| 1 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% |
| 2 |
< |
\section{Backgrounds}\label{section:BG} |
| 2 |
> |
\section{Backgrounds}\label{sec:BG} |
| 3 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% |
| 4 |
|
This section reviews our evaluation of background in the $4\ell$ analysis. We discuss expected yields and the predicted $m(4\ell)$ shapes, both of which are used in the limit and sensitivity calculations described in Section~\ref{sec:Extraction}. We estimate Electroweak (EWK) backgrounds with Monte Carlo. Our estimates of instrumental and jet backgrounds are data-driven. |
| 5 |
|
|
| 6 |
|
%_________________________________________________________________ |
| 7 |
|
\subsection{Electroweak Backgrounds}\label{sec:EWK} |
| 8 |
|
%_________________________________________________________________ |
| 9 |
< |
We use the $ZZ \rightarrow 4\ell$, $WZ \rightarrow 3\ell$ and $Z\gamma$ MC samples listed in Table~\ref{tab:MC} to estimate yields and $m(4\ell)$ shapes for these backgrounds. We correct the acceptances determined from simulation using the procedures described in Section~\ref{section:signalEff}. Background yields follow from the corrected $4e$, $4\mu$ and $2e2\mu$ acceptances ($\alpha_{c}$) for each process: |
| 9 |
> |
We use the $ZZ$, $WZ$ and $Z\gamma$ MC samples listed in Table~\ref{tab:MC} to estimate yields and $m(4\ell)$ shapes for these backgrounds. We correct the acceptances determined from simulation using the procedures described in Section~\ref{sec:Signal}. We determine background yields using the corrected $4e$, $4\mu$ and $2e2\mu$ acceptances ($\alpha_{c}$) for each process: |
| 10 |
|
|
| 11 |
|
\begin{eqnarray} |
| 12 |
< |
N^{exp}_{i} & = & \alpha^{c}_{i}\int\mathcal{L}(k_{i}\sigma_{i}) |
| 12 |
> |
N^{exp}_{i} & = & \alpha^{c}_{i}\int\mathcal{L}\sigma_{i} |
| 13 |
|
\end{eqnarray} |
| 14 |
|
|
| 15 |
< |
The cross sections and K-factors used in formula above are taken from Table~\ref{tab:xsec}. Table~\ref{tab;MCBG} lists the $\alpha_{c}$ and the expected $2.1\rm~fb^{-1}$ yields for each of the backgrounds. Figure~\ref{fig:MCshapes} shows a yield-normalized stack of the corresponding $m(4\ell)$ distributions. |
| 15 |
> |
The cross sections used in formula above are taken from Table~\ref{tab:MC}. Table~\ref{tab:MCBG} lists the $\alpha_{c}$ and the expected $2.1\rm~fb^{-1}$ yields for the diboson backgrounds. Figure~\ref{fig:MCshapes} shows a yield-normalized stack of the corresponding $m(4\ell)$ distributions. |
| 16 |
> |
|
| 17 |
> |
%------------------------------------------------- |
| 18 |
> |
\begin{figure}[htb] |
| 19 |
> |
\begin{center} |
| 20 |
> |
\includegraphics[width=0.5\linewidth]{figs/HF1.png} |
| 21 |
> |
\caption{MC Background Shapes.{\bf put the right plot here} } |
| 22 |
> |
\label{fig:MCshapes} |
| 23 |
> |
\end{center} |
| 24 |
> |
\end{figure} |
| 25 |
> |
%------------------------------------------------- |
| 26 |
|
|
| 27 |
|
%------------------------------------------------- |
| 28 |
|
\begin{table}[htb] |
| 29 |
|
\begin{center} |
| 30 |
|
\begin{tabular}{c|cc|cc|cc} |
| 31 |
+ |
{\bf Process} & $\alpha^{c}_{ee}$ & $N^{exp}_{ee}$ & $\alpha^{c}_{\mu\mu}$ & $N^{exp}_{\mu\mu}$ & $\alpha^{c}_{2e2\mu}$ & $N^{exp}_{2e2\mu}$ \\ |
| 32 |
|
\hline |
| 33 |
< |
process & $\alpha^{c}_{ee}$ & $N^{exp}_{ee}$ & $\alpha^{c}_{\mu\mu}$ & $N^{exp}_{\mu\mu}$ & $\alpha^{c}_{2e2\mu}$ & $N^{exp}_{2e2\mu}$ \\ |
| 23 |
< |
\hline |
| 24 |
< |
$ZZ*$ & ~ & ~ & ~ & ~ & ~ & ~ \\ |
| 33 |
> |
$ZZ^{*}$ & ~ & ~ & ~ & ~ & ~ & ~ \\ |
| 34 |
|
$WZ$ & ~ & ~ & ~ & ~ & ~ & ~ \\ |
| 35 |
|
$Z\gamma$ & ~ & ~ & ~ & ~ & ~ & ~ \\ |
| 36 |
|
\hline |
| 37 |
|
\end{tabular} |
| 38 |
< |
\caption{{\bf MC Background Yields.}\small{blah.}\label{tab:MCBG}} |
| 38 |
> |
\caption{MC Background Yields.} |
| 39 |
> |
\label{tab:MCBG} |
| 40 |
|
\end{center} |
| 41 |
|
\end{table} |
| 42 |
|
%------------------------------------------------- |
| 43 |
|
|
| 44 |
< |
We consider two sources of systematic uncertainties on EWK background predictions. The first is due to the uncertainty on the efficiency scale-factors, which we propagate from Tables~\ref{tab:}-\ref{tab:} to the corrected acceptance for each channel. The second uncertainty concerns the shape of the $m(4\ell)$ distribution predicted by the MC, which we estimate by reweighting our POWHEG samples at generator-level to the $m(4\ell)$ distributions predicted by MCFM. We take the relative difference in shape as the uncertainty input to the limit calculation. Table~\ref{tab:EWKsys} lists the yield uncertainties. Figure~\ref{fig:EWKshapeSys} shows the relative shape differences we obtain after reweighting. |
| 35 |
< |
|
| 36 |
< |
%------------------------------------------------- |
| 37 |
< |
\begin{figure}[htb] |
| 38 |
< |
\begin{center} |
| 39 |
< |
\includegraphics[width=0.5\linewidth]{figs/HF1.png} |
| 40 |
< |
\caption{MC Background Shapes.\label{fig:MCshapes} } |
| 41 |
< |
\end{center} |
| 42 |
< |
\end{figure} |
| 43 |
< |
%------------------------------------------------- |
| 44 |
> |
We consider two sources of systematic uncertainties on the EWK background predictions. The first is due to the uncertainty on the efficiency scale-factors, which we propagate from the tables of Section~\ref{sec:Leptons} to the corrected acceptance for each channel. {\bf still have to do this}. The second uncertainty concerns the influence of missing higher-orders on the mass shapes and kinematics predicted by the MC. We estimate the magnitude of this effect by reweighting the POWHEG samples at generator-level to the $m(4\ell)$ distributions predicted by MCFM with renormalization and factorization scales varied by $\times 2$, $/2$. We take the relative differences in shape are used as an uncertainty in the limit calculation. Figure~\ref{fig:EWKshapeSys} shows the relative shape differences we obtain after reweighting. {\bf done for ZZ, needed for the others}. |
| 45 |
|
|
| 46 |
|
%------------------------------------------------- |
| 47 |
|
\begin{figure}[bht] |
| 48 |
|
\begin{center} |
| 49 |
|
\includegraphics[width=0.5\linewidth]{figs/HF1.png} |
| 50 |
< |
\caption{EWK Shape Differences From MCFM Reweight.\label{fig:EWKshapesSys} } |
| 50 |
> |
\caption{EWK Shape Differences From MCFM Reweight.{\bf get the right plot in here} } |
| 51 |
> |
\label{fig:EWKshapeSys} |
| 52 |
|
\end{center} |
| 53 |
|
\end{figure} |
| 54 |
|
%------------------------------------------------- |
| 56 |
|
%_________________________________________________________________ |
| 57 |
|
\subsection{Instrumental/Fake Backgrounds}\label{sec:fakes} |
| 58 |
|
%_________________________________________________________________ |
| 59 |
< |
$Z+jets$ , $Zb\bar{b}/c\bar{c}$ and $t\bar{t}$ backgrounds (collectively, $\ell\ell jj$) contribute to the $4\ell$ signal region when jets in these events are either mismeasured as leptons or produce real leptons through secondary interactions. These processes are difficult to accurately simulate so we estimate their contribution from data. We assess $\ell\ell jj$ backgrounds using the ``fakeable object'' technique~\cite{fakeable}, employing ``fakerates'' defined with respect to loosely identified lepton candidates, referred to as {\it denominator objects}. Electron and muon denominator selections are defined in Table~\ref{tab:fo}. |
| 59 |
> |
$Z+jets$ , $Zb\bar{b}/c\bar{c}$ and $t\bar{t}$ backgrounds (collectively, $\ell\ell jj$) contribute to the $4\ell$ signal region when jets in these events are either mismeasured as leptons or produce real leptons through secondary interactions. These processes are difficult to accurately simulate so we estimate their contribution from data. We assess $\ell\ell jj$ backgrounds using the ``fakeable object'' technique~\cite{fakeable}. For this method we define ``fakerates'' with respect to loosely identified lepton candidates, referred to as ``denominator objects''. Electron and muon denominator selections are defined in Table~\ref{tab:fo}. |
| 60 |
|
|
| 61 |
|
%------------------------------------------------- |
| 62 |
|
\begin{table}[htb] |
| 63 |
|
\begin{center} |
| 64 |
< |
\begin{tabular}{|c|c|c|c|} |
| 65 |
< |
\hline |
| 64 |
< |
\multicolumn{2}{|c|}{Electron} & \multicolumn{2}{|c|}{Muon} \\ |
| 64 |
> |
\begin{tabular}{c|c|c|c} |
| 65 |
> |
\multicolumn{2}{c|}{Electron} & \multicolumn{2}{|c}{Muon} \\ |
| 66 |
|
\hline |
| 67 |
|
variable & requirement & variable & requirement \\ |
| 68 |
|
\hline |
| 69 |
< |
$E_{T}$ & $> 7\rm~GeV$ & $p_{T}$ & $> 5\rm~GeV$ \\ |
| 70 |
< |
$|dz|$ & $< 0.1\rm~cm$ & type & $\rm Global~||~Tracker$ \\ |
| 71 |
< |
$|\eta|$ & $< 2.5\rm~GeV$ & $|d_{0}|$ & $< 2\rm~mm$ \\ |
| 72 |
< |
$H/E$ & $< 0.12(0.1) EB(EE)$ & $Iso^{pf}_{0.3}$ & $< 3\times p_{T}$ \\ |
| 73 |
< |
$iso_{trk}$ & $<0.3$ & ~ & ~ \\ |
| 74 |
< |
$iso_{em}$ & $<0.3$ & ~ & ~ \\ |
| 69 |
> |
$E_{T}$ & $> 7\rm~GeV$ & $p_{T}$ & $> 5\rm~GeV$ \\ |
| 70 |
> |
$|dz|$ & $< 0.1\rm~cm$ & type & $\rm Global~||~Tracker$ \\ |
| 71 |
> |
$|\eta|$ & $< 2.5\rm~GeV$ & $|d_{0}|$ & $< 2\rm~mm$ \\ |
| 72 |
> |
$H/E$ & $< 0.12(0.1) EB(EE)$ & $Iso^{pf}_{0.3}$ & $< 3\times p_{T}$ \\ |
| 73 |
> |
$iso_{trk}$ & $<0.3$ & ~ & ~ \\ |
| 74 |
> |
$iso_{em}$ & $<0.3$ & ~ & ~ \\ |
| 75 |
|
$iso_{had}$ & $<0.3$ & ~ & ~ \\ |
| 75 |
– |
\hline |
| 76 |
|
\end{tabular} |
| 77 |
|
\caption{Denominator Object Definitions}\label{tab:fo} |
| 78 |
|
\end{center} |
| 86 |
|
\begin{center} |
| 87 |
|
\includegraphics[width=0.45\linewidth]{figs/bdt-medium-frpt.png} |
| 88 |
|
\includegraphics[width=0.45\linewidth]{figs/frMu.png} |
| 89 |
< |
\caption{ {\bf Muon and Electron Fake Rates.}\label{fig:FR} } |
| 89 |
> |
\caption{ Muon and Electron Fake Rates.} |
| 90 |
> |
\label{fig:FR} |
| 91 |
|
\end{center} |
| 92 |
|
\end{figure} |
| 93 |
|
%------------------------------------------------- |
| 94 |
|
|
| 95 |
|
We estimate $\ell\ell jj$ backgrounds in the signal region by applying the fakerates in events that contain a good Z1. First, we select denominator objects that fail identification/isolation to prevent bias from real leptons. Next, we loop over pairs of the denominator objects, weight each leg with $\epsilon_{FR}(p_{T},\eta)/(1-\epsilon_{FR}(p_{T},\eta))$ and apply the Z2 kinematic requirements ($12\rm~GeV < m(Z2) < 120$). The denominator in the weight term accounts for the fact that the we only consider candidates that fail full lepton selection. Weighted pairs that pass the Z2 kinematic selection are summed to obtain an estimate of the $\ell\ell jj$ background. |
| 96 |
|
|
| 97 |
< |
Table~\ref{tab:fakes} presents $\ell\ell jj$ background estimates for the $2.1\rm~fb^{-1}$ dataset. We maximize the statistical power of the small $Z1 + \ge 2\rm~denominator$ sample by integrating over the flavor of the $Z1$ leptons and then dividing the $Z1$-inclusive prediction between the $4\ell_{e,\mu}$ and $2\ell_{e,\mu}2\ell_{\mu,e}$ channels. The division is performed by assuming equal $ee$ and $\mu\mu$ $Z1$ branching ratios and using an acceptance factor ($=XXX$, measured from inclusive $Z\rightarrow ee,\mu\mu$ yields) to account for efficiency differences in the detection of electrons and muons. |
| 97 |
> |
Table~\ref{tab:fakes} presents $\ell\ell jj$ background estimates for the $2.1\rm~fb^{-1}$ dataset. We maximize the statistical power of the small $Z1 + \ge 2\rm~denominator$ sample by integrating over the flavor of the $Z1$ leptons and then dividing the $Z1$-inclusive prediction between the $4\ell_{e,\mu}$ and $2\ell_{e,\mu}2\ell_{\mu,e}$ channels. The division is performed by assuming equal $ee$ and $\mu\mu$ $Z1$ branching ratios and using an acceptance factor ($=\sim1$, measured from inclusive $Z\rightarrow ee,\mu\mu$ yields {\bf need to double check this. 1 seems strange}) to account for efficiency differences in the detection of electrons and muons. |
