1: \documentclass{JHEP3}
2: %\accepted{LU-TP}
3: \keywords{QCD, e+-e- Experiments, Phenomenological Models}
4: %\received{hep-ph/0702241}
5: %\revised{May 23, 2002}
6: %\JHEP{05(2002)046}
7: \def\revision#1{$ #1\!\!$}
8: \preprint{LU-TP 07-07\\
9: hep-ph/0702241}
10:
11: % --- This bit puts ``draft'' over everything! ---****
12: %\special{!userdict begin /bop-hook{gsave 200 30 translate
13: %65 rotate /Times-Roman findfont 240 scalefont setfont
14: %0 0 moveto 0.9 setgray (DRAFT) show grestore}def end}
15: % --- end of this bit that puts `draft' over everything ---
16:
17: \usepackage{cite}
18: \usepackage{epsfig}
19: \usepackage{color}
20: \let\normalcolor\relax
21: \usepackage{graphics}
22: \usepackage{axodraw}
23: \usepackage{inputenc}
24: \usepackage{xspace}
25: \inputencoding{latin1}
26: \renewcommand\email[1]{{\scriptsize\tt\href{mailto:#1}{#1}}}
27: \newcommand{\comment}[1]{\footnote{\textbf{\boldmath $\Longrightarrow$ #1}}}
28: \newcommand{\gtaet}{\raisebox{-0.8mm}%
29: {\hspace{1mm}$\stackrel{>}{\sim}$\hspace{1mm}}}
30: \newcommand{\ltaeq}{\raisebox{-0.8mm}%
31: {\hspace{1mm}$\stackrel{<}{\sim}$\hspace{1mm}}}
32:
33: \newcommand{\abar}{\ensuremath{\overline{\alpha}}}
34: \newcommand{\as}{\ensuremath{{\alpha}_{s}}}
35: \newcommand{\azero}{\ensuremath{{\alpha}_{0}}}
36: \newcommand{\Cbar}{\overline{C}}
37: \renewcommand{\d}{\ensuremath{\mathrm{d}}}
38: \newcommand{\pT}{\ensuremath{p_{\perp}}}
39: \newcommand{\kT}{\ensuremath{k_{\perp}}}
40: \newcommand{\kTcut}{\ensuremath{k_{\perp 0}}}
41: \newcommand{\kTpot}[1]{\ensuremath{k_{\perp}^{#1}}}
42: \newcommand{\qbar}{\ensuremath{\overline{q}}}
43: \newcommand{\sigmahat}{\ensuremath{\hat{\sigma}}}
44: \newcommand{\smallx}{S\scalebox{0.8}{MALLX}\xspace}
45: \newcommand{\cascade}{C\scalebox{0.8}{ASCADE}\xspace}
46: \newcommand{\pythia}{P\scalebox{0.8}{YTHIA}\xspace}
47: \newcommand{\charybdis}{C\scalebox{0.8}{HARYBDIS}\xspace}
48: \newcommand{\ldcmc}{\scalebox{0.8}{LDCMC}\xspace}
49: \newcommand{\leading}{\textit{leading}\xspace}
50: \newcommand{\gluonic}{\textit{gluonic}\xspace}
51: \newcommand{\standard}{\textit{standard}\xspace}
52: \def\mrm#1{\mathrm{#1}}
53: \def\sub#1{\ensuremath{_{\mrm{#1}}}}
54: \def\sup#1{\ensuremath{^{\mrm{#1}}}}
55: \def\ordo#1{\ensuremath{{\cal O}(#1)}}
56: \def\sud#1{\ensuremath{\Delta_{S_{#1}}}}
57: \def\Sud#1{\ensuremath{{\mathbf\Delta\sub{S_{#1}}}}}
58: \newcommand{\todo}[1]{\vspace*{3mm}\par\noindent\textbf{\boldmath $\Longrightarrow$ #1}\vspace*{7mm}}
59: \newcommand{\done}[1]{}
60: \def\ftwo{\ensuremath{F_2}\xspace}
61: \newcommand{\eqref}[1]{eq.~(\ref{#1})\xspace}
62: \newcommand{\eqsref}[1]{eqs.~(\ref{#1})\xspace}
63: \def\rsch{\ensuremath{r\sub{Sch}}}
64: \def\Et{\ensuremath{E_\perp}}
65: \def\shat{\ensuremath{\hat{s}}}
66: \def\sighat{\ensuremath{\hat{\sigma}}}
67: \def\mpl{\ensuremath{M\sub{P}}}
68: \def\mpln#1{\ensuremath{M\sub{P #1}}}
69: \def\mmin{\ensuremath{M\sub{min}}}
70: \def\ie{i.e.\ }
71: \def\eg{e.g.\ }
72:
73: \skip\footins = 1\bigskipamount plus 2pt minus 4pt
74:
75: \title{\boldmath String Effects on Fermi--Dirac Correlation Measurements}
76:
77: \author{Rosa Mar\'ia Dur\'an Delgado, Gösta Gustafson and Leif Lönnblad\\
78: Dept.~of Theoretical Physics,
79: S\"olvegatan 14A, S-223 62 Lund, Sweden\\
80: E-mail: \email{rositaduran@yahoo.es}, \email{Gosta.Gustafson@thep.lu.se}
81: and \email{Leif.Lonnblad@thep.lu.se}}
82:
83: \abstract{We investigate some recent measurements of Fermi--Dirac
84: correlations by the LEP collaborations indicating surprisingly
85: small source radii for the production of baryons in
86: $e^+e^-$-annihilation at the $Z^0$ peak. In the hadronization
87: models there are besides the Fermi--Dirac correlation effect also a
88: strong dynamical (anti-)correlation. We demonstrate that the
89: extraction of the pure FD effect is highly dependent on a
90: realistic Monte Carlo event generator, both for separation of
91: those dynamical correlations which are not related to Fermi--Dirac
92: statistics, and for corrections of the data and background
93: subtractions. Although the model can be tuned to well reproduce
94: single particle distributions, there are large model-uncertainties
95: when it comes to correlations between identical baryons. We
96: therefore, unfortunately, have to conclude that it is at present
97: not possible to make any firm conclusion about the source radii
98: relevant for baryon production at LEP.}
99:
100:
101:
102:
103: %****************************************************
104: \begin{document}
105:
106:
107: \section{Introduction}
108: \label{sec:introduction}
109:
110: Hanbury-Brown and Twiss used Bose--Einstein (BE) correlations between
111: photons to measure the size of distant stars
112: \cite{HanburyBrown:1956pf}. A pair of bosons produced incoherently
113: from an extended source will have an enhanced probability, $P_{12}$,
114: to be found close in momentum space when detected simultaneously, as
115: compared to if they are detected separately ($P_1$, $P_2$). If the
116: production region has a Gaussian shape with some radius, $R$, it is
117: fairly easy to show that the enhancement is given by
118: \begin{equation}
119: \label{eq:BEbasic}
120: C\sub{BE}\equiv\frac{P_{12}}{P_1P_2}=1+e^{-Q^2R^2},
121: \end{equation}
122: where $Q^2=-(k_1-k_2)^2$ is the negative of the square of their
123: four-momentum difference.
