(n I J,(O) IP> n =-
(g!A”-
9LL4” 2 W,(q2, P * 4) 4 )
+ (P, - qu Y)(P”
- q” y,
W2(q29P . a>
U-1)
can be decomposed as the sum of three pieces, the connected part C, the semidisconnected part D, and the “pair” part P, according to the nature of the intermediate states in (1. I), see Fig. 1. From the stability of the hadron target we easily derive that C can contribute only for -q2/2p * q < 1, D only for -q2/2p * q < 0 and P for -q2/2q - p < -1. C is the contribution relevant to the inelastic scattering and as we have just seen it coincides in the physical region of this channel, i.e., 0 < -(q2/2q * p) < 1 with the full matrix element of the current commutator [4]. The scaling functions F,(w) and F2(w) for the scattering channel are defined by
where in the scattering channel we define w = -q2/2v with v = p * q.
FIG. 1. Decomposition of the forward matrix element of the connected commutator connected, semidisconnected, and pair contributions.
into
DEEP
ELECTROPRODUCTION
On the other hand the correlation
AND
ANNIHILATION
function wti, for the annihilation
Wuy(q,p) = &(27~)~ S(q -p -p&O
channel,
j J,(O)! n. pXn,p / J,(O)] O>,
(1.3)
is given only by the pair contributions to the commutator in (1 .l). The scaling function in the annihilation channel are analogously defined by F,(w) = hil
F&p,
v),
F,(w) = lip-v) 1
iv2(q’, v),
(1.4)
where now w = -q2/2v, with v = -q ‘p, p being the four momentum of the detected outgoing hadron; the physical region for the annihilation channel is for l
F,(w) = &F2(w),
(1.5)
where the upper and lower sign refer to Fermi or Bose statistics of the target hadron. Our aim is to analyze whether the relations (1.5) hold (of course, for the analytic continuations); in other words, whether the knowledge of F<(w) in the physical region for scattering 0 < w ==c1 enables us to obtain through analytic continuation F,(o) in the physical region for annihilation. We know that only superrenormalizable theories exhibit scaling properties in perturbation theory; in simply renormalizable theories the scaling behavior does not hold in perturbation theory. We shall examine in detail within a superrenormalizable theory the properties of the scaling functions Fi and F, . Our analysis will consist of two parts; first, we shall determine in various specific exampIes the position of the singularities of F(w) in the w-plane to find out in which cases an analytic continuation is possible for w > 1; then we shall examine whether such analytic continuation coincides with F(w). Particular attention will be devoted to the occurrence of real cuts on the physical sheet for w > 1, which spoil the analyticity of F(w) along the positive real axis and invalidate the analytic continuation relation (1.5). It is well known that the scaling function of the box graph (we refer for simplicity to the case of a scalar photon) with suitable stability conditions for the propagated particles (see also Section 4) has a very simple analytic structure: it
4
GATTO,
MENOTTI,
AND VENDRAMIN
allows for an analytic continuation for w > 1 and this continuation gives the annihilation scaling function [5]. The scaling function for the double box graph has a much more complex structure. After setting down the general formalism in Sections 2 and 3, we shall give in Section 4 a complete analysis of the singularities of the scaling function for the double box graph. This analysis is carried through by reducing the scaling function to a double integral over Feynman parameters and giving a complete Eden-Hadamard type analysis of the singularities [6]. Both Landau and nonLandau singularities are examined. As long as the propagated particles are stable and no zero-mass particles appear in the theory the continuation of F(o) above w = 1 is possible; the calculation of Section 5, showing the coincidence of the analytically continued F(o) with -P(w), gives a direct check of the recently proposed argument by Landshoff and Polkinghorne and of the result of Drell et al. on the validity of the relation F(w) = ---F(U) for ladder graphs in which the propagated particles are stable [S]. In Section 6 we write down the scaling function for a general double box diagram in terms of the DGS spectral function; the analyticity along the positive real axis is easily evidenced as long as the stability conditions are satisfied. As the propagated particles become unstable, cuts arise on the physical sheet with real branch points for w > 1. The occurrence of real branch points for w > 1 poses the problem whether there is any simple connection between the two scaling functions F(o) and F(w) in the general case. To this propose we examine in Section 7 a model (a box diagram with dressed propagators) which possesses such real branch points for w > 1 and which is simple enough to allow for an explicit evaluation of the two scaling functions. No simple connection emerges in this case [S, 91. The breakdown of the analytic continuation rule for the scaling function is easily related to the nonvanishing double discontinuity in the parton mass and in the parton-proton energy of the parton-proton amplitude. In Section 8 we summarize the conclusions and we give a general outlook on the relations between Bjorken scaling in the scattering and the annihilation channel.
2. GENERAL EXPRESSION FOR THE SCALING FUNCTION The dominant contribution in the deep inelastic region is expected to arise from diagrams with contiguous electromagnetic vertices [ll, 121, whose general structure is shown in Fig. 2. The amplitude A describes virtual scattering on the target as illustrated in Fig. 3. We consider the forward (t = 0), virtual Compton scattering of scalar photons, interacting with scalar particles [13]. The scaling function F(U) is given by the
DEEP ELECTROPRODUCTION
AND ANNIHILATION
FIG. 2. Forward virtual Compton scattering for contiguouseleckomagnetic are indicated in parenthesis.
vertices. Masses
(fi) FIG.
3. Forward
virtual scattering amplitude.
Fourier transform of the weight function of the E(XJ 6(x2) singularity on the light cone [ 141, so that, with reference to Figs. 2 and 3 we can write F(w)
=
-
q
j eiMWS9(77x0)
dx, .
(2.1)
In Eq. (2.1) p2 = M” is the target mass, u = --(q2/2v),
b’=p.cl,
and we have introduced a four-vector n = (1; 1, 0,O).
G.2)
6
GA-I-TO,
MENOTTI,
AND
VENDRAMIN
In Eq. (2.3), A(k2, k * p) is the virtual meson-meson amplitude of Fig. 3, m, is the mass of the boson field propagated in Fig. 2. It is convenient to introduce, besides n s (1; I, 0, 0), also 5 = (- 1; 1, 0,O) and to decompose the virtual particle momentum k as K=pn+oE+&,
(2.4)
where the two-dimensional vector & gives the transverse component of k. From Eqs. (2.1), (2.3), and (2.4) exchanging the order of integration we obtain
F(f.0)= -M-g$J
dp da d2& (-4pa
-
ET2 - mo2 + ie)-’
0
* A(-4pa
&.2;
-
M(~
-
a))
. j- dxo
,9(2~+~~h
(2.5) dR (-R * A(-
R + 2Mwp;
@&J
+ 2Mwp
- mo2 + ic)-l
+ Mp),
where we have put RT2 = R. The meson-meson amplitude A(k2, k * p) can be expressed by a Nambu representation. For a graph with N internal lines and I independent loops, the Nambu representation [6] is CN-21-2
A(k2,k.p)=G~~‘~duiS(~vi--)
[Flk2 - 2F2k . p - F3 4 icC]N-21 ‘(2e6)
l=i
where C, and Fj (j = 1,2,3) are homogeneous functions of the Feynman parameters vi (i = 1 *** N) of degree 2 and I + 1, respectively, and G is a constant. The representation in Eq. (2.6), for k2 = -R + 2Mcup and k * p = &%12~ + Mp, can be inserted into Eq. (2.5), and the two denominators parametrized a la Feynman using 1 da an-l =-----*1 (2.7) llX”Y o [ax + (1 - 4 yP+l
s
We obtain, after exchange of the order of integration, F(w) = (- l)N +T~MG &
Jo1le dui S ( :
vi
-
l=i
- 1) + 11 + -~e~dp~owdR{N4&
1)
CN-2’~2
* 1’
(N
-
24
da
UN-21-1
0
2Mp[a(F2 - 04
+ a> -
+ a(M2wF2 + F3 - mo2) + mo2 - ie(aC + 1 - a)}-N+2z+1.
~1 (2.8)
DEEP
ELECTROPRODUCTION
AND
7
ANNIHILATION
After integration over R (note that the coefficient multiplying and over p, by the use of the identity +a, I -*
dp (xp + I’ -
= $gy
k)-”
R is nonnegative)
6(x),
12.9)
we find F(w) =
(-1)N 57% a ~ 1’ fi dt,i 6 (ii pi - I) CN-2’ -? iv - 21 - I amo2 o l=i da ‘+22-l
8 a-
- 1F, + ~(1 - FJ-l
F2 + w’;; - FJ 1 I) -- 11-l
[a(F, -
(2.10)
- [a(M2wF, + F3 - moz) + m02]-N+l’-‘.
Finally we perform the integration
on the variable a and obtain the result
F(w)= (-1)NT3iG ~jlfidci$+ N - 21 - 1 iimo2 o l=i ’ (F2 2 ,?Fl)
)
I)? l=i
sgW2 + 4
2
’
iEA-t
0’;; - Fl) 1
- FJI N-21-1
9
w(M2wF2 + FJ”+
(2.1 I)
mo2(F2- wF,) I
The support properties of F(w) are implicit in the &functions in Eq. (2.11). We have for
F, + ~(1 - Fl) > 0,
0 < w :< (F,/F,),
for
F2 + ~(1 - FJ < 0,
(F,/F,) < w < 0.
(2.12)
From the structure of the functions Fl and F2 in the Nambu representation, as can be deduced for instance from the set of graphical rules by Eden et al. [6], we have 1F, j < F1 , and, thus, -1 ,cw < 1.
(2.13)
In particular for “direct” graphs, i.e., for A given by any graph which can be drawn in a plane without either internal or external lines crossing, the external lines being attached round the diagram in the order of Fig. 3, we have F2 > 0 and correspondingly 0 < w < (F,I&)
< 1.
(2.14)
8
GATTO, MENOTTI, AND VENDRAMIN
3. SPECIALIZATION TO LADDER GRAPHS We shall specialize the general results obtained in the previous section to a ladder graph. We take the contribution to A(/?, k . p) from the ladder in Fig. 4 with n + 1 horizontal lines.
FIG.
4. Forward virtual scattering amplitude for ladder graph.
FIG. 5. Graph for the contracted amplitude A’ and its Feynman parameters.
DEEP
ELECTROPRODUCTION
AND
9
ANNIHILATION
We follow the technique described in the previous section taking advantage of the relation A(k2, k . p) = (fi. &)
A’W,
(3.1)
k . I’);
At(k2, k . p) is the contracted Feynman graph given in Fig. 5, where the Feynman parametrization is also illustrated. In this way we obtain for the scaling function F(w) the contribution
f’(%4 = C-1)”Gn(l@-)&a .
F'"' 2
_
Jb’ da, fi dcxi d/c?+6 ( a0 T- i
(JL~ + PJ -
I)
1 ==i
l=i
,.&Z'
'
C(n,F(")
8(w) d(Fvjn' - wFin))
2
- [o(M2wF,'"' + F;"') + mo2(F,(n' - cd+')]-'.
(3.2)
CY, Fp', Fp', and Fp' are given through the formulas [ 151 C'"'(~o
7 % =
I...,
011 ...
%a a,
+
; p1 (010
,.*., +
13,) pl,
C'-ycd,
...
2,
; p,
...
/In)
+ LYJ~~C(~-~)(~Y~ ,..., a, ; fl, ,..., /2in) - 1.1 i- x1 ‘.. cd,~,/3,
(3.3a)
and cya:, F%o
) a, 7 8, = a0 + ml + p1 ,
, '111,..., 2, ; p, ,...' Pn) = no@'(ao = 0: '2' . .. . . fY, : /3, I.' pn>,
(3.3b) (3.3c)
(3.3e)
(3.3f)
GATTO, MENOTTI, AND VENDRAMIN
10
The expression for G:“’ in Eq. (3.3e) is obtained from Fp) substituting according t0 0l.j f--) CY,-j (j = O,.**, n) and pj t) Pn+l-j (j = I ,..., n). The support property for direct graphs described at the end of Section 2 is evident since Fp’ > 0 and (Fp)/Fp)) = (q, m-ecx,jo1,,..* CL, + .**) ,< 1. 4. SINGULARITIES OF THE SCALING FUNCTION IN THE U-PLANE 4.1. Simple Box Diagram
Before examining the analytic properties of the double box graph we shall briefly recall the singularity structure of the scaling function for the simple box, as these are connected to the more complicated situation of the double box. The complete scaling function for the simple box diagram as given in Fig. 6 (with addition of the crossed diagram) can be calculated explicity [5] F(w)
=
&
+
j@)
fql
+
@l
0)
4-w)
-
w)
u2
_
w(2
5;,~2jj
+
1
1+w
02 + w(2 - (p2/M2))
+ 1 I*
(4.1)
The function F(w) in Eq. (4.1) is finite and continuous for -1 < w < 1 and vanishes linearly at both extremes of such interval. The analytic function F(U), for complex W, (4.2) S(w)= j&5 oJ2 - w(2 1-W - (p2/M2)) + 1 coincides with F(u) for 0 < w < 1.
FIG.
6. Simple box diagram for virtual Compton scattering.
DEEP
Its singularities
ELECTROPRODUCTION
AND
11
ANNIHILATlON
are two simple poles at
cc(i) = 9{2 - (p2/M2) & i(l*/M)(4
- (p2/M2))‘,‘2:.
(4.3)
For p2 = 0 the two singularities collide at w = 1. As a preliminary example [5] of more general situations the propagator for the exchanged boson (of mass CL)in Fig. 6 can be replaced by a dressed propagator obtaining for ,F(w) p(p2) d$ -- . w* - ~(2 - (p2/M2)) -!- 1
(4.4)
The cut structure of 9(w) as given by Eq. (4.4) is shown in Fig. 7. The end points C and C’ are determined by the lowest mass in the spectrum of the exchanged particle. Only when such lowest mass vanishes, C and C’ both end at w = j-1. Excluding such a possibility F(w) can be continued from the “scattering” interval 06 P w < 1, to the region w > 1. By direct calculation we verify [5] that such analytic continuation gives the scaling function F(w) for annihilation, defined for w > 1, through the relation F(w) = --F(u). This feature is related to the stability of the target hadron and of the particles appearing in the vertical propagators.
FIG. 7. Singularities of the scaling function boson replaced by a dressed propagator.
F(W) for the simple box with the exchanged
4.2. Double Box Diagram
We shall now make use of the formalism developed in the preceding sections to study the analyticity properties in w of the scaling function for the double box diagram. We refer to the diagram in Fig. 4 for the virtual scattering amplitude A(k2, li . p) taken for n = 1. To simplify the discussion at this stage we put nz = m, = 1, and at the end we shall also specialize to nz, = m. From Eqs. (3.2) and (3.3) we obtain for such a diagram
12
GATTO, MENOTTI,
After integrating
AND VENDRAMIN
over /I1 we can extract the analytic function S(w) in the form
Red,
Redo
Im dt
FIG. 8. Undistorted
region for the integral in Eq. (4.6) giving the scaling function.
The undistorted integration region is shown in Fig. 8. To study the singularities of S(w) as given in Eq. (4.6) we shall employ the usual Hadamard-Eden procedure and look for the end-point singularities and pinch-singularities. We put m, = 1 and write down the expressions for the singularity surfaces S’l’ E a1 = 0, S2’ = a&l
(4.7a)
- a>2 + w[(aO + 01~)~- (01~+ 013(2 - p2) + l] = 0, (4.7b)
and the equations specifying the boundaries of the integration
region
,z:‘l’ Es a0 = 0,
2’2’ = 010+ a1 - 1 = 0, #z(3) = wolo + q - w = 0. The singularities
(4.8a) (4.8b) (4.8c)
are given by the solutions of the set of equations &S(*) = 0 (i = 1, 2), j&‘p’ = 0 (j = 122, 3), (4.9)
DEEP
ELECTROPRODUCTION
AND
ANNIHILATIOK
13
The following conclusions are obtained from the discussion of Eqs. (4.9). The study of end-point singularities shows that w=o
(4.10)
and cG(+)= 1 - ($/2)
& +/2)(4
(4.11)
- $)ll’L
are branch-points [16]. The singularity at w = 0 also obtains in other ways (e.g. as a pinch between S(r) and P)). Obviously, such singularities lie on the physical sheet of the analytic function F(w). From pinches between SQ’ and Dz) we obtain the branch points (4.12)
3(&) = I - 2p2 * 2ip( 1 - $)1;?,
for 01~= UI~ = 4, lying on the boundary of the undistorted integration region. Such singularities are, thus, again on the physical sheet. Finally for Sc2) becoming locally cone-like we have the branch points U(i) = -1
+ (2 -
2
P2Y
! 2 -
2
P2 /+$
_ 4)1,2
(4.13)
for iy,, = CQ= (2 - p”)-l, that is for values external to the undistorted contour implying that WC*) do not belong to the physical sheet. In any case for 0 < p c 2 (defining the stability region for the decay p --f 2~2, with tn = 1) w(+) lie on the unit circle / w(~)I” = 1; whereas, for p > 2 they lie on the real axis and satisfy 0 -C ~3(-) < 1 and WC+)> 1. To study the singularities occuring for infinite values of the Feynman parameters we substitute in Eq. (4.6) y = %;I
p = yal
(4.14)
obtaining
rP - W(Y - l)l[y - (1 -r ml PCBCl -~ WI2 + o[(l
c p)” - y(1 + /I)(1 + m12 - $) -t y2”r,2];.2-. (4.15)
From pinch between SsPCl
and 22 1~ 4~
-w)2+w[(1
$8)2-y(1
+-)(I
+m,2-$)~+71,2]
- 1) - P we find, for y = 0, the singularities w L- 1
(4.16)
14
GATTO,
MENOTTI,
AND
VENDRAMIN
and w = 1 --2,
(4.17)
obviously on an unphysical sheet. The solution w = 1 - p2 is, however, fictitious as can be seen by explicitly performing one of the two integrations in the expression for S(w). In Eq. (4.6) we substitute x = (111- w(1 - CXo)
y=l-CXcu,--ol,
(4.18)
obtaining X
s(w) = -G1& s’-” dy(OJy + x)(f(x;w)y + g(x* 0 dxs’“-” , (0))’ 0 where f and g are defined later. Integration
over y gives
S(W) = -G1&/lPYdx&ln
[**$$-$-I.
10
(4.19)
>
(4.20) 3
In Eqs. (4.19) and (4.20) f(X;
w)
=
-X(1
+
w)
--0(--w
+
ml2 - p2),
(4.21b)
g(x; w) = x(1 - 0 - x) + /A, h(X;
(4.21a)
w) = X2 + wx(1 - ml2 + p2) + p2c0,
I(X; co) = (1 - 0 - x)(m12 - w - x - p2) + ~2.
(4.21~) (4.21d)
The Hadamard-Eden method can be applied to Eq. (4.20) before or after performing the derivation on m12, or to Eq. (4.19) and confirms the results already found, but shows the absence of the singularity at 1 - p2. The form of the numerator in Eq. (4.15) is responsible for such cancellation [18]. The cuts attached to the physical sheet branch points can be drawn by discussing the motion with 01~of the singularities in 01~of Eq. (4.6). From the end-point singularities in CQ= 1 - 01~we have 2 kd~o)
=
1 -
2aoclpM
ao)
4~
i
2010(lP- ao) (4a,(l - a01 - kw.
(4.22)
We have &h)(4)
4,) --+ 0-3
G(-) --t -co,
= &)
for
3
01~-+ 0+ (or a0 -+ l-).
(4.23)
DEEP
ELECTROPRODUCTION
AND
ANNIHILATION
I5
We note that, for 0 < p < 1, (G(+))* = ~5~) and j &(,)I” =: 1. For p > 1, ~3~.are real and satisfy -1 < G(+) < 0 and 63-j < -1. For 01~varying inside (0, 1) the cuts for t.~ < 1, run along the negative real axis and the fraction of the unit circle for which Re w < 1 - 2p2, while for TV> 1 they run along the real intervals (-co; +1 - 2~~ - 2p(p2 - l)‘/“) and (1 - 2~~ + 2p(pL2 - 1)lj2; 0). Finally from the end-point singularities at CY~= w( I -- a,,) we have &)(a,J with 0 < cuts -1 and
= 1 - (~72)
i +/2)((4/1
- 010)- PV.
(4.24)
&j(O) = &(+) ; for 0 < LX,< 1 they provide the cuts ending at &(+). For p < 2, i.e., in the stability region Lj(+) = &F-_) and / &q1)12 = 1. Their relative can be drawn as vertical lines, Re w = 1 - .$$. For p > 2, &) are real and 1 n < w(+) < 0, w(-) < -1. The cuts run along the real interval between &c_, d(-) and along the line Re o = 1 - 4~~.
\ I
II r--f-------) 41 I’
Rew
a FIG. 9a. The cuts on the w-planefor the function F(w), on the physical sheet for 0 < p < 1 (CL= exchangedparticlemass). .595/79/r-2
1<)1<2
I
b FIG. 9b.
The cuts on the w-plane for the function F(O) on the physical sheet for 1 < p < 2.
FIG. 9c.
The cuts on the w-plane for the function F(W) on the physical sheet for p > 2.
DEEP
ELECTROPRODUCTION
AND
ANNIHILATION
17
The situation is described for each case in Figs. 9a, 9b, and 9c. The relative positions of &) and ij(,) is as shown in the figures as G(+) is obtained from L;)(!) by substituting p with 2~. We note that, because of the location of the singularities on the unphysical sheets, it is in any case possibIe to deform the physical-sheet cuts such as to have them lying along the unit circle with branch points at & , and along the negative real axis. This is the same structure of singularities as shown in Fig. 7. 4.3. General Ladder Diagrams
The complete analysis of all singularities of the general ladder diagram looks like a rather hard task. We shall extract here a class of singularities which have a simple interpretation. These are the branch points analogous to the c&) = 1 - $(2p)2 $ ;(2p)((2p)2 --- 4)‘/2. which are obtained by replacing in cG(*) , p with 2~. Let us take for the amplitude A a ladder with n + 1 rungs, and consider the reduced graph obtained by contracting all vertical propagators in Fig. 5, i.e., by setting all pi equal to zero. We have (4.25) and also
(4.26)
Then in order to obtain a pinch between the singularity
surface (4.27)
and the contour (4.25), we have to impose that all derivatives
a IL _aa,IO orj)Iw2 -
w (2 - p2 ogj ajj + I]\
(k = CA..., n)
(4.28)
O=j
be equal. A solution of these equations is clearly given by “0 = ... = cd:,= (I/n + 1)
(4.29)
18
GATTO, MENOTTI, AND VENDRAMIN
and a,(;\ = 1 - &((n + 1) /A)2 & ((n + 1) /L/2)(((n + 1) /J)2 - 4)l’2.
(4.30)
Clearly these are branch points on the physical sheet. We see that, for increasing n, these branch points move along the circle of radius 1 toward the left and afterwards along the negative real axis with i;l_“’ -+ -co and 6:’ + O- as y1-+ co (see Fig. IO). M3) +a) wcty
,R------N
'\
: ---*--
I
: \
-4 w&l
w(+)
-$
----
41) q w(+) ' 40) t w (+I -$--------
zt77
\ \
\
i \
,'
-&-w t-1
I& $ A (1 w(- 1
/'C(2) -7231
w f-1
WC-1
FIG. 10. The branch points C& In’, for A given by the ladderwith n + 1 rungs,
5. THE SCALING FUNCTIONIN THE ANNIHILATION CHANNEL In the previous section we have shown that the deep inelastic scaling function F(w), corresponding to the double-box diagram admits an analytic continuation all along the positive real axis. In this section we shall outline the calculation which explicitly proves that for w > 1 F(w) = --F(w), with F(w) analytically continued above w = 1; F(w) is the scaling function for the annihilation diagram shown in Fig. 11 [20]. The result can be understood as follows: the expression given by Eqs. (2.1) and (2.3) (with A given by the box graph) coincides (as it can be explictly proved) with the Bjorken limit of the structure function given by the graph shown in Fig. 12, as calculated from the squared matrix element integrated over the phase-space of the outgoing particles. We go over from the scattering to the annihilation region by changing p = (M, 6) into (-A4, a). The possibility of having a relation of analyticcontinuation between the two regions is intimately connected to the fact that whenever the propagated particles are stable (the target particle is always assumed
DEEP
ELECTROPRODUCTION
AND
ANNIHILATION
19
stable) with respect to decay modes implicit in the graph under consideration, the propagators never become singular, either in the scattering or in the annihilation region. We shall in fact show in Section 6 that as soon as the stability conditions on the propagated particles are relaxed, cuts appear for w > 1, and the relation between the two scaling functions is modified.
9 ‘i --
q+k -----
k
FIG.
11. Annihilation diagram corresponding to the double box in the scattering channel.
FIG.
12. Inelastic scattering graph corresponding to the double box.
20
GATTO,
MENOTTI,
AND
VENDRAMIN
The structure function w( p, q) in the annihilation
. S(+‘((q + k)2 -
channel is given by
6”) S’+‘((l - k)2 - ,u”) S’+‘((~II - 1)~ (k2 - PZ,,~)~(1” - ml”)
@) . (5 1)
The ie can be omitted from the propagators which never become singular because of the stability of the propagated particles. We go over to light-cone and transverse-momentum variables putting
4 = an + t% k=pn+aii+&,
(5.2)
1 = pin + a@ + IT ,
p = -(M/2)(n
- it),
so that -q2 = 4&l, v =p.q
= -M(a-/I),
and for LX+ +co 6J = -(q2/2v) -+ -(2/3/M).
(5.3)
In this limit we have ~m+F(o)
8MG1 a = - 7 w s dp dp, da, d2i& d21r 1
7 S (-4(pl * 8 (h
-
- p) (a1 -
P -
+)
-
(27 -
ET)2 -
p2)
a1 + +)
* 6 (4 (7 + pl)(F
- al) - L2 - pa) 4---M
- PI + al>
1 * (-2Mwp
- ii r2 - mo2)” (--4p,a,
First we perform the pl- and p-integration transverse momentum I& .
- IT2 -
ml”)
-
and then integrate explicitly
(5.4)
over the
DEEP ELECTROPRODUCTION
21
AND ANNIHILATION
We have then CM/t&o
F(w) = -t?(w - 1) X2M $- &
jM 1
12
du, 4o,(4a,
1
I
2M IT2 + 2Ma, + 404Tp;M 4a, - 2M WW
L,4011’;M
+ Mb
- 2M)
- ml21
- .a2 + 2Mwp2
(4ul
_t
2M
+
2h4w’ -4a,
= B(w - 1) =q.o),
))
(5.5)
which shows explicitly the correct support property of the scaling function in the annihilation channel. Integrating over diT2, performing the substitution u r = (M/2)(x + w), and putting, as in Section 4, M = m, = 1, we obtain c
.F(w) = G, &
.Y
l--w
j
l2 0
dx
(
+J------
w
1(x; a)
w + x
(5.6)
g(x: w) 1'
where g, h, and I are functions of x and w given by Eqs. (4.2 1) (now we have w > 1). This shows directly that F(w) is given by the analytic continuation of F(w) (Eq. 4.20) for w > 1 along the real axis.
6. OCCURRENCE OF REAL BRANCH POINTS FOR u > 1
In this section we shall express the scaling function in term of the DGS [7] spectral function for the amplitude A. After specifying a box diagram for A we shall study the analyticity of P(w) along the whole positive real axis. We shall always assume the stability of the target hadron. We shall see that the analyticity for w > 1 is related to the stability of the vertically propagated particles. The equivalence of the present treatment with the one obtained from the Nambu representation is shown in Appendix A. The DGS representation for the amplitude A (see Fig. 2) is A(k2, P * W = j-l1 dP jlI;,, dh -k2 + 2;;A.;[)_
x _ iE .
(6.1)
For greater generality we take M2 = p2, in general different from mo2= m2. For F(w) we obtain (see Eq. (2.5))
F(w) = id j-y 4 4) I,,,
(1 - (w/B)) &0/B) &l - (w/S)) H(X; B) mo2@ -
w) + w(M2q8
+ A)
’
(6’2)
22
GATTO,
MENOTTI,
AND
VENDRAMIN
and for direct graphs (see Section 2) we have
= e(w) t?(l - W) 9 w), which defines the analytic function S(W). The simple box requires H(X, /?) for the Born term, which gives H(h, j?) = -(igz/(2r)3
8(/l - 1) 8(h - (p2 -fW))
(6.4)
l--o - w(m02 + A42 - p2) + m,2 *
(6.5)
and
SqlJ) =-e167r M2u2
Stability of the target requires M < m, + p. From Eq. (5.5) we find that (i) For /.L > M + m, the poles in o lie on the real negative axis. (ii) For m, > M + p (unstable parton) they lie on the real interval from +1 to too. (iii) Otherwise, i.e., m, - M < p < M + m, , the poles are complex conjugate and lie on the circle of radius (ma/M). This shows that for sufficiently high m, we expect singularities for w > 1; the problem of singularities for w > 1 will be considered in Section 7.
FIG. 13. Double box diagram.
DEEP
ELECTROPRODUCTION
AND
23
ANNIHILATION
We shall examine the double box shown in Fig. 13. For this double-box we need H((x, p) for the lower box in Fig. 13, which is given by
graph
where
The support of H(h, p) is easily understood, for stable target, by looking at the domain for A, X = +(01, p), as 01varies from 0 to 1. In particular it is immediately found that the lower integration bound A@) is given by the larger solution of the equation 4A
P, = (A- m2 + po2 + /3(m2 - p12 - M2))2 - 4p02(A + P2M2) = 0.
(6.8)
This has the form 4P)
= -M2P2
(6.9)
+ (PO + ( f@, M2, moZ,p12))1/2)2,
where (6.10)
f CA M2, mo2, p12) = M2/3* - p(m2 + M2 - p12) + m2, which has already been discussed previously. H(h, 8) can be explicitly evaluated by performing we obtain
the integral in Eq. (6.6) and
MA, p> = e(p) w - P)(l - p)ca/m
(6.11)
fl@, B>,
with
00 - 4P))P(k R(h~
I@
=
jGl
(d(h,
p))l/Z
[p2442
_
p(m2
P) - &Jo21 +
M2
-
p12)
+
m2]'
(6.12)
where B(X, /I) = h - m2 + P,,~ 5 p(m2 4 M* - p12). We notice that as M -c ,ul + m, p”p” - &m” + M2 - p12) + m2 > 0
for
O