COMP/CSA06DOC/analysisdemo.tex

\section {Analysis Demonstrations}

Different physics groups prepared some analysis codes to extract
physics results by running on skimmed files at Tier-2s.

\subsection {Calibration}

\input{calib_ana.tex}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection {Alignment}

A tracker alignment CSA06 exercise was carried out with the goal to
demonstrate the full work- and dataflow of the alignment process.
The exercise followed closely the ideas and concepts
developed during the T0-RTAG~\cite{tier0rtag}.
The exercise comprised the following steps:
\begin{itemize}
\item Reading of alignment constants from the offline database during prompt
reconstruction;
\item Writing dedicated AlCaReco streams for alignment;
\item Defining a misalignment scenario and insertion of the corresponding
object into the offline database;
\item Running an alignment algorithm at the Tier-0;
\item Inserting the resulting alignment corrections into the database;
\item Running re-reconstruction at a Tier-1 centre reading this alignment
object;
\item Running analysis jobs in which ideal, misaligned
and aligned distributions are compared.
\end{itemize}

The steps regarding production of the dedicated AlCaReco streams
as well as enabling the prompt reconstruction to read alignment
constants from the offline database were already described in
section~\ref{sec:offlineswalca}. In the following, a short
summary of the exercise is given. For more details, 
see~\cite{alcacsa06note}.

The following misalignment scenario was defined for this exercise: The
sensors of the Tracker Inner Barrel TIB as well as the Rods of the
Tracker Outer Barrel TOB were misaligned.  Random shifts drawing from
a flat distribution in the range $\pm 100 \rm\ \mu m$ were applied in
$u$, $v$, $w$ (local coordinate system) for layers built from double
sided modules, and in $u$ (precise coordinate) only for layers built
from single sided modules.  In addition, random rotations around all
three local coordinate axes of size $\pm 10 \rm\ mrad$ were applied to
the modules/rods of both single and double sided TIB and TOB layers. 
The pixel detector was kept fixed
in order to define a reference system. In addition, the outermost TOB
layer was also kept fixed in order to improve the convergence. This
misalignment scenario was inserted as an alignment object into the
offline database. In addition, another object corresponding to the
ideal tracker geometry was inserted, to be used during prompt
reconstruction.

\begin{figure}
\centering
\includegraphics[angle=270,width=0.8\linewidth]{figs/csa06_convergence}
\caption{Alignment exercise: The convergence of the alignment
algorithm in $\Delta u, \Delta v, \Delta w$ is shown for double sided
TIB modules as a function of the iteration number (left), as well as
projected for initial misalignment (labeled ``after 0 Iterations'')
and after 4, 7 and 10 iterations (right).  }
\label{fig:alignment_convergence}
\end{figure}

The alignment was performed running the HIP alignment
algorithm~\cite{hipnote} as implemented in CMSSW\_1\_0\_6 over approx. 
$1 \rm\ M$ AlCaReco $Z^0\rightarrow\mu^+\mu^-$ events produced during
prompt reconstruction at the Tier-0, reading the above mentioned
misalignment object from the offline database. The algorithm was run
on 20 dedicated CPUs in parallel at CERN, iterating 10 times over the
data sample. The result of the alignment was obtained in less than $5
\rm\ h$, and the corresponding tracker alignment object
was inserted into the database. Figure~\ref{fig:alignment_convergence}
illustrates the convergence of the alignment for the double sided TIB
sensors.

\begin{figure}
\centering
\includegraphics[angle=270,width=0.8\linewidth]{figs/csa06_aliexercise}
\caption{Invariant mass distribution from $Z^0\to \mu^+ \mu^-$ events,
obtained from events produced by the prompt reconstruction at
the Tier-0 (``ideal''), from events processed with misalignment
as used as input for the alignment algorithm (``misaligned'') and
from events re-reconstructed at a Tier-1 centre (PIC) using the alignment
constants derived from the alignment algorithm (``realigned'').
}
\label{fig:aliexercise}
\end{figure}

Once the alignment and calibration constants were inserted in the
database, they were deployed to the Tier-1/2 centres via Frontier.
Subsequently, re-reconstruction of some of the CSA06 datasets was
launched at various Tier-1 centres. For instance, the
$Z^0\rightarrow\mu^+\mu^-$ data set was re-reconstructed at PIC
(Barcelona) using the new alignment object. In order to demonstrate
the final missing piece of the workflow, grid analysis jobs were
submitted to PIC to process these re-reconstructed
$Z^0\rightarrow\mu^+\mu^-$ events. The reconstructed invariant di-muon
mass is presented in Figure~\ref{fig:aliexercise} for three cases:
\begin{itemize}
\item Using the ideal geometry, reading the RECO produced during prompt
reconstruction;
\item Using the misaligned geometry, reading the AlCaReco and the misalignment
scenario database object;
\item Using the realigned geometry, reading the events re-reconstructed
with the alignment database object.
\end{itemize}
As can be seen, the invariant mass resolution is degraded in the case
of misalignment. After the alignment algorithm has corrected the
tracker geometry, the resolution is recovered close to the original
value. This demonstrates that the work- and dataflow of the full
alignment was successfully carried out.


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection {Physics Analysis Exercises}

\subsubsection {Effect of tracker misalignment on track reconstruction performances}

The alignment uncertainties of the CMS Tracker detector, made of a huge amount of independent silicon sensors with an excellent position resolution, affect the performances of the track reconstruction and track parameters measurement. 
The analysis exercise performed by the team at Bari during the CSA06 had the purpose of study the effect of the CMS tracker misalignment on the performances of the track reconstruction \cite{misalignment}. Realistic estimates for the expected displacements of the tracking systems were supplied in different scenarios as specified in the following: 
 
\begin{itemize}
\item the ideal scenario with a perfect tracker geometry;
\item the short term misalignment scenario supposed to reproduce the mis-alignment conditions during the first 
data taking when the uncertainties on the position of the sub-structures of the CMS
tracker will be between $10 \, \mu$ for pixel detectors and $400 \, \mu$ for microstrip silicon detectors in 
the endcaps. Detector position and errors are read
 from the offline database at CERN by caching the needed information locally via frontier/squid
software.

\item the long term scenario when the alignment uncertainties are supposed to be a factor 10 smaller because of the
improvement obtained by using aligmnent algorithms with a high statistics of tracks.
 
\item the CSA06 aligned scenario by using the tracker module position and errors as obtained by the output of the
 alignment procedure that was run at CERN Tier-0 to verify the efficiency of the alignment procedure on the track
reconstruction. The refit of tracks is performed also in this case.
 
\end{itemize}

Track reconstruction is based on the Kalman Filter formalism \cite{Kalman} for trajectory building, cleaning and 
smoothing steps and uses hits from pixel detector as seeds to provide initial trajectory candidates. 
Because of the misalignment the analysis requires to refit tracks with a misaligned tracker geometry.
Global efficiency of track recostruction and track parameter resolutions for muons were
compared in all the cases. The association between simulated track and reconstruted tracks is performed
by comparing the corresponding track parameters at the closest approach point and choosing the pair which gives 
the minimum $\chi^2$ from the best fit procedure.

Events from CSA06 $Z\rightarrow \mu \mu $ sample were firstly skimmed by selecting events with 
 Hep MC muons from Z decay with pseudorapidity, $\eta$, in the tracker acceptance, $|\eta| < 2.55$, 
with transverse momentum larger than $5 \, \mathrm{GeV}/c^2$ and di-muons invariant mass in the following 
range aroung the Z peak: $50 < m_{\mu\mu}(\mathrm{GeV}/c^2) < 130$; the efficiency of the previous selection is 
between 50 and 60 \% mainly due to the cut on the acceptance, for a final statistics of 1 million events. 
The output files in RECOSIM format were needed for the subsequent analysis.


Jobs executing the misalignment analysis were submitted at Bari with CRAB\_1\_4\_0 in the LCG infrastructure.
A total of about 2.5 thousands jobs (45 at most in parallel) ran with a grid efficiency of 90 \% and an 
application efficiency of 80\%, by accessing detector position and errors from the offline database via 
frontier.


Some results of the misalignment analysis were summarized below. The global efficiency of track reconstruction of muons coming from
Z decay is shown in Fig.~\ref{eff} as a function of the pseudorapidity, $\eta$, in the tracker acceptance. 
In the case of a perfect geometry the global track reconstruction was not fully efficient over all the $\eta$ 
range because of the track associator algorithm itself which discards tracks with  $\chi^2$ of the fit less 
than 25. The effect of misalignment is relevant in the short term scenario and causes a partial inefficiency of the
track reconstruction; that can be recovered if the intrinsic position resolution of the tracker
 detector is combined with the alignment uncertainties to make larger the error on the position of the
reconstructed hit (called alignment position error, APE) so improving the track fit at the
expence of a larger rate of fake tracks.
                                    

The tranverse momentum resolution as a function of the transverse momentum is reported in Fig.~\ref{respt};
the degradation of the tranverse momentum resolution at large $p_{T}$ because of the misalignment is in a factor
 between 2 and 3 with respect to the perfect geometry case. At low transverse momentum (less than few $\mathrm{GeV}/c$ 
the multiple scattering is the 
most important contribution to the resolution so the effect of misalignment is overwhelmed at all.
                                                                                    
The residual of Z mass obtained as the invariant mass of muons coming from Z decay in the case of perfect
tracker geometry and in short-term and long term misalignment scenarios is shown in Fig.~\ref{mz}; the $\sigma$ of 
the Gaussian fit of the residual ditribution can be quoted as the Z mass resolution which is degradated of a factor 2 
because of the tracker misalignment in the short term scenario.
                                                                         
\begin{2figures}{hbt}
  \resizebox{\linewidth}{!}{\includegraphics{figs/Eff_eta}} &
  \resizebox{\linewidth}{!}{\includegraphics{figs/SigmapT_pT}} \\
  \caption{Global track reconstrution efficiency vs pseudorapidity for muons coming from Z decay in the case of perfect 
tracker geometry and in short-term and long term misalignment scenarios when the APE is not used.}
  \label{eff} &
  \caption{$P_{T}$ resolution vs $p_{T}$ in the case of perfect
tracker geometry and in short-term and long term misalignment scenarios.}
  \label{respt} \\
\end{2figures}


\begin{figure}[htb]
  \begin{center}
  \resizebox{0.7\linewidth}{!}{\includegraphics{figs/Residual_mZ_mu}} 
  \end{center}
  \caption{Residual of Z mass obtained as the invariant mass of muons coming from Z decay in the case of perfect
tracker geometry and in short-term and long term misalignment scenarios.}
  \label{mz} 
\end{figure}


% tau analyses:
         
\input{z_2tau.tex}

\input{taumisid.tex}

\input{tau_validation.tex}

Revision:	1.8
Committed:	Thu Nov 30 19:57:05 2006 UTC (18 years, 5 months ago) by acosta
Content type:	application/x-tex
Branch:	MAIN
Changes since 1.7:	+8 -0 lines
Log Message:	enable tau analyses
#	User	Rev	Content
1	fisk	1.1	\section {Analysis Demonstrations}
2
3	ndefilip	1.3	Different physics groups prepared some analysis codes to extract
4			physics results by running on skimmed files at Tier-2s.
5
6	fisk	1.1	\subsection {Calibration}
7
8	malgeri	1.6	\input{calib_ana.tex}
9
10	fpschill	1.5	%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
11
12			\subsection {Alignment}
13
14			A tracker alignment CSA06 exercise was carried out with the goal to
15			demonstrate the full work- and dataflow of the alignment process.
16			The exercise followed closely the ideas and concepts
17			developed during the T0-RTAG~\cite{tier0rtag}.
18			The exercise comprised the following steps:
19			\begin{itemize}
20			\item Reading of alignment constants from the offline database during prompt
21			reconstruction;
22			\item Writing dedicated AlCaReco streams for alignment;
23			\item Defining a misalignment scenario and insertion of the corresponding
24			object into the offline database;
25			\item Running an alignment algorithm at the Tier-0;
26			\item Inserting the resulting alignment corrections into the database;
27			\item Running re-reconstruction at a Tier-1 centre reading this alignment
28			object;
29			\item Running analysis jobs in which ideal, misaligned
30			and aligned distributions are compared.
31			\end{itemize}
32
33			The steps regarding production of the dedicated AlCaReco streams
34			as well as enabling the prompt reconstruction to read alignment
35			constants from the offline database were already described in
36			section~\ref{sec:offlineswalca}. In the following, a short
37			summary of the exercise is given. For more details,
38			see~\cite{alcacsa06note}.
39
40			The following misalignment scenario was defined for this exercise: The
41			sensors of the Tracker Inner Barrel TIB as well as the Rods of the
42			Tracker Outer Barrel TOB were misaligned. Random shifts drawing from
43			a flat distribution in the range $\pm 100 \rm\ \mu m$ were applied in
44			$u$, $v$, $w$ (local coordinate system) for layers built from double
45			sided modules, and in $u$ (precise coordinate) only for layers built
46			from single sided modules. In addition, random rotations around all
47			three local coordinate axes of size $\pm 10 \rm\ mrad$ were applied to
48			the modules/rods of both single and double sided TIB and TOB layers.
49			The pixel detector was kept fixed
50			in order to define a reference system. In addition, the outermost TOB
51			layer was also kept fixed in order to improve the convergence. This
52			misalignment scenario was inserted as an alignment object into the
53			offline database. In addition, another object corresponding to the
54			ideal tracker geometry was inserted, to be used during prompt
55			reconstruction.
56
57			\begin{figure}
58			\centering
59	malgeri	1.6	\includegraphics[angle=270,width=0.8\linewidth]{figs/csa06_convergence}
60	fpschill	1.5	\caption{Alignment exercise: The convergence of the alignment
61			algorithm in $\Delta u, \Delta v, \Delta w$ is shown for double sided
62			TIB modules as a function of the iteration number (left), as well as
63			projected for initial misalignment (labeled ``after 0 Iterations'')
64			and after 4, 7 and 10 iterations (right). }
65			\label{fig:alignment_convergence}
66			\end{figure}
67
68			The alignment was performed running the HIP alignment
69			algorithm~\cite{hipnote} as implemented in CMSSW\_1\_0\_6 over approx.
70			$1 \rm\ M$ AlCaReco $Z^0\rightarrow\mu^+\mu^-$ events produced during
71			prompt reconstruction at the Tier-0, reading the above mentioned
72			misalignment object from the offline database. The algorithm was run
73			on 20 dedicated CPUs in parallel at CERN, iterating 10 times over the
74			data sample. The result of the alignment was obtained in less than $5
75			\rm\ h$, and the corresponding tracker alignment object
76			was inserted into the database. Figure~\ref{fig:alignment_convergence}
77			illustrates the convergence of the alignment for the double sided TIB
78			sensors.
79
80			\begin{figure}
81			\centering
82	malgeri	1.6	\includegraphics[angle=270,width=0.8\linewidth]{figs/csa06_aliexercise}
83	fpschill	1.5	\caption{Invariant mass distribution from $Z^0\to \mu^+ \mu^-$ events,
84			obtained from events produced by the prompt reconstruction at
85			the Tier-0 (``ideal''), from events processed with misalignment
86			as used as input for the alignment algorithm (``misaligned'') and
87			from events re-reconstructed at a Tier-1 centre (PIC) using the alignment
88			constants derived from the alignment algorithm (``realigned'').
89			}
90			\label{fig:aliexercise}
91			\end{figure}
92
93			Once the alignment and calibration constants were inserted in the
94			database, they were deployed to the Tier-1/2 centres via Frontier.
95			Subsequently, re-reconstruction of some of the CSA06 datasets was
96			launched at various Tier-1 centres. For instance, the
97			$Z^0\rightarrow\mu^+\mu^-$ data set was re-reconstructed at PIC
98			(Barcelona) using the new alignment object. In order to demonstrate
99			the final missing piece of the workflow, grid analysis jobs were
100			submitted to PIC to process these re-reconstructed
101			$Z^0\rightarrow\mu^+\mu^-$ events. The reconstructed invariant di-muon
102			mass is presented in Figure~\ref{fig:aliexercise} for three cases:
103			\begin{itemize}
104			\item Using the ideal geometry, reading the RECO produced during prompt
105			reconstruction;
106			\item Using the misaligned geometry, reading the AlCaReco and the misalignment
107			scenario database object;
108			\item Using the realigned geometry, reading the events re-reconstructed
109			with the alignment database object.
110			\end{itemize}
111			As can be seen, the invariant mass resolution is degraded in the case
112			of misalignment. After the alignment algorithm has corrected the
113			tracker geometry, the resolution is recovered close to the original
114			value. This demonstrates that the work- and dataflow of the full
115			alignment was successfully carried out.
116
117
118			%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
119	fisk	1.1
120			\subsection {Physics Analysis Exercises}
121
122	ndefilip	1.2	\subsubsection {Effect of tracker misalignment on track reconstruction performances}
123
124	ndefilip	1.3	The alignment uncertainties of the CMS Tracker detector, made of a huge amount of independent silicon sensors with an excellent position resolution, affect the performances of the track reconstruction and track parameters measurement.
125			The analysis exercise performed by the team at Bari during the CSA06 had the purpose of study the effect of the CMS tracker misalignment on the performances of the track reconstruction \cite{misalignment}. Realistic estimates for the expected displacements of the tracking systems were supplied in different scenarios as specified in the following:
126
127			\begin{itemize}
128			\item the ideal scenario with a perfect tracker geometry;
129			\item the short term misalignment scenario supposed to reproduce the mis-alignment conditions during the first
130			data taking when the uncertainties on the position of the sub-structures of the CMS
131			tracker will be between $10 \, \mu$ for pixel detectors and $400 \, \mu$ for microstrip silicon detectors in
132			the endcaps. Detector position and errors are read
133			from the offline database at CERN by caching the needed information locally via frontier/squid
134	ndefilip	1.7	software.
135	ndefilip	1.3
136			\item the long term scenario when the alignment uncertainties are supposed to be a factor 10 smaller because of the
137			improvement obtained by using aligmnent algorithms with a high statistics of tracks.
138
139			\item the CSA06 aligned scenario by using the tracker module position and errors as obtained by the output of the
140			alignment procedure that was run at CERN Tier-0 to verify the efficiency of the alignment procedure on the track
141			reconstruction. The refit of tracks is performed also in this case.
142
143			\end{itemize}
144
145	ndefilip	1.4	Track reconstruction is based on the Kalman Filter formalism \cite{Kalman} for trajectory building, cleaning and
146			smoothing steps and uses hits from pixel detector as seeds to provide initial trajectory candidates.
147			Because of the misalignment the analysis requires to refit tracks with a misaligned tracker geometry.
148			Global efficiency of track recostruction and track parameter resolutions for muons were
149			compared in all the cases. The association between simulated track and reconstruted tracks is performed
150			by comparing the corresponding track parameters at the closest approach point and choosing the pair which gives
151			the minimum $\chi^2$ from the best fit procedure.
152	ndefilip	1.3
153	ndefilip	1.4	Events from CSA06 $Z\rightarrow \mu \mu $ sample were firstly skimmed by selecting events with
154			Hep MC muons from Z decay with pseudorapidity, $\eta$, in the tracker acceptance, $\|\eta\| < 2.55$,
155			with transverse momentum larger than $5 \, \mathrm{GeV}/c^2$ and di-muons invariant mass in the following
156			range aroung the Z peak: $50 < m_{\mu\mu}(\mathrm{GeV}/c^2) < 130$; the efficiency of the previous selection is
157			between 50 and 60 \% mainly due to the cut on the acceptance, for a final statistics of 1 million events.
158			The output files in RECOSIM format were needed for the subsequent analysis.
159	ndefilip	1.3
160
161	ndefilip	1.4	Jobs executing the misalignment analysis were submitted at Bari with CRAB\_1\_4\_0 in the LCG infrastructure.
162			A total of about 2.5 thousands jobs (45 at most in parallel) ran with a grid efficiency of 90 \% and an
163			application efficiency of 80\%, by accessing detector position and errors from the offline database via
164			frontier.
165	ndefilip	1.3
166	ndefilip	1.4
167			Some results of the misalignment analysis were summarized below. The global efficiency of track reconstruction of muons coming from
168			Z decay is shown in Fig.~\ref{eff} as a function of the pseudorapidity, $\eta$, in the tracker acceptance.
169			In the case of a perfect geometry the global track reconstruction was not fully efficient over all the $\eta$
170			range because of the track associator algorithm itself which discards tracks with $\chi^2$ of the fit less
171			than 25. The effect of misalignment is relevant in the short term scenario and causes a partial inefficiency of the
172			track reconstruction; that can be recovered if the intrinsic position resolution of the tracker
173			detector is combined with the alignment uncertainties to make larger the error on the position of the
174			reconstructed hit (called alignment position error, APE) so improving the track fit at the
175			expence of a larger rate of fake tracks.
176
177
178			The tranverse momentum resolution as a function of the transverse momentum is reported in Fig.~\ref{respt};
179			the degradation of the tranverse momentum resolution at large $p_{T}$ because of the misalignment is in a factor
180			between 2 and 3 with respect to the perfect geometry case. At low transverse momentum (less than few $\mathrm{GeV}/c$
181			the multiple scattering is the
182			most important contribution to the resolution so the effect of misalignment is overwhelmed at all.
183
184			The residual of Z mass obtained as the invariant mass of muons coming from Z decay in the case of perfect
185			tracker geometry and in short-term and long term misalignment scenarios is shown in Fig.~\ref{mz}; the $\sigma$ of
186			the Gaussian fit of the residual ditribution can be quoted as the Z mass resolution which is degradated of a factor 2
187			because of the tracker misalignment in the short term scenario.
188
189			\begin{2figures}{hbt}
190	malgeri	1.6	\resizebox{\linewidth}{!}{\includegraphics{figs/Eff_eta}} &
191			\resizebox{\linewidth}{!}{\includegraphics{figs/SigmapT_pT}} \\
192	ndefilip	1.3	\caption{Global track reconstrution efficiency vs pseudorapidity for muons coming from Z decay in the case of perfect
193			tracker geometry and in short-term and long term misalignment scenarios when the APE is not used.}
194			\label{eff} &
195	ndefilip	1.4	\caption{$P_{T}$ resolution vs $p_{T}$ in the case of perfect
196			tracker geometry and in short-term and long term misalignment scenarios.}
197			\label{respt} \\
198			\end{2figures}
199
200
201			\begin{figure}[htb]
202			\begin{center}
203	malgeri	1.6	\resizebox{0.7\linewidth}{!}{\includegraphics{figs/Residual_mZ_mu}}
204	ndefilip	1.4	\end{center}
205	ndefilip	1.3	\caption{Residual of Z mass obtained as the invariant mass of muons coming from Z decay in the case of perfect
206			tracker geometry and in short-term and long term misalignment scenarios.}
207	ndefilip	1.4	\label{mz}
208			\end{figure}
209
210	acosta	1.8
211			% tau analyses:
212	ndefilip	1.4
213	acosta	1.8	\input{z_2tau.tex}
214
215			\input{taumisid.tex}
216
217			\input{tau_validation.tex}
218