zea~-~.

hnlP ~ k ~
~Jr Physik C
Z. Phys. C - Particles and Fields 20, 317-329 (1983)
and
0 Springer-Verlag 1983

A General Model for Jet Fragmentation

B. Andersson, G. Gustafson, B. S6derberg 1

Department of Theoretical Physics, University of Lund, S/51vegatan 14A, S-22362 Lund, Sweden

Received 14 April i983

Abstract. We present a general condition on quark neglected, only has two free parameters. These are
fragmentation which gives a hadron distribution sat- related to the multiplicity and to the rapidity cor-
isfying Lorentz invariance and causality. The had- relations between neighbouring mesons, and it is
ronization can be described as an iterative cascade interesting to note that they also have an intuitively
process, symmetric with respect to iteration from the appealing interpretation as related to the area and
quark and the antiquark ends. The possible particle perimeter dependence of Wilson loop integrals. We
distributions are strongly restricted, with few free generalize the situation to n I flavours and then ob-
parameters related to the total multiplicity and cor- tain ( h i + l ) independent parameters, although pre-
relations in rapidity. These parameters can be given sent phenomenological applications do not seem to
an appealing interpretation in terms of the expected require more than two parameters. We feel that it is
area and perimeter dependence of Wilson loop in- particularly interesting that the very general require-
tegrals. ment given above puts such strong constraints on
the fragmentation process. Further, the stochastic
process defined in this way exhibits an appealing
symmetry in the sense that, for any part of a large
1. Introduction jet, it gives the same result independently of whether
it is applied to peel off mesons from the quark end
In this paper we will study some basic consistency or from the antiquark end.
requirements on the presently popular cascade mod- In order to formulate the general assumption
els for quark and antiquark fragmentation into fi- above in a quantitative language, we will consider a
nal state hadrons. model in one space and one time dimension with a
The dynamical picture we have in mind is the space-time structure similar to that of [1]. We con-
following: When a quark and an antiquark separate, sider a quark-antiquark pair (qoC~o), produced with
a colour field of narrow width is stretched between large energy in e.g. an e+e -annihilation event. For
them. This field (which we will represent by a rela- simplicity we assume that they are produced in the
tivistic string) can break repeatedly by the produc- space-time origin 0 (Fig. 2).
tion of qq-pairs, which combine to give a jet of
hardons emanating in two opposite directions, as
shown in Fig. 1.
We will assume that whenever a system breaks
into two parts, these will further fragment indepen-
dently of each other.
We will then show that this assumption leads to
an iterative cascade model which, for the simplified o
case of only one quark flavour and one type of
meson, and when the transverse dimensions can be Fig. 1. When a quark and an antiquark move apart with large
energies a linear colour field is stretched between them. This
1 Address after 1 Jan. 1983: NORDITA, Blegdamsvej 17, colour field breaks into pieces, hadrons, by the production of new
DK-2100 Copenhagen O, Denmark quark antiquark pairs
318 B. Andersson et aL: A General Model for Jet Fragmentation

J\\

x\\ /t 9

>

o L e O "

O
Fig. 2. The process in Fig. 1 pictured in x and t. The quarks are Fig. 3. The production of the pair q~cT~in the point V/ splits the
assumed to be massless. The hatched areas indicate nonvanishing original system in two parts, which are assumed to fragment
colour field independently of each other

As a model for the colour field between them we of a particular final state meson with e.g. the first
will use the massless relativistic string with ~c the meeting point of its constituents, it will in any frame
string tension [2]. We will for simplicity work with be the slowest mesons that are firstly produced.
massless quarks, although we will later discuss the Let us now focus our attention on a single vertex
necessary generalizations for massive ones. The orig- Vii (Fig. 3), with light-cone coordinates ( x [ , x T ) . At
inal pair will then go apart along the lightcone in this point qi and 71i are produced and afterwards
opposite directions, with q0 having the energy-mo- dragged apart by the field. This process evidently
mentum (W+,0) and 710(0, W ) in a lightcone basis, divides the system into two separate subsystems
where W e = E +_P. (7to, qi) and (qi, qo)-
The space-time points where new pairs are pro- Exactly the same situation would occur if the
duced will be called production vertices, and the two states (71o,qi) and (71~,qo) would have started in
field will after the production accelerate the quark the space-time points marked 0 L and O R in Fig. 3,
and antiquark in opposite directions. As more and with energy-momenta given by
more q71-pairs are produced, small string pieces with
a q from one vertex and a 71 from the adjoining qo: ( W+ --'~x/+, O)
vertex form stable states and go away in space-time 71i: (0, •xF) (1)
as small string " y o - y o " states [3,4] as shown in qi: ( Kx+, O)
Fig. 2. It is easy to convince oneself that the area 710: (0, W - -~cxF).
(counted as d x + d x -) of the space-time surface
spanned during one half-period of oscillation of such The independence principle above can now be for-
a state is given by m2/~c 2 where m is the mass of mulated as:
the state. The two subsystems, obtained by the breaking of
We note that in this simple model the momen- the string in the production vertex V~, will fragment
tum of a yo-yo state is always carried by the end- in the same way as if the systems had been produced
point q and 71, and that the states exhibit the proper independently at the starting points 0 L and O R, and
Lorentz contraction and time dilatation, as expected with properties as in (1).
in a relativistic model. As a further constraint we will assume that when
A fundamental property of this string-based mod- the C.M. energy W + W - of a system is very large
el is that t h e production vertices are causally dis- compared to a typical hadron mass, the vertex dis-
connected and time-ordering is therefore Lorentz tribution in the center of phase-space will approach a
frame dependent. This means, that all vertices must finite limit (this implies a central plateau in rapidity
be treated in the same way; there is no single vertex for the final state mesons).
that is more primary than any other. It should Finally, we note that in a relativistically in-
therefore be possible to consider the production pro- variant theory, the appearance of any vertex may
cess, starting at any single vertex. The vertex distri- only depend on the available Lorentz invariants. In
butions to the right and left of it should then be connection with the vertex Vii discussed above, there
mutually independent. are obviously (if we have no knowledge of the re-
We note, that if we associate the production time maining vertices) three independent Lorentz in-
B. Andersson et al.: A General Model for Jet Fragmentation 319

variants, which can be chosen as e.g,: There are only two independent nontrivial L o r e n t z
invariants available, which can be chosen as s o and
Kx + zl, and thus the distribution of the first meson
s=W+W-; F=K2x+x/- ; z=--. (2)
W+ should be given by a function f(zl, So), so that
In the next section we will consider the con- dP=f(zl,so) dz 1. (4)
sequences of these assumptions and derive a stochas-
tic process for the large energy limit. In Sect. 3 we According to our assumptions, the r e m n a n t system
focus the attention on finite energy systems, and should then fragment as an independent system
exhibit a set of useful properties of the process. In starting at the space-time point O 1 in Fig. 4, and
Sect. 4 we discuss a n u m b e r of phenomenological with a squared mass s I given by
implications, and in Sect. 5 we c o m m e n t u p o n the
similarities and differences between our a p p r o a c h sl=(1-z 0 So--~ . (5)
and the one suggested very early by A r t r u and Men-
nessier [4]. In Sect. 6 we reconsider the process in This is p r o p o r t i o n a l to the dashed area in Fig. 4.
terms of the general notion of colour coherence. The distribution of the next vertex V2 should then be
given by

2. Derivation of the Model dP =f(z2, sl) dz 2 (6)

One consequence of the assumption in Sect. 1 is that where f is the same function as in (4), and z 2 is the
the production of mesons may be described by an fraction of the remaining lightcone m o m e n t u m (1
iterative cascade principle. To see that, we consider - z ~ ) W + taken by the second meson.
a system (qo, q0) with lightcone m o m e n t a (W +, W - ) In this way the fragmentation can be described
= ( E + P , E - P ) and a squared mass so=W+W -. by a sequence of steps, each determined by the func-
We will in the subsequent discussion assume only tion f(z, s).
one flavour and one kind of meson. We focus atten- It is evident that we m a y make an analogous
tion on the vertex (V1 in Fig. 4) closest to the orig- construction for V~ and V~ in Fig. 5, and consider
inal quark q0. This vertex breaks the system into a the fragmentation process as an iterative jet from the
final state meson (c~lqo) with fixed mass m, and a C~o-end of the system. Then we would describe the
r e m n a n t system (q0q0- We note that the position process by means of steps along the negative light-
(x~,x[) of V1 is determined by the fraction z I of cone, with a fractional negative light-cone z de-
W § taken by the meson fined in analogy with z above.
Next we consider the consequences of our re-
m 2 quirement that for large energies, there should be a
x[=l(1--Zl) W+; x;-- (3)
central plateau in rapidity. The distribution of ver-
K gz 1 W +

/ "t

\ / \ \
N

\\
/

/I/ \
l!

V'1 /
~\ \'.

\ \ \

\
o
o 5/
Fig. 4. In the point V1 the original system is split into one final o
state meson and a remnant system, which decays further. This Fig. 5. The iterative cascade in Fig. 4 can equally well be treated
leads to an iterative cascade principle from the other end
320 B. Andersson et al.: A General Model for Jet Fragmentation

tices in the center must then be independent of the Introducing

hyperbolic angle y and only depend on the proper
time z. Thus, with ff:N"C 2, we must have in this z f (z) = exp g(z) (14)
limit H(F) =-exp h(F) (15)
dP we obtain from (13)
dydF - C~ H(F), S H(F) dF = 1 (7)
0
g(z+)+h(m 2 1 - z - ] (m 2 1 - z + ] (16)
with C o equal to the number of vertices per unit in \ z+z / = g ( z )+h \ z + z _ / "
y. In order to obtain such a limit for the central
production, the function f(z, s) must be independent Differentiating with respect to both z§ and z_, the
of s for large s. Thus f(z,s) must go to a limit f(z): result can be written as
lim f(z, s) : f ( z ) . (8) h'(F~) + F~h"(F~) = h'(F~+ a) + P~+a h"(F~+ 1). (17)
8400

We will study the consequences of the property that We conclude that each side of this equation must be
the fragmentation can be thought to proceed either a constant - b independent of Fi and Fi+ 1, and we
from the quark end or from the antiquark end. obtain immediately for H the solution
For a vertex V~ with lightcone coordinates H ( F ) = CF"e -br (18)
(x +, x[-) in the central region of a large mass system,
the distribution of the next point Vii+a is given by where a and C are integration constants. If this
the limit function f above. Thus the distribution in result is substituted in (16) we obtain
the positions of a pair of neighbouring vertices is
determined by b rn2
g(z + ) - a . log(1- z + ) + - -
Z+
dP = H(F~) f (z + ) dFflz + (9)
bm z
= g(z_) - a. log(1 - z_) + - - (19)
with z + = z , the fraction of the forward lightcone Z
m o m e n t u m of the system (q0, ql) taken by the meson
(eli + 1 q i)" and we may again conclude that each side is a
Alternatively, if we generate the fragmentation constant, logN, independent of z+ and z so that
from the antiquark end, the position of Vi+ 1 is de-
scribed by the distribution H(F/+l)dFi+,, and the f (z) = N (1 - z)" exp - rbn 2 (20)
"first" vertex F/ must be generated by f(z_) where Z Z

z is the fraction of the backward lightcone momen-

The two constants C and N are normalization con-
tum of the system (qi+aqo) taken by the meson,
stants while a and b are two parameters which have
leading to the pair distribution
to be determined from further dynamical input. We
will consider this question in Sect. 4.
dP = H(Fi+ 1) f (z_) dFi+ x dz_. (10)
We end this section by generalizing the above
The two expressions in (9) and (10) must evidently result to several flavours and hadron masses. We
be identical in order for the scheme to be consistent. note that the constant a is obtained as an inte-
The variables are connected by the relations gration constant. It may then depend on the flavour
produced at the vertex, in contrast to the parameter
F/= m 2 . 1 - z _ . , /]+ 1 = m 2 . 1 - z + (11) b, which must be the same for all vertices. We note,
Z+Z Z+Z_ however, that the functional form of the function H~
(with c~ indicating the flavour produced) is inde-
and we note that this implies pendent of the masses of the final state hadrons:

H~(F) = C~F "~ e -hr. (21)

dz+ dFi = -dz
- dFi+ 1. (12)
Z+ Z_
For the fragmentation function f, however, there is a
We therefore obtain the requirement mass dependence. F r o m the construction procedure
above we conclude that starting from a vertex with
flavour c~ (power G), the distribution of z for the
z + f ( z + ) H ( m Z l z + ; - ) = - z _ f ( z _ ) H ( m 2 z + zi _- z + ] ,. (13)
next vertex, with flavour fi (power a~) giving a final
B. Andersson et a1.: A General Model for Jet Fragmentation 321

state meson with mass m, is given by

f~ (z)dz=N~t3~z ~
.z 7 exp---z (22)

The symmetry of the process is particularly notice-

able when we write down the pair distribution dP of v,
(9):

dp:fo;fl(?)a~(l-z+t a13
\ z+ I
Fig. 6. An n-particle cluster is obtained from the production
.exp(_bm2tdz+ dz vertex V.. The dashed region shows the area A. defined in the text
\ z+z_/ 2.2 Z2-
_ (23)

with C ~ a normalization constant.

and in terms of these variables we define the area in
energy m o m e n t u m space A, (dashed in Fig. 6)
3. Finite M a s s S y s t e m s
A,(31,..., " mk = (28)
In the previous section we studied the fragmentation
of an initial state q0q0 with a very large energy. The
mesons can be peeled off from the end one by one. If we insert the definitions for g and z in terms of 6-
This implies that we can also study a subsystem functions and the expressions for f from (22) we
composed of a finite number of particles and analyze obtain by straight forward means
both its production and its decay into stable mesons.
We will start to consider an n particle cluster, i.e.
n particles adjacent in rank with a total invariant
mass ]/g and a total invariant energy m o m e n t u m 9e x p ( - b g l z Z ) d _ P ~ ( ~ l , . . . , 3 , ) (29)
light cone fraction z. The particle masses and the
respective lightcone fractions are m k and z k, k
= 1, ..., n and the flavour powers at the vertices are with
G, ~ = 0 ..... n in such a way that the meson labelled
1 is composed of (q0q0 etc. Then according to the
general properties of an iterative cascade jet [5] we =exp(-bA,(3j))c5 1- 3k N
obtain for the probability \ i 3k /

d P ( z l , ..., z,) 9 l ".G

lNk-~ kd3k
- . 3~k- ~--ak. (30)
k= 1 ' ~k

L - - l,k k--1 k 1 ,24 We note that in (29) the dependence on the internal
= -- ~-1 Zj 1 - ~ zj variables 3k and the "external" variable z factorizes.
j= j=l

The total mass square and lightcone fractions are

g=z i m2 (25) A. Production of a Cluster with Fixed Mass

k = l Zk
If we study the distribution in z for a cluster with
z = ~ z k. (26) fixed mass ]/~, the distribution is independent of the
k--1
internal variables and even independent of the mul-
It is useful to define the scaled "internal" lightcone tiplicity n. Thus for any cluster with mass l ~ the
energy m o m e n t a fractions 3k: distribution in z is given by
n
~k = Z; ::::>k~- 1 ~ k = l (27) ~Pz (z, g)= E,ola .b(g) 9-1- z "~ (}@~)a". e x p - bg . (31)
' "' Z 2"
322 B. Andersson et al.: A General Model for Jet Fragmentation

Here Eao,,,; b is a normalization constant given by fragmentation of a system with finite cm energy l/g.
the exponential integral This is obtained from (35) if we for fixed values of g
oo
and a, integrate over all the other variables 32, ..., 3,
Eao, a~,;b(~)= (bg)ao_,. ~ dy (y-b~)~"e_,. (32) and sum over different possible multiplicities n.
v~ Y yaO For simplicity we will in detail only consider the
dP case of a single flavour and a single meson mass and
In particular we notice that dzz in (31) has the same indicate at the end the corresponding more general
situation.
functional form as (22) for the production of a single
We integrate the expression in (35) over all the
stable meson (qoCT,) with the meson mass m replaced variables 32 . . . . . an, thereby deriving the one particle
by the cluster mass 1 ~ and with the first and final semi-inclusive (fixed multiplicity) distribution for the
flavour powers a o and a Nrespectively occurring. mass lf~. We obtain easily by similar scaling oper-
The distribution in proper time z o r / ' = t c ' c 2 for
ations as in connection with (27):
the vertex where the cluster is split off is easily
obtained from the relation
F,(al ; S)=da a dg=E,,,;b(s) exp(bg)N"g,(a, ; g). (36)
1--z
v =~ z
(33)
The expression for g.(a~;g) is
If we insert this into (31), we obtain
1 ( b i n 2] ^

dP F F~"
dF ( ; s*)oC(F + g),o+ ~ e x p ( - b F ) . (34)
with the energy dependence given by the factor
We note that for large values of g, i.e. when we have g n - 1:
gone away from the end of the jet, this distribution
n
approaches quickly the distribution H(F) in (21).
9= ~ f . e x p [ - b A , ,r -..,),)]
The fact that we obtain the same functional ex-
pression for the probability to produce a single me-
son with given flavour or a cluster of mesons with
the same "initial and final" flavours but with any
multiplicity, is a nice expression for the indepen- If this expression is summed for all values of the
dence principle that we have formulated in Sect. 1. multiplicity n we obtain the total inclusive distribu-
The breakaway of any subsystem only depends on tion F(a,;R) for the first rank particle in any cluster
the total mass, the total invariant lightcone fraction
of mass 1 ~ :
and the initial and final flavours. This result is also a
strong argument for the consistency of our general-
ization to several flavours and different hadron mas- F(3, ; g) = ~ F~(a~; g). (39)
n=l
ses in the end of Sect. 2. The production mecha-
nism is evidently the same whether the produced We note that evidently all F, with n>l/g/m must
system is a single stable heavy particle or an object vanish so that for a given ~ the sum contains only a
decaying into many lighter particles. finite number of nonvanishing terms.
Normalizing F by
gn
B. Decay of a Subsystem
f(a,~)- F(a,;~) _Z. g.(a,,e) (40)
If the distribution dP in (29) is integrated over z we
,
IF(~,,, ^s) da'~
!
2 N n ^g,(s)
n
obtain the exclusive n-particle decay distribution for
a subsystem with mass ]fg. The result is we obtain the finite energy version f(3l,g) of our
fragmentation function f in (20). It is interesting to
dP ~ dff ~ note that the distribution f ( a l , g) only depends upon
da,.., ds. d g - E ..... ; h(g) exp (b s) d a ~ da" (35) the parameter a through the normalization constants
N. These constants appear only in the weights N ~
with d/~ given in (30). for the different multiplicity contributions g,. In that
We now want to investigate the finite energy way the a-dependence does have a direct influence
version f(z,g) of the asymptotic distribution for the on the multiplicity of a state with given mass 1~-.
B. Andersson et al.: A General Model for Jet Fragmentation 323

We will now show that the parameter a also de- and therefore for large values of g, fo(~ (with the
termines the asymptotic energy dependence of the superscripts corresponding to the total system fla-
quantity g(s'): yours) will approach the asymptotic distribution f01
in (22).
N gn(S ) = N"Ig,(3, s ) ds. (411
n=l n-1
4. Comparisons with Experiments
and Heavy Quark Fragmentation
It is easily seen that the distribution f(3~, ~) in (40) is
given by (for g>m 2) In this section we would like to make a few remarks
on the phenomenological implications of the fi'ag-
f(31,g)= N (_bm2t g( (1-~l)(g-m2t)\ 311 (42) mentation function f given in (20). The general re-
~t exp 3~ I g(g) sult is that most inclusive spectra are hardly dis-
tinguishable from the results obtained in the stan-
From the scaling operations given above we obtain dard Lund model. However, rapidity correlations
an integral equation for g(s') valid for s' > m 2 become modified and the fragmentation of heavy
quarks becomes noticeably changed.
g(s')= } N ~ e x p
m21s'
(-bm2t g ((I-3)(s'-m2tt.
\ 3 l \ ~ II
(43)
We start by noting that a symmetry between
particle generation from the quark end and from the
antiquark end is by no means a trivial property in
If there is an (approximate) power-solution (s') "' to iterative cascade jets. It is well-known that the frag-
the integral equation (43) we would obtain for large mentation schemes used by Feynman-Field [5] as
values of s': well as the standard Lund model [1, 7, 8] do not
exhibit this property. Different measures of this
asymmetry have been discussed in the literature
s'a'~}Nd~exp(-bm21.(l-3?'s'"" (44) [5, 9].
o 3 1
Although this asymmetry persists even in the
Therefore there is a condition on a': case of a single flavour and a single meson mass, it
is particularly large if different masses can be pro-
1d ,, / bm2\ duced. We note that if the particles are ordered in
l=NIa3(1-3103 exp ( - ~ ) (45)
rapidity then an ordering in momentum may well be
different in this case (heavier particles will get larger
which by comparison to the normalization condition momenta).
for the asymptotic distribution f in (20) determines The basic reason for an asymmetry is not dif-
a'=a. Consequently the finite energy distribution ficult to understand. In any iterative cascade scheme
f(z, g) will approach the asymptotic distribution f, as it is possible to compute a correspondence to the
seen from (42). distribution H(F) for the proper times of the pro-
There is no difficulty to generalize to the situa- duction vertices. The center of the distribution will
tion of many flavours and meson masses. When we be along a hyperbola corresponding to a proper
produce a first rank meson qoql in a subsystem with time ZorN/K, where ~ is some median mass in the
initial and final flavours 0 and c~ and with the finite final state hadronic systems.
mass ]f~, the remaining system will have corre- If a large mass hadron is produced with a small
sponding flavours 1 and ~. Then the expression for value z of energy-momentum, the production vertex
fol corresponding to (42) is will in accordance with Fig. 7 correspond to a prop-
er time far away from this center hyperbola. Unless
,
fo(0~)(~t g) = N,,o.... b
31"~ (1-31] al ~ this is followed by a step with a very large z-value it
31 will take many steps before the proper times of the
vertices are back to the hyperbola. Thus the frag-
mentation process will be asymmetric unless the
\ ~1 l I (46) heavier particles normally take a larger fraction of
go~(~) energy-momentum.
It turns out that the asymptotic behaviour of the It is a noteworthy property of the distribution
energy factor gl~ depends only upon the "final" H(F) for our case that it is functionally independent
flavour power a~, i.e. of the mass parameters of the final state hadrons
and that the distributions f are suppressed for small
gp~(s) ~ s ~ (47) values of z, the more the larger the mass.
324 B. Andersson et al.: A General Model for Jet Fragmentation

ber F0. Then H(F)--+cS(F-Fo), i.e. all qq pair pro-

\ ,..oo ,, // ductions occur at the same invariant time, and the
ordering in rapidity agrees completely with that in
rank. F o r m o r e realistic a and b values, the rapidity
spread f f ( A y x 2 ) = ] / ( A y ~ 2 ) - ( A y 1 2 ) 2 between two
particles adjacent in rank is shown as a function of
rnLl+m• in Fig. 8b (the dependence on re•177
is smaller but not insignificant, with smallest a for
Fig. 7. If a large mass hadron is produced with a small value of z, rnll~m_Lz). Higher a and b values thus lead e.g. to
the production vertex will be far away from the hyperbola corre- stronger local charge, strangeness or b a r y o n n u m b e r
sponding to average proper times. It might then take many steps
before this hyperbola is reached again compensation.

(YcY~)
The two parameters a and b in the distributions
f are generally related to the behaviour close to z
= 0 and to z = 1. The distributions f always vanish
1.0 /
at z = 0 but afterwards peak at z~-bm 2 (if b r a e < l ) //
while the behaviour close to z = 1 is governed by (1

////////
///
-- Z) a.
F o r the subsequent comparisons with experimen-
tal data we have made use of the L u n d Jet M o n t e
Carlo, as implemented by [10]. F o r these compa- 0.5
risons it is necessary to interpret the h a d r o n masses
as the transverse masses. Experimental data are well
reproduced assuming a c o m m o n value of the param-
eter a for all flavours, and the results presented
below are obtained using this assumption. We com-
ment on the possibility of different values a i at the a o.o , , , i
,, rn~1+ m•
end of this section. 0 1 2 3 4 (GeV/c 2 )
Comparisons with experiments then suggest that
0 < a < 2 since a < 0 would give too m a n y and a > 2 a(~-u
too few particles with large momenta. Within the
region 0_<_a < 2 , b > 0 there is a rather large freedom
\
for a and b separately, but there is a strong cor- \
relation between the two: since the mean step in z
decreases and hence the mean multiplicity increases
\
when a is increased or b decreased, only values \
\
along a curve in the (a, b) plane will fit the particle
spectra and multiplicities observed in jet fragmen-
tation. Thus a g o o d agreement can be obtained with
a = l , b = l / 2 . 2 5 G e V -2 but a fit e.g. with a = 0 , b
= 1/9 G e V - e would also be possible.
We note that the collinear gluon and finite field
length corrections in the standard L u n d model [-8]
have effects very similar to the ( l - z ) a term in (20)
i.e. essentially that of suppressing z values close to 1.
Therefore these corrections are not separately im-
b o . . . . ,. % , + %
plemented in our considerations here. 0 1 2 3 4 (GeV/c 2 }
W h e n both a and b are increased (e.g. in such a
way that the m e a n multiplicity stays constant), what Fig, 8. a Rapidity difference (Yl--Y2) and b rapidity spread
happens is that the ordering in rapidity becomes ~(yl-y2)=l/((yl-y2)Z)-(yl-y2)2; both as functions of m•
m o r e and m o r e like the one in rank. The extreme + r e x 2 , for 1 and 2 of adjacent rank. Full line f(z)=J(1-z)
z
case is when a and b are allowed to go towards 1
exp(-m2/2.25z), dashed f(z)=-exp(-m2/9z) and dash-dotted
infinity in the proportions a = b F o for some fix num- standard Lund z
B. Andersson et al.: A General Model for Jet Fragmentation 325

While the spread a(Ay12) decreases for heavier

particles, the mean value <Ay12) actually increases,
as is seen in Fig. 8a. This is related to the fact that
)
small masses give small values of z so that the
remaining jet energy is not reduced much with the
production of a light particle, whereas a heavy par-
ticle will take a large z and thus not leave much
energy to subsequent particles.
Thus e.g. kaon and proton spectra should be-
come relatively harder than pion spectra. Numerical
calculations show, however, that this effect is very
small, and the ratios n/Kip are essentially the same
as in the standard Lurid model. The differences in
transverse mass simply are not large enough, in par- Fig. 9. Massive quarks move along hyperbolae in the x - t plane.
ticular since many pions stem from vector meson The mass and rapidity of the mesons are given by the asymptotes
decays. to these hyperbolae in the same way as for the massless case
If we extend our considerations to the fragmen-
tation of the heavy c-, b- and t-quarks, however,
then we do obtain a noticeable difference.
In Sect. 1 we treated the motion of massless
quarks, moving always along lightcones. Very simi-
lar results are obtained for massive constituents [8],
which however move along hyperbolae as shown in
Fig. 9. The mass and rapidity of the mesons are
given by the asymptotes to these hyperbolae in the
same way as for the massless case. For an initial
heavy qc~ pair the fragmentation looks as in Fig. 10.
If the production of quarks with mass and trans-
verse m o m e n t u m can be treated as a tunneling phe-
nomenon, which well describes the suppression of
strange quarks and large p• [8, 11] then the proba-
bility to produce a c~ pair in the field is extremely
small ( ~ 1 0 10). There is then no compelling argu-
ment that a system with an initial c-quark fragments
in the same way as a system with a c-quark pro-
duced in the field. However if we pursue the idea of
Fig. 10. The fragmentation of an initial heavy q~ pair
independence anyway, we will obtain a well-defined
result for the fragmentation of heavy quarks; this
should also be given by the function f in (20).
spaced in rapidity. A similar result was also obtained
We note that in e.g. the standard Lund scheme
by Bowler [13] from a generalization of the model
for fragmentation, the mean value of z, <z>, should
by Artru and Mennessier [4].
be close to
Our symmetric fragmentation scheme will predict
<z) ~89 (48) an even harder heavy quark fragmentation. For very
large meson masses M we obtain asymptotically
for charm and b o t t o m fragmentation. This means
that light particles produced behind the heavy one a+1
<z> ~ 1 - ~b "M (50)
will often obtain a larger rapidity. This may seem
unphysical, and actually Bjorken [12] has argued
that asymptotically, i.e. for very large masses M, Numerically we obtain for a = 1 and b
= l / 2 . 2 5 G e V -2 that the first rank hadron in a u-
(1 GeV) quark jet will in our scheme on the average be 0.28
<z) ~ 1 (49) rapidity units ahead of any other hadron, in a c-
M
quark jet 0.65 units and in a b-quark jet 1.24 units
This result can be derived from the assumption that ahead (if compared to the second rank particle only,
all .primary hadrons on the average are equally the differences are 0.55, 0.94 and 1.50, respectively).
326 B. Andersson et al.: A General Model for Jet Fragmentation

DCHARM{z) dN (c~z--->
#-+ )
dlxJ

1.5 i
#
1.0 J !

\
\
\
\
\
0.5 ~
fl o
/ i
\ !
.~ Z

0.0 0.5 1.0

d~F
sa-
i (.b GeV2) a
0.0 0.0
,
0.2
\\'~
~,~~, 0.4 0.6
,
0.8
. IxFI = 2,~1r
W

dN(bb "-*~-+}
dlxFI

0.1

0.01

2
b 0.001
0.0
,
0.5 1.0
w
= z= 2...EE

1
Fig. l l a and b. Charm fragmentation spectra for the symmetric
model, full line, and standard Lund, dashed, a in leptoproduction
filled circles CDHS, open circles E531 (preliminary) [15], b in
e+e - annihilation, filled circles Mark II, open circles TASSO, b o i
\ ~ i
~ i i i1_
%1= 21pj
triangles DELCO, open squares CLEO [16]. The former three 0.0 0.2 0.4 0.6 0.8 W
data sets are around 30 GeV, as are the Lurid model curves while
CLEO data are at 10.4 GeV, so that Zmjn=0.39. Large systematic Fig. 12a and b. Muons in a) c~ and b) bb events (W=32 GeV) as
normalization errors may be present due to uncertainties e.g. in a function of ]XF] =2]pL]/W. Full line the symmetric model, dashed
the D ~ ?z+ branching ratio. Contributions from B meson line standard Lund. In e+e annihilations c~ events are 4/1l of
decays are included in the model calculations (and are the origin the total cross section and b/; events 1/11, while u~, dd and s~
of the peaks at low z) events give no contribution to the prompt muon spectrum
B. Andersson et al.: A General Model for Jet Fragmentation 327

In Fig. 11 the resulting charm spectra are com- light heavy light
pared to experimental data for leptoproduction and r 9 9

e+e--annihilation (in the latter case we also include

0 0 0 0 @ 9 0 0 0
charm from bottom decays). The experimental data
favour a hard fragmentation spectrum but the actual
difference in <z) between the standard Lund and i i , i , i )

this model is not all that large. Only with bottom rapidity
are we in the asymptotic region where (50) applies. Fig. 14. When all the a-parameters are equal, the rapidity differ-
Since B mesons are difficult to observe, one may ence between two heavy particles is larger than that between one
look at muons (and electrons?) coming from semi- heavy and one light, which in turn is larger than that between
two adjacent light particles
leptonic decays, either by the one-particle muon
spectra, with contributions both from charm and
bottom, Fig. 12, or by same side #+/~ -pairs, domi-
nated by bottom, Fig. 13. Mark II has looked at the standard Lund model arises from the fact that the
one particle muon spectrum with contributions from field must not be too short for the energy to suffice
both bottom and charm decay, and conclude that for the quark pair creation [8]. As mentioned above
the B meson spectrum indeed is harder than the D this has an effect similar to the F a factor in (18),
meson one [17]. suppressing small F-values (and large z-values). As
We have here assumed that all the a-parameters this effect becomes stronger for larger (transverse)
are equal. The finite field length correction in the quark masses., it is tempting to assume that the a-
parameter should be larger for larger quark masses.
This would make the fragmentation of a u-quark
dN (bb "-* #*~- ) into a kaon or a proton (by production of a more
dlxFI massive Y-quark or diquark) softer and the fragmen-
tation of a c- or b-quark into a D- or B-meson even
harder than the result presented above ((50) and
Figs. 11-13). Although this cannot be excluded it has
presently no support in the experimental data.
0.08 "\
x
\
We note that if a pair of heavy quarks would be
\ produced in the field, the distribution in rapidity will
\
\ be as shown in Fig. 14 in case all a-parameters are
equal. The rapidity difference between the two heavy
0.06 particles is larger than that between one heavy and
one light, which in turn is larger than that between
two adjacent light particles. If the a-value for the
heavy quark is larger, the two heavy mesons will
come closer in rapidity and farther away from the
0.04 light particles. Similarly a smaller a-value for the
heavy quark would push the heavy mesons away
from each other. This would increase the invariant
mass of the heavy meson pair and suppress the
number of produced light particles.
0.02
\
x\
5. Comparisons with the Artru-Mennessier Model

0.00 , == IXFI= 2iPu+ PL2I The symmetric fragmentation scheme presented here
0.0 0.5 1.0 W has certain similarities with the model proposed by
Fig. 13. The total Ix~l= 2 IPL1+PLal/W of opposite sign, same side Artru and Mennessier (A-M'model) [4]. This mod-
ninon pairs in bb events ( W - 3 2 G e V ) . Full line the symmetric el was conceived very early before the use of iter-
model, dashed line standard Lund. Contributions from cE events ative fragmentation models became popular. The
are small. Note the vertical scale; only one bb-event in 40 contain breakup properties of the string in the A-M model
a same side # + # pair (if electrons also m a y be identified, this
is based on classical probability arguments. It is
rate can be quadrupled). With only one event in 11 being bb the
rate is too small for detailed studies at present energies, but at assumed that there is a constant probability P0 per
LEP it will be of interest unit time and length that the string breaks by the
328 B. Andersson et al.: A General Model for Jet Fragmentation

production of a q q-pair. No further breaks can oc-

cur in the forward lightcone with respect to a given
Poam2 1 exp
2~r 2 z+
/Poam2logz+)x
vertex point, since the field is vanishing there. This
approach then leads to an exponential decay law, i.e. Po6m 2 I'gam2 \
__ Z + ~-~K2 1). (54)
2K 2
H(F)=be -br (51)
We may now let 6rn2~O if at the same time Po ~ ~ .
with b=P0/2• 2. Note that for Po.&n2/2~c2=l we recover f ( z ) = l .
With this basic probability argument, the dif- This is the result from the original Lund Model
ferent vertex points will appear essentially indepen- (when transverse momenta and quark masses are
dent of each other so that the mass of the final state neglected) which does not give a symmetric fragmen-
hadrons will not be fixed numbers but instead be tation process.
given by a mass distribution. If one considers the
breakup closest to the q0 end, i.e. the one with the
smallest x - = t - x coordinate, one obtains
6. A Possible Interpretation of the Model
dP=e-bS=-bsdz+dz = e b"2
~+ dZ +.bdm2 (52)
Z+ There is another interesting and rather suggestive
interpretation of the fragmentation distributions in
=:~ (29) and (30). The expression in the exponent does,
as explained in connection with (28), have the in-
dP
--jbe~ b,,2
~+ dz+ =b ~ e-" dX=bEl(bm2). (53) terpretation of the area of a space-time surface. At
din2 o z.~ b,~ x this point it is useful to consider the detailed
Fig. 15a in which the string system breaks up at a
This expression is singular for m 2 ~ 0 , corre- set of vertices labelled V1.... , V,_ 1. The original pair
sponding to a large probability for the string to qoqo starts out at the space-time point O and in
break into very small pieces, unless some kind of Q C D language they correspond to a colour singlet.
lower cutoff is imposed on the mass spectrum. Such At the point 01 , q0 (which has turned around at
a cutoff is difficult to introduce in a consistent way
if one wants to keep the classical probability in-
terpretation mentioned above. Nevertheless it is pos- On 01
sible to interpret the mass spectrum in a way similar Do
to the well-known Hagedorn model (which however
has a linear dependence upon the masses in the 0il- I
exponent [6~) as corresponding to resonances etc
which may decay in later steps to the ordinary had-
rons. The A-M model which is explicitly left-right
c"-"<~ y/~
symmetric, was conceived independently of any iter-
ative framework, but can actually be given an iter-
ative structure using (52). It would thus be natural
to assume that the model presented here is a
specialization of the Artru-Mennessier model to a
a c, 0"~D"[~
physical mass spectrum, b u t this is not true. A
classical probability argument would not in the case
of fixed masses lead to an exponential decay law. To
see that we start to assume that inside a finite width
6m 2 around the mass hyperbolae we will allow for
the breakup just as in the A-M model. Then we
calculate the z+ value for which the first breakup
occurs if the production probability is a constant per
unit length and time within the band of allowed
breakup points (6z_ = c5m2-t
Z+SI b 0
Fig. 15. a "Colour coherence regions" in a gauge where A~_=0
Po and all gluons are emitted along the positive lightcone, b "Colour
f(z+)=~lc2sSz_'exp(-is6z'_2~dz'+ )
z+ coherence regions" when gluons can be emitted randomly
B. Andersson et al.: A General Model for Jet Fragmentation 329

Do), and ql (produced at V0 will meet "for the first The differentials are Lorentz invariant phase
time" to form a colour singlet final state hadron. space factors :
The surface in question is given by O D o O 1C 1 where
C~ is the point along the Cio lightcone marked out in ~o~ = dy i = d z P 6(p2 _ m2). (55)
the figure.
Inside this surface there must evidently be a
" m e m o r y " of the q0cT0 colour properties as com- In the light of this interpretation the generality of
pared to the q0c71 properties and we will call it a the parameter b is evident; it corresponds to the
"colour coherence region". The fragmentation distri- area law parameter. The parameters a t are similarly
bution for the final state hadron composed of q l q2 together with the normalization constants N~r the
(produced in the vertices V1 and V2 respectively) is perimeter law parameters and the existence of dif-
similarly related to the colour coherence region ferent possible values for different flavours is rather
C 1 V 1 0 2 C 2 and so on, so that the hadron q . - l q o is
easy to understand.
related to C. i V._~O.C. with C. the turning point
Acknowledgement. We are greatly indebted to Dr. T. Sj~Sstrand for
of q0" help with the penomenological analyses in Sect. 4.
We may, however, evidently just as well consider
the fragmentation process from the antiquark end, in
which case we will have the corresponding fragmen-
tation distributions related to the colour coherence
References
surfaces O C , O , D , 1, D , - 1 V , - I O , 1 D , - 2 etc until
D t V I O I D o. We may also split up the surface in any 1. B. Andersson, G. Gustafson, C. Peterson: Z. Phys. C - Particles
way taking steps either from the right or from the and Fields 1, 105 (1979)
left as shown in Fig. 15b. The area of the total 2. Y. Nambu: Lectures at the Copenhagen Symposium (1970)
3. X. Artru: Universit6 de Paris-Sud preprint L P T H E 78/25
surface is always equal to A, in (28).
4. X. Artru, G. Mennessier: Nucl. Phys. B70, 93 (1974)
This freedom of considering the process "in any 5. R.D. Field, R.P. Feynman: Nucl. Phys. B136, 1 (1978)
direction" corresponds in a suggestive way to the 6. R. Hagedorn: CERN Rep. 71-12
gauge freedom in 1 +l-dimensional QCD where we 7. B. Andersson, G. Gustafson: Z. Phys. C - Particles and Fields
may e.g. chose the lightcone gauge A ~_= 0 (with a 3, 223 (1980)
8. B. Andersson, G. Gustafson, T. Sj/Sstrand: Z. Phys. C - Par-
colour index) corresponding to "emission of gluons" ticles and Fields 6, 235 (1980)
only along the positive lightcone etc. When we con- 9. B. Andersson, G. Gustafson, G. Ingelman, T. SjSstrand: Phys.
sider the results for producing an n-particle cluster Rep. (in press)
with total mass g in accordance with (29), we note 10. T. SjSstrand: Comput. Phys. Commun. 27, 243 (1982)
11. A. Cosher, H. Neuberger, S. Nussinov: Phys. Rev. D20, 179
that the probability is governed by an area law [14], (1979)
corresponding to e x p ( - b A ) and a product of factors 12. J.D. Bjorken: Phys. Rev. D17, 171 (1978)
~J+'-"J corresponding to "the distances" travelled 13. M.G. Bowler: Z. Phys. C - Particles and Fields 11, 169 (1981)
along the lightcones of a q-particle (positive power) 14. K.G. Wilson: Phys. Rev. D10, 2445 (1974)
and a q-particle which, moving along the opposite 15. CDHS Collab. H. Abramowicz et al.: Z. Phys. C - Particles
and Fields 15, 19 (1982); S.M. Errede: RX-966 (Ohio State)
lightcone, should and does exhibit a negative power. 1981, Ph.D. Thesis, unpublished
The two powers 3]~ and 3,-"" correspond similarly to 16. Mark II Collab. J.M. Yelton et al.: Phys. Rev. Lett. 49, 430
the motion of the original pair along DoOt and (1982); D. Lfike: In ed. P. Petiau, M. Porneuf: 21st Int. Conf.
C , O , respectively. Evidently the whole setup is very on High Energy Physics. J. Phys. 43, C3-67 (1982); J. Kirkby,
ibid, C3-67; CLEO Collab. C. Bebek et al. Phys. Rev. Lett.
suggestive of the expected expression for Wilson
49, 610 (1982)
loop integrals with the well-known dominating area 17. Mark II Collab. G. Trilling: Contr. to 21st Int. Conf. on High
law and the corresponding perimeter dependence. Energy Physics in Paris (1982)