hep-ex0304030/analysis.tex
1: \section{Analysis method}
2: \label{sec:Analysis}
3: 
4: We select a sample of events with one \B\ meson (\Breco)
5: completely reconstructed in a variety of hadronic decay modes.
6: All the tracks and photon candidates in the event
7: not used to reconstruct the \Breco\ are associated to the other
8: \B\ meson (recoil \B) and are studied to search for a \Btaunu\ signal.
9: 
10: The advantage of having a sample of fully reconstructed
11: \Breco\ mesons is to provide a clean environment of \BpBm\ events
12: with a strong suppression of the combinatorial and continuum backgrounds. 
13: The drawback is a reduction of the data sample due to the low 
14: reconstruction efficiency.
15: 
16: \subsection{Fully reconstructed \B\ sample}
17: 
18: The \Breco\ is reconstructed in a set of hadronic modes
19: that can be summarized as $\B^+ \to D^{(\ast)0} X^+$,
20: where $D^{(\ast)0}$ is a charmed meson and 
21: $X^+$ is a system of charged and neutral hadrons 
22: composed by
23: $n_1 \pipm + n_2 \Kpm + n_3 \piz + n_4 \KS$ 
24: ($n_1=1,...5$, $n_2=0,...2$, $n_3=0, ...2$ and $n_4=0,1$).
25: The $D^{\ast 0}$ is reconstructed in the decay mode
26: $D^0\piz$ and the $D^0$ candidate is reconstructed
27: in four decay modes:
28: $D^0 \to K^-\pip,K^-\pip\piz,K^-\pip\pim\pip,K^0_S\pim\pip$.
29: 
30: The selection of the fully reconstructed 
31: \B\ candidates is made according to the values of
32: two variables:
33: \begin{equation}
34:  \DeltaE = E_B^{\ast}-E_{beam}\;,
35: \end{equation}
36: where $E_B^{\ast}$ is the energy of the \B\ meson and $E_{beam}$ is
37: the beam energy, both in the \FourS\ rest frame;
38: \mes, the energy substituted mass, defined as:
39: \begin{equation}
40: \mes =  \sqrt{[ (s/2+{\bf p}
41:       \cdot{\bf p_\B})^2 / E^2 ]
42:   - |{\bf p_\B}|^2} \; ,
43: \end{equation}
44: where $\sqrt{s}$ is the total energy of the \epem\ system in the
45: \FourS\ rest frame, and $(E, {\bf p})$ and $(E_\B, {\bf p_\B})$ 
46: are the four-momenta of the \epem\ system and the reconstructed 
47: \B\ candidate respectively, both in the laboratory frame. 
48: We require $-0.1<\DeltaE<0.08\gev$ and $\mes>5.21\gevcc$.
49: 
50: For each reconstructed \Breco\ mode $i$ 
51: the \mes\ distribution of the reconstructed \B\ candidates
52: is fit with the sum of an Argus function~\cite{ref:argus} and
53: a Crystal Ball function~\cite{ref:cb}. The Argus function models 
54: the continuum and combinatorial background whereas the Crystal Ball
55: models the signal component, which peaks at the \B\ mass.
56: The purity of the mode $i$ is defined as $S_i/(S_i+B_i)$,
57: where $S_i$ ($B_i$) is the number of signal (background) events
58: with $\mes>5.27\gevcc$, as determined by the fit. 
59: In events with more than one reconstructed charged \B\
60: candidate we select the candidate reconstructed in the mode 
61: with the highest purity.
62: %
63: Figure~\ref{fig:dat_nopresel} shows the \mes\ distribution 
64: for all \Breco\ candidates in data.
65: The yield $N_{\BpBm}$ of the sample containing one \Breco\ 
66: is determined as the area of the fitted Crystal Ball function. 
67: We obtain $N_{\BpBm} = (1.67 \pm 0.09) \times 10^5$.
68: The error on  $N_{\BpBm}$ is dominated by systematics and is
69: discussed in Section~\ref{sec:nbberror}.
70: \begin{figure}[!htb]
71:   \begin{center}
72:     \includegraphics[height=6cm]{dat_nopresel_rho.eps}
73:     \caption{Distribution of the energy substituted mass \mes\ in
74:       data for the fully reconstructed \B\ mesons (histogram).
75:       The solid curve shows the result of the fit (see text).
76:       Also shown are the signal (dashed curve) and the background
77:       (dotted curve) components.} 
78:     \label{fig:dat_nopresel}
79:   \end{center}
80: \end{figure}
81: 
82: We define the signal region on the \Breco\ side to be
83: $-0.09<\DeltaE<0.06\gev$ and $\mes>5.27 \gevcc$
84: and we use the events contained in the sideband $5.21 < \mes < 5.26
85: \gevcc$ as a control sample for continuum and combinatorial background.
86: The same \Breco\ reconstruction technique has been used in other
87: \babar\ analyses like~\cite{ref:knn}.
88: 
89: \subsection{Selection of {\boldmath \Btaunu} decays}
90: \label{sec:selections}
91: 
92: In the events where a \Breco\ is reconstructed
93: we search for decays of the recoil \B\
94: in a $\tau$ plus a neutrino;
95: the $\tau$ lepton is identified in the following decay channels:
96: \taumtoe, \taumtomu, \taumtopi, $\taum \to \pim \piz \nu$, $\taum
97: \to \pim \pip \pim \nu$.
98: All the selection criteria have been optimized to achieve the 
99: best upper limit.
100: 
101: The possible modes in which a \Breco\ meson can be reconstructed 
102: have been classified by decreasing purity. For each reconstructed
103: $\tau$ decay channel we select only \Breco\ mesons reconstructed
104: in the first $n$ modes, where $n$ has been chosen
105: for the best upper limit.
106: 
107: The event total charge $q$ is defined as the sum of the \Breco\ 
108: charge plus the $\tau$ decay products charge.
109: We consider only events with total charge $q=0$ (right sign sample).
110: The complementary sample with $|q|=2$ (wrong sign sample) contains
111: a negligible fraction of the signal and is used as a control sample to
112: test the analysis strategy and the agreement between the
113: selected events in data and the expectation from Monte Carlo
114: simulations.
115: 
116: All the physical quantities mentioned in the following, except 
117: where explicitly stated, refer to the recoil \B.
118: 
119: \subsubsection{Selection of \boldmath \taumtoe, \taumtomu and \taumtopi decays}
120: The \taumtoe, \taumtomu and \taumtopi\ channels are
121: characterized by a single charged track in the final
122: state coming from the primary vertex.
123: 
124: We require:
125: \begin{itemize}
126: \item one reconstructed charged track which has not been
127:       identified as a kaon, 
128:       no reconstructed \piz\ and  no reconstructed \KS;
129: \item at most one photon candidate with total energy in the laboratory
130:   frame less than 110~\mev. Only photons of at least 50~\mev are 
131:   considered; 
132: \item at least 1.2~\gevc of missing momentum in the laboratory frame;
133: \item the track must be identified as a lepton for the
134:   \taumtoe\ and \taumtomu\ selections;
135: \item for the \taumtopi\ selection we require the charged track to have 
136:   a momentum in the recoil \B\ rest frame of at least 1.2~\gevc and 
137:   not be identified as either an electron or a muon.
138: \end{itemize}
139: 
140: \subsubsection{Selection of \boldmath $\taum \to \pim \piz \nu$ decay}
141: The $\taum \to \pim \piz \nu$ decay proceeds via an intermediate
142: $\rho^-$ state. 
143: We require:
144: \begin{itemize}
145: \item one reconstructed track which has not been identified as 
146:   a kaon or a lepton, one reconstructed \piz and no reconstructed \KS;
147: \item at least 1.4~\gevc\ of missing momentum in the laboratory frame;
148: \item at most one photon candidate with total energy in the laboratory
149:   frame less than 100~\mev.
150:   Only photons of at least 50~\mev and not used for the \piz\
151:   reconstruction are considered; 
152: \item the invariant mass of the $\pim \piz$ pair has to be
153:   in the range $0.55 < m_{\pim \piz}< 1.0 \gevcc$;
154: \item for a further rejection of  the continuum background, we require 
155:   the cosine of the angle between the direction of the momentum
156:   of the \Breco\ and the thrust vector of the recoil \B\
157:   to be less than 0.9; the thrust orientation is chosen in order to 
158:   point in the hemisphere opposite to the direction of the recoil \B\ momentum.
159: \end{itemize}
160: 
161: \subsubsection{Selection of \boldmath $\taum \to \pim \pip \pim \nu$ decay}
162: The $\taum$ decays into three charged tracks via two
163: intermediate resonances. The full decay chain is: 
164: $\taum\to a^-_1 \nu$, $a^-_1 \to \pim \rho^0$, $\rho^0 \to \pip \pim$.
165: We require:
166: \begin{itemize}
167: \item three reconstructed charged tracks which have not been 
168:   identified as leptons or kaons, no reconstructed $\piz$ and
169:   no reconstructed \KS;
170: \item at least 1.2~\gevc of missing momentum in the laboratory frame;
171: \item at most one photon candidate with total energy in the laboratory
172:   frame less than 100~\mev.
173:   Only photons of at least 50~\mev and satisfying the quality requirements 
174:   on the lateral moment~\cite{ref:lat} to be between $0.05$ and $0.50$
175:   and on $\Sigma_9 / \Sigma_{25} > 0.9$ are considered.
176:   The lateral moment is a shape quantity for a neutral cluster and 
177:   $\Sigma_9 / \Sigma_{25}$ is the ratio of the energies deposited in the 9 
178:   and 25 crystals closest to the cluster centroid. These quality
179:   requirements are introduced to improve the description of the neutral
180:   energy distribution obtained from Monte Carlo simulations;
181: \item at least one $\pim \pip$ pair with invariant mass in the range 
182:   $0.60<m_{\pip\pim}<0.95 \gevcc$;
183: \item invariant mass of the three pions in the range 
184:   $1.1<m_{\pim\pip\pim}<1.6 \gevcc$;
185: \item total momentum of the three pions in the recoil \B\ rest frame
186:   greater than 1.6~\gevc.
187: \end{itemize}
188: 
189: \subsection{Efficiency and expected background}
190: \label{sec:effbkg}
191: The selection efficiencies for the $\tau$ decay channels we
192: consider in this analysis are determined from detailed Monte
193: Carlo simulations and are summarized in Table~\ref{tab:eff}. 
194: We compute the efficiency as the ratio of the number of events
195: surviving each of our selections and the number of events
196: where a \Breco\ has been reconstructed.
197: The efficiency for the \taumtomu\
198: channel is three times lower than the efficiency for the \taumtoe\ channel
199: because a large fraction of the muon momentum spectrum is below $1\gevc$,
200: where the muon selection efficiency is low. These muons are not recovered
201: by the pion selection because we require the pion momentum to be at least
202: $1.2\gevc$ in order to reject combinatorial and continuum backgrounds.
203: 
204: In the computation of the total efficiency for each selection 
205: we have taken into account the cross-feed from the other $\tau$ decay
206: channels reported in Table~\ref{tab:eff},
207: the requirement that the \Breco\ is reconstructed 
208: in the signal region and that the total reconstructed event charge is zero.
209: 
210: \begin{table}[!htb]
211:   \caption{Efficiency of the different selections (columns) for the 
212:     most abundant $\tau$ decay channels (rows).
213:     In case the efficiency is zero we quote a 90\% C.L. upper limit. 
214:     The last two rows show the total efficiency of the
215:     single selections, weighted by the decay branching fractions, and
216:     the total efficiency. The errors are statistical only.
217:     The total efficiency for each selection is:
218:     $\epsilon_i = \sum_{j=1}^{n_{dec}} \epsilon_i^j f_j$,
219:     where $\epsilon_i^j$ is the efficiency of the selection $i$
220:     for the $\tau$ decay channel $j$, $n_{dec}=7$ is the number of rows in
221:     the table and $f_j ={\cal
222:     B}(\tau\to j)$ are the 
223:     $\tau$ branching fractions from Ref.~\cite{ref:pdg2002}.}
224:   \begin{center}
225:     \include{eff}
226:   \end{center}
227:   \label{tab:eff}
228: \end{table}
229: 
230: The expected background is determined separately in the 
231: right sign and wrong sign samples.
232: It is composed of events from continuum and combinatorial background, 
233: and events with a correctly reconstructed \B\ meson.
234: Simulations of \BzBzb\ events have shown that
235: events where a neutral \B\ is incorrectly reconstructed as a charged
236: \B\ provide a negligible peaking component.
237: 
238: The continuum and combinatorial background is determined from the 
239: number of events in the \mes\ sideband, scaled by the ratio of the areas 
240: of the fitted Argus function in the signal and sideband regions. 
241: Since the number of background events after the full selection is
242: too small to perform a precise fit, we define for each selection
243: criterion a preselection based on the requirements on the number
244: of reconstructed charged tracks and \piz\ mentioned in
245: section~\ref{sec:selections}.
246: We  fit the \mes\ distribution after each preselection and we assume 
247: that the ratio of the fitted Argus in sideband and signal regions,
248: which we use in our estimate of the continuum and combinatorial
249: background, is unchanged after the full selection.
250: The peaking background is determined from Monte Carlo simulations
251: of \BpBm\ events. 
252: 
253: Another source of background originates from \epem\to\tautau\ events.
254: From Monte Carlo simulations we expect $5.8\pm 1.9$ events 
255: from \tautau\ that survive the \taumtopi\ selection. No \tautau\ 
256: event survives in the wrong sign sample.
257: The expected background is summarized in Tables~\ref{tab:bkg-ws}
258: and~\ref{tab:bkg-rs} for the wrong sign and right sign samples,
259: respectively. The systematic corrections on the expected background
260: are described in the next Section.
261: 
262: \begin{table}[!htb]
263:   \caption{Expected background for the wrong sign sample.
264:     The peaking component is estimated from inclusive
265:     \BpBm\ Monte Carlo and the combinatorial plus continuum component
266:     from the data sideband. If no event survives the selection we quote
267:     a 90\% C.L. upper limit on the expected background.   
268:     Systematic corrections are not included.} 
269:   \begin{center}
270: \include{bkg-ws}
271: \end{center}
272: \label{tab:bkg-ws}
273: \end{table}
274: 
275: \begin{table}[!htb]
276:   \caption{Expected background for the right sign sample.
277:     The peaking component is estimated from inclusive \BpBm\
278:     Monte Carlo and the combinatorial plus continuum from the
279:     data sideband. The contribution from the \tautau\ background
280:     is also shown. If no event survives the selection we quote
281:     a 90\% C.L. upper limit on the expected background.
282:     Systematic corrections are not included.}
283:   \begin{center}
284: \include{bkg-rs}
285: \end{center}
286: \label{tab:bkg-rs}
287: \end{table}
288: 
289: In Fig.~\ref{fig:final} we show the neutral energy distribution for
290: events in data and for the expected background. Each distribution
291: refers to a different selection and is obtained applying all the 
292: requirements except the one on the neutral energy. 
293: The plots show no evidence of signal in data.
294: 
295: \begin{figure}[!htbp]
296:   \begin{center}
297:     \begin{tabular}{cc}
298:       \includegraphics[height=5cm]{1pro_05_ecalgam_final.eps} &
299:       \includegraphics[height=5cm]{1pro_06_ecalgam_final.eps} \\
300:       (a) & (b) \\
301:     \end{tabular}
302:     \begin{tabular}{cc}
303:       \includegraphics[height=5cm]{1pro_07_ecalgam_final.eps} &
304:       \includegraphics[height=5cm]{3pro_08_ecalgam_final.eps} \\
305:       (c) & (d) \\      
306:     \end{tabular}
307:     \begin{tabular}{c}
308:       \includegraphics[height=5cm]{rho_10_eneut_final.eps} \\
309:         (e) \\      
310:     \end{tabular}
311:     \caption{Neutral energy distribution in the laboratory frame after all
312:     the selection requirements except the one on the neutral energy for
313:     the channels (a)\taumtoe, (b)\taumtomu, (c)\taumtopi, 
314:     (d)$\tau\to\pim\pip\pim\nu$ and (e)$\tau\to\pim\piz\nu$.
315:     The shaded histogram is the continuum plus combinatorial component 
316:     of the expected background; the solid histogram represents the peaking
317:     component of the expected background and the background from \tautau\ events;
318:     the dots are the data in the \mes\ signal region; the red light shaded 
319:     histogram represents the distribution for Monte Carlo simulated
320:     signal events scaled to \BR(\Btaunu)=$10^{-3}$. The vertical arrow is
321:     the requirement on the neutral energy in each selection.}
322:     \label{fig:final}
323:   \end{center}
324: \end{figure}
325: 
326: \section{Systematic uncertainties}
327: \label{sec:Systematics}
328: The main sources of uncertainty in the determination of the \Btaunu\ branching
329: fraction are:
330: \begin{itemize}
331: \item uncertainty in the determination of the efficiency $\epsilon_i$
332:   for each selection channel;
333: \item uncertainty in the determination of the number of \BpBm\ events
334: with one reconstructed \Breco\, $N_{\BpBm}$;
335: \item uncertainty in the determination of the number of expected
336:   background events $b_i$ in each selection channel.
337: \end{itemize}
338: 
339: \subsection{\boldmath Uncertainty in the selection efficiencies}
340: The main contributions to the systematic uncertainties in the
341: determination of the efficiencies come from systematic uncertainty on
342: tracking efficiency, neutral reconstruction 
343: efficiency and particle identification (PID).
344: Uncertainty in the \piz\ reconstruction efficiency introduces 
345: an additional 5\% contribution to the systematics in the
346: $\taum \to \pim \piz \nu$ selection .
347: The different contributions to the systematic 
348: uncertainties on the selection efficiencies are 
349: reported in Table~\ref{tab:syseff}.
350: 
351: \begin{table}[!htb]
352:   \caption{Contributions to the systematic uncertainty
353:         on the efficiency of the different selections. In the $\piz\nu$
354:         channel the contribution to the neutral systematic uncertainty
355:         due to \piz\ reconstruction is reported explicitly.}
356:   \begin{center}
357:   \include{sys_eff}
358:   \end{center}
359:   \label{tab:syseff}
360: \end{table}
361: 
362: \subsection{\boldmath Uncertainty in the determination 
363:         of $N_{\BpBm}$ }
364: \label{sec:nbberror}
365: We determine $N_{\BpBm}$ as the area of the Crystal Ball function 
366: fitted to the \mes\ distribution (see Fig.~\ref{fig:dat_nopresel}).
367: Using a Gaussian function as an alternative description of 
368: the peak, we obtain a value of $N_{\BpBm}$ which is smaller by 4.5\%. 
369: We assume this relative difference as the 
370: systematic uncertainty on $N_{\BpBm}$. 
371: Using the product of a third order polynomial times an Argus 
372: function as an alternative model for the background,
373: the change in $N_{\Bp\Bm}$ is 0.6\%.
374: 
375: \subsection{\boldmath Uncertainty in the 
376: expected background and systematic corrections}
377: 
378: To take into account possible dependencies of the 
379: fitted Argus shape on a given variable used in the
380: selections, we compute a correction factor as the ratio 
381: of the expected background events passing the requirement
382: on it using two different approaches.
383: In the first approach we use a single sideband to signal
384: scaling factor (see Section~\ref{sec:effbkg}) determined
385: from a \mes fit over the full variable range.
386: In the second approach we divide the range of the variable
387: into bins and determine different scaling factors for
388: each bin.
389: To each correction factor we assign 100\% of the deviation
390: from unity as a systematic uncertainty.
391: 
392: The expected number of background events after the correction
393: is shown in Tables~\ref{tab:bkgcor-ws} and~\ref{tab:bkgcor-rs}
394: for the wrong sign and right sign samples, respectively.
395: It agrees with the number of selected events in data.
396: The total systematic uncertainty amounts
397: to  8.3\% for the \taumtomu\ and \taumtoe\
398: channels, 9.4\% for the \taumtopi\ channel, 9.9\% for the 
399: $\taum \to \pim \piz \nu$ channel, and 6.1\% for the 
400: $\taum \to \pim \pip \pim \nu$ channel.
401: 
402: \begin{table}[!htb]
403:   \caption{Corrected expected background for the wrong sign sample
404:     compared to the number of the selected 
405:     data candidates. The errors are the statistical and systematic
406:     uncertainties.} 
407:   \begin{center}
408:     \include{bkgcor-ws}
409:   \end{center}
410: \label{tab:bkgcor-ws}
411: \end{table}
412: 
413: \begin{table}[!htb]
414:   \caption{Corrected expected  background for the right sign sample
415:    compared to the number of the selected data candidates.
416:    The errors are the statistical and systematic
417:     uncertainties.}
418:   \begin{center}
419:     \include{bkgcor-rs}
420:   \end{center}
421: \label{tab:bkgcor-rs}
422: \end{table}
423: 
424: \section{Upper limit extraction}
425: In order to extract the upper limit on the branching fraction
426: for \Btaunu\ we combine the results of the different selections.
427: 
428: We use the likelihood ratio estimator:
429: \begin{equation}
430:   Q=\frac{{\cal L}(s+b)}{{\cal L}(b)} \: ,
431: \end{equation}
432: where ${\cal L}(s+b)$ and ${\cal L}(b)$ are the likelihood functions
433: for signal plus background and background only hypotheses, respectively. 
434: The likelihood functions ${\cal L}(s+b)$ and ${\cal L}(b)$ are defined
435: as:
436: \begin{eqnarray}
437:   {\cal L}(s+b) & = &
438:   \prod_{i=1}^{n_{ch}}\frac{e^{-(s_i+b_i)}(s_i+b_i)^{n_i}}{n_i!} \: ,
439:   \label{eq:lsb}
440:   \\
441:   {\cal L}(b) & = &
442:   \prod_{i=1}^{n_{ch}}\frac{e^{-b_i}b_i^{n_i}}{n_i!}  \: ,
443:   \label{eq:lb}
444: \end{eqnarray}
445: where $n_{ch}$ is the number of selection channels, $s_i$ and $b_i$
446: are the expected number of signal and background
447: events respectively and $n_i$ is the number of selected events
448: in each channel.
449: In particular, $s_i$ can be written in terms of \BRBtaunu\ as:
450: \begin{equation}
451:   s_i = s \epsilon_i = 
452:   N_{\Bp\Bm}\BRBtaunu\epsilon_i\ ,
453: \end{equation}
454: where $s$ is the total expected number of \Btaunu\ events,
455: $\epsilon_i$ is the selection efficiency for the $i$-th channel, 
456: $N_{\Bp\Bm}$ is the number of \BpBm\ events with one 
457: reconstructed \Breco.
458: 
459: We have no evidence of signal
460: and we set a $90\%$ C.L. upper limit using a fast 
461: parametric Monte Carlo
462: generating random experiments for different values of the
463: branching fraction \BRBtaunu. The confidence level for the signal
464: hypothesis can be computed as:
465: \begin{equation}
466:   {\rm C.L.}_s = \frac{{\rm C.L.}_{s+b}}{{\rm C.L.}_b} = 
467:   \frac{N_{Q_{s+b}\le Q}}{N_{Q_b\le Q}} \: ,
468: \end{equation}
469: where $N_{Q_{s+b}\le Q}$ and $N_{Q_b\le Q}$ are the number of
470: the generated experiments which have a likelihood ratio less than
471: or equal to the measured one, in the background plus signal and
472: background only hypothesis respectively. 
473: The 90\% C.L. upper limit to the branching fraction is the value
474: for which ${\rm C.L.}_s=1-0.9$. We determine:
475: \begin{equation}
476:   \BRBtaunu < 6.3 \times 10^{-4}\:,\:\: 90\%\:{\rm C.L.}
477: \end{equation}
478: 
479: In the extraction of the above limit we have included 
480: the uncertainty on the efficiency by reducing
481: the efficiencies by one standard deviation (adding in 
482: quadrature the statistical and systematic uncertainty), 
483: and we have assumed conservatively the estimate
484: of $N_{\Bp\Bm}$ obtained with a Gaussian model
485: instead of a Crystal Ball.
486: 
487: The statistical and systematic uncertainties on the expected 
488: background can be included in the likelihood definition by folding it 
489: with a Gaussian distribution 
490: having as standard deviation the combined statistical
491: and systematic error on the estimate of $b_i$. 
492: The effect of the uncertainty on the
493: expected background is shown in Fig.~\ref{fig:clerr}.
494: \begin{figure}[!htb]
495:   \begin{center}
496:     \begin{tabular}{cc}
497:     \includegraphics[height=5.5cm]{logqerr.eps} &
498:     \includegraphics[height=5.5cm]{clerr.eps} \\
499:     \end{tabular}
500:     \caption{Distributions of the likelihood ratio (left) and confidence level
501:         (right) as a function of \BRBtaunu. The dashed (solid) curve corresponds
502:         to the case in which the uncertainty on the expected background is included
503:         (not included).}
504:     \label{fig:clerr}
505:   \end{center}
506: \end{figure}
507: Including this uncertainty the upper limit becomes:
508: \begin{equation}
509:   \BRBtaunu < 7.7 \times 10^{-4}\:,\:\: 90\%\:{\rm C.L.}
510: \end{equation}
511: 
512: The central value of the branching fraction corresponds to
513: the minimum in the likelihood ratio distribution. Using
514: $N_{\Bp\Bm}$ obtained with a Crystal Ball model and the
515: central values of the efficiencies,
516: we determine $\BRBtaunu=(1.1^{+3.8}_{-1.1}\times 10^{-4})$.
517: 
518: If we let the number of selected events in each channel
519: fluctuate according to Poisson distribution with 
520: the number of observed events as a mean
521: we obtain the distribution of the possible upper
522: limits shown in Fig.~\ref{fig:cldistrerr}.
523: The central value of this distribution ($7.1\times 10^{-4}$) 
524: represents our sensitivity to the upper limit.
525: 
526: \begin{figure}[!htb]
527:   \begin{center}
528:     \includegraphics[height=6cm]{cldistrerr.eps}
529:     \caption{Distribution of the upper limit
530:       obtained by generating the selected data events according to 
531:       Poisson distributions. The statistical and systematic uncertainties
532:       on the expected background are taken into account.
533:       The dashed line indicates the nominal sensitivity, whereas the
534:       solid line shows the upper limit extracted from the data.}
535:     \label{fig:cldistrerr}
536:   \end{center}
537: \end{figure}
538: 
539: The \babar\ Collaboration performed also another search for the 
540: \Btaunu\ decay using a statistically independent 
541: sample~\cite{ref:wisconsin}.
542: The sample is defined by one \Bp\ meson decaying in
543: $\bar{D}^0 \ell^+\nu_{\ell}X$ final state where $X$
544: is either a photon, \piz\ or nothing.
545: The two upper limits have been combined using the statistical
546: technique described above to combine several channels.
547: The combined upper limit is:
548: 
549: \begin{equation}
550:   \BRBtaunu < 4.1 \times 10^{-4}\:,\:\: 90\% \:{\rm C.L.}
551: \end{equation}
552: 
553: \section{Summary}
554: \label{sec:Summary}
555: A search for \Btaunu\ has been performed
556: in the recoil of a fully reconstructed \Breco\ sample.
557: The analysis uses the following $\tau$ decay channels: 
558: \taumtoe, \taumtomu, \taumtopi, $\taum \to \pim \piz \nu$
559: and $\taum \to \pim \pip \pim \nu$.
560: The results of the search in the different channels have been
561: combined using a likelihood approach.
562: No signal is observed and an upper limit has been set:
563: \[
564:   \BRBtaunu < 7.7 \times 10^{-4}\:,\:\: 90\%\: {\rm C.L.}
565: \]
566: The upper limits set by the two independent 
567: \Btaunu\ searches in the \babar\ experiment have been combined
568: using the statistical technique described in this paper to obtain
569: the following result:
570: \[
571:   \BRBtaunu < 4.1 \times 10^{-4}\:,\:\: 90\%\: {\rm C.L.}
572: \]
573: All results are preliminary.
574: