1. Search of Higgs boson in vector boson fusion
qq → qq H0 → qq qq eνe:
a CMS simulation
I.N.F.N. Turin

2. V-V fusion: the process
“tag” quarks
correlated with incoming quarks and protons, expected at higher η, one forward, the other backward q1 q2
q't,1
q't,2 qV e νe
q'V ? V W
leptons
expected central in η, from a W decay hopefully coming from Higgs boson
“central” quarks
expected more central in η, from decay of the other W or Z from Higgs boson
Interaction scheme and labels
13/12/2006

3. Software and data formats
“hard scattering” qq → qq qq eνe, 4-momentum, particle ID, colour stream information, scale data
PHASE: 770 bytes/event
Generation
hadrons and leptons from hard scattering and underlying event; 4-momentum, particle ID, history tree, vertex information
PYTHIA: 13 kb/event
Fragmentation
ROOT tree with ORCA (C++) objects describing both generated particles and reconstructed objects, with their specific characteristics
FAMOS/ExRootAnalysis: 100 kB/event
Reconstruction
Currently 4-momentum, charge, relations information; in our hands!
“VVfusionFW”: some kB/event
Analysis
13/12/2006

4. Software and data formats
PHASE events
FAMOS...
FAMOS reconstructed events
generated objects
reconstructed objects
Tagging
Combinations...
roles recognition
χ2-like quantities
roles assignment
reconstruction information
goodness of roles assignment
13/12/2006

5. PHASE Find the Truth™ q't,1 e q1 νe qV q2 q'V q't,2 H0
Higgs boson mass 200 GeV/c GeV/c2 ∞ (no Higgs)
Available events
Years with L = 100 fb-1/year ½
Analysed events (0.89) (0.72) (0.69)
Layout fitting V-V fusion (0.66) (0.53) (0.50)
“topped” event (0.32) (0.32) (0.32)
WWVV vertex 3062 (0.03) 2778 (0.04) 2870 (0.02)
V-V fusion event (0.29) (0.17) (0.14)
Higgs event (0.24) (0.12) (0.08) q1 q2
q't,1
q't,2 qV e νe
q'V H0
Selection of samples and flags
Starting numbers and x-sec in PHASE samples
PHASE
Mass windows used: W±: ( ± 15) GeV/c2; Z0: ( ± 15) GeV/c2; t/t: (174.3 ± 20) GeV/c2
Cross section in picobarn of each subsample is indicated in brackets.
13/12/2006

6. A background: p p ―> t t
The most similar background with the same final status as signal is p p → t t → b b W W, where at least one W boson decays hadronically: qV ℓ
νℓ/q'
q'V W
ℓ/q b t p
W decay semileptonic fully hadronic
Total cross section 100 pb 324 pb
Available events
The sample was generated by PYTHIA from the core QCD process p p → t t .
Process layout, generic cross sections
13/12/2006

7. Reconstruction algorithms
FAMOS reconstruction computes some objects from the generated particles given as input, which are the result of fragmentation of the six fermions of PHASE event and the remnants of the protons. The following algorithms have been used in this analysis:
electrons are reconstructed as ECAL superclusters (“EGSClus”);
jets: HCAL and ECAL energy deposits in a cone around “seeds” are assigned to a jet; the maximum size of this cone is measured by its angular radius Δr2 = Δη2 + Δφ2 < R2 the radius used is R = 0,5; different calibrations are available, among them: no calibration (“JetIC5A”) and γ-jet calibration (“JetIC5C”), which is used here;
neutrino: the starting hypotesis is that the total transverse momentum of each event is null, while nothing can be inferred about longitudinal one; we assume all transverse momentum missing for this to happen is due to just one neutrino; it is computed from the other reconstructed objects via a cone algorithm (“METIC”).
To recognize which of the many objects reconstructed are direct manifestation of generated fermions (four quarks, one electron and one neutrino), for each event a χ2-like quantity is minimized:
Then, each particle is labelled as “well reconstructed” if its contribution is χi < χTHR = 0,5 .
The contribution for neutrino is a little simpler: χν2 = (ΔE┴/E┴)2 + Δφ2.
 2 = 𝑖  𝐸 𝑖,𝑟𝑒𝑐 − 𝐸 𝑖,𝑔𝑒𝑛 𝐸 𝑖,𝑔𝑒𝑛  2    𝑖,𝑟𝑒𝑐 −  𝑖,𝑔𝑒𝑛  2    𝑖,𝑟𝑒𝑐 −  𝑖,𝑔𝑒𝑛  2   𝑖,𝑟𝑒𝑐 −  𝑖,𝑔𝑒𝑛  2
Process layout, generic cross sections
13/12/2006

8. Electron: ECAL superclusters
Electrons are reconstructed from ECAL data only; reconstruction provides:
hit, a crystal where energy has been deposited;
cluster, a set of neighboring hits;
super-cluster, a set of clusters along the same η ring, which can gather both an electron and its irradiated photons, which are not bent from axial magnetic field;
electron candidate, a supercluster plus a pointing track from inner trackers
This analysis assumes an electron to be represented well enough by ECAL superclusters.
What is a cluster
Number of clusters
Pt rank, Pt distribution
Isolation
H/E, E/p
13/12/2006

9. Electron: reconstruction
electron chi2; dE/E; dR
sample good reconstruction
mH [GeV/c2] full sample Higgs signal
/ /2690
/ /1656
no Higgs 9751/ /1553
13/12/2006

10. Electron: transverse momentum
Sorting reconstructed ECAL superclusters by transverse momentum, we see that the one generated by the electron from PHASE is the one with the biggest p┴ on 65% of events, in the first two superclusters with highest p┴ in 90% of events, in the first three in 95% of events.
One sample: Pt rank; all samples: Pt distribution of reconstructed electron (including backgrounds)
13/12/2006

11. Neutrino: reconstruction
 2 ≡  𝐸 ⊥𝑟𝑒𝑐 − 𝐸 ⊥𝑔𝑒𝑛 𝐸 ⊥𝑔𝑒𝑛  2    𝑟𝑒𝑐 −  𝑔𝑒𝑛  2   𝑟𝑒𝑐 −  𝑔𝑒𝑛  2
formula!
neutrino chi2; dET/ER; dPhi
sample good reconstruction
mH [GeV/c2] full sample Higgs signal
/ /2690
/ /1656
no Higgs 5947/ /1553
13/12/2006

12. Missing transverse energy
The ratio of missing transverse energy over the “scalar missing energy” (the sum of transverse energy of all reconstructed objects) is an indicator of how important is the measured missing energy in the energetic balance of the event.
Pt distribution
MET/SET
13/12/2006

13. Lepton-decaying vector boson
W m┴
W p┴
From truth reconstruction: eta, Pt, mass, transverse mass (definition!)
13/12/2006

14. Cuts on leptons and W reconstruction
sample PHASE p p → t t
mH = 200 GeV/c2 mH = 500 GeV/c2 no Higgs leptonic hadronic
not H0 H0 not H0 H0 not H0 H0
total:
events after cuts:
(“purity”: 6175(51%) 3711(55%) 2485(50%) )
ε∙σ [pb]
p┴ rank < 2
e±: p┴ > 20 GeV/c
νe: p┴ > 30 GeV/c
m┴W є [40, 105]
mW є [60, 100]
MET/SET > 0.07 q1 q2
q't,1
q't,2 qV e νe
q'V ? V W
Tables of cuts (with purity where available).
Influence of mass windows (mass and transverse mass).
13/12/2006

15. What is that “purity” Tables of cuts (with purity where available).
In the previous slide, a “purity” is quoted; usually, the purity is defined as events of expected kind over all events in sample. In our case, we can't build such a simple ratio:
PHASE sample has a certain number of events with six fermions each; let's assume we can identify each event as belonging to signal or background;
generated event is reconstructed: some sets of objects (clusters, track etc.) are computed;
for each event, the first step is to “connect” the generated particles to the reconstructed objects: six among the reconstructed objects must be associated to them; the way it's done (via a χ2) almost always produces a result, but sometimes that result is so different from generated data (high χ2) that we can't trust it; this event is probably beyond recognition since it's badly reconstructed, be it signal or background;
then, the true analysis algorithm is applied; it splits data into two sets: signal candidates and background candidates; this is the only discrimination we have in real life; furthermore, it assigns to each role in V-V fusion event a reconstructed object;
we can look into the signal candidates set, now, and see whether they were signal or not; that is the true purity; which will be very low, since we are not trying to separate signal from background yet;
we can look among events known to be signal, and see what happened to them:
event has been reconstructed badly, and the result of assignment will be screwed;
event has been well reconstructed:
algorithm has not found any suitable assignment;
algorithm has found an assignment, which, alas, is the wrong one
algorithm has found an assignment, which, blessed it, it's the correct one
Tables of cuts (with purity where available).
Influence of mass windows (mass and transverse mass).
13/12/2006

