| 1 |
\setcounter{table}{0}
|
| 2 |
\renewcommand\thetable{\Alph{section}.\arabic{table}}
|
| 3 |
\setcounter{figure}{0}
|
| 4 |
\renewcommand\thefigure{\Alph{section}.\arabic{figure}}
|
| 5 |
\section{Trigger Efficiency Model}
|
| 6 |
\label{sec:appendix_trigger}
|
| 7 |
|
| 8 |
As described in Section~\ref{sec:trigSel} we rely on a
|
| 9 |
mixture of single and double lepton triggers. The trigger
|
| 10 |
efficiency is very high because for most of the phase space
|
| 11 |
we have two leptons each of which can fire a single lepton
|
| 12 |
trigger -- and the single lepton triggers are very efficient.
|
| 13 |
|
| 14 |
We apply to MC events a simplified model of the trigger efficiency
|
| 15 |
as a function of dilepton species ($ee$, $e\mu$, $\mu\mu$), the $P_T$
|
| 16 |
of the individual leptons, and, in the case of muons, the $|\eta|$
|
| 17 |
of the muons. We believe that this model is adequate for
|
| 18 |
the trigger efficiency precision needed for this analysis.
|
| 19 |
|
| 20 |
The model assumptions are the following:
|
| 21 |
|
| 22 |
\begin{itemize}
|
| 23 |
|
| 24 |
\item Muon and electron trigger turn-ons as a function of $P_T$
|
| 25 |
are infinitely sharp. %{\color{red} Can we add references?}
|
| 26 |
|
| 27 |
\item All electron triggers with no ID have essentially 100\%
|
| 28 |
efficiency for electrons passing our analysis cuts\cite{ref:evans}.
|
| 29 |
|
| 30 |
\item Electron triggers with (Tight(er))CaloEleId have 100\%
|
| 31 |
efficiency with respect to our offline selection. This we
|
| 32 |
verified via tag-and-probe on $Z\to ee$.
|
| 33 |
|
| 34 |
\item Electron triggers with EleId have somewhat lower
|
| 35 |
efficiency. This was also measured by tag-and-probe.
|
| 36 |
|
| 37 |
\item The single muon trigger has 40\% efficiency for
|
| 38 |
$|\eta|>2.1$~\cite{ref:evans}.
|
| 39 |
|
| 40 |
\item If a muon fails the single muon trigger, it
|
| 41 |
will also fail the double muon trigger. This is actually
|
| 42 |
a conservative assumption. See Section~\ref{sec:emutrg}
|
| 43 |
for a discussion of $e\mu$ triggers.
|
| 44 |
|
| 45 |
\item The double muon trigger has efficiency
|
| 46 |
equal to the square of the single muon efficiency. This is
|
| 47 |
also a conservative assumption.
|
| 48 |
|
| 49 |
\item The $e\mu$ triggers have no efficiency if the muon has $|\eta|>2.1$.
|
| 50 |
Again, this is conservative.
|
| 51 |
\end{itemize}
|
| 52 |
|
| 53 |
The model also uses some luminosity fractions and some trigger
|
| 54 |
efficiencies.
|
| 55 |
|
| 56 |
\begin{itemize}
|
| 57 |
|
| 58 |
\item $\epsilon_{\mu}$=93\%, the single muon trigger efficiency plateau
|
| 59 |
for $|\eta|<2.1$~\cite{ref:evans};
|
| 60 |
|
| 61 |
\item $\epsilon'_{\mu}$=40\%, the single muon trigger efficiency plateau
|
| 62 |
for $|\eta|>2.1$~\cite{ref:evans};
|
| 63 |
|
| 64 |
\item $f9$=0.215: fraction of data with the Mu9 trigger unprescaled.
|
| 65 |
(run$\le 147116$).
|
| 66 |
|
| 67 |
\item $f11$=0.273 fraction of data with the Mu9 trigger prescaled and
|
| 68 |
the Mu11 trigger unprescaled.
|
| 69 |
(147196 $\leq$ run $\leq$ 148058).
|
| 70 |
|
| 71 |
\item $e10$=0.002: fraction of data with the 10 GeV unprescaled electron triggers.
|
| 72 |
(run$\le 139980$).
|
| 73 |
|
| 74 |
\item $e15$=0.086: fraction of data with the 15 GeV unprescaled electron triggers.
|
| 75 |
(139980 $<$ run $\leq$ 144114).
|
| 76 |
|
| 77 |
\item $e17$=0.127: fraction of data with the 100\% efficient 17 GeV unprescaled electron triggers.
|
| 78 |
(144114 $<$ run $\leq$ 147116).
|
| 79 |
|
| 80 |
\item $e17b$=0.273: fraction of data with 17 GeV unprescaled electron triggers
|
| 81 |
with efficiency $\epsilon_e^b=90\%$ (as measured by tag-and-probe).
|
| 82 |
(147116 $<$ run $\leq$ 148058).
|
| 83 |
|
| 84 |
\item $emess$=0.512: the remainder of the run with several different electron
|
| 85 |
triggers, all of $P_T>17$ GeV. For this period we measure the
|
| 86 |
luminosity-weighted
|
| 87 |
trigger efficiency $\epsilon(P_T)$ via tag and probe to be 99\%
|
| 88 |
($17<P_T<22$, 97\% ($22<P_T<27$), 98\% ($27<P_T<32$) and
|
| 89 |
100\% ($P_T>32$).
|
| 90 |
|
| 91 |
\end{itemize}
|
| 92 |
|
| 93 |
The full trigger efficiency model is described separately for
|
| 94 |
$ee$, $e\mu$, and $\mu\mu$.
|
| 95 |
|
| 96 |
\subsection{$ee$ efficiency model}
|
| 97 |
\label{sec:eemodel}
|
| 98 |
|
| 99 |
This is the easiest. Throughout the 2010 run we have always
|
| 100 |
had dielectron triggers with thresholds lower than our (20,10)
|
| 101 |
analysis thresholds. Since electron triggers are very close
|
| 102 |
to 100\% efficient\cite{ref:evans},
|
| 103 |
the trigger efficiency for $ee$ is 100\%. We have verified that
|
| 104 |
the efficiency of the dielectron trigger is 100\% with respect
|
| 105 |
to the single electron trigger using $Z \to ee$ data.
|
| 106 |
|
| 107 |
\subsection{$\mu\mu$ efficiency model}
|
| 108 |
\label{sec:mmmodel}
|
| 109 |
|
| 110 |
We consider different cases.
|
| 111 |
|
| 112 |
\subsubsection{Both muons in $|\eta|<2.1$ and with $P_T>15$ GeV}
|
| 113 |
This is the bulk of the $\mu\mu$.
|
| 114 |
|
| 115 |
\begin{center}
|
| 116 |
$\epsilon = 1 - (1-\epsilon_{\mu})^2$
|
| 117 |
\end{center}
|
| 118 |
|
| 119 |
\subsubsection{Both muons in $|\eta|<2.1$, one muon with $11<P_T<15$ GeV}
|
| 120 |
In this case there must be a muon with $P_T>20$ GeV. The single muon
|
| 121 |
trigger is operative for the full dataset on this muon. Some loss
|
| 122 |
of efficiency can be recovered when the 2nd muon fires the trigger.
|
| 123 |
But this can happen only for a fraction of the run. The dimuon trigger
|
| 124 |
cannot fire in our model to recover any of the efficiency lost by
|
| 125 |
the single muon trigger on the high $P_T$ muon.
|
| 126 |
|
| 127 |
\begin{center}
|
| 128 |
$\epsilon = \epsilon_{\mu} + (1-\epsilon_{\mu})\epsilon_{\mu}(f9+f11)$
|
| 129 |
\end{center}
|
| 130 |
|
| 131 |
\subsubsection{Both muons in $|\eta|<2.1$, one muon with $10<P_T<11$ GeV}
|
| 132 |
Same basic idea as above.
