MSUGRA Topology Selector in inclusive search

1
MSUGRA Topology Selector in inclusive search
V. Zhukov University Karlsruhe
Some introduction and clarifications to the
ICHEP06 talk SUSY discovery potential at LHC
2
Summary of the CMS PTDR discovery reaches
3
Selection efficiency for the PTDR analysis
Use selection cuts defined in PTDR analysis and
apply them in FAMOS scans. Most of the studies
are tuned to m060, m1/2250 bulk region LM1
point.
tan?50 A00
No tau reconstruction efficiency
4
Discovery reaches 10 fb-1
Taking the Nbkg from analysis, can reproduce the
disovery reaches...
Significance S/sqrt(B)
5s
5
Event selection in inclusive search
Can we do better?
Standard analysis

• Only one parameter (Sign) characterize a
  hypothesis
• Can not say much about the model parameters
• Very sensitive to the bkg uncertainties

MSUGRA region dependent selection
Preselector -reduce the bkg Selectori -optimized
for i region
signal
Selector i optimized for i-topology
ns
ns
S
Si
Selector i optimized for i-topology
Ns
Sign
Nb
backgrounds
Bkg suppression should be region independent!
Preselector
nb
/- uncert.

• Extends the discovery reach by cut
  optimization()
• Different selectors can be used for mSUGRA region
  separations, effectively increases significance
  by vetoing on other regions (only one region
  exist)()
• Strong model dependency(-)

6
Cross sections
Fractions of different production channels
Main production channels
Total mSUGRA cross sections
s LOtot pb
The sparticles decays according to the mass
spectrum and couplings. Expect region dependent
signal topologies MET Njets Nleptons Scan
mSUGRA with FAMOS (10kev/point), standard
reconstruction algorithms, and calculate the
average observables.
tan?50 A00
mSUGRA (ISASUGRA LO) PYTHIA NLO (PROSPINO)
KNLO1.3-1.8 (m0100-2000)
7
MSUGRA observables Jets
Reconstructed (FAMOS) averaged observables in
m0-m1/2 plane
ltNjgt ETgt30GeV
ltNBjgt B jets
mgmq long cascades
c02-gt H0 -gtbbar
Second jet ltETgt
Highest jet ltETgt
200
250
Hardest jet in q decays
120
Third jet ltETgt
Cosf between first and second
Cosf between first and MET
Cosf between second and MET
tan?50 A00
8
more observables MET
S ET from CaloTowers
MET from CaloTowers
2 body
Low MET region
MinvminvH1minvH2/2 invariant mass per hemisphere
Heff METS ET jets S ETleptons
9
And more Leptons
ltN gt electrons
ltN gt muons
ltN gt SameSignSameFlavor
ltN gt OppositeSignSameFlavor
Highest muon ltPTgt m1/2
Cosf OSSF and MET
back2back
Nss/Nss-
ltMinvgt OSSF
Zpeak
10
Trigger on MSUGRA
Trigger is emulated in FAMOS using reconstructed
objects (as for the scans).
Trigger efifciency is droping for large m0 low
m1/2 region from 98 to 60
L1
HLT
Inclusive efficiencies
HLT/L1
11
Region separation.
Use Neural Network (NN) for region separation.
Train NN for region i against other regions
j-i
Sig2/bkg13
Preselection efficiency
ExampleTeacher out
Selection efficiencies of 3 NNs trained for
Sig1
Sig2
Sig3
12
Constrain production channels fractions
Can we separate channels and identify the
fractions(f.ex. gg, qq, etc)? Train NNs for
individual channels against others for one test
point (m0500 m1/2200)
MC and Rec fractions
NN selection efficiencies ( m0500 m1/2200)
The selection efficiency E is a convolution of
NN and preselection efficiencies Eie (presl)ki
e (NN)k . Find the fractions F i from the
observed Di DiEi X Fi and compare with the
expectation from Mc(normalized)
Another point(m02500 m1/2600) with the same
NN( m0500 m1/2200)
Larger errors, but still can constrain the ratios.
13
Bkg suppression
Preselect data samples and use NN to train ttbar
against different mSUGRA regions, Produce many
NNs(100) and check efficiency for the bkg and
signal.
Similar patern, Ppreselection efficiency
dominates? Are there improvement in discovery?
Up to 10 spread, is it acceptable?
14
Summary and Plans
The mSUGRA topology selection based on NN is
working and region separation
is possible already in inclusive search. In
combinations with the kinematic edges and cross
sections, this can improve reconstruction
of mSUGRA parameters at low statistics (high mass
regions) This topology selector also can be used
for testing non SUSY models.
Under investigation - optimize region dependent
bkg suppression - sensitivity to the
uncertainties, selection of NN parameters -
example of the BLIND analysis with multiregion
selectors - extraction of the channel fractions
for the whole plane - Use of the Genetic
Algorithm instead NN. - Combination with other
measurements by using Markov chain. - check with
other models
15
Back up slides
16
Jets ET in PYTHIA and ALPGEN
Simulate Wjets in PYTHIA (PS) and ALPGEN(ME)
(Nj0,1,2,3,4) (using standard CMS procedure and
jets matching in CMKIN ) Simulate detector with
FAMOS using standard reconstruction Compare
reconstructed highest jet ET(iterativeCone) and
MET (CaloTowers)
Significant difference in tails, as expected,
will decrease SUSY discovery reaches where the
MET selection is critical