16. Jets
Jets have been reconstructed with a iterative cone algorithm with ΔR = 0.5 around seed. The energy has been calibrated by simulating γ/jet events.
Number of jets (with different Pt cuts)
Eta; invariant mass; Pt ranking; eta differences
Isolation
13/12/2006

17. Jets: reconstruction
For one sample (200): central and tag jets chi2; dE/E; dR
well reconstructed in full sample well reconstructed in Higgs signal
[GeV/c2] both tag jets both central jets both tag jets both central jets
mH = / / / /2690
mH = / / / /1656
no Higgs 7632/ / / /1553
13/12/2006

18. Jets: transverse momentum
Higgs events
Number of jets (with different Pt cuts)
Eta; invariant mass; Pt ranking; eta differences
Isolation
not Higgs events
13/12/2006

19. Jets: transverse momentum rank
Higgs events
Number of jets (with different Pt cuts)
Eta; invariant mass; Pt ranking; eta differences
Isolation
not Higgs events
13/12/2006

20. Jets: pseudorapidity Number of jets (with different Pt cuts)
Eta; invariant mass; Pt ranking; eta differences
Isolation
13/12/2006

21. Hadronic-decaying weak boson
sample vector bosons fusion p p → t t
mH = 200 GeV/c2 mH = 500 GeV/c2 no Higgs leptonic hadronic
not H0 H0 not H0 H0 not H0 H0
total:
“known” events:
after cuts:
p┴ rank < 5 (“purity”: 2561(40%) 1737(45%) 1115(39%) )
ε∙σ [pb]
tags: p┴ > 30 GeV/c; 1st c: p┴ > 15 GeV/c, 2nd c: p┴ > 10 GeV/c
ηFW > –1, ηBW < +1, 1stc: |ηc1| > 2.5, 2nd c: |ηc2| > 2; ηFW > {ηc1,ηc2} > ηBW tags: mj >10 GeV/c2; central jets: mj > 5 GeV/c2
|Δηtags| є [3, 8], |Δηcentral| < 1.5
mW є [60, 100]
mtags > 200 GeV/c2 q1 q2
q't,1
q't,2 qV e νe
q'V H0
Distributions of true W: mass, eta
13/12/2006

22. Hadronic weak boson resolution
Distributions of true W: mass, eta
13/12/2006

23. Higgs (after ~2 months) Table of events with bot cuts.
sample (10000 events) vector bosons fusion p p → t t
mH = 200 GeV/c2 mH = 500 GeV/c2 no Higgs leptonic hadronic
not H0 H0 not H0 H0 not H0 H0
total:
Higgs candidates:
ε∙σ [pb] α×1.910 α×0.504
Table of events with bot cuts.
Invariant mass of Higgs boson, transverse mass.
13/12/2006

24. Higgs (after ~1 year) Table of events with bot cuts.
sample (~ events) vector bosons fusion p p → t t
mH = 200 GeV/c2 mH = 500 GeV/c2 no Higgs leptonic hadronic
not H0 H0 not H0 H0 not H0 H0
total:
Higgs candidates:
ε∙σ [pb] α×1.910 α×0.504
Table of events with bot cuts.
Invariant mass of Higgs boson, transverse mass.
13/12/2006

25. Biggest ToDos
samples from different generators must have cross sections weighted
a new sample with an Higgs mass bigger than 200 GeV/c2 is required
a shy attempt to cut top background by an invariant mass cut on three jets
tracker information is not extensively used yet (but it's used in jets construction anyway)
pile-up must be added to make this even vaguely believable
Normalization between samples
13/12/2006

26. Top mass
Invariant mass of the two jets reconstructed as central (from W) and each of the two tag jets are plotted (two entries per event).
Normalization between samples
13/12/2006