Calorimeter Response Rjet The energy response to particle jets is smaller than unity due to energy loss in material in front of the calorimeter, uninstrumented regions, lower response to hadrons compared to EM objects and module-to-module inhomogeneities. It is a function of jet energy,
Rcone andηdet.
Showering Correction Sjet The showering corrections account for the fraction of energy deposited
outside the jet cone by particles originating from the corresponding particle jet as well as energy
deposited inside the cone by particles not belonging to the particle jet [54]. The net correction is typically smaller than unity. Sjet depends strongly on Rcone and ηdet but only mildly on the jet energy.
Since the Monte Carlo simulation does not model all of the effects described above precisely, different jet corrections are applied to data and Monte Carlo. The jet energy scale at D0 is approved for two different cone sizes Rcone=0.7 and Rcone=0.5 [55]. In this analysis, Rcone=0.5 jets are used and the so-called final p17 JES is applied.
Heavy flavor jets are expected to have different JES corrections because their harder fragmentation leads to a different response and showering correction. So far, there is no dedicated b jet energy scale certified by D0. The only correction applied is concerning jets containing a muon within Rcone. In these cases, the muon is supposed to stem from a semi-leptonic b decay and the jet is corrected to account for the momentum carried by the muon and the neutrino. In this instance the neutrino is assumed to have the same momentum as the muon.
3.6
Identification of Bottom Quark Jets
At hadron colliders, multijet events have the largest production cross sections and hence constitute a major source of background to many decay signatures. Multijet events can be significantly suppressed by identification of b jets in signal final states containing b quarks. Due to their relatively long lifetime
b hadrons can travel several millimeters before decaying. The three main ways to identify b jets are
i) secondary vertex reconstruction from tracks, ii) large impact parameter significance of tracks with respect to the primary vertex and iii) a muon reconstructed within the jet cone [56, 57]. In the past, four tools were certified at D0 to identify (tag) b jets:
Counting Signed Impact Parameters (CSIP) Is based on the number of tracks matched to jets (i.e.
lying within Rcone) with a large impact parameter significance (IPsig) with respect to the primary vertex [58–61].
Tag: IPsig>3 for≥2 tracks or IPsig>2 for≥3 tracks.
Jet Lifetime Probability Tagger (JLIP) Combines impact parameter information from all tracks
belonging to a jet into a single variable, giving the probability of all tracks to originate in the primary vertex [62–65].
Tag: Small probability (e.g.≤0.002≡very tight tag)
Secondary Vertex Tagging (SVT) Uses tracks significantly displaced from the primary vertex to
reconstruct secondary vertices [66–70].
fake jtnn 0 0.5 1 # of events 10 102 103 104 c jtnn 0 0.5 1 10 102 103 b jtnn 0 0.5 1 10 102 103
Figure 3.1: Distribution of the neural network output for light, c- and b jets. The cut values of the 12 certified operating points are indicated by the solid lines.
Soft Lepton Muon Tagging (SLT) Identifies muons within Rconeof the jet [71].
Tag: Associated muon found in jet.
All taggers calculate different variables containing valuable information on the likelihood of a jet to originate from a b quark and hence lead to a powerful discrimination between u,d,s,g and b,c jets.
Combining such variables by means of multivariate techniques can further enhance the descrimination power. In the neural network (NN) b-tagger implemented at D0, seven variables from SVT, JLIP and CSIP enter [72–74]. The SLT does not provide input as it is used later in the determination of b-tagging efficiencies (see below) as independent tagger and only contains information for a small fraction of jets. In order to be independent of any particular jet reconstruction algorithm, variables that depend on the details of the algorithm such as the jet ptare avoided. The seven input variables ranked by their discrimination power are:
1. Decay length significance of the secondary vertex with respect to the primary vertex (SVT). 2. Combination of all positive and negative decay length significance variables with different
weights (CSIP).
3. Probability of a jet to originate from the primary vertex (JLIP). 4. χ2per degree of freedom of the secondary vertex (SVT).
5. Number of tracks used for the secondary vertex reconstruction (SVT).
6. pt corrected mass of the secondary vertex (combined rest mass of tracks, assuming pions and correcting for neutral particles) (SVT).
7. Number of secondary vertices reconstructed within Rconeof the jet (SVT).
The output of the neural network is a single value for each jet constrained between 0 and 1 (cf. Fig. 3.1). The distribution for light quark jets (i.e. jets with a fake tag) peaks at values near 0, while that for b quark jets peaks near 1 and c jet distributions exhibit an intermediate behavior. The NN tagger is evaluated and certified for 12 operating points, defined by a lower cut on the NN output.