1: An unbinned likelihood fit is performed to mass, $ct$, and $\vrho$

2: in order to extract the decay amplitudes $A_{0,\parallel,\perp}$ and

3: decay widths $\Gamma$ ($\Gamma_H$ and $\Gamma_L$ in the case of the $B^0_s$).

4: Inclusion of the mass information in the fit is crucial for separation

5: of the signal from the background.  The mass distribution is modeled

6: with a Gaussian for the signal peak and a linear shape for the

7: background.  The mass-measurement uncertainty is incorporated for each

8: candidate.  The probability density function for $ct$ includes positive

9: exponentials for the signal, a $\delta$-function for the prompt

10: background (which is about 85\% of the total) and a set of exponentials

11: for positive and negative decay lengths which describe a short-lived

12: background component due to mis-measured vertices and a long-lived

13: contribution due to incorrectly reconstructed heavy-flavor decays.  Each

14: contribution is convoluted with a Gaussian resolution function, the

15: width of which is proportional to the uncertainty of the candidate's

16: $ct$ measurement.  To allow for a systematic underestimate of the

17: uncertainties, the mass and the $ct$ uncertainties are multiplied by

18: scale factors determined in the fit.  The $\vrho$

19: distribution of the signal is parameterized in accordance with the

20: equations above.  The background distributions in $ct$ and $\vrho$ are

21: assumed to be uncorrelated.  The latter is described by a shape similar

22: to that of the signal, but with an independent set of amplitudes.

23: %\begin{equation*}

24: %\begin{split}

25: %\frac{d^3\mathcal{P}(\vrho)}{d\vrho} = \, &

26: %|B_0|^2  \cdot f_1(\vec\rho)  + \,

27: %|B_\prl|^2 \cdot f_2(\vec\rho) \\ +  \, &

28: %|B_\perp|^2 \cdot f_3(\vec\rho) + \,

29: %{\rm Re}(B^*_0B_\prl) \cdot f_5(\vec\rho).

30: %\end{split}

31: %\end{equation*}

32: The relationship of mass, $ct$, and $\vrho$ of the $B^0_d$ candidates

33: with $K\pi$ mis-assignment to those of correctly reconstructed candidates

34: is established via Monte Carlo simulation.

35: 

36: Distributions in $\vrho$ are distorted by the detector acceptance, the

37: trigger efficiency and, most importantly, the kinematic selection

38: criteria.  We use the method developed for the CDF Run~I measurement of

39: transversity amplitudes~\cite{THESIS2, cdf00} to account for this

40: distortion.  With as little as six constants extracted from Monte Carlo

41: decays generated uniformly in $\vrho$, this method allows one to avoid

42: the need for explicit parameterization of the distortion in the

43: likelihood.  All aspects of the fitting are extensively verified using

44: Monte Carlo simulations.

45: 

46: Data and fit projections in mass and $ct$ for the $B^0_s$ and $B^0_d$ are

47: shown in Figures~\ref{fig:massfit} and~\ref{fig:lifefit}.  Fits in the

48: transversity sub-space are illustrated by Figure~\ref{fig:ang-proj}.

49: \begin{figure}[t]

50: \centering

51: %\begin{tabular}{@{}c c@{}}

52: %\epsfig{file=figure_1a.eps, clip=, angle=0, width=0.49\linewidth}&

53: %\epsfig{file=figure_1b.eps, clip=, angle=0, width=0.49\linewidth}

54: \epsfig{file=figure_1a.eps, clip=, angle=0, width=1.0\linewidth}

55: \epsfig{file=figure_1b.eps, clip=, angle=0, width=1.0\linewidth}

56: %\end{tabular}

57: \caption{Mass distribution with the fit projection overlaid:

58:   (a)~\BsDec\ , (b)~\BdDec\ .}

59: \label{fig:massfit}

60: \end{figure}

61: 

62: \begin{figure}[tph]

63: \centering

64: %\begin{tabular}{@{}c c@{}}

65: %\epsfig{file=figure_2a.eps, clip=, angle=0, width=0.49\linewidth}&

66: %\epsfig{file=figure_2b.eps, clip=, angle=0, width=0.49\linewidth}

67: %\end{tabular}

68: \epsfig{file=figure_2a.eps, clip=, angle=0, width=1.0\linewidth}

69: \epsfig{file=figure_2b.eps, clip=, angle=0, width=1.0\linewidth}

70: \caption{$ct$ distribution with the fit projection for the

71: signal and background (bkg.) overlaid:

72:   (a)~\BsDec\ , (b)~\BdDec\ .}

73: \label{fig:lifefit}

74: \end{figure}

75: 

76: %\begin{figure}[pth]

77: \begin{figure*}

78: \epsfig{file=figure_3.eps, clip=, angle=0, width=0.9\linewidth}

79: \caption{Projections of the fit onto transversity variables for the

80: mass-sideband-subtracted acceptance-corrected signal: \BsDec\ (top row),

81: \BdDec\ (bottom row). A $ct > 0$ cut is applied.}

82: \label{fig:ang-proj}

83: \end{figure*}

84: 

85: The single largest source of systematic uncertainty in the measurement

86: of the transversity amplitudes of $B^0_{d}$ and $B^0_{s}$ is the choice

87: of parameterization of the background distribution in $\vrho$.  The

88: $B^0_s$ transversity amplitudes receive a small contribution to their

89: systematic uncertainty from a 3.5\% contamination from $B^0_{d}$.  Two

90: other sources contribute to the uncertainty in $B^0_{d}$ amplitudes: the

91: way candidates with incorrect $K\pi$ assignment are handled and the

92: potential contribution of $\mu^+\mu^-K^+\pi^-$ final states that are not

93: due to $B^0_d\rightarrow\jpsi\kst(892)$ decays, which is estimated to be

94: less than 4\% of the total signal.  

95: The systematic

96: uncertainty in the lifetimes receives contributions from the choice of

97: the $ct$ parameterization and from 

98: the SVX alignment.

99: Slightly larger contributions come from the choice

100: of the background parameterization in $\vrho$ and, in the case of

101: $B^0_s$, from $B^0_{d}$ contamination.  For $\dg_s/\Gamma_s$ the only

102: two sources of systematic uncertainty are from the choice of the

103: background $\vrho$ parameterization and from $B^0_{d}$ contamination.

104: Other potential sources of systematic uncertainty, including those from

105: the method of correcting for distortion of the signal distribution in

106: $\vrho$ and potential contribution of $B^0_s\rightarrow\jpsi f^0(980)$

107: decays, were found to be negligible.  The precision of all of the

108: results of this analysis is statistically limited.

109: