hep-ph0507008/jpg.ltx
1: \documentstyle[12pt,epsfig]{article}
2: \textheight 23.0cm \textwidth 16.0cm \topmargin -2.0cm
3: \oddsidemargin -1.0cm \evensidemargin -1.0cm \pagestyle{plain}
4: 
5: \begin{document}
6: \begin{titlepage}
7: \begin{center}
8: \vspace*{2cm}
9: 
10: { \Large  \bf  ON THE HIGH ORDER \\~
11: \\ MULTIPLICITY MOMENTS } \vspace{2cm}
12: 
13: \begin{author}
14: \Large K. Fia{\l}kowski\footnote{e-mail address:
15: uffialko@th.if.uj.edu.pl}, R. Wit\footnote{e-mail address:
16: wit@th.if.uj.edu.pl}
17: 
18: \end{author}
19: 
20: \vspace{1cm}
21: 
22: {\sl M. Smoluchowski Institute of Physics\\ Jagellonian University \\
23: 
24: 30-059 Krak{\'o}w, ul.Reymonta 4, Poland}
25: 
26: \vspace{2cm}
27: 
28: \begin{abstract}
29: The description of multiplicity distributions in terms of the
30: ratios of cumulants to factorial moments is analyzed both for data
31: and for the Monte Carlo generated events.  For the
32: PYTHIA generated events the moments are investigated for the
33: restricted range of phase-space and for the jets reconstructed from
34: single particle momenta. The results cast doubts on the validity of
35: extended local parton-hadron duality and suggest the possibility of
36: more effective experimental investigations concerning the origin of
37: the
38:  observed structure in the dependence of moments on their order.
39: 
40: \end{abstract}
41: 
42: \end{center}
43: \vspace{1cm}
44: 
45: PACS: 12.90.+b, 13.66.Bc \\
46: 
47: {\sl Keywords:}  Multiplicity moments, parton -  hadron duality  \\
48: 
49: \vspace{0.5cm}
50: 
51: 
52: \end{titlepage}
53: 
54: 
55: \section{Introduction}
56:    ~~~~For decades the investigation of the moments of multiplicity distributions for high energy
57:     hadroproduction processes has been limited to the lowest three orders. It was generally believed
58:     that the experimental errors make the results for the fourth and higher cumulants meaningless. The
59:     investigations of higher normalized factorial moments in terms of intermittency \cite{BP} was possible by
60:     means of averaging the distributions over many bins in phase space, thus efficiently increasing
61:     the statistics. Still, even in this case the fourth and higher moments seemed to be
62:     determined to large extent by the values of the first three moments.
63: \par   The breakthrough came with the analysis of Pavia group \cite{Pavia}, who have shown that the analysis can be
64:    significantlly extended for a new kind of moments: the ratios of cumulants to the factorial moments
65:    $$H_q = K_q/F_q.$$
66:     Here the factorial moments are defined in the standard way
67:     $$F_q=\sum_{n} \frac {n!}{(n-q)!}P(n),$$
68: and the cumulants may be calculated from a recursive formula
69: $$K_q=F_q-\sum_{i} \frac{(q-1)!}{(i-1)!(q-i)!}K_{q-i}F_i.$$
70: For
71:  the theoretical models it is sometimes easier to calculate moments
72: from the generating function of the multiplicity distribution
73: $$g(z) = \sum_{n} z^n P(n).$$
74:  The corresponding formulae read
75:  $$F_q=\frac{d^q}{dz^q}g(z)|_{z=1};~~~K_q=\frac{d^q}{dz^q}\log g(z)|_{z=1}.$$
76:  If the values of the $F_q$ moments of the order $q>3$ are determined by the values of lower order moments,
77:  the values of the $K_q$ moments are consistent with zero. In the Pavia group analysis the $H_q$ moments
78:  were shown to be significantly different from zero for $q=4,~5$ and $6$, and possibly also for $q>8$.
79: The reason why
 the new moments of higher orders seem to have much smaller
80: relative errors than the cumulants of the same order is
81: the cancellation of some contributions to the overall errors.
82: Fluctuations in the high multiplicity tail affect in a similar way
83: the cumulants and factorial moments, and the resulting
84: fluctuations of the values of their ratios are damped.
85: \par
86:    The behaviour  of the moments found by the Pavia group initiated much interest because of their apparent 
87:    compatibility with the
88:     predictions of perturbative QCD at the NLLA level \cite{D1},\cite{D2}. In both cases the dependence
89:     of moments on their order
90:     was found  to be non-monotonical: after a minimum (with negative value) at $q=5$, a hint of oscillations
91:     at higher $q$ was seen. Later, however, some doubts appeared about the
92:     origin of the observed structure. In particular, a simple cut in the smooth multiplicity
93:     distribution (e.g. of the negative binomial type) was shown to produce similar effects \cite{UGL}.
94:    In this note we analyze the behaviour of the moments calculated from the PYTHIA generator \cite{PYTHIA}
95:    for
96:  the electron-positron annihilation at LEP-I energy and compare them, where possible, with the
97:    experimental L3 results \cite{L3}, \cite{MA}. We include also some proposals for future investigations.
98: 
99: \section{Data and Monte Carlo results for Z decays}
100: ~~~~ To measure reliably higher multiplicity moments, one needs
101: high statistics and negligible systematic
102:      errors. The LEP-I data, where millions of events have been collected by each experiment using detectors covering
103:      almost $4\pi$ solid angle with very high efficiency, seem to be ideal for this purpose. The results of L3
104:      collaboration \cite{L3} were published recently (and compared with JETSET \cite{J} and HERWIG \cite{H}
105:       Monte Carlo results).
106:      In this section we remind these results and compare them with the moments calculated from PYTHIA.
107:      The comparison is only qualitative, as we do not use the L3 detector simulation program, which served
108:      to unfold the experimental multiplicity distribution \cite{L3}. Nevertheless, we will see that some
109:      surprising features of data and MC results are confirmed.
 We should note also that the overall event
110:      characteristics is properly described by PYTHIA: e.g., the values of the average charged multiplicity 
111:      (above $18$) and of the dispersion (about $5.9$) are reproduced correctly. 
112:      
113:    \par
114:    The values of $H_q$ moments calculated from the unfolded multiplicity distributions are reproduced in
115:    Fig.1a for the order $q=4\div 19$ together with the expectations of JETSET (the points for $q=2,3$
116:    are never shown in such plots, because they lie much higher). It is obvious that
117:    JETSET results agree with data well (we do not reproduce here the unsuccessful comparison of data with
118:    HERWIG).
119:     However, it is equally obvious that the errors
 read out directly from the figures 
120:    seem overestimated. The JETSET results (obtained without tuning the parameter values) never deviate
121:    from data by more than one standard deviation, and for the majority of
122:    moments agree with data within less than half of SD. This is statistically improbable and suggests that the
123:    values shown cannot be interpreted as simple uncorrelated statistical errors. Indeed, they are just the
124:    diagonal elements of the covariance matrix which is influenced by large bin-by-bin correlations.
125:    \par
126:     The authors seem to recognize this problem, and they do not use
127:    directly the multiplicity distributions obtained from the data by an unfolding procedure.
128:     Instead they use the multiplicity distributions truncated
129:    in such a way, that multiplicities with relative
130:    error on $P(n)$ greater than $50 \% $ are rejected.   This corresponds to the cut at highest
131:    multiplicities removing about $0.035 \% $ of events. With such a truncation the errors are
132:    visibly reduced and the clear structure appears both in data and MC results shown in Fig.1b:
133:    a minimum with negative values of $H_q$ at $q=5\div6$, and possible oscillations with maxima for $q$ around
134:    $10\div11$ and $16\div17$.
135:    
136:         \begin{figure}[h]
137: \centerline{\epsfig{figure=Fig1a.eps, height =8cm}}
138: \caption{\label{fig1} {\small \sl The L3 results \cite{L3} for $H_q$ moments calculated from full (a) and truncated
139:  (b) multiplicity distribution. Black points represent the data, and solid curve the events generated from JETSET.
140:  In Fig. 1b additional points and curves correspond to the separated $udsc$ and $b$ quark contributions.}}
141: \end{figure}
142: 
143:     \par
144:      In our opinion the lack of fluctuations above one standard deviation and the reduction of errors after the
145:      removal of some data suggest strongly that the error estimate  may be misleading. Thus for our simulations
146:      using PYTHIA we do not calculate errors from standard statistical formulae (relating the errors of the
147: q-th moment to some moments of order $2q$), but use instead the
148: uncertainty evaluated from the spread of results for different
149: samples of independently generated events.
150:  We use the statistics, for which the errors are comparable to the
151: size of symbols in the figures.
152:    \par
153:    The results for samples of $4$ million  events (more than twice the experimental sample of L3)
154:    are shown in Fig. 2. We have checked that the choice of PYTHIA parameters (default values,
155:    or the values tuned to L3 data) does not influence significantly our results. 
156:    We restricted our calculations to $q<11$, as the results for
157:    higher $q$ seem unstable even for such a huge statistics. We confirm the shallow minimum in the untruncated data,
158:    which becomes deeper after introducing the cut. Cutting off bigger part of the high multiplicity tail makes
159:    the minimum still more pronounced. 
160:    
161:  \begin{figure}[h]
162: \centerline{\epsfig{figure=Fig2a.eps, height =6cm}}
163: \caption{\label{fig2} {\small \sl The $H_q$ moments calculated
164: from PYTHIA events for  L3 parameters
. Crosses represent full distributions and $x$-s the
165: distributions with the highest multiplicity tail (0.035\% of cross
166: - section) removed.}}
167: \end{figure}
168:    
169:    \par
170:    We checked that the minimum in untruncated data does not change significantly if we change the number of
171:    generated events between $2$ and $4$ million; thus it is not due to the "natural cut" resulting form the finite
172:    statistics. This confirms once more the results from many other investigations:
173:    the existence of minimum
174:  seems to be universal, but its shape depends strongly on the multiplicity cuts.  If the statistics  is low,
175:    such cuts are "naturally" present in data; for high statistics experiments they may result from detector
176:    deficiencies.
177: 
178: \par
179:       The origin of the presence of a structure in the dependence of the $H_q$ moments on their order $q$
180:  for the PYTHIA events is not clear. We think that these effects are not due to the NNLO order QCD
181:  perturbative effects \cite{D1}, \cite{D2}, as the PYTHIA generator does not contain explicitly such components.
182:  A more plausible explanation is to relate the structure more generally to the multicomponent character
183:  of the production process \cite{D3}. An example of such  "multicomponent expansion" for Z decays may be a separation
184:  of two-, three- and multijet classes of events. It seems natural to assume a smooth (e.g. negative
185:  binomial - NBD) distribution to the two-jet events \cite{GU}. For such a distribution characterized by two parameters,
186:  $\overline n$ and $k$
187: $$ P_n^{NBD}(\overline{n},k) =
188: \frac{\Gamma(n+k)}{n!\Gamma(k)}\bigg(\frac{\overline
189: n}{\overline{n}+k}\bigg)^n\bigg(\frac{k}{\overline{n} +
190: k}\bigg)^k$$
191:  one can easily calculate the factorial moments
192: $$F_q^{NBD}(\overline{n}, k) =
193: \bigg(\frac{\overline{n}}{k}\bigg)^q\frac{\Gamma(k+q)}{\Gamma(k)}.
194: $$
195: \par
196: The distributions for three- and multijet events should be also
197: related to NBD. In a toy model, in which only two- and
198: three-jet events are taken into account, and the parameters in
199: this two classes are scaled in the $2:3$ ratio
 we have
200: $$ P_n = \alpha P_n^{NBD}(\overline n, k)
201: +(1-\alpha)P_n^{NBD}(3\overline n /2,3k/2).
202: $$
203: Obviously, in such a toy model we get for the factorial moments 
204: $$ F_q = \alpha F_q^{NBD}(\overline n, k)
205: +(1-\alpha)F_q^{NBD}(3\overline n /2,3k/2)
206: $$
207: and from the formulae given in the Introduction one may easily
208: write down the explicit formulae for the $H_q$ moments of any
209: order.
210:    We have calculated these moments for $q<16$ for the wide range of model parameters.
211:    We found that for the values of $\overline n$ and $k$ around $20$ (as suggested by the experimental
212:    values of the average multiplicity and dispersion), and for $a\sim  0.85$ (corresponding to a $15\%$
213:    admixture of three-jet events, as suggested by data) a minimum for $q=5$ (with a value around $-0.0005$)
214:    appears naturally. We checked also that a cut removing highest multiplicity events (at the level of $0.01\%$)
215:    enhances such a minimum. Thus any generator reproducing correctly the two- and three-jet components may be expected
216:    to reproduce also qualitatively the observed structure in the dependence of $H_q$ moments on their order $q$.
217: 
218:    \par
219:    We conclude that the minimum in the dependence of moments $H_q$ on their order $q$ seems to occur naturally at
220:    $q=5\div6$ for the hadron multiplicity distributions in the full phase-space. Its shape depends, however, on the
221:    details of the generating procedure. In particular, the cuts removing high multiplicity tail enhance the
222:    minimum quite strongly.
223: 
224: \section{Higher moments in restricted parts of phase-space}
225: \par
226: 
227:     The data and simulations discussed above concerned the full phase-space multiplicity distributions,
228:     unless limited by the experimental conditions. It is well known, however, that the multiplicity
229:     distributions in limited regions of phase-space depend significantly on the definition of the limits.
230:     For the early collider data \cite{UA5} the successful negative binomial (NB) fits were found both for full
231:     available phase-space and for the intervals of CM rapidity. However, the $1/k$ parameter (equal to second
232:     normalized cumulant for NB distributions) increases significantly for a decreasing size of the rapidity interval.
233:     It was approximately described by the hypothesis of $\overline{n}/k$ scaling \cite{VH} or by the minimal 
234:     model \cite{KF},
 in
235:  which the only significant correlations were those reflected by the multiplicity distributions for
236:     full phase-space. The data were not accurate enough to measure higher order cumulants and the
237:     statistics was insufficient to investigate small intervals (which was done later for intermittency
238:     effects using the averaging over many intervals \cite{BP}).
239: 
240:     \par
241:    The LEP1 data have sufficient accuracy and statistics to investigate higher moments for various regions
242:    of phase-space. Thus we have performed calculations of such moments for the events generated with PYTHIA,
243:    defining phase-space regions by cuts in values of CM momenta. It should be interesting to see if the
244:    patterns revealed by such calculations will be confirmed by future data analysis.
245:    \par
246:    We have generated samples of $4M$ events both for the default values of PYTHIA parameters and for the values
247:    used by L3 collaboration. We calculated the moments for orders $2\div 10$ for regions defined by simple
248:    inequality for the values of the CM three-momenta $p$: $p<\epsilon n$, where $\epsilon =0.2GeV$, and $n=1\div 10$.
249:   \par
250:    Although the results for two choices of parameters are different, there is a common
251:    pattern in them; thus we show only the results for $L3$ parameters.
252:     The minimum (with negative values) in the dependence of moments $H_q$ on their order
253:    $q$ occurs for small phase space regions very late (at $q$ around $8$) and shifts gradually to smaller
254:    values of $q$ for increasing range of CM momenta. However, even for the widest range of momenta investigated,
255:    the position and shape of minimum differs significantly from that for full phase-space.
256:     The results are shown in Fig.3a; for transparency we show only points for $n$ =4,6,8 and 10.
257: 
258:             \begin{figure}[h]
259: \centerline{\epsfig{figure=fig4.eps, height =5.5cm}}
260: \caption{\label{fig3} {\small \sl The  $H_q$ moments calculated
261: from PYTHIA events in limited parts of phase-space (see text) for
262:  L3 values of the parameters without (a) and with (b) a cut removing highest multiplicities.
263:   Crosses, $x$-s, stars and squares represent the results for $n$ = 4, 6, 8 and 10,
264: respectively.}}
265: \end{figure}
266: 
267:    \par
268:     In the previous chapter we have seen that the minimum for full phase-space was strongly enhanced by
269:     introducing cuts in the multiplicity distributions, although for the LEP1 data the minimum cannot be
270:     explained just by such a naturally occurring cut due to the finite statistics. Thus we repeated our
271:     calculations cutting off the highest multiplicities contributing $0.001$ to the full cross-section.
272:     The results, shown in Fig. 3b, are quite surprising: now the position of minimum is practically
273:     independent on the size of phase-space region selected. It will be very interesting to see if the
274:     data show the same patterns when investigated with- and without extra cuts in multiplicity.
275: 
276: \section{Higher moments for hadrons and for jets}
277: 
278: \par
279:    As noted in the introduction, the existence of minimum (with negative values) in the $q$-dependence
280:    of moments $H_q$ was related to the higher order corrections in
281:    perturbative QCD. It is obvious, however, that the perturbative QCD calculations yield the
282:    multiplicity distributions for gluons, and not for single hadrons. The identifications of
283:    higher moments for those two distributions means a rather bold extension of the assumption of
284:    parton-hadron duality, usually applied only to the average values.
285:    \par
286: To estimate the reliability of such an extension we applied to the
287: generated events the default clustering algorithm of PYTHIA (PYCLUS) and investigated the
288: multiplicity distributions of reconstructed jets (which may be
289: expected to correspond to the gluon distribution more closely,
290: than the distribution of single hadrons).
291:     The average number of jets depends strongly on the values of two parameters. One of them defines the maximal
292:     phase-space distance between the particles added to the existing jet; the other one defines the maximal CM momentum
293:     for the  slowest particles in CM, which form a separate cluster/jet.
294:     \par
295:      We calculated the ten lowest moments
296:     for the distributions of jets defined with those parameters spanning a range of $0.02-0.1$ GeV, 
297:     for the L3 parameters, with a cut in the multiplicity. An additional degree
298:     of freedom was introduced by including in the clustering algorithm not only the charged stable hadrons
299:     (stable in the experimental sense, i.e. coming to the detectors; this is the condition used in all the distributions), but
300:     also stable neutral particles. These are mainly photons (coming from the ${\pi}^0$ decays), but also
301:     (anti)neutrons and long-living
302:     neutral kaons. This option allows to compare the distribution of hadrons and jets with the same
303:     average multiplicity.
304:    The results of our calculations are shown in Fig.4. It is obvious that the values of moments
 for
305:  jets are very different from those for hadrons, even for relatively small jets (when the multiplicity
306:    in both cases is similar); e.g., the fourth moment is now negative.
307: 
308:             \begin{figure}[h]
309: \centerline{\epsfig{figure=Fig5a.eps, height =5.5cm}}
310: \caption{\label{fig4} {\small \sl The $H_q$ moments for jets
311: reconstructed from hadrons for PYTHIA events generated with L3
312: parameters. In Fig. (a) only charged hadrons are used; in Fig. (b)
313: all stable particles are counted. Crosses, $x$-s, stars, open and
314: closed squares correspond to the increasing average number of
315: particles in a jet (from about 1.1 to 1.5 in Fig. (a) and from 1.3
316: to 2 in Fig. (b).}}
317: \end{figure}
318: 
319:   \par
320:    When we count all the stable particles, we may select the values of
321:    jet defining parameters
322:    for which their average multiplicity is the same as the average number of charged stable hadrons. Also
323:    in this case we observe a striking difference between the values of higher moments for jets
324:    and hadrons. With decreasing
325:    average jet multiplicity (and increasing average number of particles in a single jet) even the second
326:    cumulant (not shown) becomes negative and all the pattern of the dependence of moments $H_q$ on their
327:    order $q$ does not
328:    resemble even roughly the pattern found for the single charged stable hadrons.
329:    \par
330:    An experimental investigation of the $H_q$ moments for jets has been
331:    performed by the L3 collaboration. The results are available in
332:    the D.J. Mangeol Ph.D. Thesis \cite{MA}. Although the jet
333:    definition used (Durham algorithm) was different than the default
334:    algorithm from PYTHIA used by us, the results are very similar
335:    to those presented in Fig. 4a: with increasing "jet size"
336:    the pattern of moments changes significantly.
337:    \par
338:    We regard this dependence as a suggestion that the
339:    extended local parton-hadron duality (ELPHD), in which one identifies the
340:    (charged) hadron multiplicity distribution with the parton distribution at small fixed
341:    virtuality, is not very reliable. We should note,
342:    however, that the opposite conclusion was drawn from the same
343:    data \cite{BFO}. The dependence of the $H_q$ moments on their
344:    order $q$ was investigated there for the parton jets in a MC
345:    model for varying jet resolution parameter $Q_c$. A  good
346:    description of L3 data was found for jets as well as for hadrons. This was regarded as a support for ELPHD.
347:    \par
348:    The agreement with the L3 data is very good indeed for $Q_c \geq 1GeV$. For small $Q_c$
349:    the agreement is reasonable, although not perfect. In
350:    particular, the ELPHD prediction for hadrons is practically
351:    identical to that for jets with $Q_c=100 MeV$, whereas in data
352:    there is a marked difference between the moments. This
353:    difference is similar to that seen for our PYTHIA events: a clear
354:    minimum for $q=5$ seen for hadrons (Fig. 2) moves already for smallest jets (Fig.4a) so that the values of
355:    $H_4$ and $H_5$ are almost equal. Thus there seems to be no
356:    clear "hadron limit" for parton MC, although the differences are not big.
357: 
358: 
359: \begin{figure}[h]
360: \centerline{\epsfig{figure=Fig6a.eps, height =7cm}}
361: \caption{\label{fig6} {\small \sl The $H_q$ moments  for PYTHIA
362: events generated with L3 parameters. Crosses denote the results
363: for charged hadrons, $x$-s are for all stable particles and stars
364:  for positive hadrons.}}
365: \end{figure}
366: \par
367: To verify further the possibility of describing hadron data by
368: ELPHD we have compared the moments calculated for charged hadrons
369: (pions, kaons, protons and antiprotons) with the moments
370: calculated for all stable particles (thus adding photons, neutral
371: kaons, neutrons and antineutrons). The results, shown in Fig.5,
372: suggest that a relatively deep minimum at $q=5$ is to a large
373: extent due to the charge conservation effect. Indeed, the results
374: for positive hadrons (shown also  in Fig.5) do not exhibit a
375: pattern seen for all charged hadrons. A deep minimum appears for
376: $q=4$, and the small oscillations with a very short period follow.
377: In this case it is rather obvious that the data will confirm our
378: findings, as the charge conservation is correctly built in PYTHIA
379: and the charged-neutral correlations are known to be well
380: described by this generator.
381:  \par
382: Obviously, there is no reason why the multiplicity distribution of
383: partons should be reflected in the distribution of all charged
384: hadrons more closely than in the distribution of all stable
385: particles. Thus the observed difference between the moments
386: calculated for these two choices seems to confirm our doubts
387: concerning the ELPHD.
388: \par
389: The choice of charged hadrons of both signs as a hadronic
390: counterpart of partons (mostly gluons) in ELPHD is in fact rather
391: surprising.  The measured cross section for odd multiplicities of
392: charged hadrons should be zero for an ideal detector, whereas the
393: distribution of gluons should be smooth without any discrimination
394: of odd multiplicity values. This suggests strongly that the
395: agreement of moments calculated for partons and charged hadrons is
396: to a large extent accidental and is due mostly to the similar
397: influence of cuts and the two-component mechanism (mentioned in
398: section 2) in both distributions. In our opinion, one should
399: rather try to find parameters, for which the parton MC would agree
400: with the distribution of positive (or negative) hadrons.
401: \\
402: 
403: \section{Conclusions and outlook}
404: 
405:    Using the PYTHIA
406:  generator we have investigated
407:  the multiplicity distributions for hadrons coming from the $Z$ decay. We have calculated the ratios of cumulants
408:   to factorial moments. The results for charged hadrons in full phase-space are compatible with the L3 data.
409:     We confirm the earlier findings of a minimum at $q=5,6$ when the moments $H_q$
410:    are considered as  functions of their order $q$. We confirm also that the truncation of the multiplicity
411:    distribution causes a significant enhancement of this minimum. However, increasing the statistics of generated
412:    events beyond $2M$ events does not change visibly the results. This excludes the "natural" truncation due
413:    to the final statistics as the main source of the minimum.
414:    \par
415:    We find that the moments calculated from the multiplicity distributions in the limited part of phase-space
416:    differ significantly from that for the full phase-space. In particular, the minimum in the dependence of $H_q$
417:    on $q$ shifts to higher values of $q$ for smaller range of momenta. However, introducing an universal truncation
418:    on the multplicity distributions for different ranges of momenta, we recover a stable position of the minimum.
419:    It will be interesting to see if this pattern of generated events will be confirmed in the real data.
420:    \par
421:    We have also investigated the moments for the jets, defined as clusters of particles close in momentum space.
422:    We find that even for very small jets, containing in average only slightly more than one hadron, the pattern of moments
423:    is different from that for hadrons. The same is true if we use the neutral particles to form jets, for
424:    which the average multiplicity is the same, as for charged hadrons. Since the distributions of such jets may be
425:    expected to correspond more closely to the distributions of partons, one may wonder if it is meaningful to
426:    compare the moments measured for the observed hadrons to the moments calculated for partons. We feel that our
427:    results cast doubts on the possibility of such an extension of "local parton-hadron duality" beyond the
428:    average quantities (moments of order one). Again,  more  investigation on the real data would be
429:    highly desirable.
430: 
431: 
432: 
433: 
434: \section{Acknowledgements}
435: \par
436:  We are grateful to A Bia{\l}as and A. Kota{\'n}ski for reading the manuscript and for helpful
437:  remarks. Fruitful critical remarks by W. Ochs, W. Metzger and by the Referees are gratefully
438:  acknowledged.
439: 
440: 
441: \begin{thebibliography}{99}
442: \bibitem{BP} A. Bia{\l}as and R. Peschanski, {\it Nucl. Phys.} {\bf B273}, 703 (1985).
443: \bibitem{Pavia}I.M. Dremin {\sl et al.}, {\it Phys. Lett.} {\bf B336}, 119 (1994).
444: \bibitem{D1} I.M. Dremin, {\it Phys. Lett.} {\bf B313}, 209 (1993).
445: \bibitem{D2} I.M. Dremin, {\it JETP Lett.} {\bf 58}, 945 (1993).
446: \bibitem{UGL} R. Ugoccioni, A. Giovannini and S. Lupia, {\it Phys. Lett.} {\bf D59}, 094020 (1999).
447: \bibitem{PYTHIA}T. Sj\"ostrand {\sl et al.}, {\it Comp. Phys. Comm.}
448: {\bf 135}, 238 (2001).
449: \bibitem {L3} P. Achard {\sl et al.}, L3 Collaboration, {\it Phys. Lett.} {\bf B577}, 109 (2003).
450: \bibitem{MA} D.J. Mangeol, Correlations in the Charged-Particle
451: Multiplicity Distribution, Ph.D. thesis, e-print hep-ph/0110029.
452: \bibitem {J} T. Sj\"ostrand, {\it Comp. Phys. Comm.} {\bf
453:  82}, 74 (1994).
454: \bibitem {H} G. Marchesini {\sl et al.}, {\it Comp. Phys. Comm.} {\bf
455:  67}, 465 (1992).
456: \bibitem{D3} I.M. Dremin, e-print hep-ph/0404092, proceedings of the Moriond meeting 2004.
457: \bibitem{GU} A similar model for hadron-hadron collisions was first suggested by
458: A. Giovannini and R. Ugoccioni, {\it Phys. Rev.} {\bf B336}, 119
459: (1994).
460: \bibitem{UA5} G.J. Alner {\sl et al.}, UA5 collaboration, {\it Phys. Lett.} {\bf B160}, 193 (1985).
461: \bibitem{VH} L. Van Hove, {\it Physica} {\bf 147A}, 19 (1987).
462: \bibitem{KF} K. Fia{\l}kowski, {\it Phys. Lett.} {\bf B173}, 197 (1986).
463: 
464: \bibitem{BFO} M.A. Buican, C. F\"orster and W. Ochs, {\it Eur.
465: Phys. J.} {\bf C31}, 57 (2003); W. Ochs, {\it Acta Phys. Pol.} {\bf
466: B35}, 429 (2004).
467: \end{thebibliography}
468: \end{document}
469: