hep-ex0408068/mxq2.tex
1: \section{Two-dimensional \mx-\Q\ analysis}
2: \label{sec:mxq2}
3: The \mx\ analysis is systematically limited by the dependence on the 
4: shape function. This can be overcome by selecting a phase space region where the 
5: shape function effects are small, namely the region at large
6: \Q\ values~\cite{Bauer:2001yb}. 
7: Therefore we find a trade-off between the statistical 
8: and theoretical uncertainties by loosening the \mx\ cut and applying a
9: \Q\ one.
10:  Moreover, since most of the theoretical uncertainties are due to the 
11: extrapolation from a selected kinematic region to the full phase space,  
12: measurements of partial branching fractions in different regions of phase space and their extrapolation to the 
13: full phase space can serve as tests of the theoretical calculations and models. 
14: 
15: 
16: In order to extract the partial charmless semileptonic branching ratio in a given region of the 
17: \mx-\Q\ plane $\Delta \BR(\Bxulnu)$,  we define as signal
18: the events with true values of the kinematic variables in the chosen
19: region, treating as background those that migrate from outside this region
20: because of the resolution.
21:  This means that in applying Eq.~\ref{eq:ratioBR} we include
22:  the \btoulnu\ events outside the signal region in $BG_u$ and the quoted efficiencies refer only to events 
23: generated in the chosen (\mx-\Q) region. These efficiencies are computed on Monte Carlo, and therefore are based on 
24: the DFN model. However, the associated theoretical uncertainty on the final result is small compared to the 
25: extrapolation error to the full phase space. We divide the events into two-dimensional bins of \mx\ and \Q, 
26: we fit the \mes\ distribution to extract the yield in each bin, and 
27: we perform a two-dimensional binned fit of the entire \mx-\Q\ distribution in order to extract the signal and background
28:  components. The result of the fit 
29: is shown in Fig.~\ref{fig:mxq2fit}.
30: %
31: \begin{figure}
32: \centerline{\epsfig{figure=plots/q2profile.eps,width=16.cm}}
33: \caption{Distributions of \Q\ in bins of \mx. Points are data, the blue, light gray and yellow histograms  
34: represent  the fitted contributions from \btoulnu\ events with
35: true  \mx$<1.7\gevcc$, \Q$>8\gevccsq$ ,  \btoulnu\
36: events not satisfying these requirements,  and background events, respectively.
37: }
38: \label{fig:mxq2fit}
39: \end{figure}
40: %
41: 
42:  Fig.~\ref{fig:mxq2dbr}a and Table~\ref{tab:partial} show, for $\mx<1.7\gevcc$,
43: the partial branching fraction $\Delta \BR(\Bxulnu)$ as a function of the minimum \Q\ cut.
44: 
45: 
46: 
47: \begin{table}[!b]
48: \begin{center}
49: \caption{Partial branching fraction measurements (in 10$^{-3}$ units) for 
50: $\mx<1.7$ \gevcc and $\Q>\Q_{cut}$, as a function of $\Q_{cut}$. 
51: The different sources of uncertainty (as described in \ref{sec:sys}) are also reported.}
52: \vspace{0.1in}
53: \begin{tabular}{|l|c|c|c|c|c|c|c|c|} 
54: \hline
55: $\Q_{cut}>$ & $\Delta \BR(\Bxulnu)$ & $\sigma_{stat}$ & $\sigma_{det}$ & $\sigma_{breco}$ & $\sigma_{bkg}$ & $\sigma_{theo}$ & $\sigma_{ul\nu}$ & $\sigma_{MCstat}$ \\ \hline
56: 0     &   1.68    &  0.22 &  0.15 &  0.08 &  0.12 & -0.045 +0.035 &  0.137  &  0.08 \\
57: 2     &   1.52    &  0.20 &  0.16 &  0.07 &  0.11 & -0.028 +0.036 &  0.110  &  0.07 \\
58: 4     &   1.33    &  0.18 &  0.11 &  0.06 &  0.10 & -0.040 +0.026 &  0.116  &  0.06 \\
59: 6     &   1.10    &  0.16 &  0.14 &  0.05 &  0.08 & -0.022 +0.018 &  0.083  &  0.05 \\
60: 8     &   0.88    &  0.14 &  0.09 &  0.04 &  0.06 & -0.028 +0.009 &  0.053  &  0.05 \\
61: 10    &   0.55    &  0.11 &  0.03 &  0.02 &  0.04 & -0.006 +0.019 &  0.027  &  0.03 \\
62: 12    &   0.41    &  0.09 &  0.04 &  0.02 &  0.03 & -0.010 +0.000 &  0.033  &  0.03 \\
63: 14    &   0.21    &  0.06 &  0.01 &  0.01 &  0.02 & -0.012 +0.012 &  0.018  &  0.02 \\
64: 
65: \hline
66: \end{tabular}
67: \label{tab:partial}
68: \end{center}
69: \end{table}
70: 
71: 
72: We convert the measured $\Delta \BR(\Bxulnu)$ into \Vub\ by
73: \begin{equation}
74: \label{eq:dbrvub}
75:  |V_{ub}| = \sqrt{\frac{192 \pi^3}{\tau_B G_F^2 m_b^5}\frac{\Delta \BR(\Bxulnu)}{G}} 
76: \end{equation}
77: where $\tau_B = 1.61 ps$ and $G$ is a theoretical parameter calculated in the BLL approach~\cite{Bauer:2001yb}. 
78: The first factor under the square root is 
79: 192$\pi^3/(\tau_B G_F^2 m_b^5)=0.00779$. 
80: The measured \Vub\ as a function of the \Q\ cut is shown in
81: Fig.~\ref{fig:mxq2dbr}b for the acceptances computed by BLL and by
82: DFN. Note that, since the operator product expansion breaks down when going to low \Q, the BLL calculation is only possible for higher values of
83: \Q. 
84: 
85: 
86: \begin{figure}
87: \centerline{\epsfig{figure=plots/pbf_theworks__s_12.eps,height=7cm}}
88: \centerline{\epsfig{figure=plots/vubpbf_theworks__s_12.eps,height=7cm}}
89: \begin{center}
90: \caption{
91: (a) Measured partial branching ratio for $\mx<1.7$ \gevcc and $\Q>\Q_{cut}$, as a function of $\Q_{cut}$. 
92: The error bar is the sum in quadrature 
93: of statistical, systematical and theoretical uncertainties. 
94: (b) Measured value of \Vub for $\mx<1.7\gevcc$ as a function of the \Q\ cut
95: applied when using acceptances from DFN (open points) and BLL
96: (solid points). The error bars include the statistical, systematic and theoretical 
97: uncertainties, added in quadrature.}
98: \label{fig:mxq2dbr}
99: \end{center}
100: \end{figure}
101: 
102: The error on the acceptance as computed by BLL increases for tighter cuts on 
103: \Q. For smaller values of \Q, the shape function effects increase. 
104: In the signal region $\Q>8\gevccsq$, $\mx<1.7\gevcc$
105: we obtain: 
106: \begin{equation}
107:  \Delta \BR(\Bxulnu,\mx<1.7 \gevcc, \Q>8  \gevccsq) = (0.88 \pm 0.14 (\rm stat.) \pm 0.13 (\rm sys.) \pm 0.02 (\rm theo.)) \times 10^{-3}. 
108: \label{mxq2res}
109: \end{equation}
110: To extract \Vub, we take $G$ as computed by BLL and rescale it to the 
111: $b$-quark mass as measured by \babar\cite{Aubert:2004aw}, 
112: obtaining $G = 0.282 \pm 0.053$, corresponding to an acceptance $\epsilon_{BLL}=0.325\pm 0.061$. 
113: Eq.~\ref{eq:dbrvub} yields
114: \begin{equation}
115: |V_{ub}|  = (4.92 \pm 0.39 (\rm stat.) \pm 0.36 (\rm sys.) \pm 0.46 (\rm theo.)) \times 10^{-3}.
116: \end{equation}
117: 
118: In the DFN model the calculated acceptance is $\epsilon =  0.337^{+0.037}_{-0.074}$
119: and by using Equation~\ref{eq:vubextr} we obtain 
120: $\Vub= (4.85 \pm 0.39 (\rm stat.) \pm 0.36 (\rm sys.) $ $^{+0.54}_{-0.29} (\rm theo.)) \times 10^{-3}$, in  
121: agreement with the extraction based on BLL, as well as with the result form the one-dimensional \mx\ fit. 
122: Figure~\ref{fig:mxq2dbr} shows the measured values for \Vub\ as a function of the \Q\ cut for 
123: $\mx<1.7$ \gevcc, showing good consistency between the different cuts and theoretical framework. 
124: Checks were done also with a looser ($\mx<1.86$ \gevcc) 
125: and a tighter ($\mx<1.5$ \gevcc) 
126: cut on \mx, and they give consistent results. 
127: 
128: \section{Results based on Belle's estimate of the SF parameters} 
129: \label{sec:belle}
130: We report here the results obtained with the SF parameters as estimated from the  $b\ra s \gamma$ photon 
131: energy spectrum measured 
132: by Belle (see Sec.~\ref{sec:theosys}) and we reinterpret our results by using them in the \mx\ and \mx-\Q\ analyses. 
133: The  acceptance obtained for the DFN model is lower, and therefore
134:  the charmless semileptonic branching fraction and  \Vub\ are higher. 
135: The theoretical systematics due to the shape function parameters is reduced, due to the 
136: significantly better precision. 
137: 
138: 
139: For the \mx\ analysis we get 
140: 
141: \begin{equation}
142: \rusl=(2.81\pm 0.32(\rm stat.)\pm 0.31(\rm sys.)^{+0.23}_{-0.21} (\rm theo.))\times 10^{-2}
143: \end{equation}
144: 
145: which translates into 
146: 
147: \begin{equation}
148:  \Vub=(5.22\pm 0.30(\rm stat.)\pm 0.31 (\rm sys.)^{+0.22}_{-0.20}(\rm SF)\pm 
149: 0.25(\rm pert+1/mb^3))10^{-3}.
150: \end{equation}
151: \begin{figure}
152: \includegraphics[width=0.48\textwidth]{plots/spec_976belle.eps} \hfill
153: \includegraphics[width=0.48\textwidth]{plots/int_976belle.eps}
154: \caption[Unfolded \mx\ spectrum corrected for bias]{The unfolded spectrum
155: (left) and its cumulative distribution (right) as a function of \mx. 
156: The spectrum and the cumulative distribution from MC with the Belle best fit $\lbarsf=
157: 0.66\gevcc$ and $\lonesf = -0.4\gevccsq$ is shown in green. 
158: The orange and violet spectra and cumulative distributions correspond to the two extreme points in the 
159: Belle ellipse (see Fig.~\ref{fig:ellipses}), $\lbarsf=
160: 0.600\gevcc$ and  $\lbarsf=
161: 0.748\gevcc$ and to $\lonesf = -0.66(\gevcc)^2$  and  $\lonesf =
162: -0.28(\gevcc)^2$, respectively. 
163: In the left plot 
164: black errors are only statistical, while the red ones always include systematics.}
165: \label{unf:result2}
166: \end{figure}
167: 
168: As far as the unfolding is concerned, 
169: Fig.~\ref{unf:result2} compares the measured spectra with the
170:  distributions corresponding to the SF parameters measured by Belle.
171: 
172: The  partial branching fraction measurements as a function of the \Q\ cut obtained by the 
173: \mx-\Q analysis are reported, for
174: \mx$<1.7$ \gevcc, in Table~\ref{pbfbelle}. 
175: 
176: \begin{table}[!b]
177: \begin{center}
178: \caption{BELLE ellipse: Partial branching fraction $\Delta\BR(\Bxulnu)$ measurements (in 10$^{-3}$ units) 
179: for different \Q\ cuts. \mx\ is required to be less than 1.7\gevcc. 
180: The different sources of uncertainties are also reported.}
181: \vspace{0.1in}
182: \begin{tabular}{|l|c|c|c|c|c|c|c|c|} 
183: \hline
184: $\Q_{cut}>$ & $\Delta \BR(\Bxulnu)$ & $\sigma_{stat}$ & $\sigma_{det}$ & $\sigma_{breco}$ & $\sigma_{bkg}$ & $\sigma_{theo}$ & $\sigma_{ul\nu}$ & $\sigma_{MCstat}$ \\ \hline
185: 0  & 1.740 &  0.231 &  0.159 &  0.078 &  0.129 & -0.042 +0.026 &  0.158 &  0.078 \\
186: 2  & 1.584 &  0.205 &  0.165 &  0.071 &  0.117 & -0.037 +0.027 &  0.145 &  0.068 \\
187: 4  & 1.381 &  0.186 &  0.113 &  0.062 &  0.102 & -0.036 +0.026 &  0.140 &  0.061 \\
188: 6  & 1.135 &  0.161 &  0.144 &  0.051 &  0.084 & -0.025 +0.017 &  0.105 &  0.053 \\
189: 8  & 0.896 &  0.143 &  0.091 &  0.040 &  0.066 & -0.017 +0.012 &  0.064 &  0.047 \\
190: 10 & 0.566 &  0.113 &  0.026 &  0.025 &  0.042 & -0.006 +0.013 &  0.041 &  0.036 \\
191: 12 & 0.406 &  0.085 &  0.038 &  0.018 &  0.030 & -0.002 +0.003 &  0.034 &  0.026 \\
192: 14 & 0.207 &  0.059 &  0.014 &  0.009 &  0.015 & -0.007 +0.002 &  0.026 &  0.019 \\
193: \hline
194: \end{tabular}
195: \label{pbfbelle}
196: \end{center}
197: \end{table}
198: 
199: The measurement of the partial branching fraction $\BR(\Bxulnu)$ 
200: in the region limited by 
201: $\mx < 1.7 \gevcc, ~~~~ \Q>8 (\gevcc)^2$ is 
202: \begin{eqnarray*}
203: \Delta\BR(\Bxulnu,\mx<1.7 \gevcc, \Q>8 \gevccsq)=(0.90 \pm 0.14(\rm stat.) \pm 0.14(\rm sys.)^{+0.01}_{-0.02}(\rm theo.))\times 10^{-3}.\\
204: \end{eqnarray*}
205: 
206: By using $G = 0.282 \pm 0.053$ from BLL, we get 
207: \begin{eqnarray*}
208: |V_{ub}| & = & (4.98 \pm 0.40(\rm stat.) \pm 0.39(\rm syst.) \pm 0.47(\rm theo.)) \times 10^{-3}. \\
209: \end{eqnarray*}
210: The DFN acceptance computed at \mx$<1.7$ \gevcc\ and \Q$>8$ (\gevcc)$^2$ with the Belle SF parameters is 
211: $\epsilon =  0.300^{+0.023}_{-0.028}$. This gives in the DFN framework 
212: $|V_{ub}| = (5.18 \pm 0.41_{stat} \pm 0.40_{syst} ~^{+0.25}_{-0.20~~~theo}) \times 10^{-3}$. 
213: 
214: 
215: Figure~\ref{vubscanbelle} shows the results for \Vub as a function of the \Q\ cut for \mx$<1.7$ \gevcc, for both DFN and BLL. The 
216: two models are still consistent within the present accuracies. 
217: The stability of the result and the agreement between the two methods seems to indicate that OPE is still valid in this \Q\ range.
218: \begin{figure}[!t]
219: \begin{centering}
220: \centerline{\epsfig{figure=plots/bellevub.eps,height=7cm}}
221: \caption{BELLE ellipse: Values for $|$V$_{ub}|$ as a function of the \Q\ cut for \mx$<1.7$ \gevcc\ by taking 
222: acceptances from Bauer, Ligeti and Luke (points) and De~Fazio-Neubert (squares). The error is the sum 
223: in quadrature of the statistical, systematic and theoretical uncertainties.}
224: \label{vubscanbelle}
225: \end{centering}
226: \end{figure}
227: 
228: