Z. Phys.

C 69, 475480 (1996)

Collinear asymptotic dynamics for massive particles
Regge behaviour and eikonalization
J. Boguszynski, H. D. Dahmen, R. Kretschmer, L. Lukaszuk, L. Szymanowski
Siegen University, Fachbereich Physik, Siegen, Germany, and Soltan Institute for Nuclear Studies, Warsaw, Poland
Received: 19 December 1994
Abstract. The dynamics of massive particles in the col-
linear high momentum regime is investigated. Methods
hitherto exploiting large time asymptotic Hamiltonians in
the Dirac picture for the treatment of infrared divergences
are adapted to the collinear asymptotic dynamics. The
essential role of time ordering of the dressing operators is
brought out.
1 Introduction
For the treatment of infrared divergencies and the prob-
lem of the Coulomb phase it has been known for a long
time that perturbative contributions up to innite order
have to be taken into account. Kinoshita [1], Lee and
Nauenberg [2] showed that maintaining unitarity in the
approximation is the key for solving this problem.
Khulish and Faddeev [3] pointed out that both phe-
nomena are dominated by the asymptotic dynamics of
QED. They separated out that part of the Hamiltonian
which is essential for large time scales.
The dynamics at large times is dominated by the
interaction involving the long wave length photons. In the
Dirac picture the temporal variation of their contribution
to the interaction Hamiltonian is small. Actually, for the
energy of the photons going to zero the corresponding
interaction Hamiltonian density in momentum space be-
comes stationary.
Similarly, photons collinear to electrons lead to parts
in the interaction density stationary in the limit of high
momentum [4].
In the case of massive particles only in the high mo-
mentum limit the stationary regime is sustained.
Adapting the methods of asymptotic dynamics
[5, 6, 7] to the massive case we shall show that they lead
This work has been supported under the German-Polish agreement
on bilateral scientic and technological cooperation
to Regge behaviour [8, 9, 10]. It turns out that the most
important contributions are the leading logarithms. While
in the treatment of infrared dynamics time ordering of the
dressing operator did not come into play, here it will be of
essential importance.
We shall also show that the other typical high energy
approximation, i. e. eikonalization [11], can be obtained
using our methods.
In this paper for developing the appropriate adapta-
tion of our method we study the ` theory, since it is the
simplest model for which Regge behaviour has been dem-
onstrated.
2 Approximate unitarity
Calculations of bremsstrahlung processes require in a
unitary approximation the inclusion of an innity of or-
ders in the coupling constant to ensure a nite total cross
section.
One way of reaching this aim is through an approxi-
mate calculation of Dysons time evolution operator
(t, t
"
) determined by the equation
i
c
ct
(t, t
"
)"H
'
(t) (t, t
"
) (2.1)
with the well known formal solution
(t, t
"
)"Texp

!i
R
j
R

H
'
(t') dt'

, (2.2)
where H
'
(t) is the interaction Hamiltonian in the Dirac
picture.
For deniteness we exemplify our method on the
` model. The interaction Hamiltonian
H
'
"!
g
3!
j : `(x): d`x (2.3)
can be written as
H
'
"!
g
3!
j ''`(x)#'>'`(x)#3''`(x) '>' (x)
#3'' (x) '>'`(x) d`x (2.4)
by splitting the eld operator (x)"'' (x)#'>' (x)
into the creation and annihilation parts.
As in the argument of Kulish and Faddeev [3] in the
case of bremsstrahlung and Coulomb phase we select
those contributions to the interaction that are most im-
portant at large time scales. At rst we neglect those terms
of H
'
that create or annihilate three particles and get
HI
'
"H'>'
'
#H''
'
(2.5)
with
H'>'
'
"!
g
2
j '' `(x) '>' (x) d`x,
H''
'
"!
g
2
j '' (x) '>'`(x) d`x. (2.6)
or in momentum space
H'>'
'
"!
g
4
j
dkI dpJ
E(p!k)
aR (p!k)
aR(k) a(p) expi [E(k)#E(p!k)!E(p)]t (2.7)
where
dkI"
d`k
(2)`2k"
. (2.8)
Collinear asymptotic dynamics (for p<m, where p" p )
will be incorporated by restricting the k-integration to
a region D(p) in which E(k)#E(p!k)!E(p) tends to
zero for pPR. A convenient choice for D(p) is given by
the requirement
kIl
I
4o` (2.9)
where lI is the momentum vector of the second created
particle, lI"(E(p!k), p!k), and o is a scale parameter.
For o'm the region D(p) dened in this way restricts
k
]
and k
N
to an ellipsoid

k
]
!
p
`
`
a`
#
k
N2
b`
41 (2.10)
with the principal axes a and b given by
a`"
p`
4
o`!m`
o`#m`
#
o`!m`
2
and b`"
o`!m`
2
.
(2.11)
For p<o<m we have
a+
p
2
. (2.12)
The calculation of (t, t
"
) is greatly simplied if the time-
ordered exponential is suitably factorized. This is facili-
tated by a time dependent similarity transformation. For
two operators A(t), B(t), not necessary Hermitean, one has
Texp

(!i)
R
j
R

A(t' ) dt'#(!i)
R
j
R

B(t' ) dt'

"Texp

!i
R
j
R
A(t' ) dt'

Texp

!i
R
j
R

(t', t
"
) B(t' )

(t', t
"
) dt'

(2.13)
with

(t, t
"
)"Texp

!i
R
j
R

A(t' ) dt'

. (2.14)
The validity of the factorization of the T-product follows
from a factored ansatz "

for the time evolution

operator leading to the dierential equation
i
c
ct
(t, t
"
)"

(t, t
"
) B(t)

(t, t
"
) (t, t
"
) (2.15)
having again as solution a time-ordered exponential
(t, t
"
)"Texp

!i
R
j
R

(t', t
"
) B(t')

(t', t
"
) dt'

.
(2.16)
For further use we quote a formula already given by
Dyson [12]

(t', t
"
) B(t')

(t', t
"
)
"
`

L"
iL
R
j
R

dt
L
R
L
j
R

dt
L
2
R

j
R

dt

[A(t

),
[A(t
`
),
2
[A(t
L
), B(t)]
2
]. (2.17)
In our context A(t), B(t) will be identied with H'>'
'
(t),
H''
'
(t), respectively. We will use the truncated expression

"B(t)#i
R
j
R

[A(t

), B(t)] dt

#O(g`). (2.18)
Again applying (2.13) we can introduce

`
(t, t
"
)"Texp

!i
R
j
R

B(t' ) dt'

(2.19)
and

`
(t, t
"
)"Texp

R
j
R

dt
`
R

j
R

dt

[A(t

), B(t
`
)]

(2.20)
(note that the time ordering in
`
refers to t
`
). In this way
we get an approximate expression for the time evolution
operator
(t, t
"
)"

(t, t
"
)
`
(t, t
"
)
`
(t, t
"
) . (2.21)
It fullls an approximate unitarity relation of the kind
R(t, t
"
) (t, t
"
)"expO(g`). (2.22)
A similar approach has been successfully used in brems-
strahlung calculations [4]. In what follows we are going
476
to show that the above approximation produces Regge
behaviour and eikonalization of two-body amplitudes for
elastic scattering.
Let us give the explicit formula for
`
using
A(t)"H'>'
'
(t), B(t)"H''
'
(t):

`
(t, t
"
)"Texp

g`
4
R
j
VR

d"x
V
j
WR

d"y (''`(y) '>'`(x)

;A>(y!x)!4'' (x) '' (y) '>' (x)
;'>' (y) A>(x!y))

(2.23)
with
A>(x!y)"['>' (x), '' (y)]. (2.24)
In the expression for
`
we have neglected a part that
corresponds to renormalization corrections which will not
be considered here since they are not leading.
3 Regge behaviour
In the case of soft electromagnetic bremsstrahlung the
asymptotic interaction Hamiltonian is linear in the
photon eld. The photon creating part of this Hamil-
tonian corresponding to our H'>'
'
(t) commutes with itself
at dierent times, since the electromagnetic current can be
well approximated by the classical one. Thus, the time-
ordering of the exponential is superuous.
In contradistinction, H'>'
'
(t), (2.6), is a product of
''` and '>' and therefore does not commute with itself
at dierent times, so that the time-ordering in

remains
essential. A similar comment holds true for
`
.
Let us start with a suitable formula for the elastic
amplitude T

(p

#p
`
Pp
`
#p
"
):
T

"# , (3.1)
where
"!ij d"xd"y e
!ip
`
y#ip
"
x
G(x"!y")
;p
`
, out j
'
(x) j
'
(y) p

, in, (3.2)
and
"(p
`
!p
"
) . (3.3)
In (3.2) the Heisenberg current is denoted by the index
H.
The matrix element can be written as
"!i jd"xd"y e
!ip
`
y#ip
"
x
G(x"!y")
;p
`
R (x", !R) j (x) (x", y") j ( y)
;(y", !R) p

,. (3.4)
In what follows we shall consider small transfers
t"(p

!p
`
)` at high energies. It will be convenient to
work in the laboratory frame of the particle with mo-
mentum p
`
; p

, p
`
are chosen to be large momenta. De-
tailed calculations will be made in the interaction picture
and we expect that the leading contribution to the reg-
geized amplitude is real and comes from the quasi-stable
multiparticle congurations. Therefore it is reasonable to
replace
(x", y")P1. (3.5)
Inserting a complete set of states between j(x) and j(y)
yields
"(2)"o(p

#p
`
!p
`
!p
"
) M
`
(3.6)
with
M
`
"(2)`
`

L
1
n!

L
, (3.7)
where

L
"j dkI

2
dkI
L
o(p

#p
`
!L
G
k
G
)
p"

#p"
`
!L
G
k"
G
#ic
G*
L
(p
`
) G
L
(p

)
(3.8)
and
G
L
(p)"k

,
2
, k
L
j(0) (0,!R) p,. (3.9)
As we are working in the laboratory frame of the particle
with momentum p
`
, the current j (0) is connected with the
interaction of this slow particle. Since we expect that
(0, !R) alone accounts for the production of energetic
particles, we replace
j (0)"g

'' (0) '>' (0)#

''`(0)#'>'`(0)
2
(3.10)
by
j (0)"g'' (0)'>' (0) (3.11)
in (3.9).
Next we substitute the approximation for (0, !R),
cf. eq. (2.21). The operators
`
and
`
act trivially on the
one-particle state p,:

`
p," p,. (3.12)
So we are left with
G
L
(p)"gk

,
2
, k
L
('' (0) '>' (0))
Texp

!i
"
j
`
dt H'>'
'
(t)

p,. (3.13)
Let us remark that we can limit ourselves to the term of
order (n!1) in the time ordered expansion in (3.13).
Lower orders lead to disconnected graphs which should
not contribute to the full amplitude in the asymptotic
collinear dynamics considered here. The terms higher
than (n!1) necessarily lead to vertex and/or mass correc-
tions neglected throughout this paper.
Therefore
G
L
(p)"g(!i)L
"
j
`
dt

2
dt
L
G(t

,
2
, t
L
)
;0a(k

) 2a( k
L
) '' (0) '>' (0) H'>'
'
(t

)
;2H'>'
'
(t
L
) p,, (3.14)
where the symbol G(t

,
2
, t
L
) denotes
G(t

,
2
, t
L
)"G(t

!t
`
)2G(t
L`
!t
L
) . (3.15)
477
It is useful to rewrite G
L
as
G
L
(p)"
L

P
gL
P
(p) (3.16)
where
gL
P
"g(!i)L
"
j
`
dt

2dt
L
G(t

,
2
, t
L
)
;0 a(k

) 2a(k
P
) a(k
P>
)2a(k
L
) '>' (0) H'>'
'
(t

)
;2H'>'
'
(t
L
) p,. (3.17)
In evaluating the matrix elements in (3.17) we have to
commute the a(k
G
) with H
'
(t
H
). Thus we shall encounter the
following commutators of a(k):
D(t, k)"!i [a(k), H'>'
'
(t)]"ig j
VR
d`x
;'' (x) '>' (x) eIV , (3.18)
d(t, k, k' )"[a( k), D(t, k')]"ig j
VR
d`x '>' (x) e 'I>I'V ,
(3.19)
[a(k'' ), d(t, k, k')]"0. (3.20)
The operator d is linear in '>'. Therefore we should
consider the commutator of d with H'>'
'
again. Since
d depends on the sum of two intermediate momenta it
leads to vertex and mass corrections whichas already
stated aboveare beyond the scope of our work. As
a consequence, d can be treated as a commuting quantity
and it is easy to see that for this reason it does not
contribute at all. In this way we obtain
(!i)L0 a(k

) 2a(k
P
) a(k
P>
) 2a(k
L
) '>'(0)
H
'
(t

)2H
'
(t
L
) p,
"

P
0 '>' (0)D(t

, k
L''
)2D(t
L
, k
L'L'
) p,, (3.21)
where the summation runs over all permutations of the
numbers 1, 2,
2
, r!1, r#1,
2
, n.
Applying consecutively
D(t, k) p,"ig
e
i(E(p!k)#k"!p")t
2E(p!k)
p!k,, (3.22)
and integrating, we arrive at
gL
P
(p)"gL

P
1
2E(p!k
L'L'
)
;
1
E(p!k
L'L'
)#k"
L'L'
!p"
;
1
2E(p!k
L'L'
!k
L'L'
)
;
1
E(p!k
L'L'
!k
L'L'
)#k"
L'L'
#k"
L'L'
!p"
;2
;
1
2E(p!L
1, iOr
k
L'G'
)
;
1
E(p!L
1, iOr
k
L'G'
)#L
1, iOr
k"
L'G'
!p"
. (3.23)
Having this form of gL
P
, we now look for the kinematical
region that gives a dominant contribution. Note that for
massive particles the denominators in gL
P
never vanish.
We shall calculate the dominant contributions in the
leading logarithmic approximation, so we consider xed
transverse momenta and
k
]
L'Q'
<(k
N
`
L'Q'
#m` (3.24)
q
]
P Q
<(q
N
`
P Q
#m` (3.25)
for
s"1, 2,
2
, r!1, r#1,
2
, n (3.26)
and with
q
P Q
"p!
L

i"s, i Or
k
L'G'
. (3.27)
Then one can show that the terms
1
2E(q
P Q
) [E(q
P Q
)!q"
P Q
]
(3.28)
in (3.23) are dierent from zero for
k
]
L'Q'
<q
]
P Q
(3.29)
or
k
]
L'Q'
"O(q
]
P Q
) . (3.30)
The second possibility does not contribute in leading
logarithmic order, so we are left with the conditions (3.29).
These, together with momentum conservation (see (3.8))
k
]
P
"q
]
P
(3.31)
lead to strong ordering of the ks
p
]
9k
]
L'L'
<k
]
L'L'
<2<k
]
L'P>'
<k
]
L'P'
<2<k
]
L''
<k
]
P
<m. (3.32)
It means that in the expression (3.8) (with p
N

"0) only the

same orderings in both G
L
enter, i.e.

L
"g`L j dkI

2dkI
L
o(p

!L

k
Q
)
p"

#p"
`
!L
G
k"
G
L

P
f L
P
()
;
1
m`#(p
N
`
!k
N
L'L'
)`
;
1
m`#k
N
`
L'L'
2
;
1
m`#(p
N
`
!L
1, sOr
k
N
L'Q'
)`
;
1
m`#(L
1, sOr
k
N
L'Q'
)`
(3.33)
where f L
P
() is unity in the region dened by (3.32) and zero
otherwise.
Integrating over transverse momenta we arrive at

L
"
g`
2(2)`
(g`K(p
N
`
`
))L
L

IL
P
() (3.34)
where
IL
P
()"j
dk
]

2dk
]
L
k"

2k"
L
;
f L
P
() o(p
]

!L

k
]
G
)
p"

#p"
`
!L

k"
G
(3.35)
478
and
K"
1
2(2)`
j
d`k
N
(m`#k
N2
)(m`#(p
N
`
!k
N
)`)
. (3.36)
The longitudinal integration yields the same result for
each IL
P
:
IL
P
()"
2
s
(lns)L
(n!1) !
1#O

1
lns
(3.37)
where s+2mp
]

, so that

L
"n!
g`
(2)`
(g`Klns)L
(n!1)!
1
s
. (3.38)
Then using (3.7) results in
M
`
"
g`
s
`

L
(g`Klns)L
(n!1) !
"g`s
!1#g`K
, (3.39)
i. e. we have obtained Regge behaviour [9]. Of course this
result agrees with that obtained by use of the standard
method based on summation of the ladder Feynman
graphs in the leading logarithm approximation [8, 9].
4 Eikonalization
In this chapter we apply the factorization formula (2.21) to
derive the familiar eikonal form of the scattering ampli-
tude.
We start with
p
`
, p
"
S p

, p
`
,,
S"(R, !R)
"

(R, !R)
`
(R, !R)
`
(R, !R) . (4.1)
In this matrix element the part

`
can be replaced by
unity; other terms of

`
either give a one-particle
s-channel contribution that can be neglected, or vanish
identically, because
`
conserves the particle number.
The rst term in the exponent of
`
, (2.23), is also non-
leading for large c.m.s. energy.
In this way we arrive at
p
`
, p
"
S p

, p
`
,"p
`
, p
"

Texp

`
j
`
r(t) dt#O(g`)

, p
`
,, (4.2)
where
r(t)"!g`
R
j
`
dy" j
VR
d`x d`y '' (x)'' (y)'>' (x)
'>' (y) A>(x!y) . (4.3)
In fact, for large energies the time ordering in (4.2) is
irrelevant. To see this, consider any two-particle state
k

, k
`
, in the center of mass system of k

and k
`
. Then
we have
r(t) r(t') k

, !k

,"!
g"
4
jdqJ dqJ' aR (k

#q#q')
aR(!k

!q!q') 0,
;
e
!2i (k"

!E(k

#q)) t
(E(k

#q))`[k"

!q"!E(k

#q)#ic ]
;
e
!2i (E(k

#q)!E(k

#q#q')) t'
(E(k

#q#q' ))`[E(k

#q)!q'"!E(k

#q#q' )#ic]
.
(4.4)
In the limit k"

PR, eq. (4.4) reduces to

r(t) r(t') k

, !k

,
k"

PR
"!
g"
4(k"

)"
j dqJ dqJ'
;aR(k

#q#q' ) aR(!k

!q!q' ) 0,
;
e`O
]
R
q"#q
]
!ic
;
e`O
]
R
q'"#q'
]
!ic
, (4.5)
from which it is evident that
[r(t), r(t')] k

, !k

,
k"

PR
"0. (4.6)
As a consequence, (4.2) can be written as
p
`
, p
"
S p

, p
`
,"p
`
, p
"
exp

`
j
`
r(t) dt

, p
`
,
(4.7)
with
`
j
`
r (t) dt p

, p
`
,"i
g`
8(2)`(p"

)`
j
d`q
N
q
N2
#m`
;aR (p

#q
N
) aR(p
`
!q
N
) 0,. (4.8)
From (4.8) it is straightforward to derive the well-known
eikonal formula
M

"2isjd`b e
ip
N
`
b

exp

ig`
4s
K
"
(m b)

!1

. (4.9)
5 Concluding remarks
We have shown how the collinear asymptotic dynamics
described by the Hamiltonian eq. (2.5) restricted to the
kinematical region D, eq. (2.9), leads to both Regge beha-
viour and eikonalization of two-particle scattering. While
Regge behaviour is dominated by three-particle collinear
dynamics, eikonalization corresponds to that kinematical
region of the asymptotic Hamiltonian where only two
particles are collinear. In contradistinction to infrared
dynamics [3, 6], the time ordering of the dressing operator
is essential for three-particle collinear dynamics which
leads to Regge behaviour.
Acknowledgement. Three of us (HDD, RK, LS) would like to thank
Prof. Lev Lipatov for discussions.
References
1. T. Kinoshita: J. Math. Phys. 3 (1962) 650
2. T. D. Lee, M. Nauenberg: Phys. Rev. 133 (1964) B1549
479
3. P. Kulish, L. Faddeev: Theor. Math. Fiz. 4 (1970) 153
4. H. D. Dahmen et al.: Z. Phys. C50 (1991) 75
5. S. D. Drell, D. J. Levy, T.-M. Yan: Phys. Rev. 187 (1969) 2159;
ibid. D1 (1970) 1035, 1617, 2402
6. H. D. Dahmen, B. Scholz, F. Steiner: Nucl. Phys. B202 (1982)
365
7. H. D. Dahmen, W. Krzyzanowski, M.-L. Larsen: Phys. Rev.
D33 (1986) 1726; H. D. Dahmen, T. Mannel, P. Manakos:
Z. Phys. C40 (1988) 425; H. D. Dahmen, T. Mannel, P.
Manakos: Phys. Rev. D38 (1988) 1176
8. R. J. Eden et al.: The Analytic S-matrix, Cambridge University
Press (1966)
9. P. D. Collins: An Introduction to Regge Theory and High
Energy Physics, Cambridge University Press (1977)
10. J. Koplik, A. H. Mueller: Phys. Rev. D12 (1975) 3638
11. H. M. Fried: Functional Methods and Models in Quantum
Field Theory, The Massachusetts Institute of Technology
(1972)
12. F. J. Dyson: Phys. Rev. 75 (1949) 486 and 1736
.
480