Vous êtes sur la page 1sur 41




4555495 (1975)

Use of the Wigner


in Scattering


Department of Pkysics, College of William and Mary, Williamsburg, Virginia 23185 Received May 5, 1975

The basic equations of quantum scattering are translated into the Wigner representation. This puts quantum mechanics in the form of a stochastic process in phase space. Instead of complex valued wavefunctions and transition matrices, one now works with real-valued probability distributions and source functions, objects more responsive to physical intuition. Aside from writing out certain necessary basic expressions, the main purpose of this paper is to develop and stress the interpretive picture associated with this representation and to derive results used in applications published elsewhere. The quasiclassical guise assumed by the formalism lends itself particularly to approximations of complex multiparticle scattering problems. We hope to be laying the foundation for a systematic application of statistical approximations to such problems. The form of the integral equation for scattering as well as its multiple scattering expansion in this representation are derived. Since this formalism remains unchanged upon taking the classical limit, these results also constitute a general treatment of classical multiparticle collision theory. Quantum corrections to classical propogators are discussed briefly. The basic approximation used in the Monte Carlo method is derived in a fashion that allows for future refinement and includes bound state production. The close connection that must exist between inclusive production of a bound state and of its constituents is brought out in an especially graphic way by this formalism. In particular one can see how comparisons between such cross sections yield direct physical insight into relevant production mechanisms. Finally, as a simple illustration of some of the formalism, we treat scattering by a bound two-body system. Simple expressions for single- and double-scattering contributions to total and differential cross sections, as well as for all necessary shadow corrections thereto, are obtained. These are compared to previous results of Glauber and Goldberger.


This paper developes the elements of what may be called the Wigner representation of quantum scattering. It is well known that quantum mechanics can be formulated entirely in terms of density operators and linear operators (often called superoperators in this context) that act upon the densities [l, 2, 91. The Wigner representation of densities puts them in the form of real functions of the coordinates
* Supported
Copyright Au rights

in Part by N.A.S.A.

Contract NASI-11808.

0 1975 by Academic Press, Inc. of reproduction in my form reserved.






and momenta of the systems particles. In other words, density operator matrix elements are functions defined in the classical phase space of the system. Schroedingers equation, correspondingly transformed, is a linear equation for the densitys time dependence. It looks like the equation governing a Markovian stochastic process in phase space. The integral form of Schroedingers equation, the Lippmann-Schwinger equation in the usual Hilbert space representation of the theory is transformed into an integral form of the equation of continuity. The equation of continuity relates densities, currents, and sources. Thus, upon translation into the Wigner representation, the fundamental equations of quantum scattering appear in terms of these physically graphic and transparent objects. The main purpose of this paper is to develop and stress this picture and, in particular, to indicate how its graphic intuitively appealing nature may be exploited in the formulation of approximations to complex multiparticle scattering problems. The formalism encompasses both classical and quantum scattering. In fact, the Wigner representation is especially constructed to go over smoothly to the classical limit. In taking this limit, none of the equations change their form nor do any of the symbols appearing in them change in interpretation or role. Schroedingers equation goes over into the Liouville equation and its integral form similarly emerges as the classical integral equation of motion obtained here in its most general form [IO]. Quantum mechanics assumes a classical guise in this formalism; in this, lies the source of its intuitive power. Nevertheless, since we have here merely another representation of ordinary quantum mechanics, all of the latters wave and interference properties must be hidden within. What has happened is that this translation of quantum mechanics necessarily gives birth to a nonclassical stochastic process. Densities are not positive definite and in addition exhibit long-range oscillatory behavior-just that needed to produce interference. Thus, although the total structure of the formalism developed here and the various roles played by its elements (densities, sources, etc.) are isomorphic to classical theory, the particular functions needed to represent these elements will differ considerably in some respects from those of classical stochastic theory. These qualitative remarks point to the circumstances in which it may or may not be useful to employ this representation. The property being observed should not depend critically on high-order interference effects or concomitantly on the existence of certain long-range order or correlations. There also should be a significant advantage in being able to visualize at least some of the main participants in the process under consideration as having simultaneously defined positions and momenta while moving along trajectories in classical phase space. We shall call such circumstances quasiclassical, to be distinguished from the more restrictive term, semiclassical. The latter requires that wavelengths be short compared to




interaction length parameters in which case particles can be considered to move as if on a continuous classical path both in the force field and out of it. Quasiclassical is a broader and less sharply defined category that depends on the property being observed as well as on the state of the system. The most commonly known quasiclassical problem is that of obtaining thermodynamic and transport properties of gases under ordinary external conditions. Maxwell-Boltzmann statistics and often billiard ball models may be used even when the typical molecular wavelength is comparable to molecular size-a condition that violates the semiclassical criterion. Generally, quasiclassical scattering problems have been difficult to treat in the usual Hilbert space representations and we may hope that the approximations engendered by the Wigner phase space representation will be complimentary to those previously established. An example satisfying these conditions occurs in medium and high-energy collisions between various projectiles and nuclei. Elastic scattering, most of it diffractive, is treatable by the Glauber approximation, the multiple scattering expansion, the optical model, or various combinations of these. This is a highly coherent process depending critically on high-order interference effects. Inelastic scattering, in which the target is left in one, or possibly a few, well-defined final states, also may be treated by such methods. This leaves about half of the possible reactions unaccounted for-the nonelastic collisions-in which many nuclear species may be produced in assorted multiplicities and momenta. Since these final states are so complex, one generally observes averages over them such as inclusive cross sections and multiplicities. This washes out most high-order interference effects. Since this formalism is written directly in terms of the density matrix, such averaging may be done automatically with ease. Finally, perhaps the most important attribute this formalism brings to such a problem is its quasiclassical guise. These collisions involve may particles, high orders of multiple scattering, and possibly, collective motion. It is very difficult to see such phenomena in terms of waves in multidimensional position or momentum space. The minds eye is irresistably drawn to hydrodynamic, thermodynamic, or transport theory type pictures in which joint average distribution functions in position and momentum play a central role. In contrast to previous ad hoc application of such concepts to such problems of scattering, they occur here as natural and systematic approximation procedures of ordinary quantum mechanics. Another field that may be mentioned is chemical reactions. Here again complexity is most often the rule, statistical averaging is the natural ally, the important steric properties of the compounds are easily representable and, an added positive factor, short wavelengths often made the problem semiclassical as well as quasiclassical. The Wigner representation is especially suited to the semiclassical limit. There are other potential areas of application [ll] that will not be discussed since those already mentioned should serve to illustrate practical reasons and


E. A.


requisite criteria for use of the representation. However, it should be noted finally that the Wigner representation has had a long history of use in, especially, transport theory [12]. What is being developed here is a version suited to collision phenomena on complex yet microscopic systems. Section 2 gives elementary definitions, examples, and theorems connected to the Wigner representation. A bra-ket notation is introduced to designate elements of the vector space of density operators. The mapping of operators from their usual representation into the new one is given. Schroedingers equation is then transformed into the new representation. In Section 3, the semiclassical limit is discussed very briefly. The propagator for finite time translations on densities (not pure states) is obtained as a perturbation series in terms of its classical, not free, value. Thus, just as one thinks of waves propagating freely between successive scatterings by a potential in the usual representation of perturbation theory, here a particle moves on its classical trajectory between successive quantum jumps. These ideas lead to practical formalas for computing quantum corrections to semiclassical processes, but this branch of the subject will not be developed in this paper. Section 4 introduces the most basic concepts used in scattering in the context of potential scattering. The time-independent integral equation for scattering is derived. The effect of the potential on the incoming stationary flow of particles from the accelerator is expressed by a source distribution function in phase space. This gives the net rate of production of particles being produced by the potential with a certain position and momentum according to the heuristic quasiclassical interpertation of the symbols in this formalism. Source functions are in many ways analogous to transition matrices, but here they occupy an even more central role in the theory. The relation between the solution to the integral equation and the observed scattering cross section is established by noting that the spatial integral of the local production rate of particles of a certain momentum is the total production rate, which in turn, is the incident flux times the differential cross section. In Section 5, the results of the previous section are extended to the multiparticle problem where, it is expected, the most effective use of the formalism is to be found. Formulas continue to be analogous to standard results of scattering theory in multiparticle Hilbert space. In Section 6, the multiple scattering expansion is derived. The result is quite similar to that of Watson. Here the role of the fully off-shell transition matrix is taken by a jump operator. The general properties of the jump operator for two particles interacting via a phenomenological potential are fully as mysterious as those of its counterpart, the transition matrix. Much energy has been expended but little physical intuition can illuminate the off-shell properties of the latter complex function. In



Section 7, two limiting forms of the jump operator are examined, the classical and the dilute. The relation between a particle moving on a classical trajectory and the concept of sources previously developed is explained. In the dilute limit, in which all particles are very far from each other compared to all other length parameters of the problem, we obtain the basis of the Monte Carlo method [4]. However, having obtained this not as an ad hoc procedure but as a well-defined quantum mechanical approximation, one sees immediately how its range of application may be properly extended to include bound final states [5] and, in addition, how systematic improvements in the approximation may be made for less dilute systems. In Section 8, the multiple scattering formalism is developed further and then applied to find expressions for various simple but important production cross sections. In additions, the close relation between inclusive production of a bound state and of its unbound constituents is discussed. It is pointed out how comparisons between these measurements can provide information on production mechanisms. In Section 9, as a final illustration of the formalism we look in some detail at the simplest problem to which it might usefully be applied, scattering of an elementary projectile by a two body bound state in the dilute limit. The lowest-order term in the multiple scattering expansion gives immediately both the differential cross section formula first derived by Goldberger for quasifree scattering by a bound particle [6] as well as a corresponding expressing for the total cross section. The second-order term gives an obvious shadow correction to Goldbergers formula (the effects of which may be seen quite dramatically in reducing the backward peak in proton-deuteron eIastic scattering at intermediate energies) [13], a shadow correction to the total cross section, which is compared to that of Glauber [7] and, the expected double scattering contribution to the differential crosssection. All these results are straightforward to obtain and transparent in physical meaning. They are written in terms of observed cross sections. They are extendable in a straightforward fashion to more complex systems and also to nondilute systems [8]. 2. ELEMENTS OF THE WIGNER REPRESENTATION We begin with a resume of previous results concerning the Wigner representation. Details of certain standard derivations that may be found in the earlier literature are omitted. Initially, discussion will be confined to the case of one spinless particle. The extension to more than one spinless particle is straightforward and will be used in later sections. Throughout the paper, we ignore spin and:indistinguishability of particles.





Let 0 be any operator on single-particle Hilbert space. Its Wigner representative is defined to be [l]: Ou(x, p) = 1 dy eiP.y(x - :Y I 0 I x + i!Y)
(1) = s dq eiq.x(p + &q j 0 1p - +q).

We will give some simple examples. Let X and P denote the ordinary position and momentum operators and let 0(X, P) denote some functions of these operators; then (mm& 69 P) = O(x) and
(w%u (XTP) = 060. (2)

When 0 depends on products of conjugate operators, then their ordering will have due effect. Another instructive example is the density operator of a normalized Gaussian wave packet state in Hilbert space. The Wigner representative of this operator is

where 52and @are averages and dx, Apa = +. For any operator 0, <x I 0 I x> = j 4 CW3
Owtx, P), (4) <P I 0 I P> = 1 dx (2~)-~ OttAx, P), j 0 I = j- dx (x I 0 I x) (5) = s dx dp (27r)-3 Ow(x, p).

In particular, if 0 is a density operator p, then the normalization


condition is

It can be seen that pw appears to be a sort of joint probability tion in phase space, the volume element of which is given by
dx dp (27~)-~ = dx dp h-3






(since h = 1). We find it convenient to denote a point in phase space by a single symbol v = 1% P> and correspondingly a volume element as
drp = dx dp (2~)-~.

It is straight forward to show in general that for any two operators 0 and 0,

I 00 I = j- dg, O,(T) O,(T).


In particular, the expectation value of any 0 in the state described by p is given by

I PO I = s 6 pdv) ow(~).

This is consistent with a probability density interpretation of pw . What is inconsistent with this interpretation, however, at least in the ordinary sense, is the fact that although pro is necessarily real, it is not necessarily positive everywhere. Thus, Jde dp, pw(y) cannot refer to the probability of a realizable measurement (measuring whether the particle is in dv) for arbitrary regions, LI v, of phase space. This is consistent with the quantum mechanical fact that to every region of phase space there does not necessarily correspond a physically realizable measurement. In particular, dp, might violate the uncertainty principle. As we shall see, this fact does not appear to affect the heuristic value in thinking of pw as a joint probability distribution function in the quantum scattering formalism to be developed. Equation (1) maps Hilbert space operators onto phase space functions. The inverse mapping may be performed with the help of the operator
R(x, P> = 1 4 eiaex I p .$q)(p + $q 1 (9)

which has Wigner representative

(ax, PNW w, P) = cw3 6(x - x> %P - P)

or (~(FNtlJ (94 = SC% v>*



Thus, Z?(F) maps into the Dirac delta density in phase space. Although it is not a true quantum mechanical density operator, since it does not describe a physically realizable state of the system, it will be a convenient heuristic device to speak of it as a density.


E. A.


We not note some useful properties of R(q);

s 4 R(v) = 1, I W9J)l = 1,
Ow(v) = I OR(v I R(v) R(F)l = V% 90. The above equations imply


(13) (14) (15)

In fact, it is easy to see that this integral has the same Wigner representative as 0. Therefore, it must equal 0 as long as distinct Hilbert space operators map into necessarily distinct phase space functions. The latter statement is true because the Wigner representation is accomplished via a Fourier transformation on the matrix elements of 0. Another bit of notation is now introduced which serves, among other things, to enhance the analogy between the usual quantum scattering formalism and the representation of it to be developed here. 0, is considered as an element of a linear vector space [2] and written in braket notation as I 0). IO) = 0, (17)

We also write

I R(v)) = I v>,
from which one gets


CT I 0) = Odd (Y I 9) = XT, T? s dp, I v>(v I = 1.

(20) (21) (22)

Quantum operators appear in this formalism in two distinct ways: as kets, of the type just introduced, or as operators on kets. Generally, density operators are



463 space operators (p generally

mapped into kets. Now let 0 and p be any Hilbert being a density operator), we write
OLl p) = lop), ORIP) = IPO).


Previous authors have shown [3], in effect, that the matrix elements of these operators are given by

where fl is the Poisson Bracket operator, (25) Alternatively, we note

(Y I OL I F> = <p I ONF)) = I R(F) OR(v

Similarly, it is easy to see that
(p' 1 OR I q'> = I R(#)OR(y)l.



Thus, they are related by transposition


19) = ($1
= (9'

OLIT) 1 OR I q'>.

We are now prepared to consider the form that dynamics takes in the Wigner representation. Schroedingers equation is

a/atp = -i(Hp - pH)

as written in the usual Hilbert becomes


space. In terms of Wigner representatives, this

a/at I p) = -i 1Hp - pH)

= -WL HR)I






It is easy to see from the preceding that the hermeticity

H,* = H R

of H implies (31) from which is

(32) (33)

(complex conjugation obtained In addition, we note

taken with respect to the q~ representation),

D = D. D = -DT.

The antisymmetry of the time evolution operator is sufficient to prove conservation of probability and time reversal invariance. If (p [ p) were a true positive semidefinite probability density, then p(t) would correspond to a true stochastic process. One aim of this paper is to demonstrate that the analogy to statistical physics implied by a formal probability density interpretation of (g 1p) may be extended to all the basic equations of quantum scattering theory, which thereupon assumes most naturally the guise of a stochastic process in classical phase space. The situation encountered here with p is somewhat reminiscent of that of the Dirac delta function. The latter is not a true function (nor can the various commonly used distributions containing it correspond to true quantum states). Nevertheless, it is extremely useful to treat it formally as a function (and to use nonnormalizable eigenfunctions) during intermediate stages of a calculation. One knows that such distributions have meaning when finally multiplied by test functions and integrated. Similarly, (p 1p), while not a true probability density may be formally treated as such, and usefully treated as such during intermediate stages of a calculation. At the end, one must not ask of the formalism quantum mechanically disallowed questions such as: What is the probability for finding a particle infinitely sharply localized in phase space? Provided only that sensible questions are asked, quantum mechanically correct and sensible answers will be obtained. Mathematically, this again will entail intergration with test functions of a certain class.




The dynamical equation can be reexpressed as alat <V I P> = H(d(--2 where H(v) is just the classical Hamiltonian. sine gives the classical Liouville equation sin WWF
I P>,


The first term in the expansion of the CT I P>(35)

alat <p I P> = --H($n




Let us call D, the operator that generates classical time evolution;


It is clear that the solution to the classical problem is given formally by

eDctI 9J>= I d% t>>,


where vC is the function that gives the coordinates of the phase point of the classical trajectory at time t such that y&y, 0) = y. Let A(t) be the operator that corrects the classical propagator,
e Dt = d(t) eDCt. (38)

It satisfies
a/at n(t) = eDt(D - DC) eeDCt = deDCt(D - DC) eCDCf. (39)

Its perturbative

solution with initial condition d(0) = 1, is d(t) = 1 + 1 dt eDCt(D - DC) epDCt


+ Iot dt Jot dt eDct(D - DC) eeDCt . eDct(D _ DC) e-Dct + . . . . (40)

The first quantum correction to the matrix element of the time evolution operator is thus given by
cv' 1 ,Dt e%t I v> = Iot dt' (w(F', -0 I D DC I wAy, t t')> + a*-


using Eq. (37). The interpretation of this equation is that a particle may travel from q~to v in a time t via an infinite variety of paths in phase space. Each path is itself made up of some number of classical path segments, the first beginning at q~and the last ending at q (Fig. 1). The particle moves classically along the first segment, performs a quantum jump through phase space to the second when it reaches the end of the first, and proceeds in like manner until the end. Each quantum jump, say, that going from point y1 to vz, occurs with a probability (vz / D - D, 1 c&. This probability may be negative. In fact, due to the antisymmetry of both D and D, , the time-reversed jump always occurs with a sign opposite that of the time-direct jump.



T = t-t'

FIG. 1. This illustrates the first-order correction to the classical propogator given in Eq. (41). A particle is known to be at p initially and one asks for the probability of it being at p at a time t later. In addition to the classical value of this probability there is a contribution corresponding to the particle traveling a time I - t along its initial classical trajectory and then performing a quantum jump to a new classical trajectory which brings it to I in the required time 1. The trajectories are pictured in position space emphasizing the fact that for a local potential there is a jump in momentum only as with a classical stochastic impulse. The probability that the particle makes the jump is given by Eq. (41).

It is interesting at this point to exhibit in more detail the jump probability operators matrix elements for potential scattering. If H = X2/2&f t V(P) then it is easy to see that only V can contribute to D - D, . We have (9) I VL I 90 = I NT) = VR(p,)l

dy df ,+~Y+ip~Y s -Ilx+~y)(x-~yIvIx+~y)(x- +y)



= 6(x - x) j dy e+~-~)v(x

so that

(p' I D - DC I q~'>= S(x - x'> Ij dy ei(p--p)T(V(x + $y) - V(x - iy))

- (21T)3 (a/ax V(x) * a/ap S(p - p.))/. (43)

This shows that a particle may experience a finite jump in momentum but not in position, which is what is expected for a particle subject to a random force. One can show that the average of this random force is just the classical force.


In this section, we wish to develop the essential ideas of this paper in the simplified context of potential scattering. These results will be extended subsequently to the general multiparticle multichannel case. The plan here is first to derive the integral equation for scattering-the Lippmann-Schwinger equation in the Wigner representation. Next, we want to relate the solution of this equation to the cross




section, and finally, to interpret the symbols appearing in these results in a way that brings out clearly the analogy to statistical physics. Let D, = -i[(P/2M)L - (P2/2M),] (44) be the time evolution operator in the absence of interaction. expansion in terms of Poisson bracket operators, it is (9 I D, = --v * a/ax (F I , where v = p/M. Let 1p(t)) represent any density of freely moving particles, alat 1p(t)> = 4 1f(t)> and consider a density I p(+)(t))
/ ,.4+(t)> = / p(t))


using the


+ J:, dt eD@)(D D,) / p+(t)).

(47) (48) (49)

Then, clearly

a/at 1p(+yt)j =

D I p(+yt))

pm (I P+(t)> - I p(t)>> = 0.

Thus, 1p(+)(t)) is that solution to the potential scattering problem with the boundary condition that it approach a specific free density 1p(t)) as t + - cc. One already knows that solutions to this integral equation exist in which the densities correspond to wavepackets. More generally, here they can be mixtures of packets. We now assume that an exponential convergence factor will suffice to allow limits in which p(t) is stationary to be taken in the integral equation. This leads to the definition
1 p(+)(O)) = I p) + = 1 p) [j-4, dt e-(Do-Z)] (D, z)- (D (D D,) ! p+(O))

D,) 1 p-(O))

with D, I p) = 0.

If, in addition, the limit as z - 0 of the solution of this time independent equation exists, algebra suffices to derive
D / pf) = 0,

where li+i / p+(o)) SE 1PC+).


E. A.


Another representation of the time-independent have a very interesting and important interpretation

equation that will be seen to is obtained by noting that


eDOt I x, p) = I x + vt, P>

from which, by transposition, one obtains

(x, p I emDot = (x + vt, p [.


This, inserted into the integral equation yields (limit notation implicit)
(x, P I P+) = <x, P I p> + sf, dt (x + vt, p I D - D, I p+). (54)

One can now verify by direct substitution e.g., momentum pE and unit density

that when p represents a plane wave of,

I p> - I 0 = j- dx I x, PE)
= Gw3


(I pE), a Hilbert space ket), then in fact

I E+) = G-I3 ICIP%(+PE



satisfies the time-independent equation in the limit z -+ 0, 1pk:) being the outgoing wave solution to the Schroedinger equation. Thus, the limit of this equation is not vacuous and possesses the physically expected solutions. We next show how to obtain the correct formula for the cross section based on the assumption that (p I EC+)) may be treated as a density. The purpose in doing this is to develop and emphasize the heuristic power of the symbols in the formalism. In addition, the resulting procedure is remarkably simple. A rigorous derivation for potential scattering, leading to the same result, has been given elsewhere [5]. Under this assumption then, the current density of particles of momentum p is

JE(x, PI = v<x, P I E+)



(27~)~ dp s, JE . d.9, where Y is a surface totally enclosing the region of interaction,is the total outgoing flux of particles within the interval dp about p. Let a(p, pE) dp be the probability




that an incident particle of momentum corresponding interval. Thus,

pE be scattered by the potential

into the

Q(p, PE) ZE = s JE . dy (27r)-3. s


Gausss theorem, we get

G(P, PE) VE = (27F j. dx SE(X, P), (59)

where the integral extends over all space and the source function SE is defined by

P> =



p 1 EC+))

v . a/ax

= -(x,plD,/E+) = (x, p / D D, j Et+).

The equation of continuity when applied to a stationary distribution shows that the divergence of the current, the source function as defined here, does indeed give the local rate of production of particles within the element ds, = dx dp (2~)-~. Therefore, our equation for the cross section merely states that the total rate of production of particles in interval dp is the integral over all space of the local production rate. It is easy to see that the usual differential cross section is given by
U(P, ~4 = I P I- %I P I - I PE I) d4dQ. (61)

The equation for the source function in terms of the density function can be rewritten as
SE(q) = s dq (g, 1D D, 1 $j(p;' ! Et+)\. (62)

Now (y / D 1 IJJ) is the total probability per unit time that a particle at y will jump to cp, while (40 I D,, / cp'j is the contribution to that probability due to simple free particle streaming. Thus, by the above equation, since dq( F j E(+j is the number of particles in steady state in dy, S,(v) dg, is the rate at which particles are jumping into dg, due to the potential. This is equal and opposite to the rate at which particles are jumping into dzp due to streaming. The net rate at which particles are jumping into dp, is zero since we are in steady state, i.e.,
f dcp (p, I D I y)( y I EC+) = 0. (63)



The Wigner representation of the Lippmann-Schwinger equation for the time independent wavefunction of the scattering problem can be rewritten as
<x, P I E+> = <x, P I E) + /:m dt &(x + vt, PI. (64)

This expresses the number of particles to be found in dp, as being equal to those that would be there in the absence of the potential plus the net number (possibly negative) of arrivals of those streaming in to this interval after jumping in somewhere upstream because of having undergone an interaction. In the usual Hilbert space representation of scattering, an equation for the transition matrix may be derived from the equation for the wavefunction by operating on the latter with the potential operator. A similar maneuver here yields a similar result:
sdd = (9, I D Do I E) + j dv' j" --m dt (9' I D Do I q~'> 5'E(To(~', t));

tpo(cp, t) = x + vt, p.


In summary, what we have shown is that the elementary relations of scattering theory translate, under the Wigner representation, into elementary relations of transport theory. The only thing that distinguishes quantum from classical theory in all of this lies in the details of the jump probability function (v 1D - Do j q~) and in the occurrence of negative probability densities during intermediate stages of the calculation.

Our first step will be to generalize the density operator R(v) describing a completely localized elementary particle to obtain a corresponding operator for bound systems of elementary particles. Consider as a simplest example a bound state D of particles 1 and 2. Let I pD) be the Hilbert space ket corresponding to a momentum eigenstate of D. Then
I PO) = 1 dp, dp, I ~1, P&P I D> XP, + ~2 - PD), (66)

where Thus, (p 1D) is the momentum of this system and space wavefunction describing the internal structure

<D I W = j 4 @ I P><PI D> = 1.




We find it useful to define, as before,

RD(yD) = s dq eiqxD I PO + k>(PD - h I. (68)

This is now a two-body operator, but it has much the same properties as before. In particular, I dw, RD(VD) = SD I RD(wJI = 1, I RD(TD) RD(FD)~ = a(~, >TO), (69) (70) (71)

where pD = xD , PO, and gD is the projection operator for D. Using an obvious extension of the expansion theorem written in terms of the elementary particle density operator R(v) that was discussed previously, we have

Carrying through, the algebra yields

RD(YD) = jdg, wD(?) R,(R) RAoA


where x1 , p1 , xz , pz are the linear combinations center-of-mass transformation and

of Xn , Pn , x, p appropriate to the

wD(x, p) = j dq eiq (P - $q I D)@

I P + iq>


is just the Wigner representative of the internal structure wavefunction of D. The set of internal phase coordinates of a bound system will be given the generic symbol [. Thus, in the simple case we have discussing
bD > PD ; x, p> = i?D ; $Dh

(2~)-~ dXD dpD (27~-~ dx dp = dcfD df, , etc. If the system D were to be instead, a three-body system, the set 50 would denote two internal phase points and so on. In all cases of N elementary particles bound into a state D we write (75) (76)





Consider next the integral equation of the stationary density describing a beam of particles B scattered by a target A in the laboratory frame. Both A and B in general may be bound states. Assume the whole system to be made up of N elementary particles. The incoming channel will be denoted by the subscript AB. The incoming channel Hamiltonian is HAB . In the absence of any interaction between A and B, the incoming density operator satisfies the Schroedinger equation

a/atP,m(t)= -~(HABPAB(~) - PAB@ HAB)

alat 1 pAB(t)) = DAB 1 PABct)).


Proceeding as before, we obtain


for the stationary scattering state. To obtain an expression for the cross section, we again construct a standard density describing the unperturbed system
1 PAB) = I AB) = 1 RA@ 0) J ~XB &(XB 3 PB)).


At this point, the objection might well be raised that since this operator does not really represent a physically realizable quantum mechanical state of the system, an error may be made in its use as the unperturbed initial state in a quantum scattering problem. A more detailed derivation, given in the Appendix, shows that this is not the case. It is yet another example of a phenomenon noted earlier; the formalism is indifferent to the quantum nature of the problem. This density describes a target particle A localized at the origin with zero momentum, while at the same time a steady stream of bombarding particles B flows in with momentum pB and flux
UB = 1 VB / = 1 PB/MB / .

Due to the interaction,


this distribution

is altered to become
] AB+). (80)

= 1AB) + [J:m dt eMDAB] (D - &)

Suppose one were interested in the inclusive cross section, A + B -+ C + anything, where C is some, in general, bound state of the system. Asymptotically far from the origin, the components of C, if it is a bound state, will have a negligable




probability of being close together unless in fact they are in their bound state. Thus, at large distances (yc / AI++)), where I rpc) = I M~cD

is the probability density for finding C within &I,. Note that this inner product, since it is in fact the Hilbert space trace of R,(yc) with the stationary density describing the collision, includes an integration over unobserved final state particles and an integration over [c , the internal coordinates of C. The current of particles C far from the region of interaction is




Proceeding exactly as in potential scattering yields, now for the inclusive production rate


Sc:,a(~)c) = a/& . JAB(~ = vc . a/ax, (qJc I ABf).

Note next that if Hc is the Hilbert practicles in C such that space Hamiltonian PC), acting on the subspace of (64) pc, then it is straight(85)

Hc I PC) = (P~/~~c)I where / p,-) is the Hilbert forward to prove that --iVWc(q~c) This implies

space eigenket of momentum

- Rc(w) Hc) = Vc . a/a% &(yc).

DCI yc) = vc . a/h I vc, cqc I DC= -Vc . a/a% (FC ! /Ct 1& = j e-HCt&(vc)e-=Ct)
= I R&c + vd, PC)) = I xc + vet, PC). Further, note that









where DC is the generator for all particles not included in the cluster C. This is true because the symbol (vc 1 as used here contains an implicit integration over all the phase coordinates of these unobserved particles; this fact, in conjunction with Eq. (12), yields the result. Combining Eqs. (87), (88), and (83) yields

&:AB(w) = <w I - DC IAB+)

= (cpc 1D - DC - Dc j ABC+)

or &4B(W) = <yc I Ic I AB+)


where Ic is the sum of all interactions between members of C and members of C, the set of all unobserved particles in the final state. It might be noted that this equation is very similar to its analog in the ordinary representation of scattering theory except that it is simpler having eliminated all reference to unobserved final state channels by means of the closure idea. Cross sections for two or more particles in the final state may be derived in a similar manner. For example, consider the cross section for A + B --+ C + Df anything. The pertinent two-particle phase space probability density is (WPD I A@+)). Consider surfaces Ye and LYD, exterior to and enclosing the region of interaction. The probability that C be found in dxc dpc about xc, pc and, at the same time, that D be found anywhere in the volume VD surrounded by 9 with momentum in dp, about PD iS dpc dP, dxc (277k3
fyb dXD



1 A@+).


The total number of such pairs of particles being produced per unit time is
~CD;,l&PD ;

Pe) 1B / dP, dPD

(24-3 (q&b 1AB+). (91)

= dp, h-b fyc (2n)k3 dYc * vc &ym j-r, &

This can be converted by use of Gausss theorem to

= lim



* a/ax, fvDdX, syc dxc (2~)-~ vc



1 A@+).


The order in which the limits are to be taken when the integrand is written in this



form is important. In fact, were the order to be reversed, the integral would vanish since for fixed xD , (vcyD 1A?(+)) must rapidly vanish as 1xc 1+ co, and vice versa, due to the necessarily correlated nature of the particles emissions. We can take advantage of this fact and add to the above integrand the perfect differential VD * w&l <qwpD I AB+)

without changing the value of the integral. This allows us to write

uCD:AB(k, PD ; PB) DB


dXD (2~)-3(VC

* wxc

+ VD * v+b>

<qwpLl I AB).


We have deleted the symbols limYC-)m limTD+m because in this form the order is clearly immaterial. Further, using (vc . wxc + VD - vxD)<qwpD = (qwJrl I(-& - w, = <WPD I D, = 0, I (94) (95)

where CD is the set of particles neither in C nor D, one obtains, in complete analogy to the one particle inclusive case,
uCD:AB(P, PO ; PB) OB = dxC (2n)-3 b (271)-3 sCD;AB(~c9)D)


where &D;AB(SPCP)D) = (qw/b I GD I AB+>, (97)

and I&, is the sum of all interaction between members of the set CD and the unobserved set a as well as interactions between members of C and members of D. Thus, we have obtained the general integral equation for scattering and shown how to write down the expression for any desired cross section. The formulas are simple generalizations of those obtained in potential scattering and analogous to those of ordinary scattering theory.




The analog of the Lippman-Schwinger given in Eq. (80), may be rewritten as

equation for the multiparticle


j ABC+) = / AB) + GaBZAB 1AB),





GAB = f

dt eeDABt

= -(DAB- z)-1; z-++o

lAB = D - DA;


are the Greens function and interaction operators, respectively. As in the case of the Hilbert space representation of the problem, there is an alternative form for this equation: 1 ABC+) = 1AB) + GZAB / AB) ww where G = -(D - z)-. The full Greens function G may be expanded in the usual manner. We write the interaction as a sum over pairwise parts

z= CL, a Ja= I, + I,G,J, ,

- z)-.

(102) (103)

where 01runs over all pairs of particles in the system A + B. Now define a jump operator by where
Go = -(Do


Thus, the two-body jump operators take the place of the two-body transition operators in the usual development. However, there is one interesting difference between the two representations. Neither the Greens function nor, consequently, the jump operators used here have the energy of the A + B system as a parameter as is the case with the transition matrices in the Watson expansion. Thus, there is no possibility of an off-shell or half off-shell type jump operator and, in fact, the same jump operator is valid for any energy and any system in which the corresponding two particles appear. This fact does not effect the algebraic structure of the equations, which is the same here as in the usual analysis so that we get finally G=

+ G,,JGo ,

(106) (107) (108)


J = Z + ZG,J, Jo = I + JG 0ZL1,
c OL J(a) = ;

Jcol) = J, + c JbG,J,




GL = Gda -k GoJGd,
= G& + Go[J(a) - Im] = G,,J(o). (111)

The last result inserted into Eq. (101) yields

where the prime on the summation indicates that a is restricted to all pairs not interacting in the AB channel. The perturbation expansion of Eq. (110) inserted into Eq. (112) is the multiple scattering equation:
I AB+> = I AB) + c GoJ,,I AB) + c c G,,J,G,J, / AB) + ... .

0: Bfcr



Before extending the multiple scattering theory begun in the previous section further, it will be useful to examine the properties of the elementary operators appearing in the expansion given in Eq. (113). Consider first potential scattering and the matrix element (Y I 0 I cp> = <F I G?G I F,>

=s According to CT I eDt 1 v> was localized up at t at v.

OD dt v . a/ax (p 1eDt 1F>.


the probability interpretation of the symbols developed in this paper, is the probability density of a particle to be at v at a time t after it at v. The integrand in (114) is thus the divergence of the current set Applying the equation of continuity we get

= j-= dt S(v, t> + CT I v>,

where S(y, t) describes the sources due to the interaction and p(y, t), the particle probability density at any time t. Since S(v, t) is a time rate of production, its time integral equals the net production. We can now describe the situation as follows: An external source pumps a net of one particle into the system at 9). The system





over a period of time reacts and redistributes this source distribution from ( y [ y) to (v 1 u 1 CJJ). Since u is in the nature of a probability one must have

This is in fact satisfied by the theory already since, using Eq. (106). u = 1 + JG,, and s
d# (F I J = 0 (116)


identically. The relation between the operator CJand the function SE(~) already introduced in the discussion of time-independent potential scattering (Eq. (60)) is SE(TJ) = j db (9, I u I m) UB,

where xE = (bE , zE) , and bE is the impact parameter two-vector perpendicular to the entering momentum pE , and zE -+ - co, along the beam line, where the accelerator is situated. The equation says that the accelerator is a source distributed randomly over the plane zE = -cc with an average density of one per unit area. The system reacts to this external probe, by redistributing this density. Thus a particle introduced at TE will reappear at F with probability density (p ( u I TE). The cross section is
O(P, PE> = j dx W7F3 db <p, I u

I FE).


In principle, a could be computed using

<v I (7I F> = v * a/ax jm o dt / I?(#) e-iHtR(v) e+iHt I


if the spectral decomposition of H is known. This seems to be a difficult task. It will be seen that one can often avoid such an explicit computation. The classical limit is of practical interest and can be discussed in some generality. We have, as discussed in Section 3,

(@ I uc / q} = jm dt
0 = s0



(c$ [ eDCt/ v) (121)

mdt (9 I - DoeDCt I y).




To see what this says, imagine a numerical integration of the classical equation of motion. Time is segmented; the first interval is 0 < t < t, ; the next is tI S t < tz , etc. During the ith interval, the trajectory is approximated by a straight line xi(t) = x,(O) + v*(t - tg) X,(G) = Xi+,(O) while the momentum pi is constant. The momentum changes discontinuously, if by an impulsive force, at the instants t, , tz ,. .. . Thus,


= <qJ' I fp> + 1 [-(4 I xi+1 ,

Pi> + CT I xi+1 9 Pi+21


having used,
-DoeDot = -(d/dt) eDot,

(Xl 9 PI) = (5 P) = VThus, the reaction of the system in the classical limit to an imposed unit source at C+I is a line distribution along the classical trajectory of sources and sinks. At each spatial point subsequent to x, there is a unit sink for the old momentum going into the point and a unit source for the new mementum leaving the point (Fig. 2).

P ttlb,t1/ .:. P I 2
x=x, PI x, Pz 3 x, Ps x, 4 p4

FIG. 2. This illustrates the classical limit of scattering in terms of source functions as given in Eq. (123) and discussion subsequent to it. A circled portion of the spatial trajectory vo is enlarged. The continuous force is approximated by a sequence of inpulses. A particle starts at x1 with momentum pl. Thus at x1 there is a unit source of such particles. At x8 it receives an impulse, changing its momentum from p1 to pz . Thus, at x2 there is a unit sink for p1 and source for pz , and so on. In this picture, classical motion is a consequence of the potentials creation of a line of sources and sinks in response to the externally imposed source at x1, pI .





Another limit of considerable practical and conceptual interest that can be discussed in some generality may be termed dilute. This will be applied to a multiple scattering problem when the distance between successive collisions is large compared to other characteristic lengths of the problem, wavelengths and ranges of potentials of particles. In particular this limit implies no necessary ordering between wavelengths and ranges, whereas the classical limit requires wavelengths to be smaller than all other characteristic lengths. Consider, for simplicity, scattering by a system of fixed scatters centered at points x, . The multiple scattering series is still given by Eq. (113); the subscripts now range over fixed scatterers indices. In any one of the terms of the series in which the orth jump operator J, appears it operates on some particle density function; call it 1p). In the dilute limit, 1p) is set up by sources centered far away from x, . This leads to the conjecture that any such density must be slowly varying over the scale of distance set by the size of the olth scatterer, the range of its potential. We could then, expand I p) about x, and, in the most favorable case, keep only the lowest order term; thus,


However, quantum mechanics implies the existence, in principle, of substantial oscillations in 1p) associated with interference. These would vary on the scale of a wavelength and contradict the conjecture. This question is resolved in the further quantum mechanical requirement that sources be distributed over a certain minimum region in phase space. Oscillations due to distributed sources wash out increasingly with distance; in the dilute limit they disappear. When the condition of diluteness is not strong enough to allay all worries associated with the approximation of Eq. (124) other factors may assist. Randomness is often invoked to accomplish phase averaging. A relatively density packed medium such as a liquid or a nucleus will possess only short-range order. The approximation in Eq. (124) might be good provided only that I p> contains no contribution in which a near neighbor of cywas the last scatterer. Terms in the multiple scattering series that describe successive scatterings by near neighbors might be treated separately, leading to an expansion in powers of correlation functions of the medium. Another escape hatch opens if the wavelength of the particles (or particles) being scattered is short compared to a long-range component of the potentials doing the scattering. It is straightforward to show that the multiple scattering series can be rewritten so that the classical propagator Gc replaces G, . Concurrently, the jump




operator is replaced by one that represents the stochastic quantum jump correction to the classical motion, as discussed in Section 3. This correction is due mainly to the short-range component of the potential. Such cores may be spaced widely enough apart for the diluteness condition to apply. J, 1p) is the source distribution created by the olth scatterer when 1p} impinges. This must always be integrated over another density, say (p 1 , to calculate an observable probability (p j J, / p). But the arguments applied to J, 1p), also apply to ip 1J, , using time-reversal invariance. Thus, one can write


= s dg, <P I XaP><p, I J, *

Combining Eqs. (124) and (125),

CT I J, I F> cs %x - x,) 6(x - XJ j dx dx (y / J, I cp).

In contrast with the general expression for u given in Eq. (120), this approximation for the closely related jump operator is easy to evaluate and interpert. Note that sdx I x, p) is the density for a unit incoming plane wave state. Therefore, it can be written as



where / p> is a Hilbert space ket. Next, note that J, = G;GI, which, in combination ,

with our integral equation for scattering, Eq. (101) yields P+)<P+)

J, j dx I xpl = (27d3 G% Hence, s dx dx Cx, P I Ja I x, P>

I)? - !(i PXP WI.


= (2d3 j dx (~7 P I GMl P+%P+ = (2~)~ j dx v - a/ax KP+ I W,

I)> - I(1 P>(P


P) I P+) - <P I R(x, P) I P>I>





using Eqs. (5), (7), (18), and (19). Inserting an expansion for R(x, p) such as that given in Eq. (9) and writing (P + &I I p(+> = (P + &I I P>
+ (C(P) + io - 4 + $q))-l TdP + $a; p), Ta*@ - $q; p), (131) (132)

<PC+)I P - 4q) = <P I P - $4)

+ (e(p) - i0 - c(p - an))-

for the usual expression relating outgoing wave solutions to Greens function and transition matrices, one gets, s & dx <x, P I J, I x, P>
= (2T)3 1 dx dq - {<p I p V'

. a/ax eiQx + i0 4-f + h)Y TAP + hi; P)

$4) (40

+ <P + Bq I P> (E(P) - io - 4P - iq))- + (c(p) + i0 - l (p + $a))- (e(p) - i0 * TAP + 4q; P) T,*(P iq; P)>.

T,*(P - gq; P) l (p - $a))-l (133)

The terms linear in T are the complex conjugates of each other. The first of these integrates over dq to become
(27r)3 / dx eiqx @iv * q)(&) + i0 - 41 + q)Y TAP + q; p) (134)

with q = 2(p - p). The integral over dx will give a a-function distribution so that we can write inside the integral

in q

+) - E(p + q') A --Q' - ar/ap = -q' sV.

Integration over dx now gives
- isb - ~)(27V TAP; P).


The sum of the linear terms in T is

XW %P - P) Im TAP; P). (137)

The quadratic term may be evaluated easily by noting again that the x integration restricts the integrand to the neighborhood q SW0. Thus, we get
(29~)~ I T,(p; p)j2 lii iv . q (e(p) - i0 c(p ;a))- (138)

. (c(p) + i0 - c(p + +q))-l

= (277Y (277-I %c(P) - 4~)) I TAP; P)I.




Hence we obtain in the dilute limit,

Xv I J, I F> = 86 - x,)G7d66(x - x,1

{a(~ - P) 2 Im TAP; P) + %e(P) - c(P)) 2n I T,(P; P)l>. Using the optical theorem and the relation between the transition differential cross section ToTAL(p) = -2(277)3 T(p; p) VU, WJp; P) = (27? &C(P) - C(P)) I T(p; P)12, (139) matrix and the

(140) (141)

where a,(~; p) is related to the differential cross section by Eq. (61) this result can be rewritten as

. (243


- p) Uo:oTAL(p) + YAP;



This equation (142) says something that could have been written down right away using elementary considerations. In particular, a density ( p) impinging upon the c&h fixed scatterer gives rise to a source distribution function

(y I J, I P> = (2n13w
. W(P

- 4 J dP {(257)-3 L.&Y 9 P I P>I

- P) zoTAL(P) + %(PG PH. (143)

The first bracket is the incoming flux per unit area of particles between p and p + dp at x, . This is multiplied by two factors in the second bracket; the first gives the number produced per unit dp at p, a positive source term; the second gives the number lost to the incoming beam, a negative sink term. This result can be immediately generalized to include nonfixed scatterers. Let 01 now refer to a pair of particles, e.g., particles 1 and 2 Then we have


m2 -


X12) I





. {- I Vl
where in the written The is that

S(PL -



I Vl



9 PI,)),


X,, , P,, refer to phase coordinates of the center of mass while xl1 , p12 apply center of mass. This result may be obtained from the formalism but can be down directly on physical grounds. advantage gained in having derived this approximation via this formalism it is now part of quantum mechanical scattering theory. The possibility of



systematic refinements to any approximation is thereby created. We have already discussed quantum corrections to the classical limit in Section 3. The possibility of combining the classical approximation (for motion through the long-range part of a potential) with the dilute approximation (for transitions induced by short-range potential cores) has been mentioned. In addition, finite size corrections to the dilute approximation, arising when the range of the potential is not negligably small compared to the distance between scatterers, have been caiculated [8]. The relation between the dilute approximation and Monte Carlo calculations has been discussed elsewhere [ 151.

8. MULTIPLE SCATTERING EXPANSIONS OF INCLUSIVE CROSS SECTIONS The previous section developed certain limiting cases that are useful for future applications and serve to illustrate and clarify the meaning of the basic jump operators that appear in the multiple scattering expansion. We now return to more general developments of the theory. Formulas for the simplest inclusive production processes are rewritten in a multiple scattering form, which facilitates their interpertations and computation in many cases of interest. It is useful to define a symbol (@).V for the sum of all scatterings beginning with the 01and ending with the ,f3pair. Simple expansion in powers of the jump operators will verify that it satisfies, (B)P) = .I$,, + c JB(l - 6,,) G,,(Y)J(OL) - a,,,) G,,J, . (145)

= J,&, + ;. @J(l

Furthermore, using Eqs. (106) and (107), and some strictly algebraic manipulations, one can show that cO)J = IaS,, + I,GZ, . (146) Particle B impinges on a composite target A. Consider first the inclusive source distribution for B observed in the final state. Application of Eq. (89) shows

&: AB(P) = (w I 1~ I ABC+) = (9)~ I b(1 + Gfs-m) I A&.

Since IB = IAB , the sum of all interactions immediately obtains
&;&cpB) =


between B and constituents of A, one


c (Q)B I cBa)JBa I AB) a*a




where a and a run over all constituents of A. If, for example, A were a two-body bound state composed of particles labeled 1 and 2, the multiple scattering expansion of Eq. (148) yields
~B;AB(TB') = (9~' IJBI + Jm + JB&,JB~ + JBI'&JB~ t .'. I A@


These expressions will be evaluated in the dilute limit in the next section. Expressions for the inclusive production of a single particle other than the incident one (knockout), follow in a similar manner. These source functions are, of course, to be integrated over space to give the total production rate or, equivalently, the cross section. Several examples of practical interest involve two or more particles observed in the final state. They may be in bound or unbound condition. Our intuition would like to relate, for example, a two-particle inclusive production rate to that of a bound state of the same two particles. The present formalism brings this out in an especially graphic manner. The inclusive source function for the bound state production of D, made up of constituents 1 and 2 originally in A is given by Eq. (89): S,,~P),,) = (qD I ZD 1ABC+:,. (150)

That for 1 and 2 unbound is given by Eq. (97):

ID = 4, + while


4, + 12, +




z;, = Zo + Z12.


Thus, we see that the basic source functions required to calculate bound and unbound production differ only in the term (~i~~ 1I,, 1AB(+)). This term represents final state interaction between the observed constituents. It is the sum of all contributions to the multiple scattering series in which the very last interaction is between the constituents of the observed final state. This final state interaction must be deleted in the computation of bound state productionincluded in the unbound. Except for kinematics that especially emphasize the role of the final state interaction (e.g., at low energies and low relative momenta pi - p2 of the observed pair), this term is unimportant. It is useful to define a source function that neglects this term, %:AB~+&) = (91q32 I ID I AB+j. (154)





uD;AB(pD) ZB S12;AB(&&)

= i/ dxD (27~)-~ f dx dp (27~)-~ WD@, p)


dx, (2?~)-~ dx, (2~r))-~ &AB(Q)&$),


where 6jlzzABis the cross section less final state interaction above formulas we have used


In the

<TO I = j- dx 4 (W-3
PD = PI + P2,


P) <~l~z I,

dx, dx, = dx dxDr,

P = (KPl

- M,P,Y(M,

+ NJ.

These equations say that the source functions 3, giving the spatial probability distribution of the origin of the particles produced in the interaction already contains all the information necessary to compute the bound state production rate. One just convolutes L?with the Wigner density wD describing the bound state. Thus, for example, a Monte Carlo calculation approximating the multiple scattering series, and hence, providing an estimate of 3, without further effort can, provide an approximation to the bound state production rate. If one further approximation, often valid in practice, is made, these relations become even more striking. The dependence of s on p is assumed to be much less rapid than that of wD , which is peaked at p = 0. In this case, one has

%;AB(P1, s

j- dXD' (27T)-3j- dx 1#D(x)[s,,;,,

p2) ZB


; X;,



j- dx, (27~~~dx, (27~)~~ &;2;.,fB(x1,

; x2, $).


The integrands differ only in the substitution

1 #D(x)? = $ dp

of the spatial probability

wD(xl, P)





in one for a factor of (27~)~~in the other. Thus, the ratio ~D:AB(PD)/u12:AB(PDI2, PD/2) is related to the volume from which particles 1 and 2 emanate, and hence, the mechanism of their production.






The multiple scattering expression for inclusive scattering of B by a two-body bound state A is now evaluated in the dilute limit. The methods and results of this formalism may be connected with others in the context of this well known example. The contribution first order in JB, to this process is given by (see Eq. (149)),



Since we are in the laboratory system, p1 + pz = 0. As before

are the relative coordinates and momenta of 1 and 2. The dilute expression for JB1 can be written in the present notation, Eq. (144), as


(fPlP29)B I JBl I 9)lwPB) = CR I ~zx@;ll I @Bl) GBl) ~(xd (277)-31 VB

vl 1 {-%&?l



3 PI)




@,, = X,, , PB1 are the phase coordinates of the Bl center of mass; FBI = xB1, PB1 refer to relative coordinates, specifically xB1 = xB - x1 and
PSI = (MIPB MBk)/(Ml + MB);

&(PB,P1), which really only depends on psi, is the total cross section for B on 1; UB~(PB, PI; PB , Pi), which really on depends on PBl and p;ll, is the differential cross section for scattering into dp;, . Its relation to the usual center-of-mass

differential integrating




cross section is given by Eq. (61). Combining over spatial functions yields - Pm)
PSI) &pB / VB 2 PI) + (TB1(PB', PI)

these equations


s dPl dP dx @71r)-3 w4(x, P) w%,

. 1 vB = --6(I),' v1 1 {-6(Pil PB) j & 1 &4(p)i2


PB , PI))

v1 j &pB,

+ j dP I #A( where now



1 (TB1(PB',





PI = P





The first term gives the contribution to the total cross section due to collisions of B with 1 only. If we add to this the corresponding contribution from 2, the first-order estimate of the total cross section is

The second term in Eq. (163) is the well-known obtained by Goldberger [6], j
dp I $A(p)I" 17 1 UBdpB',

single scattering

term originally





to which one should add the corresponding term for scattering of B on 2. The contributions second order in the jump operators provide shadow corrections to the first order results as well as double scattering. The contribution from the JBzGJBl term in Eq. (149) is f
dXB'(2r)-3 &'1'&'2'& dxB wx4(9)) (%'~'z'~B' I JB,G,JB, 19)1P)2P)B). (166)


over intermediate
Go 1~1, 932,

states and using

.-a> = jam dt I ~)o(vl, t), ~(972, 0, . ..>


where wl(% t) = x + vt, PY




we get (plyz~a I JB&~JB~ I RRFB>

=.r u; dt f dy; dy; d& (P)~P)~P)B I JB2 I Y&V; 3 t) . ..> kmi I JBl I 9)lWB) =.c ccdt s dy$ (F~F~ I JB2 I v,,(Y~ 7 t>> R,(v; >t)> *!~ob,~ -11, f?J;I JBl I g, Q.


Inserting this result into Eq. (166) and integrating over the spatial functions, we get s0
m dt

dpl dpz dp dpz, dx (27r-3 wA(x, p) I * 6(x - v,t + v;;t>{l 6 - vz I %PB + Pz - P;; - Pz) * F&P;; , Pz) %P& - P&J + %B(PB, Pz; P;; 3 PAI
I VB b&B VI 1 a(& 9 PI) s(& + PI PBd PB + PI) aBl(pii 2 Pl'; PB 3 Pd).


The spatial 6 function requires that at some time t after collision of B with 1, the position of B, x1 + vit, coincides with the position of 2, x, + v,t. The only new notation used in Eq. (169) is p;lz for the relative momentum of pi and pz , pbz for that of PB and p2, piI for that of pi and pl and, PSI for that of PB and p1 . The term quadratic in Us represents a shadow correction to the total cross section; the terms linear in uT represent shadow corrections to single scattering; the term quadratic in the differential cross sections represent double scattering. We consider each in turn. The shadow correction to the total cross section given by this expression is - j dp (h--3

Irn dt / VB
, -P) VB 17 -

+ v

I I 1iA(-(vB

V) f,


1 &pB

9 Ph


which must be added to a similar term coming from collisions in the reverse order to get the total shadow correction to second order. This expression simplifies somewhat in the high-energy limit, where it can be compared to the corresponding correction due to Glauber. Using
1 VB f v 1 = LB

UB dt = dz s dp (277Y WA@, P) = I &&)I:





for the bound state spatial density, we get,

Taking into acount that the contribution to the shadow correction coming from the reverse order of collision, one finds that the total shadow correction in the dilute limit is twice that obtained by Glauber [7] in his simple approximation to black sphere scattering. The reason for the difference in these results may be traced to the fact that they each hold true in different nonoverlapping regimes. Our result is accurate in the dilute limit in which the Glauber approximation is invalid. It should be reemphasized at this point that the basic formalism, the Wigner representation of scattering, is not limited to the dilute limit but is susceptible to other approximations that may lead to simple formulas valid in other regimes. The shadow correction to single scattering is composed of two parts. The particle casting the shadow may lie either between the accelerator or between the detector and the single scatterer. In either case one gets a reduction in the single scattering, a unitarity correction since it is merely a manifestation of probability conservation. The shadow correction to single scattering by 1, to be added to the uncorrected term given in Eq. (164) is obtained by straightforward application of the equations just discussed yielding again in the dilute limit,

- Jam dtJ4


]I VI? +

v 1 w4(-(VB

+ v)

t, p)



1 VB

1 w&B


& P>



PB > P>


where pl is again given as in Eq. (164). A similar term reduces single scattering on the other particle. The physical origin of these terms is transparent. See Fig. 3. We might note that the effect of such corrections is quite striking in reducing the backward peak observed in proton-deuteron elastic scattering (~50 % correction) [13]. It may be that a simple formula, such as that derived here (perhaps somewhat refined to take into account the not completely dilute nature of nuclear matter), will be quite accurate in calculating such shadow corrections for many nuclear targets, including the deuteron. This would be especially helpful at those



FIG. 3. This illustrates two phenomena generated by the second-order term of the multiple scattering expansion for scattering of projectile B by a two particle bound state A (see Eq. (149)). When B hits constituent 1 it happens, in this instance, to have momentum x and be displaced from constituent 2 by x. The solid line shows the trajectory of B. Between collisions B, having gotten intermediate momentum pi, displaces itself along vit while constituent 2, since A was initally at rest in the laboratory frame, has momentum -p and has moved -pt. The collision between B and 2 at a time t after the first collision has two consequences: First it reduces the number of (single scattering) events into pi, i.e., particle 2 shadow events shining from particle 1. This phenomenon is described by the first bracketed term in Eq. (172). Second, it increases the number of (double scattering) events into p B. This is described by Eq. (173). Note that the small circles denote the particles positions and the small arrows through them denote their momenta, not spin.

intermediate energiesat which the Glauber approximation begins to break down. We also note that, as before, this formula simplifies considerable when vg is much greater than Y, the Fermi velocity. Finally, the double scattering term obtained from Eq. (169) is j dp (27~)-~dp;; Jrn dt ~)B~w~(-(v;; + v) t,


. / 6 + v 1aB2(pBP2; p;; , -p) 1VB- v 1 o,,(p;;p,; PB, p),

where now momentum conservation gives PI = ps + p - p;; Pi?= p;; - p - PB.


To this must be added the corresponding term from the reversed order of collision to get the total double scattering. We have here a five-dimensional integral since there are two energy conserving a-functions implied in the elastic cross-sections.When uB is large compared to v, we can again effect a considerable simplification by integrating analytically over dp. One is then left with a two-dimensional integral. In either casethis integral. In either casethis integral is especially suited to Monte Carlo integration techniques.






The intent here is the derivation of the general relation between densities and cross sections. Particular cases were derived in Section 5 based on the assumption that the Wigner representative of the density could be formally considered as a probability. Here, the formula is derived along conventional lines. The fact that the same results are obtained in both ways is then further evidence that the aforementioned formal identification of the density is both safe and effective. Let y index possible subsystems and C index possible channels of y. Thus, we write Y=y10yzOf, c = (n 7 yz Y), (4.1)

where each yi is a set containing either a single elementary particle or a bound state of more than one. The possibility of multiple bound states of a yi involves a simple extension of this notation and will be ignored. Let jj denote the compliment of y so that y @ jj = the whole system in question. The channel Hamiltonian for C is denoted as Kc. This is the same notation as in [6]. The total Hamiltonian for y is denoted as H, . The channel and total Hamiltonians are not identical except when C is comprised of a single fragment, i.e., when C = (y). We now define
Hc = Kc + HP


H = Hc + Vc =K+V


where His the total Hamiltonian of the system. Thus, Hc contains the sum of all the kinetic energy operators of the entire system (- K), all interactions amongst elements of 7 (- V,), and all interactions necessary to hold the fragments yi in C together: v, = c V,< . (A.5) The residuum is

V,= v-




The direct product of the plane wave channel eigenstates of C is denoted by IW = I k,,) 0 I k,) 0 ... (A.7)




and the projection Evidently then

operator onto these is wk) = I k&kc I= 0. (A.8) (A.9) and, a unit

[Hc ) sykc)]

Denoting again the target and projectile by A and B respectively normal packet centered about zero momentum by g we obtain
/ gA , PB) = j dpA gcpA) / PA, PB)

(A.lO) (A.1 1)


= 1 epiH 1gA , p$?))]t=o

[0B(27T)-3] oc;AB(kC) = [a/a? (@)gA , pB / eiHt8(kc)


In this expression, the cross section uC;AB has the meaning given in the text with the proviso that C may more generally denote an arbitrary channel with kc = k,, , k,, ,... being the set of channel momenta. The flux factor is z1~(27r-~. Finally, I g, > P2) = -Q,B I g,4 2 PB), (A.13)

where fiAB is the Moller operator for incoming channel AB. Using Eq. (A.4) we get wC:/&c) = (27~)~ i((+)gA
> PB 1 vc'gP(kc) 1 g, , I&+')


+ complex conjugate. One may now expand over intermediate states using the set j kc, k,), which is complete in 7. Taking the limit in which g is infinitely narrow, one obtains, %?~c;AB(kC) = (2~)~ i 1 dk, &K, + x 7-c*& where &= CL asp kp = @a); (k, , kc 1PA, Pl;t)) = %K, + KC = W,

Kc - PB)
, k, ; 0,
PB) + c.c.,

, kc ; 0, PB) 4,


Kc = 1 hi i

- pa) w(k, , kc ;



(A.16) (A.1 7)

(ki, , kc I vc I PA , P%

+ Kc - PA - ps) T,(k,

, kc ; PA , 14.



On the other hand, the integrand in the expression for the flux can be rewritten as

which leads to

= CW3 i j- 4, I ~(k,Wc,
and thence to



> PB)<-



1 %,]I



= j- dx,, (2?~)-~ dx,, (2~)-~ ... (q,, , yrz 1 Ic 1Ml+),

(A.19) (A.20)

1; = --i Wc)r - (~cI)BI

and I AB+)

= I pa,),
&(O, 0) jdXB




, PB>


= (243 j- 4x4 1&A , PB)(- $9, , PB1.

Eq. (A.19) is the result to have been demonstrated.

ACKNOWLEDGMENTS I wish to thank Dr. J. Hornstein and Dr. A. P. Sathe for several useful discussions and for a careful reading of an earlier version of this work.

REFERENCES 1. E. WIGNER, Phys. Rev. 40 (1932), 749. 2. SEWELL, Lectures in TheoreticalPhysics,Vol. 10, p. 289,Interscience, New York, 1967. 3. K. IMRE, E. OZIZMIR, M. ROSENBAUM, AND P. F. ZWEIFEL, J. Math. Phys. 8, 1097. 4. H. W. BERTINI, Phys. Rev. C6 (1972),631. 5. E. A. REMLER AND A. P. SATHE, Ann. Phys. (N.Y.) 91 (197%295. 6. M. L. GOLDBERGER AND K. M. WATSON, CollisionTheory, Wiley, New York, 1964.




7. R. J. GLAUBER, Lectures in Theoretical Physics, Vol. 1, p. 315, Interscience, New York, 1959. 8. E. A. REMLER, unpublished. 9. U. FANO, Rev. Mod. Phys. 29 (1957), 74. 10. J. R. N. MILES AND J. S. DAHLER, J. Chem. Phys. 52 (1970), 616. 11. P. CARRUTHERSAND F. ZACHARISEN,Transport Equation Approach to Multiparticle Production Processes, A. I. P. Conf. Proc., Vol. 23, p. 481, American Institute of Physics, Williamsburg, Va., 1974. 12. H. Mom, I. OPPENHEIM,AND J. Ross, The Wigner Function and Transport Theory, Studies in Statistical Mechanics, Vol. 1, p. 217, Interscience, New York, 1962. 13. E. A. REMLER AND R. A. MILLER, Ann. Phys. (N.Y.) 82 (1974), 189.