1: The prediction of vector boson \pt\ distributions at hadron colliders
2: has long been an active subject~\cite{th:wzpt3,th:wzpt2,th:wzpt1}.
3: It is also a crucial input
4: for the \W mass analysis, especially when using the \ptl\
5: observable. We discuss below the impact of \ptw\ uncertainties on
6: the \W mass determination in this hypothesis.\\
7:
8: \noindent
9: The measurable \ptw\ and \ptz\ distributions are the result of several
10: effects, most notably the repeated, partly non-perturbative parton
11: radiation occurring in the transition from the low-$Q^2$ proton
12: towards the hard process (commonly referred to as parton showers, or
13: soft gluon resummation). Another source is the transverse momentum intrinsic
14: to the partons in the proton. We choose not to discuss these effects
15: separately. Rather, reckoning that although \W and \Z production differ
16: in several respects (the coupling to initial partons is different in
17: both phase space and flavour), the non-perturbative mechanisms are
18: universal, we evaluate how precisely their combined effect can be
19: measured in neutral current events, and how this improves the \W
20: predictions. Notice that heavy flavour PDF have caused only a small
21: decorrelation between \W and \Z events in the previous section; this is
22: assumed to remain true in this discussion.\\
23:
24: \noindent
25: First, the relation between the bias in the modeling of \ptw\ and the
26: measurement of \MW\ is investigated by applying scaling factors to
27: the \ptw\ distributions in our pseudo-data, deducing the
28: corresponding \ptl\ distributions, and fitting \MW\ against
29: un-distorted templates. The bias in \MW\ appears to be a linear
30: function of the \ptw\ mis-modelling, with a slope of order 0.3, meaning
31: a 3~\MeV\ bias on \ptw\ results in a 1~\MeV\ bias on \MW, when exploiting the
32: \ptl\ distribution. When \mtw\ is used, the effect is negligible.\\
33:
34: \noindent
35: Neutral current dilepton events allow to measure the \ptll\
36: distribution, as a function of mass, over a large
37: mass range. Assuming usual selections, this distribution will be
38: measured precisely for $30 < M_{\ell\ell} < \sim 200$~\GeV. This large lever
39: arm, in addition to the very precise determination of the \ptll\
40: distribution on the \Z peak, provides a precise control of $d\sigma/d\ptll$ when
41: $M_{\ell\ell} \sim \MW$. This is illustrated in Figure~\ref{ptw}, which
42: displays the dilepton mass dependence of its average transverse
43: momentum, $<\ptll>$, as predicted by {\tt PYTHIA}.\\
44:
45: \begin{figure}
46: \begin{center}
47: \includegraphics[width=0.6\textwidth]{figures/dSdM.eps}
48: \includegraphics[width=0.6\textwidth]{figures/pt_vs_m.eps}
49: \caption{\label{ptw} Top : Dilepton invariant mass spectrum,
50: from inclusive neutral current events ($\gamma$ and \Z exchange
51: are included). Bottom : dilepton average \pt\ as a function of the
52: dilepton invariant mass. The \W-mass region is strongly
53: constrained by the lever arm provided by the \Z peak and the
54: Drell-Yan rise at low mass (note the improved precision in these
55: regions). The points correspond to a measurement with 10~\ifb.}
56: \end{center}
57: \end{figure}
58:
59: \noindent
60: On the \Z peak, \ptll\ will be known to about 7 \MeV\ with an
61: integrated luminosity of 10 fb$^{-1}$. Thanks to the Drell-Yan
62: continuum, the accuracy in the region of \MW\ is still
63: $\sim$8~\MeV. This precision can be used to constrain the
64: non-perturbative parameters governing the parton shower or resummation
65: computations, and to predict the \ptw\ distribution with similar accuracy.
66: This leads to an uncertainty on \MW\ of about 3 \MeV.\\
67:
68: \noindent
69: Arguably, the \ptw\ distribution cannot be summarized by its mean
70: value. However, in the low \ptw\ region (selected by the recoil cut,
71: cf. Section~\ref{sec:methods}), it can be empirically described by a
72: two-parameter function. As an exercise, the mass-dependence of the
73: parameters were determined on Drell-Yan events, their values and
74: uncertainties in the \MW\ region were used to produce \ptl\
75: pseudo-data as above, and corresponding fits to \MW\ were
76: performed. The spread in \MW\ resulting from the uncertainty in the
77: empirical parameters was found compatible with the above estimate.
78: