Fragmentation and Hadronization

Bryan R. Webber
Theory Division, CERN, 1211 Geneva 23, Switzerland, and
Cavendish Laboratory, University of Cambridge, Cambridge CB3 0HE, U.K.1

1 Introduction
Hadronic jets are amongst the most striking phenomena in high-energy physics,
and their importance is sure to persist as searching for new physics at hadron
colliders becomes the main activity in our field. Signatures involving jets almost
always have the largest cross sections, but are the most difficult to interpret and
to distinguish from background. Insight into the properties of jets is therefore
doubly valuable: both as a test of our understanding of strong interaction dy-
namics and as a tool for extracting new physics signals in multi-jet channels.
In the present talk, I shall concentrate on jet fragmentation and hadroniza-
tion, the topic of jet production having been covered admirably elsewhere [1, 2].
The terms fragmentation and hadronization are often used interchangeably, but
I shall interpret the former strictly as referring to inclusive hadron spectra, for
which factorization ‘theorems’2 are available. These allow predictions to be made
without any detailed assumptions concerning hadron formation. A brief review
of the relevant theory is given in Section 2.
Hadronization, on the other hand, will be taken here to refer specifically to
the mechanism by which quarks and gluons produced in hard processes form
the hadrons that are observed in the final state. This is an intrinsically non-
perturbative process, for which we only have models at present. The main models
are reviewed in Section 3. In Section 4 their predictions, together with other
less model-dependent expectations, are compared with the latest data on single-
particle yields and spectra.
One of the most important objectives of jet studies is to understand the dif-
ferences between jets initiated by different types of partons, especially quarks
versus gluons, which could also be valuable in new physics searches. The wealth
of recent data on quark-gluon jet differences is discussed in Section 5. Another
1
Permanent address.
2
I put this term in inverted commas because proofs of factorization in fragmentation do not
really extend beyond perturbation theory [3].

594
Bryan R. Webber Fragmentation and Hadronization

goal is to compare jets produced in different processes such as e+ e− annihila-

tion and deep inelastic scattering (DIS). This is being done by the H1 and ZEUS
collaborations at HERA; some of their results are discussed in Section 6.
A good understanding of heavy quark jets is especially important because
these are expected to be copiously produced in new processes such as Higgs
boson decay. Section 7 discusses some recent results on heavy quark fragmen-
tation.
Finally, Sections 8 and 9 deal with correlation effects, first those of Bose-
Einstein origin and then those expected in fully hadronic final states from e+ e− →
W + W − . A summary of the main points is given in Section 10.
Regrettably, I have no space to discuss many other relevant and interest-
ing topics, such as: fragmentation function parametrizations [4]; photon frag-
mentation function; polarization in hadronization; event shapes and power cor-
rections [5]; fluctuations and intermittency; dynamical two-particle correlations;
heavy quark production in jets (g → cc, bb); transverse energy flow in deep inelas-
tic scattering; underlying event in DIS and hadron-hadron collisions; identified
particle production in DIS; tests of QCD coherence; jet profiles and substruc-
ture [1]. Apologies also to all those whose work I have omitted or mentioned
only superficially.
In citing the latest experimental data (mostly still preliminary) I have relied
on the reference system of the EPS-HEP99 conference in Tampere, Finland [6],
because few experimental papers were explicitly submitted to LP99. In the case of
the large collaborations, papers submitted to EPS-HEP99 and/or LP99 can usually
be found easily via the collaboration web pages [7]–[14].

2 Jet fragmentation—theory
Let us start by recalling the basic factorization structure of the single-particle
inclusive distribution, e.g. in e+ e− → hX (Fig. 1):

 1 dz
h
F (x, s) = Ci (z, αS (s))Dih (x/z, s)
x z
i

s = q2 , x = 2ph · q/q2 = 2Eh /Ecm

where Ci are the coefficient functions for this particular process (including all
selection cuts etc.) and Dih is the universal fragmentation function for parton
i → hadron h.
The fragmentation functions are not perturbatively calculable, but their s-

595
Bryan R. Webber Fragmentation and Hadronization

qµ i h pµ
Ci Di h

Figure 1: Factorization structure of e+ e− → hX.

dependence (scaling violation) is given by the DGLAP equation:

1
∂ dz
s Dih (x, s) = Pji (z, αS (s))Djh (x/z, s) .
∂s j
x z

Thus, they can be parameterized at some fixed scale s0 and then predicted at
other energies [4].
In certain kinematic regions, higher-order corrections are enhanced by large
logarithms, which need to be resummed. At small x, log x enhanced terms can
be resummed by changing the DGLAP equation to

1
∂ dz
s Dih (x, s) = Pji (z, αS (s))Djh (x/z, z2 s) .
∂s j
x z

This is commonly known as the modified leading-logarithmic approximation (MLLA)

[15, 16, 17]. The effect of resummation is to generate a characteristic hump-
backed shape in the variable ξ = ln(1/x), with a peak at ξp ∼ 14 ln s.
Large logarithms of ratios of invariants may also appear inside the coeffi-
cient functions Ci , for example in three-jet events when the angles between jets
become small. In some cases, these can be absorbed into a change of scale in the
fragmentation functions. Examples will be encountered in Sections 4 and 5.
Although universal, fragmentation functions are factorization scheme depen-
dent. The splitting functions Pji are also scheme dependent in higher orders. To
specify the scheme requires calculation of the coefficient functions to (at least)
next-to-leading order. This has only been done in a few cases. Thus, there is need
for theoretical work to make full use of the data on fragmentation functions.

