hep-ex0508053/data-analysis-2nd/doc.tex
1:  Even after the fiducial volume cut, some fraction of the 
2: ambient background still
3: remains. It is mainly due to mis-reconstruction of the vertex position.
4: In order to remove the remaining background, several cuts to evaluate the
5: quality of the reconstructed vertex were applied.
6: 
7: \paragraph{Fit stability cut}
8: The goodness value of the vertex was calculated for points in the 
9: region of the reconstructed vertex position and the sharpness of 
10: the goodness distribution as a function of detector coordinates was evaluated. 
11: First of all, the goodness at 
12: $\sim300$ grid points around the original vertex are calculated, and 
13: their difference
14: from the goodness at the reconstructed vertex ($\Delta_g$) is determined.
15: The number of test points which give a $\Delta_g$ more than some threshold
16: was counted; the threshold is as a function of energy and vertex position.
17: The ratio of the number of points over this threshold to the total number of 
18: grid points 
19: is defined as $R_{grid}$. The $R_{grid}$ distribution of the data
20: is shown in Figure~\ref{fig:gringo} along with simulated 
21: $^8$B Monte Carlo events.
22: Events with $R_{grid} > 0.08$ are rejected as background.
23: The systematic error of this reduction was evaluated using LINAC data and MC
24: and the systematic error for flux measurement is estimated to be $\pm1.0\%$.
25: %
26: \begin{figure}[htbp]
27:   \begin{center}
28:   \includegraphics[scale=0.4]{./data-analysis-2nd/gringo.eps}
29:   \caption{$R_{grid}$ distribution for data (solid line) and MC (dashed line).
30:            The normalization is done by scaling to the first bin. 
31:            The dotted line
32:            shows the reduction criteria.}
33:   \label{fig:gringo}
34:   \end{center}
35: \end{figure}
36: %
37: 
38: \paragraph{Hit pattern cut}
39: Mis-reconstructed events often do not have the expected Cherenkov ring pattern 
40: when the hit PMT's are viewed from the reconstructed vertex. Also, some 
41: fraction
42: of spallation products like $^{16}$N emit multiple gammas in addition to 
43: an electron and so do not fit a single Cherenkov ring pattern very well. 
44: In order to remove those kind of events, the following likelihood 
45: function is defined:
46: %
47: \begin{equation}
48:   L(E, \vec x ) \equiv \frac{\sum_{i=1}^{N_{hit}} \log (f{(\cos \theta_{dir},E, \vec x )})_i}
49:                {N_{hit}},
50:   \label{eq:patlik}
51: \end{equation}
52: %
53: here, \(f{(\cos \theta_{dir}E,\vec x)}\) is the same function as
54: that derived from MC in Eq.~(\ref{eq:dirlike}), but here it depends on  
55: energy and vertex position. Figure~\ref{fig:patlik} shows 
56: the likelihood distribution
57: Eq.~(\ref{eq:patlik}) for data and $^8$B MC. The events with a likelihood less
58: than $-1.85$ are rejected. The systematic error of this cut is estimated using
59: solar neutrino MC and a sample of very short lived spallation products; 
60: it is $+1.0\%$ and $-0.5\%$ for the flux measurement.
61: %
62: \begin{figure}[htbp]
63:   \begin{center}
64:   \includegraphics[scale=0.4]{./data-analysis-2nd/patlik.eps}
65:   \caption{Likelihood distribution for data (solid line) and MC (dashed line).
66:            The MC is normalized to the peak bin. The dotted line
67:            shows the reduction criteria.}
68:   \label{fig:patlik}
69:   \end{center}
70: \end{figure}
71: %
72: 
73: \paragraph{Fiducial volume cut using the second vertex fitter}
74: The 2~m fiducial volume cut using the second vertex reconstruction described
75: in Section~\ref{sec:vertex} is applied. Figure~\ref{fig:cluscut} shows the
76: distance from the wall distribution of data and MC using the second vertex
77: reconstruction. Table~\ref{tab:reduc} also shows the reduction step summary
78: described in this section.
79: %
80: \begin{figure}[htbp]
81:   \begin{center}
82:   \includegraphics[scale=0.4]{./data-analysis-2nd/cluscut.eps}
83:   \caption{The distance from the wall distribution using the second vertex
84:            reconstruction for data (solid) and MC (dashed).
85:            The normalization of MC is done by the total number of events.
86:            The dotted line shows the reduction criteria.}
87:   \label{fig:cluscut}
88:   \end{center}
89: \end{figure}
90: %
91: 
92: 
93: 
94: 
95: 
96: 
97: 
98: 
99: 
100: 
101: 
102: 
103: