1: \section{\label{sec:CPT}Tests of CPT Conservation}
2:
3: To probe CPT conservation with MINOS data, we use two tests. After discussing the systematic uncertainties, we first compare the strength of the oscillation signals for the neutrino-induced $\mu^-$ and $\mu^+$ samples individually. Second, we divide the charge-separated muons into low and high momentum samples and further test for differing rates of neutrino disappearance. We show that this second test is the one that minimizes systematic uncertainties.
4:
5: \subsection{\label{sec:systematicsCPT}Systematic Uncertainties}
6:
7: Here we discuss the additional sources of systematic uncertainties associated with the CPT analysis. One additional source of systematic uncertainty is due to misidentified charge sign or the charge purity of the data sample. We determined the charge purity systematic by comparing the underground cosmic ray muon charge ratio for events passing the $\chiLine$ cut with the charge ratio found by MINOS, $N_{\mu^{+}}/N_{\mu^{-}} = 1.371$~\cite{m&r}. For this comparison we divided the cosmic ray muons into two data samples, one with muons having momenta $< 30$~GeV/$c$ and the other with momenta in the range 30~--~100~GeV/$c$. For cosmic ray muons in the lower momentum sample, we calculated the difference in the charge ratio to be 0.047, a difference that can be attributed to an impurity of 5\%. For cosmic ray muons in the high momentum sample, we calculated the difference in the charge ratio to be 0.12, which can be attributed to an impurity of 14\%. By taking a weighted average of these impurities, where the weights are given by the number of neutrino-induced muons in each momentum range, we calculated the impurity in the neutrino-induced muon data to be 6\%. Using similar methods we found the weighted average impurity for the neutino-induced Monte Carlo events to be 1\%. These weighted impurities are given in Table~\ref{table:cptsystematics3} (4).
8:
9: There are also additional sources of systematic uncertainty due to the physics model. These arise from uncertainties in the ratios of the $\numbar$ cross section, $\sigma_{\numbar}$, to the $\num$ cross section, $\sigma_{\num}$, and we estimate the uncertainties to be 9\%~\cite{Sanjib2} for energies $<30$ GeV~(9) and 2\%~\cite{PDBook2} for energies $>30$ GeV~(10). The uncertainty in the ratio of the number of $\num$ to $\numbar$~(11) is 1\%~\cite{Barr:2006it}.
10:
11: \subsection{\label{sec:CPTRatio}Charge-Separated Event Ratio}
12:
13: Our first test of CPT conservation comes from the comparison of the ratio of the total number of $\mu^-$ events to $\mu^+$ events in the data and the Monte Carlo simulation including backgrounds. In the data, this ratio of charge separated muons is given by
14: \begin{equation}
15: R^{data}_{\mu^-/\mu^+} = \sum (N_{\mu^-}){\Large /} \sum (N_{\mu^+}).
16: \label{eq:rdata1}
17: \end{equation}
18: The Monte Carlo ratio is defined in a similar way. If the $\num$ oscillate with the same parameters as the $\numbar$ then the ratio of the data and Monte Carlo ratios, $\hat{\mathcal{R}}_{CPT}$, will be consistent with unity. The systematic uncertainties in $\hat{\mathcal{R}}_{CPT}$ are shown in Table~\ref{table:cptsystematics3}. This table shows how changes in the default parameters corresponding to the sources of systematic uncertainty changes the value of $\hat{\mathcal{R}}_{CPT}$. The main contributions to the uncertainty in $\hat{\mathcal{R}}_{CPT}$ are the purity, cross section ratio for neutrinos with energies $<30$ GeV, and the ratio of $\num/\numbar$. Since the ratio in Eq.~(\ref{eq:rdata1}) is greater than one, more negative than positive muons will be misidentified and the measured charge ratio decreases towards unity. Consequently, charge misidentification leads to a one-sided (positive) error on the measured charge ratio. The uncertainty due to the impurity is $(\Delta \hat{\mathcal{R}}_{CPT})_{4} = 0.07$. The cross section ratio for neutrinos with energies $<30$ GeV changes the ratio by $(\Delta \hat{\mathcal{R}}_{CPT})_{9} = \pm 0.04$. The uncertainty in the ratio of the number of $\num$ to $\numbar$ corresponds to $(\Delta \hat{\mathcal{R}}_{CPT})_{11} = \pm 0.01$. We find from Table~\ref{table:events} and Table~\ref{table:cptsystematics3}
19: \begin{equation}
20: \hat{\mathcal{R}}_{CPT} = \frac{R^{data}_{{\mu^-}/{\mu^+}}}{R^{MC}_{{\mu^-}/{\mu^+}}} = \mtop^{+\mtopUp}_{-\mtopDn}(\text{stat})\mtopSys(\text{syst}),
21: \label{eq:totalRatio}
22: \end{equation}
23: where the statistical uncertainties were calculated using the method for Eq.~(\ref{eq:lowhiresult}). This value of $\hat{\mathcal{R}}_{CPT}$ is consistent with CPT conservation.
24:
25: \begin{table}
26: \caption{\label{table:cptsystematics3}Sources of systematic uncertainty considered and their effects on the ratio $\hat{\mathcal{R}}_{CPT}$.}
27: \begin{tabular}{|l|c|c|}
28: \hline\hline
29: & & \\
30: \multicolumn{1}{|c|}{Source}& ~~$\sigma$~~ & $(\Delta\mathcal{\hat{R}}_{CPT})_k$\\
31: \hline\hline {\bf Reconstruction systematics}: & \multicolumn{2}{c|}{ } \\
32: \hline ~~~(1)~$\chi^{2}_{fit}/ndf<1.5$ & 0.01 & $<5\times10^{-4}$\\
33: \hline ~~~(2)~$\chi^{2}_{1/\beta}/ndf<3.0$ & 0.01 & $<1\times10^{-4}$\\
34: \hline ~~~(3)~$\chiLine<10$ ~ & 0.27 & $<5\times10^{-3}$ \\
35: \hline ~~~(4)~Charge Purity (Data,MC)~ & (0.06,0.01) & +0.07 \\ %changes by -0.15
36: \hline\hline {\bf Model systematics}: & \multicolumn{2}{c|}{ } \\
37: \hline ~~~(5)~Normalization & 0.15 & $<1\times10^{-4}$\\
38: \hline ~~~(6)~Spectral Index & 0.03 & $<1\times10^{-4}$ \\ %changes by 0.02
39: \hline ~~~(7)~Cross Section ($E<30$ GeV) & 0.07 & $<5\times10^{-3}$\\ %changes by 0.01
40: \hline ~~~(8)~Cross Section ($E>30$ GeV) & 0.02 & $<5\times10^{-3}$\\
41: \hline ~~~(9)~$\sigma_{\numbar} / \sigma_{\num}$ ($E<30$ GeV) & 0.09 & 0.04\\ %changes by 0.11
42: \hline ~~(10)~$\sigma_{\numbar} / \sigma_{\num}$ ($E>30$ GeV) & 0.02 & $<5\times10^{-3}$\\ %changes by 0.01
43: \hline ~~(11)~$\num/\numbar$ & 0.01 & 0.01\\
44: \hline\hline$ \Delta \mathcal{\hat{R}}_{CPT} = \sqrt{ \sum (\Delta\mathcal{\hat{R}}_{CPT})_k^2}$ & & -0.04, +0.08 \\
45: \hline\hline
46: \end{tabular}
47: \end{table}
48:
49: \subsection{\label{sec:CPTeventRatio}Charge-Separated, Low-to-High Momentun Event Ratio}
50:
51: Our second test of CPT conservation is based on the charge separated ratios of $\lowMom$ to $\hiMom$ events compared to the Monte Carlo expectation including backgrounds. First define ratios of data to Monte Carlo expectation for the $\num$ and $\numbar$ individually,
52: \begin{eqnarray}
53: {\cal R}_{-} & = & \biggl(\frac{R^{data}_{{L}/{H}}}{R^{MC}_{{L}/{H}}}\biggr)_{\mu^-}
54: \nonumber \\
55: & = & \lowHiM^{+\lowHiMUp}_{-\lowHiMDn}(\text{stat})\pm0.07(\text{syst}),
56: \label{eq:cptresult1}
57: \end{eqnarray}
58: \noindent and
59: \begin{eqnarray}
60: {\cal R}_{+}& = & \biggl(\frac{R^{data}_{{L}/{H}}}{R^{MC}_{{L}/{H}}}\biggr)_{\mu^+}
61: \nonumber \\
62: &=&\lowHiP^{+\lowHiPUp}_{-\lowHiPDn}(\text{stat})\pm0.07(\text{syst}).
63: \label{eq:cptresult2}
64: \end{eqnarray}
65: The systematic uncertainties in these ratios are listed in Table~\ref{table:cptsystematics2}, and the statistical uncertainties were again calculated using the same method as for Eq.~(\ref{eq:lowhiresult}). We define the ratio of ratios $\tilde{\mathcal{R}}_{CPT}$,
66: \begin{equation}
67: \tilde{\mathcal{R}}_{CPT}= {\cal R}_{-}/ {\cal R}_{+} =\cpt^{+\cptUp}_{-\cptDn}(\text{stat})\pm\cptSys(\text{syst}).
68: \label{eq:cptresult}
69: \end{equation}
70: We found the upper and lower limits for the statistical uncertainty in Eq.~(\ref{eq:cptresult}) by fitting the observed L and H event rates for $\mum$ and $\mup$ to a model which used $\tilde{\mathcal{R}}_{CPT}$ and three of the four event rates as inputs. The data and model were compared using a $\chi^{2}$ function appropriate for Poisson distributed data~\cite{PDBook1}. The 68\% C.~L.~interval quoted includes all values of $\tilde{\mathcal{R}}_{CPT}$ which produce a $\chi^{2}$ within 1 of the best fit value. Again within the statistical and systematic uncertainties, the value of $\tilde{\mathcal{R}}_{CPT}$ is consistent with unity and CPT conservation.
71:
72: The value of this test is clearly demonstrated in Table~\ref{table:cptsystematics2} where it can be seen that several sources of systematic uncertainty cancel in the calculation of $\tilde{\mathcal{R}}_{CPT}$. These cancellations occur because the $\mum$ and $\mup$ ratios are affected similarly for these systematic uncertainties and so divide out in the ratio. The uncertainty in the ratio of the number of $\num$ to the number of $\numbar$ does not contribute to the uncertainty in $\tilde{\mathcal{R}}_{CPT}$. As seen in Table~\ref{table:cptsystematics2}, the only sources of uncertainty which do not cancel are those in the relative magnitude of the $\num$ and $\numbar$ cross sections at low (9) and high (10) energies.
73:
74: This approach should prove valuable in high statistics accelerator-based tests of CPT conservation because it minimizes systematic uncertainties so effectively.
75:
76: \begin{table*}
77: \caption{\label{table:cptsystematics2}Sources of systematic uncertainty considered and their effects on the ratios $\mathcal{R_{-}}$, $\mathcal{R_{+}}$, and $\mathcal{\tilde{R}_{CPT}}$.}
78: \begin{tabular}{|l|c|c|c|c|c|c|}
79: \hline\hline
80: & & \multicolumn{2}{c|}{ } & \multicolumn{2}{c|}{ }&\\
81: \multicolumn{1}{|c|}{Source}& ~~$\sigma$~~ & \multicolumn{2}{c|}{$\Delta\mathcal{R}_{-}$} & \multicolumn{2}{c|}{$\Delta\mathcal{R}_{+}$} &
82: $(\Delta {\mathcal {\tilde{R}}_{CPT}})_k$\\
83: \hline & & $-1\sigma$ & $+1\sigma$& $-1\sigma$&$+1\sigma$ &\\
84: \hline\hline {\bf Reconstruction systematics}: & \multicolumn{6}{c|}{ } \\
85: \hline ~~~(1)~$\chi^{2}_{fit}/ndf<1.5$ & 0.01 & $<5\times10^{-4}$ & $<5\times10^{-4}$ & $<5\times10^{-4}$ & $<5\times10^{-4}$ & --\\
86: \hline ~~~(2)~$\chi^{2}_{1/\beta}/ndf<3.0$ & 0.01 & $<1\times10^{-4}$ & $<1\times10^{-4}$ & $<1\times10^{-4}$ & $<1\times10^{-4}$ & --\\
87: \hline ~~~(3)~$\chiLine<10$ ~ & 0.27 & 0.05& 0.05 & 0.05 & 0.05 & -- \\
88: \hline ~~~(4)~Charge Purity (Data,MC) & (0.06,0.01) & $<1\times10^{-4}$ & $<1\times10^{-4}$ & $<1\times10^{-4}$ & $<1\times10^{-4}$ & -- \\
89: \hline\hline {\bf Model systematics}: & \multicolumn{6}{c|}{ } \\
90: \hline ~~~(5)~Normalization & 0.15 & $<1\times10^{-4}$ & $<1\times10^{-4}$ & $<1\times10^{-4}$ & $<1\times10^{-4}$ & --\\
91: \hline ~~~(6)~Spectral Index & 0.03 & 0.04 & 0.03 & 0.04 & 0.04 & -- \\
92: \hline ~~~(7)~Cross Section ($E<30$ GeV) & 0.07 & 0.02 & 0.02 & 0.02 & 0.02 & --\\
93: \hline ~~~(8)~Cross Section ($E>30$ GeV) & 0.02 & 0.01 & 0.01 & 0.01 & 0.01 & --\\
94: \hline ~~~(9)~$\sigma_{\numbar} / \sigma_{\num}$ ($E<30$ GeV) & 0.09 & -- & -- & 0.02 & 0.02 & 0.03\\
95: \hline ~~(10)~$\sigma_{\numbar} / \sigma_{\num}$ ($E>30$ GeV) & 0.02 & -- & -- & 0.01 & 0.01 & 0.01\\
96: \hline\hline $\Delta {\mathcal {\tilde R}}_{CPT} = \sqrt{ \sum ( \Delta {\mathcal {\tilde R}}_{CPT})_k^2}$ & \multicolumn{5}{c|}{ } & $\cptSys$ \\
97: \hline\hline
98: \end{tabular}
99: \end{table*}
100:
101: