COMP/CSA06DOC/zee.tex

\subsubsection{Reconstruction of Z$\rightarrow e^+e^-$ events}

This exercise has multiple goals: one is the selection of
Z$\rightarrow e^{+} e^{-}$ out of the ``Electroweak Soup'' (EWK) and their
use to evaluate the performance of the electron reconstruction;
and another one is the demonstration of the workflow for running offline
ECAL calibration at a remote Tier-2 centre, starting from an AlcaReco
stream produced at the Tier-1 centre. 

For this reason a special skim filter job has been designed, producing as output two streams for the same events: one is the standard electron AlcaReco stream, as described in section \ref{calibtool}, the other one is the standard RECOSIM output.
The events are selected on the basis of Monte Carlo generator
information, requiring only Drell-Yan $e^+e^-$ events in a mass
range from 50 to 130 GeV/c$^2$. The electrons are also required to
have $p_T$ greater than 5 GeV/c and $|\eta|<2.7$. 
This selection has an efficiency over the Electroweak Soup of 3.7$\%$.

Skim jobs were processed at the CNAF Tier-1, starting from less than 3
million EWK soup events (not the full EWK dataset have been processed
using the CMSSW\_1\_0\_5 release). Out of these events, a sample of
76494 events have been selected and the two created streams were
transferred to the Rome Tier-2. The total amount of data amounts to
about 33.5GB, of which around 500MB are occupied by the AlcaReco
electron stream.  
Analysis jobs are sent to the Rome Tier-2 using CRAB\_1\_3\_0 and CMSSW\_1\_0\_5 on both the RECOSIM and the AlcaReco stream. 

The main purpose is to validate the electron reconstruction, evaluating the electron reconstruction performance.
The electron reconstruction efficiency both versus $p_T$ and $\eta$ is shown in Fig.~\ref{fig:zee-eleeff}.

\begin{figure}[!hbtp]
  \centering
  {\includegraphics[width=0.49\textwidth]{figs/zee-eleeffpt.pdf}} 
  {\includegraphics[width=0.49\textwidth]{figs/zee-eleeffeta.pdf}} 
  \caption{ (left) Electron reconstruction efficiency as a function of $p_T$ and (right) $\eta$.\label{fig:zee-eleeff} }

\end{figure}

The two main ingredients for the electron reconstruction, the
supercluster and the electron track, are then evaluated separately. 
The ratio of reconstructed supercluster energy over the true energy as
a function of $\eta$ is visible in Fig.~\ref{fig:zee-erecetrue},
showing a well-known problem in the CMSSW\_1\_0\_5 release for what
concerns the endcap supercluster reconstruction. The projected
distribution for the barrel is presented in the right plot of
Fig.~\ref{fig:zee-erecetrue}. 

\begin{figure}[!hbtp]
  \centering
  {\includegraphics[width=0.49\textwidth]{figs/zee-erecetrueeta.pdf}} 
  {\includegraphics[width=0.49\textwidth]{figs/zee-erecetruebarrel.pdf}} 
  \caption{ (left) Distribution of the electron reconstructed supercluster energy over the true energy as a function of $\eta$ and (right) the corresponding distribution in the ECAL barrel acceptance.  \label{fig:zee-erecetrue} }

\end{figure}

The quality of the track reconstruction is evaluated looking at the E/p distribution, using the track parameters both at the vertex and at the outermost state, as it can be seen in Fig.~\ref{fig:zee-eoverp}. 

\begin{figure}[!hbtp]
  \centering
  {\includegraphics[width=0.49\textwidth]{figs/zee-eoverpvertex.pdf}} 
  {\includegraphics[width=0.49\textwidth]{figs/zee-eoverpoutermost.pdf}} 
  \caption{Distribution of the E/p variable using track parameters at the (left) vertex (right) outermost state. \label{fig:zee-eoverp} }

\end{figure}

At the time of the CMSSW\_1\_0\_5, the reconstructed electron track
was not the track after the smoothing step, and for this reason the
quality of the E/p distribution with the track parameters at the
vertex is rather poor. A refitting of the electron track has been
tried in the offline analysis, improving the quality of the E/p
matching, as it visible in Fig.~\ref{fig:zee-refittedeoverp}. 

\begin{figure}[htbp]
  \centering
  \includegraphics[width=0.6\textwidth]{figs/zee-refittedeoverp.pdf}
  \caption{Distribution of the E/p variable using refitted track parameters at the vertex. \label{fig:zee-refittedeoverp} }

\end{figure}

The invariant mass computed from the electron pair nearest two the nominal Z mass is presented in Fig.~\ref{fig:zee-invariantmass}.
\begin{figure}[h]
  \centering
    \includegraphics[width=0.6\textwidth]{figs/zee-invariantmass.pdf}
  \caption{Electron pair invariant mass in  Z$\rightarrow e^{+} e^{-}$ events. \label{fig:zee-invariantmass}}

\end{figure}

The distribution of the difference between the reconstructed mass and
the generated mass is shown in
Fig.~\ref{fig:zee-relativemassdifference}, where the right plot
displays the difference as a fuction of $\eta$. The distribution shows
the expected 
behaviour due to the effect of the increasing tracker material towards
$\eta$ over the electron reconstruction.  

\begin{figure}[!hbtp]
  \centering
  {\includegraphics[width=0.49\textwidth]{figs/zee-relativemassdifference.pdf}} 
  {\includegraphics[width=0.49\textwidth]{figs/zee-relativemassdifferenceeta.pdf}} 
  \caption{ (left) Relative difference between the reconstructed and the generated mass in  Z$\rightarrow e^{+} e^{-}$ events and (right) its bahaviour as a function of $\eta$. \label{fig:zee-relativemassdifference} }

\end{figure}

%It can be argued that 
Using the corrections present in CMSSW\_1\_0\_5, the invariant mass peak is about 1.4$\%$ off, while the Z invariant mass resolution is about 1.8$\%$, 


Revision:	1.3
Committed:	Sun Jan 28 19:14:17 2007 UTC (18 years, 3 months ago) by acosta
Content type:	application/x-tex
Branch:	MAIN
CVS Tags:	HEAD
Changes since 1.2:	+42 -11 lines
Log Message:	major edits from DA
#	User	Rev	Content
1	meridian	1.1	\subsubsection{Reconstruction of Z$\rightarrow e^+e^-$ events}
2
3	acosta	1.3	This exercise has multiple goals: one is the selection of
4			Z$\rightarrow e^{+} e^{-}$ out of the ``Electroweak Soup'' (EWK) and their
5			use to evaluate the performance of the electron reconstruction;
6			and another one is the demonstration of the workflow for running offline
7			ECAL calibration at a remote Tier-2 centre, starting from an AlcaReco
8			stream produced at the Tier-1 centre.
9	meridian	1.1
10			For this reason a special skim filter job has been designed, producing as output two streams for the same events: one is the standard electron AlcaReco stream, as described in section \ref{calibtool}, the other one is the standard RECOSIM output.
11	acosta	1.3	The events are selected on the basis of Monte Carlo generator
12			information, requiring only Drell-Yan $e^+e^-$ events in a mass
13			range from 50 to 130 GeV/c$^2$. The electrons are also required to
14			have $p_T$ greater than 5 GeV/c and $\|\eta\|<2.7$.
15			This selection has an efficiency over the Electroweak Soup of 3.7$\%$.
16
17			Skim jobs were processed at the CNAF Tier-1, starting from less than 3
18			million EWK soup events (not the full EWK dataset have been processed
19			using the CMSSW\_1\_0\_5 release). Out of these events, a sample of
20			76494 events have been selected and the two created streams were
21			transferred to the Rome Tier-2. The total amount of data amounts to
22			about 33.5GB, of which around 500MB are occupied by the AlcaReco
23			electron stream.
24	acosta	1.2	Analysis jobs are sent to the Rome Tier-2 using CRAB\_1\_3\_0 and CMSSW\_1\_0\_5 on both the RECOSIM and the AlcaReco stream.
25	meridian	1.1
26	acosta	1.3	The main purpose is to validate the electron reconstruction, evaluating the electron reconstruction performance.
27	meridian	1.1	The electron reconstruction efficiency both versus $p_T$ and $\eta$ is shown in Fig.~\ref{fig:zee-eleeff}.
28
29			\begin{figure}[!hbtp]
30			\centering
31			{\includegraphics[width=0.49\textwidth]{figs/zee-eleeffpt.pdf}}
32			{\includegraphics[width=0.49\textwidth]{figs/zee-eleeffeta.pdf}}
33			\caption{ (left) Electron reconstruction efficiency as a function of $p_T$ and (right) $\eta$.\label{fig:zee-eleeff} }
34
35			\end{figure}
36
37	acosta	1.3	The two main ingredients for the electron reconstruction, the
38			supercluster and the electron track, are then evaluated separately.
39			The ratio of reconstructed supercluster energy over the true energy as
40			a function of $\eta$ is visible in Fig.~\ref{fig:zee-erecetrue},
41			showing a well-known problem in the CMSSW\_1\_0\_5 release for what
42			concerns the endcap supercluster reconstruction. The projected
43			distribution for the barrel is presented in the right plot of
44			Fig.~\ref{fig:zee-erecetrue}.
45	meridian	1.1
46			\begin{figure}[!hbtp]
47			\centering
48			{\includegraphics[width=0.49\textwidth]{figs/zee-erecetrueeta.pdf}}
49			{\includegraphics[width=0.49\textwidth]{figs/zee-erecetruebarrel.pdf}}
50			\caption{ (left) Distribution of the electron reconstructed supercluster energy over the true energy as a function of $\eta$ and (right) the corresponding distribution in the ECAL barrel acceptance. \label{fig:zee-erecetrue} }
51
52			\end{figure}
53
54			The quality of the track reconstruction is evaluated looking at the E/p distribution, using the track parameters both at the vertex and at the outermost state, as it can be seen in Fig.~\ref{fig:zee-eoverp}.
55
56			\begin{figure}[!hbtp]
57			\centering
58			{\includegraphics[width=0.49\textwidth]{figs/zee-eoverpvertex.pdf}}
59			{\includegraphics[width=0.49\textwidth]{figs/zee-eoverpoutermost.pdf}}
60			\caption{Distribution of the E/p variable using track parameters at the (left) vertex (right) outermost state. \label{fig:zee-eoverp} }
61
62			\end{figure}
63
64	acosta	1.3	At the time of the CMSSW\_1\_0\_5, the reconstructed electron track
65			was not the track after the smoothing step, and for this reason the
66			quality of the E/p distribution with the track parameters at the
67			vertex is rather poor. A refitting of the electron track has been
68			tried in the offline analysis, improving the quality of the E/p
69			matching, as it visible in Fig.~\ref{fig:zee-refittedeoverp}.
70	meridian	1.1
71			\begin{figure}[htbp]
72			\centering
73			\includegraphics[width=0.6\textwidth]{figs/zee-refittedeoverp.pdf}
74			\caption{Distribution of the E/p variable using refitted track parameters at the vertex. \label{fig:zee-refittedeoverp} }
75
76			\end{figure}
77
78	acosta	1.3	The invariant mass computed from the electron pair nearest two the nominal Z mass is presented in Fig.~\ref{fig:zee-invariantmass}.
79	meridian	1.1	\begin{figure}[h]
80			\centering
81			\includegraphics[width=0.6\textwidth]{figs/zee-invariantmass.pdf}
82			\caption{Electron pair invariant mass in Z$\rightarrow e^{+} e^{-}$ events. \label{fig:zee-invariantmass}}
83
84			\end{figure}
85
86	acosta	1.3	The distribution of the difference between the reconstructed mass and
87			the generated mass is shown in
88			Fig.~\ref{fig:zee-relativemassdifference}, where the right plot
89			displays the difference as a fuction of $\eta$. The distribution shows
90			the expected
91			behaviour due to the effect of the increasing tracker material towards
92			$\eta$ over the electron reconstruction.
93	meridian	1.1
94			\begin{figure}[!hbtp]
95			\centering
96			{\includegraphics[width=0.49\textwidth]{figs/zee-relativemassdifference.pdf}}
97			{\includegraphics[width=0.49\textwidth]{figs/zee-relativemassdifferenceeta.pdf}}
98			\caption{ (left) Relative difference between the reconstructed and the generated mass in Z$\rightarrow e^{+} e^{-}$ events and (right) its bahaviour as a function of $\eta$. \label{fig:zee-relativemassdifference} }
99
100			\end{figure}
101
102	acosta	1.2	%It can be argued that
103			Using the corrections present in CMSSW\_1\_0\_5, the invariant mass peak is about 1.4$\%$ off, while the Z invariant mass resolution is about 1.8$\%$,
104	meridian	1.1
105
106
107
108
109
110
111
112