124:
125: It was also early proposed to use a similar analysis to gain
126: information on the geometry of the production region for pions in
127: high-energy collisions \cite{Goldhaber:1960sf}. The assumption of
128: completely incoherent production in a Gaussian region is obvious when
129: considering photons from a star, and also very reasonable for the
130: production of pions in central heavy-ion collisions. For hadronic
131: collisions or $e^+e^-$ annihilation it may, however, seem much less
132: natural. The assumption of incoherent production implies that the
133: source is undisturbed by the emission, and thus not affected by the
134: enhanced radiation. In $e^+e^-$ annihilation the source disappears in
135: the hadronization process. Energy--momentum conservation is an
136: important constraint, and the source is not even approximately
137: constant. The distribution of pions is also far from isotropic,
138: usually concentrated in narrow jets, and further complicated by the
139: fact that the pions often come from decays of long-lived heavier
140: resonances. In spite of all these problems, introducing a so-called
141: caoticity parameter in eq.~(\ref{eq:BEbasic}), and still assuming a
142: Gaussian production region, all $e^+e^-$ experiments arrive at a
143: remarkably consistent value for the size of the production region:
144: $R\approx0.5-1$~fm.
145:
146: With a confining force, or string tension, of the order 1~GeV/fm this
147: might have been regarded as a small production region for collisions
148: at e.g.\ LEP energies where the $q\bar{q}$ pair is separated by about
149: 90~fm before they are stopped. However, in successful models based on
150: strings or cluster chains, hadrons which are close in momentum space
151: originate from regions which are also close in coordinate space.
152: Although the origin for the correlation is not fully understood, a
153: production region for pions or kaons of the order of 1~fm is therefore
154: quite reasonable.
155:
156: An attempt to explain the observed correlation as an effect of quantum
157: interference between different contributions to the production
158: amplitude in the string hadronization process is presented in
159: refs.~\cite{Andersson:1985qn, Andersson:1997hs}. Although this
160: approach gives a qualitatively correct
161: result, quantitative predictions at LEP energies have been hampered
162: by technical difficulties. Within this approach it has been argued that the
163: correlation between string pieces separated by a gluon should be strongly
164: reduced. When the center-of-mass energy is increased the number of gluons is
165: also large, and the mass of a straight string piece between two
166: gluons is kept relatively small. The result is therefore sensitive
167: to the hadronization of small string systems.
168: The iterative solution to the Lund string hadronization model
169: \cite{Andersson:1983jt} is only exact for high mass systems, and
170: although the corrections due to finite energy normally can be neglected for
171: \emph{inclusive distributions}, they do have a large impact on
172: \emph{correlations} \cite{Sandipanprivate}.
173:
174: \done{BE correlations in stars and multi-particle production,
175: measuring the size of the production region}
176:
177: A natural parallel to BE correlations is to look at the corresponding
178: correlation between identical fermions. Here one would expect a
179: depletion of fermion pairs close in momentum space, and using the same
180: assumptions for the production region as above we arrive at
181: \begin{equation}
182: \label{eq:FDbasic}
183: C\sub{FD}=1-\lambda e^{-Q^2R^2},
184: \end{equation}
185: where we now have explicitly included the ``caoticity parameter''
186: $\lambda$. Recently three of the LEP experiments have published
187: results on such Fermi--Dirac (FD) correlations for $pp$,
188: $\bar{p}\bar{p}$, $\Lambda\Lambda$ and $\bar{\Lambda}\bar{\Lambda}$,
189: again finding consistent results giving $R\sim0.15$~fm. This result is a
190: bit disturbing, not only because the size of the production is much
191: smaller than the one obtained for mesons, but also because the size is
192: smaller than the baryons themselves. It is therefore important to
193: thoroughly investigate possible theoretical and/or experimental problems
194: in the analyses.
195:
196: \done{FD correlations could be an important cross-check}
197:
198: \done{Measurments indicate a production region for baryons, smaller than
199: that of mesons, and smaller than the size of the baryons
200: themselves.}
201:
202: An essential problem in extracting the correlation is the estimate
203: of the reference distribution $P_1P_2$ in eq.~(\ref{eq:BEbasic}).
204: The distributions $P_1$ and $P_2$ must here correspond to exactly
205: identical events, i.e.\ events with the same emission of gluon
206: radiation (and the same orientation). If there are other correlations
207: beside the BE or FD effect, the distribution $P_1P_2$ should be
208: replaced by a reference distribution
209: corresponding to the two-particle distribution in a hypothetical world
210: without BE or FD correlations. If $N(Q)$ and $N_{\mathrm{ref}}(Q)$
211: represent the number of pairs in the real and the reference sample
212: respectively, the correlation is determined by the ratio
213: \begin{equation}
214: C(Q)=\frac{N(Q)}{N_{\mathrm{ref}}(Q)}.
215: \label{eq:CQ}
216: \end{equation}
217:
218: Such a reference sample is obviously not directly observable in an
219: experiment. Different methods have been applied for the construction
220: of reference samples, but they all suffer from serious limitations.
221: One possibility is to use a phenomenological hadronization model
222: without BE or FD correlations, implemented in a Monte Carlo event
223: generator. Besides a strong model dependence this method is also
224: sensitive to any imperfection in the event-generator implementation of
225: the underlying hadronization model.
226:
227: To reduce the model sensitivity it is therefore preferable to
228: construct reference samples directly from data. Two different methods
229: have been used for this purpose: To use \emph{opposite sign pairs} and
230: to use pairs from \emph{mixed events}. When studying correlations in
231: $\pi^+\pi^+$ or $\pi^-\pi^-$ pairs it may seem reasonable to use a
232: reference sample of $\pi^+\pi^-$ pairs, which are free from the BE
233: effect. The $\pi^+\pi^-$ pairs have, however, strong correlations due
234: to resonance contributions, and when using the ratio
235: $N_{\pi^+\pi^+}(Q^2)/N_{\pi^+\pi^-}(Q^2)$ to determine the BE effect,
236: it is therefore necessary to cut out the resonance regions in the fit,
237: or else to estimate their contributions to the distribution. The
238: method to use a reference sample with pairs of particles from
239: different events (a mixed reference sample) has the problem that the
240: hadronization is dependent on the gluon radiation, which differs from
241: event to event. This bias can be reduced (but not eliminated) by a cut
242: in thrust, which limits the radiation, and by orienting the events to
243: align the thrust axes.
244:
245: In this paper we want to discuss the special problems encountered when
246: trying to extract the effect of FD correlations in pairs of
247: (anti-)protons and $\Lambda$s. Naturally there are great experimental
248: difficulties in a determination of $\bar{p}\bar{p}$ or
249: $\Lambda\Lambda$ correlations, which are associated with acceptance
250: limitations, an admixture of pions in the (anti-)proton sample and
251: limited statistics in the $\Lambda\Lambda$ sample. We will here show
252: that the experimental analyses of momentum correlations also depend
253: very strongly on a realistic hadronization model, which can introduce
254: large errors due to uncertainties both in the model used and in its
255: implementation in \pythia\cite{Sjostrand:2003wg,Sjostrand:2006za}. We
256: therefore conclude that it presently is not possible to confirm the
257: very small production regions presented in the literature. It should
258: be mentioned that there is also a model-independent method to study
259: the FD effect, which is based on the spin correlation in
260: $\Lambda\Lambda$ pairs. This method is, however, limited by low
261: statistics, which also here prevents a definite conclusion.
262:
263: The layout of the paper is as follows: In section
264: \ref{sec:meas-fermi-dirac} we describe how the experiments extract FD
265: effects in correlations between identical baryons. Then in section
266: \ref{sec:bary-prod-lund} we present the basics of the Lund string
267: hadronization model, where we in particular discuss baryon production
268: and baryon--baryon correlation. In sec.~\ref{sec:uncertainties} we
269: discuss in more detail the uncertainties in the Lund model and the
270: approximations in its implementation in the \pythia event generator,
271: which have an impact on the baryon correlations. Finally our main
272: results are summarized in sec.~\ref{sec:conclusions}.
273:
274: \section{Measurements of Fermi--Dirac correlations}
275: \label{sec:meas-fermi-dirac}
276:
277: As mentioned in the introduction there are two different method,
278: which have been used to determine
279: the FD correlations between identical baryons. One is based on
280: momentum correlations similar to the analyses of BE correlations
281: between meson pairs discussed above, while the other uses
282: spin correlations in $\Lambda\Lambda$ pairs. While the first is
283: model dependent the second is hampered by low statistics. We
284: will in this section discuss the results of both methods.
285:
286: \subsection{Momentum correlations}
287: \label{sec:mom-corr}
288:
289: Experimental measurements (and theoretical expectations) show strong
290: correlations between a proton and an anti-proton or between a $\Lambda$
291: and an $\bar\Lambda$.
292: In the recent measurements of FD correlations between protons and
293: $\Lambda$s, the reference samples have therefore typically been constructed using
294: pairs of particles coming from different events, rather than opposite charge
295: particles.
296: To reduce the bias due to gluon radiation the events have frequently
297: been oriented so that the thrust axes end event planes coincide, and
298: sometimes also a lower cut on the thrust value has been applied.
299:
300: Associated with the baryon--anti-baryon correlations there are in
301: current hadronization models also very strong correlations between
302: identical baryons, besides those caused by gluon emission. These
303: correlations must be separated before the true FD effect can be
304: extracted. A way to reduce the bias due to both the gluon radiation
305: when using a mixed reference sample, and the dynamical correlations
306: which are not due to FD statistics, is to compare with expectations
307: from a Monte Carlo (MC) simulation program. This is usually done by
308: studying the "double ratio"
309: \begin{equation}
310: \label{eq:BEdouble}
311: C(Q)=\left(\frac{N^{\mathrm{exp}}(Q)}
312: {N_{\mathrm{ref}}^{\mathrm{exp}}(Q)}\right)\bigg/
313: \left(\frac{N\sup{MC}(Q)}{N_{\mathrm{ref}}\sup{MC}(Q)}\right).
314: \end{equation}
315: Here $N\sup{MC}$ and $N_{\mathrm{ref}}\sup{MC}$ represent the MC
316: generated real pairs and pairs from a generated reference sample of mixed
317: events. With a realistic MC this method should isolate the true FD effect.
318:
319: We want to emphasize that this method necessarily has the problem,
320: that the result is very sensitive to the description of the
321: correlations in the MC, which should be free of FD effects. These
322: correlations are much more uncertain than the inclusive distributions,
323: which have been accurately tuned to experimental data (see e.g.\
324: \cite{Hamacher:1995df}). Note also that they cannot be constrained by
325: data, as it is not possible to switch off FD effects in the
326: experiment. It is exactly the difference between the data and the
327: FD-free MC, which is interpreted as the FD effect. If the model and/or
328: its MC implementation is imperfect the result will be wrong. We here
329: also note that if the MC is tuned to correctly describe the single
330: proton spectrum, then it automatically also reproduces the pairs in
331: the mixed events. This implies that the double ratio in
332: eq.~(\ref{eq:BEdouble}) is actually very close to the single ratio,
333: with the MC result as the reference sample in eq.~(\ref{eq:CQ}). As
334: far as we know, there has been no measurements of correlations between
335: non-identical baryons, e.g.\ $p\Lambda$ and $\bar{p}\bar{\Lambda}$.
336: Such a measurement would be a very effective tool for validating the
337: baryon--baryon correlations in the hadronization models in the absence
338: of FD effects.
339:
340: The method with double ratios has been used in analyses by the three LEP
341: experiments ALEPH, OPAL, and DELPHI. Their results are presented in
342: table \ref{table}, and we see that they all find similar results
343: with a production radius of the order of 0.15~fm.
344:
345: \TABLE{
346: \begin{tabular}{c c c c l}
347: \hline
348: & \textbf{$R \: (fm)$} & \textbf{$\lambda$} & \textbf{Experiment}\\
349: \hline
350: $\bar{p}\bar{p}$ & $0.14 \pm 0.06$ & $0.76 \pm 0.33$ & OPAL\footnotemark[1]
351: & \cite{opal} \\
352: & $0.11 \pm 0.01$ & $0.49 \pm 0.09$ & ALEPH & \cite{Schael:2004qn} \\
353: & $0.16 \pm 0.05$ & $0.67 \pm 0.25$ & DELPHI & \cite{delphi1,delphi2}\\
354: \hline
355: $\Lambda\Lambda$ & $0.11 \pm 0.02$ & $0.59 \pm 0.10$ & ALEPH &
356: \cite{Barate:1999nv} \\
357: & $0.17 \pm 0.14 $ & Spin Analysis & ALEPH & \cite{Barate:1999nv} \\
358: & $0.19
359: \begin{array}[c]{c}
360: +0.37\\[-2mm]-0.07
361: \end{array}
362: $ & Spin Analysis & OPAL & \cite{Alexander:1996jq} \\
363: & $0.11
364: \begin{array}[c]{c}
365: +0.05\\[-2mm]-0.03
366: \end{array}
367: $ & Spin Analysis & DELPHI\footnotemark[1] & \cite{Lesiak:1998ij} \\
368: \hline
369: \end{tabular}
370: \caption{\small{Experimental results for $\lambda$ and $R$ from FD
371: correlations between baryon-pairs produced in $e^{+}e^{-}$
372: annihilations at the LEP collider.}}
373: \label{table}
374: }
375:
376: \footnotetext[1]{Note that these results has only been presented as a
377: preliminary.}
378: \subsection{Spin--spin correlations}
379: \label{sec:spin-corr}
380:
381: An alternative way to study the FD effect, which does not rely on
382: theoretical models and Monte Carlo simulations, is offered by the fact that
383: $\Lambda$ particles reveal their spin in the orientation of their decay
384: products. A $\Lambda\Lambda$ pair with total spin 1 must
385: have an antisymmetric spacial wave function and is therefore expected to show
386: a suppression for small relative momenta $Q$. $\Lambda\Lambda$ pairs with
387: total spin 0 has a symmetric spacial wave function, and should therefore
388: show an enhancement for small $Q$, similar to the correlation for bosons.
389: Therefore one expects pairs with small $Q$-values to be dominantly $S=0$.
390: This ought to be revealed in a preference for the protons from the
391: decaying $\Lambda$s to be more back to back in the di-$\Lambda$
392: center-of-mass system for small $Q$.
393:
394: % \FIGURE{
395: % \includegraphics[angle=0, scale=0.5]{./aleph.ps}
396: % \caption{\small{The corrected distributions $dN/dy^{*}|_{corr}$
397: % obtained for the $(\Lambda\Lambda)$ and
398: % $(\Lambda\bar{\Lambda})$ data, the latter free from FD
399: % correlations, by the ALEPH Collaboration \cite{Barate:1999nv}.}}
400: % \label{fig:2}
401: % }
402:
403: Analyses at LEP \cite{Alexander:1996jq,Abbiendi:1998ux,Barate:1999nv}
404: do indicate such an effect. An example is the distribution in
405: $dN/dy^*$ as obtained by the ALEPH collaboration (figure 4 in
406: \cite{Barate:1999nv}). Here $y^*$ is the cosine of the angle between
407: two protons in the di-$\Lambda$ center of mass system. When fitted to
408: a straight line the resulting slope does increase for low $Q$-values,
409: thus favouring back-to-back correlations. Results from three LEP
410: experiments are presented in table \ref{table}, and the fitted
411: production radii are consistent with those from analyses of momentum
412: correlations. This type of fit to an expected form gives a very small
413: error, but looking at the result in \cite{Barate:1999nv}, this error
414: cannot represent the real uncertainty in the data, which can also be
415: well fitted by a horizontal line for all values of $Q$. Unfortunately
416: we have to conclude that the statistics is too limited for a reliable
417: determination of the range of the FD effect using this method.
418: (Although in principle model independent, also this method needs Monte
419: Carlo simulations to correct for losses due to acceptance and for
420: background contributions.)
421:
422: \section{Baryon production in the Lund String Model}
423: \label{sec:bary-prod-lund}
424:
425: The most successful model of hadron production is the Lund string
426: fragmentation model. In this model it is easy to show that two
427: identical baryons cannot be produced close in rapidity along a jet
428: and, hence, with small $Q$, since flavour-number conservation requires
429: that an anti-baryon is produced in between. This need not be the case
430: in other models. In e.g.\ the cluster hadronization model, two nearby
431: clusters may both decay isotropically into baryon--anti-baryon pairs
432: resulting in two identical baryons close in momentum space. It should
433: be noted that although the models can be tuned to reproduce inclusive
434: particle spectra with high accuracy, there are large uncertainties in
435: the description of particle correlations in general and baryon
436: correlation in particular.
437:
438: As the Lund string model is used in the experimental analyses, we will
439: here describe this model in some detail. The Lund model is
440: based on the assumption that the colour-electric field is
441: confined to a linear structure, analogous to a vortex line in
442: a superconductor. The model contains two basic components:
443: A model for the breakup of a straight force field and a model for a gluon
444: as an excitation on the string-like field.
445:
446: \subsection{Breakup by $q\bar{q}$ pair production.}
447:
448: \FIGURE[t]{
449: % \begin{picture}(250,260)(0,-50)
450: % \LongArrow(-50,-20)(-30,-20)
451: % \LongArrow(-50,-20)(-50,0)
452: % \Text(-20,-20)[]{\footnotesize{$x$}}
453: % \Text(-50,10)[]{\footnotesize{$t$}}
454: % \Line(0,80)(100,-20)
455: % \Line(100,-20)(200,80)
456: % \Line(0,80)(50,130)
457: % \Line(50,130)(130,50)
458: % \Line(130,50)(180,100)
459: % \Line(200,80)(180,100)
460: % \Line(180,100)(230,150)
461: % \Line(180,100)(160,120)
462: % \Line(160,120)(210,170)
463: % \Line(230,150)(200,180)
464: % \Line(210,170)(220,180)
465: % \DashLine(130,50)(150,30){3}
466: % \DashLine(130,50)(80,0){3}
467: % \Text(110,10)[]{\footnotesize{$\kappa^{2}\:\Gamma_{1}^{2}$}}
468: % \Text(125,25)[]{$A$}
469: % \Text(170,70)[]{\footnotesize{$m_{1}^{2}$}}
470: % \Text(70,70)[]{\footnotesize{$m_{rem}^{2}$}}
471: % \LongArrow(150,10)(210,70)
472: % \LongArrow(150,10)(110,-30)
473: % \Text(145,-10)[l]{\footnotesize{$p_{+0}$}}
474: % \LongArrow(200,40)(220,60)
475: % \LongArrow(200,40)(170,10)
476: % \Text(200,30)[lu]{\footnotesize{$p_{+1}=z_{1} \cdot p_{+0}$}}
477: % \LongArrow(200,100)(210,90)
478: % \LongArrow(200,100)(191,109)
479: % \Text(210,110)[l]{\footnotesize{$p_{-1}=\frac{m_{1}^{2}}{p_{+1}}=
480: % \frac{m_{1}^{2}}{z_{1} \cdot p_{+0}}$}}
481: % \LongArrow(40,20)(90,-30)
482: % \LongArrow(40,20)(-10,70)
483: % \Text(30,10)[r]{\footnotesize{$p_{-0}$}}
484: % \Text(125,50)[r]{\footnotesize{$(x_{1},t_{1})$}}
485: % \Vertex(130,50){1.5}
486: % \end{picture}
487: \includegraphics*[bb=160 300 460 500]{./lunddecay.ps}
488: \caption{\label{fig:breakup} Schematic space--time picture of hadron
489: production in the Lund string model.}}
490:
491: The breakup of a high energy $q\bar{q}$ system is illustrated in
492: fig.~\ref{fig:breakup}.
493: We study first a simplified one-dimensional world with only one meson mass
494: and no baryon production. The probability for a final state
495: with $n$ mesons with mass $m$ and momenta $p_i$ is then given by the relation
496: \begin{equation}
497: dP \propto \prod_i^n\Bigl[N d^2 p_i\delta(p_i^2 - m^2)\Bigr]
498: \delta\!\left(\sum p_i - P_{\mathrm{tot}}\right) \times \exp(-b A).
499: \label{eq:breakup}
500: \end{equation}
501: Here $A$ is the space-time area indicated in fig.~\ref{fig:breakup}, while
502: $N$ and $b$ are two free parameters of the model.
503: The expression in eq.~(\ref{eq:breakup}) is a product of a phase space
504: factor and the exponent of a "colour coherence area", $A$, which can be
505: interpreted as the imaginary part of an action.
506: The phase space is specified by the parameter $N$, where a large
507: $N$-value favours many particles and a small $N$ few particles.
508: The value of $b$ specifies the strength of the imaginary action,
509: and here a large value favours early breakups and correspondingly
510: few particles in the final mesonic state.
511:
512: For a \emph{high energy system} the result in eq.~\ref{eq:breakup}
513: can be generated in an iterative way. The mesons can be "peeled
514: off" one by one from one end, where the $i$th meson takes the fraction
515: $z_i$ of the remaining (light-cone) momentum. The fractions $z_i$
516: are determined by the "splitting function"
517: \begin{equation}
518: f(z)=N\frac{(1-z)^a}{z} \exp(-\,\frac{b m^2}{z}).
519: \label{eq:f}
520: \end{equation}
521: Here the parameters $N$ and $b$ are the same as in
522: eq.~(\ref{eq:breakup}) (if measured in units such that the string
523: tension is 1) and $a$ is determined by the normalization constraint
524: $\int f(z) dz =1$. (In practice, $a$ and $b$ are treated as the free
525: parameters and $N$ is determined from normalization.)
526:
527: In the real world there are different quark species and different
528: meson masses. This is simulated by different weights $N_i$ in
529: eq.~(\ref{eq:f}), representing a suppression of strange quarks
530: and of the heavier vector mesons relative to pseudo-scalar mesons.
531: The parameter $b$ has to be a universal constant, but
532: $a$ can in principle vary depending on the quark flavour at the breakup,
533: although most fits to data assume a single value. It is also necessary to
534: include transverse momenta, where the meson mass, $m$, in
535: eqs.~(\ref{eq:breakup}, \ref{eq:f}) is replaced by the transverse
536: mass $m_\perp=\sqrt{m^2+p_\perp^2}$.
537:
538: We emphasized that the iterative procedure works only when the energy
539: is large. This implies that it is a bad approximation at the end of
540: the generation, when only little energy is remaining. In the \pythia
541: program this problem is solved by peeling off mesons randomly from
542: both ends of the string, making two hadrons when the remaining invariant
543: mass is small enough. This implies that the error from the
544: ``junction'' between the two halfs will be spread out and not visible
545: in inclusive distributions. This method will, however, not remove the
546: error on the correlations. They will still remain, as we will discuss
547: further below.
548:
549: \subsection{Gluon emission}
550:
551: The second basic feature of the Lund model is the assumption
552: that in three dimensions the dynamics of the confined force
553: field is well represented by the massless relativistic string,
554: and that the gluons act as transverse excitations on
555: this string \cite{Andersson:1979ij, Andersson:1980vj}. This
556: assumption implies angular asymmetries, which
557: were first observed by the JADE collaboration at the PETRA
558: accelerator \cite{Bartel:1981kh}.
559: A most important feature of this gluon model is
560: infrared stability. Soft or collinear gluons give
561: only small modifications of the string motion, and hence also
562: on the final hadronic state.
563:
564: Although the breakup of a string, which is bent by many gluons, should
565: also be determined by the area law in eq.~(\ref{eq:breakup}), there
566: are here ambiguities and technical problems. Is the projection of a
567: bent string piece onto a meson state the same as for a straight string
568: piece? The generalization of eq.~(\ref{eq:breakup}) and its
569: formulation in an iterative process in an event generator also imply
570: ambiguities \cite{Sjostrand:1984ic}. Although an error here does not
571: show up in inclusive distributions, it could well have effects on
572: correlations, and thus be important in analyses of the FD effects.
573:
574: \subsection{Baryon production}
575:
576: Baryon--anti-baryon pairs can be produced when the string breaks by the
577: production of a diquark--antidiquark pair in a $\bar{3}\!-\!3$ colour
578: state. The weights must here be adjusted so that the baryons become
579: fully symmetric spin-flavour states, which also preserves isospin
580: invariance. This mechanism gives strong correlations between the
581: baryon and the anti-baryon, which must have two quark flavours in
582: common. This correlation is stronger than what is observed, and for
583: this reason we have to imagine that the diquarks can be produced in a
584: step-like manner, allowing a meson to be produced between the baryon
585: and the anti-baryon, as indicated in fig.~\ref{fig:popcorn}
586: \cite{Andersson:1984af}. This so called "popcorn" mechanism also
587: implies that the baryon and the anti-baryon come farther apart in
588: rapidity and momentum space.
589:
590: \FIGURE[t]{
591: \includegraphics[angle=0, scale=0.5]{./pop}
592: \caption{\label{fig:popcorn} Representation of colour fluctuations
593: in the string. The wrongly coloured pair
594: $q_{1}(b)\bar{q}_{1}(\bar{b})$ together with
595: $q_{2}(g)\bar{q}_{2}(\bar{g})$ form an effective
596: diquark--antidiquark pair, yielding a baryon and an anti-baryon,
597: neighbors in rank. At the left-hand side we show the production
598: of a baryon and an anti-baryon with a meson between them, arising
599: from two breakups in a colour fluctuation region.}
600:
601: }
602:
603: A more elaborate model for baryon production is developed in
604: ref. \cite{Eden:1996xi}. As we find no significant differences
605: between this model and the standard popcorn model for the
606: distributions of interest in this paper, we will not discuss
607: it further here.
608:
609: \subsection{Baryon--anti-baryon correlations}
610:
611: The ordering of the hadrons along the string, the so called rank
612: ordering, agrees on average with the ordering
613: in rapidity, with an average separation, $\Delta y$, of the order of half
614: a unit in rapidity. As baryon production is suppressed compared to
615: meson production, a baryon--anti-baryon pair frequently originates from
616: a single diquark--antidiquark breakup. The baryon and the anti-baryon are
617: then produced as neighbours in rank, or with one (or a few) mesons in between,
618: which implies that they are not far away from each other in momentum
619: space. Two baryons must necessarily come from two different $B\bar{B}$
620: pairs, and must always be separated in rank by at least one anti-baryon
621: (and normally also with one or more mesons). This will give a strong
622: anti-correlation between two baryons in rapidity and in momentum.
623: This is illustrated in
624: fig.~\ref{fig:pp-par}, which shows correlations in $pp$ and $p\bar{p}$
625: pairs. We see that there is a very strong positive correlation between
626: protons and
627: anti-protons, which are frequently neighbours in rank, but a negative
628: correlation between two protons.
629:
630: \FIGURE[t]{
631: \includegraphics[angle=270, scale=0.4]{./corr.ps}
632: \caption{\label{fig:pp-par} Monte Carlo results for the ratio
633: $C_{MC}(Q) = MC(Q)/MC_{\mathrm{mix}}(Q)$ for $pp$- ($\times$) and
634: $p\bar{p}$-pairs (+).}
635: }
636:
637: In fig.~\ref{fig:pp-par} we see that the range for the $pp$ correlation
638: is given by $Q \sim 1.5\,\mathrm{GeV} \sim 1/(0.15\,\mathrm{fm})$. We
639: note that this corresponds exactly to the correlation length reported in
640: the experiments. The strength of the correlation is, however, smaller in
641: the simulation than in the data.
642:
643: \emph{This raises the question: Is the difference between data and
644: \pythia really a FD effect, or could the event generator
645: underestimate the strength of the correlation?}
646:
647: In the model there is a strong correlation between momentum and space
648: coordinates for the produced hadrons. Thus two identical baryons are
649: (in the model) well
650: separated also in coordinate space, and we would from this picture
651: expect Fermi--Dirac correlations
652: to correspond to a radius $\sim 2 - 3$~fm. If the dynamical
653: anti-correlation is indeed underestimated in \pythia, we would in the model
654: expect that the real FD correlation corresponds to a larger radius, and
655: therefore show up for $Q$-values around $1/(2\,\mathrm{fm}) \approx 0.1
656: \, \mathrm{GeV}$. Since the phase space is suppressed for these small
657: $Q$-values, this effect would be impossible to observe in the LEP
658: experiments.
659:
660: Although the fundamental nature of BE correlations is not understood,
661: its effects have been fairly well reproduced by a model in which the
662: momenta of the produced hadrons are shifted so that identical mesons
663: come closer in momentum space \cite{PYBOEI,Lonnblad:1995mr}. This
664: model is implemented in the program \texttt{PYBOEI} and included in
665: the \pythia package. With minor modifications, \texttt{PYBOEI} can
666: also be used to simulate FD correlations, in which case one would see
667: an expected drop for very small $Q$-values below 0.5 GeV assuming a
668: source radius of 1~fm.
669:
670: % \FIGURE[t]{
671: % \includegraphics[angle=270, scale=0.45]{./pyboei}
672: % \caption{\label{fig:pyboei} Modeled FD correlations (with crosses)
673: % compared to the input function \mbox{$f(Q) =\mathcal{N}[
674: % 1-\lambda\exp(-R^{2}Q^{2})]$} with $R=1\:fm$, $\lambda=1$.}}
675:
676: \section{Uncertainties in the event generator model}
677: \label{sec:uncertainties}
678:
679: There are a number of sources for uncertainty in the \pythia program:
680: \vspace{1mm}
681:
682: (1) There are two fundamental parameters, $a$ and $b$, in the
683: splitting function in eq.~(\ref{eq:f}).
684: The hadron multiplicity depends essentially on the ratio $(a+1)/b$,
685: This ratio is therefore well determined by experiments, but
686: $a$ and $b$ separately are more uncertain. Small values of $a$ and $b$
687: correspond to a wide distribution $f(z)$, and a wide distribution in
688: the separation, $\Delta y$, between hadrons which are neighbours in rank.
689: Large values of $a$ and $b$ imply a narrow $\Delta y$-distribution
690: and therefore lower probability for two particles to be close in
691: momentum space. The effect of varying $a$ and $b$ keeping the
692: multiplicity unchanged is shown in fig.~\ref{fig:abdep}. We see that
693: larger $a$- and $b$-values give a stronger anti-correlation for
694: small $Q$.
695:
696:
697: \FIGURE[t]{
698: \includegraphics[angle=270, scale=0.4]{./corrab.ps}
699: \caption{\label{fig:abdep} The ratio $C_{MC}(Q)=
700: MC(Q)/MC_{\mathrm{mix}}(Q)$ for different values of the parameters
701: $a$ and $b$. Larger $(a,b)$-values give a stronger correlation and
702: a deeper dip for small $Q$. The default values in the \protect\pythia
703: program is $a=0.3$ and $b=0.58$~GeV$^{-2}$}
704: }
705:
706: \FIGURE[t]{
707: \includegraphics[angle=270, scale=0.4]{./JUNCTION10.ps}
708: \caption{\label{fig:junc} A single jet \emph{without} a junction has
709: a large dip in $C_{MC}(Q)= MC(Q)/MC_{\mathrm{mix}}(Q)$ for small
710: $Q$ ($\times$). This correlation is reduced by the approximate
711: treatment in the Monte Carlo of the small mass systems close to the
712: "junction" (+).}
713: }
714:
715: \FIGURE[t]{
716: \includegraphics[angle=270, scale=0.4]{./PIONS4.ps}
717: \caption{\label{fig:pion} $C_{MC}(Q)= MC(Q)/MC_{\mathrm{mix}}(Q)$
718: with (+) and without ($\times$) a 15\% admixture of pions.}
719: }
720:
721: (2) The parameter $b$ is a universal constant, but $a$ may be different
722: for baryons, although data are well fitted by a universal $a$-value.
723: This gives some extra uncertainty.
724:
725: (3) As discussed in the previous section the \pythia program does not
726: exactly reproduce the Lund hadronization model. The splitting
727: function in eq.~(\ref{eq:f}) gives a correct result when the remaining
728: energy in the system is large. To minimize the error at the end of the
729: cascade, when the remaining invariant mass is small, the MC cuts off
730: hadrons from both ends randomly, and joins the two ends of the system
731: by producing two single hadrons. It is possible to adjust the cutoff
732: so that this method works well for inclusive distributions, but it
733: implies that the correlations do not correspond to the model
734: prediction for particles close to the ``junction''. To estimate the
735: error from this approximation we show in fig.~\ref{fig:junc} the $pp$
736: correlations in a single high energy jet without a junction, and
737: compare it with the standard result for a 91 GeV $q\bar{q}$-system.
738: In both cases no gluon radiation is included. We note here that a
739: significant part of the anti-correlation present in the model has
740: disappeared in the standard Monte Carlo implementation, as the
741: approximation in (and around) the junction allows baryons to be
742: produced with momenta close to each other.
743:
744: (4) Gluon emissions imply that straight string pieces are small compared to
745: the mass of a $B\bar{B}B$ system. This gives also extra uncertainty.
746: The gluon corners on the string imply ambiguities and approximations
747: in the hadronization model \cite{Sjostrand:1984ic}, which even if not
748: seen in inclusive distributions may be important for correlations.
749: This may well be a major reason why $\pi \pi$ correlations are not so well
750: reproduced by \pythia (see e.g.\ \cite{DeWolf:1992sx, Andersson:1994xd}).
751:
752: (5) Besides these model uncertainties, there are uncertainties in the
753: experimental correction procedure, such as the problems with event
754: mixing and thrust alignment mentioned in section
755: \ref{sec:meas-fermi-dirac}. In addition the identification of
756: anti-protons is not perfect in the experimental data. As an example
757: the D\textsc{elphi} anti-proton sample contains 15\% pions. As
758: mentioned above, the pion correlations are not perfectly reproduced by
759: \pythia. Fig.~\ref{fig:pion} shows results with and without a 15\%
760: pion admixture. We see that an error in the simulation of the
761: pion--pion or pion--proton correlations also will affect the estimated
762: proton--proton correlations.
763:
764:
765: In summary we see that there are many effects which make the Monte
766: Carlo predictions for $pp$ or $\Lambda\Lambda$ correlations quite
767: uncertain. As the experimental determination of the correlations rely
768: so strongly on a correct Monte Carlo simulation, it is therfore at
769: present premature to conclude that the production radius has the very
770: small value around 0.15~fm. As the expected dynamical $BB$
771: correlation has the same range ($Q\sim 1.5 \,\mathrm{GeV}$) as the
772: published results for the FD effect, we believe that it is more likely
773: that the strength of the dynamical $BB$ correlation is
774: underestimated in the \pythia event generator. The true FD effect may
775: then correspond to a larger production region and therefore show up at
776: $Q$-values too small for experimental observation.
777:
778: \section{Conclusions}
779: \label{sec:conclusions}
780:
781: The reported results on the production radius for baryon pairs is
782: clearly not consistent with the conventional picture of string
783: fragmentation. In fact, a production radius of $0.15$~fm, which is
784: smaller than the size of the baryons themselves seems difficult to
785: reconcile with any conceivable hadronization model.
786:
787: There are principal problems in the construction of a reference sample
788: which contains all dynamical baryon--baryon correlations but not the
789: effects of Fermi--Dirac statistics. When an event generator is tuned
790: to inclusive distributions, the use of the double ratio in
791: eq.~(\ref{eq:BEdouble}) implies that the result depends critically on
792: a perfect simulation program.
793:
794: In this article we have noted a number of uncertainties in the Lund
795: string fragmentation model and its implementation in the \pythia event
796: generator used in the correction of the data when extracting the
797: Fermi--Dirac effects. In particular we noted that the strong dynamical
798: correlation between baryons in the model appears in the same range of
799: $Q$ as the claimed FD correlations in the data, and that the model
800: uncertainties for these correlations are large and have not been
801: independently constrained by data.
802:
803: In conclusion we feel that it is premature to claim that the observed
804: discrepancy between data and Monte Carlo is really only due to
805: Fermi--Dirac effects, which would indicate a new kind of production
806: mechanism. To make such claims one would first have to demonstrate
807: that the models correctly describe baryon--baryon correlations in the
808: absence of FD effects, e.g.\ by comparing model predictions for
809: $p\Lambda$ and $\bar{p}\bar{\Lambda}$ correlations to experimental
810: data.
811:
812:
813: \bibliographystyle{utcaps}
814: \bibliography{/home/beckett/leif/personal/lib/tex/bib/references,refs}
815: \end{document}
816: