|
[Rivet-svn] r3873 - schools/2012-CERN/handouts/day3blackhole at projects.hepforge.org blackhole at projects.hepforge.orgThu Jul 26 08:23:36 BST 2012
Author: dgrell Date: Thu Jul 26 08:23:36 2012 New Revision: 3873 Log: Copied handouts Added: schools/2012-CERN/handouts/day3/day3.tex - copied unchanged from r3869, schools/2011-Kyoto/handouts/day3/day3.tex schools/2012-CERN/handouts/day3/rivet-classes.pdf - copied unchanged from r3869, schools/2011-Kyoto/handouts/day3/rivet-classes.pdf Copied: schools/2012-CERN/handouts/day3/day3.tex (from r3869, schools/2011-Kyoto/handouts/day3/day3.tex) ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ schools/2012-CERN/handouts/day3/day3.tex Thu Jul 26 08:23:36 2012 (r3873, copy of r3869, schools/2011-Kyoto/handouts/day3/day3.tex) @@ -0,0 +1,279 @@ +\documentclass{article} + +\usepackage[a4paper]{geometry} +\usepackage[sc]{mathpazo} +\usepackage{amsmath} +\usepackage{microtype} +\usepackage{fullpage} + +\def\ttbar{\ensuremath{t \bar t}} + +%%%%%%%%%%%%%%%%%%%%%% +% Code +\makeatletter +\def\code#1{\texttt{#1}} + +\def\codeblock% + {\@verbatim\frenchspacing\@vobeyspaces + \@ycodeblock{$\hookrightarrow$}\@xcodeblock} +\def\endcodeblock% + {\if at newlist \leavevmode\fi\endtrivlist} +\begingroup + \catcode `§=13 + \catcode `[= 1 \catcode`]=2 + \catcode `\{=12 \catcode `\}=12 + \catcode `|=0 \catcode`\\=12 + |gdef|@xcodeblock#1\end{codeblock}[#1|end[codeblock]] + |gdef|@ycodeblock[|catcode`§=13 |def§] +|endgroup +\makeatother +%%%%%%%%%%%%%%%%%%%%%% + +\begin{document} + +\title{IPMU-YITP School 2011 --\\A simple Top analysis with Rivet} +\date{7 September 2011} +\maketitle + + +\section{Introduction} + +Today we will dive a bit deeper into Rivet and create a simple \ttbar{} +mass analysis. More explicitly you are going to look at the electron and muon +channel in semi-leptonic \ttbar{} events, reconstructing the hadronic +top. + +\subsection{Physics} + +The top quark is the heaviest of the six quarks, its mass is about +173~GeV. It also has an extremely short life-time, so it doesn't have +time to hadronise before decaying. Top quarks almost exclusively decay +into a $W$ boson and a $b$ quark (the CKM matrix element $V_{tb}$ is +basically 1). Therefore their decay channels can be characterised by the +$W$ decay: If the $W$ decays into an electron, muon, or tau, and the +associated neutrino, we talk about a leptonic $W$ decay, and if the $W$ +decays into a quark-antiquark pair, the decay is called hadronic. The +quarks in the final state evolve into jets of hadrons. + +This yields three categories of final states for top pairs: In the +dilepton channel the $W$ bosons from both top quarks decay into leptons, +in the semi-leptonic (or lepton+jets) channel one $W$ decay is hadronic +and the other one is leptonic, and in the alljets (or hadronic) channel +both $W$ bosons decay into quarks. + +Each of the three decay modes has its own advantages and disadvantages. +The dilepton channel features a clear signature because of its two $b$ +jets and the two high energy charged leptons, but suffers from low +statistics --~only 10.3\,\% of the top pairs decay in this channel~-- +and kinematic ambiguities because of the momentum carried away by the +undetected neutrinos. + +The semi-leptonic channel has a much higher branching ratio of 43.5\,\% +and only one neutrino, so the statistics is higher than in the dilepton +channel and the kinematic ambiguities are smaller. On the other hand +this channel is affected by a large background from events with one $W$ +boson and additional jets. The signature of this channel is one high +energy lepton, two $b$ jets, two light quark jets, and missing +momentum from the escaping neutrino in the plane transverse to the beam +pipe. + +The alljets channel finally has the largest branching ratio (46.2\,\%) +and no neutrino at all, so there is plenty of events which are +kinematically well constrained (but suffer from combinatoric +ambiguities). The $W$+jets background can be neglected when the $b$ jets +are identified, but the background from QCD multijet production is +overwhelming. The alljets channel's signature is six high energy jets, +two of which are $b$ jets, and no lepton. + +\subsection{Mass measurement in the semi-leptonic channel} + +To measure the mass in the semi-leptonic channel it is necessary to +identify \ttbar{} events and then reconstruct the hadronic top to get +its invariant mass. The signature of those events are a high momentum +lepton from the $W$ decay, two $b$-jets from the top decay, and two +light jets from the decay of the second $W$. All jets are coming from +decays of heavy particles, so they are relatively hard -- harder than +typical jets from QCD showering. + +After selecting events with four jets (two of which are $b$-jets) and a +hard lepton, you might want to reconstruct the hadronic $W$ and apply a +mass window cut around the $W$ mass. Then combine the reconstructed $W$ +with the two $b$-jets; this gives you a correct and a wrong combination, +so your mass distribution will consist of a top mass peak and a broader +combinatoric background. + +You can also run $W+\text{jets}$ Monte Carlo through your analysis to +check how your selection is performing on background. Because it is very +hard to get multiple additional jets together with the $W$ from a parton +shower, and because those jets would be too soft anyhow, it's best to +use a matrix element generator to get parton level events and let them +evolve with a shower. We have prepared Les Houches event files with +MadGraph for this purpose. Les Houches files are a common standard to +exchange such data between generators. + + +\section{Getting started} + +Before you continue, generate some event data that you can use to +refine your analysis. Since you'll probably want to rerun the analysis +again and again while you're developing it, we save the HepMC data to +a file \texttt{top-signal.hepmc}. + +% GENERATOR-SPECIFIC +Choose the command(s) for your generator: +\begin{itemize} +\item Herwig++ +\begin{codeblock} + Herwig++ read TopSignal.in + Herwig++ run TopSignal.run -N4000 +\end{codeblock} +\item Pythia +\begin{codeblock} + main32.exe topsignal.cmnd top-signal.hepmc +\end{codeblock} +\item Sherpa +\begin{codeblock} + cd ttbar + Sherpa HEPMC2_GENEVENT_OUTPUT=top-signal.hepmc EVENTS=4000 +\end{codeblock} +\end{itemize} + +You can interrupt the run with \texttt{CTRL-c} if it takes too +long. While you wait for the results, please continue with looking at +the Rivet analysis in section \ref{analysis}. + +We have prepared background samples for the most problematic background +in this channel: A $W$ boson associated with two $b$-jets and two other +jets, with the $W$ decaying into $e/\mu + \nu$. This final state has the +same signature as our signal events. Since it is hard to get four jets +from the parton shower, we have prepared Les Houches event files with +MadGraph for Herwig++ and Pythia. + +% GENERATOR-SPECIFIC +You can process them like this +\begin{itemize} +\item Herwig++ +\begin{codeblock} + Herwig++ read TopBackground.in + Herwig++ run TopBackground.run -N4000 +\end{codeblock} +\item Pythia +\begin{codeblock} + main32.exe topbackground.cmnd top-background.hepmc +\end{codeblock} +\item Sherpa +\begin{codeblock} + cd wbbjets + Sherpa HEPMC2_GENEVENT_OUTPUT=top-background.hepmc EVENTS=4000 +\end{codeblock} +\end{itemize} + +Again, while you wait, please continue with looking at +the Rivet analysis in section \ref{analysis}. +The HepMC files will be huge, so please do not upload them! If you run +into space problems, reduce the number of events accordingly. + +\subsection{Running the analysis}\label{analysis} + +We have prepared a Rivet analysis template which you can extend: +\begin{codeblock} + cd school/day3/rivet-plugin +\end{codeblock} +Rivet supports plugin analyses which are compiled as a library object +and loaded during runtime. The Makefile just contains the command +necessary to compile this library, and the file \code{MC\_TOP.cc} +contains the actual analysis code. + +The \code{MC\_TOP.cc} file has four methods: +\begin{itemize} + \item The \code{MC\_TOP::init()} method is called once at the + beginning of the run. It is used to + initialise the projections used in the analysis. We will be + using the \code{ChargedLeptons} projection to extract the + leptons from the final state particles, and the \code{FastJets} + projection for jet reconstruction. + This is also the place to book histograms. + + \item The \code{MC\_TOP::analyze()} method is called for each + event. Here you find the main analysis code. + + \item The \code{MC\_TOP::finalize()} method is called once at the + end of the run. Here you can for example normalise histograms. +\end{itemize} + +Have a look at the files and try to understand what they are doing. +There are plenty of comments in the source code. +You can compile the library by calling +\begin{codeblock} + make +\end{codeblock} +and run it as follows: +\begin{codeblock} + export RIVET_ANALYSIS_PATH=$PWD + rivet -a MC_TOP -H ttbar.aida top-signal.hepmc +\end{codeblock} +% $ +We will +only be looking at the shapes of the distributions today, +the histograms are scaled by the number of events only. In +reality at LHC energies, the cross-section for our signal process is +much higher than the background we look at. We ignore this here to make the +construction of a good analysis slightly more challenging. We haven't +included other backgrounds since they are even easier to suppress. + + +%%% HERE +\subsection{Plotting the histograms} + +Your signal histograms are saved in the file \code{ttbar.aida}. +We now run the same Rivet analysis, this time on the background dataset: +\begin{codeblock} + rivet -a MC_TOP -H background.aida top-background.hepmc +\end{codeblock} +and compare signal and background with +\begin{codeblock} + rivet-mkhtml ttbar.aida background.aida +\end{codeblock} +You can look at the plots using any browser, for example: +\begin{codeblock} + firefox plots/index.html +\end{codeblock} + + +\section{Improve the analysis} + +When you have understood what the analysis is currently doing, it's +time to improve it. + +\begin{enumerate} +\item Can you get a better signal efficiency? +\item A better background discrimination? +\item Why does the $W$ mass window cut give such a large improvement? +\item Add a lepton isolation cut to the analysis, \emph{i.e.}~only + select jets which are well separated from the hard charged lepton. +\item You can also adjust the jet $p_T$ cuts and the cut on the lepton $p_T$ +(but keep in mind that the jets in the background sample have been +created with a minimum $p_T$ of 20~GeV at parton level, so better keep +your cuts above $\sim$30~GeV). Think about how this biases the $W$ and +the top mass. +\item Try to add more observables. Things like $H_T = \sum_\text{jets}{E_T}$ +or Centrality $C = H_T/\sum_\text{jets} E$ might be worth looking at. +Try just including the four light and $b$-jets you are using, or all +jets above some $p_T$ cut. You also can include the lepton into this +calculation. If you cut on $H_T$, how does this bias your $W$ or top +mass distributions? +\item What other observables do you come up with? +\end{enumerate} + +\section{Compare with other generators and groups} + +Share your results with students who run one of the other generators. +By simply adding more ``.aida'' files to the \code{rivet-mkhtml} +command, you can compare the different generators directly to each +other. + +Once you are happy with your analysis, you can also +generate more statistics than just 4000 events for +those plots. + +\end{document} Copied: schools/2012-CERN/handouts/day3/rivet-classes.pdf (from r3869, schools/2011-Kyoto/handouts/day3/rivet-classes.pdf) ============================================================================== Binary file (source and/or target). No diff available.
More information about the Rivet-svn mailing list |