|
| 133 |
|
| 134 |
\begin{center}
|
| 135 |
$\epsilon = \epsilon_{\mu} + (1-\epsilon_{\mu})\epsilon_{\mu}f9$
|
| 136 |
\end{center}
|
| 137 |
|
| 138 |
\subsubsection{Both muons with $|\eta|>2.1$}
|
| 139 |
This is a very small fraction of events.
|
| 140 |
%In our model they can only be triggered by the dimuon trigger.
|
| 141 |
|
| 142 |
\begin{center}
|
| 143 |
$\epsilon = \epsilon_{\mu}^2 + \alpha (1-\epsilon_{\mu}) \epsilon'_{\mu}$
|
| 144 |
\end{center}
|
| 145 |
|
| 146 |
\noindent where $\alpha=2$ if both muons are above 15 GeV, $\alpha=(1+f9+f11)$ if
|
| 147 |
one of the muons is between 11 and 15 GeV, and $\alpha=(1+f9)$ if one of the muon
|
| 148 |
is below 11 GeV.
|
| 149 |
|
| 150 |
\subsubsection{First muon with $P_T>15$ and $|\eta|<2.1$; second muon
|
| 151 |
with $|\eta|>2.1$}
|
| 152 |
The single muon trigger is always operative. If it fails the double muon
|
| 153 |
trigger also fails.
|
| 154 |
|
| 155 |
\begin{center}
|
| 156 |
$\epsilon = \epsilon_{\mu} + (1-\epsilon_{\mu})\Delta_{\mu}$
|
| 157 |
\end{center}
|
| 158 |
|
| 159 |
\noindent where
|
| 160 |
|
| 161 |
\begin{center}
|
| 162 |
$\Delta_{\mu} = \epsilon'_{\mu}$ ~~~~(2nd muon with $P_T \geq 15$ GeV) \\
|
| 163 |
$\Delta_{\mu} = (f9+f11)\epsilon'_{\mu}$ ~~~~(2nd muon with $11 \leq P_T < 15$ GeV) \\
|
| 164 |
$\Delta_{\mu} = f9\epsilon'_{\mu}$ ~~~~(2nd muon with $9 \leq P_T < 11$ GeV) \\
|
| 165 |
\end{center}
|
| 166 |
|
| 167 |
|
| 168 |
\subsubsection{First muon with $11<P_T<15$ and $|\eta|<2.1$; second muon
|
| 169 |
with $|\eta|>2.1$ and $P_T>20$}
|
| 170 |
The single muon trigger at low $\eta$ is fully operative only for a fraction of the run,
|
| 171 |
this efficiency is captured by the first term below.
|
| 172 |
For the remaining fraction, we rely on the double muon trigger as well as the
|
| 173 |
single muon trigger at high $\eta$ (2nd term in the equation).
|
| 174 |
|
| 175 |
\begin{center}
|
| 176 |
$\epsilon = (f9+f11)(\epsilon_{\mu} + (1-\epsilon_{\mu})\epsilon'_{\mu})
|
| 177 |
+ (1-f9-f11)(\epsilon_{\mu}^2 + (1-\epsilon_{\mu})\epsilon'_{\mu})$
|
| 178 |
\end{center}
|
| 179 |
|
| 180 |
\noindent which reduces to
|
| 181 |
|
| 182 |
\begin{center}
|
| 183 |
$\epsilon = (f9+f11)\epsilon_{\mu} + (1-f9-f11)\epsilon_{\mu}^2
|
| 184 |
+ (1-\epsilon_{\mu})\epsilon'_{\mu}$
|
| 185 |
\end{center}
|
| 186 |
|
| 187 |
\subsubsection{First muon with $10<P_T<11$ and $|\eta|<2.1$; second muon
|
| 188 |
with $|\eta|>2.1$ and $P_T>20$}
|
| 189 |
Same basic idea as above.
|
| 190 |
|
| 191 |
\begin{center}
|
| 192 |
% $\epsilon = f9~\epsilon_{\mu} + (1-f9)\epsilon_{\mu}^2$
|
| 193 |
$\epsilon = f9\epsilon_{\mu} + (1-f9)\epsilon_{\mu}^2
|
| 194 |
+ (1-\epsilon_{\mu})\epsilon'_{\mu}$
|
| 195 |
\end{center}
|
| 196 |
|
| 197 |
\subsection{$e\mu$ efficiency model}
|
| 198 |
\label{sec:emumodel}
|
| 199 |
|
| 200 |
This is the most complicated case. The idea is that the muon trigger
|
| 201 |
is used to get the bulk of the efficiency. Then the single electron
|
| 202 |
trigger(s) and the $e\mu$ triggers are used to get back some of the
|
| 203 |
efficiency loss. The various cases are listed below.
|
| 204 |
|
| 205 |
\subsubsection{Muon with $|\eta|<2.1$ and $P_T>15$}
|
| 206 |
This is the bulk of the acceptance.
|
| 207 |
|
| 208 |
\begin{center}
|
| 209 |
$\epsilon = \epsilon_{\mu} + (1-\epsilon_{\mu})\Delta_1$
|
| 210 |
\end{center}
|
| 211 |
|
| 212 |
where $\Delta_1$ is the efficiency from the electron trigger:
|
| 213 |
\begin{itemize}
|
| 214 |
\item $P_T(ele)<15 \to \Delta_1=e10$
|
| 215 |
\item $15<P_T(ele)<17 \to \Delta_1=e10+e15$
|
| 216 |
\item $P_T(ele)>15 \to \Delta_1=e10+e15+e17+\epsilon_e^b~e17b+\epsilon(P_T)~emess$
|
| 217 |
\end{itemize}
|
| 218 |
|
| 219 |
|
| 220 |
\subsubsection{Muon with $|\eta|<2.1$ and $11<P_T<15$}
|
| 221 |
|
| 222 |
This is the similar to the previous case, except that the muon
|
| 223 |
trigger is operative only for a subset of the data taking period.
|
| 224 |
|
| 225 |
\begin{center}
|
| 226 |
$\epsilon = (f11+f9)\epsilon_{\mu} + \Delta_2 + \Delta_3$
|
| 227 |
\end{center}
|
| 228 |
|
| 229 |
Here $\Delta_2$ is associated with the period where the muon
|
| 230 |
trigger was at 15 GeV, in which case we use electron triggers or
|
| 231 |
$e\mu$ triggers. Note that the electron in this case must be
|
| 232 |
above 20 GeV. This can happen only in the latter part of the run,
|
| 233 |
thus we write
|
| 234 |
\begin{center}
|
| 235 |
$\Delta_2 = (1-f11-f9)~(\epsilon_{\mu}~+~
|
| 236 |
(1-\epsilon_{\mu})\epsilon(P_T))$
|
| 237 |
\end{center}
|
| 238 |
\noindent where the first term is for the $e\mu$ trigger and the
|
| 239 |
second term corresponds to $e\mu$ trigger failures, in which case we have
|
| 240 |
to rely on the electron trigger.
|
| 241 |
|
| 242 |
Then, $\Delta_3$ is associated with muon trigger failures in early runs,
|
| 243 |
{\em i.e.}, run $<148819$. In this case the electron trigger picks it
|
| 244 |
up and the $e\mu$ trigger does not help.
|
| 245 |
|
| 246 |
\begin{center}
|
| 247 |
$\Delta_3 = (f11+f9)(1-\epsilon_{\mu}) \cdot \epsilon_e$
|
| 248 |
\end{center}
|
| 249 |
|
| 250 |
\noindent where $\epsilon_e$ is the efficiency of the electron
|
| 251 |
trigger for $P_T>20$. This is 100\% up to run 147716 (fraction
|
| 252 |
$(e10+e15+e17)/(f11+f9)$; then it is somewhat lower up to
|
| 253 |
run 148058, then it becomes very close to 100\% again.
|
| 254 |
For this latter part of the run we approximate it as $\epsilon_e^b$.
|
| 255 |
Thus:
|
| 256 |
|
| 257 |
\begin{center}
|
| 258 |
$\epsilon_e = (e10+e15+e17)/(f11+f9) +
|
| 259 |
\epsilon_e^b(f11+f9-e10-e15-e17)/(f11+f9)$
|
| 260 |
\end{center}
|
| 261 |
|
| 262 |
\subsubsection{Muon with $|\eta|<2.1$ and $9<P_T<11$}
|
| 263 |
|
| 264 |
Identical to the previous case, but replace $(f11+f9)$ with $f9$ everywhere.
|
| 265 |
|
| 266 |
\subsubsection{Muon with $|\eta|>2.1$}
|
| 267 |
|
| 268 |
This is a 10\% effect to start with.
|
| 269 |
The first term is from the electron efficiency. The 2nd term is the correction
|
| 270 |
due to the single muon efficieny.
|
| 271 |
|
| 272 |
\begin{center}
|
| 273 |
$\epsilon = \Delta_1 + (1-\Delta_1)\Delta_{\mu}$
|
| 274 |
\end{center}
|
| 275 |
|
| 276 |
\subsection{Summary of the trigger efficiency model}
|
| 277 |
\label{sec:trgeffsum}
|
| 278 |
|
| 279 |
We take the trigger efficiency for $ee$ as 100\%. The trigger efficiency
|
| 280 |
for the $e\mu$ and $\mu\mu$ final states is summarized in
|
| 281 |
Figures~\ref{fig:emuModel} and~\ref{fig:mumuModel}.
|
| 282 |
We estimate the systematic uncertainties on the trigger modeling
|
| 283 |
to be at the few percent level.
|
| 284 |
|
| 285 |
\begin{figure}[htb]
|
| 286 |
\begin{center}
|
| 287 |
\includegraphics[width=0.99\linewidth]{emuModel.png}
|
| 288 |
\caption{\label{fig:emuModel}\protect Trigger efficiency for the
|
| 289 |
$e\mu$ pair as a function of the $P_T$ of the two leptons.
|
| 290 |
The top table corresponds to $|\eta(\mu)| < 2.1$, the bottom
|
| 291 |
table to $|\eta(\mu)| > 2.1$.}
|
| 292 |
\end{center}
|
| 293 |
\end{figure}
|
| 294 |
\clearpage
|
| 295 |
|
| 296 |
|
| 297 |
\begin{figure}[tbh]
|
| 298 |
\begin{center}
|
| 299 |
\includegraphics[width=0.99\linewidth]{mumuModel.png}
|
| 300 |
\includegraphics[width=0.99\linewidth]{mumu24Model.png}
|
| 301 |
\caption{\label{fig:mumuModel}\protect Trigger efficiency for the
|
| 302 |
$\mu\mu$ pair as a function of the $P_T$ of the two muons.
|
| 303 |
The top table corresponds to both muons having $|\eta| < 2.1$;
|
| 304 |
the middle table has one of the muon with $|\eta|<2.1$ and the
|
| 305 |
other muon with $|\eta|>2.1$; the bottom table has both muons with
|
| 306 |
have $|\eta|>2.1$.}
|
| 307 |
\end{center}
|
| 308 |
\end{figure}
|
| 309 |
|
| 310 |
\clearpage
|
| 311 |
|
| 312 |
\subsection{$e\mu$ trigger study}
|
| 313 |
\label{sec:emutrg}
|
| 314 |
The $e\mu$ cross triggers were introduced late in the run and are
|
| 315 |
not well studied (as far as we know). We performed a study to verify
|
| 316 |
the key assumptions of the model:
|
| 317 |
\begin{enumerate}
|
| 318 |
\item the $\mu$ component of the $e\mu$ trigger is at least as efficient
|
| 319 |
as the single muon trigger;
|
| 320 |
\item the electron component of the $e\mu$ trigger is essentially 100\% efficient,
|
| 321 |
as already determined by tag and probe.
|
| 322 |
\end{enumerate}
|
| 323 |
|
| 324 |
To this end, we select $e\mu$ events in the run ranges were the
|
| 325 |
$e\mu$ triggers were operational as follows:
|
| 326 |
\begin{itemize}
|
| 327 |
\item An offline muon of $P_T > 10$ GeV satisfying the standard
|
| 328 |
requireemnts.
|
| 329 |
\item This muon must have fired one of the unprescaled single muon triggers.
|
| 330 |
\item An offline electron of $P_T > 20$ GeV satisfying the standard requirements.
|
| 331 |
%\item This electron must have fired one of the unprescaled electron triggers,
|
| 332 |
%which are known to be essentially 100\% efficient on real electrons
|
| 333 |
%from tag \& probe studies.
|
| 334 |
\end{itemize}
|
| 335 |
|
| 336 |
|
| 337 |
|
| 338 |
\begin{figure}[hbt]
|
| 339 |
\begin{center}
|
| 340 |
\includegraphics[width=\linewidth]{emu_trigger.png}
|
| 341 |
\caption{\label{fig:emutrg}\protect $e\mu$ trigger efficiency
|
| 342 |
as a function of run number. We plot the trigger
|
| 343 |
efficiency of a given variant of the $e\mu$ trigger only for
|
| 344 |
the runs when said trigger was enabled.}
|
| 345 |
\end{center}
|
| 346 |
\end{figure}
|
| 347 |
|
| 348 |
|
| 349 |
We then defined the $e\mu$ trigger efficiency as the probability
|
| 350 |
for $e\mu$ trigger to fire on these events.
|
| 351 |
The results are shown in Figure~\ref{fig:emutrg}. We find 100\%
|
| 352 |
trigger efficiency (as expected) for {\tt Mu5\_Ele9}, {\tt Mu5\_Ele13},
|
| 353 |
and {\tt Mu5\_Ele17\_V2}. The other triggers ({\tt Mu5\_Ele5},
|
| 354 |
{\tt Mu8\_Ele8}, and {\tt Mu5\_Ele17\_V1}) do not work very well.
|
| 355 |
We note that the working triggers are seeded at L1 by both Mu and EG.
|
| 356 |
On the other hand, {\tt Mu5\_Ele5} and {\tt Mu8\_Ele8} are only seeded
|
| 357 |
by Mu. We have since learned\cite{ref:harper} that the inefficiency
|
| 358 |
of these triggers is due to an HLT bug that has recently been understood.
|
| 359 |
The source of inefficiency of {\tt Mu5\_Ele17\_V1}
|
| 360 |
is not understood
|
| 361 |
at the moment. Nevertheless, the working $e\mu$ triggers safely
|
| 362 |
cover the run ranges with the problematic $e\mu$ triggers.
|
| 363 |
|
| 364 |
Thus, our key trigger model assumptions are verified, {\em i.e.}, the
|
| 365 |
muon trigger piece of the $e\mu$ trigger is at least as efficient
|
| 366 |
as the single muon trigger, and the electron piece is essentially
|
| 367 |
100\% efficient.
|
| 368 |
|
| 369 |
%Note that if we calculate the efficiency
|
| 370 |
%relaxing the requirement on the single electron trigger, we find
|
| 371 |
%$e\mu$ trigger efficiencies of order 90\% instead of 100\% (for the
|
| 372 |
%working set of triggers). This must be due to the fact that we
|
| 373 |
%are measuring the efficiency on ``junk'' electrons, and the
|
| 374 |
%electron trigger efficiency on these electrons is not exactly
|
| 375 |
%100\% (the tag \& probe efficiency measurement on electrons from
|
| 376 |
%$Z \to ee$ gives essentially 100\%). |