| 98 |
|
|
| 99 |
|
%------------------------------------------------- |
| 100 |
|
\begin{table}[htb] |
| 104 |
|
\multicolumn{2}{c}{Z1-Inclusive $\ell\ell jj$ Yields} \\ |
| 105 |
|
\hline |
| 106 |
|
$Z1 + \mu\mu$ & $0.057 \pm X$ \\ |
| 107 |
< |
$Z1 + ee$ & $X \pm Y$ \\ |
| 107 |
> |
$Z1 + ee$ & $1.8 \pm 1$ \\ |
| 108 |
|
\hline |
| 109 |
|
\multicolumn{2}{c}{Per-Channel $\ell\ell jj$ Yields} \\ |
| 110 |
|
\hline |
| 111 |
|
$4\mu$ & $0.044 \pm X$ \\ |
| 112 |
< |
$4e$ & $X \pm Y$ \\ |
| 113 |
< |
$2e2\mu$ & $(0.013 + Z) \pm Y$ \\ |
| 112 |
> |
$4e$ & $0.4 \pm 0.3$ \\ |
| 113 |
> |
$2e2\mu$ & $(0.013 + 1.4) \pm 0.9$ \\ |
| 114 |
|
\hline |
| 115 |
|
\end{tabular} |
| 116 |
< |
\caption{{\bf Expected $ell\ell jj$ Events.}\label{tab:fakes}} |
| 116 |
> |
\caption{Expected $\ell\ell jj$ Events.} |
| 117 |
> |
\label{tab:fakes} |
| 118 |
|
\end{center} |
| 119 |
|
\end{table} |
| 120 |
|
%------------------------------------------------- |
| 121 |
|
|
| 122 |
< |
It is difficult to predict shapes for the $m(4\ell)$ distributions of $\ell\ell jj$ background with the limited number of events containing a good $Z1$ and two failing denominator objects. We increase sample size by loosening the denominator and $Z2$ selections. For the muon-channel, we relax the isolation requirement in the denominator definition and remove the opposite-sign requirement in the $Z2$ selection. For electrons we ... {\bf XXX}. With these modifications we obtain the $m(4\ell)$ distributions shown Figure~\ref{fig:mufakeshapes}. We fit the observed shapes with Landau distributions and obtain an acceptable goodness-of-fit. Consequently, we use Landau distributions to model the $m(4l)$ distribution of our $\ell\ell jj$ predictions, also shown in Figure~\ref{fig:mufakeshapes}. |
| 122 |
> |
It is difficult to predict $m(4\ell)$ and kinematic shapes for $\ell\ell jj$ background with the limited number of events containing a good $Z1$ and two failing denominator objects. Although loosening the denominator and $Z1,2$ selections helps, these requirements must not be made so loose that distributions from the control region no longer resemble those of the signal region. As an alternative, we study shapes using $Z+jets$, $t\bar{t}$ and $Zb\bar{b}$ MC events that pass our nominal selections. Figure~\ref{fig:fakeshapes}, for example, shows the cross section normalized $m(4\ell)$ distributions from these processes. |
| 123 |
|
|
| 124 |
|
%------------------------------------------------- |
| 125 |
|
\begin{figure}[tbp] |
| 126 |
|
\begin{center} |
| 127 |
|
\includegraphics[width=0.45\linewidth]{figs/muFakeShape-4m.png} |
| 128 |
< |
\includegraphics[width=0.45\linewidth]{figs/muFakeShape-2m.png} |
| 129 |
< |
\includegraphics[width=0.45\linewidth]{figs/muFakeShape-4m.png} |
| 128 |
< |
\includegraphics[width=0.45\linewidth]{figs/eleFakeShape-inclusive.png} |
| 129 |
< |
\caption{ Predicted $m(4\ell)$ Distributions for $\ell\ell jj$ Events.\label{fig:mufakeshapes} } |
| 128 |
> |
\caption{ Predicted $m(4\ell)$ Distributions for $\ell\ell jj$ Events. {\bf this is an old data-driven plot. Put the MC one here.}} |
| 129 |
> |
\label{fig:fakeshapes} |
| 130 |
|
\end{center} |
| 131 |
|
\end{figure} |
| 132 |
|
%------------------------------------------------- |
| 152 |
|
|
| 153 |
|
%This difference can be understood as a result of differences in the composition of the prediction sample (mainly light flavor) and that used to measure the fakerate (a mix of light and heavy flavor). |
| 154 |
|
|
| 155 |
< |
For electrons ... |
| 155 |
> |
For electrons, in order to isolate a pure sample of W+jets where the jet fakes an electron, a cross-flavor same-sign sample was chosen with a well-identified muon satisfying $iso_{PF}/P_{T} < 0.025$ and $p_{T}>25$. This region also contains Z+jets and dijet backgrounds. The W+jets and Z+jets components are represented by templates from Monte Carlo while the dijet is modelled by a Rayleigh distribution. The three are fitted to the observed same-sign events on the left of Figure~\ref{fig:ssEle}. The region with $MET>30~GeV$ is enriched in W+jets, so is chosen to test closure in the mass spectrum. The center of the same figure shows this observation compared to the fake prediction computed in the same way as in the muon case, and shows an overall under-prediction of $4\%$. The right of this figure shows that the observed shape discrepancy comes from an under-prediction of $33\%$ for $p_{T}<20$ and an over-prediction of $20\%$ for $p_{T}>20$. We thus take an overall systematic of $33\%$ on the fake prediction. |
| 156 |
> |
|
| 157 |
> |
An additional cross-check of the electron fake rates was performed on a sample of single-Z plus one fakeable object in both data and $Z+jets$ monte carlo. The largest observed discrepancy between observation and fake rate prediction of these four estimates was $20\%$, well within the systematic used above. |
| 158 |
|
%For electrons, charge misidentification is significant enough to result in a noticible Z-peak. The jet background is however easily estimated from a fit with a same-sign MC Z template and an exponential background PDF. Events selected in data are shown in Figures~\ref{fig:ssMuon} and (\ref{fig:ssEle}) as points. Table~\ref{tab:ssfakes} lists the total number of observed events in the muon-channel and the electron-channel background determined from the fit. |
| 159 |
|
|
| 160 |
|
%------------------------------------------------- |
| 162 |
|
\begin{center} |
| 163 |
|
\includegraphics[width=0.45\linewidth]{figs/ssEleMET.png} |
| 164 |
|
\includegraphics[width=0.45\linewidth]{figs/ssEleMZ1.png} |
| 165 |
< |
\caption{ Fakerate Predictions for Same-sign Electron Events.}\label{fig:ssEle} |
| 165 |
> |
\includegraphics[width=0.45\linewidth]{figs/ssElePtLoose.png} |
| 166 |
> |
\caption{ Fakerate Predictions for Same-sign Electron Events.} |
| 167 |
> |
\label{fig:ssEle} |
| 168 |
|
\end{center} |
| 169 |
|
\end{figure} |
| 170 |
|
%------------------------------------------------- |
| 171 |
|
|
| 172 |
< |
Table summarizes the results of this section. We take $47.2\%$ ($X\%$) as the systematic uncertainty on the muon (electron) fakerate to account for potential biases in our prediction due to differences in light flavor composition. |
| 172 |
> |
Table summarizes the results of this section. We take $47.2\%$ ($15\%$) as the systematic uncertainty on the muon (electron) fakerate to account for potential biases in our prediction due to differences in light flavor composition. |
| 173 |
|
|
| 174 |
|
%------------------------------------------------- |
| 175 |
|
\begin{table}[tbh] |
| 176 |
|
\begin{center} |
| 177 |
< |
\begin{tabular}{|c||c|c||c|} |
| 177 |
> |
\begin{tabular}{c|c|c|c} |
| 178 |
|
\hline |
| 179 |
< |
channel & observed & predicted & systematic \\ |
| 179 |
> |
channel & observed & predicted & systematic \\ |
| 180 |
|
\hline |
| 181 |
< |
\hline |
| 182 |
< |
$same sign~\mu\mu$ & $159$ & $108.04$ & $47.2\%$\\ |
| 183 |
< |
$same sign~ee$ & $X$ & $Y$ & $Z\%$ \\ |
| 181 |
> |
${\rm same~sign} \mu\mu$ & $159$ & $108.04$ & $47.2\%$\\ |
| 182 |
> |
${\rm same~sign} ee (total) $ & $5333$ & $5132$ & $33\%$ \\ |
| 183 |
> |
${\rm same~sign} ee (p_{T}<20)$ & $2783$ & $1993$ & $33\%$ \\ |
| 184 |
> |
${\rm same~sign} ee (p_{T}>20)$ & $1550$ & $3138$ & $20\%$ \\ |
| 185 |
|
\hline |
| 186 |
|
\end{tabular} |
| 187 |
< |
\caption{{\bf Same-sign Control Yields.}\label{tab:ssfakes}} |
| 187 |
> |
\caption{Same-sign Control Yields and Systematic} |
| 188 |
> |
\label{tab:ssfakes} |
| 189 |
|
\end{center} |
| 190 |
|
\end{table} |
| 191 |
|
%------------------------------------------------- |
| 193 |
|
%_________________________________________________________________ |
| 194 |
|
\subsubsection{Cross Check and Systematics : Heavy Flavor }\label{sec:hflavor} |
| 195 |
|
%_________________________________________________________________ |
| 196 |
< |
Backgrounds from $t\bar{t}$ and $Zb\bar{b}/c\bar{c}$ involve real leptons from heavy flavor decays. As with light flavor, a difference in the fraction of heavy flavor in the fakerate and prediction samples can lead to errors in estimation. We assess the potential impact of heavy flavor composition differences by applying our fakerate in a sample of relatively pure $Zb\bar{b}/c\bar{c}$ and $t\bar{t}$. |
| 196 |
> |
Backgrounds from $t\bar{t}$ and $Zb\bar{b}/c\bar{c}$ involve real leptons from heavy flavor decays. As with light flavor, a difference in the fraction of heavy flavor in the fakerate and prediction samples can lead to errors in signal region background estimation. We assess the impact of heavy flavor composition differences by applying our fakerate in a sample of relatively pure $Zb\bar{b}/c\bar{c}$ and $t\bar{t}$. |
| 197 |
|
|
| 198 |
< |
The control region consists of events that contain a pair of leptons passing the $Z1$ selection and at least two additional denominator objects with $\sigma_{IP_{3D}}/IP_{3D} > 4$. Denominators are defined according to the requirements of Table~\ref{tab:fo}. We make no requirement on denominator charge or flavor. The leftmost plot of Figure~\ref{fig:ZHF} compares the observed $m(Z1)$ distributions for events passing this selection in data with cross section normalized predictions from MC. We observe $71$ events and predict $66.3 \pm 2.0$ with $Zb\bar{b}$ and $t\bar{t}$ MC. Thus we confirm that the data sample is indeed dominated by heavy flavor. |
| 198 |
> |
The control region consists of events that contain a pair of leptons passing the $Z1$ selection and at least two additional denominator objects with $\sigma(IP_{3D})/IP_{3D} > 4$. Denominators are defined according to the requirements of Table~\ref{tab:fo}. We make no requirement on denominator charge or flavor. The leftmost plot of Figure~\ref{fig:ZHF} compares the observed $m(Z1)$ distributions for events passing this selection in data with cross section normalized predictions from MC. We observe $71$ events and predict $66.3 \pm 2.0$ with $Zb\bar{b}$ and $t\bar{t}$ MC, which confirms that the data sample is indeed dominated by heavy flavor. {\bf update numbers, they're like 62 and 58 now ...} |
| 199 |
|
|
| 200 |
|
Next, we require the high-IP denominator objects to additionally pass the more stringent lepton ID and isolation criteria used in our nominal Z2 selection. We estimate $0.81 \pm 0.21$ events from MC and observe 2. Electron and muon fakerates are then applied to the denominator objects in the original $71$ events and, following the procedures described in Section~\ref{sec:lflavor}, we predict $0.84 \pm 0.10$ events. Given the consistent results, we assign no additional systematic uncertainty on our predicted $\ell\ell jj$ background yields. |
| 201 |
|
|
| 211 |
|
\end{figure} |
| 212 |
|
%------------------------------------------------- |
| 213 |
|
|
| 214 |
< |
We determine a shape for heavy flavor background in the signal region from the distribution of $m(4\ell)$ from the $Z1 + 2\times$ denominator events. The rightmost plot of Figure~\ref{fig:ZHF} compares the $m(4\ell)$ distributions for this selection in data and (cross-section normalized) simulation. We fit both distributions with Landaus and compare the normalized PDFs in Figure~\ref{fig:HFshape}. |
| 214 |
> |
%We determine a shape for heavy flavor background in the signal region from the distribution of $m(4\ell)$ from the $Z1 + 2\times$ denominator events. The rightmost plot of Figure~\ref{fig:ZHF} compares the $m(4\ell)$ distributions for this selection in data and (cross-section normalized) simulation. We fit both distributions with Landaus and compare the normalized PDFs in Figure~\ref{fig:HFshape}. |
| 215 |
|
|
| 216 |
|
%------------------------------------------------- |
| 217 |
< |
\begin{figure}[htbp] |
| 218 |
< |
\begin{center} |
| 219 |
< |
\includegraphics[width=0.5\linewidth]{figs/HFshape.png} |
| 220 |
< |
\caption{$m(4\ell)$ shapes in the Heavy Flavor control region.}\label{fig:HFshape} |
| 221 |
< |
\end{center} |
| 222 |
< |
\end{figure} |
| 217 |
> |
%\begin{figure}[htbp] |
| 218 |
> |
%\begin{center} |
| 219 |
> |
%\includegraphics[width=0.5\linewidth]{figs/HFshape.png} |
| 220 |
> |
%\caption{$m(4\ell)$ shapes in the Heavy Flavor control region.}\label{fig:HFshape} |
| 221 |
> |
%\end{center} |
| 222 |
> |
%\end{figure} |
| 223 |
|
%------------------------------------------------- |
| 224 |
|
%_________________________________________________________________ |
| 225 |
|
\subsection{Cross Check and Systematics: $WZ$ } |
| 226 |
|
%_________________________________________________________________ |
| 227 |
< |
The estimate of $WZ$ background in Table~\ref{tab:EWK} is entirely MC-based. In addition to the leptons from $W$ and $Z$ decay, an additional ``fake'' lepton is needed for this process to contribute in the $4\ell$ signal region. We cross-check MC predictions with an estimate obtained from the fakeable object method. |
| 227 |
> |
The estimate of $WZ$ background in Table~\ref{tab:MCBG} is entirely MC-based. In addition to the leptons from $W$ and $Z$ decay, an additional ``fake'' lepton is needed for this process to contribute in the $4\ell$ signal region. We cross-check MC predictions with an estimate obtained from the fakeable object method. |
| 228 |
|
|
| 229 |
< |
We begin by requiring three fully selected leptons (two from the Z1 plus one additional) and $1+$ denominator objects. We then perform a single loop to associate the denominator objects with the third lepton. As before, we weight the denominators with $\epsilon_{FR}(p_{T},\eta)/(1-\epsilon_{FR}(p_{T},\eta))$, apply opposite-sign, same-flavor and kinematic selections and sum. The additional, ID'ed lepton with which the denominators are paired is either a fake (from $Z+jets$) or a real lepton (from $WZ$ or $ZZ$ where one of the leptons is not reconstructed). In order to extract the $WZ$ component of the measurement, we need to subtract off the $3\ell$ contribution predicted by MC for $ZZ$ as well as the double-fake estimate described in Section~\ref{sec:fakes}. The latter is double-counted when performing a single denominator loop. |
| 229 |
> |
We begin by requiring three fully selected leptons (two from the Z1 plus one additional) and $1+$ denominator objects. We then perform a single loop to associate the denominator objects with the third lepton. As before, we weight the denominators with $\epsilon_{FR}(p_{T},\eta)/(1-\epsilon_{FR}(p_{T},\eta))$, apply opposite-sign, same-flavor and kinematic selections and sum. The additional, identified lepton with which the denominators are paired is either a fake (from $Z+jets$) or a real lepton (from $WZ$, $Z\gamma$, or $ZZ$ where one of the leptons is not reconstructed). In order to extract the $WZ$ component of the measurement, we need to subtract off the $3\ell$ contribution predicted by MC for $ZZ$ and $Z\gamma$, as well as the double-fake estimate described in Section~\ref{sec:fakes}. The latter is double-counted when performing a single denominator loop. |
| 230 |
|
|
| 231 |
|
\begin{eqnarray} |
| 232 |
< |
N(WZ) &=& \ell\ell\ell~\Sigma_{i=0}^{Nd}~\frac{\epsilon(\eta^{i},p_{T}^{i})}{1-\epsilon(\eta^{i},p_{T}^{i})} - 2\times \ell\ell~\Sigma_{i=0}^{Nd}\Sigma_{j=i+1}^{Nd}~\frac{\epsilon(\eta^{i},p_{T}^{i})}{1-\epsilon(\eta^{i},p_{T}^{i})}~\frac{\epsilon(\eta^{j},p_{T}^{j})}{1-\epsilon(\eta^{j},p_{T}^{j})} - N(WZ) |
| 232 |
> |
N(WZ) &=& \ell\ell\ell~\Sigma_{i=0}^{Nd}~\frac{\epsilon(\eta^{i},p_{T}^{i})}{1-\epsilon(\eta^{i},p_{T}^{i})} \\ |
| 233 |
> |
