1: The \W rapidity distribution is essentially driven by the proton parton
2: density functions (PDFs).
3: Our study is based on the CTEQ6.1 structure functions sets~\cite{th:cteq6},
4: which provide, in addition to the global best fit, PDFs corresponding
5: to the variation of each diagonal parameter (i.e, the linear
6: combination of input parameters that diagonalize the covariance
7: matrix) within its estimated uncertainty. The PDF-induced
8: uncertainty for an observable is obtained by computing its value
9: with all sets, taking the central value as given by the best fit, and
10: quadratically summing the biases (w.r.t the best fit value) obtained
11: from the uncertainty sets.\\
12:
13: \noindent
14: As illustrated in Figure~\ref{fig:pdfsyst} (see also~Ref.~\cite{lhc:cmsMw}), the
15: current PDF uncertainties induce an uncertainty in the \W rapidity
16: distributions which, through acceptance effects, propagates a
17: systematic uncertainty on the \W mass determination of $\sim$25~\MeV. We
18: present below an attempt to estimate how this will improve with the LHC data.\\
19:
20: \begin{figure}
21: \begin{center}
22: \includegraphics[width=0.6\textwidth]{figures/MwPdf.eps}
23: \caption{\label{fig:pdfsyst} Bias on \MW\ obtained when varying the
24: proton PDFs within their uncertainties. Each point on the abscissa
25: correponds to a given PDF set: set 0 is the best fit, and gives 0
26: bias up to the statistical uncertainty of the fit; sets 1-40 are the
27: uncertainty sets, each inducing a given bias on \MW. The total
28: uncertainty on \MW\ is given by the quadratic sum of the biases,
29: giving $\delta\MW \sim 25 \MeV$.}
30: \end{center}
31: \end{figure}
32:
33: \noindent
34: At the LHC, \W and \Z particles are essentially produced through sea
35: quark interactions; the influence of valence quarks is small. Low-$x$,
36: high-$Q^2$ sea quarks mainly evolve from higher $x$, lower $Q^2$
37: gluons, and a consequence from perturbative QCD flavour symmetry is
38: that up to initial asymmetries and heavy-quark mass effects, the
39: different quark flavours should be represented democratically. This
40: then implies that the impact of sea quark PDF uncertainties on \W and \Z
41: production should be very similar. In other words, when varying PDFs
42: within their uncertainties, one expects a strong correlation
43: between the induced variations of the \W and \Z distributions.\\
44:
45: \noindent
46: This is confirmed by Figure~\ref{pdf1}~\footnote{This plot is
47: reminiscent of Figure~2 in~\cite{th:Nadolsky2004}, displaying
48: similar correlations in the production rates. Note that for our
49: purpose, normalizations are irrelevant and we are interested only
50: in the distributions.}.
51: On the left, the correlation between the widths of the \W and \Z boson
52: rapidity distributions is displayed. We choose to use the
53: distributions RMS, denoted $r_y^\W$ and $r_y^\Z$, to quantify
54: their width. The
55: current CTEQ6.1 prediction, $r_y^\Z = 2.16 \pm 0.03$, will be
56: refined to a precision of $\delta r_y^\Z = 0.001$. Exploiting
57: Figure~\ref{pdf1} (right), which quantifies the correlation between
58: $r_y^\W$ and $r_y^\Z$, this can be translated into a prediction of
59: the \W boson rapidity distribution, $\delta r_y^\W = 0.0013$, to
60: be compared to the current prediction $r_y^\W = 2.24 \pm 0.03$.\\
61:
62: \begin{figure}
63: \begin{center}
64: \includegraphics[width=0.48\textwidth]{figures/ywyz.eps}
65: \includegraphics[width=0.48\textwidth]{figures/ywzratio.eps}
66: \caption{\label{pdf1} Left : correlation between the spreads (RMS)
67: $r_y^\W$ and $r_y^\Z$ of the \W and \Z rapidity
68: distributions, when varying the CTEQ6.1 PDFs within their estimated
69: uncertainties. The fitted pseudo-data are scaled to an integrated
70: luminosity of~10~fb$^{-1}$. Right : distribution of the ratio
71: $r_y^\W/r_y^\Z$, again varying the PDFs within their
