV. We plan to also have a mass dependence plot.
Note 2: Tevatron Higgs combination uses Moch and Uwer, http://arxiv.org/abs/0807.2794 This is a NNLO calculation with MRST PDF's which for a mass of Mt=172.4 GeV the cross section is 7.884 pb.
Uncertainties are collected together for 10% theory error on it (bigger than Moch and Uwer's quoted uncertainty)

Table of Systematics

This is the table to be used for the CDF-D0 combination of ttbar cross sections. Each analysis in either experiment has to provide a single uncertainty for each of the categories, and so analyses will report something like this:
Sigma=6.5 +\- 0.7 (stat) +\- 0.4 (det modeling) +\- 0.5 (dJES) +\- ... and so on.

Added a
- Method uncertainty.

Glossary of terms:
  -not used = category not used
  -included = will be set to zero because it is already included in another category.

Table used for CDF-D0 ttbar cross section combination. Click on the column headers to sort the table.
Source of
systematic		 corr. between
D0 channels 		corr. between
CDF channels		 corr. between
CDF & D0 		Tevatron
categories
Data quality y not used n det modeling
Deltaz y y n det modeling
Prim. vertex y y n det modeling
electron ID y y n det modeling
muon ID y y n det modeling
muon track/track ID y y n det modeling
muon isolation y included n det modeling
lepton Pt y y n det modeling
jet ID y not used n det modeling
JES y y n dJES
JSSR y y n dJES
JER y included n det modeling
trigger y y n det modeling
PDF y y y signal modeling
W+jets HF SF y/channel y/channel y/channel or n background from MC (bkg)
or det modeling (check both correlation and uncorrelated)
Z+jets HF SF y/channel y/channel y/channel or n background from MC (bkg)
or det modeling (check both correlation and uncorrelated)
b-fragm y included y bJES
semilept Br y included y bJES
b-overlaid-JES y y n aJES
Z pt RW y/channel y/channel y/channel background from MC
b TRF y y n btagging
c TRF y y n btagging
light TRF y y n btagging
Taggability y not used n btagging
Tag. flav. dep. y not used n btagging
lumi RW y y n MHI
Part of lumi y y y lumi from xsec
Rest of lumi y y n lumi from det
Z XS theory not used y y Z cross section theory
MET model y included n det modeling
bkg xsection y y y signal modeling
Branching ratio y y y signal modeling
tau corr n not used n taus
tau id n not used n taus
W+jets norm n not used n taus
tau energy sc. n not used n taus
stat n n n stat
stat L-T y/channel not used n bkg from data
eps_qcd/sig mu+jets y/channel y/channel n bkg from data
eps_qcd/sig e+jets y/channel y/channel n bkg from data
Fake EM stat y/channel y/channel n bkg from data
Fake MU stat y/channel y/channel n bkg from data
Fake EMU/MU fit y/channel not used n bkg from data
stat SS y/channel included n bkg from data
Higher OE y not used y signal modeling
ISR/FSR y y y signal modeling
had and UE / MC gen. y y y signal modeling
CR y y y CR
Q^2 y/channel y/channel y/channel background from MC
Method n n n Methodology
So far 14 different categories and counting...
Added :
-lepton Pt in "det modeling".
-Z XS theory in "Z cross section theory".

Explanation of systematic uncertainties

Data quality

DO: Efficiency of the calorimenter event data quality flags (noise events that should be removed from the analysis but that weren't from the luminosity calculation).
CDF: does not have this.

Deltaz

D0: MC/Data efficiency of the cut between the primary vertex z and the lepton tracks z. Actually a MC/Data scale factor which has an uncertainty which we propagate.
CDF: Uses 5cm. Under investigation.

Prim. vertex

D0: Prim. vertex: MC/Data efficiency of the primary vertex selection cuts. Actually a MC/Data scale factor which has an uncertainty which we propagate.
CDF: About 0.96 +/- 2 which we call systematic error "Z0".

Electron ID

CDF: Obtained from Z->ee data sample acquired with a single high-pt lepton trigger, eff= ntight_2leg/n_2leg (all lepton pairs considered)
Scale Factor obtained comparing to Pythia. Does not include tracking efficiency, we compute that separately.

Muon ID

CDF: Obtained from Z->mumu data sample acquired with a single high-pt lepton trigger, eff= ntight_2leg/n_2leg (all lepton pairs considered)
Scale Factor obtained comparing to Pythia. Does not includes tracking efficiency, we compute that separately.

Muon track / Track ID

D0: Muon track: errors on the MC/Data scale factor for the track selection. Explicitly quoted for the muon since the "muon id" systematics is only including the systematics related to the muon system.
CDF: Track ID: errors on the MC/Data scale factor for the track selection, both for muons and electrons.

Muon isolation

CDF: Both ele and muo isolation requirements are included in the lepton ID and reconstruction factor efficiency.

Lepton Pt

CDF: Error due to a 1% variation of lepton Pt.

Jet ID

D0: MC/Data jet identification selection efficiency.
We remove some fraction of jets from MC to match jet ID efficiency in data. This fraction has an uncertainty. So we vary the removal rate up and down within the uncertainty.
CDF: We do not apply this. Our cuts are hard, so it is hard to think of an effect which would make 20GeV jets easier to be reconstructed in the Monte Carlo then in data. 20GeV jet means tens of illuminated towers in the calorimeter, due to tens of particles traversing it. Calorimeter threshold is 1GeV for seed, 0.1GeV for inclusion in JetClu, no resolution effect can contribute too here.

JES

CDF: error in the result while varying JES up/down by one sigma.
D0: error in the result while varying JES up/down by one sigma.

JSSR

D0: Jet Shifting, Smearing and Removal.
quark/gluon JES MC/Data error (same as for the mass)
CDF: not explicitly included in the cross section measurements.

JER

DO: smear jets in MC.
CDF: We do not apply it as we think it is included in the Pythia-Herwig MC generator systematic. From gamma+jets Pythia undershoots the JER by a few %, while Herwig gets it right. A mass systematic from this would be below 10 MeV, but we think we cover it with the MC generator systematic.

Trigger

CDF & D0: Trigger efficiency calculated from Drell-Yan process.

PDF

D0: 20 eigenvectors of CTEQ6M added in quadrature.
CDF: Look at 20 eigenvectors 1<=n<=20; compare the difference between CTEQM1n and CTEQ6M, and all up in quadrature.
Look at the difference between MRST72 and CTEQ5L PDF sets and add in quadrature. Look at the difference between MRST72 and MRST75 and add this in quadrature.

W+jets HF SF

Not clear how much it comes from MC/theory and how much it is detector related.
Decision to try both the CDF-D0 correlation and no correlation

Z+jets HF SF

Not clear how much it comes from MC/theory and how much it is detector related.
Decision to try both the CDF-D0 correlation and no correlation

b-fragm

CDF: Not a separate category, because this enters our b-tagging scale factor and its uncertainty.

Semilept Br

CDF: Not a separate category, because this enters our b-tagging scale factor and its uncertainty.

b-overlaid-JES

CDF: +\- 1 % change in JES from hepg b quarks.

Lumi RW

D0: reweighting of the MC luminosity profile that comes from the minbias data events overlaid to the MC to match the luminosity profile in data
CDF: reweight of mc to correct for higher instantaneous lum. We don't add it, but check that the effect is very small.

Z pt RW

D0: Systematic due to reweight of the pt of the Z. Done in each jet-bing separately.
CDF: Possibly strong correlation with Njets scale factor to match Drell-Yan MC to data in dilepton channel.

b TRF

D0: Tag rate function (TRF): uncertainty of the tagging probability for b-jets.
CDF: similar to our scale factor for b's. We called it "b-tagging" scale factor.it 95% +\- 5%

c TRF

D0: Tag rate function (TRF): uncertainty of the tagging probability for c-jets.
CDF: similar to our scale factor for c's. We called it "c-tagging" scale factor it 95% +\- 10%

Light TRF

D0: Tag rate function (TRF): uncertainty of the tagging probability for light jets.
CDF: We use the mistag matrix with its uncertainties, we call this "mistag".

Taggability

D0: Taggability in data: probability of a jet to be taggable (ie. ~ matched to tracks)
CDF: Evaluates the b-tag scale factor per jet (not per jet-taggable), so we don't need this systematic as it is included in the b-tagging scale factor.

Tag. flav. dep.

D0: Flavor-dependence of taggability: different parametrization of the probability of a b-jet or c-jet to be taggable.
CDF: Not used, we don't split at taggable.

Part of lumi

Systematic from luminosity coming from inelastic cross section.
Typically 4% for the ppbar inelastic cross section.
CDF:
  A- for analyses not using the ratio to Z cross section use 4%.
  B- for analyses using the ratio to Z cross section put here 0.

Rest of lumi

Systematic from luminosity coming from detector effects.
CDF:
  A- for analyses not using the ratio to Z cross section use 4.2%.
  B- for analyses using the ratio to Z cross section put here 0.

Z cross section theory

CDF: Systematic from error on the Z cross section theory, only for analysis normalizing to the Z cross section.
D0 : Does not normalize to Z cross section.

MET model

D0: MC/Data difference in the MET distribution (measured in the low MET region, propagate in the high signal region). Elizaveta: This is for fake MET as far as I remember, in the dilepton analysis. So you can think about it as an uncertainty on the fake rate, I believe.
CDF: This uncertainty is already included in our QCD part of the "background from data", see eps_qcd sytematics.

Bkg xsection

D0: We can discuss it, it should either be in signal modeling or bkg from MC because it has to be correlated between D0 and CDF.
We have put it into signal modeling since we would like it to be correlated between ljets and dilepton even if this is not 100% correct as some of the bkg are only in one of the channel but this is conservative.

Branching ratio

D0: Propagation of the W->lepton BR uncertainty assuming lepton universality. Elizaveta: We always require the process of interest on the truth level. In this case you do not depend on the possible different branching fractions built in into different generators. Instead we take BR uncertainty from PDG. The one you consider can be smaller or larger than PDG depending on what generators use. So it is possible that you underestimate this uncertainty.
CDF: We have it included in MC generator syst. The difference between Pythia and Herwig BR's are larger than the pdg errors.

Tau corr

CDF: taus are not used.

Tau id

CDF: taus are not used.

W+jets norm

CDF: taus are not used.

Tau energy sc.

CDF: taus are not used.

Stat

CDF & D0: statistical uncertainty in the measurement

Stat L-T

DO: Statistics of the Loose-Tight sample
CDF: not used

Eps_qcd/sig mu+jets

D0: From Matrix element method, systematic derived from isolation efficiency for fake events (muons).
CDF: uncertainty in the fraction of QCD events in the signal region derived from data (electron and muons alltogether)

Eps_qcd/sig e+jets

D0: From Matrix element method, systematic derived from isolation efficiency for fake events (electrons).
CDF: uncertainty in the fraction of QCD events in the signal region derived from data (electron and muons alltogether)

Fake EM stat

DO : Dilepton channel fake background (electrons).
CDF: Dilepton channel fake background (electrons and muons alltogether).

Fake MU stat

DO: Dilepton channel fake background (muons).
CDF: Dilepton channel fake background (electrons and muons alltogether).

Fake EMU/MU fit

DO:
CDF: not used

Stat SS

D0: Statistics of the same sign events for dilepton channels.
CDF: Not used, this number is included in the Fake Em/MU stat.

Higher OE

DO:Comparison of signal MC@NLO to default sample to understand higher order loops.
MC@NLO+Herwig is compared to Alpgen+Herwig (as opposed to +Pythia)
CDF: We don't use it.

ISR/FSR

D0: We plan to use the same method as for the mass (ie. CTEQ5 with your up and down ISR/FSR variations).
For the moment, it was evaluated with the R method (as it was for the mass before).
CDF: Evaluate shift with ISR/FSR more and less samples as for the mass.

Had and UE / MC generator

D0: Had and UE.
CDF: MC generator: Pythia - Herwig difference in the cross section result. Regardless of whether Herwig's BRs were reweigthed or not as they enter in the signal modeling anyway.

CR

CDF & D0: difference between pythia tune A and pythia tune ACR

Q^2

CDF: We include it.
D0: not included for the moment.

Method

CDF: Uncertainty coming from the methodology used. We considered it uncorrelated with all the other measurement and experiments

Z veto, Conversion, cosmics

Chi2

CDF: For muons, a track chi2 cut efficiency is included in the lepton ID.