1:
2: \subsubsection{Proper time determination}
3:
4: The position of the \BtoDs\ ("decay") vertex is reconstructed as described above from the
5: $\ell ,\pstar$ tracks and the beam spot. The decay point of the other \B\ ("tag" vertex)
6: is determined using the remaining tracks in the event, constrained to the decay vertex in
7: the plane orthogonal to the beam axis ($x,y$). This constraint exploits the fact that in the
8: $x$-$y$ view the separation between the \Bz\ and the \Bzb\ is small compared to the
9: experimental resolution; it is applied to remove the tracks from charm or hyperon
10: decay from the vertex fit, which would bias the $\Delta z$ reconstruction. To reduce the additional bias, peculiar to
11: inclusive analyses, due to the un-resolved particles from the \Dz\ produced at the decay vertex,
12: all the tracks lying within a one radian cone around the \pstar\ are rejected from the tag vertex fit.
13: In this way, about 60\% of the \Dz\ tracks are removed, while still retaining $\sim 70\%$ of
14: the \B\ tag tracks.
15:
16: The \Bz\ lifetime is determined by measuring event by event the distance along $z$
17: between the decay and the tag vertex, as already described in section 3.2.4.
18:
19: % which, neglecting the \Bzb\ boost in the $xy$ direction,
20: %is connected to the proper time difference $t$ by the relation:
21: %\be
22: %\DelZ = \beta \gamma c t
23: %\eeq
24:
25: In the Monte Carlo simulation the $\Delta z$ resolution for the signal events is described by
26: two Gaussian distributions: the narrow (wide) component containing about 65\% (35\%)
27: of the events, with a width of 120 (370)\mum\ and a small bias of about 20 (65)\mum.
28: The pull distribution is parametrised by the sum of two Gaussian distributions, the narrow (wide) component
29: contained 82\% (18\%) of the events, with a width of 1.07 (2.4) and a bias of 0.19 (0.7).
30: The same resolution was assumed for \Bp\ events.
31:
32: \subsubsection{Determination of \boldmath \tBz}
33:
34: The \Bz\ lifetime is determined by means of an unbinned maximum likelihood fit, accounting for the
35: event-by-event error determined by the vertex reconstruction algorithm.
36: The fit function was the sum of the probability density function (pdf) for \Bz,\Bp\ and combinatorial
37: background multiplied by the relative fractions for each type in the sample.
38:
39: The $\Delta z$ probability density function for \Bz\ events is described by the
40: convolution of the proper time exponential with the pull resolution function described above.
41: The same function is assumed for \Bp\ events, with a lifetime fixed to the present world average
42: $\tau_\Bp =~ 1.65 \pm 0.04$\ps.
43:
44: The combinatorial background is described by the sum
45: of a zero-lifetime and a non-zero lifetime component, both convoluted with a double Gaussian resolution function.
46: All parameters (lifetime and fraction of non-zero lifetime events, widths, fractions and biases of
47: the resolution function) were determined independently in the data and in the simulation
48: by fitting wrong-charge events with the same program used to determine \tBz. Good consistency is
49: found in the data and in the simulation, except for the rms width of the resolution function, which
50: is 1.06 times larger in data. This discrepancy is attributed to resolution effects.
51:
52: The method was first tested on the dedicated sample of simulated \BtoDs\ events after applying the selection requirements
53: described above. Due to the influence of the residual tracks from the un-reconstructed \Dz, the
54: measured lifetime in Monte Carlo is smaller than the generated one.
55: The resulting reconstruction bias $\DTau\ = 0.217 \pm 0.010$\ps\ for the lifetime is
56: defined as the difference between the generated value of 1.560\ps\ and the fitted
57: value $\tBz^{raw,MC} = 1.343 \pm 0.010$\ps.
58:
59: The final lifetime result for the data \tBz\ was obtained by
60: adding the bias \DTau\ determined previously in the Monte Carlo to the raw fit result
61: $\tBz^{raw,Dt} = 1.405 \pm 0.016$\ps.
62: The lifetime obtained in this way is:
63: \be
64: \tBz = 1.622 \pm 0.016 ~\ps
65: \eeq
66: where the error is statistical only. Figure~\ref{f:lds_fitres} shows a comparison between the
67: data and the fit result.
68:
69: \begin{figure}[!htb]
70: \begin{center}\begin{tabular}{cc}
71: \includegraphics[height=7cm]{lds/fit_lin.eps}&
72: \includegraphics[height=7cm]{lds/fit_log.eps}
73: \end{tabular}
74: \caption{ Lifetime fit result on a linear (left) and a logarithmic scale (right).
75: The inner dark area is the \Bp\ contribution, the lighter area is the combinatorial background,
76: the line is the result of the fit. Dots with error bars represent the data.}
77: \label{f:lds_fitres}\end{center}\end{figure}
78: