1: Final state radiation causes significant distortions of the naive,
2: lowest order \pt\ spectrum of the \W decay leptons. We estimate
3: the stability of the theoretical calculation below, using the {\tt
4: PHOTOS} program~\cite{gen:photos} as a benchmark.\\
5:
6: \noindent
7: The numerical importance of final state radiation is illustrated in
8: Figure~\ref{fsr:fig1}, which displays the distribution of the measured
9: lepton energy fraction (relative to their energy in the absence of
10: FSR). For electrons, measured $via$ calorimetric energy clusters, most
11: of the (collinearly radiated) photon energy is collected in the
12: cluster. The momentum of muons tracks, on the contrary, is measured
13: independently of any photon radiation. The average values of the
14: distributions lie at about 99\% of the original value, meaning that ignoring the effect
15: entirely would cause a bias on the \W mass of about 800 \MeV. The
16: theoretical stability of the calculation is thus of critical
17: importance.\\
18:
19: \begin{figure}
20: \begin{center}
21: \includegraphics[width=0.8\textwidth]{figures/EconoverEbef_distrib_exp.eps}
22: \caption{\label{fsr:fig1} Distribution of the measured lepton energy
23: fraction (\ie\ relative to their energy in absence of FSR). {\tt
24: PHOTOS} is run in exponentiated mode. The energy of electrons is
25: measured dressed with all photon energy radiated within a cone of
26: radius 0.1, corresponding to the size of reconstructed EM clusters.
27: Muon momentum is measured bare, after FSR.}
28: \end{center}
29: \end{figure}
30:
31: \noindent
32: In recent versions of {\tt PHOTOS}, it is possible to switch between
33: several theoretical assumptions. In particular, \W and \Z
34: boson decays can be simulated with photon emission up to {\cal
35: O($\alpha$)}, {\cal O($\alpha^2$)}, {\cal O($\alpha^4$)}, or with
36: photon emission exponentiation~\cite{th:yfs}. To study the model
37: differences, we have generated about $10^6$ events for each setting,
38: and for each production and decay channel ($\W \ra \ell\nu$,
39: $\Z \ra \ell\ell$, for $\ell = e, \mu$).\\
40:
41: \noindent
42: The average values of the energy fractions discussed above are shown
43: in Figure~\ref{fsr:fig2}, for successive theoretical refinements.
44: The different average values for electrons and muons
45: reflect the different ways their energy or momentum is measured.
46: The calculation appears stable to about 1-2$\times 10^{-4}$, the
47: residual differences being compatible with coming from the finite sample
48: statistics only. It is unfortunately not practical to further increase
49: the samples sizes and quantify the stability to better precision.\\
50:
51: \begin{figure}[tp]
52: \begin{center}
53: \includegraphics[width=0.8\textwidth]{figures/EconoverEbef_meanWandZ.eps}
54: \caption{\label{fsr:fig2} Averages of the distributions of
55: Figure~\ref{fsr:fig1}, for various {\tt PHOTOS} settings (see text).}
56: \end{center}
57: \end{figure}
58:
59: \noindent
60: To improve on the above argument, consider the \Z boson
61: mass measurement at LEP1~\cite{ex:lep1final}. Similarly to our case, QED
62: corrections, in the form of initial state radiation off the electron
63: beams, have a large impact on the \Z lineshape, inducing a
64: decrease of the cross-section of about
65: 30\%, and a shift of the peak position of about
66: 100~\MeV. Nevertheless, the theoretical uncertainty on these effects
67: are estimated to 0.3~\MeV, compared to a total measurement uncertainty
68: of 2.1~\MeV. The theory of QED radiation thus carries negligible
69: uncertainty.\\
70:
71: \noindent
72: For the QED induced \MW\ uncertainty to be as small, the event
73: generators used to produce our templates thus need to have similar
74: theoretical accuracy, with the additional complication that the
75: present analysis requires an exclusive description of the final state
76: (i.e, a complete description of the photon distributions), whereas
77: the \Z lineshape analysis only relies on the effective energy of the
78: beams after radiation. In~Ref.~\cite{gen:photosnlo}, the accuracy of the
79: {\tt PHOTOS} algorithm is upgraded to NLO accuracy. Similarly, the
80: {\tt HORACE} event generator~\cite{gen:horace} contains QED and weak
81: corrections to NLO accuracy. Both programs implement photon emission
82: exponentiation.\\
83:
84: \noindent
85: We thus assume that ultimately $\delta {\MW}(QED)\leq 1$~\MeV\ can be
86: reached. This assumption is conditioned by the availability of the
87: necessary tools in time for the measurement.\\
88:
89: \noindent
90: Let us finally note that \W and \Z events behave differently
91: under QED radiation. The
92: average energy fraction in \Z events is 5-7$\times 10^{-3}$
93: smaller than in \W events, depending on the final state. The
94: energy scale measurement (\cf\ Section~\ref{subsec:lepscale}) and the
95: \W mass measurement should properly account for the difference in
96: the respective QED radiation patterns. We will come back to this point
97: in Section~\ref{sec:correlations}.
98: