| 1 |
< |
The dominant (excluding rare decays to kaons, which are not considered here.) |
| 2 |
< |
hadronic decays of taus consist of a varying number of charged and neutral |
| 3 |
< |
pions. The neutral pions undergo prompt decay to photon pairs. These decays |
| 4 |
< |
proceed through intermediate resonances, given in |
| 5 |
< |
table~\ref{table:decay_modes}. Each of these decay modes uniquely maps to a |
| 6 |
< |
tau final state multiplicity, and each resonance has a different invariant |
| 7 |
< |
mass. This implies that the problem of hadronic tau identification can be |
| 8 |
< |
reframed from a global search for collimated hadrons under the tau mass into an |
| 9 |
< |
ensemble of searches for single production of the various decay resonances |
| 10 |
< |
given in table~\ref{table:decay_modes}. In this paper, we present a novel |
| 11 |
< |
algorithm, the ``Tau Neural Classifier'' (TaNC) which uses this approach to |
| 12 |
< |
improve on traditional tau-ID strategies. |
| 1 |
> |
The tau identification strategy described previously can be extened by examining |
| 2 |
> |
the underlying physics of the different hadronic tau lepton decays. The |
| 3 |
> |
dominant hadronic decays of taus consist of a one or three charged $\pi^{\pm}$ |
| 4 |
> |
mesons and up to two $\pi^0$ mesons and are enumerated in |
| 5 |
> |
table~\ref{table:decay_modes}. The majority of these decays proceed through |
| 6 |
> |
intermediate resonances and each of these decay modes maps directly to a tau |
| 7 |
> |
final state multiplicity. Each intermediate resonance has a different invariant |
| 8 |
> |
mass (see figure~\ref{fig:trueInvMass}). This implies that the problem of |
| 9 |
> |
hadronic tau identification can be re-framed from a global search for collimated |
| 10 |
> |
hadrons satisfying the tau mass constraint into a ensemble of searches for |
| 11 |
> |
single production of the different hadronic tau decay resonances. The Tau |
| 12 |
> |
Neural Classifier algorithm implements this approach using two complimentary |
| 13 |
> |
techniques: a method to reconstruct the decay mode and an ensemble of neural |
| 14 |
> |
network classifiers used to discriminant the individual decay modes. |
| 15 |
|
|
| 16 |
|
\begin{table} |
| 15 |
– |
\caption{Visible products of hadronic tau decays} |
| 17 |
|
\centering |
| 18 |
|
\begin{tabular}{ l c r r } |
| 19 |
< |
Visible Decay Products & Resonance & Mass (M$e$V/$c^2$) & Fraction \\ |
| 19 |
> |
Visible Decay Products & Resonance & Mass (M$e$V/$c^2$) & |
| 20 |
> |
Fraction~\cite{pdg} \\ |
| 21 |
|
\hline |
| 22 |
< |
$\pi^{-}$ & n/a & 135 & fixme \\ |
| 23 |
< |
$\pi^{-}\pi^0$ & $\rho$ & 770 & fixme \\ |
| 24 |
< |
$\pi^{-}\pi^0\pi^0$ & $a1$ & 1200 & fixme \\ |
| 25 |
< |
$\pi^{-}\pi^{-}\pi^{+}$ & $a1$ & 1200 & fixme \\ |
| 26 |
< |
$\pi^{-}\pi^{-}\pi^{+}\pi^0$ & $a1$ & 1200 & fixme \\ |
| 22 |
> |
$\pi^{-}$ & n/a & 135 & 10.9\% \\ |
| 23 |
> |
$\pi^{-}\pi^0$ & $\rho$ & 770 & 25.5\% \\ |
| 24 |
> |
$\pi^{-}\pi^0\pi^0$ & $a1$ & 1200 & 9.3\% \\ |
| 25 |
> |
$\pi^{-}\pi^{-}\pi^{+}$ & $a1$ & 1200 & 9.03\% \\ |
| 26 |
> |
$\pi^{-}\pi^{-}\pi^{+}\pi^0$ & $a1$ & 1200 & 4.5\% \\ |
| 27 |
|
\hline |
| 28 |
< |
Total & & & 65\% \\ |
| 28 |
> |
Total & & & 59.2\% \\ |
| 29 |
> |
\hline |
| 30 |
> |
Other hadronic modes & & & 5.59\% \\ |
| 31 |
|
\end{tabular} |
| 32 |
< |
\label{table:decay_modes} |
| 32 |
> |
\label{tab:decay_modes} |
| 33 |
> |
\caption{Resonances and branching ratios of the dominant hadronic decays of |
| 34 |
> |
the tau lepton.} |
| 35 |
|
\end{table} |
| 36 |
|
|
| 37 |
+ |
\begin{figure}[t] |
| 38 |
+ |
\begin{center} |
| 39 |
+ |
\includegraphics[width=90mm]{figures/truthIMvsDM.pdf} |
| 40 |
+ |
\end{center} |
| 41 |
+ |
\caption{The invariant mass of the visible decay products in hadronic tau |
| 42 |
+ |
decays. The decay mode $\tau^{-} \rightarrow \pi^{-} \nu_\tau$ is omitted. |
| 43 |
+ |
The different decay modes have different invariant masses corresponding to |
| 44 |
+ |
the intermediate resonance in the decay.} |
| 45 |
+ |
\label{fig:trueInvMass} |
| 46 |
+ |
\end{figure} |