3 Hadronization Models

3.1 General ideas

Local parton-hadron duality [18]. Hadronization is long-distance process, involv-
ing only small momentum transfers. Hence, the flows of energy-momentum and

596
Bryan R. Webber Fragmentation and Hadronization

flavor quantum numbers at hadron level should follow those at parton level. Re-
sults on inclusive spectra and multiplicities support this hypothesis.
Universal low-scale αS [19, 20, 21]. Perturbation theory works well down to low
scales, Q ∼ 1 GeV. Assume therefore that αS (Q2 ) can be defined non-perturbatively
for all Q, and use it in the evaluation of Feynman graphs. This approach gives a
good description of heavy quark spectra and event shapes.

3.2 Specific models

The above general ideas do not try to describe the mechanism of hadron forma-
tion. For this we must so far resort to models. The main current models are
cluster and string hadronization (Fig. 2). We describe briefly the versions used in
the HERWIG and JETSET event generators, respectively.

1. Cluster model [22]-[26]. The model starts by splitting gluons non-perturbatively,

g → qq, after the parton shower. Color-singlet qq combinations have lower
masses and a universal spectrum due to the preconfinement [27, 28] prop-
erty of the shower (Fig. 3 [29]). These color-singlet combinations are as-
sumed to form clusters, which mostly undergo simple isotropic decay into
pairs of hadrons, chosen according to the density of states with appropri-
ate quantum numbers [23]. This model has few parameters and a natural
mechanism for generating transverse momenta and suppressing heavy par-
ticle production in hadronization. However, it has problems in dealing with
the decay of very massive clusters, and in adequately suppressing baryon
and heavy quark production.

2. String model [30]-[34]. This model is based on the dynamics of a relativistic

string, representing the color flux stretched between the initial qq. The
string produces a linear confinement potential and an area law for matrix
elements:
|M(qq → h1 · · · hn )|2 ∝ e−bA
where A is the space-time area swept out (Fig. 4). The string breaks up into
hadrons via qq pair production in its intense color field. Gluons produced
in the parton shower give rise to ‘kinks’ on the string. The model has extra
parameters for the transverse momentum distribution and heavy particle
suppression. It has some problems describing baryon production, but less
than the cluster model.

3. UCLA model [35]-[37] is a variant of the JETSET string model which takes the
above area law for matrix elements more seriously, using it to determine
the relative rates of production of different hadron species. This results in

597
Bryan R. Webber Fragmentation and Hadronization

1010
00
11
10
11
00 00
11
11111
00000 00000
11111 00 111
11 000
00000
11111 00000
11111 000
111
00000
11111 10 00000
11111 000
111
00000
11111 10
000
111 00000
11111
00000
11111 00000
11111 11
00
00000
11111 00000
11111
00000
11111 00000
11111
0000001
11111 00000
1111101 00
11
1111
0000 00000
11111 11
00
00
11 00000
11111 111
000 00
11
111
000
0000
1111 0011111
1100000
0011111
1100000 0000
11 1111
0000 101011111
00000
00000
11111
0000
1111 000
111
000
111
11 0000
1111 000
111
0000
1111
0000
1111 000
111 0000
1111 000
111 111
000
10111 0000
1111 000
111
0000
1111
11
00
0000
1111 10000
000
111 00
11
00
11 0000
1111
11
00 01111
000 1000
0000
1111
0000
1111
000
111
000
111
00
11
00
11 0000
1111
0000
1111 000
111
000
111 111011
00 00
11
01
0000
1111 000
111
000
111 0000
1111 000
111
1010
000
111 111
000
0000
1111 01
000
111 0000
1111 000
111
000
111
000
111 0000
1111 000
111 00
11
00010
111 000
111
11
00 000
111 000
111 00
11
0001011
111 00 000
111
00
11
000
111 0000111
111 000
111
00
11 000
111 00 11
00
000
111
00
11 00
11 000
111
00
11 00
11
00
11 00
11 00
11
00 11
11 00
00
11 00
111010
00
11 00 000
11 111
00
11
Figure 2: Cluster and string hadronization models.
Frequency

1.2 Singlet 50 GeV

Random 500 GeV

1 5000 GeV

50000 GeV
0.8

0.6

0.4

0.2

0
0 1 2 3 4
2
log10 Mass (GeV/c )

Figure 3: Cluster model: mass distribution of qq pairs.

heavy particle suppression without extra parameters, the mass-squared of

a hadron being proportional to its space-time area. At present, the model
still uses extra parameters for pT spectra, and again has some problems
describing baryon production.

598
Bryan R. Webber Fragmentation and Hadronization

hn
h1
h2 ..... hn-1

-q
t q A

Figure 4: String model: space-time picture.

4 Single-particle yields and spectra

Tables 1 and 2 compare predictions of the above models3 with data on Z0 de-
cay from LEP and SLC.4 Of course, the models have tunable parameters, but the
overall agreement is encouraging. As stated earlier, the main problems are in the
baryon sector, especially for HERWIG.

It is remarkable that most measured yields (except for the 0− mesons, which
have special status as Goldstone bosons) lie on the family of curves

n = a(2J + 1)e−M/T

where M is the mass and T  100 MeV (Fig. 5 [39]). This suggests that mass,
rather than quantum numbers, is the primary factor in determining production
3
rates. Note that, surprisingly, the orbitally-excited J = 2 baryon Λ(1520) (not yet
included in models) is produced almost as much as the unexcited J = 32 baryon
Σ(1385) [40, 41].
At other energies, model predictions for identified particle yields are in broad
agreement with e+ e− data (Fig. 6 [42]), but statistics are of course poorer. Charged
particle spectra at low x agree well with the resummed (MLLA) predictions [15,
16, 17] over a wide energy range, as illustrated in Fig. 7 [43].
In pp → dijets [44] the relevant scale is taken to be Q = MJJ sin θ where MJJ
is the dijet mass and θ is the jet cone angle (Fig. 8). Results are then in striking
√
agreement with theory and with data from e+ e− annihilation at Q = s (Fig. 9).
3
Using the recent ALEPH HERWIG tuning with strangeness suppression 0.8 [38].
4
Updated from Ref. [29].

599
Bryan R. Webber Fragmentation and Hadronization

Particle Multiplicity HERWIG JETSET UCLA Expts

5.9 7.4 7.4
Charged 20.96(18) 20.95 20.95 20.88 ADLMO
π± 17.06(24) 17.41 16.95 17.04 ADO
π0 9.43(38) 9.97 9.59 9.61 ADLO
η 0.99(4) 1.02 1.00 0.78 ALO
ρ(770)0 1.24(10) 1.18 1.50 1.17 AD
ω(782) 1.09(9) 1.17 1.35 1.01 ALO
η
(958) 0.159(26) 0.097 0.155 0.121 ALO
f0 (980) 0.155(8) 0.111 ∼0.1 — ADO
a0 (980)± 0.14(6) 0.240 — — O
φ(1020) 0.097(7) 0.104 0.194 0.132 ADO
f2 (1270) 0.188(14) 0.186 ∼ 0.2 — ADO
f2 (1525) 0.012(6) 0.021 — — D
K± 2.26(6) 2.16 2.30 2.24 ADO
K0 2.074(14) 2.05 2.07 2.06 ADLO
K (892)±
∗ 0.718(44) 0.670 1.10 0.779 ADO
K∗ (892)0 0.759(32) 0.676 1.10 0.760 ADO
K∗2 (1430)
0 0.084(40) 0.111 — — DO
D± 0.187(14) 0.276 0.174 0.196 ADO
D0 0.462(26) 0.506 0.490 0.497 ADO
D (2010)±
∗ 0.181(10) 0.161 0.242 0.227 ADO
D± s 0.131(20) 0.115 0.129 0.130 O
B∗ 0.28(3) 0.201 0.260 0.254 D
B∗∗
u,d 0.118(24) 0.013 — — D
J/ψ 0.0054(4) 0.0018 0.0050 0.0050 ADLO
ψ(3685) 0.0023(5) 0.0009 0.0019 0.0019 DO
χc1 0.0086(27) 0.0001 — — DL

Table 1: Meson yields in Z0 decay. Experiments: A=Aleph, D=Delphi, L=L3,

M=Mark II, O=Opal. Bold: new data this year. Underlined: disagreement with
data by more than 3σ .

New SLD data include hadron spectra in light quark (distinguished from an-
tiquark) fragmentation, selected by hemisphere using the SLC beam polarization
[45]. One sees strong particle–antiparticle differences in the expected directions
(Fig. 10), bearing in mind the predominance of down-type quarks in Z0 decay.

600
Bryan R. Webber Fragmentation and Hadronization

Particle Multiplicity HERWIG JETSET UCLA Expts

5.9 7.4 7.4
p 1.04(4) 0.863 1.19 1.09 ADO
∆++ 0.079(15) 0.156 0.189 0.139 D
0.22(6) 0.156 0.189 0.139 O
Λ 0.399(8) 0.387 0.385 0.382 ADLO
Λ(1520) 0.0229(25) — — — DO
Σ± 0.174(16) 0.154 0.140 0.118 DO
Σ0 0.074(9) 0.068 0.073 0.074 ADO
Σ ± 0.0474(44) 0.111 0.074 0.074 ADO
Ξ− 0.0265(9) 0.0493 0.0271 0.0220 ADO
Ξ(1530)0 0.0058(10) 0.0205 0.0053 0.0081 ADO
Ω− 0.0012(2) 0.0056 0.00072 0.0011 ADO
Λ+
c 0.078(17) 0.0123 0.059 0.026 O

Table 2: Baryon yields in Z0 decay. Legend as in Table 1.

Figure 5: Particle yields in Z0 decay.

601
Bryan R. Webber Fragmentation and Hadronization

Figure 6: Particle yields in e+ e− annihilation.

1/N dn/dξp

ξ*

4.2
simultaneous Fong-Webber Fit (without Z data)
10
Λeff = 210 MeV 4 MLLA ( = 200 MeV)
189 GeV DELPHI DELPHI DLA
3.8
183 GeV phase space
8
161/172 GeV 3.6 DELPHI
133 GeV ALEPH
3.4
6
91 GeV TASSO
58 GeV TOPAZ
3.2
44 GeV TASSO
4 35 GeV 3

22 GeV DELPHI
2.8
14 GeV
2 2.6

2.4

0
0 1 2 3 4 5 6 7 10 20 30 40 50 60 70 80 90100 200
ξp √s (GeV)

Figure 7: Low-x fragmentation in e+ e− annihilation.

602
Bryan R. Webber Fragmentation and Hadronization

Figure 8: Cone angle in pp → dijets

CDF Preliminary
CDF Preliminary

-- - -
K = 0.477 + 0.005 + 0.058 --- -
K = 0.525 + 0.004 +0.051 -- - -
K = 0.542 + 0.004 + 0.054
Qeff = 0.2207 + 0.0022 + 0.0350 Qeff = 0.2300 +0.0026 + 0.0379 Qeff = 0.2405 + 0.0006 +0.0388
x2/ndf = 157.7 / 31 x2/ndf = 130.4 / 36 x2/ndf = 117.9 / 41
x

1 dN
Nevent dxx
x
x

1
x=log(_ )
x

Figure 9: Low-x fragmentation in pp → dijets.

5 Quark and gluon jets

DELPHI [46] selects gluon jets by anti-tagging heavy quark jets in Y and Mercedes
three-jet events (Fig. 11). As expected, the higher color charge of the gluon (CA =
3 vs. CF = 4/3) leads to a softer spectrum and higher overall multiplicity (Fig. 12).
In general, the relative multiplicities of identified particles are consistent with
those of all charged, with no clear excess of any species in gluon jets (Fig. 13). In
particular there is no enhanced φ(1020) or η production:

DELPHI [46] : Ng (φ)/Nq (φ) = 0.7 ± 0.3

603
Bryan R. Webber Fragmentation and Hadronization

0
1.0 p Λ 1.0

Normalized Difference

Normalized Difference
0.5 0.5

0.0 0.0

−0.5 −0.5
− −
1.0 π K 1.0

Normalized Difference

Normalized Difference
JETSET
UCLA
0.5 HERWIG 0.5

0.0 0.0

−0.5 −0.5
0.0 0.2 0.4 0.6 _
0
K* 1.0

Normalized Difference
xp

0.5

0.0

−0.5
0.0 0.2 0.4 0.6 0.8
xp

Figure 10: Normalized particle–antiparticle differences in quark jet fragmenta-

tion. Results are from SLD [45].

a) b)
Jet 3
Jet 3

Jet 1 θ2 Jet 1 θ2
θ1 θ1
θ3 θ3

Jet 2
Jet 2

Y events Mercedes events

Figure 11: Selection of gluon jets by DELPHI.

OPAL [47] : Ng (η)/Nq (η) = 1.29 ± 0.11 .

OPAL [48] selects gluon jets recoiling against two tagged b-jets in the same
hemisphere. Monte Carlo studies indicate that such jets should be similar to
those emitted by a point source of gluon pairs. The qualitative message from
the data is again clear (Fig. 14): Gluon jets have softer fragmentation than light
quark jets, and higher multiplicity. The precision of the data is now such that
next-to-leading order calculations of the relevant coefficient functions, taking
into account the experimental selection procedures, are needed to check the uni-
versality of the extracted gluon fragmentation function.

604
Bryan R. Webber Fragmentation and Hadronization

Y topology

(1/NJet) dN/dxE(ch)
10 2 DELPHI
Quark jets
Gluon jets
10 Jetset 7.4
Ariadne 4.08
Herwig 5.8C
1

-1
10

-2
10

-3
10
Gluon / Quark

1.4 Gluon / Quark

1.2
1
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8
xE(ch)

Figure 12: Charged particle spectra in quark and gluon jets.

The ratio of gluon/quark multiplicities at low rapidity (large angle) is close

to the ratio of color charges r ≡ CA /CF = 2.25, in agreement with local parton-
hadron duality:

OPAL: rch (|y| < 1) = 1.919 ± 0.047 ± 0.095 .

Monte Carlo studies [48] suggest that a better measure of CA /CF is obtained
by selecting low-momentum hadrons with relatively large pT (i.e. low rapidity).
This gives

OPAL: rch (p < 4, 0.8 < pT < 3 GeV) = 2.29 ± 0.09 ± 0.015 .

DELPHI [49] has observed scaling violation in quark and gluon jet fragmenta-

605
Bryan R. Webber Fragmentation and Hadronization

Gluon/duscb (g/duscb)P / (g/duscb)’All ch.’

Ratio
± DELPHI ±
π Jetset 7.4 π
Ariadne 4.08
2 Herwig 5.8C

a) d)
0
± ±
K K
2

b) e)
0
_ _
pp pp
2

c) f)
0
1 10 1 10
p [GeV/c]

Figure 13: Comparisons of particle spectra in quark and gluon jets.

10 3 8
(a) OPAL 7 (a) OPAL
10 2 gincl. jets gincl. jets
uds jets 6 uds jets
10
Jetset 7.4 5 Jetset 7.4
1 dnch.

1 dnch.
N dxE

1 Herwig 5.9 Herwig 5.9

N dy

4
-1 Ariadne 4.08
10 3 AR-2
-2 Ariadne 4.08
10 2 AR-3
AR-2
-3
10 AR-3 1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6
xE Rapidity y
2
(b) OPAL (b) OPAL
gincl./uds jets 2 gincl./uds jets
1.5 Jetset 7.4 Jetset 7.4
Rg/q N dx ch.

Rg/q N dy ch.
E

Herwig 5.9 Herwig 5.9

1 dn

1 dn

1.5
Ariadne 4.08 Ariadne 4.08
1
AR-2 AR-2
1
AR-3 AR-3
0.5
0.5

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6
xE Rapidity y

Figure 14: Momentum fraction and rapidity distributions in quark and gluon jets.

606
Bryan R. Webber Fragmentation and Hadronization

tion separately (Fig. 15) by studying the dependence on the scale

1
κH = Ejet sin(θ/2)  sy3
2
where θ is the angle to the closest jet and y3 is the Durham jet resolution [50]
at which 3 jets are just resolved. This is expected to be the relevant scale when
y3 becomes small. One sees clearly that there is more scaling violation in gluon
jets (Fig. 16). The ratio provides another measure of CA /CF :

DELPHI: [Link]. = 2.23 ± 0.09 ± 0.06 .

A crucial point in the DELPHI analysis is that 3-jet events are not selected
using a fixed jet resolution ycut , but rather each event is clustered to precisely 3
jets. This avoids ‘biasing’ the gluon jet sample by preventing further jet emission
above ycut .
The same point is well illustrated in analyses of average multiplicities in 2-
and 3-jet events [51, 52, 53]. If Nqq (s) is the ‘unbiased’ qq multiplicity, then in
events with precisely 2 jets at resolution ycut there a rapidity plateau of length
ln(1/ycut ) (see Fig. 17) and the multiplicity is

N2 (s, ycut )  Nqq (sycut ) + ln(1/ycut )Nqq (sycut )

where N  (s) ≡ sdN/ds. Clustering each event to 3 jets we get this multiplicity
with y3 in place of ycut , plus an unbiased gluon jet:

1
N3 (s)  N2 (s, y3 ) + Ngg (sy3 ) .
2

Thus, one can extract the unbiased gg multiplicity, plotted in Fig. 18 vs. p1T ∼
√
sy3 [54]. The ratio of gg/qq slopes gives yet another measure of CA /CF [53]:

rmult = 2.246 ± 0.062(stat.) ± 0.080(sys.) ± 0.095(theo.) .

6 Current and target fragmentation in DIS

H1 [55] and ZEUS [56] have studied the distributions of xp = 2|p|/Q in the
current and target hemispheres in the Breit frame (Fig. 19).
In the current hemisphere, one expects fragmentation of the current jet (C in
Fig. 20), similar to half an e+ e− event. In the target hemisphere, the contribution
T1 is similar to C, T2 gives extra particles with xp < 1, while T3 gives xp ≥ 1,
generally outside the detector acceptance.

607
Bryan R. Webber Fragmentation and Hadronization

xE xE
1/NJetdNch/dxE

1/NJetdNch/dxE
interval interval
10 2 DELPHI 10 2 DELPHI 0.03 ... 0.08
0.03 ... 0.08

0.08 ... 0.15

0.08 ... 0.15
10 10
0.15 ... 0.23 0.15 ... 0.23

0.23 ... 0.33

0.23 ... 0.33
1 0.33 ... 0.45 1
0.45 ... 0.60 0.33 ... 0.45

-1 -1
10 0.60 ... 0.75 10 0.45 ... 0.60

0.75 ... 0.90

-2 -2 0.60 ... 0.75
10 10
0.75 ... 0.90
Quark Gluon
-3 -3
10 10
All events All events
1st order DGLAP-evolution 1st order DGLAP-evolution
-4 -4
10 10
5 6 7 8 9 10 20 30 5 6 7 8 9 10 20 30
κH [GeV] κH [GeV]

Figure 15: Scale dependence of quark and gluon fragmentation.

dlog DH(xE,κ) / dlog κ

0.4 DELPHI
0.2

-0.2

-0.4

-0.6
Quark
-0.8 Gluon

-1 Tasso (Quark)

DGLAP evolution in 1st order

-1.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
xE

Figure 16: Logarithmic gradients of quark and gluon fragmentation.

1 ln s 1 ln 1 1 ln 1 1 ln s
2 2 y cut 2 y cut 2

Figure 17: Rapidity plateau in 2-jet events.

608
Bryan R. Webber Fragmentation and Hadronization

35

<Nch>
<N>gg Eden et al.
DELPHI <N>gg = 2(<N>3 -<N>qq)
30 CLEO Υ decays
OPAL (recoiling gluon)
25
<N>qq
e+e- experiments
20

15

10

2
10 10
Scale = √s, p1T [GeV]

Figure 18: Average qq and gg multiplicities deduced from 2- and 3-jet events.

pt

(1-x)Q/2x

-Q/2
Q/2 p
L

CURRENT TARGET

Figure 19: Breit frame current and target regions in DIS.

e+
;
;;
;;
e+ ;
- ;











 - C

6

  T1






















 
T2  

















 

     T3


















 

  














- 
proton  remnant

Figure 20: Contributions to the final state in DIS.

609
Bryan R. Webber Fragmentation and Hadronization

Figure 21: Charged multiplicity in current hemisphere.

Figure 22: Fragmentation in DIS. Upper data (heavy curve) target region, lower
data (light curve) current region.

610
Bryan R. Webber Fragmentation and Hadronization

1. In the current hemisphere, the charged multiplicity is indeed similar to e+ e−

(Fig. 21 [56]). Differences at low Q2 are consistent with the expected boson-
gluon fusion contribution. The distribution of ξ = ln(1/xp ) is also similar
to e+ e− , e.g. close to Gaussian with little Bjorken x dependence (Fig. 22).
At low Q2 , there is evidence of strong subleading corrections. The distri-
bution is skewed towards higher values of ξ (smaller xp ), contrary to MLLA
predictions (Fig. 23). The quantity plotted is
   3
3 2 2
Skewness ≡ (ξ − ξ) / (ξ − ξ) .

On the other hand, the data lie well below the fixed-order perturbative pre-
diction [57] at low xp and Q2 (Fig. 24). Discrepancies could be due to power-
suppressed (1/Q2 ) corrections, of dynamical and/or kinematical origin. The
bands in Fig. 24 correspond to an ad hoc correction factor
  
2 −1
1 + meff  (0.1 < meff < 1 GeV).
Qxp

2. In the target hemisphere, there is also disagreement with MLLA [56], possi-
bly due to the T3 contribution “leaking” into the region xp < 1. If anything,
Monte Carlo models predict too much leakage (Fig. 25). Little Q2 depen-
dence is evident.

7 Heavy quark fragmentation

New data on b → B fragmentation from SLD [58], using high-precision vertexing,
discriminate between parton-shower plus hadronization models (Fig. 26). Note
that the data have not yet been corrected for detector effects.
Including more perturbative QCD leads to a reduction in the amount of non-
perturbative smearing required to fit the data. Non-perturbative effects are con-
ventionally parameterized by εb in the Peterson function [59]
−2
1 1 εb
f (z) = 1− − (z = xB /xb ) .
z z 1−z
In various models, the expectations for εb are

Pure Peterson [58] : εb = 0.036

JETSET ( LLA QCD) + Peterson [58] : εb = 0.006
NLLA QCD + Peterson [60] : εb = 0.002 (Fig. 27) .

611
Bryan R. Webber Fragmentation and Hadronization

Figure 23: Skewness in current fragmentation region.

Figure 24: Scaling violation in DIS fragmentation.

612
Bryan R. Webber Fragmentation and Hadronization

Figure 25: Target fragmentation compared with models.

300 JETSET + BCFY JETSET + Bowler JETSET + CS SLD

83/16 17/15 103/16
Entries

200

100

0
300 JETSET + Kart. JETSET + Lund JETSET + Peterson
34/16 17/15 62/16
Entries

200

100

0
300 HERWIG cld=0 HERWIG cld=1 UCLA
460/17 94/17 25/17
Entries

200

100

0
0 0.5 1 0 0.5 1 0 0.5 1
xrec
B xrec
B xrec
B

Figure 26: SLD data on b → B fragmentation compared with models.

613
Bryan R. Webber Fragmentation and Hadronization

SLD Preliminary
3.5

3 Average SLD unfolded data

Inner error bar = stat.

2.5 Outer error bar = stat. + RMS

1/NdN/dxB 2

1.5

0.5

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
xB

Figure 27: SLD [58] and ALEPH [61] data on b → B fragmentation, the latter
compared with NLLA QCD [60].

Expt RL (fm) RT (fm) RL /RT

DELPHI 0.85 ± 0.04 0.53 ± 0.04 1.61 ± 0.10
L3 0.74 ± 0.04 0.56+0.03
−0.06 1.23 ± 0.03+0.40
−0.13
OPAL 0.935 ± 0.029 0.720 ± 0.045 1.30 ± 0.12

Table 3: Longitudinal and transverse source radii.

In the universal low-scale αS model, the perturbative prediction is extrap-

olated smoothly to the non-perturbative region, with no Peterson function at
all [20].

8 Bose-Einstein correlations

Studies of π ± π ± correlations which distinguish between directions along and

perpendicular to the thrust axis find definite evidence for elongation of the source
region along that axis (Fig. 28 and Table 3 [62, 63, 64]). This has a good explana-
tion in the Lund string model, in terms of the change of the space-time area A in
Fig. 4 when identical bosons are interchanged [65].
ALEPH [66] has clear evidence of Fermi-Dirac anticorrelation in ΛΛ (S=1). Plots
A–C in Fig. 29 correspond to different comparison (no-correlation) samples. The
source size appears to decrease with increasing particle mass (Table 4 [66, 67,
68]). However, some of this effect is kinematic [69].

614
Bryan R. Webber Fragmentation and Hadronization

1.4
(a) L3
1.2

0.8
0 0.2 0.4 0.6 0.8 1
QL [GeV]
1.4
(b)
1.2
R2
1

0.8
0 0.2 0.4 0.6 0.8 1
Qout [GeV]
1.4
(c)
1.2

0.8
0 0.2 0.4 0.6 0.8 1
Qside [GeV]

Figure 28: Bose-Einstein correlations with respect to axes along and perpendicu-
lar to the thrust axis.

PRELIMINARY ALEPH
C(Q)

0.5
A)

0.5
B)

0.5
C)

0
0 1 2 3 4 5 6 7 8 9 10

Q [GeV]

Figure 29: Fermi-Dirac correlation for ΛΛ.

615
Bryan R. Webber Fragmentation and Hadronization

Particles Rsource (fm)

ππ 0.65 ± 0.04 ± 0.16
KK 0.48 ± 0.04 ± 0.07
ΛΛ 0.11 ± 0.02 ± 0.01

Table 4: Comparison of source radii.

9 WW fragmentation
In e+ e− → WW, we would expect correlations between W hadronic decays due to
overlap of hadronization volumes. This occurs mainly in the central region, and
is orientation-dependent (Fig. 30). These reconnection effects have been searched
for in single-particle distributions. One would expect discrepancies between the
distributions in semi-leptonic and fully hadronic decays, especially at low mo-
menta. There is no firm evidence yet for such effects in the xp distribution
(Fig. 31 [70]). However, DELPHI [42] report a possible small (∼ 2σ ?) effect in
the distribution of pT relative to the thrust axis (Fig. 32).
Bose-Einstein correlations between hadrons from different W’s are also be-
ing looked for. They would lead to an increase in the correlation function for
WW relative to that for a single W. There is no sign of any increase at present
(Fig. 33 [71]).

10 Summary
1. Detailed fragmentation studies need more theoretical input in the form of
coefficient functions that take account of selection procedures, especially
for gluon jets in e+ e− final states.

2. Hadronization studies suggest that particle masses, rather than quantum

numbers, are the dominant factor in suppressing heavy particle production.
Baryon production is not yet well described by any model.

3. Quark and gluon jets have the expected differences and these can be used
to measure the ratio of color factors CA /CF . There is no strong evidence yet
for different particle content in gluon jets.

4. Fragmentation in DIS shows disagreements with perturbative predictions.

It is not yet clear whether these are due to higher-order or non-perturbative
effects.

616
Bryan R. Webber Fragmentation and Hadronization

Figure 30: Hadronization volumes in WW decay.

ALEPH PRELIMINARY 189GeV

6

0
1 2 3 4 5 6
-ln(Xp)

10

0
1 2 3 4 5 6
-ln(Xp)

1.3
1.2
1.1
1
0.9
0.8
1 2 3 4 5 6
-ln(Xp)

Figure 31: Distribution of momentum fraction in WW decay.

617
Bryan R. Webber Fragmentation and Hadronization

DELPHI DELPHI
183 GeV 189 GeV
a) a)

PYTHIA PYTHIA

DELPHI DELPHI
183 GeV 189 GeV
b) b)

PYTHIA PYTHIA

Figure 32: Distribution of transverse momentum in WW decay.

Figure 33: Bose-Einstein correlations in WW decay.

5. New precise b quark fragmentation data test models and suggest that per-
turbative effects dominate.

6. Bose-Einstein (Fermi-Dirac) correlations show elongation of the source along

the jet axis and source shrinkage with increasing mass.

7. WW fragmentation still shows no firm evidence for correlation between the

618
Bryan R. Webber Fragmentation and Hadronization

decay products of the two W’s.

References
[1] J. Womersley, these proceedings.

[2] M. L. Mangano, plenary talk at 1999 EPS Conference, Tampere, Finland, July,
1999, hep-ph/9911256.

[3] J. C. Collins and D.E. Soper, Ann. Rev. Nucl. Part. Sci. 37, 383 (1987).

[4] See the fragmentation web pages of the European network “Hadronic Physics
with High Energy Electromagnetic Probes” (HaPHEEP) for much useful infor-
mation on this topic:
[Link]

[5] M. Beneke, these proceedings.

[6] [Link]

[7] Aleph Collaboration, [Link]

[8] DELPHI Collaboration, [Link]

delsec/conferences/tampere99/

[9] L3 Collaboration, [Link]

[10] OPAL Collaboration, [Link]

[11] SLD Collaboration, [Link]

[12] H1 Collaboration,
[Link]

[13] ZEUS Collaboration, [Link]

[14] D∅ Collaboration, [Link]

[15] Y. I. Azimov, Y. L. Dokshitzer, V. A. Khoze and S. I. Troian, Zeit. Phys. C31,

213 (1986).

[16] C. P. Fong and B. R. Webber, Phys. Lett. B229, 289 (1989); Nucl. Phys. B355,
54 (1991).

[17] Y. L. Dokshitzer, V. A. Khoze, A. H. Mueller and S. I. Troian, Basics of pertur-

bative QCD (Editions Frontieres, Gif, 1991).

619
Bryan R. Webber Fragmentation and Hadronization

[18] Y. I. Azimov, Y. L. Dokshitzer, V. A. Khoze and S. I. Troian, Z. Phys. C27, 65

(1985).

[19] Y. L. Dokshitzer and B. R. Webber, Phys. Lett. B352, 451 (1995)

hep-ph/9504219.

[20] Y. L. Dokshitzer, V. A. Khoze and S. I. Troian, Phys. Rev. D53, 89 (1996)

hep-ph/9506425.

[21] Y. L. Dokshitzer, G. Marchesini and B. R. Webber, Nucl. Phys. B469, 93 (1996)

hep-ph/9512336.

[22] G. Marchesini and B. R. Webber, Nucl. Phys. B238, 1 (1984).

[23] B. R. Webber, Nucl. Phys. B238, 492 (1984).

[24] G. Marchesini and B. R. Webber, Nucl. Phys. B310, 461 (1988).

[25] G. Marchesini, B. R. Webber, G. Abbiendi, I. G. Knowles, M.H. Seymour and

L. Stanco, Comput. Phys. Commun. 67, 465 (1992).

[26] G. Marchesini, B. R. Webber, G. Abbiendi, I. G. Knowles, M. H. Seymour and

L. Stanco, hep-ph/9607393.

[27] D. Amati and G. Veneziano, Phys. Lett. 83B, 87 (1979).

[28] G. Marchesini, L. Trentadue and G. Veneziano, Nucl. Phys. B181, 335 (1981).

[29] I. G. Knowles and G. D. Lafferty, J. Phys. G G23, 731 (1997) hep-ph/9705217.

[30] B. Andersson, G. Gustafson, G. Ingelman and T. Sjostrand, Phys. Rept. 97,

31 (1983).

[31] T. Sjostrand, Comput. Phys. Commun. 39, 347 (1986).

[32] T. Sjostrand and M. Bengtsson, Comput. Phys. Commun. 43, 367 (1987).

[33] T. Sjostrand, Comput. Phys. Commun. 82, 74 (1994).

[34] T. Sjostrand, CERN-TH-7112-93, hep-ph/9508391.

[35] C. D. Buchanan and S. B. Chun, Phys. Rev. Lett. 59, 1997 (1987).

[36] S. B. Chun and C. D. Buchanan, Phys. Lett. B308, 153 (1993).

[37] S. Chun and C. Buchanan, Phys. Rept. 292, 239 (1998).

[38] G. Rudolph, private communication.

620
Bryan R. Webber Fragmentation and Hadronization

[39] P. V. Chliapnikov, Phys. Lett. B462, 341 (1999).

[40] G. Alexander et al. [OPAL Collaboration], Z. Phys. C73, 569 (1997).

[41] DELPHI Collaboration, EPS-HEP99 paper 3_147.

[42] DELPHI Collaboration, EPS-HEP99 paper 1_225.

[43] DELPHI Collaboration, EPS-HEP99 paper 1_225.

[44] CDF Collaboration, EPS-HEP99 paper 1_600.

[45] K. Abe et al. [SLD Collaboration], hep-ex/9908033.

[46] DELPHI Collaboration, EPS-HEP99 paper 3_146.

[47] OPAL Collaboration, EPS-HEP99 paper 1_4.

[48] G. Abbiendi et al. [OPAL Collaboration], Eur. Phys. J. C11, 217 (1999)
hep-ex/9903027.

[49] DELPHI Collaboration, EPS-HEP99 paper 1_571.

[50] S. Catani, Y. L. Dokshitzer, M. Olsson, G. Turnock and B. R. Webber, Phys.

Lett. B269, 432 (1991).

[51] S. Catani, B. R. Webber, Y.L. Dokshitzer and F. Fiorani, Nucl. Phys. B383, 419
(1992).

[52] P. Eden, G. Gustafson and V. Khoze, Eur. Phys. J. C11, 345 (1999)
hep-ph/9904455.

[53] P. Abreu et al. [DELPHI Collaboration], Phys. Lett. B449, 383 (1999)
hep-ex/9903073.

[54] DELPHI Collaboration, private communication.

[55] C. Adloff et al. [H1 Collaboration], Nucl. Phys. B504 (1997) 3 hep-
ex/9707005.

[56] J. Breitweg et al. [ZEUS Collaboration], Eur. Phys. J. C11, 251 (1999)
hep-ex/9903056.

[57] D. Graudenz, Phys. Lett. B406, 178 (1997) hep-ph/9606470.

[58] K. Abe et al. [SLD Collaboration], hep-ex/9908032.

[59] C. Peterson, D. Schlatter, I. Schmitt and P. Zerwas, Phys. Rev. D27, 105 (1983).

621
Bryan R. Webber Fragmentation and Hadronization

[60] P. Nason and C. Oleari, hep-ph/9903541.

[61] D. Buskulic et al. [ALEPH Collaboration], Phys. Lett. B357, 699 (1995).

[62] DELPHI Collaboration, EPS-HEP99 paper 1_221.

[63] L3 Collaboration, EPS-HEP99 paper 3_280.

[64] OPAL Collaboration, EPS-HEP99 paper 3_64.

[65] B. Andersson and M. Ringner, Phys. Lett. B421, 283 (1998) hep-ph/9710334.

[66] ALEPH Collaboration, EPS-HEP99 paper 1_389.

[67] D. Decamp et al. [ALEPH Collaboration], Z. Phys. C54, 75 (1992).

[68] P. Abreu et al. [DELPHI Collaboration], Phys. Lett. B379, 330 (1996).

[69] M. Smith, hep-ph/9912250.

[70] ALEPH Collaboration, EPS-HEP99 paper 1_387.

[71] ALEPH Collaboration, EPS-HEP99 paper 1_388.

Discussion

Charles Buchanan (UCLA): To elaborate on the UCLA approach to hadroniza-

tion: We find that the idea of a spacetime area law (as suggested by strong QCD)
works very well as an organizing principle in the stage of the soft strong-coupled
hadronization part of the process—that is, it gracefully predicts “easy” data such
as light-quark meson production rates and distributions in e+ e− with few parame-
ters and forms an attractive basis for studying the relation with the perturbative
stage and more complicated phenomena such as baryon formation and pT ef-
fects. To study these latter in detail, we have joined BaBar where we will use the
108 high quality events to be collected in the next 2–3 years to study for baryon-
meson-antibaryon 3-body corrections, pT correlations, etc. We invite physicists
interested in the area to contact us (at buchanan@[Link]).

George Hou (National Taiwan University): Regarding particle content of gluon

jets, is there any result on the η content? This particle is more naturally associ-
ated with gluons than η or φ.

622
Bryan R. Webber Fragmentation and Hadronization

Webber: As far as I know, there are no results available yet on the η content of
gluon jets.

Michael Peskin (SLAC): You have shown that, when quark and gluon jets are
carefully selected, their average properties are clearly distinguished. Of course,
what one really wants is a variable which allows one to separate quark and gluon
jets (if only statistically) in the Tevatron or LHC environment. What is the best
choice for this purpose?

Webber: This is difficult to do because, although the average properties are

different, the fluctuations are large.

Tom Ferbel (University of Rochester): Actually, D0 has used the differences

between quark and gluon jets very effectively, in a statistical manner (with neural
networks) to improve the signal to background in the analysis of tt production
in the all-jets channel.

623