~ &-& 2\times \ell\ell~\Sigma_{i=0}^{Nd}\Sigma_{j=i+1}^{Nd}~\frac{\epsilon(\eta^{i},p_{T}^{i})}{1-\epsilon(\eta^{i},p_{T}^{i})}~\frac{\epsilon(\eta^{j},p_{T}^{j})}{1-\epsilon(\eta^{j},p_{T}^{j})} \\ |
| 234 |
> |
~ &-& N(ZZ) |
| 235 |
|
\end{eqnarray} |
| 236 |
|
|
| 237 |
< |
Table~\ref{tab:} provides values for the terms in the equation above. The result XXX |
| 237 |
> |
Table~\ref{tab:WZfake} lists values for the terms in the equation above. The result ... {\bf fill in the table and quote a systematic for WZ}. |
| 238 |
|
|
| 239 |
|
%------------------------------------------------- |
| 240 |
|
\begin{table}[tbh] |
| 241 |
|
\begin{center} |
| 242 |
< |
\begin{tabular}{|c|c|c|} |
| 242 |
> |
\begin{tabular}{|c|c|c|c|} |
| 243 |
|
\hline |
| 244 |
< |
$4e$ & $4\mu$ & $2e2\mu$ \\ |
| 244 |
> |
& $4e$ & $4\mu$ & $2e2\mu$ \\ |
| 245 |
|
\hline |
| 246 |
< |
$X\pm Y$ & $Z\pm Y$ & $Z\pm Y$ \\ |
| 246 |
> |
single loop & $9.1\pm 2$ & $Z\pm Y$ & $6.6 + Z\pm 1.7 + Y$ \\ |
| 247 |
|
\hline |
| 248 |
< |
\end{tabular} |
| 249 |
< |
\caption{{\bf Data-driven Expected $WZ$ Yields.}\small{blah.}\label{tab:WZfake}} |
| 250 |
< |
\end{center} |
| 243 |
< |
\end{table} |
| 244 |
< |
%------------------------------------------------- |
| 245 |
< |
|
| 246 |
< |
|
| 247 |
< |
%_________________________________________________________________ |
| 248 |
< |
\subsubsection{Data-Driven Systematics Summary}\label{sec:fakesys} |
| 249 |
< |
%_________________________________________________________________ |
| 250 |
< |
|
| 251 |
< |
We summarize the systematic uncertainties on $\ell\ell jj$ and $\ell\ell\ell j$ background yields in Table~\ref{tab:fakessyssummary}. The relative yields differences discussed in Sections~\ref{sec:lflavor} and~\ref{sec:hflavor} are added in quadrature to address potential biases due to sample dependence. The difference in MC and data-driven predictions are assigned as a modeling uncertainty for $WZ$. |
| 252 |
< |
|
| 253 |
< |
%------------------------------------------------- |
| 254 |
< |
\begin{table}[tbh] |
| 255 |
< |
\begin{center} |
| 256 |
< |
\begin{tabular}{|c|c|c|} |
| 248 |
> |
double loop & $2.6\pm 0.5$ & $Z\pm Y$ & $2.2 + \pm 0.4 + Y$ \\ |
| 249 |
> |
\hline |
| 250 |
> |
$ZZ$ & $0.49\pm 0.005$ & $Z\pm Y$ & $0.67 + Z\pm 0.01 + Y$ \\ |
| 251 |
|
\hline |
| 252 |
< |
$4e$ & $4\mu$ & $2e2\mu$ \\ |
| 252 |
> |
$Z\gamma$ & $0.9\pm 0.4$ & $Z\pm Y$ & $0.8 + Z\pm 0.5 + Y$ \\ |
| 253 |
|
\hline |
| 254 |
< |
$X\pm Y$ & $Z\pm Y$ & $Z\pm Y$ \\ |
| 254 |
> |
WZ estimate & $2.5\pm 2.1$ & $Z\pm Y$ & $0.7 + Z\pm 1.8 + Y$ \\ |
| 255 |
|
\hline |
| 256 |
|
\end{tabular} |
| 257 |
< |
\caption{{\bf Data-driven Expected $WZ$ Yields.}\small{blah.}\label{tab:fakesyssummary}} |
| 257 |
> |
\caption{Data-driven Expected $WZ$ Yields} |
| 258 |
> |
\label{tab:WZfake} |
| 259 |
|
\end{center} |
| 260 |
|
\end{table} |
| 261 |
|
%------------------------------------------------- |
| 262 |
|
|
| 263 |
< |
The central shapes used for the $\ell\ell jj$ backgrounds are the Landaus fit to the $1.6\rm~fb^{-1}$ predictions. Our alternate shapes are the high-statistics Landaus of Section~\ref{sec:} and the distributions from the high-IP control region. We take the larger of the differences between the central shapes and the alternatives as a shape systematic. Figure~\ref{fig:fakeshapesummary} shows the central shape and corresponding uncertainty envelope. |
| 269 |
< |
We take the high-statistics distribution as our are included in our evaluation of systematic uncertainties (Section~\ref{sec:fakesys}). |
| 270 |
< |
%------------------------------------------------- |
| 271 |
< |
\begin{figure}[tbp] |
| 272 |
< |
\begin{center} |
| 273 |
< |
\includegraphics[width=0.5\linewidth]{figs/m4l-HF.png} |
| 274 |
< |
\caption{Heavy Flavor $m(4\ell)$ Shape.\label{fig:fakeshapesummary} } |
| 275 |
< |
\end{center} |
| 276 |
< |
\end{figure} |
| 277 |
< |
%------------------------------------------------- |
| 263 |
> |
|
