Return to ObjectID_Bkg_and_Dataset.tex CVS log

Up to [Nicadd] / ttbar / p20_taujets_note

Annotation of ttbar/p20_taujets_note/ObjectID_Bkg_and_Dataset.tex, revision 1.1.1.1

1.1       uid12904    1: \section{Object Identification \label{sec:objects}}
                      2: 
                      3: \noindent In this section we describe the main objects used in this study: \met, 
                      4: jets and hadronic tau candidates.
                      5: 
                      6: 
                      7: \subsection{\label{sub:tau--ID}\boldmath Taus}
                      8: 
                      9: %\subsubsection{Tau decay modes}
                     10: 
                     11: \noindent Taus are reconstructed in the D{\O} detector from energy in the calorimeter and one or more 
                     12: tracks.The tau reconstruction algorithm uses a cone of 
                     13: $\Delta R = \sqrt{{\delta \eta}^{2} + {\delta \phi}^{2}} < 0.5$ and an inner cone of $\Delta R < 0.3$
                     14: is used to calculate tau isolation variables. 
                     15: 
                     16: For us, the most important discriminating variables for $\tau$-leptons are \cite{tau-id}:
                     17: 
                     18: \begin{itemize}
                     19: \item Profile - $\frac{E_{T}^{1}+E_{T}^{2}}{\sum_{i}E_{T}^{i}}$, where
                     20: $E_{T}^{i}$ is the $E_{T}$ of the $i^{\rm th}$ highest $E_{T}$ tower in
                     21: the cluster.
                     22: \item Isolation, defined as $\frac{E(0.5)-E(0.3)}{E(0.3)}$, where $E(R)$
                     23: is the energy contained in a $y,\phi$ of radius $R$ around the calorimeter cluster centroid.
                     24: \item Track isolation, defined as scalar sum of the $p_{T}'$ of non-$\tau$ tracks in a $\eta,\phi$
                     25: cone of 0.5 around the calorimeter cluster centroid divided by similar sum for tracks associated with $\tau$.
                     26: \end{itemize}
                     27: 
                     28: Such variables are chosen based on the possible tau decays:
                     29: 
                     30: 
                     31: \begin{itemize}
                     32: \item electron or muon ($\tau\rightarrow e\nu_{e}\nu_{\tau}$ or $\tau\rightarrow\mu\nu_{\mu}\nu_{\tau})$,
                     33: BR = 35\% .
                     34: \item single charged hadron and no neutral hadrons ($\tau\rightarrow\pi^{-}\nu_{\tau}$), BR =12\% .
                     35: \item single charged hadron + $\geq1$ neutral 
                     36: hadron (i.e.,  $\tau\rightarrow\rho^{-}\nu_{\tau}\rightarrow (\pi^0+\pi^{-})\nu_{\tau}$)
                     37: , BR = 38\% .
                     38: \item 3 charged hadrons + $\geq0$ neutral hadrons, BR = 15\% (so-called
                     39: {}``3-prong'' decays). 
                     40: \end{itemize}
                     41: 
                     42: 
                     43: \noindent which leads us classificate reconstructed taus into three different types
                     44: depending on the number of tracks and electromagnetic (EM) clusters \cite{PDG}:
                     45: 
                     46: \begin{enumerate}
                     47: \item {\bf Type1}: calorimeter cluster, one matched charged track and no associated EM subcluster. 
                     48: Mainly $\tau\rightarrow\pi^{-}\nu_{\tau}$.
                     49: \item {\bf Type2}: calorimeter cluster, one matched charged track and one or more associated EM subclusters. 
                     50: Mainly $\tau\rightarrow\rho^{-}\nu_{\tau}\rightarrow \pi^0\pi^{-}\nu_{\tau}$.
                     51: \item {\bf Type3}: calorimeter cluster, two or more matched charged tracks and with or without EM subcluster. 
                     52: Mainly $\tau\rightarrow\pi^{-}\pi^{-}\pi^{+}(\pi^{0})\nu_{\tau}$.
                     53: \end{enumerate}
                     54: 
                     55: In order to provide an optimal tau identification,three Neural Networks (NNs) are trained to
                     56: identify the three types of the taus (1,2 and 3).
                     57: 
                     58: The output of these NNs provides a set of three variables ({\tt{nnout}} = 1,2,3)
                     59: to be used to select the tau in the event. The types roughly 
                     60: correspond to the $\tau$ lepton decay modes. High values 
                     61: of NN correspond to the physical taus, while low ones
                     62: should indicate jets misidentified as taus (fakes). 
                     63: 
                     64: \subsection{\label{sub:jet--ID}\boldmath Jets}
                     65: 
                     66: \noindent Jets are identified using the RunII cone algorithm \cite{jet-id} with cone size of
                     67: $\Delta R < 0.5$. The jet algorithm T42 \cite{t42} is ran before jet reconstruction
                     68: to remove isolated small energy deposits due to noise. D\O\ standard jet quality cuts \cite{jet-qual}
                     69: include L1 Trigger information, calorimeter electromagnetic fraction and coarse hadronic fraction.
                     70: 
                     71: Jets used in this analysis are required to have at least two primary vertex tracks associated to them
                     72: (vertex confirmed jets). It implies that although a calorimeter cluster
                     73: is still reconstructed as a jet, it will be discarded if it has less than 2 associated PV tracks.
                     74: In order to correct the energies of reconstructed jets in data and MC back to
                     75: parton-level energies, we apply certified jet energy scale correction (JES)\cite{jes}. Additionally,
                     76: jets containing a muon with $\Delta R(\mu , jet) < 0.5$ from a $b$-quark decay are corrected to take into
                     77: account the momentum carried away by the muon and the neutrino \cite{jesmu}.
                     78: 
                     79: 
                     80: \subsection{\label{sub:met-id}\boldmath \met}
                     81: 
                     82: \noindent Presence of neutrinos in an event causes an imbalance of energy in the transverse plane (\met).
                     83: This quantity is calculated from the transverse energies of all calorimeter cells that pass the 
                     84: T42 algorithm, except those of the coarse hadronic layers due to high noise level. However, they are included
                     85: in the case that they are clustered within a reconstructed jet. This raw \met is corrected for the energies
                     86: of other objects like photons, electrons, taus and jets. As muons deposit only a small portion of their
                     87: energy in the calorimeter, their momenta is subtracted from the \met vector.
                     88: 
                     89: 
                     90: \subsection{The Neural Network b-tagging Algorithm \label{sec:nntag}}
                     91: 
                     92: Being QCD and $W$ + jets the main sources of backgrounds in this analysis, requiring
                     93: the presence of at least one jet coming from a $b$-quark is a very powerful method of background 
                     94: rejection. The $b$-tagging algorithm used in this measurement is a
                     95: Neural Network (NN) tagging algorithm developed by the b-ID group \cite{bID-p20},
                     96: which combines 7 characteristic variables of SVT, JLIP and CSIP tagging algorithms into the NN 
                     97: discriminant. As in the previous analysis we have chosen the operating
                     98: point TIGHT, which is equivalent to requiring the NN discriminant output to be greater than 0.775.
                     99: Both the average efficiency and fake rate are comparable between this p20 version of the
                    100: algorithm and the version used in p17 \cite{bID-p17}.
                    101: 
                    102: \paragraph{$b$-tagging efficiency}
                    103: 
                    104: In data we apply the b-tagging algorithm directly to jets selected in our sample. In MC such ``direct tag''
                    105: is not done. Instead we have to apply a certain efficiency to MC samples. This inclusive b-decay efficiency ($\epsilon_{b}$)
                    106: is measured in data and it is the product of the probability to tag a b-jet in an MC sample ($\epsilon_{b}^{MC}$)
                    107: containing inclusive decays of the b quark times a scale factor. This data/MC scale factor is given by the ratio
                    108: of data semileptonic efficiency ($\epsilon^{DATA}_{b\rightarrow \mu}$) and a MC semileptonic efficiency ($\epsilon^{MC}_{b\rightarrow \mu}$).
                    109: This scale factor, that measures the effect on the tagging rate
                    110: caused by the differences in tracking between data and MC, is then used to properly scale
                    111: the MC-derived efficiency. It is assumed that such factor could be applied to any MC tagging efficiency \cite{bID-p20}.
                    112: 
                    113: \paragraph{$c$-tagging efficiency}
                    114: 
                    115: It is assumed that the same procedure adopted in the b-jet case is also valid for c-jets,
                    116: namely, a c-jet scale factor ($\epsilon^{DATA}_{b\rightarrow \mu}$)/($\epsilon^{MC}_{b\rightarrow \mu}$)
                    117: multiplies the probability to tag a c-jet in an MC sample ($\epsilon_{b}^{MC}$) to get the
                    118: c-tagging efficiency.
                    119: 
                    120: 
                    121: \paragraph{Light jet tagging efficiency}
                    122: 
                    123: The $b$-tag fake rate from light quarks is computed by measuring the
                    124: negative tag rate as defined in \cite{b_fake}. The method uses fits of
                    125: b-, c- and light jets tagging rates to binned data combined with sample
                    126: composition estimated from data. Binned NN output ($NN_{out}$) distributions
                    127: for b, c and light jets is fitted to data distribution using b- and c-jet efficiencies
                    128: provided by standard bID TRF's.
                    129: 
                    130: 
                    131: 
                    132: \paragraph{Taggability}
                    133: 
                    134: $b$-tag algorithm can't be applied to any jet, but only the ones that contain tracks. 
                    135: Such jets are called ``taggable'' and are defined by matching within $\Delta R$ $<$ 0.5 to 
                    136: a track jet, composed of at least two tracks. 
                    137: Just like b-tagging, taggability is different in data and MC, due to imperfect 
                    138: simulation of tracking system. To account for this, taggability rate functions are applied
                    139: to MC events. Such functions are parametrized in terms of jet $p_{T}$, jet $\eta$ and
                    140: primary vertex $z$. They were derived on the single top loose data samples, where the isolation
                    141: quality of the lepton is loose.The validity of these functions is tested by comparing observed
                    142: and predicted taggability to tight samples. The results show a good agreement as described in \cite{single top}. 
                    143: 
                    144: \clearpage