light jet energy scale determination with top events after rome workshop status
DESCRIPTION
Light jet energy scale determination with Top events After Rome Workshop status. D. PALLIN 15/12/05. Rome Workshop. Extract the light jet energy scale E parton -> E jet MC -> E jet recons no hypothesis on calib funtion jets > 40 GeV ; W sample from top sample 85%purity . - PowerPoint PPT PresentationTRANSCRIPT
Light jet energy scale determination with Top
eventsAfter Rome Workshop status
D. PALLIN 15/12/05
D. Pallin TOP meeting DEc 2005 2
Rome Workshop Extract the light jet energy scale E parton -> E jet MC -> E jet recons no hypothesis on calib funtion jets > 40 GeV ; W sample from top sample 85%purity
Calibration function (Ejet) = Eparton / EjetAOD
Cone 04
EPa
rt /
E E
An example: Rome AOD (Ejet) = 1. to 1.1 ‘cone 04’
D. Pallin TOP meeting DEc 2005 3
Rome Workshop Ingredients Constraint 1 Rq: angle well reproduced if jet masses used Constraint 2 : 1 and 2 have the same dependance in function of E ( f(E) ) give the correlation between all W mass reconstructed => Build MW distributions in function of E to keep correlation
each MW distrib gives <MW> (E)
<MW> (E) depends on calib
jeti
parti
iRECW
PDGW
EE
withEEMEEM )(),( 212121
MW
E
D. Pallin TOP meeting DEc 2005 4
Rome Workshop Methods Use MW distributions in function of E Find best 1 and 2 such that <MW> (E) = MW
PDG(E)
=> 1/ 2 FIT <MW> (E) = MWPDG(E)
=> 2/ iterative method without fit
D. Pallin TOP meeting DEc 2005 5
Iterative Method to extract the E scale Constraint R=1
compute R for k bins in E apply k factors on R and recompute R n times =>
jeti
parti
iWPDGW E
EwithMMR 21/
1 2k j j True nk k
n
R
E
AODW Recons.
No comb BKG
D. Pallin TOP meeting DEc 2005 6
Iterative Method to extract the E scale
compute R for k bins in E apply k factors on MW and recompute R n times =>
jeti
parti
iWPDGW E
EwithMMR 21/
1 2k j j True nk k
n
E
EPa
rt /
E
E
EPa
rt /
E
D. Pallin TOP meeting DEc 2005 7
Method1 Results after recalibration
Function found with any
‘a priori’ hypothesis
EPa
rt /
E
E
before
after
AOD4000 W Recons.No comb BKG
Mw
D. Pallin TOP meeting DEc 2005 8
After Rome Workshop
Reduce known systematics
Try to get calibration below 40 GeV
try to explain the connections existing between the know systematics on calibration method and the observed Mtop dependence in function of Pt
D. Pallin TOP meeting DEc 2005 9
Systematic 1 E_jet / E_part % E_parton MC calib=1
D. Pallin TOP meeting DEc 2005 10
Systematic 1 E_jet / E_part % E_jet MC calib=1
!! No more calibrated ???
Bias is within 1% above 40 GeV
But need to be corrected
huge effect below 40 GeV
D. Pallin TOP meeting DEc 2005 11
Systematic 1 Why ?
E_jet
E_parti i+1i-1
j
Ej=ai Epi overestimated
(E/Ep)j = Ej / (ai Epi) underestimated
D. Pallin TOP meeting DEc 2005 12
Systematic 1 How to correct? origine of bias : purely statistical from : E shape+ E
resolution negligeable if resolution(E)<<E important when resolution(E)~E
Need to know shape and resolution Use full Top MC to extract bias Compute statistical function = corr_det
1;)1(
)(
*),0(1
j
i
i
ji
j
jj
binj
binj iiji
EpEpa
EpE
EpE
nGa
D. Pallin TOP meeting DEc 2005 13
Systematic 1 E_part / E_jet % E_jet after bias correction E_jet =E_jet X corr_det
Corr_det Works for any generated resolution and E shape
Shape taken as the observed E shape
D. Pallin TOP meeting DEc 2005 14
Effects on Mw MW
REC-cor = MWREC X sqrt(corr_det (E1) )
BUT additionnal effect (effect 2)
Mw MC calib=1Mw=79.93 ±0.08 GeV ;
470 MeV from the generated W mass
D. Pallin TOP meeting DEc 2005 15
Effects on Mw (effet 2) Mw MC calib=1
)1)(1()cos1(2 2121rrMEEM PDG
jjW
• 1+r1+r2 almost gaussian, centered on 1
• r1r2 not linear, introduce a bias
• effect seen in SM group (Z resonnance) and computed recently by Kramner. Same as our result
• depends on E
• Alreaydy known effect but negligeable (0.5%) for our purpose
MeVM
Bias EW 4808
12
D. Pallin TOP meeting DEc 2005 16
Effects on Mw (effect 1) Mw % E_parton MC calib=1
D. Pallin TOP meeting DEc 2005 17
Effects on Mw (systematic 1) Mw % E MC calib=1
Meth1 on AOD: Calibration obtained is a convolution of calib X corr_det MW%Ejet , Ejet/Ep%Ejet distrib are flat MW%Ep , Ejet/Ep%Ep distrib no more flat
Same effect as on
Ejet/Epart
Same correction apply
MWrec
x corr-det is
flat in function of E
=> Same type of correction apply to correct MW in function of Pt cut
EPa
rt /
E
MW
Ejet
Ejet
D. Pallin TOP meeting DEc 2005 18
Method 1 remove systematics Energy resolution and distribution needed
Good estimate of the shape given by Ejet shape E Resolution could be given by MW resolution
OR/AND from outside ( Z+jets) Suppose resolution and shape known :
Correct bias with fonction corr-det Apply method1 as previously result OK for E>40GeV Correlation found between calib and corr-det below
40GeV.
D. Pallin TOP meeting DEc 2005 19
E resolution f rom MW width
MW /MW do not depend on calib use MW distributions for different Energy bins => Measure MW /MW in function of E Extract E/E from the constraint :
result is enough precise to be used in coor_det
2222
1
1 )()22
()2(
W
W
W
W
MM
EE
EE
MM
D. Pallin TOP meeting DEc 2005 20
Conclusion for calib
to calibrate jets below 40 GeV, knowledge of E and E distributions are needed
Still possible to extract jet energy scale without using MC or calib function hypothesis
OR use MC apply E and calibration on partons find the best E and calibration reproducing the observed
MW %E and E distribution.
D. Pallin TOP meeting DEc 2005 21
Mw and Mtop in function of pt cut
for the same reasons (effets from E and E distributions) Mw or Mtop % Pt cut are flat . No dependence Mw or Mtop % Ptjet cut are no flat. Huge effect on Mtop
Easy to handle for MW. Tricky for Mtop Mtop : use MC to get the Mtop value for a given Ecut
use measurement of E and calibration from MW for light jets find the best E and calibration for bjets reproducing the observed Mtop
%ptcut and ptb distributions. This
D. Pallin TOP meeting DEc 2005 22
Mw and Mtop in function of pt cut
for the same reasons (effets from E and E distributions) Mw or Mtop % Pt cut are flat . No dependence Mw or Mtop % Ptjet cut are no flat. Huge effect on Mtop
Easy to handle for MW. Tricky for Mtop Mtop : use MC to get the Mtop value for a given Ecut To be tested
use measurement of E and calibration from MW for light jets find the best E and calibration for bjets reproducing the observed Mtop
%ptcut and ptb distributions. How sensible is the dependence to the b jet calib. Could we calib also b
jets form the Top ?