72: uncertainties. The spread of the ratio distribution is $4 \times
73: 10^{-4}$.}
74: \end{center}
75: \end{figure}
76:
77: \begin{figure}
78: \begin{center}
79: \includegraphics[width=0.6\textwidth]{figures/yz_discr.eps}
80: \caption{\label{pdf2} The line histograms represent two extreme
81: predictions for the \Z rapidity distribution, as given by the CTEQ6.1
82: PDF sets. The points are pseudo-data, obtained with the central
83: set, and scaled to an integrated luminosity of~10~fb$^{-1}$.}
84: \end{center}
85: \end{figure}
86:
87: \noindent
88: One thus expects an improvement on the \Z rapidity distribution by a
89: factor $\sim$30. This is also illustrated in Figure~\ref{pdf2}, where
90: two extreme predictions (with current knowledge) of the \Z rapidity
91: distribution are compared with an example distribution representing
92: the same measurement. Given the residual decorrelation between the \W
93: and \Z distributions, this translates into an improvement on the \W
94: rapidity distribution by a factor $\sim 23$.\\
95:
96: \noindent
97: Starting with $\delta \MW(y_\W) \sim$25~\MeV, putting in a precise
98: measurement of the \Z rapidity distribution at the LHC, and exploiting
99: the strong correlation between the \W and \Z production mechanisms, we
100: thus anticipate a final uncertainty from the description of the \W
101: rapidity distribution of $\delta \MW(y_\W) \sim$1~\MeV.\\
102:
103: \noindent
104: In practice, the analysis will of course proceed $via$ a formal QCD
105: analysis to the LHC data: the measured \Z differential
106: cross-section ${\mathrm d}\sigma/{\mathrm d}y$, together with
107: other measurements (see below), will be fed to parton distribution fits, and
108: the systematic $\delta\MW(\yw)$ from the improved PDF sets will be
109: evaluated as above. The present discussion however allows to estimate
110: the expected improvement while avoiding these complications.\\
111:
112: \noindent
113: Let us also note that \Z rapidity distribution can be analyzed over a
114: domain that fully includes the range relevant for \W
115: production. In ATLAS (as in CMS), the usual \Z acceptance, given by
116: $|\eta_\ell| < 2.5$ for both decay leptons, can be extended in the electron
117: channel by allowing one of the electrons to be detected within $|\eta_e|
118: < \sim 4.9$. In addition, high-rapidity \Z events will be produced and
119: detected at LHCb (for example, the geometric acceptance of the muon
120: detector is approximately $2.1 < |\eta_\mu| < 4.8$). Accounting for
121: this, and as illustrated in Figure~\ref{pdf3}, the \W rapidity
122: range selected for the \MW\ measurement is entirely included in the \Z
123: one. This remains true in terms of the parton momentum fractions.\\
124:
125: \begin{figure}
126: \begin{center}
127: \includegraphics[width=0.6\textwidth]{figures/wz_acc.eps}
128: \caption{\label{pdf3} Upper plot: the outer histogram represents the complete
129: rapidity distribution for \W production at the LHC; the inner
130: histogram represents the range selected by the condition $|\eta_\ell| <
131: 2.5$. Lower plot: the outer histogram represents the complete
132: rapidity distribution for \Z events. The innermost histogram is
133: obtained requiring two decay leptons within $|\eta_\ell| < 2.5$; the
134: intermediate histogram is obtained when allowing one electron within
135: $|\eta_\ell| < 4.9$. The two symmetric histograms at high rapidity
136: correspond to the LHCb muon acceptance.}
137: \end{center}
138: \end{figure}
139:
140: \noindent
141: We conclude this section with some caveats. The above results partly
142: are a consequence of the assumed flavour and charge symmetry in the
143: low-$x$ proton; notably, the parton parametrisations used in the fits
144: used above assume that $d(x) = \bar{d}(x) = u(x) = \bar{u}(x)$ at
145: low-$x$, and $s = \bar{s}$ at all $x$. This implies the strong
146: correlation discussed above, since the \Z production rate is
