1: \section{Analysis}
2: \label{sec:analysis}
3:
4: The initial selection of semileptonic decay candidates follows the single-tag analysis described
5: in Ref. \cite{Aubert:2004bn}.
6: For each $\Dstarp \to \Dz \pi^+; \Dz \to K^{(*)} e \nu$ candidate
7: (charge conjugation is implied in all signal and tagging modes),
8: we calculate the $\Dstarp - \Dz$ mass difference $\Delta M = m(K e \pi ) - m(Ke)$
9: and the proper lifetime, as well as the output of an event selection neural network (NN).
10: We then require that a high-momentum $D$ decaying hadronically be fully reconstructed in the opposite hemisphere of the event.
11: This ensures that the underlying production mechanism is $e^+ e^- \to c \overline c$ and provides a second production flavor tag.
12: We implement additional candidate selection criteria based on studies of alternate background samples in data and a Monte Carlo (MC) simulated event
13: sample (the ``tuning" sample) to reduce various sources of background.
14: The quark fragmentation in $e^+ e^- \to c \overline c$ MC events
15: is simulated using JETSET \cite{Sjostrand:2000wi},
16: the detector response is simulated via
17: GEANT4 \cite{Agostinelli:2002hh},
18: and the resulting events are reconstructed in the same way as are real data.
19:
20: To minimize bias, we use a MC sample (the ``unbiased'' sample)
21: disjoint from the tuning sample, to obtain all MC based estimates of efficiencies and backgrounds.
22: We study this sample,
23: with effective luminosity roughly equivalent to 603 fb${}^{-1}$
24: ($ \approx 1.75 \times $ data)
25: only after all selection criteria and the
26: full analysis method have been established.
27: After the expected background rates are determined from the unbiased MC,
28: we examine the signal region in the data and determine the net
29: number of observed RS and WS
30: signal events ($ n_{\rm{RS}} $ and $ n_{\rm{WS}} $).
31: The measured mixing rate is then determined as $r_{\rm{mix}}=n_{\rm{WS}}/n_{\rm{RS}}$,
32: corrected for the relative efficiency of the WS and RS signal selection criteria.
33:
34: \input reco
35:
36: \subsection{The Hadronic Tagging Samples}
37: We use the flavor of fully reconstructed charm decays
38: in the hemisphere opposite the semileptonic signal to
39: additionally tag the production flavor of the
40: semileptonic signal, and thus significantly reduce
41: the rate of wrongly tagged candidates.
42: We use five hadronic tagging samples.
43: Three samples explicitly require $\Dstarp$ decays:
44: $\Dstarp\rightarrow \Dz\pi^+$ where
45: $\Dz\rightarrow K^-\pi^+$,
46: $\Dz\rightarrow K^- \pi^+ \pi^0$, and
47: $\Dz\rightarrow K^- \pi^+ \pi^+ \pi^-$;
48: while the other two samples are not related to
49: $\Dstarp$ decays:
50: $\Dz\rightarrow K^-\pi^+$ and
51: $\Dp\rightarrow K^-\pi^+ \pi^+$.
52: Candidates from the $\Dstarp $ sample are explicitly excluded from the more
53: inclusive $\Dz \to K^- \pi^+$ sample to ensure that the tagging samples are disjoint.
54:
55: \begin{figure*}[thb]
56: \begin{center}
57: \begin{minipage}{1.0\textwidth}
58: \epsfig{file=./plots/hadtagplots_final_2.eps,width=0.8\textwidth}
59: \end{minipage}
60: \caption[]{\label{TaggingFigure}
61: Invariant mass distributions of hadronic tagging candidates in data events with RS semileptonic
62: candidates passing the initial semileptonic-side selection.
63: The dark shaded regions denote hadronic candidates used as tags.
64: Sideband ``1" events are used to characterize ``false tag" rates and
65: sideband ``2"
66: events are used for background normalization in optimizing the hadronic mass selection.
67: Top row: (a) $\Dz\rightarrow K^-\pi^+$, (b) $\Dp\rightarrow K^-\pi^+ \pi^+$.
68: Bottom row: $\Dstarp\rightarrow \Dz\pi^+$ events in final states
69: (c) $\Dz\rightarrow K^-\pi^+$, (d) $\Dz\rightarrow K^- \pi^+ \pi^+ \pi^-$,
70: (e) $\Dz\rightarrow K^- \pi^+ \pi^0$.
71: }
72: \end{center}
73: \end{figure*}
74:
75: The selection criteria for the tagging samples, such as the $ \Delta M $ ranges for the
76: $ D^* $ modes or the use of production and
77: decay vertex separation for the $ D^+ $ mode,
78: vary from channel to channel to balance high purity against high statistics.
79: Potential criteria are studied using candidate events from a RS sample chosen with
80: loose requirements on the semileptonic side.
81: To eliminate candidates from $\BB$ events, we require the CM momentum of the tag-side
82: $D$ be at least $2.5 \gevc$.
83: The individual $\Dstarp$, $\Dz$ and $\Dp$ tagging
84: candidate invariant mass distributions for the final RS sample are shown in Figure~\ref{TaggingFigure}.
85: The purities of these tagging samples, defined as the ratio of signal
86: in the selected mass range to the total number of candidates in that range
87: (the darkly shaded entries in each histogram), vary from about $92\%$
88: for $\Dz \to K^- \pi^+$ which do not come from $\Dstarp $ to $97.5\%$ for $\Dz \to K^- \pi^+$ from $\Dstarp$.
89: In the very few events where there are multiple hadronic tag candidates,
90: we use the tag coming
91: from the highest purity tagging sample.
92: We reject events with multiple semileptonic
93: signal candidates, requiring that one and only one candidate, whether RS or WS,
94: be present after all tagging and basic semileptonic side selections are imposed.
95: This requirement rejects approximately $13\%$
96: of signal candidates and a similar fraction of background
97: candidates.
98:
99: \begin{table*}[thb]
100: \caption{\label{ProgressiveCuts}
101: Effects of additional semileptonic side selection criteria. Approximate cumulative acceptance rates in the double-tag unbiased MC for signal and WS
102: background as additional selection criteria are applied, relative to acceptances following the initial semileptonic-side selection. The signal
103: acceptances are the same for RS and WS signal samples except for the decay time cut where the entry is that for the WS sample.}
104: \begin{center}
105: \begin{tabular}{ccc} \hline
106: Criterion & Signal Retained & WS Background Retained \\ \hline
107: $ e^{\pm} $ conversion and Dalitz pair veto & 100\% & 82\% \\
108: $ \pi_s $ $\dedx $ cut & 85\% & 66\% \\
109: $ \pi_s $ $ p_T $ and $ p_L $ selection & 72\% & 36\% \\
110: $ m(Ke) > 0.8 \gevcc$ & 71\% & 30\% \\
111: $ ( M(Ke), \Delta M ) $ kinematic cut & 70\% & 20\% \\
112: 600 $ < t < $ 3900 fs & 55\% & 10\% \\ \hline
113: \end{tabular}
114: \end{center}
115: \end{table*}
116:
117: \subsection{Additional Semileptonic Side Selection Criteria}
118: Double-tagging the production flavor of the neutral $D$ mesons effectively eliminates the
119: WS background due to real semileptonic $D$ decays paired with false slow pions from putative $\Dstarp $ decay.
120: From studies of background events in the tuning MC sample,
121: we find that kaon and electron
122: candidates are almost always real kaons and electrons, respectively, with correctly assigned charges.
123: We also find that many fake slow pion candidates are electrons produced as part of conversion pairs,
124: or Dalitz decays of $\piz$ or, to a much lesser extent, $\eta$ mesons.
125: These processes also contribute background tracks to the pool of electron candidates
126: used to create $Ke$ vertices combinatorically.
127: We consequently implement selection criteria to reject tracks
128: which may have
129: originated in such processes by requiring that neither an electron nor slow pion candidate
130: form a conversion pair when combined with an oppositely charged track treated as an electron
131: (whether identified as such or not). We further require that electron candidates not form
132: a $\pi^0$ candidate when combined with a photon candidate and an oppositely charged track treated as an electron.
133: (After applying all event selections, we find no contribution in the tuning MC sample from $\eta$ Dalitz decays.)
134: Rejecting photon conversions and Dalitz decays reduces the total RS and WS backgrounds by
135: about $20\%$ each, and has a negligible effect on signal efficiency.
136:
137: To reduce backgrounds from kaons misidentified as electrons,
138: we require that the laboratory momentum of electron candidates be greater than $600 \mevc$.
139: This reduces the signal efficiency by about $15\%$ and the background rate by about $35\%$.
140: To further reduce the number of electrons that are considered as slow pions, we veto tracks
141: where $\dedx$ in the SVT is consistent with that of an electron.
142: This reduces the signal efficiency by about $15\%$ and the background rate by about $25\%$.
143:
144: We study kinematic distributions that discriminate between signal and
145: background using data and MC events with two fully reconstructed hadronic
146: decays of charm mesons.
147: As a result, we require that the slow pion CM longitudinal momentum
148: (along the axis defined by the direction opposite the
149: tagging $D$'s CM momentum) lie in the range $150-400 \mevc$ and its
150: transverse momentum
151: be less than $80 \mevc$.
152: This reduces the background by approximately $40\% $ and the signal efficiency by
153: approximately $15\%$.
154:
155: We require that the electron-kaon invariant mass be greater than $800 \mevcc$;
156: this reduces the signal efficiency by a few percent while it reduces the
157: background rate by approximately $20\% $.
158: When we count the final number of signal candidates, we also require that
159: $ (M(Ke), \Delta M )$ lies inside the kinematic boundary
160: expected for
161: $ D^{*+} \to D^0 \pi^+; \ D^0 \to K e \nu $ decays where the neutrino
162: momentum is ignored.
163: This has essentially no effect on signal efficiency, and reduces WS backgrounds by about $35\% $.
164: The cumulative effects of the additional semileptonic side selection criteria are summarized
165: in Table \ref{ProgressiveCuts}.
166: (The selection for electron momentum $>600 \mevc$ is applied
167: prior to calculating the acceptances
168: listed in the table.)
169: The effects of these additional selection criteria on
170: $ \Dz \to K^{*} e \nu $ events
171: are reasonably consistent with those for
172: $ \Dz \to K e \nu $ events.
173: The combination of the slow pion longitudinal and transverse momenta and the $(M(Ke), \Delta M)$
174: selections will hereinafter be referred to as the ``double-tag kinematic selection''.
175: Figure~\ref{fig:dmRSsgnl} shows the $ \Delta M $ distributions of signal events in unbiased MC scaled
176: to the luminosity of the data both before and after imposing the double-tag kinematic selection.
177: The marginal efficiency resulting from applying these last selection criteria to
178: signal events is $84 \pm 1\% $.
179:
180: \begin{figure}[tb]
181: \begin{center}
182: \begin{minipage}{.5\textwidth}
183: \epsfig{file=./plots/rsdmj_mc_scaled_3.eps,width=0.9\textwidth}
184: \end{minipage}
185: \caption[]{\label{fig:dmRSsgnl}
186: RS signal $\Delta M$ distribution in unbiased MC scaled to the luminosity of the data before (line) and after (solid) applying the double-tag kinematic
187: selection.}
188: \end{center}
189: \end{figure}
190:
191: The decay time distributions of the RS and WS signals should differ,
192: as shown in Equation \ref{eqn:semileptime}.
193: The RS sample is produced with an exponential decay rate, while the WS sample should
194: be produced with the same exponential rate
195: modulated by $ t^2 $.
196: Figure~\ref{fig:taursws} shows the normalized lifetime distributions
197: for reconstructed simulated RS and WS signal events passing the final tag and signal-side selection.
198: To improve sensitivity, we select only WS candidates with measured lifetimes between 600 fs $ ( \approx 1.5 \tau_{\Dz} ) $ and 3900 fs
199: $ ( \approx 9.5 \tau_{\Dz} ) $, which accepts approximately $80\%$
200: of signal and less than 30\% of background.
201: Because the RS signal-to-background ratio is comparatively very large,
202: we accept RS candidates
203: across the full range shown in Figure~\ref{fig:taursws}.
204: This WS/RS relative efficiency
205: has a 2\% systematic uncertainty due to imperfect knowledge of the decay
206: time resolution function.
207: This is determined from changes in
208: the WS/RS efficiency observed when varying the signal resolution function
209: according to the difference between resolution functions observed in RS data and MC samples.
210: \begin{figure}[tb]
211: \begin{center}
212: \begin{minipage}{.5\textwidth}
213: \epsfig{file=./plots/lifetime_final_4.eps,width=0.9\textwidth}
214: \end{minipage}
215: \caption[]{\label{fig:taursws}
216: Normalized RS (dashed) and WS (solid) reconstructed simulated signal
217: lifetime distributions.
218: The solid vertical lines mark the range for the selection of the WS events.}
219: \end{center}
220: \end{figure}
221:
222: Figure~\ref{fig:nn}(b) shows the NN event selector output for RS signal,
223: RS backgrounds and WS backgrounds in the unbiased MC sample
224: passing the additional semileptonic side selection criteria (scaled to the
225: luminosity of the data).
226: The effectiveness of the additional semileptonic-side criteria in suppressing WS
227: backgrounds while simultaneously retaining good signal efficiency can be seen by
228: comparing Figures~\ref{fig:nn}(a) and (b).
229: Figure~\ref{fig:dmWSbkgd} shows the $\Delta M$ distribution of WS backgrounds passing
230: the decay time selection in unbiased MC scaled to the luminosity of the data both
231: before and after the double-tag kinematic selection.
232: A total of 2.85 background candidates, the sum of the luminosity-scaled events in the
233: solid histogram shown in the figure, is expected after all event selection criteria are applied.
234:
235: \begin{figure}[tb]
236: \begin{center}
237: \begin{minipage}{.5\textwidth}
238: \epsfig{file=./plots/wsdmj1_mc_scaled_5.eps,width=0.9\textwidth}
239: \end{minipage}
240: \caption[]{\label{fig:dmWSbkgd}
241: WS $\Delta M$ distribution for background events passing the WS decay time selection in unbiased MC scaled to the luminosity of the data before (line)
242: and after (solid) applying the double-tag kinematic selection.}
243: \end{center}
244: \end{figure}
245:
246: \subsection{Measuring Signal Yields}
247:
248: To determine the mixing rate,
249: we first determine
250: the number of RS signal candidates by fitting the RS
251: $\Delta M$ distribution,
252: as described in detail below.
253: We then estimate the expected rate of WS background events
254: in the signal region of the
255: data from the unbiased MC sample.
256: Using several background control samples drawn from both data and MC,
257: we estimate how well MC events describe real data events.
258: Using a statistical procedure with good frequentist coverage,
259: we combine
260: the number of candidates observed in the WS sample,
261: the expected background rate,
262: and the estimated systematic uncertainty in the expected background rate
263: to obtain a central value for the mixing rate and
264: 68\% and 90\% confidence intervals.
265: This procedure is described in detail in the appendix.
266:
267: We extract the number of RS signal events from the $\Delta M$ distribution of the
268: RS sample selected without the double-tag kinematic selection using an
269: extended maximum likelihood fit.
270: The likelihood function includes probability density functions
271: (PDF's) for the signal,
272: the background events which peak in the signal region,
273: and the combinatorial background.
274: The PDF for each event class is assigned using the functional
275: forms described in Ref. \cite{Aubert:2004bn}.
276: The shape parameters for the combinatoric background are determined using the following
277: technique: $\Dz$ signal candidates in the data from one event are combined with $\pi_s$
278: candidates from another event to model the shape of this PDF. Based on MC studies,
279: the shape of the peaking $\Delta M$ background is assumed to be the same shape as the signal.
280: Its relative level is also determined from MC studies.
281: The shape parameters of the signal PDF, as well as the number of RS signal events and
282: the number of combinatorial background events, are then obtained from the likelihood fit of the data.
283:
284: The main plot in Figure~\ref{FinalRSFig} shows the $\Delta M$ fit of the RS data
285: before applying the double-tag kinematic selection, with the signal and background
286: contributions overlaid.
287: The fitted RS signal yield in this sample is $5748 \pm 90$ events, with $\chi^2=77$ for 60 bins,
288: where six parameters are determined from the fit.
289: The inset plot of Figure~\ref{FinalRSFig} shows the RS data
290: $\Delta M$ distribution
291: after the double-tag kinematic selection is imposed. As noted above, the efficiency
292: of this selection is $0.84 \pm 0.01$, giving a final RS signal yield of $4780 \pm 94$,
293: which is used as the normalization in calculating the mixing rate.
294:
295:
296: \begin{figure}[tb]
297: \begin{center}
298: \begin{minipage}{.5\textwidth}
299: \epsfig{file=./plots/sldmfit_rs_final_6.eps,width=0.9\textwidth}
300: \end{minipage}
301: \caption[]{\label{FinalRSFig}
302: RS data $\Delta M$ distribution.
303: The main plot shows the RS data (points) before imposing the double-tag kinematic selection,
304: and the projections of the total fit PDF (solid line) and the background PDF (dashed line).
305: The inset plot shows the RS $\Delta M$ distribution after the double-tag kinematic selection
306: criteria are applied.}
307: \end{center}
308: \end{figure}
309:
310: To determine the number of WS mixed events, we consider three regions of $\Delta M$:
311: the signal region, $\Delta M \le 0.20 \gevcc$;
312: the near background region, $0.20 < \Delta M \le 0.25 \gevcc$;
313: and the far background region, $0.25 < \Delta M \le 0.35\gevcc $.
314: These $ \Delta M $ ranges are shown in Figure ~\ref{FinalWSFig},
315: and are respectively labeled ``1", ``2" and ``3" in the plot.
316: To avoid potential bias,
317: we examine neither the signal region nor the near background region
318: in the WS data sample until all of the selection criteria and the procedure for calculating
319: confidence intervals are determined.
320: The WS signal region may contain both signal and background events after applying the final
321: event selection critera.
322: As discussed above, we determine the expected number of background events from the unbiased
323: MC sample: we observe 5 events,
324: which scales to 2.85 for the luminosity of the data.
325: To estimate the possible non-$c \overline c$
326: background rate, we also examine events which satisfy the semileptonic-side selection criteria
327: but fail the tagging-side criteria because the mass of the hadronic $D$ candidate falls outside
328: the accepted window.
329: Since we had examined the data events in the ``far" sidebands
330: (sidebands ``2") of Figure~\ref{TaggingFigure} while optimizing hadronic
331: side selection criteria,
332: we also examine those in the ``near sidebands (sidebands ``1")
333: to estimate the number of these ``false tag" events:
334: we find no WS candidates in the near or far $\Delta M$ sideband regions in either the data or
335: unbiased MC sample.
336: Given the agreement between data and the unbiased MC sample, we determine the central value of
337: the number of WS signal events by subtracting the luminosity-scaled number of unbiased MC WS
338: background events in the signal region from the number of candidates observed in the data there.
339:
340: The dark shaded entries in Figure~\ref{FinalWSFig} denote the $\Delta M$ distribution of WS
341: candidates in the data after all event selection, where we observe $3$ WS candidates
342: in the signal region and none in the sideband regions.
343: Given the expected WS background of 2.85~events shown in the solid histogram of
344: Figure~\ref{fig:dmWSbkgd}, we calculate a net WS signal yield of 0.15 events.
345: We discuss below the total error associated with the estimated number
346: of WS background events.
347:
348: \begin{figure}[tb]
349: \begin{center}
350: \begin{minipage}{.5\textwidth}
351: \epsfig{file=./plots/wsdm_final_7.eps,width=0.9\textwidth}
352: \end{minipage}
353: \caption[]{\label{FinalWSFig}
354: WS data $\Delta M$ distribution.
355: The dark histogram shows WS events in the data passing all event selection.
356: The light histogram shows WS events passing all selections except the double-tag
357: kinematic selection.
358: Region ``1'' is the signal region, ``2'' is the near sideband, and ``3'' is the far sideband.
359: }
360: \end{center}
361: \end{figure}
362:
363: \subsection{Systematics and Confidence Intervals}
364:
365: To calculate confidence intervals for the number of mixed events observed,
366: we first determine a systematic uncertainty associated with the
367: WS background estimate.
368: To do this,
369: we compare 10 background control samples in data with the
370: corresponding MC samples.
371: The results of this comparison are shown in
372: Table \ref{ProxyBackgroundTable}.
373: The first line compares the number of WS events observed in the
374: far background region of the data and the tuning MC sample.
375: The second line compares the same numbers for the data and for
376: the unbiased MC sample.
377: The remaining table entries compare the number of events observed
378: in two types of doubly-charged (DC) background samples obtained
379: from data with those observed from the same sources in
380: unbiased MC events.
381: In both of the DC background samples, the kaon and the electron have
382: the same charge sign, and are reconstructed exactly as neutral
383: $ K e $ vertex candidates are, except for the differing charge correlation.
384: In those additionally labeled WS, the slow pion has the same charge
385: as the kaon, while in those additionally labeled RS,
386: the slow pion has the opposite charge.
387:
388: \begin{table*}[tb]
389: \caption{\label{ProxyBackgroundTable}
390: Comparison of MC and data background yields. The doubly-charged (DC) MC entries refer to MC event samples disjoint from those used to optimize event
391: selection. The ``kinematic selection" refers to the double-tag kinematic selection.
392: }
393: \begin{center}
394: \begin{tabular}{cccccc} \hline
395: Entry & Data Sample & $\Delta M$ Range (\gevcc) &
396: kinematic selection & Data & \phantom{mm} MC \phantom{mm} \\ \hline
397: 1 & WS, tuning MC & $0.25 \le \Delta M \le 0.35$ & no & $2\pm 1.4$ & $2.1 \pm1.5$ \\
398: 2 & WS, unbiased MC & $0.25 \le \Delta M \le 0.35$ & no & $2\pm1.4 $ & $3.4 \pm1.4$ \\
399: 3 & DC, RS & $\Delta M \le 0.20$ & yes & $37 \pm 6$ & $40 \pm 5.1$ \\
400: 4 & DC, WS & $\Delta M \le 0.20$ & yes & $36 \pm 6$ & $51 \pm 5.8$ \\
401: 5 & DC, RS & $\Delta M \le 0.20$ & no & $42 \pm 7$ & $47 \pm 5.5 $\\
402: 6 & DC, WS & $\Delta M \le 0.20$ & no & $ 55 \pm 8$ & $64 \pm 6.5$ \\
403: 7 & DC, RS & $0.20 < \Delta M \le 0.25$ & no &
404: $20 \pm 5$ & $ 24 \pm 3.9$ \\
405: 8 & DC, WS & $0.20 < \Delta M \le 0.25$ & no &
406: $13 \pm 4$ & $ 19 \pm $ 3.5\\
407: 9 & DC, RS & $0.25 \le \Delta M \le 0.35$ & no & $20 \pm 5$ & $31 \pm 4.5$\\
408: 10 & DC, WS & $0.25 \le \Delta M \le 0.35$ & no & $23 \pm 5$ & $18 \pm 3.4$ \\ \hline
409: \end{tabular}
410: \end{center}
411: \end{table*}
412:
413: Ignoring the correlations between entries 3,5 and 4,6 in
414: Table \ref{ProxyBackgroundTable}, we estimate the consistency
415: between the data and MC samples by calculating a
416: summed $\chi^2 $ for all the entries:
417:
418: \begin{eqnarray}
419: \chi^2 (\mbox{data, MC}) = \sum\limits_{i=1}^{10} \biggl [\frac{(x_i^{data}-x_i^{MC})^2} {(\sigma_i^{data})^2 + (\sigma_i^{MC})^2} \biggr] = 11.4
420: \end{eqnarray}
421:
422: The value $\chi^{2}=11.4$ is consistent with 1 per degree of freedom.
423: Taken together, these observations indicate that the MC estimate for
424: the background rate in the signal region of the WS sample is reasonably accurate.
425: We conservatively assign the largest discrepancy between the data and MC rates,
426: 50\%, as the systematic uncertainty associated with the ratio between the MC
427: estimate of the background rate and its true value.
428:
429: To determine confidence intervals for the number of WS mixed events,
430: we adapt a suggestion made in Ref. \cite{Porter:2003ui}.
431: The complete statistical procedure is described in detail in the appendix;
432: it is summarized here.
433: We start with a likelihood function, $ {\cal L} (n, n_b; s, b) $, for the number of events
434: observed in the signal region of the WS data sample, $ n $, and the corresponding number
435: observed in the MC sample, $ n_b $. $ {\cal L} (n, n_b; s, b) $ depends upon the true signal
436: rate $ s $ and the true background rate $ b $ in the signal region, and also accounts for the
437: systematic uncertainty in the ratio of the true background rate in data to that estimated
438: from MC.
439: The value of $ (s, b ) $ which maximizes the likelihood function,
440: $ {\calL}_{\rm max} $, is denoted by $ (\hat s , \hat b ) $.
441: As one expects naively, $\hat b $ is equal to $ n_b $ times
442: the ratio of data and MC
443: luminosities while $\hat s = n - \hat b $. We then search for the values of $ s $
444: where $ - {\rm ln} {\cal L} (s) $ changes by 0.50 [1.35]; here $ {\cal L} (s) $
445: denotes the likelihood at $ s $ maximized with respect to $ b $.
446: The lower and upper values of $ s $ which satisfy this condition define the
447: nominal 68\% [90\%] confidence interval for $ s $.
448: As discussed in the appendix, for the range of parameters relevant for this analysis,
449: the confidence intervals produced by this procedure provide frequentist coverage which
450: is accurate within a few percent.
451:
452: \subsection{Final Results and Conclusion}
453:
454: We observe 3 candidates for $\Dz-\Dzb$ mixing,
455: compared to 2.85 expected background events, where
456: we ascribe a 50\% systematic uncertainty to this
457: expected background rate.
458: We find the central value for the number of
459: WS signal events to be 0.15, with 68\% and 90\% confidence intervals
460: $ (-2.2, 2.8) $
461: and $ (-5.2, 4.7) $, respectively.
462: Accounting for the ratio of WS and RS signal efficiencies due to the cut on the measured
463: WS decay time ($0.80 \pm 0.02$), we find the central value of
464: $ r_{\rm mix} $ to be $ 0.4 \times 10^{-4} $,
465: with 68\% and 90\% confidence intervals
466: $ (-5.6, 7.4) \times 10^{-4}$ and $ (-13, 12) \times 10^{-4} $, respectively.
467: We ignore variations in the RS yield due to statistical error and systematic effects in the RS
468: fit as they are negligible relative to the statistical errors associated with the WS data and MC rates as well as the 50\% systematic error assigned to
469: the ratio of MC and data WS background rates.
470:
471: The sensitivity of this double-tag analysis is comparable to that
472: expected for a single-tag analysis, see Table \ref{PreviousLimitsTable}.
473: Future analyses should be able to combine these two approaches to significantly improve overall sensitivity to charm mixing using semileptonic final
474: states. Improved methods for reconstructing and selecting semileptonic signal candidates, the use of more hadronic tagging modes, and the additional
475: use of semi-muonic decay modes
476: may allow semileptonic charm mixing analyses to approach the $ r_{\rm mix} $ sensitivity of analyses using hadronic final states.
477:
478:
479: