hep-ex0601009/chAnalysis.tex
1: \chapter{Data and analysis}
2: \section{Overview}
3: 
4: The left-right transverse single-spin asymmetry, $A_N$, of
5: mid-rapidity neutral pions was measured from the data taken during
6: the first polarized proton run at RHIC, in late 2001 and early 2002.
7: To make a single-spin measurement with two polarized beams, the spin
8: states of only one beam at a time were taken into account, averaging
9: over the spin states of the other. Neutral pions were reconstructed
10: via their decay to two photons. The general procedure was to obtain
11: the spin-dependent $\pi^0$ yields for each machine fill, calculate
12: the raw (uncorrected) asymmetries for each fill as a function of
13: $p_T$, make fill-by-fill polarization corrections, and then average
14: the asymmetries over all fills. The results were corrected for
15: estimated contributions to the asymmetry from the background under
16: the \piz\ invariant-mass peak by measuring and subtracting the
17: asymmetry of the background immediately around the peak in mass.
18: Various studies were subsequently performed as checks. The final
19: results of this analysis have been published in \textit{Physical
20: Review Letters} \cite{Adler:2005in}.
21: 
22: \section{Data selection and quality}
23: 
24: Prior to this spin-asymmetry analysis, a polarization-averaged cross
25: section measurement was made for neutral pion production at PHENIX
26: and published in \cite{Adler:2003pb}.  The analysis completed for
27: this thesis started with the data sample utilized in the cross
28: section analysis, and then additional quality cuts relevant to a
29: spin-dependent analysis were made.
30: 
31: The data used were from ERT \trig-triggered events. The \trig\
32: energy threshold was $\sim0.8$ GeV in the 2001-02 run. The triggered
33: sample had much better statistics at higher transverse momenta than
34: the MB sample, but the MB sample was also analyzed for comparison
35: purposes. 18.7 million triggered events were analyzed, corresponding
36: to approximately 880 million sampled MB events. The detector
37: subsystems involved were the BBC and the EMCal, including both the
38: lead scintillator (PbSc) and lead glass (PbGl). Additionally, the
39: drift chamber and pad chambers were used to veto clusters in the
40: EMCal produced by charged particles. Basic checks on the quality of
41: output from all involved subsystems were performed.
42: 
43: \subsection{Fill and run selection}
44: 
45: In making $\pi^0 \rightarrow \gamma + \gamma$ spin-asymmetry or
46: cross section measurements, it is essential to ascertain that the
47: EMCal was working properly when the analyzed runs were taken. The
48: uppermost sector of PbSc in the West arm had been omitted from the
49: cross section analysis due to problems with electronics noise during
50: data taking. A number of runs had been eliminated from that analysis
51: because of EMCal towers with unusually high or low numbers of hits
52: that were not among the known and understood hot or dead EMCal
53: towers. In addition, a sequence of runs had been omitted because of
54: the ERT.  The energy threshold for the ERT \trig\ trigger had been
55: adjusted in the early part of the data-taking period, and only runs
56: taken after the threshold was stable at $\sim 0.8$~GeV were
57: included.  Runs from a total of twenty machine fills were included
58: in the final cross section analysis.
59: 
60: For the present analysis, six of the twenty fills used in the cross
61: section analysis were removed. Two were removed because no
62: polarization measurement was available. Three additional fills were
63: removed because a number of bad runs from these fills were found in
64: a study investigating the stability of the MB trigger. One fill was
65: removed because it had an unusual spin pattern.
66: 
67: \subsection{Event and crossing selection}
68: 
69: An offline BBC event vertex within $\pm $30~cm of the nominal
70: interaction point was required for all events.  The acceptance of
71: the central arms is approximately constant for collisions taking
72: place in this region. The BBC vertex resolution was $\sim 2$~cm. The
73: online vertex cut for the MB trigger in the 2001-02 run was $\pm
74: $75~cm.
75: 
76: Spin-sorting the events requires keeping track of the direction of
77: the spin vector for the polarized bunch in the bunch crossing that
78: produced the event (simply the term \emph{crossing} will be used
79: henceforth). Note that the same pairs of bunches in the two rings
80: collide at the same interaction points each time, but the same pairs
81: of bunches do not necessarily collide at the different interaction
82: points, i.e.~at different experiments. Data quality was checked on a
83: crossing-by-crossing basis.  For the entire 2001-02 $p+p$ data set,
84: four crossings out of a nominal 60 were consistently removed from
85: this analysis.  These four crossings had unusually low luminosities
86: because they were regularly affected by either injection or steering
87: activities in the ring. In addition, the ten crossings in which only
88: one beam had filled bunches while the other had a five-crossing
89: beam-abort gap were removed from all fills. Any events occurring in
90: these crossings were beam-background collisions rather than
91: beam-beam collisions.
92: 
93: There were two bunch-by-bunch spin patterns utilized in the 2001-02
94: run, one in the initial period of data taking and the other in the
95: latter period. The second pattern included bunches with zero
96: polarization so that they would be available for systematic checks,
97: in particular for the $p$C polarimeter.  For fills with the second
98: polarization pattern, one additional crossing was eliminated from
99: this analysis because it had zero polarization.  The other
100: zero-polarization crossings corresponded to two of the
101: low-luminosity bunches that had already been eliminated.  A
102: schematic illustration of the first spin pattern is given in
103: Figure~\ref{figure:spinPattern}.  Note that the spin patterns are
104: selected to provide approximately equal numbers of same-spin and
105: opposite-spin crossings for each experiment.  They are also chosen
106: so that the spin combinations are different every crossing.  This
107: rapid change in spin combinations, occurring on the order of every
108: hundred nanoseconds, greatly reduces potential time-dependent
109: systematic uncertainties.
110: 
111: \begin{figure}
112: \centering
113: \includegraphics[height=0.25\textheight]{%
114: bunchpattern.eps} \caption[Schematic illustration of the spin
115: pattern in RHIC.]{Schematic illustration of the first spin pattern,
116: with no zero-polarization crossings.  In one beam, alternate bunches
117: have opposite spin directions; in the other beam, the spin direction
118: changes every two bunches.} \label{figure:spinPattern}
119: \end{figure}
120: 
121: Crossing-by-crossing luminosity measurements from the $p$C
122: polarimeter provided a fill-dependent bad crossing list.  A total of
123: four individual crossings from two fills were discarded from the
124: analysis based on this information.
125: 
126: 
127: \section{EMCal-RICH trigger}
128: 
129: All data used for the final asymmetry results were from events
130: accepted by the \trig-tower tile trigger for high-$p_T$ (high energy
131: at mid-rapidity) photons. The \trig\ trigger had an average
132: rejection factor of 47, i.e.~only accepted on average one in 47 MB
133: events. It had a 78\% efficiency for neutral pions above $p_T
134: \approx 3.5$~GeV/$c$, as can be seen in
135: Fig.~\ref{figure:trigEfficiency}.
136: 
137: \begin{figure}
138: \centering
139: \includegraphics[angle=0,height=0.5\textheight]{%
140: trigEffic.eps} \caption[Trigger efficiency for neutral pions.]{a)
141: The \trig\ trigger efficiency for neutral pions, as a function of
142: pion transverse momentum.  The dashed line shows a Monte Carlo
143: simulation based on trigger tile efficiencies, and the solid line
144: indicates an upper limit on the $\pi^0$ efficiency based on the
145: number of active trigger tiles.  b) The fraction of the $\pi^0$
146: yield satisfying the MB trigger condition.  The solid line is the
147: fit of the data to a constant. The figure is taken from
148: \cite{Adler:2003pb}.} \label{figure:trigEfficiency}
149: \end{figure}
150: 
151: 
152: The effects of the trigger on the mean $p_T$ of the photon pairs
153: falling under the $\pi^0$ peak can be seen in
154: Table~\ref{table:meanPtMBTrig}.  The trigger raised the mean $p_T$
155: in the 1-2~GeV/$c$ bin significantly because it was still well below
156: its maximum efficiency in this transverse momentum range.  The
157: trigger had little effect on the mean $p_T$ of higher $p_T$ bins.
158: 
159: 
160: \begin{table}[tbp] \centering
161: \begin{tabular}{|c|c|c|}
162:   \hline
163:   % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...
164:   $p_T$ bin & $<p_T>$ MB & $<p_T>$ \trig \\
165:   (GeV/$c$) & (GeV/$c$) & (GeV/$c$) \\
166:   \hline
167:   1-2 & 1.27 & 1.38 \\
168:   2-3 & 2.32 & 2.33 \\
169:   3-4 & 3.33 & 3.35 \\
170:   4-5 & 4.39 & 4.36 \\
171:   \hline
172: \end{tabular}
173: \caption[Trigger effects on mean $p_T$ of photon pairs.]{Mean
174: $p_{T}$ of photon pairs under the $\pi^{0}$ mass peak for MB and
175: \trig-triggered data.} \label{table:meanPtMBTrig}
176: \end{table}
177: 
178: \section{Reduction of background}
179: In order to understand the background contribution to the \piz\ mass
180: peak and obtain \piz\ yields, the invariant-mass spectrum for photon
181: pairs was fitted.  The mass peak was fitted to a Gaussian and the
182: combinatorial background to a second-degree polynomial. The \piz\
183: yields per $p_{T}$ bin, given in Table~\ref{table:yieldsWidths},
184: were obtained by subtracting the background from the total number of
185: pairs in the peak.  An example fitted and subtracted invariant-mass
186: spectrum for $1 < p_T < 2$~GeV/$c$ is shown in
187: Figure~\ref{figure:invMass}. Mass bins of 10 MeV/$c^2$ were used in
188: this analysis.
189: 
190: \begin{table}[tbp] \centering%
191: \begin{tabular}{|c|c|c|} \hline
192: 
193: $p_T$ (GeV/$c$) & $\pi^0$ yield & Peak width (MeV/$c^2$) \\
194: \hline
195: 
196: 1-2 & 658k & 13.2 \\ \hline
197: 
198: 2-3 & 143k & 11.2 \\ \hline
199: 
200: 3-4 & 22k & 10.4 \\ \hline
201: 
202: 4-5 & 4k  & 10.6 \\ \hline
203: 
204: \hline
205: \end{tabular}%
206: \caption[$\pi^0$ yields and peak widths.]{$\pi^0$ yields obtained
207: after background subtraction; 1$\sigma$ \piz\ peak widths from a
208: Gaussian fit.}%
209: \label{table:yieldsWidths}
210: \end{table}%
211: 
212: \begin{figure}
213: \centering
214: \includegraphics[angle=0,height=0.6\textheight]{%
215: peakFit1To2GeV.eps} \caption[Fitted photon-pair invariant-mass
216: spectrum.]{Top panel: Invariant-mass spectrum for $1 < p_T <
217: 2$~GeV/$c$ photon pairs. Middle panel: Fitted spectrum. Bottom
218: panel: Subtracted spectrum.} \label{figure:invMass}
219: \end{figure}
220: 
221: The 1$\sigma$ peak widths from a Gaussian fit to the 120-160
222: MeV/$c^{2}$ mass region are shown in Table \ref{table:yieldsWidths}.
223: The transverse momentum spectrum for the pairs falling under the
224: \piz\ mass peak can be seen in Figure~\ref{figure:momSpectrum}.
225: 
226: 
227: \begin{figure}
228: \centering
229: \includegraphics[angle=0,height=0.45\textheight]{%
230: momSpectrum2x2.eps} \caption[$p_T$ spectrum for photon pairs falling
231: under the \piz\ mass peak.]{Raw transverse momentum spectrum for
232: photon pairs falling under the \piz\ mass peak.}
233: \label{figure:momSpectrum}
234: \end{figure}
235: 
236: 
237: Several cuts were made in order to reduce the background in the
238: photon pair sample.  Mismatched true photons, coming from two
239: different particles, were the main source of background. Other
240: sources of background included electrons, hadrons that deposited
241: energy in the EMCal, or secondary particles not coming from the
242: event vertex, all of which could lead to false combinatorial pairs
243: whose mass fell under the \piz\ mass peak, either in combination
244: with each other or with true photons from neutral pions.  The cuts
245: were:
246: 
247: \begin{itemize}
248: \item Minimum energy cut of 0.1 GeV in the PbSc, 0.2 GeV in the PbGl.  This cut
249: effectively eliminated pairs in which one photon carried nearly all
250: the energy and the other very little.  The same minimum energy cut
251: was used for the $\pi ^{0}$ $A_{LL}$ analysis, the results of which
252: were published in \cite{Adler:2004ps}.
253: 
254: \item Charged veto cut.  All EMCal clusters within a 10-cm radius of a
255: projected charged-track position onto the EMCal were excluded.
256: 
257: \item Shower shape cut in the PbSc to select clusters displaying the expected
258: shape for energy deposits from photon hits. The analogous
259: information for the PbGl was not available.
260: \end{itemize}
261: 
262: While a time-of-flight (TOF) cut could have potentially offered
263: additional hadron-photon discrimination, the EMCal TOF was not well
264: calibrated in the 2001-02 data set, so no timing cut was performed.
265: 
266: The background fraction, $r$, was obtained by taking the ratio of
267: the fitted background to the total number of pairs falling within
268: the mass peak. This fraction before and after the extra cuts were
269: performed is shown in Table~\ref{table:backgroundReduction}.
270: Significant reduction was achieved in the lowest two $p_T$ bins.
271: 
272: \begin{table}[tbp] \centering%
273: \begin{tabular}{|c|c|c|} \hline
274: $p_T$ (GeV/$c$) & \multicolumn{2}{|c|}{Background fraction (\%)} \\
275: \hline
276:  & Before cuts & After cuts \\ \hline
277: 1-2 & 58 & 34 \\ \hline
278: 
279: 2-3 & 23 & 12 \\ \hline
280: 
281: 3-4 & 12 & 6 \\ \hline
282: 4-5 &  9 & 5 \\
283:  \hline
284: 
285: \end{tabular}%
286: \caption[Reduction of background contribution.]{Reduction of
287: background contribution to the \piz\ mass peak, taken as 120-160
288: MeV/$c^2$, before and after background-removal cuts.}%
289: \label {table:backgroundReduction}
290: \end{table}%
291: 
292: 
293: 
294: \section{Asymmetry calculation}
295: \subsection{Overview}
296: The two counter-circulating RHIC beams are frequently referred to as
297: "blue" and "yellow," named after the colored stripes painted on the
298: respective magnet systems. The blue beam orbits clockwise, the
299: yellow beam counter-clockwise. Both beams are typically polarized,
300: as they were in the 2001-02 data-taking period. In order to make a
301: single-spin asymmetry measurement, the spin direction of the bunches
302: in only one beam was considered at a time, averaging over the spin
303: direction of the bunches in the other. Results from the blue and
304: yellow beams were obtained separately and subsequently combined.
305: 
306: The formula used to calculate asymmetry values is given in
307: Eq.~\ref{eq:lumiFormula},
308: 
309: \begin{equation}
310: \label{eq:lumiFormula}
311:   A_N = \frac{1}{P_{\textrm{beam}}}\frac{1}{\langle |\cos \varphi| \rangle}\frac{N^{\uparrow
312: }-\mathcal{R}N^{\downarrow }}{N^{\uparrow }+\mathcal{R}N^{\downarrow }} \\
313: \end{equation}
314: in which $P_{\textrm{beam}}$ is the beam polarization,
315: $\frac{1}{\langle |\cos \varphi| \rangle}$ is an azimuthal
316: acceptance correction factor (see below), $N^\uparrow$
317: ($N^\downarrow$) is the neutral pion yield from crossings with the
318: polarized bunch spin up (down), and $\mathcal{R} =
319: \mathcal{L}^\uparrow / \mathcal{L}^\downarrow$ is the relative
320: luminosity between crossings having the polarized bunch with spin up
321: versus down. As $A_N$ is a left-right asymmetry,
322: Eq.~\ref{eq:lumiFormula} must be used separately for the two
323: detector arms. As given, it applies to yields to the left of the
324: polarized beam; an overall minus sign is required for yields
325: observed to the right of the polarized beam.  Asymmetry results for
326: the left and right detector arms were obtained separately and then
327: combined.  The correction factors and relative luminosity are
328: discussed further below.
329: 
330: 
331: The beam polarization varies fill by fill.  Thus the asymmetry is
332: determined for every fill, then averaged over all fills. An example
333: of fill-by-fill asymmetries is given in
334: Figure~\ref{figure:asymVsFill}; this figure also gives an indication
335: of the fill-by-fill stability of the measured asymmetry. Large
336: observed variation among fills could indicate systematic errors.
337: (Note that all uncertainties given in figures and tables are
338: statistical unless stated otherwise.)  Table \ref{table:lumi2x2}
339: shows results obtained from Eq.~\ref{eq:lumiFormula} for the two
340: detector arms for triggered events. Note that the uncertainties are
341: slightly smaller for the yellow beam due to higher average
342: polarization. Combined results for both detector arms are shown in
343: Figure~\ref{figure:combinedLeftRight}, and the values for both arms
344: and beams combined are given in Table~\ref{table:combinedLumi}. All
345: results given in this section are before correction to the
346: asymmetries for the asymmetry of the background, which is described
347: in Section~\ref{section:backgroundSubtraction}.
348: 
349: \begin{figure}
350: \centering
351: \includegraphics[angle=0,height=0.5\textheight]{%
352: asymLumiVsFill1To2GeV.eps} \caption[Fill-by-fill
353: asymmetries.]{Fill-by-fill asymmetry results for the two beams. }
354: \label{figure:asymVsFill}
355: \end{figure}
356: 
357: 
358: \begin{table}  [tbp] \centering
359: \begin{tabular}{|c|c|c|c|c|c|} \hline
360:  $p_T$ (GeV/$c$) & Beam & \multicolumn{2}{|c|}{Left} &
361:  \multicolumn{2}{|c|}{Right} \\ \hline
362:   & & $A_N$ & $\sigma_{A_N}$ & $A_N$ & $\sigma_{A_N}$  \\
363: \hline
364: 
365: 1-2 & Blue & -0.008 & 0.015 & -0.012 & 0.019
366:  \\
367: 
368:  & Yellow & -0.005 & 0.015 & 0.003 & 0.011
369:  \\ \hline
370: 
371: 2-3 & Blue & 0.028 & 0.035 & -0.066 & 0.039 \\
372: 
373:  & Yellow & -0.016 & 0.030 & -0.003 & 0.028 \\ \hline
374: 
375: 3-4 & Blue & -0.101 & 0.094 & 0.106 & 0.099 \\
376: 
377:  & Yellow & 0.033 & 0.077 & -0.092 & 0.073 \\ \hline
378: 
379: 4-5 & Blue & -0.02 & 0.23 & 0.16 & 0.22 \\
380: 
381:  & Yellow & 0.01 & 0.17 & 0.14 & 0.18 \\ \hline
382: \end{tabular}%
383: \caption[Asymmetry results.]{Asymmetry results for particles
384: observed in the left detector arm (west arm for blue, east arm for
385: yellow) and right detector arm (east for blue, west for yellow).}
386: \label{table:lumi2x2}
387: \end{table}%
388: 
389: 
390: \begin{table}  [tbp] \centering
391: \begin{tabular}{|c|c|c|} \hline
392:  $p_T$ (GeV/$c$) & $A_N$ & $\sigma_{A_N}$   \\
393: \hline
394: 
395: 1-2 & -0.006 & 0.008
396:   \\ \hline
397: 
398: 2-3 & -0.014 & 0.017
399:   \\ \hline
400: 
401: 3-4 & -0.013 & 0.043
402:   \\ \hline
403: 
404: 4-5 & 0.070 & 0.101
405:   \\ \hline
406: 
407: \end{tabular}
408: \caption[Combined results for both beams and production to the left
409: and right.]{Combined asymmetry results for the two beams and
410: particle production on the two sides of the polarized beam.}
411: \label{table:combinedLumi}
412: \end{table}
413: 
414: 
415: 
416: \begin{figure}
417: \centering
418: \includegraphics[angle=0,height=0.5\textheight]{%
419: asymLumi2x2Combined.eps} \caption[Combined results for both beams
420: and production to the left and right.]{Combined asymmetry results
421: for the two beams and particle production to the left and right. }
422: \label{figure:combinedLeftRight}
423: \end{figure}
424: 
425: 
426: \subsection{Determination of relative luminosity}
427: 
428: The relative luminosity between crossings with spin-up bunches and
429: crossings with spin-down bunches for the polarized beam in
430: consideration was obtained from the number of MB events recorded by
431: the BBC. A typical value in this analysis was $\mathcal{R} =
432: \mathcal{L}^\uparrow / \mathcal{L}^\downarrow =$~1.09 for the yellow
433: beam, $\mathcal{R} =$~0.92 for the blue beam. Error in the relative
434: luminosity as measured in this case for azimuthal transverse
435: single-spin asymmetries could potentially come from an azimuthal
436: dependence of the BBC efficiencies.  The error would be proportional
437: to the physics asymmetry of the particles hitting the BBC (measured
438: to be $\lesssim 1$\%) times the difference in the efficiency of the
439: left and right halves of the detector. There is no current
440: measurement of this value, but it is expected to be small, also at
441: the level of a few times $10^{-2}$ or less. This would lead to a
442: potential systematic error on $\mathcal{R}$ on the order of a few
443: times $10^{-4}$, but this is only a rough estimate. Rather than
444: providing a quantitative error directly on the relative luminosity
445: measurement, an alternative method of asymmetry calculation that
446: does not rely upon measurement of the relative luminosity is used to
447: estimate the uncertainty on the asymmetry values calculated using
448: Eq.~\ref{eq:lumiFormula}. See Section~\ref{section:sqrtFormula} for
449: a description of this alternative method.  Refer to
450: Appendix~\ref{section:relLumi} for a discussion of potential sources
451: of error in the determination of the relative luminosity for
452: different single- and double-spin asymmetry measurements.
453: 
454: \subsection{Fill-by-fill polarization correction}
455: 
456: The average beam polarization in the 2001-02 run was $15\pm 5$\%,
457: with the 5\% representing a systematic scale uncertainty, discussed
458: below. Unpolarized protons in the beam act to dilute the physics
459: asymmetries being measured. It is necessary to correct for this
460: dilution, which is done by the factor of
461: $\frac{1}{P_{\textrm{beam}}}$ in Eq.~\ref{eq:lumiFormula}.
462: 
463: The beam polarization varies fill by fill and is typically different
464: for the two beams.  In the 2001-02 run, the yellow beam frequently
465: had slightly higher polarization than the blue beam.  The
466: polarization values for both beams for all fills included in this
467: analysis are given in Table~\ref{table:polarizationValues}.
468: Statistical uncertainties on the beam polarization were on the order
469: of $10^{-3}$ in absolute polarization.  These uncertainties were
470: negligible compared to the statistical uncertainties on the yields
471: as well as compared to the systematic uncertainty on the
472: polarization (see separate discussion below) and were not
473: incorporated into the final error on the asymmetry values.
474: 
475: \begin{table}\centering
476: \begin{tabular}{|c|c|c|} \hline
477: Fill & $P_{\textrm{blue}}$ & $P_{\textrm{yellow}}$ \\
478: \hline 2222 & 0.12 & 0.14 \\ \hline 2226 & 0.22 & 0.23
479: \\ \hline 2233 & 0.17 & 0.19 \\ \hline 2235 & 0.16 & 0.24 \\ \hline 2244 & 0.14 & 0.18 \\
480: \hline 2251 & 0.09 & 0.16 \\ \hline 2266 & 0.09 & 0.10
481: \\ \hline 2275 & 0.12 & 0.14 \\ \hline 2277 & 0.08 & 0.11 \\ \hline 2281 & 0.13 & 0.11
482: \\ \hline 2289 & 0.15 & 0.15 \\ \hline 2290 & 0.17 & 0.21 \\ \hline 2301 & 0.15 & 0.17 \\
483: \hline 2304 & 0.09 & 0.16 \\ \hline
484: \end{tabular}
485: \caption[Fill-by-fill beam polarization values.]{Fill-by-fill beam
486: polarization values for the 14 RHIC fills used.}
487: \label{table:polarizationValues}
488: \end{table}
489: 
490: 
491: \subsection{Acceptance correction}
492: \label{section:acceptanceCorrection}
493: 
494: The transverse single-spin asymmetry, $A_N$, is an azimuthal or
495: "left-right" asymmetry.  One can consider only particle production
496: to the left or right of the polarized beam and calculate the
497: asymmetry in production from spin-up versus spin-down bunches, as in
498: Eq.~\ref{eq:lumiFormula}.  Performing the calculation in this way,
499: detector acceptance effects cancel.  Alternatively, one can consider
500: only particle production from up- or down-polarized bunches and
501: calculate the asymmetry in production to the left versus the right,
502: as given in Eq.~\ref{eq:LRAsym}
503: 
504: \begin{equation}
505: \label{eq:LRAsym} A_N = \frac{1}{P_{\textrm{beam}}}\frac{1}{\langle
506: |\cos \varphi|
507: \rangle}\frac{N_{L}-\mathcal{R}_{\textrm{acc}}N_{R}}{N_{L}+\mathcal{R}_{\textrm{acc}}N_{R}}
508: \end{equation}
509: for bunch polarization in the upward direction. Here $N_{L}$ and
510: $N_{R}$ are the number of neutral pions produced in the left and
511: right detector arms with respect to the polarized beam direction,
512: and $\mathcal{R}_{\textrm{acc}} = \frac{\alpha_L}{\alpha_R}$ is the
513: relative acceptance of the left and right detector arms. In this
514: way, luminosity effects cancel. From Eq.~\ref{eq:LRAsym}, the
515: left-right nature of the asymmetry is clear.  However, maximal
516: effects, i.e.~the greatest and least particle production, are at
517: $90^\circ$ from the direction of the spin vector, which was vertical
518: in the entire 2001-02 run.  Thus integrating particle yields over
519: the entire azimuthal coverage of the central arm spectrometers would
520: lead to a dilution of the true physics asymmetry.  The factor of
521: $\frac{1}{\langle |\cos \varphi| \rangle}$ in both
522: Eq.~\ref{eq:lumiFormula} and \ref{eq:LRAsym} corrects for this
523: dilution.  Note that $\varphi = 0^\circ$ is in the horizontal plane,
524: implying no dilution at $\varphi = 0^\circ$ or $\varphi =
525: 180^\circ$.
526: 
527: The value of $\langle |\cos \varphi| \rangle $ over the idealized
528: azimuthal coverage of the detector arms can be calculated
529: analytically by Eq.~\ref{eq:phiIntegral}.
530: 
531: \begin{equation}
532: \label{eq:phiIntegral}
533:  \langle |\cos \varphi| \rangle =\frac{\int
534: |\cos \varphi | d\varphi}{\int d\varphi}
535: \end{equation}
536: To account more carefully for dead areas in the EMCal and in the
537: ERT, which would require detailed attention to the distribution of
538: these areas in order to make an analytical calculation, the average
539: values of $|\cos \varphi |$ used in the analysis were calculated
540: directly from the data using Eq.~\ref{eq:phiSum}, where $j$ is a sum
541: over all photon pairs with an invariant mass which fell under the
542: \piz\ mass peak. A comparison of results for $\langle |\cos \varphi
543: |\rangle $ determined analytically and from the data is given in
544: Table~\ref{table:cosPhi}.
545: 
546: \begin{equation}
547: \label{eq:phiSum}
548:  \langle |\cos \varphi |\rangle =\frac{\sum
549: \limits_{j=1}^{N} |\cos \varphi _{j}|}{N}
550: \end{equation}
551: 
552: \begin{table}[tbp] \centering%
553: \begin{tabular}{|c|c|c|c|}
554: \hline $\langle |\cos \varphi |\rangle $ & Ideal & Actual: MB &
555: Actual: \trig
556: \\ \hline
557: West Arm & 0.943 & 0.955 & 0.949 \\
558: East Arm & 0.883 & 0.880 & 0.874 \\
559: Both Arms & 0.909 & 0.920 & 0.913 \\ \hline
560: \end{tabular}%
561: \caption[Azimuthal acceptance correction factors.]{Comparison of
562: results for the average value of $|\cos \varphi |$ for ideal and
563: actual detector acceptances. Results for the west and east arms are
564: significantly different because the uppermost sector was not
565: included in the west.  Differences between the MB and triggered data
566: are due to the distribution of dead or masked tiles in the ERT.}
567: \label{table:cosPhi}
568: \end{table}
569: 
570: 
571: \subsection{Subtraction of background asymmetry}
572: \label{section:backgroundSubtraction}
573: 
574: Subtraction of the asymmetry of the background is performed as given
575: in Eq.~\ref{eq:backgroundSubtraction}, taking into account the
576: fraction of background under the $\pi^{0}$ mass peak.
577: $A_N^{\textrm{peak}}$ indicates the asymmetry of all photon pairs
578: falling under the \piz\ peak; note that it has generally been
579: written simply as $A_N$ up until this point.
580: Equation~\ref{eq:backgroundSubtractionError} gives the prescription
581: for calculation of the final statistical uncertainty on the \piz\
582: asymmetry after subtraction of the background asymmetry.
583: 
584: \begin{equation}
585: A_{N}^{\pi^{0}}=\frac{A_{N}^{\textrm{peak}}-rA_{N}^{\textrm{bg}}}{1-r}
586: \label{eq:backgroundSubtraction}
587: \end{equation}
588: 
589: \begin{equation}
590: \sigma _{A_{N}^{\pi ^{0}}}=\frac{\sqrt{\sigma
591: _{A_{N}^{\textrm{peak}}}^{2}+r^{2}\sigma
592: _{A_{N}^{\textrm{bg}}}^{2}}}{1-r}
593: \label{eq:backgroundSubtractionError}
594: \end{equation}
595: The same technique to handle background in a \piz\ asymmetry
596: analysis was used for the longitudinal double-spin asymmetry
597: \cite{Adler:2004ps}. Refer back to
598: Table~\ref{table:backgroundReduction} for the fraction of background
599: in each $p_{T}$ bin.  After cuts to reduce the background, it ranged
600: from 34\% in the 1-2 GeV/$c$ $p_T$ bin to 5\% in the 4-5 GeV/$c$
601: bin.
602: 
603: It is not possible to measure the asymmetry of the background under
604: the peak directly.  Differentiation between true neutral pions and
605: combinatorial background is only possible statistically and not on
606: an event-by-event basis.  Therefore it must be estimated in order to
607: correct for it. The asymmetries of two different background
608: invariant-mass regions were studied as estimates of the asymmetry of
609: the background under the \piz\ peak: 50-MeV/$c^2$ regions
610: immediately around the $\pi^0$ mass peak (60-110 MeV/$c^2$ and
611: 170-220 MeV/$c^2$), and the invariant-mass region above the $\pi^0$
612: but below the $\eta$ (250-450 MeV/$c^2$). The final results
613: published in \cite{Adler:2005in} subtracted the asymmetry of the two
614: 50-MeV/$c^2$ regions.  It was felt that this invariant-mass region,
615: being closer to that directly under the peak, was likely to reflect
616: the asymmetry of background under the peak more accurately. However,
617: the background region used was found to have little effect on the
618: final results. The similarity in the background asymmetries for the
619: two different invariant-mass regions lent confidence to their
620: validity in estimating the asymmetry of the background under the
621: \piz\ peak. See Section~\ref{section:backgroundStudy} for further
622: comparison of the results obtained from the two different background
623: regions.
624: 
625: The asymmetries measured for photon pairs falling in the background
626: invariant-mass region immediately surrounding the peak are shown in
627: Table~\ref{table:sidebands}.  The asymmetries after
628: background-asymmetry subtraction for the two 50-MeV/$c^2$ regions
629: are given in Table~\ref{table:subtractedSidebands}.  The asymmetry
630: results before and after subtraction of the background asymmetry are
631: shown in Figure~\ref{figure:peakSideband}.  It can be seen that
632: correction for the background asymmetry made only a small difference
633: in the results.  The mean $p_{T}$ in each bin was adjusted to
634: account for possible differences between the mean $p_T$ of the true
635: neutral pions and the background pairs under the peak. The final
636: value is calculated by Eq.~\ref{eq:meanPt}
637: \begin{equation}
638: p_{T}^{\pi^{0}}=\frac{p_T^{\textrm{peak}}-rp_{T}^{\textrm{bg}}}{1-r}
639: \label{eq:meanPt}
640: \end{equation}
641: and given for each $p_T$ bin in Table~\ref{table:meanPtFinal}.
642: 
643: 
644: \begin{table}  [tbp] \centering
645: \begin{tabular}{|c|c|c|} \hline
646:  $p_T$ (GeV/$c$) & $A_N^{\textrm{bg}}$ & $\sigma_{A_N^{\textrm{bg}}}$ \\
647: \hline
648: 
649: 1-2 & -0.007 & 0.009  \\ \hline
650: 
651: 2-3 & -0.031 & 0.034
652:   \\ \hline
653: 
654: 3-4 & 0.036 & 0.123
655:   \\ \hline
656: 
657: 4-5 & 0.42 & 0.39
658:   \\ \hline
659: 
660: \end{tabular}
661: \caption[Background asymmetries.]{Asymmetry results of background
662: photon pairs falling within 50-MeV/$c^2$ regions around the
663: $\pi^{0}$ mass peak.} \label{table:sidebands}
664: \end{table}
665: 
666: 
667: \begin{table}  [tbp] \centering
668: \begin{tabular}{|c|c|c|} \hline
669:  $p_T$ (GeV/$c$) & $A_N^{\pi^0}$ & $\sigma_{A_N^{\pi^0}}$  \\
670: \hline
671: 
672: 1-2 & -0.005 & 0.012
673:  \\ \hline
674: 
675: 2-3 & -0.012 & 0.020
676:  \\ \hline
677: 
678: 3-4 & -0.016 & 0.047
679:  \\ \hline
680: 
681: 4-5 & 0.052 & 0.109
682:  \\ \hline
683: 
684: \end{tabular}
685: \caption[Background-subtracted \piz\
686: asymmetries.]{Background-subtracted \piz\ asymmetries, using
687: 50-MeV/$c^2$ mass regions around the $\pi^{0}$ peak as the
688: background.}
689: \label{table:subtractedSidebands}%
690: \end{table}%
691: 
692: \begin{figure}
693: \centering
694: \includegraphics[angle=0,height=0.5\textheight]{%
695: asymLumiPeakSubtracted.eps} \caption[Asymmetry results before and
696: after correction for background.]{Asymmetry results before and after
697: correction for the asymmetry of the background in the invariant-mass
698: regions immediately surrounding the \piz\ peak. Background-corrected
699: points are shifted down by 50 MeV/$c$ from the center of the bin for
700: readability.} \label{figure:peakSideband}
701: \end{figure}
702: 
703: 
704: \begin{table}[tbp] \centering%
705: \begin{tabular}{|c|c|c|} \hline
706: \multicolumn{3}{|c|}{$\langle p_T \rangle$ (GeV/$c$)} \\ \hline
707: 
708: $\pi ^{0}$ peak & Background & Final \\ \hline
709: 
710: 1.40 & 1.31 & 1.45 \\ \hline
711: 
712: 2.34 & 2.28 & 2.34 \\ \hline
713: 
714: 3.35 & 3.33 & 3.36 \\ \hline
715: 
716: 4.38 & 4.37 & 4.38 \\ \hline
717: 
718: \end{tabular}%
719: \caption[Mean $p_T$ values of the background and after correction
720: for background.] {Mean $p_T$ values of the background and of neutral
721: pions after correction for background.} \label{table:meanPtFinal}
722: \end{table}%
723: 
724: 
725: 
726: \subsection{Calculation of statistical uncertainties}
727: 
728: In order to calculate the statistical error, accounting for the fact
729: that multiple neutral pions could be produced per collision, the
730: multiplicity distributions of $\pi^{0}$'s per event for the
731: different detector arms were determined.  See Figures
732: \ref{figure:mult1To2GeV} and \ref{figure:mult2To3GeV} for sample
733: multiplicity distributions in the west arm. Note that these
734: distributions in fact indicate the multiplicity per event of photon
735: pairs with an invariant mass between 120 and 160 MeV/$c^{2}$, thus
736: including both real $\pi^{0}$'s as well as false combinatorial
737: pairs.  From these multiplicity distributions, the degree to which
738: the distribution is non-Poisson was calculated in a simplistic way
739: by taking the ratio of the distribution's RMS to the square root of
740: its mean. For a Poisson distribution, this ratio would be 1.
741: 
742: \begin{figure}
743: \centering
744: \includegraphics[angle=0,height=0.4\textheight]{%
745: multBBCWest1To2GeV.eps} \caption[$\pi^0$ multiplicity per MB event,
746: $1 < p_T < 2$~GeV/$c$]{Number of photon pairs per MB event in
747: $\pi^0$ invariant-mass range, $1 < p_T < 2$~GeV/$c$, west arm. }
748: \label{figure:mult1To2GeV}
749: \end{figure}
750: 
751: \begin{figure}
752: \centering
753: \includegraphics[angle=0,height=0.4\textheight]{%
754: multBBCWest2To3GeV.eps} \caption[$\pi^0$ multiplicity per MB event,
755: $2 < p_T < 3$~GeV/$c$]{Number of photon pairs per event in $\pi^0$
756: invariant-mass range, $2 < p_T < 3$~GeV/$c$, west arm. }
757: \label{figure:mult2To3GeV}
758: \end{figure}
759: 
760: 
761: Table \ref{table:multFactors} shows the degree to which the photon
762: pair yield is non-Poisson, calculated as $\sigma _{k}/\sqrt{<k>}$,
763: where $<k>$ is the mean number of pairs per triggered event and
764: $\sigma _{k}$ is the RMS of this multiplicity distribution.  These
765: values were determined for the MB data, and then the uncertainty on
766: each individual yield, $N$, was taken to be $ \frac{\sigma
767: _{k}}{\sqrt{<k>}}\sqrt{N}$.  The 4-5 GeV/$c$ background bin suffers
768: from very low statistics. Rather than using the calculated value of
769: 1.41 for the west arm, 1.04, the same as for the 3-4 GeV/$c$
770: background bin, was used.
771: 
772: 
773: \begin{table}[tbp] \centering%
774: \begin{tabular}{|c|c|c|c|c|} \hline
775: 
776: $p_T$ (GeV/$c$) & \multicolumn{2}{|c|}{$\pi^0$ peak} & \multicolumn{2}{|c|}{Background} \\
777: \hline
778:     & West & East & West & East \\ \hline
779: 
780: 1-2 & 1.08 & 1.06 & 1.16 & 1.11 \\ \hline
781: 
782: 2-3 & 1.02 & 1.02 & 1.08 & 1.05 \\ \hline
783: 
784: 3-4 & 1.01 & 1.01 & 1.04 & 1.02 \\ \hline
785: 
786: 4-5 & 1.02 & 1.00 & 1.41 & 1.00 \\ \hline
787: 
788: \end{tabular}%
789: \caption[Deviation of particle production from a Poisson
790: distribution.]{Degree to which the yield is non-Poisson for MB
791: photon pairs falling in the 120-160 MeV/$c^2$ and (60-110 or 170-220
792: MeV/$c^2$) invariant-mass regions.}%
793: \label{table:multFactors}
794: \end{table}%
795: 
796: \subsection{Asymmetry scale uncertainty}
797: 
798: In the 2001-02 run, only the $p$C polarimeter was available in RHIC.
799: As stated in Section~\ref{section:polarimetry}, the analyzing power
800: in the process utilized by the $p$C polarimeter to measure the beam
801: polarization was originally measured by AGS experiment E950 to $\pm
802: 30\%$ \cite{Alekseev:2002ym}.  The total systematic error on the
803: measurement of the beam polarization was derived from a relative
804: systematic uncertainty on the RHIC beam measurement of 15\%, the
805: 30\% relative uncertainty on the analyzing power of the process, and
806: a relative uncertainty of 10\% in the change in analyzing power from
807: a beam energy of 22 GeV at the AGS to 100 GeV at RHIC
808: \cite{Jinnouchi:2003cp}.  Adding these uncertainties in quadrature
809: gave a total relative systematic uncertainty on the beam
810: polarization of $\pm 35\%$.  This uncertainty is a scale
811: uncertainty; it affects asymmetry values and statistical errors,
812: generally proportional to
813: $\frac{1}{P_{\textrm{beam}}}\frac{1}{\sqrt{N}}$, in the same way,
814: preserving the significance of each point from zero.
815: 
816: 
817: 
818: \section{Studies and checks}
819: \subsection{Alternative asymmetry calculation}
820: \label{section:sqrtFormula}
821: 
822: As described above in Section~\ref{section:acceptanceCorrection},
823: azimuthal transverse single-spin asymmetries can be considered for a
824: single polarization direction comparing particle production to the
825: left and right of the beam, or for particle production on a single
826: side of the beam comparing different polarization directions.
827: Equation~\ref{eq:sqrtFormula} combines yields from up- and
828: down-polarized bunches and from the left and right halves of the
829: detector such that systematic errors are reduced.
830: \begin{equation}\label{eq:sqrtFormula}
831: A_{N} =\frac{1}{P_{\textrm{beam}}}\frac{1}{\langle |\cos \varphi|
832: \rangle}\frac{\sqrt{N_{L}^{\uparrow }N_{R}^{\downarrow }}-
833: \sqrt{N_{L}^{\downarrow }N_{R}^{\uparrow }}}{\sqrt{N_{L}^{\uparrow
834: }N_{R}^{\downarrow }}+\sqrt{N_{L}^{\downarrow }N_{R}^{\uparrow }}}
835: \end{equation}
836: In particular, the acceptance and luminosity asymmetries cancel out
837: to several orders. See \cite{Spinka:1999vv} for a detailed
838: discussion of this and other methods of calculation for transverse
839: single-spin asymmetries.  It should be noted that while
840: Eq.~\ref{eq:lumiFormula} is mathematically exact,
841: Eq.~\ref{eq:sqrtFormula} is an approximation, albeit an excellent
842: one for the purposes of this analysis. It should also be noted that
843: Eq.~\ref{eq:sqrtFormula} is only suitable for transverse single-spin
844: analysis, while Eq.~\ref{eq:lumiFormula} has an equivalent for
845: longitudinal double-spin analysis (see
846: Eq.~\ref{eq:lumiFormulaDoubleLongitudinal}).
847: 
848: 
849: The asymmetry is determined for every fill, then averaged over all
850: fills, as in the calculations using Eq.~\ref{eq:lumiFormula}.
851: Table~\ref{table:sqrtAsym} shows the results obtained from
852: Eq.~\ref{eq:sqrtFormula} for triggered events.
853: 
854: 
855: \begin{table}  [tbp] \centering
856: \begin{tabular}{|c|c|c|c|} \hline
857:  $p_T$ (GeV/$c$) & Beam & $A_N$ & $\sigma_{A_N}$   \\
858: \hline
859: 
860: 1-2 & Blue & -0.010 & 0.012 \\
861: 
862:  & Yellow & -0.001 & 0.009  \\ \hline
863: 
864: 2-3 & Blue & -0.017 & 0.026  \\
865: 
866:  & Yellow & -0.009 & 0.020  \\ \hline
867: 
868: 3-4 & Blue & -0.002 & 0.068  \\
869: 
870:  & Yellow & -0.032 & 0.053  \\ \hline
871: 
872: 4-5 & Blue & 0.06 & 0.16  \\
873: 
874:  & Yellow & 0.08 & 0.12  \\ \hline
875: \end{tabular}
876: \caption[Asymmetry results, alternative calculation.]{Asymmetry
877: results as determined by Eq.~\ref{eq:sqrtFormula}.}
878: \label{table:sqrtAsym}
879: \end{table}
880: 
881: The results of the asymmetry calculations obtained using
882: Eq.~\ref{eq:lumiFormula} and Eq.~\ref{eq:sqrtFormula} can be seen
883: together in Figure~\ref{figure:sqrtLumi}.  The figure is for the
884: yellow beam; results for the blue beam are similar.  The two methods
885: agree extremely well. The dominant systematic uncertainty in the
886: results from Eq.~\ref{eq:lumiFormula} is expected to be from the
887: determination of the relative luminosity; therefore, systematic
888: uncertainties are calculated from a direct quantitative comparison
889: of the asymmetry results obtained from these two methods of
890: calculation.
891: 
892: \begin{figure}
893: \centering
894: \includegraphics[angle=0,height=0.5\textheight]{%
895: sqrtLumi2x2.eps} \caption[Comparison of two methods of asymmetry
896: calculation.]{Comparison of asymmetry results obtained using
897: Eq.~\ref{eq:lumiFormula} (Method 1) and Eq.~\ref{eq:sqrtFormula}
898: (Method 2), shown here for the yellow beam. Points for Method 2 are
899: shifted down by 50 MeV/$c$ from the center of the bin for
900: readability.} \label{figure:sqrtLumi}
901: \end{figure}
902: 
903: A systematic uncertainty, $\sigma_{sys}$, was calculated for each
904: bin from $A_N^{1} - A_N^{2} \equiv \Delta$, $\sigma_{A_N^{1}}$, and
905: $\sigma_{A_N^{2}}$.  The index '1' refers to results from
906: Eq.~\ref{eq:lumiFormula}; '2' to results from
907: Eq.~\ref{eq:sqrtFormula}.  The uncertainty on the difference was
908: calculated following a prescription for results obtained by applying
909: two different methods to data sets that are 100\% correlated
910: (exactly the same data) \cite{Barlow:2002yb}. In this case,
911: $\sigma_{\Delta} = \sqrt{| \sigma_{A_N^{1}}^2 + \sigma_{A_N^{2}}^2
912: -2\rho \sigma_{A_N^{1}} \sigma_{A_N^{2}}|}$, with the correlation
913: $\rho = 1$.  $\sigma_{sys} = |\Delta| - \sigma_{\Delta}$ was taken
914: in cases where $|\Delta|
915: > \sigma_{\Delta}$, and $\sigma_{sys} = 0$ was taken in cases where
916: $|\Delta| < \sigma_{\Delta}$.  See
917: Tables~\ref{table:lumiSqrtComparisonBlue} and
918: \ref{table:lumiSqrtComparisonYellow} for the values used in the
919: calculation and the results. The systematic error was also
920: calculated as the square root of the difference of the squares,
921: $\sigma_{sys} = \sqrt{|\Delta|^{2} - \sigma_{\Delta}^{2}}$, to see
922: if it affected the results.  At the level of $10^{-3}$, which is the
923: precision quoted in \cite{Adler:2005in}, it did not.  Final
924: systematic uncertainties calculated using this method, averaged over
925: the two beams, are shown in Table~\ref{table:sysError}.
926: 
927: \begin{table}
928: \begin{tabular}{|c|c|c|c|c|c|c|c|} \hline
929: $p_T$\  & $A_N^{1}$  & $A_N^{2}$  & $\Delta$ & $\sigma_{A_N^{1}}$  & $\sigma_{A_N^{2}}$ & $\sigma_{\Delta}$ & $\sigma_{sys}$\\
930: (GeV/$c$) & ($\times 10^2$) & ($\times 10^2$) & ($\times 10^2$) &
931: ($\times 10^2$) & ($\times 10^2$) & ($\times 10^2$) & ($\times
932: 10^2$)  \\ \hline  1-2       & -0.291 & -0.292 & 0.001 & 0.816 &
933: 0.816 & 0.000 & 0.001 \\ \hline 2-3       & -1.79  & -1.79  & 0.00 &
934: 2.07  & 2.07  & 0.00  & 0.00 \\ \hline 3-4       &  1.61  &  1.61 &
935: 0.00  & 5.51  & 5.51  & 0.00  & 0.00 \\ \hline 4-5       &  4.23 &
936: 4.17  & 0.06  & 13.0  &13.0   & 0.00  & 0.06 \\ \hline
937: \end{tabular}
938: \caption[Agreement of asym. results from the two methods, blue
939: beam.]{Agreement of asymmetry results from the two methods of
940: calculation, blue beam. Note that these values were obtained before
941: final cuts were performed.  See text for further explanation.}
942: \label{table:lumiSqrtComparisonBlue}
943: \end{table}
944: 
945: \begin{table}
946: \begin{tabular}{|c|c|c|c|c|c|c|c|} \hline
947: $p_T$  & $A_N^{1}$  & $A_N^{2}$  & $\Delta$ & $\sigma_{A_N^{1}}$  & $\sigma_{A_N^{2}}$ & $\sigma_{\Delta}$ & $\sigma_{sys}$\\
948: (GeV/$c$) & ($\times 10^2$) & ($\times 10^2$) & ($\times 10^2$) &
949: ($\times 10^2$) & ($\times 10^2$) & ($\times 10^2$) & ($\times
950: 10^2$)  \\ \hline
951: 
952: 1-2       &  -0.974 & -0.975 &  0.001 & 0.636 &  0.635 & 0.001 &
953: 0.000 \\ \hline
954: 
955: 2-3       &  -2.50  & -2.50 &  0.00  &  1.61  & 1.61  & 0.00  & 0.00
956: \\ \hline
957: 
958: 3-4       & -3.29  & -3.30  &  0.01 &  4.27  &  4.28  & 0.01  & 0.00
959: \\ \hline
960: 
961: 4-5       &   6.81  & 6.99   & -0.18  & 10.0 & 10.1   & 0.1   & 0.1
962: \\ \hline
963: \end{tabular}
964: \caption[Agreement of asym. results from the two methods, yellow
965: beam.]{Agreement of asymmetry results from the two methods of
966: calculation, yellow beam.  The data samples are 100\% correlated, so
967: agreement at better than the statistical level is expected. Note
968: that these values were obtained before final cuts were performed.
969: See text for further explanation.}
970: \label{table:lumiSqrtComparisonYellow}
971: \end{table}
972: 
973: \begin{table} \centering
974: \begin{tabular}{|c|c|} \hline
975: $p_T$   & Avg $\sigma_{sys}$ \\
976: (GeV/$c$) & ($\times 10^2$) \\ \hline
977: 
978: 1-2       & 0.000 \\ \hline
979: 
980: 2-3       & 0.00  \\ \hline
981: 
982: 3-4       & 0.00  \\ \hline
983: 
984: 4-5 & 0.1 \\ \hline
985: \end{tabular}
986: \caption[Systematic uncertainty on the asymmetry.]{Systematic
987: uncertainty on the neutral pion asymmetry, calculated for each bin
988: and averaged over both beams.} \label{table:sysError}
989: \end{table}
990: 
991: While the comparison of the asymmetry results from the two different
992: methods of calculation provided a strong check on the relative
993: luminosity, the main identified potential source of systematic error
994: present in the asymmetries as calculated by
995: Eq.~\ref{eq:lumiFormula}, a number of other checks on the results
996: were performed.  These other checks are described in the following
997: sections.
998: 
999: \subsection{Left and right detector arms}
1000: 
1001: Calculating the asymmetry using Eq.~\ref{eq:lumiFormula} necessarily
1002: gives separate results for the left and right sides of the polarized
1003: beam.  There is no overlap in the particle yields from the two
1004: detector arms; therefore, the expected agreement between the two
1005: results is that for uncorrelated samples. The uncertainty on the
1006: difference in results is taken to be $\sigma_{\Delta} = \sqrt{|
1007: \sigma_{A_N^{\textrm{left}}}^2 + \sigma_{A_N^{\textrm{right}}}^2
1008: -2\rho \sigma_{A_N^{\textrm{left}}}
1009: \sigma_{A_N^{\textrm{right}}}|}$, with the correlation $\rho = 0$.
1010: As can be seen from Table~\ref{table:leftRightAgreement}, the
1011: difference in the results from the two detector arms was within the
1012: uncertainty on the difference, and the results were in agreement.
1013: Results for both detector arms can be seen together for the blue and
1014: yellow beams in Figures~\ref{figure:leftRightAgreementBlue} and
1015: \ref{figure:leftRightAgreementYellow}, respectively.
1016: 
1017: \begin{table} \centering
1018: \begin{tabular}{|c|c|c|c|c|c|c|}
1019:   \hline
1020:   % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...
1021:   $p_T$ (GeV/$c$) & $A_N^{\textrm{left}}$ & $A_N^{\textrm{right}}$ & $\Delta$ & $\sigma^{\textrm{left}}$ & $\sigma^{\textrm{right}}$ & $\sigma_\Delta$ \\
1022:   \hline
1023: 
1024:   1-2 & -0.014 & -0.007 & -0.007 & 0.011 & 0.009 & 0.0142 \\
1025:   \hline
1026: 
1027:   2-3 & -0.034  & -0.021 & -0.013 & 0.026 & 0.024  & 0.0354 \\
1028:   \hline
1029: 
1030:   3-4 &  0.002  & -0.072 & 0.074 & 0.068 & 0.065 & 0.0941 \\
1031:   \hline
1032: 
1033:   4-5 &  0.01  &  0.13 & -0.12 & 0.15 & 0.16  &  0.219 \\
1034:   \hline
1035: \end{tabular}
1036: \caption[Agreement of results for production to the left and
1037: right.]{Evaluation of agreement of asymmetry results from particle
1038: production to the left and right of the polarized beam, shown for
1039: the yellow beam. $\Delta$ is the difference between the asymmetries
1040: for the two arms. $\sigma_\Delta$ is the uncertainty on the
1041: difference.} \label{table:leftRightAgreement}
1042: \end{table}
1043: 
1044: \begin{figure}
1045: \centering
1046: \includegraphics[angle=0,height=0.5\textheight]{%
1047: leftRightBlue.eps} \caption[Asym. comparison to the left and the
1048: right, blue beam polarized.]{Comparison of results for particle
1049: production to the left and right side, blue beam polarized. Points
1050: for the right of the beam are shifted down by 50 MeV/$c$ from the
1051: center of the bin for readability. Statistical agreement is
1052: expected.} \label{figure:leftRightAgreementBlue}
1053: \end{figure}
1054: 
1055: \begin{figure}
1056: \centering
1057: \includegraphics[angle=0,height=0.5\textheight]{%
1058: leftRightYellow.eps} \caption[Asym. comparison to the left and the
1059: right, yellow beam polarized.]{Comparison of results for particle
1060: production to the left and right side, yellow beam polarized.
1061: Points for the right of the beam are shifted down by 50 MeV/$c$ from
1062: the center of the bin for readability. Statistical agreement is
1063: expected.} \label{figure:leftRightAgreementYellow}
1064: \end{figure}
1065: 
1066: 
1067: \subsection{Two independent beams}
1068: 
1069: While the results for the blue and yellow beams use the same events
1070: and yields, they are combined in a different way, taking into
1071: account the spin direction of either one beam or the other, leading
1072: to (nearly) statistically independent measurements.  As the
1073: correlation between results from the two beams is believed to be
1074: small but is unknown, it is assumed to be zero in evaluating the
1075: agreement of the results. Thus as for the case of comparing results
1076: from the two detector arms, the uncertainty on the difference of the
1077: results from the two beams is taken to be $\sigma_{\Delta} = \sqrt{|
1078: \sigma_{A_N^{\textrm{blue}}}^2 + \sigma_{A_N^{\textrm{yellow}}}^2
1079: -2\rho \sigma_{A_N^{\textrm{blue}}}
1080: \sigma_{A_N^{\textrm{yellow}}}|}$, with $\rho = 0$.  It can be seen
1081: in Table~\ref{table:blueYellowAgreement} that the results from the
1082: two beams agree as expected.  Results for both beams can be seen
1083: together and evaluated by eye in
1084: Figure~\ref{figure:blueYellowAgreement}.
1085: 
1086: \begin{table} \centering
1087: \begin{tabular}{|c|c|c|c|c|c|c|}
1088:   \hline
1089:   % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...
1090:   $p_T$ (GeV/$c$) & $A_N^{\textrm{blue}}$ & $A_N^{\textrm{yellow}}$ & $\Delta$ & $\sigma^{\textrm{blue}}$ & $\sigma^{\textrm{yellow}}$ & $\sigma_\Delta$ \\
1091:   \hline
1092:   1-2 & -0.01035 & -0.00129 & -0.00906 & 0.01184 & 0.00922 & 0.01501 \\
1093:   \hline
1094:   2-3 & -0.0190 &  -0.0090 &  -0.0100 &  0.0264 &  0.0205 &  0.03342 \\
1095:   \hline
1096:   3-4 &  0.0026 &  -0.0292 &   0.0318 &  0.0681 &  0.0529 &  0.08623 \\
1097:   \hline
1098:   4-5 &  0.0662 &   0.0741 &  -0.0079 &  0.160  &  0.124  &  0.20243 \\
1099:   \hline
1100: \end{tabular}
1101: \caption[Agreement of results for individual beams.]{Evaluation of
1102: agreement of asymmetry results from blue and yellow beams. $\Delta$
1103: is the difference between the asymmetries for the two beams.
1104: $\sigma_\Delta$ is the uncertainty on the difference.}
1105: \label{table:blueYellowAgreement}
1106: \end{table}
1107: 
1108: \begin{figure}
1109: \centering
1110: \includegraphics[angle=0,height=0.5\textheight]{%
1111: blueYellow.eps} \caption[Asym. comparison for the two polarized
1112: beams.]{Comparison of results obtained for the two polarized beams.
1113: Points for the yellow beam are shifted down by 50 MeV/$c$ from the
1114: center of the bin for readability. Statistical agreement or better
1115: is expected.} \label{figure:blueYellowAgreement}
1116: \end{figure}
1117: 
1118: 
1119: \subsection{Triggered and minimum-bias data}
1120: 
1121: Results for the physics asymmetry, $A_N$, can also be compared for
1122: the triggered and minimum-bias data samples.  In this case the
1123: correlation between the samples is poorly understood.  In the higher
1124: $p_T$ bins, where the trigger was more efficient, nearly all pions
1125: in the MB sample should have fired the trigger and been present in
1126: the triggered sample as well, making the MB sample nearly a direct
1127: subset of the triggered sample. In the lower $p_T$ bins, this
1128: correlation should be lower but still significant.   Because of the
1129: unknown correlation and the fact that the MB sample is severely
1130: statistically inferior to the triggered sample, no direct evaluation
1131: of their agreement was performed.  However, it can be seen in
1132: Figure~\ref{figure:bbc2x2Agreement} that the MB and triggered
1133: results do not exhibit notable disagreement.
1134: 
1135: \begin{figure}
1136: \centering
1137: \includegraphics[angle=0,height=0.5\textheight]{%
1138: BBC2x2.eps} \caption[Asym. comparison for minimum-bias and triggered
1139: events.]{Comparison of results for minimum-bias and triggered
1140: events, shown here for the yellow beam and east detector arm. Points
1141: for the triggered data are shifted down by 50 MeV/$c$ from the
1142: center of the bin for readability. The correlation between the
1143: samples is unknown.} \label{figure:bbc2x2Agreement}
1144: \end{figure}
1145: 
1146: 
1147: \subsection{Different background regions in invariant mass}
1148: \label{section:backgroundStudy}
1149: 
1150: As described above, the asymmetry of two different background mass
1151: regions in invariant mass was investigated:  50-MeV/$c^2$ regions
1152: around the $\pi^{0}$ mass peak (60-110 MeV/$c^2$ and 170-220
1153: MeV/$c^2$) and the mass region between the $\pi^{0}$ and $\eta $
1154: (250-450 MeV/$c^2$). Table~\ref{table:backgroundComparison} shows
1155: the background asymmetries as calculated by Eq.~\ref{eq:lumiFormula}
1156: for the two background regions.  The asymmetries are similar.  The
1157: background asymmetries are consistent with zero for both background
1158: regions for $2 < p_T < 5$. In the 1-2~GeV/$c$ $p_T$ bin, both
1159: regions suggest a slightly negative asymmetry.
1160: 
1161: \begin{table}[tbp] \centering
1162: \begin{tabular}{|c|c|c|c|c|}
1163:   \hline
1164:   % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...
1165:   $p_T$ (GeV/$c$) & \multicolumn{2}{|c|}{$A_N$} &
1166:   \multicolumn{2}{|c|}{$\sigma_{A_N}$} \\ \hline
1167: 
1168:    & bg 1 & bg 2 & bg 1 & bg 2 \\
1169:    \hline
1170: 
1171:   1-2  & -0.008 & -0.014 & 0.005  & 0.004 \\ \hline
1172:   2-3  & -0.006 & 0.008 & 0.020  & 0.013 \\ \hline
1173:   3-4  & -0.012 & 0.015 & 0.079  & 0.055 \\ \hline
1174:   4-5  &  0.00  & 0.03  & 0.21   & 0.14 \\
1175:   \hline
1176: \end{tabular}
1177: \caption[Comparison of background asymmetries.]{Asymmetry results of
1178: photon pairs falling within the two 50-MeV/$c^2$ regions around the
1179: mass peak (bg 1) and within 250-450 MeV/$c^2$ (bg 2). Note that this
1180: check was performed before the final cuts on the data sample; thus,
1181: the results from bg 1 shown here differ from the final background
1182: results.} \label{table:backgroundComparison}
1183: \end{table}
1184: 
1185: In Figure \ref{figure:subtractedBackgrounds} a direct comparison of
1186: the asymmetry results after subtraction of the two background
1187: asymmetries is shown.  The background region used has little effect
1188: on the final asymmetry.
1189: 
1190: \begin{figure}
1191: \centering
1192: \includegraphics[height=0.5\textheight]{%
1193: asymMinusBackgrounds.eps} \caption[Asym. comparison after
1194: subtracting two different background regions.]{Comparison of
1195: asymmetries obtained after subtracting the asymmetry of two
1196: different background invariant-mass regions. "bg 1" indicates the
1197: asymmetry after subtraction of the asymmetry around the peak; "bg 2"
1198: is after subtraction of the asymmetry of the 250-450 MeV/$c^2$
1199: background region.  Points for bg 2 are shifted down by 50 MeV/$c$
1200: from the center of the bin for readability. Note that this check was
1201: performed before the final cuts on the data sample; thus the results
1202: differ from the final ones.} \label{figure:subtractedBackgrounds}
1203: \end{figure}
1204: 
1205: 
1206: \subsection{Different neutral pion invariant-mass integration regions}
1207: 
1208: As an additional check on the sensitivity of $A_N$ to the background
1209: under the \piz\ peak, three different integration regions for the
1210: \piz\ mass were examined:  120-160 MeV/$c^2$, 110-170 MeV/$c^2$, and
1211: 100-180 MeV/$c^2$.  Figure~\ref{figure:massWindowAgreement} shows a
1212: comparison of asymmetries obtained for peak integration from
1213: 120-160~MeV/$c^2$ and 100-180~MeV/$c^2$.  There is little effect on
1214: the result from the amount of background included, providing
1215: additional evidence that the background under the peak does not
1216: affect the asymmetry greatly.
1217: 
1218: \begin{figure}
1219: \centering
1220: \includegraphics[angle=0,height=0.5\textheight]{%
1221: massWindows.eps} \caption[Asym. comparison for different \piz\ peak
1222: integration regions.]{Comparison of results for different \piz\ peak
1223: integration regions.  Points for the wider integration region are
1224: shifted down by 50 MeV/$c$ from the center of the bin for
1225: readability.} \label{figure:massWindowAgreement}
1226: \end{figure}
1227: 
1228: 
1229: \subsection{Bunch shuffling}
1230: A technique called "bunch shuffling" can be utilized to check for
1231: uncorrelated bunch-to-bunch and fill-to-fill systematic errors.  For
1232: each bunch crossing the spin direction is reassigned randomly, and
1233: then the new asymmetry with false spin dependence is recalculated.
1234: This procedure is repeated many times. With random reassignment of
1235: the spin direction to each crossing, a Gaussian asymmetry
1236: distribution centered around zero is expected. One must take care in
1237: choosing the exact procedure used for this study.  If done
1238: correctly, the root-mean-square (RMS) width of the "shuffled"
1239: (non-physics) asymmetry distribution should correspond to the
1240: statistical uncertainty on the physics asymmetry, since fluctuations
1241: in the calculated asymmetry should not be due to any spin dependence
1242: but rather to statistical fluctuations in event-by-event particle
1243: production. If the RMS width of the shuffled asymmetry distribution
1244: is larger than the statistical uncertainty on the physics asymmetry,
1245: it should reflect the presence of elements that broaden the
1246: distribution beyond statistical fluctuations in particle production,
1247: e.g.~some source of bunch-to-bunch systematic error in the physics
1248: asymmetry.  It should be noted, however, that the expected
1249: quantitative agreement between the statistical uncertainty and RMS
1250: width of the shuffled distribution is not entirely understood.  As
1251: is discussed in Section~\ref{section:shufflingMC}, a study done for
1252: this thesis found that it is also possible to obtain widths smaller
1253: than the statistical uncertainties.
1254: 
1255: It is easier to consider the validity of various procedures assuming
1256: large asymmetries, for example 1 (100\%) .  Taking one bunch at a
1257: time and randomly assigning its spin direction to be up or down,
1258: with no further constraints, it is theoretically possible to make
1259: the spin assignments exactly as the true, original spin directions
1260: ($\textrm{up}_{\textrm{phys}} \rightarrow
1261: \textrm{up}_{\textrm{shuf}}$, $\textrm{down}_{\textrm{phys}}
1262: \rightarrow \textrm{down}_{\textrm{shuf}}$), or exactly opposite
1263: ($\textrm{up}_{\textrm{phys}} \rightarrow
1264: \textrm{down}_{\textrm{shuf}}$, $\textrm{down}_{\textrm{phys}}
1265: \rightarrow \textrm{up}_{\textrm{shuf}}$), yielding a shuffled
1266: asymmetry distribution that can range from -1 to +1, even with a
1267: wealth of statistics. This method would generally give distribution
1268: widths wider than the statistical uncertainties on the physics
1269: asymmetry values. Selecting any half of the bunches to be assigned
1270: up and the other half to be assigned down creates a similar
1271: situation for true relative luminosity values close to 1, i.e.~a
1272: nearly equal number of events coming from bunches with spin up and
1273: spin down: it is possible to assign the shuffled spins to be nearly
1274: the same as the original physics spin directions.
1275: 
1276: The procedure utilized in this analysis was to assign half of the
1277: original up spins to down and keep the others as up, and similarly
1278: assign half of the original down spins to up while keeping the
1279: others down. Thus the particle production gets redistributed evenly,
1280: at the \emph{crossing} (not event) level, between up and down spins,
1281: and even for large physics asymmetries, shuffled asymmetry will
1282: always be (nearly) zero. Repeating this procedure many times should
1283: yield a distribution around zero that is due to
1284: \emph{event-by-event} statistical fluctuations in particle
1285: production (i.e. 0, 1, 2, etc. neutral pions produced in a given
1286: event, with a certain probability distribution).
1287: 
1288: 
1289: For this analysis, Eq.~\ref{eq:sqrtFormula} was used to calculate
1290: the shuffled asymmetries, as it avoided the complication of
1291: recalculating the relative luminosity for each iteration. The bunch
1292: shuffling procedure was performed 1000 times. Refer to Figure
1293: \ref{figure:shuffledAsym} for examples of shuffled asymmetry
1294: distributions from the data set used for this analysis; the expected
1295: increase of the distribution width for the smaller-statistics
1296: (higher-$p_T$) bins can be clearly seen.  In Table
1297: \ref{table:shuffledMeans} it can be seen that the mean of the
1298: shuffled asymmetry distributions was zero at the level of $10^{-4}$
1299: or better. The $\chi^2$ distributions for a fit to a constant
1300: asymmetry across all shuffled fills can be seen in Figure
1301: \ref{figure:shuffledChiSq}. The dotted line shows the expected
1302: $\chi^2$ distribution for 13 degrees of freedom (asymmetry values
1303: for 14 fills, fit to a constant), which are in agreement with the
1304: data.
1305: 
1306: 
1307: \begin{figure}
1308: \centering
1309: \includegraphics[angle=0,height=0.5\textheight]{%
1310: shuffledAsymBlue.eps} \caption[Bunch-shuffled asymmetry
1311: distributions.]{Bunch-shuffled asymmetry distributions for triggered
1312: data, shown here for the blue beam. } \label{figure:shuffledAsym}
1313: \end{figure}
1314: 
1315: 
1316: \begin{figure}
1317: \centering
1318: \includegraphics[angle=0,height=0.5\textheight]{%
1319: shuffledChiSqBlue.eps} \caption[Bunch-shuffled $\chi^2$
1320: distributions.]{Bunch-shuffled $\chi^2$ distributions from triggered
1321: data, shown here for the blue beam. The dashed lines indicate the
1322: expected distributions.} \label{figure:shuffledChiSq}
1323: \end{figure}
1324: 
1325: 
1326: \begin{table}  [tbp] \centering
1327: \begin{tabular}{|c|c|c|} \hline
1328:  $p_T$ (GeV/$c$) & Beam & $\overline{A}_N^{\textrm{shuf}}$  \\ \hline
1329: 
1330: 1-2 & Blue & $-9.3 \times 10^{-6}$ \\
1331: 
1332:  & Yellow &  $-2.2 \times 10^{-5}$ \\ \hline
1333: 
1334: 2-3 & Blue & $-7.8 \times 10^{-7}$ \\
1335: 
1336:  & Yellow & $-2.2 \times 10^{-5}$ \\ \hline
1337: 
1338: 3-4 & Blue & $3.8 \times 10^{-5}$ \\
1339: 
1340:  & Yellow & $1.0 \times 10^{-4}$ \\ \hline
1341: 
1342: 4-5 & Blue & $3.0 \times 10^{-4}$ \\
1343: 
1344:  & Yellow & $-4.3 \times 10^{-4}$ \\ \hline
1345: \end{tabular}
1346: \caption[Shuffled asymmetry distribution means.]{Mean values of
1347: shuffled asymmetry distributions.} \label{table:shuffledMeans}
1348: \end{table}
1349: 
1350: Table \ref{table:shuffledAgreement} shows the statistical
1351: uncertainties on the physics asymmetries compared to the RMS widths
1352: of the shuffled asymmetry distributions for the blue and yellow
1353: beams. In order for the uncertainties on the physics asymmetries to
1354: reflect simply statistical distributions in particle production, the
1355: beam polarization was not factored in.  For this reason, the
1356: statistical uncertainties are exactly the same for the blue and
1357: yellow beams, the results for each using the same total set of
1358: events.  The \% difference is calculated as $100 \times
1359: (\sigma_{A_N^{\textrm{shuf}}} -
1360: \sigma_{A_N^{\textrm{phys}}})/\sigma_{A_N^{\textrm{phys}}}$.
1361: 
1362: \begin{table}[tbp] \centering%
1363: \begin{tabular}{|c|c|c|c|c|}
1364: \hline $p_T$ (GeV/$c$) & Beam & $\sigma_{A_N^{\textrm{phys}}}$ &
1365: $\sigma_{A_N^{\textrm{shuf}}}$ & \% difference \\ \hline
1366: 1-2 & Blue & 0.0014 & 0.0012 & -14 \\
1367:  & Yellow & 0.0014 & 0.0012 & -14 \\ \hline
1368: 2-3 & Blue & 0.0032 & 0.0030 & -6.3 \\
1369:  & Yellow & 0.0032 & 0.0028 & -13 \\ \hline
1370: 3-4 & Blue & 0.0082 & 0.0070 & -15 \\
1371:  & Yellow & 0.0082 & 0.0069 & -16 \\ \hline
1372: 4-5 & Blue & 0.0191 & 0.0179 & -6.3 \\
1373:  & Yellow & 0.0191 & 0.0179 & -6.3 \\ \hline
1374: \end{tabular}%
1375: \caption[Comparison of stat. uncertainties and widths of shuffled
1376: distributions.] {Comparison of statistical uncertainties on the
1377: physics asymmetries ($\sigma_{A_N^{\textrm{phys}}}$) and RMS widths
1378: of shuffled asymmetry distributions
1379: ($\sigma_{A_N^{\textrm{shuf}}}$). } \label{table:shuffledAgreement}
1380: \end{table}%
1381: 
1382: The shuffled widths are systematically \emph{smaller} than the
1383: statistical uncertainties calculated for the physics asymmetry. This
1384: outcome was at first a surprise, as the expectation was that the RMS
1385: width should have been the same as or greater than the physics
1386: statistical uncertainty. However, a subsequent Monte-Carlo study of
1387: the bunch shuffling technique, discussed below, corroborated the
1388: tendency for the width of shuffled distributions for a zero physics
1389: asymmetry to be narrower than the statistical uncertainty.
1390: 
1391: 
1392: \subsubsection{Bunch shuffling Monte Carlo}
1393: \label{section:shufflingMC}
1394: 
1395: A Monte Carlo study of the bunch shuffling technique was performed
1396: to better quantitatively understand the expected agreement between
1397: the RMS widths of shuffled asymmetry distributions and the
1398: statistical uncertainties of the physics asymmetry values, for
1399: different simulated physics asymmetry values.
1400: 
1401: In the Monte Carlo study, 14 fills were assumed, the same number as
1402: in the actual data set. In order to simulate the variation in the
1403: total particle yield per fill, the actual total \piz\ yields per
1404: fill for the 1-2 GeV/$c$ $p_T$ bin were used as the simulated yields
1405: per fill.  The number of bunches per beam was taken to be 48,
1406: selected because it was very close to the typical value of 46 in the
1407: actual data set, and because it was divisible by four, so that it
1408: was possible to assume half of the bunches were spin up and half
1409: were spin down, then easily reassign exactly half in each spin group
1410: to the wrong spin direction.  The desired physics asymmetry to
1411: simulate was selected, which determined the average yield per bunch
1412: with spins up and down in each fill. Simulated yields for each bunch
1413: crossing were produced by sampling from Poisson distributions around
1414: these asymmetry-dependent averages.  In this way, the uncertainty on
1415: the yields themselves was purely Poisson and well understood.  The
1416: uncertainty on the simulated asymmetry was purely statistical and
1417: calculated by performing straightforward error propagation on the
1418: asymmetry formula assuming only uncertainties on the yields; there
1419: were no effects due to detectors, triggers, polarization
1420: measurements, or any other factors incorporated. The simulated
1421: yields were then input into the same software program written to
1422: handle the bunch shuffling of the data. The RMS width of the
1423: resulting shuffled asymmetry distribution was then compared to the
1424: statistical error on the simulated "physics" asymmetry.
1425: 
1426: The effects of varying the simulated physics asymmetry, the total
1427: statistics, and the number of bunches per beam were studied.  In
1428: Table~\ref{table:shufflingMCMain} the results for two different
1429: statistical sample sizes as well as three different simulated
1430: asymmetry values are shown.  There are several items to note here.
1431: Despite incorrectly assigning the spin direction for exactly one
1432: half of the bunches, the width of the shuffled distribution
1433: increases noticeably as a function of the simulated physics
1434: asymmetry. The statistical uncertainty on the physics asymmetry is
1435: itself weakly but \emph{inversely} dependent on the physics
1436: asymmetry, leading to a shuffled width that is more than double the
1437: simulated uncertainty for a 90\% simulated asymmetry. For the case
1438: of a zero simulated asymmetry, corresponding to the results of the
1439: present analysis (see Section~\ref{section:results}), a shuffled
1440: width smaller than the simulated uncertainty is obtained for a
1441: statistical sample equivalent to the \piz\ yield in the 1-2 GeV/$c$
1442: $p_T$ bin in the present analysis. For ten times this statistical
1443: sample, the study suggests that the shuffled widths tend to increase
1444: with respect to the smaller sample, for at least the zero and 30\%
1445: simulated asymmetry values. In order to investigate the possibility
1446: that the smaller width for zero simulated asymmetry was due to
1447: correlations in obtaining shuffled asymmetries many times for a
1448: limited number of bunch crossings, a study with only 10 bunches per
1449: beam was performed, the results of which are given in
1450: Table~\ref{table:shufflingMCNumCrossings}. Performing 10,000
1451: reassignments of the spin direction of five out of ten bunches
1452: should have led to many duplicate reassignments and thus duplicate
1453: shuffled asymmetry values in the distribution. (Note that the
1454: difference in the number of iterations in the data versus the
1455: simulation was due simply to technical reasons.)  For the smaller
1456: statistical sample size, the shuffled distribution width was only
1457: slightly smaller for 10 bunches per beam than 48 bunches; for the
1458: larger sample size the reduction was more significant. While the
1459: degree of correlation should not depend on the value of the physics
1460: asymmetry, it may be that the reduction of the shuffled distribution
1461: width due to correlations only makes it less than the statistical
1462: uncertainty in the case of zero asymmetry because for larger
1463: asymmetries, the dependence of the width on the asymmetry is a
1464: larger effect.
1465: 
1466: As an additional means of investigating correlations as the cause of
1467: shuffled distribution widths being smaller than the statistical
1468: uncertainties, an attempt was made to remove all correlations.  To
1469: achieve this, the yield for each crossing was resampled for each
1470: iteration of the reassignment of spin directions.  This procedure
1471: effectively incorrectly assigned the spin direction for the yields
1472: from the bunches in many different fills ($\sharp$ effective fills =
1473: $\sharp$ shuffling iterations), all having the same physics
1474: asymmetry, rather than repeatedly reassigning the spin direction for
1475: the yields in a single fill or small number of fills. The results of
1476: this exercise are shown in
1477: Table~\ref{table:shufflingMCNoCorrelations}; the shuffled width is
1478: within 1\% of the statistical uncertainty on the simulated physics
1479: asymmetry, suggesting that the smaller shuffled distribution widths
1480: may be due to correlations.  Note that this procedure is not
1481: applicable to real data, due to the fact that there is always a
1482: limited number of fills.
1483: 
1484: 
1485: \begin{table} \centering
1486: \begin{tabular}{|c|c|c|c|c|}
1487:   \hline
1488:   % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...
1489:   Statistics & $A_N^{\textrm{sim}}$ & $\sigma_{A_N^{\textrm{sim}}} \times 10^4$ & $\sigma_{A_N^{\textrm{shuf}}} \times 10^4$ & \% difference \\
1490:   \hline
1491:   1 & 0 &  9.70 & 9.16 & -5.5 \\
1492:   \hline
1493:   1 & 0.3 & 9.20 & 9.18 & -0.2 \\
1494:   \hline
1495:   1 & 0.9 & 4.20 & 9.37 & 123 \\
1496:   \hline
1497:   10 & 0 & 3.05 & 3.09 & 1.3 \\
1498:   \hline
1499:   10 & 0.3 & 2.91 &  3.09 & 6.2 \\
1500:   \hline
1501:   10 & 0.9 & 1.33 & 2.99 & 125 \\
1502:   \hline
1503: \end{tabular}
1504: \caption[Shuffling results for different sample sizes and simulated
1505: asym. values.]{Bunch shuffling results for two different statistical
1506: sample sizes and three different simulated asymmetry values,
1507: assuming 48 bunches per beam. The "Statistics" column indicates the
1508: factor times the 1-2 GeV/$c$ $p_T$ bin \piz\ yields in the actual
1509: data sample.  10,000 shuffles were performed.}
1510: \label{table:shufflingMCMain}
1511: \end{table}
1512: 
1513: \begin{table} \centering
1514: \begin{tabular}{|c|c|c|c|c|c|}
1515:   \hline
1516:   % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...
1517:   \# bunches & Statistics & $A_N^{\textrm{sim}}$ & $\sigma_{A_N^{\textrm{sim}}} \times 10^4$ & $\sigma_{A_N^{\textrm{shuf}}} \times 10^4$ & \% difference \\
1518:   \hline
1519:   10 & 1 &  0  & 9.65 &  9.01 &  -6.6 \\
1520:   \hline
1521:   48 & 1 & 0 &  9.70 & 9.16 & -5.5 \\
1522: \hline
1523:   10 & 10 & 0  & 3.05 &  2.89 &  -5.2 \\
1524:   \hline
1525:   48 & 10 & 0 & 3.05 & 3.09 & 1.3 \\
1526:   \hline
1527: \end{tabular}
1528: \caption[Comparison of simulated shuffling results for 10 and 48
1529: bunches.]{Comparison of simulated shuffling results for 10 and 48
1530: bunches per beam. 10,000 shuffles were performed.}
1531: \label{table:shufflingMCNumCrossings}
1532: \end{table}
1533: 
1534: \begin{table} \centering
1535: \begin{tabular}{|c|c|c|c|c|}
1536:   \hline
1537:   % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...
1538:   Statistics & $A_N^{\textrm{sim}}$ & $\sigma_{A_N^{\textrm{sim}}} \times 10^4$ & $\sigma_{A_N^{\textrm{shuf}}} \times 10^4$ & \% difference \\
1539:   \hline
1540:   1 &  0 &  9.66 &  9.74 &  0.9 \\
1541:   \hline
1542: \end{tabular}
1543: \caption[Modified bunch shuffling simulation with correlations
1544: removed.]{Modified bunch shuffling simulation with correlations
1545: removed (see text).} \label{table:shufflingMCNoCorrelations}
1546: \end{table}
1547: 
1548: While the Monte Carlo investigation confirmed the pattern seen in
1549: applying the bunch shuffling technique to the actual data, the
1550: merits and limitations of the technique are still not completely
1551: understood.  Improved procedures for randomly reassigning the spin
1552: direction of the bunches in order to obtain better agreement between
1553: statistical uncertainties on the physics asymmetries and shuffled
1554: asymmetry distribution widths may exist. A more thorough future
1555: study would be valuable. This method of checking for uncorrelated
1556: bunch-to-bunch and fill-to-fill systematic errors is particularly
1557: important for double-spin asymmetry measurements, in which many of
1558: the checks available to a single-spin analysis are not available.
1559: 
1560: 
1561: 
1562: \section{Results}
1563: \label{section:results}
1564: 
1565: Final asymmetry results for mid-rapidity neutral pions from 200-GeV
1566: polarized $p+p$ collisions for $1 < p_T < 5$~GeV/$c$, as published
1567: in \cite{Adler:2005in}, are given in Figure~\ref{figure:finalAsym}
1568: and Table~\ref{table:finalPi0Results}. They utilized the triggered
1569: data sample and the 50-MeV/$c^2$ regions in invariant mass around
1570: the \piz\ peak for the background correction.  The asymmetries are
1571: consistent with zero within a few percent for all $p_T$ bins. For
1572: further discussion of these results and their implications, see
1573: Chapter~\ref{section:conclusions}.
1574: 
1575: \begin{figure}
1576: \centering
1577: \includegraphics[height=0.5\textheight]{%
1578: finalNeutral.eps} \caption[Final \piz\ asymmetry results.]{ Final
1579: mid-rapidity neutral pion transverse single-spin asymmetry. The
1580: error bars represent statistical uncertainties.}
1581: \label{figure:finalAsym}
1582: \end{figure}
1583: 
1584: 
1585: \begin{table} \centering
1586: \begin{tabular}{|c|c|c||c|c|c|} \hline
1587: $p_T$   & $\langle p_T \rangle$  & $r$  & $A_N^{\textrm{peak}}$  & $A_N^{\textrm{bg}}$ & $A_N^{\pi^0}$  \\
1588: (GeV/$c$) & (GeV/$c$)                 & (\%) & (\%)          & (\%)
1589: & (\%)          \\ \hline
1590: 
1591: 1-2       & 1.45 & 34   & -0.6$\pm$0.8  & -0.7$\pm$ 0.9 & -0.5$\pm$
1592: 1.2 \\ \hline
1593: 
1594: 2-3 & 2.34 & 12   & -1.4$\pm$1.7  & -3.1$\pm$ 3.4 & -1.2$\pm$ 2.0 \\
1595: \hline
1596: 
1597: 3-4 & 3.36 & 6 &  1.3$\pm$4.2  & 3.6$\pm$12.2 & -1.6$\pm$ 4.7
1598: \\ \hline
1599: 
1600: 4-5 & 4.38 &  5   & 7.0$\pm$10.1 & 42  $\pm$39   & 5.2$\pm$10.9 \\
1601: \hline
1602: \end{tabular}
1603: \caption[Asymmetry results.]{Neutral pion transverse single-spin
1604: asymmetry values and statistical uncertainties for all photon pairs
1605: falling within the $\pi^0$ mass peak, for the background
1606: ($\textrm{bg}$), and for the $\pi^0$ background-corrected.  The
1607: third column ($r$) indicates the background contribution under the
1608: $\pi^0$ peak.  An $A_N$ scale uncertainty of $\pm$35\% is not
1609: included.} \label{table:finalPi0Results}
1610: \end{table}
1611: 
1612: 
1613: \section{Comparison to charged hadron asymmetry results}
1614: \label{section:charged}
1615: 
1616: A similar analysis of the transverse single-spin asymmetry, $A_N$,
1617: of inclusive charged hadrons at mid-rapidity for $0.5 < p_T <
1618: 5.0$~GeV/$c$ was performed by F. Bauer for the 2001-02 data.  The
1619: results of the neutral pion and charged hadron measurements have
1620: been published together in \cite{Adler:2005in}.  For more details on
1621: the charged hadron analysis, refer to the publication. Charged
1622: hadron asymmetry results were obtained separately for positively and
1623: negatively charged particles; they are presented in
1624: Table~\ref{table:chargedAsymmetryResults}. While a clear charge
1625: dependence has been observed in transverse single-spin asymmetries
1626: for forward production (refer back to
1627: Section~\ref{section:transverseStructure}), the results for
1628: mid-rapidity neutral pions and both charges of hadron were all found
1629: to be similarly consistent with zero.
1630: Figure~\ref{figure:finalAsymChargedNeutral} shows the observed
1631: asymmetries for both neutral pions and charged hadrons together.
1632: 
1633: \begin{table} \centering
1634: \begin{tabular}{|c|c||c|c|} \hline
1635: $p_T$  & $\langle p_T \rangle$ & $A_N^{h^-}$  & $A_N^{h^+}$      \\
1636: (GeV/$c$) & (GeV/$c$)                & (\%)           & (\%)
1637: \\ \hline
1638: 
1639: 0.5-1     & 0.70  & -0.38$\pm$0.42 & -0.09$\pm$0.41
1640: \\ \hline
1641: 
1642: 1-2       & 1.32 & -0.12$\pm$0.82 & -0.54$\pm$0.78 \\ \hline
1643: 
1644: 2-5 & 2.56 & -2.1$ \pm$2.7  & -3.1 $\pm$2.6  \\ \hline
1645: \end{tabular}
1646: \caption[Charged hadron asymmetry results.]{Charged hadron
1647: transverse single-spin asymmetry values and statistical
1648: uncertainties.  An $A_N$ scale uncertainty of $\pm$35\% is not
1649: included.} \label{table:chargedAsymmetryResults}
1650: \end{table}
1651: 
1652: 
1653: \begin{figure}
1654: \centering
1655: \includegraphics[height=0.5\textheight]{%
1656: finalChargedNeutral.eps} \caption[Comparison of $A_N$ for
1657: mid-rapidity neutral pions and charged hadrons.]{ Mid-rapidity
1658: neutral pion and charged hadron transverse single-spin asymmetries
1659: versus mean $p_T$. Points for positive hadrons have been shifted
1660: down by 50 MeV/$c$ to improve readability. The error bars represent
1661: statistical uncertainties.} \label{figure:finalAsymChargedNeutral}
1662: \end{figure}
1663: