Monte Carlo Reweighting: Monte Carlo Reweighting: an Alternative to 10TeV an Alternative to 10TeV Production? Production? Claire Gwenlan, UCL Claire Gwenlan, UCL Recently investigated reweighting existing FullSim samples to 10 TeV Recently investigated reweighting existing FullSim samples to 10 TeV My motivation was to look at the potential for SUSY discovery with small My motivation was to look at the potential for SUSY discovery with small amounts of 10 TeV, BUT given the decisions currently being made on which amounts of 10 TeV, BUT given the decisions currently being made on which samples to produce at 10 TeV, perhaps method is relevant and/or samples to produce at 10 TeV, perhaps method is relevant and/or interesting to this audience? interesting to this audience? Just a couple of slides shown here, taken from a recent talk I gave Just a couple of slides shown here, taken from a recent talk I gave in the ATLAS-UK SUSY meeting (full talk appended in Backups) in the ATLAS-UK SUSY meeting (full talk appended in Backups)
23
Embed
Monte Carlo Reweighting: an Alternative to 10TeV Production? Claire Gwenlan, UCL Recently investigated reweighting existing FullSim samples to 10 TeV My.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Monte Carlo Reweighting: Monte Carlo Reweighting: an Alternative to 10TeV Production?an Alternative to 10TeV Production?
Claire Gwenlan, UCLClaire Gwenlan, UCL
Recently investigated reweighting existing FullSim samples to 10 TeV Recently investigated reweighting existing FullSim samples to 10 TeV My motivation was to look at the potential for SUSY discovery with small amounts My motivation was to look at the potential for SUSY discovery with small amounts
of 10 TeV, BUT given the decisions currently being made on which samples to of 10 TeV, BUT given the decisions currently being made on which samples to produce at 10 TeV, perhaps method is relevant and/or interesting to this audience?produce at 10 TeV, perhaps method is relevant and/or interesting to this audience?
Just a couple of slides shown here, taken from a recent talk I gave in the Just a couple of slides shown here, taken from a recent talk I gave in the ATLAS-UK SUSY meeting (full talk appended in Backups)ATLAS-UK SUSY meeting (full talk appended in Backups)
If no 10 TeV Monte Carlo … can If no 10 TeV Monte Carlo … can re-weightre-weight existing 14 TeV samples existing 14 TeV samples hadronic cross section given by:hadronic cross section given by:
For a particular set of MC events – with particular incoming parton momenta (pFor a particular set of MC events – with particular incoming parton momenta (p 11, p, p22) –) –the hard cross section is *fixed*. To change to a different beam energy, need to the hard cross section is *fixed*. To change to a different beam energy, need to adjust the *probability* that partons of those momenta were found in the proton in the adjust the *probability* that partons of those momenta were found in the proton in the
first place (i.e the PDFs)first place (i.e the PDFs) just a PDF re-weight just a PDF re-weight
xi : x of parton i at 14 TeV; x i’: x of parton i at 10 TeV
Re-weight 14 TeV MC to 10 TeV using:Re-weight 14 TeV MC to 10 TeV using:
22
hard sub-process cross section
What you need to know:What you need to know:incoming parton momenta for:incoming parton momenta for:
x = |px = |pzz|/E|/Ebeambeam
(different for different beam (different for different beam energies)energies)
QQ22, the fact. scale of the process , the fact. scale of the process (same for both beam energies)(same for both beam energies)
incoming parton flavoursincoming parton flavours
Reweighting : 14 TeV Reweighting : 14 TeV 10 TeV 10 TeV
Check of Reweight MethodCheck of Reweight Method
Good agreement between reweighted and 10 TeV samplesGood agreement between reweighted and 10 TeV samples
EG:EG: 50000 SU3 SUSY events generated in ATLFAST 50000 SU3 SUSY events generated in ATLFAST
pTJet,1,2 >150,100
GeV
Meff = pTJet,1 + pT
Jet,2 + MET
Effective Mass
example weights (all < 1)example weights (all < 1)
33
PDF weight
Other example checksOther example checks a few other example checks a few other example checks (currently low statistics, except for J8MET)(currently low statistics, except for J8MET) good good agreementagreement
Results: SU3 SUSY vs. BkgResults: SU3 SUSY vs. Bkg
Integrated L = 100 pbIntegrated L = 100 pb-1-1
BEFOREBEFORE reweighting reweighting
AFTERAFTER reweighting reweighting
14 TeV 14 TeV 10 TeV: 10 TeV: SU3 signal reduced by ~ 2.9SU3 signal reduced by ~ 2.9
SM bkg reduced by ~ 2.3SM bkg reduced by ~ 2.3
55
Example: CSC5 0Example: CSC5 0lepton lepton + MET dijet analysis+ MET dijet analysis
SummarySummaryThe PDF re-weight method The PDF re-weight method seemsseems to work to work
Note some caveats however Note some caveats however method is not adjusting the amount of method is not adjusting the amount of energy left in the beam remnant, so the underlying event treatment is energy left in the beam remnant, so the underlying event treatment is not quite rightnot quite right
however, all checks done so far indicate this isn’t a big effect however, all checks done so far indicate this isn’t a big effect
Could this method be useful, given the large amount of 10 TeV Could this method be useful, given the large amount of 10 TeV simulation we are about to set going? simulation we are about to set going?
At the very least, it could be used by individuals whose favourite At the very least, it could be used by individuals whose favourite samples are not planned to be produced at 10 TeV samples are not planned to be produced at 10 TeV
But perhaps it could even be more helpful than this? But perhaps it could even be more helpful than this? i.e. an alternative to some 10 TeV production, saving lots of time, CPU, … ?i.e. an alternative to some 10 TeV production, saving lots of time, CPU, … ?
Presumably this would all have to be verified and validated Presumably this would all have to be verified and validated
66
BackUps - (Full Talk)BackUps - (Full Talk)
SUSY Discovery at 10 TeV ?SUSY Discovery at 10 TeV ?
10TeV run coming soon (2-3 months? 100 pb10TeV run coming soon (2-3 months? 100 pb-1-1 possible?) possible?)
Studies show that, if SUSY is relatively light, it could be discovered Studies show that, if SUSY is relatively light, it could be discovered very early at the LHC (based on 14TeV Monte Carlo)very early at the LHC (based on 14TeV Monte Carlo)
– – so what about at 10TeV – could we hope to see something?so what about at 10TeV – could we hope to see something?
99
“Commissioning to 10 TeV should be fast, no quench being anticipated, giving us confidence that the experiments will be recording data at record high energies by the summer.
In 1989, it was only a matter of weeks before LEP produced its first profound result – a measurement of the number of light neutrino families. In this respect at least, history will not be repeating itself. The LHC is a discovery machine, and the discoveries it is chasing will require a little more patience.”
Robert Aymar, on the 10TeV run
1
0
ˆ 2 21 2 1 1 2 2
flav
= dx dx PDF(x ,Q ,flav ) × PDF(x ,Q ,flav ) ×
PDFs PDFs from 14TeV from 14TeV 10TeV 10TeV
Currently no MC… BUT can re-weight existing 14TeV samplesCurrently no MC… BUT can re-weight existing 14TeV samples hadronic cross section given by:hadronic cross section given by:
For a particular kinematic configuration For a particular kinematic configuration
with s, t, u, should just need to change the with s, t, u, should just need to change the
probability that incoming partons *had* probability that incoming partons *had*
that configuration i.e. just a PDF re-weightthat configuration i.e. just a PDF re-weight
proton PDFs
' 2 ' 21, 2
2 21 2
PDF x Q ,flav1 ×PDF x ,Q ,flav2Weight=
PDF x ,Q ,flav1 ×PDF x ,Q ,flav2
' 2 ' 21, 2
2 21 2
PDF x Q ,flav1 ×PDF x ,Q ,flav2Weight=
PDF x ,Q ,flav1 ×PDF x ,Q ,flav2
Plot from James Stirling: DIS08
^ ^ ^ ^ ^̂
xi : x of parton i at 14TeV; x i’: x of parton i at 10TeV
Re-weight MC to 10TeV using:Re-weight MC to 10TeV using:
1010
Check of Reweight MethodCheck of Reweight Method
Good agreement between reweighted and 10TeV samples (bkgs also checked)Good agreement between reweighted and 10TeV samples (bkgs also checked)
EG:EG: SU3 SUSY events generated in ATLFAST SU3 SUSY events generated in ATLFAST
pTJet,1,2 >150,100
GeV
Meff = pTJet,1 + pT
Jet,2 + MET
Effective Mass
example weights (all < 1)example weights (all < 1)
1111
PDF weight
EG: CSC5-style, inclusive 2- and 3-Jet analyses (full details in note)EG: CSC5-style, inclusive 2- and 3-Jet analyses (full details in note) results shown in main part of talk are for the 2-Jet case (3-Jet results in backups)results shown in main part of talk are for the 2-Jet case (3-Jet results in backups)
Numbers are maximum Numbers are maximum significances (taken above significances (taken above some Msome Meffeff threshold) threshold)
(don’t take too much notice of the (don’t take too much notice of the actual values – it’s just to get a actual values – it’s just to get a rough feeling)rough feeling)
* Zn is a measure of the significance (as used in CSC5) which tries to take into account systematic
uncertainties on the bkg measurements. The numbers in the table above are calculated assuming
50% uncertainty on QCD and 20% on all other bkgs – these are not necessarily the “right”
numbers – dedicated bkg studies needed for those! 1717
Numbers are maximum Numbers are maximum significances (taken above significances (taken above some Msome Meffeff threshold) threshold)
(don’t take too much notice of the (don’t take too much notice of the actual values – it’s just to get a actual values – it’s just to get a rough feeling)rough feeling)
SummarySummary
Is there potential for discovery with small amounts of 10 TeV data?Is there potential for discovery with small amounts of 10 TeV data?
YES – there does seem to be potential!YES – there does seem to be potential!
A 5x increase in centre-of-mass energy compared to previous A 5x increase in centre-of-mass energy compared to previous experiments is still a lot!!! – and the discovery of light SUSY may experiments is still a lot!!! – and the discovery of light SUSY may not need much data (it doesn’t take much to give large S/not need much data (it doesn’t take much to give large S/B values B values for the models considered here)for the models considered here)
BUT that data still needs to be understoodBUT that data still needs to be understood
This was really just a quick look for fun – to see if anything is even This was really just a quick look for fun – to see if anything is even potentially feasible! It looks like it could be, but the limiting factor potentially feasible! It looks like it could be, but the limiting factor will of course be how well we can will of course be how well we can determine and understanddetermine and understand the the backgrounds with the small amount of data we expect.backgrounds with the small amount of data we expect.
1818
BackUpsBackUps
Integrated L = 100 pbIntegrated L = 100 pb-1-1
Results: Results: other mSUGRA pointsother mSUGRA points