147: proportional to $u\bar{u} + d\bar{d} + \ldots$, and the \W rate
148: is proportional to $u\bar{d} + d\bar{u} + \ldots$. It is thus
149: important to quantify the dependence of our result on these hypotheses.\\
150:
151: \noindent
152: The anti-quark flavour asymmetry $\bar{u}-\bar{d}$ was measured to be non-0
153: in the region $0.015 < x < 0.35$, and $Q^2 \sim 50
154: \GeV^2$~\cite{ex:udbarNA51,ex:udbarE866}, in contradiction with the
155: flavour symmetry assumption. The relative asymmetry,
156: $(\bar{u}-\bar{d})/((\bar{u}+\bar{d})$, is however of the order $\sim
157: 10^{-2}$, decreasing towards higher $Q^2$. Starting from
158: $\bar{u}=\bar{d}$ and full correlation between \W and \Z
159: production (\ie\ \W and \Z distributions have the same
160: rate of change under PDF variations), $\bar{u}\ne\bar{d}$ induces a
161: decorrelation of order $(\bar{u}-\bar{d})/(\bar{u}+\bar{d})\times
162: (u-d)/(u+d)$, where both factors are of order $10^{-2}$ (see for
163: example Figure~1 in~\cite{th:cteq6}). Hence, even in the presence of
164: non-vanishing $\bar{u}-\bar{d}$, the freedom of the \W
165: distributions is very limited once \Z ones have been precisely
166: measured. We thus assume that our estimates remain correct;
167: nevertheless, measurements of the \W charge asymmetry, sensitive
168: to $\bar{u}-\bar{d}$, will allow to verify this hypothesis.
169: Additional information will be provided by measuring \MW\ in $\W^+$ and $\W^-$
170: events separately.\\
171:
172: \noindent
173: The proton strangeness asymmetry, $s(x) - \bar{s}(x)$, is constrained
174: by neutrino scattering
175: data~\cite{ex:ssbarNUTEV,ex:ssbarBPZ,ex:ssbarCTEQ}. The relative asymmetry
176: is rather small, even at low $Q^2$: $(s-\bar{s})/(s+\bar{s}) \sim
177: 10^{-2}$ at $Q^2 = 10 \GeV^{2}$. It will only become smaller at $Q^2
178: \sim \MW^{2}$, where most of the strange sea is generated
179: radiatively. We consider, as above, that the contribution of the
180: asymmetry is small in terms of the overall \W production and its
181: uncertainty. However, the impact on the \MW\ measurement would need to
182: be studied specifically. At the LHC, the analysis of $\W^{-/+} +
183: c/\bar{c}$ production should provide additional insight.\\
184:
185: \noindent
186: Finally, one may argue that the influence of heavy quark PDFs on \W and
187: \Z production is different, thus a source of decorrelation between the
188: two processes. The charm quark contribution to \W
189: production is significant ($\sim (V_{cs}c\bar{s} + V_{cd}c\bar{d} + c.c.)$), but
190: smaller for \Z production ($\sim c\bar{c}$). On the other hand, the $b$-quark
191: content contributes to \Z production ($\sim b\bar{b}$), but
192: negligibly to \W production ($\sim (V_{cb}c\bar{b} + c.c.)$), due to the smallness of
193: the off-diagonal third generation CKM matrix elements. These
194: differences are however accounted for by the present analysis, since
195: the heavy quark PDFs are included the CTEQ6.1 PDF sets; heavy flavours
196: are actually understood to cause in part the small decorrelation
197: between the \W and \Z boson distributions. Our conclusions thus remain
198: unchanged. \\
199:
200: \noindent
201: The present study has been repeated using the MRST2001 PDF
202: sets~\cite{th:mrst2001}. The same correlation is observed between $r_y^\W$
203: and $r_y^\Z$, and the same result is obtained. Non-global parton
204: density fits, such as those performed by the H1 and Zeus experiments,
205: are based on similar hypotheses and claim slightly smaller
206: uncertainties~\cite{heralhcA}, again preserving our result.
207: Finally, during the course of this work, CTEQ6.5 PDF sets became
208: available~\cite{th:cteq65}, which improves on the treatment of heavy
209: quark masses in the QCD evolution. The flavour symmetry assumptions
210: are however unchanged, so that the present discussion is not affected.
211: