Skip to main content

Full text of "Event weights for simulating Bose-Einstein correlations"

See other formats


manchester 
particle physics 



MAN/HEP/2001/2 
April 30, 2001 



Event weights for simulating Bose-Einstein correlations 

V. KartvelishviliQ 

Department of Physics and Astronomy, University of Manchester, UK 

and 
R. Kvatadze 

High Energy Physics Institute, Tbilisi State University, Georgia. 



Abstract 

An event weighting method for simulating Bose-Einstein effects in hadronic final states 
is presented. The weight for an event depends on the momentum distribution of identical 
bosons in the event. By using a theoretically motivated parametrisation allowing weights 
below as well as above unity, the necessity of a weight-rescaling procedure is eliminated. 
A single parameter is used to adjust the average event weight to unity. Once adjusted, the 
same value of the parameter gives average event weights that are essentially independent of 
energy, initial quark flavour, multiplicity and jet topology. The influence of Bose-Einstein 
correlations on various measurable quantities in W pair production is found to be small. 
In particular, none of the scenarios considered resulted in a W mass shift larger than 20 
MeV. 



Present address: Department of Physics, Lancaster University, UK 



1 Introduction 



There are two main reasons for the renewed interest in Bose-Einstein correlations (BEC) in 
particle physics. One is the quark-gluon plasma search in high energy heavy ion collisions, 
where BEC are providing important information about the space-time development of the final 
hadron formation in the dense matter. The other is connected to the precision measurement of 
the W mass in e + e~ annihilation, which can be used to constrain the allowed range of the Higgs 
boson mass in the Standard Model, or restrict the parameter space of any other "new physics" . 
However, it was suggested that in the fully hadronic channel, e + e~ — > W + W~ — > qqqq, BEC 
and colour reconnection effects could lead to significant uncertainties in the determination of W 
mass, up to (9(100 MeV) which can effectively render this channel useless and significantly 
reduce the precision on the W mass achievable at LEP2. 

Existing Monte Carlo simulation programs for hadronic final states are based on the factori- 
sation property of the QCD amplitudes ||: the cross sections are defined by the perturbative 
parton level amplitudes, while the hadronisation process of the final quark states is simulated 
in the framework of a particular model, assuming that it does not change the probability of the 
perturbative part. Various parameters of the hadronisation models have been finely tuned to 
reproduce many aspects of the data, with a notable exception of Bose-Einstein correlations, 
which cannot be simulated in this approach in principle. 

Several attempts have been made to implement Bose-Einstein effects a posteriori, so that 
the characteristic BEC are reproduced without breaking down the good description of other 
aspects of the data. At the moment, the most popular approach is the one developed in [g] and 
implemented in the PYTHIA Monte Carlo generator |4]]. This model is based on the assumption 
that the Bose-Einstein effects are local in phase space and are introduced as shifts in final-state 
boson momenta (hence the rather misleading name of "local reweighting"). The advantage of 
this method is that the QCD factorisation is explicitly preserved, and, thus, cross- sect ions of 
the processes are not affected. This procedure, however, does not conserve energy. In an early 
approach, energy conservation was restored by rescaling all final-state hadron momenta, while 
in later algorithms the energies were corrected locally. In this model, a W mass shift arises as a 
consequence of the final-state boson momentum re-distribution. Numerical predictions depend 
on the details of the rescaling procedure and vary from up to 100 MeV. The width of the W 
boson is also increased by up to 40 MeV. A major disadvantage of this method is that particle 
momentum re-distribution necessitates the re-tuning of the hadronisation parameters in order 
to reproduce the data, which makes the BEC effects rather difficult to extract. 

Another approach to the implementation of Bose-Einstein effects is the event weighting 
method (referred to as "global reweighting" in [[J). Here, BEC are introduced by assigning 
weights to the events according to the momentum distributions of identical hadrons in the 
final state. Within certain simplifying assumptions, this procedure can be justified using the 
formalism of Wigner functions 0. A number of such algorithms have been used recently to 
study the Bose-Einstein effects in the reaction e + e~ — > W + W _ (see also |2|,|TT| for the 

comparative analysis of different methods). Although the various methods differ significantly 
in the prescriptions for weight calculation, one thing in common to all of them is a relatively 
small BEC-induced shift in the W mass, less than about 20 MeV. Apart from being much 
more appealing theoretically, event weighting has also another advantage compared to "local 
reweighting": it can, in principle, be applied to the existing Monte Carlo samples. 

However, the event weighting methods also have some serious shortcomings. The distribu- 
tion of event weights is usually very broad (if not divergent); average weights, if taken literally, 
are usually much larger than unity, and in order to keep the cross sections intact a rather ar- 
bitrary procedure of weight rescaling is used. Average weights may also vary for various event 



1 



classes such as different initial quark ffavours, number of jets in the event, multiplicity etc. 
Thus, factorisation is not guaranteed, and is usually preserved by applying an ad hoc weight 
rescaling procedure for each class of events separately. 

These difficulties can be traced back to the fact that event weights were larger than unity 
by construction, implying that BEC enhance configurations where identical bosons are close 
to each other in the phase space. However, Bose symmetry can generate repellent forces too, 
which may become dominant in some areas of the phase space (e. g. identical pions may not 
exist in a P-wave, so the decay p — > 7r°7r° is forbidden), and give rise to event weights below 
unity. We use a theoretically motivated parametrisation which allows some event weights to fall 
below unity, and thus avoid the necessity of weight rescaling. In our method the average event 
weight in e + e~annihilation events is adjusted to unity using a single parameter, which appears 
to be independent of energy, initial quark flavour, number of jets or particle multiplicity in the 
event. Inclusive spectra of various hadrons also remain unaffected by the weighting procedure. 

The theoretical motivation and description of our method is presented in the following 
section. The choice of the model parameters is discussed in Section 3. In Section 4 the method 
is applied to e + e~annihilation into hadrons in the energy range from 30 to 200 GeV, with the 
region around the Z peak studied in detail. The influence of Bose-Einstein effects on the process 
of W pair production is analysed in Section 5. Some conclusion are drawn in Section 6. 



2 Motivation and algorithm description 

Let M be the matrix element describing the production of a hadronic final state which, among 
other, non-identical particles, contains n identical bosons. This amplitude consists of n! terms, 
each corresponding to a particular permutation P of the n identical particles in the final state: 

M = ^M P . (1) 

p 

When this process is simulated, the probabilistic treatment of the hadronisation stage means 
that the interference between different amplitudes is not included in the simulation: 

\M\ 2 MC = Y,\M P \ 2 ^\M\ 2 . (2) 



As shown in ||12|| , in order to take interference terms into account and thus restore the correct 
symmetry properties of the process, a weight wp has to be assigned to each event: 

\M\ 2 = Y,vopWp\\ (3) 



where 



Wp 



. ^ 2Re (M P M*,) 
\M P \ 2 + \M P ,\ 2 

v 2Re(M P M*,) 

^ \M P \ 2 + |Mp/l 2 ' ' ' 



p'^p 



The sum contains n\ terms and depends on the kinematical properties of the event. However, in 
order to be useful, the above formula needs to be implemented in a recipe for weight calculation. 



2 



Consider a simple parametrisation for the matrix element Mp, based on the Lund model of 
string hadronisation Jl2|. lj| : 

M P = exp[(m - b/2)A P ] . (5) 

Here Ap stands for the integral over the space-time area of the string fragmentation, while k and 
b are constants describing string tension and its breaking probability, respectively. Substituting 
(U) into one obtains: 

^ cosjnAApp,) 

where AApp/ = Ap — Api . 



In [T2|] it was argued that the dimensionless combination nAAppi between the two configura- 
tions labelled P and P' can be estimated as the scalar product of the differences in 4-momentum 
and in the space-time. The event weights were calculated at the stage of event generation by 
the JETSET program, taking the transverse motion of hadrons into account. The resulting 
weights were found to be well-behaved and described several manifestations of Bose-Einstein 
correlations in two-jet events, but the calculation process is rather labourous and is not easy 
to generalise to include more complex jet topologies. 

We propose a significantly simplified method of calculating event weights according to eq. 
(H), which, in principle, can be applied a posteriori to pre-generated event samples. We suggest 
that the combination nAAppi can be estimated as the product of an average interaction radius 
R and the "relative momentum" Q, which characterises the difference in kinematics between 
the two permutations P and P 1 . If the configuration P' is obtained from the configuration P 
by permuting n identical bosons with masses m and momenta Pi, ■ ■ ■ ,p n , then Q 2 is defined as 

Q 2 = (p 1 + ...+p n ) 2 -n 2 m 2 J (7) 

which coincides with the usual definitions Q\ 2 = ~(p\ — P2) 2 and Qf 23 = —(pi — P2) 2 — (pi — 
Pz) 2 ~~ G°2 — P3) 2 for n = 2 and n = 3, respectively. So, we propose the folowing replacement: 

nAAppi -» RQ, 

h -AA PP , - £i?Q, (8) 

where £ is a parameter whose value is to be determined phenomenologically. This leads to the 
weight calculated as 

cos(i?Q) , , 

For example, in the simplest case of two identical particles, the weight is 

cos{RQ 12 ) . . 

W2 = 1 + u CP n \ • ( 10 ) 

As noticed in [0], for £ values around 1 the weight (|T0D , shown as the solid line in Figure 1, is 
fairly close to the Gaussian-type function 

w G = 1 + exp{-R 2 Q 2 12 ) (11) 



(dashed line in Figure 1), used as the basic weight in a number of previous studies 
The important difference is that the new basic weight ([T^) goes slightly below unity for some 
intermediate values of Q, while the Gaussian weight (ITT]) always lies above unity. When the 



3 



'3 



1.75 



1.5 



1.25 



0.75 



0.5 



0.25 



I 1 1 1 I 1 1 1 I 1 1 1 I 1 1 1 I 




l+exp(-R 2 Q 2 ) 



l+cos(RQ)/cosh(^RQ) 



0.2 0.4 0.6 O.i 



1.2 1.4 1.6 1.8 2 

Q, GeV 



Figure 1: The basic weight (10) used in this paper, compared to the Gaussian basic weight (11), for 
R = 0.9 fm and f = 1.125. 



weight of an average event is built as a product of many terms of the type fllPp or (jTlf) , the 
latter may result in very large event weights, while the former tends to yield event weights close 
to unity. 

Consider, for example, the reaction e + e~— > W + W~, with both Ws decaying hadronically, 
and three 7r + mesons in the final state. Let the pions 1 and 2 come from the W + decay while 
the pion 3 comes from the W~. The weight for such an event has the following structure: 



» ! 3 



1 + (12) + (13) + (23) + 2 x (123) 



121 



where each term stands for one of the six possible permutations. It has the form cos(RQ^) / 
cosh(^i?Q( Q ,)), where the numbers in brackets, (a) = (12), (13), . . . show which pions have been 
permuted. The unity corresponds to no permutation, i.e. initial configuration P. The second 
term describes the only permutation if no inter- W correlations are allowed, while the following 
two terms stand for the two 2-particle inter- W permutations. The last term describes two 
3-particle permutations (corresponding to the so called "genuine" 3-boson correlations |l4jl). 

In a final state with 4 identical pions (1 and 2 from W + , 3 and 4 from W~) there are 4! 
terms: 



w A = 1 + (12) + (34) + (12) (34) + 

(13) + (14) + (23) + (24) + (13)(24) + (14)(23) 
2 x [(123) + (124) + (234) + (134)] + 6 x (1234) 



(13) 



The first line contains only permutations of pions originating from the same Ws, while the re- 
maining terms are either inter- W, or mixed. Note a new term type, e.g. (12) (34), corresponding 
to the simultaneous permutation within two pairs of pions. 



4 




Figure 2: Maximum rapidity difference in a cluster of identical charged pions vs. the variable Q 
characterising the cluster. 

In the hadronic final states produced in high energy collisions, number of identical mesons 
of each type can be rather big (e.g. 20 or more in hadronic WW events). Of course, only 
those mesons should be allowed to participate in BEC, which were produced directly during 
or shortly after the hadronisation phase. In the weight calculation we have included only 
those mesons whose parents have travelled less than d max in the centre of mass frame. This 
excludes mesons from the decays of long-lived parents, such as B and D mesons, r lepton 
etc., leaving on average about 40% of all mesons. The remaining mesons should be subject 
to BEC, but a straightforward application of the procedure described above would still lead 
to serious computational difficulties due to the big number of permutations. However, most 
of these permuted configurations would have a near- zero contribution to the weight, as their 
respective values of the "distance" in the momentum space, Q, tend to be high. The weight of 
each event is essentially determined by clusters of bosons with small values of Q. 

In order to eliminate unnecessary calculations, we have ordered all participating mesons of 
a particular type according to their rapidity y = (1/2) ha[(E + p z ) / (E —p z )] (calculated against 
the thrust axis of the event), and used the strong correlation existing between the value of Q 
characterising the cluster and the maximum rapidity difference Ay between the mesons in the 
cluster (see Figure 0). A new cluster was started if the rapidity difference between a meson 
and the first meson of the current cluster exceeded A?/ max . However, no cluster was allowed to 
contain more than n max mesons. The total weight for a system of mesons of a particular type 
was calculated as the product of the cluster weights. 

Separate weights corresponding to 9 types of mesons (tt + , 7r°, 7T~ -,K + ,K°,K ,K~ r], 7]') were 
calculated for each event, and the event as a whole was assigned a weight equal to the product 
of these 9 weights. 



5 



3 Choice of parameters 



In order to apply the algorithm described above to simulate BEC in Monte Carlo generated 
events, the values for the parameters R, d max , n max , Ay max and £ have to be fixed. R essentially 
describes the effective radius of BEC, which has been measured in various studies at LEP fl5HlT| 
to be within 0.5 and 1.0 fm. In the following, unless stated otherwise, we use R = 0.9 fm. 

The distance d max should be small enough to exclude from BEC the decay products of long- 
lived resonances. In our studies we used <i max = 10 fm. Note that in our approach the usual 
parameter A, which governs the "strength" of BEC, is missing altogether: identical bosons 
either fully participate in BE correlations (if they are produced early during hadronisation), or 
do not participate at all (if they come from long-lived parents). So the choice of d max , which 
determines the fraction of participating bosons, also determines the "effective strength" A e g, 
measured by experiments. 

The maximum cluster size n max should be chosen large enough to allow 2- and 3-particle 
correlations measured by experiments, but small enough to keep calculations manageable. It 
was found that the choice n max = 4 gives the best overall results, and we have used this value 
in our calculations. The maximum allowed rapidity difference in a cluster, Ay max , was chosen 
to be equal to 6/n, where n is the number of identical bosons of a particular type in an event. 
This means Ay max < 1 — 1.5, which is fairly harmless, as the clusters with larger Ay typically 
have rather large Q (see Figure and their contribution to the event weight is small. 

The parameter £, which is defined by the ratio of the two scales, k and b, characterising the 
hadronisation process, determines the value of the argument RQ for which the basic weight 
becomes smaller than unity. In typical e + e~ events the Q-distribution of identical meson pairs 
subject to BEC (see Figure §) is such that one can find a value of £ for which the average 
event weight equals unity. In practice, finding this value of £ may involve some trial- and-error 
and interpolation, and is only possible up to a certain precision, determined by Monte Carlo 
statistics and the variance of the weight distribution. However, once found, it appears to be 
fairly stable under variation of other parameters. 

4 Influence of event weighting on Z properties 

As long as the average event weight for Z hadronic decays is equal to unity, the cross section 
of this process is not changed by event weighting. However, other measurable properties of the 
Z could be affected. 

Since the parameters of the Monte Carlo hadronisation models, which do not explicitly 
include BEC effects, have been carefully tuned to reproduce various measured distributions, 
uncritical application of event weights may lead to large inconsistencies with measured partonic 
branching ratios, different jet topologies, final hadron multiplicities etc. |7j. Also, the average 
event weight adjusted to unity at one energy may deviate from unity at other energies, thus 
potentially affecting such parameters as the mass and the width of the Z boson. 

In order to study how serious these effects are and to judge what consequences they have 
for the analysis of the WW events, we have compared various weighted and unweighted distri- 
butions. We have used PYTHIA 6.125 Monte Carlo [|J to generate a sample of 10 5 hadronic 
events at yfs = M% = 91.2 GeV, with £ adjusted so that the average event weight is equal to 
1 (within statistical errors). The distribution of event weights is shown in Figure [|. It peaks 
close to its average value and is fairly narrow, with an rms of about 0.5. 

Table 1 presents average event weights and respective rms values for various initial quark 
flavours and different jet topologies (as determined by the PYCLUS jet finding algorithm with 



6 




Figure 3: Q-distribution of identical charged pions subject to BEC in Z decays, generated with 
PYTHIA Monte Carlo (see Section 4). 



WW 4 : 



entries 100000 
mean 1 .003 
rms 0.531 



10 



"A 



1 



i mi 



8 10 

Weight 



Figure 4: Distribution of event weights in Z decays, for = 0.9 fm and £ = 1.125. 



7 




30 40 50 60 

Charged multiplicity 



Figure 5: The charged multiplicity distribution for events weighted according to our recipe (data 
points) compared to the unweighted distribution (solid line). The error bars shown correspond to the 
errors specific to the event weighting process, and do not include the statistical uncertainties common 
to both distributions. 

default parameters [|J). As seen from the table, the average weights are essentially independent 
of the initial flavour and number of jets in the event. This means that the implementation of 
event weights in the form (§) does not cause any noticeable changes in either the partonic 
branching ratios or the jet activity in Z decays. This is not trivial, as the patterns of the heavy 
and light quark fragmentation are rather different, and the multiplicity of low- momentum 
particles is strongly correlated with the number of jets in the event. 

Figure 5 shows a very good agreement between the charged particle multiplicity distribu- 
tions at the Z peak with (data points with errors) and without (solid line) event weighting. 
Only the errors specific to the weighting process are shown on the plot. The means of the 
two distributions differ by An c h = 0.07 ± 0.003, well within the combined experimental error 
obtained by four LEP experiments, ±0.11 ||18|| . Similarly, a very good agreement between the 
weighted (data points) and unweighted (solid line) momentum distributions of 7r + mesons at 
the Z peak is shown in Figure 6. The same is true for other hadron types. 

In order to study the dependence of the average weight upon initial energy, four more 
samples of events e + e~— >■ 7*/Z* — > hadrons of the same size were generated at 30, 131, 161 and 
200 GeV, using the same value of £. Average event weights for these energies are also presented 
in Table 1. The weights are fairly independent on the initial energy of the collision, although 
a decrease of about 1% is seen at the highest energy, 200 GeV. This could be connected to the 
fact that at higher energies a slightly larger percentage of hadronic resonances (such as p and 
K*) escape the 10 fm limit, which means that their decay products no longer contribute to the 
weight. 



S 



> I 

4) 




Momentum, GeV 



Figure 6: The momentum distribution of tt + mesons, with events weighted according to our recipe 
(data points) compared to the unweighted distribution (solid line). As in Fig. 5, the error bars shown 
correspond to the errors specific to the event weighting process, and do not include the statistical 
uncertainties common to both distributions. 



qq 


dd 


1111 


ss 


cc 


bb 


(w) 


1.016 ±0.004 


1.010 ±0.005 


1.000 ±0.004 


1.002 ±0.004 


0.991 ±0.003 


rms 


0.573 


0.595 


0.520 


0.482 


0.463 




2 


3 


4 


5 


6 


(w) 


1.009 ±0.003 


0.997 ±0.002 


1.000 ±0.003 


1.015 ±0.007 


1.068 ±0.023 


rms 


0.482 


0.429 


0.559 


0.713 


1.053 


E 


30 GcV 


91.2 GeV 


131 GeV 


161 GeV 


200 GeV 


(w) 


0.995 ±0.001 


1.003 ±0.002 


0.994 ±0.002 


0.991 ±0.002 


0.988 ±0.002 


rms 


0.343 


0.531 


0.526 


0.503 


0.494 



Table 1: Average weights, (w), and rms values for different quark flavours, jet topologies and centre 
of mass energies. 



9 




Figure 7: The correlation function C(Q), calculated as the ratio of weighted and unweighted Q 
distributions, for same-sign charged pions (a) and opposite-sign charged pions (b). The line in (a) 
shows the fit result using the parametrisation (|14|), with parameter values given in Table 2 (event 
weight). 



In order to study possible effects of the event weighting upon the mass and the width 
of the Z resonance, four more samples of 10 5 events were generated at ^/s = M% ± 2 GeV 
and ± 4 GeV, with the same value of the parameter £. A Breit-Wigner fit to these five 
points with and without event weighting yielded no significant shifts in the Z mass and width: 
AM^ = 0.4 ± 0.5 MeV and Ar^? = 1.7 ± 1.7 MeV. The experimental uncertainties on these 
extremely precisely measured quantities are 2.2 MeV and 2.6 MeV, respectively [fTj|. 



The correlation functions C(Q) were constructed as ratios of the Q distributions of particle 
pairs for weighted and unweighted events. The ratios for same- and opposite-sign charged pion 
pairs are shown in Figure 7. A clear enhancement is seen for the same-sign pion pairs at small 
values of Q, while the distribution for the opposite-sign pairs is flat and close to 1. Also shown 
is the result of the fit to the same-sign pair correlation function of the form 

C(Q) = N(l+ f3Q)(l + \ cS eM-Q 2 R 2 cS )), (14) 

which is often used to parametrise the experimentally observed correlation function in Z decays. 
The values obtained for the parameters (for a fit range of 0-1 GeV in Q) are presented in Table 
2, together with the results of a similar fit to the input pair weight (the latter is merely a fit 
of the form flUD to the basic weight described by (0)). The statistical errors in the fitted Q 
distributions were increased by 40%, in order to account for the bin-to-bin correlations, arising 
from the fact that each boson in an event can contribute to several combinations. 

Fit results for input (pair weight) and output (event weight) parameters are in a reasonable 
agreement with each other, the only noticeable difference being that the observed enhancement 



10 





Event weight Pair weight 


A e ff 


0.181 ±0.012 0.186 ±0.011 


N 


0.978 ±0.008 0.951 ±0.009 


(3 (GeV- 1 ) 


0.046 ±0.013 0.053 ±0.013 


Res (fm) 


0.902 ±0.056 0.758 ±0.038 



Table 2: The fitted values of the output (event weight) and input (pair weight) correlation function 
parameters at the Z peak. 

at small Q is slightly narrower than in the input distribution, leading to a larger effective value 

Of Reft- 

So, the event weighting method described above reproduces the BE correlation functions 
without introducing any significant modification of the properties of the Z boson, and no no- 
ticeable energy dependence of the average weight, using the same value for the single adjustable 
parameter £ for all types of events across the whole energy range considered. 

We have repeated most of our studies for R = 0.6 fm with very similar results. The average 
event weight was equal to unity (within statistical errors) for £ = 0.98, with no significant 
dependence on energy, initial quark flavour or number of jets. The fitted values for the pa- 
rameters of the correlation function (|14|) were: A e fr = 0.29 ± 0.03 and R e s = 0.61 ± 0.04. The 
values of these parameters, obtained by LEP experiments studying Z hadronic decays, vary in 
the intervals 0.2-0.6 for A e fr and 0.5-1.0 fm for R e ft, depending on the analysis [|T~5| — P~7f| . More 
meaningful and detailed comparisons of our results with real data will only be possible when the 
real-world analyses are applied to the generator-level Monte Carlo samples, weighted according 
to our recipes. 

5 BEC in W pair production 

We have studied possible influence of inter-W BE effects on the apparent mass and other 
measured properties of the W boson. The PYTHIA 6.125 event generator |4[] was again used to 
simulate the process e + e~ — > W + W~ — > qqqq, and the weighting method described above was 
applied to implement Bose-Einstein effects. For obvious reasons, final-state meson distributions 
in this process are quite different from those in Z decays considered above, and one should expect 
to obtain the average weight equal to unity for a different value of the parameter £. 

In weight calculations we have used the same source radius as for Z studies, R = 0.9 fm. 
Three different weighting schemes were considered: 

- Only identical bosons originating from different Ws were included in Bose-Einstein cor- 
relations (labelled as DW scheme). 

- Only bosons from the same W are subject to BEC (labelled as SW scheme). 

- All identical bosons from both the same and different Ws are allowed to participate in 
Bose-Einstein effects (labelled as SW±DW). 

Two samples of 25000 events were generated at the energies 161 and 200 GeV. We have 
tuned £ to obtain the average value of the event weight approximately equal to unity at the 
energy 161 GeV in each of the three schemes separately, and then used the same values of £ at 



11 





£ (w) rms AM W Mw An c/l 

(MeV) (MeV) 


161 GeV 




DW 


1.008 0.989 ±0.003 0.491 13 ± 5 -11 ± 18 0.12 ±0.005 


SW 


1.048 1.007 ±0.004 0.679 -2 ± 6 -45 ± 26 0.19 ±0.005 


SW+DW 


1.094 1.002 ±0.006 0.930 11 ± 9 -63 ± 33 0.27 ±0.005 


(SW+DW)-(SW) 


13 ±6 -18 ±20 0.08 ±0.005 


200 GeV 




DW 


1.008 0.983 ±0.003 0.430 1 ±4 3 ±11 0.08 ± 0.005 


SW 


1.048 0.986 ±0.004 0.623 -5 ± 5 -34 ±15 0.11 ±0.005 


SW+DW 


1.094 0.986 ±0.005 0.814 -7 ± 6 -37 ± 20 0.17 ±0.005 


(SW+DW)-(SW) 


-2 ±4 -3 ± 13 0.06 ±0.005 



Table 3: Parameter £, average weights and their rms for W + W hadronic decay events at 161 and 
200 GeV, together with shifts in W mass, width and charged multiplicity caused by event weighting. 

200 GeV. The application of the parameter £, fixed at 161 GeV, to the higher energy, 200 GeV, 
reduces average event weights by about 1.5%, apparently because at higher energies, due to the 
boost of the W decay frame, more resonances travel beyond the allowed limit (i max . A similar 
reduction of the average weight was observed when we applied our algorithm to semileptonic 
WW events, e + e~ — > W + W~ — > qqlv at 200 GeV, using the value of £ adjusted for Z decays. 

Table 3 presents the values of £, average event weights and their rms, together with the 
shifts in the mass and the width of the W and the average charged particle multiplicity in 
hadronic W decays, compared to the case with no event weighting. The mass and the width 
were determined by fitting a Breit-Wigner parametrisation to the invariant mass distribution 
of the W decay products. 

The fact that the shifts in W parameters for the SW scheme differ from zero shows that 
our implementation of Bose-Einstein effects is not perfect, as we do not know any valid reason 
why the inclusion of BEC only for bosons originating from same W should change any of them 
(see a similar discussion on Z properties in Section 4). However, we expect that the differences 
in these quantities between SW±DW and SW represent a valid estimate of the effects of inter- 
W BE correlations, alongside with the predictions of the DW scenario. Note that these two 
sets of shifts (the first and the last rows for each energy in Table 3) are consistently close to 
each other. Thus, averaging over these two scenarios, our simulations show that the shifts in 
the W mass, width and average charged multiplicity in W decays respectively are 13 ± 5 MeV, 
-15±18 MeV and 0.10±0.005 at 161 GeV, reducing correspondingly to 0±4 MeV, 0±11 MeV 
and 0.07 ± 0.005 at 200 GeV. Typical experimental errors on these quantities are at present 
significantly larger: 56 MeV, 50 MeV and 0.4 JT|]. 

Table 4 presents values of the average event weight and the rms of the event weight dis- 
tribution, for varous jet topologies in the DW scenario. The average weights are essentially 
independent of the number of jets. Hence, the implementation of event weights does not 
change significantly the jet multiplicity distribution in the W + W~ production process. The 
rms of the weight distribution increases slightly with the number of jets, as in Z decays. This 
is connected to the increase of particle multiplicity with increasing number of jets. The event 
weight dependence on the flavours of the initial quarks in W decays is very weak and does not 
change partonic branching ratios. We have also checked that the introduction of event weights 



12 





4 


5 


6 


7 


8 


161 GeV 




(w) 


0.964 ±0.008 


0.969 ±0.005 


0.982 ±0.005 


1.001 ±0.007 


1.030 ±0.012 


rms 


0.352 


0.410 


0.443 


0.558 


0.641 


200 GeV 




(w) 


0.968 ±0.009 


0.970 ±0.005 


0.978 ±0.005 


0.984 ±0.005 


1.000 ±0.008 


rms 


0.319 


0.344 


0.393 


0.430 


0.483 



Table 4: Average event weights and the rms for different number of jets at 161 and 200 GeV, in the 
DW scheme. 





DW SW SW±DW 


161 GeV 






0.073 ±.012 0.117 ±.013 0.201 ± .015 


R c s (fm) 


1.026 ±.126 1.006 ±.086 1.053 ±.057 


200 GeV 




A c fT 


0.057 ±.010 0.116 ±.012 0.182 ±.014 


R cS (fm) 


0.902 ±.107 0.994 ±.081 0.993 ± .055 



Table 5: The fitted values of the correlation function parameters A e g and R e s for W pair production 
in (DW), (SW) and (SW+DW) schemes at yfs = 161 and 200 GeV. 



does not alter multiplicity distributions and inclusive momentum spectra for various hadrons in 
W + W~ production. Similar results were obtained also for SW±DW and SW event weighting 
schemes. 

On the experimental side, it is still not clear at the moment whether the Bose-Einstein 
correlations between mesons originating from different Ws exist or not 



IS -El. Therefore, 



we 



have studied the correlation functions for identical bosons in all three scenarios (DW, SW and 
SW±DW). The correlation functions were constructed as ratios of particle pair four-momentum 
difference distributions for weighted and unweighted W + W~ hadronic decay events. They are 
plotted in Figure 8 for the charged pion case. In all three scanarios, the figure shows a clear 
enhancement at small Q, characteristic of Bose-Einstein correlations. 

Values for the effective strength of correlations A e ff and the observed effective source radius 



R c s, obtained by fitting the form (14]) to these distributions, are given in Table 5. The strength 
of the correlations in the DW case is significantly smaller compared to the full SW±DW sce- 
nario, because a large number of identical bosons come from the same W, and the pairs of 
identical bosons from different Ws are on average farther away from each other, in the Q space, 
than those from the same Ws. Hence, the experimental observation of Bose-Einstein corre- 
lations for charged pion pairs originating from different Ws at LEP2 would indeed be very 
difficult, requiring high statistics, and a careful control of systematics coming from the choice 
of the reference sample. As in the case of hadronic Z decays, the effective input (pair weight) 
and output (event weight) values for parameters \ e s and R e s are reasonably close to each other. 



13 



§1.3 
U 

1.2 






(a) 


i i | i | i | i | i i | i _ 

DW 1 


1.1 




L 






1 









■ 







0.2 0.4 


0.6 0.8 1 1.2 1.4 1.6 1.8 J 

Q, GeV 


§1.3 
O 

1.2 




i 


1 i i i | i 

(b) 


1 i i i | i i i | i i i | i i i | i i i | i i i | i i i _ 

sw ] 


1.1 










1 








, 1 , , , ] , , , 1 , , , 1 , , , 1 , , , 1 , , , 1 , , , ■ 



Q, GeV 





L< c > 




1 1 1 i 1 1 1 1 1 

SW+DW^ 




i , , , i , , , 


,1,1,1,,!,- 



r i 

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 



Q, GeV 



Figure 8: The correlation functions C(Q), for the three different scenarios described in the text: a) 
only bosons from different Ws participate in BEC (DW); b) only bosons from the same W participate 
in BEC (SW); c) all identical bosons participate in BEC (SW+DW). The lines show the fit results 
using the parametrisation (|14|), with parameter values given in Table 5. 



14 



6 Conclusions 



We have developed a method for modelling Bose-Einstein correlations using event weighting, 
with a theoretically motivated parametrisation for the basic weight, based on the string frag- 
mentation picture. Our approach differs from some other "global weighting" schemes, as the 
event weights in our case are distributed both above and below unity with a rather narrow 
distribution, and the average event weight is easily adjusted to unity using a single parameter 
£. This eliminates the additional rescaling of event weights, necessary in models where all event 
weights are larger than one. 

By weighting Monte Carlo events in accordance with our prescriptions, the experimentally 
observed characteristic Bose-Einstein enhancement at small relative momenta is reproduced. 
Good agreement was found between the input and output values of the parameters A e ff and 
R c g in the de-facto standard Gaussian parametrisation. By fine-tuning the values of our model 
parameters rf max and R (and subsequent re-adjustment of £ in order to keep the average weight 
equal to unity) one can bring \ cS and -R e ff closer to the values obtained by particular experi- 
ments. However, any detailed comparison of our results with the real data will only be possible 
when the real-world analyses are applied to the Monte Carlo samples, weighted according to 
our recipes. 

The main weakness of all BEC implementations via event reweighting is the possible viola- 
tion of factorization between the hard perturbative part of the process and the non-perturbative 
hadronisation stage, essential to all Monte Carlo generators. Our model is practically free from 
these difficulties. We have made extensive checks by comparing our predictions with unweighted 
distributions, which have been tuned and tested to reproduce very precise experimental data 
on Z decays. We have found no significant shifts in the mass, width and partonic branching 
fractions of the Z boson due to event reweighting, within the estimated errors which are well 
below the level of existing experimental uncertainties. The same is true for charged multiplicity 
distributions and inclusive spectra of various final-state hadrons. 

In the process e + e~ — * W + W~ — > qqqq, the introduction of event weights leads to small 
shifts in the values of W mass, width and charged multiplicity, less than 15 MeV, 20 MeV 
and 0.10, respectively at y/s = 161 GeV, and even smaller at 200 GeV. These values are well 
below currently existing experimental errors. Hence, at the generator level the Bose-Einstein 
correlations as implemented in our model do not introduce large additional uncertainties in the 
determination of W characteristics in the fully hadronic channel. However, as in the case of Z 
decays, in order to assess the effects of BEC on the experimentally measured W parameters, 
the detector simulation and the actual fitting procedures used by the LEP experiments have 
to be applied to the weighted event samples. This is not too difficult because there is no need 
to generate special Monte Carlo samples, as the weighting can be applied a posteriori to the 
existing Monte Carlo events. 

We have also studied the correlation functions for charged pions in fully hadronic WW 
events in three separate scenarios, depending on which pairs of identical bosons were allowed 
to participate in BEC: only from different Ws (DW), only from the same W (SW) and all pairs 
(SW+DW). The characteristic enhancement at Q < 0.2 GeV was seen in all three scenarios. 
However, the effective value of the parameter A e fr is the smallest in the (DW) scenario, suggest- 
ing that the direct observation of BEC for pions originating from different Ws with the available 
statistics from LEP2 would be difficult, requiring a very careful control of systematics. 

Helpful discussions with N. Kjaer, G.Lafferty and L. Lonnblad are thankfully acknowledged. 
One of us (R.K.) would like to thank the Particle Physics Group at the University of Manchester, 
and IPPP (Durham) for their kind hospitality. 

This work was supported in part by the INTAS grant 99-0558. 



15 



References 

[1] J. Ellis and K. Geiger, Phys. Rev. D52 (1995) 1500; Phys. Rev. D54 (1996) 1755; Phys. 
Lett. B404 (1997) 230. 



[2] 
[3] 

[4] 
[5] 
[6] 
[7] 
[8 

[9 
[10 

[11 
[12 

[13 

[14 

[15 

[16 

[17 

[18 

[19 

[20 

[21 

[22 



L. Ldnnblad and T. Sjostrand, Phys. Lett. B351 (1995) 293; Eur. Phys. J. C2 (1998) 165. 

See e.g. G. Sterman, An Introduction to Quantum Field Theory, Cambridge University 
Press, Cambridge, 1993. 

T. Sjostrand, Computer Physics Comm. 82 (1994) 74. 

A. Bialas and A. Krzywicki, Phys. Lett. B354 (1995) 134. 

S. Jadach and K. Zalewski, Acta Phys. Polon. B28 (1997) 1363. 

V. Kartvelishvili, R. Kvatadze and R. Moller, Phys. Lett. B408 (1997) 331. 

K. Fialkowski and R. Wit, Acta Phys. Polon. B28 (1997) 2039. 

S. Todorova-Nova and J. Rames, [hep-ph /971 0280| . 

J. Hakkinen and M. Ringer, Eur. Phys. J. C5 (1998) 275. 

K. Fialkowski and R. Wit, |hep-ph/9805"223 . 



B. Andersson and M. Ringner, Nucl. Phys. B513 (1998) 627. 

B. Andersson and W. Hofmann, Phys. Lett. B169 (1986) 364. 

See e.g. DELPHI Coll., Phys. Lett. B355 (1995) 415 and references therein. 

OPAL Coll., Phys. Lett. B267 (1991) 143. 

ALEPH Coll., Z. Phys. C54 (1992) 75. 

DELPHI Coll., Phys. Lett. B286 (1992) 201; Z. Phys. C63 (1994) 17. 
Review of Particle Physics, D. E. Groom et al., Eur. Phys. J. C15 (2000) 1. 
ALEPH Coll., Phys. Lett. B478 (2000) 50; ALEPH 2000-024 CONF. 
OPAL Coll., Eur. Phys. J. C8 (1999) 559. 
L3 Coll., Phys. Lett. B493 (2000) 233. 

DELPHI Coll., CERN-OPEN-99-456; DELPHI 2000-115 CONF 414. 



16