Revista Brasileira de Ensino de Fı́sica, v. 36, n.

1, 1310 (2014)
www.sbfisica.org.br

Variable phase equation in quantum scattering

(Equação da fase variável no espalhamento quântico)

Vitor D. Viterbo1 , Nelson H.T. Lemes2 , João P. Braga1

1
Departamento de Quı́mica, Universidade Federal de Minas Gerais, Belo Horizonte, MG, Brasil
2
Instituto de Quı́mica, Universidade Federal de Alfenas, Alfenas, MG, Brasil
Recebido em 3/6/2013; Aceito em 6/9/2013; Publicado em 6/2/2014

This paper presents the derivation and applications of the variable phase equation for single channel quan-
tum scattering. The approach was first presented in 1933 by Morse and Allis and is based on a modification
of the Schrödinger equation to a first order differential equation, appropriate to the scattering problem. The
dependence of phase shift on angular momentum and energy, together with Levinson’s theorem, is discussed.
Because the variable phase equation method is easy to program it can be further explored in an introductory
quantum mechanics course.
Keywords: phase equation, scattering matrix, phase shift.

Este artigo apresenta a dedução e aplicações da equação da fase variável para o caso de um canal no espal-
hamento quântico. Esta abordagem foi apresentada pela primeira vez em 1933 por Morse e Allis e baseia-se numa
modificação da equação Schrödinger para uma equação diferencial de primeira ordem, adequada para o problema
de espalhamento. A dependência do deslocamento de fase com o momento angular e a energia, juntamente com
o teorema de Levinson, é discutida. A equação resultante do método da fase variável é de fácil programação e
pode ser explorado em cursos introdutórios de mecânica quântica.
Palavras-chave: equação da fase, matriz de espalhamento, deslocamento de fase.

1. Introduction will be no need to apply boundary conditions or to cal-

culate Bessel functions. This simplifies the approach
The variable phase method was first introduced in 1933 considerably, presenting a very efficient and straight-
by Morse and Allis [1] who established the basic equa- forward approach to elastic scattering. Numerical ex-
tion for zero angular momentum scattering. In 1949, amples will be considered for model potential energy
Drukarev extended their results in the book ”The The- functions, together with the angular momentum and
ory of Electron-Atom Collision” [2]. In this book, the energy dependence for the phase shift. A discussion of
method was established in a more general form for an- the Levinson theorem will also be carried out.
gular momentum different from zero. A presentation
of this method is given by F. Calogero [3] in his 1967 2. Basic quantum scattering theory
book, although the simplicity of the method is not
fully explored. The present discussion will be based on In a central field collision process, one can seek a solu-
a single channel problem, but the method can be gen- tion of the Schrödinger equation by expressing the total
eralised to several channels [3, 4]. The variable phase wavefunction as partial waves,
method has been used before in several applications, ∞
such as in detection of bound and meta-stable states [5]. ∑ Pl (cos(θ))
ψ(R) = ul (R) (1)
R
l=0
Due to its simplicity, the variable phase method rep-
resents an attractive tool, both for the theoretical and in which Pl (cos(θ)) are the Legendre polynomials. If
numerical understanding of the scattering process. The the interaction between the particles is described by a
objective of the present work is to emphasise this sim- potential energy function Ep (R), it is appropriate to
2
plicity. The original Morse and Allis [1] derivation for define an effective potential, U = 8πh2µ Ep (R) + l(l+1)
R2 ,
zero angular momentum can be adequately adapted for with h the Planck’s constant, µ the system reduced
any angular momentum, as will be discussed. There mass’s, R the scattering coordinate and l the angular
2 E-mail: [email protected].

Copyright by the Sociedade Brasileira de Fı́sica. Printed in Brazil.

1310-2 Viterbo et al.

momentum. Substitution of Eq. (1) into Schrödinger From Schrödinger equation one obtains for the log-
dϕρ (R)
equation then gives for ul (R), derivative wavefunction, Yρ (R) = ϕρ1(R) dR ,
( 2 )
d l(l + 1) 8π 2 µ dYρ (R)
+k −
2
− 2 Ep (R) ul (R) = 0. (2) + k 2 − Uρ (R) + Yρ2 (R) = 0. (6)
dR2 R2 h dR
Using Eq. (5) one can develop
For a given potential energy function one has to find the
solution ul (R). Because this potential energy function −k 2 − k dR
dδ
cos2 (kR + δ(ρ))
2 + k 2 − Uρ (R) + k 2 2 = 0,
goes to zero at large distances, the appropriate bound- sin (kR + δ(ρ)) sin (kR + δ(ρ))
ary condition will be (7)
or (using cos2 x = 1 − sin2 x),
lπ
ul (R) ∝ e−(kR− 2 ) − Se+(kR− 2 ) ∝ sin(kR −
lπ lπ
+ δl ), dδ(ρ) Uρ (R)
2 =− sin2 (kR + δ(ρ)), (8)
(3) dR k
in which S is the scattering matrix, Sl = e2iδl , where which is the variable phase equation for the one-
δl denotes the phase shift. The phase shift, or the scat- dimensional case. This is essentially the Morse and
tering matrix, gives complete information about the Allis deduction but is developed here for the effective
collision process, including the differential and total potential considering angular momentum different from
cross sections. The relation between the phase shift zero. This extremely simple proof will be enough to
and cross section can be found in quantum scattering understand the basic concepts in elastic atomic colli-
textbooks [6, 7]. sion.
The theoretical information about a collision pro-
cess is completely described by the scattering matrix or Equation (8) requires an initial condition to be prop-
phase shift. In a time independent formalism, this scat- agated. To avoid numerical instability, it is convenient
tering matrix is calculated in a three steps procedure, to start integration at a point, R0 , close to the ori-
involving the following: a) initial conditions for the gin. Assuming the wavefunction to be zero at this
wavefunction; b) numerical solution of the Schrödinger point and for zero angular momentum, one must satisfy
equation and c) boundary conditions with Bessel func- u0 (R0 ) = 0. From Eq. (3), it is clear that δ = −kR0 ,
tions at the end point. There are a variety of methods which can be used as an initial condition. Changing
to solve the Schrödinger equation, and a comparison the angular momentum to larger values will shift the
between two common algorithms, log-derivatives and potential energy function to the right, making this ini-
Numerov methods [8], is presented in the literature. tial condition appropriate for any angular momentum.
These numerical procedures are simplified by using an The initial condition is then
important scattering matrix property: the matrix S
δ(R0 ) = −kR0 . (9)
is a ratio of amplitudes, and the overall wavefunction
normalisation is irrelevant, a fact made clear in Eq. The phase shift calculated from Eq. (8) will also
(3). This is an important point to be explored in the carry information about the centrifugal term. To clar-
variable phase approach. ify this point, consider the solution of Eq. (2) for
a very large scattering coordinate, in a region with
Ep (R) = 0. In this case, solutions will be given by
the Riccati-Bessel function, with asymptotic behaviour
3. The variable phase equation of sin(kR − lπ2 ) and cos(kR − 2 ). Consequently, so-
lπ

In the variable phase approach, the potential energy lutions of the Schrödinger equation for zero potential
function is divided into two regions as energy will carry a phase of − lπ2 due to the centrifu-
{ gal contribution. Therefore, the phase for the potential
2
Uρ (R) 0 ≤ R ≤ ρ, U = 8πh2µ Ep (R) can be calculated from the phase shift
U (R) = (4) 2
0 R > ρ. for the potential U = 8πh2µ Ep (R)+ l(l+1)
R2 by subtracting
In an analogous way, the solution is considered in these the centrifugal term contribution, − lπ 2 . The required
two regions as ϕ(R) for 0 ≤ R ≤ ρ and ϕρ (R) for R ≥ ρ. phase shift will be
This method seeks the solution ϕ(R), because the solu- lπ
δ = lim δ(R) + . (10)
tion for R ≥ ρ corresponds to a free particle wavefunc- R→∞ 2
tion conveniently written as Consequently, usage of the variable phase equation can
be summarised by three equations,
ϕρ (R) = α(ρ) sin(kR + δ(ρ)). (5)
dδ
dR = − U (R) 2
k sin (kR + δ),
The amplitude and phase for this free particle solution δ(R0 ) = −kR0 ,
(11)
will carry information about the inner region and, as a δ = limR→∞ δ(R) + lπ
2 .
consequence, must depend on ρ.
Variable phase equation in quantum scattering 1310-3

Implementation of this approach can be performed in given by S = ei2δ = cos(2δ) + i sin(2δ). The results in-
any computer language, for it will involve the numerical dicate that the variable phase method gives essentially
propagation of a first order differential equation. exact answers and can be further explored.

4. Model potential Table 1 - Scattering matrix comparison.

l Scalculated Sexact
A Morse potential energy function [9],
0 -0.2521-i0.9677 -0.2521-i0.9677
1 -0.5164-i0.8563 0.5164+i0.8563
Ep (R) = De (1 − e−α(R−Re ) )2 − De (12) 2 -0.9452-i0.3266 -0.9452-i0.3265
4 0.3851+i0.9229 0.3852+i0.9228
with De =1136 Å−2 , α = 2.4 Å−1 and Re =0.74 Å [10],
which are approximate values for hydrogen-hydrogen Phase shift convergence against the scattering coor-
interaction, will be used to illustrate the method. Units dinate is shown in Fig. 1. If the potential is positive,
for energy are in Å−2 as, for example, in the l(l+1)
R2 term. the phase shift will be negative. From Eq. (8) one
These units are obtained by first calculating energy in may infer the sign of the phase shift, because it is pro-
atomic units and then converting to Å−2 . portional to −Ep (R). If an approximation is used for
This prototype potential energy function is meant to the phase shift derivative, one can write that for Eq.
be a model potential and does not precisely describe the (8), δ(R + h) ≈ δ(R) − U (R) 2
k sin (kR + δ). Then, for a
H2 molecule. In fact, any model potential can be used, positive potential energy function, the phase shift, to-
as long as results are compared with a more precise gether with the initial condition, will give negative val-
calculation. This will be done here by comparing scat- ues. However, if the potential changes sign, the phase
tering matrix values calculated by the variable phase shift will also carry this information, as exemplified in
approach with those calculated by the Renormalized Fig. 1. The potential energy function changes sign at
Numerov method [8, 11]. R = 0.46 Å, exactly at the point at which the phase
shift changes its behaviour. The constant phase shift
5. Results and discussion value for large R, shown in Fig. 1, is also evident from
Eq. (8). In this region, the potential energy function
Propagation of the variable phase equation for dδ
will go to negligible values and, consequently, dR ≈ 0.
k 2 = 25 Å−2 and several values of angular momentum
will be discussed. The initial condition was taken at
R0 = 10−3 Å, but the final integration coordinate has to
be tested against convergence. For example, for l = 0,
the maximum scattering coordinate was Rmax = 10 Å.
For angular momentum different from zero, integration
has to be carried out to large scattering coordinates,
because one must approximately cancel the centrifugal
contribution. The maximum integration point will de-
pend on the angular momentum and can be estimated
by imposing the condition ε = l(l+1) R2 , in which ε is a
small number. For a typical value of ε = 10−4 and
l = 30, one obtains the maximum integration point
at 3000 Å, which is the consequence if Bessel function
boundary conditions are not considered. Figure 1 - Phase shift (full line) and potential energy function in
Numerical integration was performed using a atomic units (dashed line) plotted scattering coordinates. The
Runge-Kutta fifth and sixth order method with variable parameters used in the phase shift calculations were l = 0 and
k2 = 25 Å−2 .
step size, as in Forsythe and Moler [12]. This variable
step size is very important because for low collision en- As the angular momentum is increased, the cen-
ergy, the phase shift will have a step function behaviour. trifugal term will become more important, making the
Additionally, for large scattering coordinates, consider- potential energy function negligible. Hence, the total
able computer time can be saved by using larger step potential for large angular momentum will be approx-
sizes. imately given by the centrifugal contribution. In this
The reliability of the present approach can be in- case, phase shift must go to zero, because the reference
ferred by comparing phase shifts with another more pre- potential in the present formulation is the centrifugal
cise scheme. The scattering matrices calculated from potential. In fact, that is the reason for subtracting the
the variable phase method and the very precise Renor- phase − lπ2 from the computed phase shift. At approx-
malized Numerov method are presented in Table 1. The imately l = 30, the phase shift goes to zero, and no
real and imaginary parts of the scattering matrix are more scattering processes can take place, as shown in
1310-4 Viterbo et al.

Fig. 2. This maximum angular momentum can be esti- angular momentum has to be considered. Numerical
mated from a classical analysis. The maximum impact integration will provide information on the number of
parameter, b, above which there will be no scattering, bound states, and because transitions in infrared spec-
can be estimated to be equal to the potential range, tra are due to vibrational mode excitation, one can infer
Rmax . Because total angular momentum is given by consequences, such as the number of lines, about the in-
L = bk, one has lmax ≈ Rmax k. Considering √ the po- frared spectrum [10]. The results are shown in Table 2.
tential range to be 6 Å one has lmax ≈ 6 × 25 = 30, There is a clear tendency to show that the prototype
as confirmed numerically. Additionally, this maximum molecule can accommodate 14 bound states. In fact, at
angular momentum estimation is important to calcu- the limit for k 2 = 10−4 Å2 , it was found that πδ = 13.98,
late cross section because it will involve a summation confirming this tendency. Thus, the prototype poten-
on phase shifts for different angular momenta. tial represents a molecule with 14 bound states. This
Oscillations in Fig. 2 can also be explored to clar- procedure is the same as the procedure adopted for a
ify the connection between classical and quantum scat- realist potential energy function and was used to detect
tering. In a semiclassical context, the derivative of bound and meta-stable states of rare gas hydrides [5].
the phase shift is proportional to the scattering angle.
For this reason, the maximum in Fig. 2 corresponds Table 2 - Numerical solution of Levinson’s theorem.
to a concentration of trajectories, and consequently, δ
k2
greater intensity in the differential cross section, an ef- π
25 9.7
fect known as the rainbow effect in atomic and molec- 2.5 12.34
ular collision [13]. 0.25 13.42

40
Further theoretical and computational aspects of
the method can be explored. For example, the first
35 Born approximation [6] is a special case of the variable
phase method. For small phase shift values, such that
30 sin2 (kR + δ) ≈ sin2 (kR) is a valid approximation, the
phase differential equation reduces to
25

dδ U (R)
Phase shift

≈− sin2 (kR) (14)

20
dR k
15 which is the first Born approximation in differential
form. Integration of this differential equation gives
10 the∫ usual presentation for this approximation, δ ≈
∞
− 0 U (R) 2
k sin (kR).Thus, usage of the variable phase
5
approach can provide a simple and convincing proof of
the first Born approximation.
0
0 5 10 15 20 25 30 35 40
l

6. Conclusion
Figure 2 - Phase shift convergence for several angular momenta
and k2 = 25 Å−2 .
In contrast with numerical methods to calculate elastic
Levinson’s theorem [6] states that at the limit of scattering that require knowledge of Bessel functions,
zero collision energy the phase shift is a multiple of π, a simple approach based on the variable phase method
was discussed. The algorithm discussed is very sim-
lim δ = nb π, (13)
k→0 ple to implement and allows several important conse-
quences to be explored. Calculation of the scattering
in which nb is the number of bound states that the
matrices were conducted and compared with results
molecule can support. If there are bound states with
obtained using the Renormalized Numerov method.
zero energy and angular momentum different from zero,
Phase shift behaviour as a function of energy and angu-
then Eq. (13) has to be modified to limk→0 δ =
lar momentum was discussed, together with numerical
(nb + 12 )π [5,6]. As confirmed numerically, these bound
examples of Levinson’s theorem.
states with zero energy were not detected here. Equa-
The variable phase method as presented here can be
tion (13) is thus used in the present study.
explored further to calculate other quantum properties.
Levinson’s theorem is a powerful and elegant the-
orem that makes a connection between the continuum
and the discrete states of system. The variable phase Acknowledgment
equation is appropriate to investigate this theorem nu-
merically. Because the energy will be small, only zero We would like to thank CNPq for financial support.
Variable phase equation in quantum scattering 1310-5

References [7] J.N. Murrell and S.D. Bosanac, Introduction to the

Theory of Atomic and Molecular Collisions (John Wi-
[1] P.M. Morse and W.P. Allis, Phys. Rev. 44, 269 (1933). ley, Chichester, 1989).
[2] G.F. Drukarev, The theory of Electron-Atom Collision [8] J.P. Braga, J. Comput. Chem. 10, 413 (1989).
(Academic Press, London, 1965). [9] P.M. Morse, Phys. Rev. 34, 57 (1929).
[3] F. Calogero, Variable phase approach to potential scat- [10] G. Herzberg, Molecular Spectra and Molecular Struc-
tering (Academic Press, New York, 1967). ture (Van Nostrand Reinhold, New York, 1950).
[4] R. Martinazzo, E. Bodo and F.A. Gianturco, Comput. [11] B.R. Johnson, J. Chem. Phys. 67, 4086 (1977).
Phys. Commun. 151, 187 (2003).
[12] G.E. Forsythe, M.A. Malcolm and C.B. Moler,
[5] J. P. Braga and J.N. Murrell, Molecular Physics 53, Computer Methods for Mathematical Computations
295 (1984). (Prentice-Hall, Englewood Cliffs, 1977).
[6] C.J. Joachain, Quantum collision theory (North- [13] J.R. Pliego, J.P. Braga and S.D. Bosanac, Phys. Rev.
Holland, Amsterdam, 1975). A 52, 342 (1995).