Home

Search

Collections

Journals

About

Contact us

My IOPscience

Space-time evolution of a decaying quantum state

This content has been downloaded from IOPscience. Please scroll down to see the full text.
1999 J. Phys. A: Math. Gen. 32 6347
(http://iopscience.iop.org/0305-4470/32/35/311)
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 132.248.209.200
This content was downloaded on 06/10/2016 at 18:58

Please note that terms and conditions apply.

You may also be interested in:

Simulation of atomic ionization followingthe decay of the nucleus
F Kataoka, Y Nogami and W van Dijk
Delay time for a single resonance
M Moshinsky and G Monsivais
Quantum excitations due to the passage of a particle
W van Dijk, K A Kiers, Y Nogami et al.
Survival probability of a single resonance
Gastn Garca-Caldern, Vernica
Riquer and Roberto Romo
Atomic physics with super-high intensity lasers
M Protopapas, C H Keitel and P L Knight
Time dependence and line shape of spontaneous quantum tunneling
B U Felderhof
Photo-induced processes in laser irradiation of a metal surface
Ph Martin, S Vivirito and G Petite

J. Phys. A: Math. Gen. 32 (1999) 63476360. Printed in the UK

PII: S0305-4470(99)02094-6

Space-time evolution of a decaying quantum state

W van Dijk, F Kataoka and Y Nogami
Redeemer College, Ancaster, Ontario, Canada L9K 1J4
Department of Physics and Astronomy, McMaster University, Hamilton, Ontario, Canada
L8S 4M1
E-mail: [email protected]
Received 24 February 1999, in final form 9 July 1999
Abstract. We examine a one-dimensional model of a decaying quantum system, e.g., one that
could simulate decay. A particle is initially confined in a region and leaks out, tunnelling through
a potential barrier. The same model has been examined before, primarily to study decay properties
which are determined by the wavefunction of the particle within the potential barrier. However, the
wavefunction of this model (and of any similar models) outside the potential barrier has not been
obtained explicitly as a function of position and time. We obtain the wavefunction outside as well as
inside the potential barrier by solving the time-dependent Schrodinger equation, and explore various
features of the space-time development of the system. In an early study by Winter it was pointed
out that the probability current just outside the barrier, after a long time, may fluctuate and can be
negative, i.e., inward. We find that under certain circumstances the amplitude of the fluctuations
increases significantly as the distance from the barrier increases. We propose a simple approximate
wavefunction that works well when the decay is very slow. The Gamow wavefunction, commonly
used to describe the -decay process, is not appropriate at large distances. Our wavefunction can
be used anywhere and is normalized in the usual manner.

1. Introduction
The evolution of a decaying quantum state is one of the oldest problems in quantum mechanics.
As a typical case let us consider the decay of a quantum system of the type initiated by
Gamow [1] and by Condon and Gurney [2]. The model assumes that a particle is initially
confined and at a certain time, t = 0, it begins to leak out by tunnelling through a potential
barrier. The model has been studied extensively and various ingenious methods have been
developed. The model is successful in explaining the exponential decay law which has
ample experimental support. More detailed scrutiny revealed interesting deviations from the
exponential law for very small t as well as for very large t. We still feel, however, that our
understanding of the decay problem is far less than complete. This is because, as far as we
know, the wavefunction of the particle of the model over the entire space has not been worked
out explicitly as a function of position and time. All work done so far on the decaying state
focused on the wavefunction at and inside the potential barrier, and very little is known about
the behaviour of the wavefunction outside the potential barrier. The so-called Gamow state
or Gamow wavefunction is often used [1, 3]. The Gamow wavefunction, however, increases
exponentially in magnitude at large distances. Therefore, despite its merits for determining
-particle decay rates, etc, the Gamow wavefunction is not a desirable wavefunction over all
space because it is not square integrable. This difficulty stems from the fact that, in the Gamow
wavefunction, the initial condition that the decay process begins at t = 0 is not fully taken into
0305-4470/99/356347+14$30.00

1999 IOP Publishing Ltd

6347

6348

W van Dijk et al

account [4]. The main purpose of this paper is to find the correct wavefunction at all distances
by explicitly solving the time-dependent Schrodinger equation.
A problem in which such a time-dependent wavefunction at large distances is needed is
the quantum mechanical treatment of atomic ionization caused by decay of the nucleus.
Traditionally this ionization problem has been treated, after Migdal, by regarding the emitted
particle as a point charge that obeys Newtons equation of classical mechanics [5, 6]. In a
fully quantum mechanical calculation, the particle should be treated as a wave that leaks
out of the nucleus. We are aware of only one calculation in which the particle is treated
quantum mechanically [7]. In that calculation, however, the particle is described by means of
a quasi-stationary state with complex energy. Since decay is an intrinsically time-dependent
phenomenon, the validity of such quasi-stationary treatment is not entirely obvious. In this
paper we focus on the decay process itself but in a forthcoming paper we plan a fully quantum
mechanical calculation for a model that simulates the atomic ionization process. In order to do
such a calculation we have to be able to first solve the time-dependent Schrodinger equation
for a decaying system.
Instead of a detailed realistic calculation we consider a simple one-dimensional model.
This is the same model that has been examined by a number of people, in particular,
by Petzold [8] and Winter [9], and more recently by Garca-Calderon et al [1012].
Although simple, the model illustrates many interesting features as well as difficulties that
are characteristic of the decay process in quantum mechanics. The model consists of a particle
which is subject to an infinite repulsive wall at x = a and a repulsive barrier represented by
a -function potential at x = 0. The particle is initially confined in the region of a < x < 0
and begins to leak out at t = 0 tunnelling through the barrier. As shown by Petzold [8],
by Winter in more detail [9] and also by Garca-Calderon et al [1012], the time-dependent
wavefunction (x, t) can be expressed in the form of an integral with an integrand that can be
written down explicitly. The integrand, however, is highly oscillatory with large amplitude. It
is difficult to carry out the integration numerically except for small values of x. Probably for
this reason, (x, t) of the model has not been obtained explicitly outside the barrier, x > 0.
Of course, the wavefunction inside the potential barrier contains a significant amount of
information on the decay process. Winter examined (0, t) and the associated current j (0, t)
at the barrier. He illustrated small deviations from the exponential decay law for extremely
small t and extremely large t. A particularly remarkable and curious feature that he found is
that j (0, t) fluctuates as a function of t in a certain large t region. It can be negative, meaning
that the current can flow inward. It would be interesting to see how j (x, t) behaves for x > 0.
In section 2 we set up the model and briefly summarize some of Winters results. We
then extend Winters calculation by carrying out the integral for (x, t) for larger values of
x. We have been able to do this only to a certain limit. We discuss the nature of the difficulty
that is encountered. We show that the amplitude of fluctuations that Winter found for j (0, t)
becomes much larger as x increases. In section 3 we integrate the time-dependent Schrodinger
equation numerically and obtain (x, t) in the entire space. In section 4 we propose a simple
approximate wavefunction that works well when the decay process is very slow. The conclusion
and a discussion are given in section 5. The appendix enlarges on the negative local current
density which is present in the model.

2. Model
Consider a one-dimensional, non-relativistic model, with a particle of mass m in a potential
that consists of an infinite repulsive wall at x = a and a repulsive -function potential at

Space-time evolution of a decaying quantum state

x = 0 [812],

(
V (x) =

(g/2m)(x)

x < a
x > a

6349

(1)

where a > 0 and g > 0. We use units such that h

= 1, throughout. In numerical illustrations,
we further set a = 1 and 2m = 1. Let us follow Winters elementary approach rather than the
Laplace transform method used in [8, 1012].
Since V (x) has no attractive part, there is no bound state. The spatial part of the
wavefunction of the stationary continuum state of energy E = k 2 /2m can be written as

1 sin(ka + k ) sin k(x + a)

a < x < 0

sin ka

(2)
k (x) =
1

x>0
sin[k(x + a) + k ]

where the phase shift k is given by [13]



ka
k = ka + tan1
G + ka cot ka

G = ga.

(3)

The quantity G is a dimensionless constant. It is understood that the wavefunction always

vanishes for x 6 a. We normalize the stationary wavefunctions in the usual manner [14] as
Z
k (x)k0 (x) dx = (k k 0 ).
(4)
a

For the strength of the -function potential at x = 0, Winter took G = 6 and 20. We,
however, take G = 6 and 100. In this way we can see the contrast between fast and slow
decay processes more clearly. The phase shift k may go through /2 (mod ) upwards for a
certain value of k or E. This means that there is a resonance at that energy. Such a resonance
corresponds to one of the poles of the integrand that we examine shortly. When G = 6 there
is only one resonance due to a pole labelled by = 1. When G = 100 there are 15 resonances
with = 1, 2, . . . , 15.
We are interested in the time-dependent wavefunction that is subject to the initial condition
(p
2/a sin(n x/a)
a < x < 0
(5)
(x, 0) =
0
0<x
where n = 1, 2, . . . . In the illustrations that follow we take n = 1. There is no particular
difficulty in taking n = 2, 3, . . . , or a combination of components with different values of
n. The function (x, 0) is normalized. The expectation value of the energy in this state is
(n/a)2 /2m. The wavefunction (x, t) at a later time t can be expressed as a superposition
of stationary states. Winter writes it as
Z
f (q) dq
(6)
(x, t) =
0

where

2 eiq T q sin q[q sin q( + 1) + G (x) sin q sin q ]

.
f (q) = 2n
a
(q 2 n2 2 )(q 2 + Gq sin 2q + G2 sin2 q)
2

(7)

We have used dimensionless variables defined by

q = ka

T = t/(2ma 2 )

= x/a.

(8)

6350

W van Dijk et al

Figure 1. The integration contour of equation (9) in the complex q-space.

The is Winters l. Petzold obtained a different but equivalent expression for (x, t) by means
of the Laplace transform [8]. Garca-Calderon et al [1012] also used Laplace transforms. We
find Winters formula somewhat simpler.
To evaluate the wavefunction (x, t) of equation (6) is not easy. To our knowledge such
a wavefunction has not been obtained except for the special case of x = 0 which was done
by Winter [9]. He calculated (0, t) and the probability current j (0, t) at the barrier and
examined the nature of the decaying system. We are interested in (x, t) and j (x, t) for
x 6= 0, in particular, for large values of x. Following Winter we try to evaluate the integral of
equation (6) by using a closed contour in the complex plane, as shown schematically in figure 1.
2
2
2
If we replace q with z = sei , the factor eiq T in equation (7) becomes eis T cos 2 es T sin 2 .
This factor tends to zero as s if 0 > > /2. The integral along the arc becomes
negligible in this limit. Therefore, the integral for the wavefunction can be reduced to the
integral along the line N and contributions from the poles enclosed by the contour,

Z
X
f (z) dz + 2 i
(residues) .
(9)
(x, t) =
N

The factor (z2 n2 2 ) of the denominator of f (z) does not give rise to any poles. This
is because of the sin z of the numerator. The poles of f (z) are determined by
z2 + Gz sin 2z + G2 sin2 z = 0.

(10)

If G  1, sin z 0 and hence z where = 1, 2, . . . , or more accurately



1
i
1
z 1 + 2 2
G G
G

(11)

where we have ignored terms like 1/G3 , etc. Only the poles with a negative imaginary part
lie inside the contour. The energy E and width 0 of the resonances are determined by
z 2 /2ma 2 = E i0 /2. If G  1, we obtain
E



1  2
3
2
1 + 2
2m a
G G

1 4( )3
.
2m G2 a 2

(12)

Space-time evolution of a decaying quantum state

6351

Among the poles, the z with = n contributes dominantly to the integral. If we single
out this pole contribution, we obtain
r


zn (x + a)
zn x
zn x
2 iEn 1 0n t

2
e

cos
a < x < 0
sin

a
Ga
a
(13)
(x, t) r a
 z 

2
1

n
iE

0
t
iz
x/a

e n 2 n
e n
x>0
a
G
which is essentially the Gamow wavefunction for the -function potential model. The
amplitude of this (x, t) decays exponentially in time. Outside the barrier, the wavefunction
is an outgoing wave. For a given t the amplitude grows exponentially as x increases. This
is because of the imaginary part of zn . This wavefunction is not normalizable in the usual
manner over the entire coordinate space. Other potential models yield Gamow wavefunctions
with similar properties [4].
For G = 6, expansions like equations (11) and (12) are not very accurate. A more accurate
estimate gives 0 /E 0.13 which is not very small. For G = 100, the expansions in 1/G
works well. The resonances are sharp as can be seen from 0 /E 4/G2 (1.3103 ).
The width increases as increases. In extending Winters calculation for the wavefunction,
we have to determine the poles with great accuracy. This can be done as follows. Taking the
real and imaginary parts of equation (10) we obtain two coupled equations. The problem is
then reduced to finding intersections of two curves in the complex z-plane [15]. Starting with
the approximate solution (11), we can solve the coupled equations by iteration to any desired
accuracy.
2
2
For the line N it is convenient to choose = /4 so that eiz T = e2s T . This factor
appears to make the integrand decay rapidly for finite t. However, the integrand also contains
the factors e s , cos s, sin s, as well as a factor that varies slowly with s and falls off as 1/s 2
for large s. The integral cannot be done analytically. It converges but its numerical integration
provides a challenge. For large values of x, the wavefunction, as we will show in the next
2
section, is very small. However, when T = 2 and = x/a = 50, for example, e2T s + s has
a maximum value of 5 1067 . With values of the integrand this large, numerical quadratures
will not be able to give results of small values of the integral because of the limited precision
normally provided by computers. For values of = x/a up to about 20 (when T = 2), any
reasonably sophisticated quadrature or Fourier transform routines give reliable results. But it is
difficult to go further. It should be noted that to obtain a reliable wavefunction, the contributions
due to the poles, which blow up at large x, must be combined with the contribution due to the
line integral along N.
In figure 2, we show the modulus of the wavefunction |(x, t)| for G = 6 that we
obtained by using equations (6) and (9). In this figure and in dealing with numerical values
in the following, we take units such that a = 1 and 2m = 1 as we said below equation (1).
Then we have = x and T = t. The wavefunction of figure 2 is for t = 2. It is not reliable
beyond x 20. Figure 3 shows |(x, t)| for G = 6 and t = 2 that we obtained by solving
the time-dependent Schrodinger equation numerically as we will explain in the next section.
This wavefunction agrees very well with that of figure 2 up to x 20.
One of the very interesting features that Winter obtained is the oscillations in the probability
current, for G = 6 at x = 0 around t = 10. The current is negative at times. He found similar
oscillations for G = 20. In figure 4 we reproduce Winters result for t = 012, and in figures 5
and 6 we obtain the probability current profile for the time intervals from t = 715 when x = 3
and 10. Surprisingly, the oscillations not only persist but their amplitudes (and their maximum
negative values) increase by orders of magnitude as x increases. We also confirm Winters
results using the numerical method described in section 3. We are therefore convinced that the

6352

W van Dijk et al

Figure 2. The modulus of the wavefunction |(x, t)| for G = 6 and t = 2, obtained by using
equation (6). Units are such that h
= 1, 2m = 1 and a = 1. The wavefunction is reliable up to
x 20, beyond which the results are marred by huge rounding errors.

Figure 3. The modulus of the wavefunction |(x, t)| for G = 6 and t = 2, obtained by solving
the time-dependent Schrodinger equation numerically (section 3). The units are the same as in
figure 2. The dotted curve is that of figure 2. The two curves overlap for x < 20.

negative currents are physical and not due to numerical uncertainties. We return to this point
in section 3 where we find out more about the details of the wavefunction.
3. Integration of the time-dependent Schrodinger equation
In order to obtain the wavefunction of the decaying system let us solve a difference-equation
version of the time-dependent Schrodinger equation. Such calculations have been done for
scattering systems [16, 17], but not for a decaying state. We assume that space and time are

Space-time evolution of a decaying quantum state

6353

Figure 4. The current at the barrier x = 0, for G = 6, from t = 012. This was obtained by
Winter [8]. The units are the same as in figure 2.

Figure 5. The current at x = 3 for G = 6 from t = 716. The units are the same as in figure 2.

discretized with small intervals of 1x and 1t, and (x, t + 1) and (x, t) are related by
(x, t + 1t) =

1 2i H 1t
1 + 2i H 1t

(x, t).

(14)

Here H is the Hamiltonian which we interpret as

H =

1 2
D + V (x)
2m

(15)

where D 2 is the operator such that

D 2 (x, t) =

1
[(x + 1x, t) 2(x, t) + (x 1x, t)].
(1x)2

(16)

6354

W van Dijk et al

Figure 6. The current at x = 10 for G = 6 from t = 716. The units are the same as in figure 2.

The wavefunction can be expressed as an array of i,n (xi , t), where xi are the discretized
points with equal interval 1x. Then equation (14) can be put into a form of a system of linear
equations for i,n+1 = (xi , t + 1t). We do this by operating 1 + 2i H 1t on equation (14)
from the left.
Transformation
P (14) is unitary and consequently the normalization of the wavefunction
in the sense of i |(xi , t)|2 1x = 1 is exactly maintained. Let us add that if one uses
Simpsons rule rather than the trapezoidal summation formula, the normalization may not be
exactly maintained. Furthermore, the expectation value of the Hamiltonian,
X
(xi , t)(D 2 + V )(xi , t)1x
(17)
hH i =
i

remains also constant for all t. This is so even if 1t and 1x are not small. We have also verified
that the discretized version of the Ehrenfest relation hpi = m dhxi/dt is valid for sufficiently
small 1x and 1t. For the initial wavefunction
we take the discretizedPversion of (x, 0) of

equation (5) with the same factor of 2/a. It is exactly normalized as i |(xi , 0)|2 1x = 1.
As is the case with wavepacket scattering [16] a higher-order formula of D 2 is not required to
give sufficiently accurate results.
For 1x and 1t in numerical illustrations, we take 1x = 0.01 and 1t = 0.005. We
replace the -function potential at x = 0 with a square barrier of width 1x and height G/1x.
Effects of finite meshes are almost invisible in the figures that we show. For t we start with
t = 0 and go up to t 50. In order to have a feel for the space-time scale, let us note that the
radius of the model nucleus is a = 1, the decay half-life is 1/2 = ln 2/ 0 1.08 for G = 6
and 1/2 56 for G = 100. For the range of x, we take [1, 999]. The xmax = 999 is large
enough such that its finiteness has no discernible effect on our numerical solution.
We have already shown |(x, t)| for G = 6 and t = 2 in figure 3. Figure 7 shows
|(x, t)| for G = 6, t = 5, 10, 15, 20, 25, and 30. Figure 8 shows the same quantities as those
of figure 7, except that G = 100 this time. In figure 8 we can clearly see a wavefront structure
of the wavefunction. The wavefront proceeds with speed v = . We can also see small
crests proceeding with greater speeds of 2 and 3. These small crests can be understood as
follows. When the initial wavefunction (x, 0) is written as a superposition of stationary states
k (x), in addition to the main contribution from the resonance of n = 1, higher resonances of

Space-time evolution of a decaying quantum state

6355

Figure 7. The modulus of the wavefunction |(x, t)| for G = 6 and t = 0, 5, 10, 15, 20, 25, and
30. The units are the same as in figure 2.

n = 2, 3, . . . also contribute. In the case of G = 6 the structure of the wavefunction is more

complicated.

Figure 8. The modulus of the wavefunction |(x, t)| for G = 100 and t = 0, 5, 10, 15, 20, 25,
and 30. The units are the same as in figure 2.

In figures 46 we showed the current as a function of time for G = 6 at x = 0, 3 and 10.

We noted that the amplitude of the fluctuations is larger for larger values of x. This can be
related to the wavefunction shown in figure 7. Note that the major peak at x 50 in the curve
for t = 10 is followed by many small peaks with about the same intervals. Every time such a
peak passes through a point, the current at that point increases. The height of the peak is larger
for larger x. Figure 7, by itself, does not show that the current is negative at certain points.
This could be done by comparing the wavefunction at t and the one at a slightly different time.
To shed further light on the current fluctuations, we plot the current at t = 8.25 as a
function of x in figure 9. The curve for the current in the main part of the graph is calculated
using the numerical solution of the Schrodinger equation. The local fluctuations of the current
are due to the fact that x has been discretized with a finite interval 1x. We note that the
current follows the probability density (dotted curve), i.e., they have local extrema at the same
values of x. The inset shows the analytical calculation (of section 2) of the current for smaller
values of x than those of the main peak. This shows that the current is negative at its minimum
values, but only slightly so, compared with the overall amplitude of the current. The fact that
the current has negative values locally does not contradict the idea that the overall wave pattern

6356

W van Dijk et al

Figure 9. The current density j (x, t = 8.25) as a function of x is shown as the solid curve, when
G = 6. The dotted curve is the probability density |(x, 8.25)|2 . The inset gives an expanded view
of the current calculated by the method of section 2 in the region where it has negative minima.

is travelling outward. The negative currents are very small and occur where the amplitude of
the wave is already a small fraction of the main peak and at times of the order of ten half-lives
of the decaying system. See the appendix in this connection. The amount by which the current
becomes negative increases as x increases up to the second last local minimum before the main
peak of the wave pattern. Let us add that the current exhibits fluctuations also at very short
times. This is related to the part of the wavefunction that propagates ahead of the main peak.
All these calculations can be repeated using a Gaussian potential in place of the -function
potential. We find very little difference in the results if a Gaussian with a width of 0.1 at the
nuclear boundary is used starting with the same initial state. This more realistic model of
the Coulomb barrier does not affect the features discussed in this paper.
4. Approximate wavefunction for x > 0
Since we wish to eventually obtain a wavefunction for the decaying system that can be used in
the quantum mechanical calculation of the ionization probability due to the emitted particle,
we propose an approximate form of the wavefunction for x > 0. Such a wavefunction is
needed since it is difficult to integrate numerically the time-dependent Schrodinger equation
for the particle to distances comparable to the size of the atom. We are particularly interested
in such a wavefunction appropriate to a long-lived emitter.
Let us consider a situation such that the particle inside the potential barrier is in a very sharp
resonance state, the decay rate is very small and the exponential decay law holds. We assume
that the probability distribution flows outwards with a constant speed v, which is also the speed
of the emerging particle. Then the probability density (0, t) and current j (0, t) just outside
the barrier are related by j (0, t) = v(0, t). The current is equal to (d/dt)e0t = 0e0t ,
which leads to
0
(18)
(0, t) = e0t (t)
v
where (t) ensures that there was no leakage of the probability before t = 0. If we assume

Space-time evolution of a decaying quantum state

6357

that the probability flows out uniformly with speed v, we obtain


x
0
x
(19)
(x, t) = e0(t v ) t
v
v
R
where it is understood that x > 0. This (x, t) is normalized as 0 (x, t) dx = 1 e0t . If
there are a large number, N0 , of decaying samples at t = 0 and all at x = 0 and if one counts
the number of emitted particles detected at a certain position x from time t to t + 1t, one will
find it to be N0 (x, t)v1t.
For the wavefunction outside the barrier it would be reasonable to assume, apart from a
possible additional phase factor,
p
(20)
(x, t) = (x, t)eiEt eikx
where E = k 2 /2m and k = mv. This wavefunction is normalized to 1 e0t and
conforms to j (0, t) = 0e0t . By substituting this wavefunction into the Schrodinger equation
i/t = (1/2m) 2 /x 2 for 0 < x < vt, we find that we need additional phase factors.
In this way we arrive at
r
x
0 i 0 (t x ) i[E 1 ( 0 )2 ]t ikx 
2m 2v
e t
e e 2 v e
(21)
(x, t) =
v
v
where the constant phase factor ei is related to the choice of the initial wavefunction. Keeping
one further term of the expansion in 1/G of the approximation of 0 in equation (12), we obtain
the expressions for v and 0 which we used to evaluate the wavefunctions shown in figure 10,
i.e.,
p
(22)
v 2 1 2/G
and



1
4 3
(23)
0 2 1
G
G
where a = 2m = 1. It should be noted that equation (21) is an exact solution of the timedependent Schrodinger equation for 0 < x < vt.
Figure 10 compares for G = 100, the modulus of the approximate wavefunction and
the one obtained by solving the time-dependent Schrodinger equation numerically. The
approximate wavefunction works well for G = 100. However, it does not give a good
representation of the wavefunction for G = 6. Recall that for G = 6, 0/E 0.13 which is
not very small. It is clear that we cannot assume a uniform speed v in this case. In contrast to
this, for G = 100, 0/E 1.3 103 . As we point out in section 5, the value of 0/E is much
smaller in the actual -decay processes. The wavefunction (21) will work even better in such
situations.
Figure 10 shows only the modulus of the wavefunction. For the phase of the wavefunction,
if we set ei = 1, the phase of the approximate wavefunction for G = 100 becomes virtually
indistinguishable from that of the wavefunction obtained by solving the time-dependent
Schrodinger equation. The reason for this choice of is the following. The (x, 0) is negative
just inside the barrier. When it leaks out through the barrier the wavefunction remembers its
sign, that is, Re (1x, 1t) < 0.
We have obtained the approximate wavefunction for outside of the potential barrier. For
the inside, a first approximation is
(x, t) = ei(E 2 0)t (x, 0)
i

for

a <x <0

(24)

where (x, 0) is that of equation (5). We have confirmed that this is a good approximation
when G  1. This (x, t) for the inside together with that of equation (21) for the outside

6358

W van Dijk et al

Figure 10. The absolute value of the approximate wavefunction (x, t) of equation (20) for
G = 100 at t = 20 plotted as a solid line. The modulus of the actual wavefunction (x, t) as one
of the curves in figure 8 is plotted as the dashed curve.

is correctly normalized to unity. The (x, t) of equation (24), however, vanishes at x = 0.

Therefore, there is a slight mismatch at x = 0 between the inside and outside parts of this
wavefunction.
5. Conclusion and discussion
We have examined a one-dimensional model that simulates, e.g., the -decay process. A
particle is initially confined within a nucleus and begins to leak out at t = 0 tunnelling
through a potential barrier. We were interested in obtaining the wavefunction of the model at
all distances. We first attempted to extend Winters calculation for the wavefunction. We were
only able to do so up to a certain distance from the nucleus. We examined the current outside
the potential barrier. We found that the fluctuations of the current that Winter discovered
become much larger as the distance from the potential barrier increases. These fluctuations
can be understood in terms of the fluctuations of the wavefunction preceding and following
the main peak. We then obtained the time-dependent wavefunction (x, t) in the entire space
by solving the time-dependent Schrodinger equation numerically. As far as we know this is
the first time that such a wavefunction has been obtained explicitly. When the potential barrier
is very strong and the decay rate is very low, the (x, t) obtains a simple structure. It can be
well represented by the approximate wavefunction (x, t) which we obtained in a heuristic
manner. Unlike the commonly used Gamow wavefunction, this wavefunction (and of course
the solution of the time-dependent Schrodinger equation) is normalized in the usual manner.
Our approximate wavefunction is expected to be useful in quantum mechanical calculations
of the atomic ionization process caused by nuclear decay.
For the strength of the potential barrier that controls the decay rate, we considered two
cases, G = 6 and 100. For the main resonance (of n = 1) that is involved, 0/E 0.13 for
G = 6 whereas 0/E 1.3 103 for G = 100. These choices of the value of G represent
two typical situations: the resonance is rather broad in one and is very sharp in the other. In
the actual -decay processes, the ratio 0/E is much, much smaller. As an example let us

Space-time evolution of a decaying quantum state

6359

consider Po212 as the parent nucleus. Its decay rate is the fastest among the usually listed
-decay examples, e.g., see [18]. The energy of the emitted particle is E = 8.95 MeV and
0 = 2.31 106 s1 . This means the dimensionless ratio of 0/E = 1.70 1016 . Note that in
natural units (c = 1, h
= 1), 1 s1 = 6.58 1022 MeV. The ratio can be reproduced by our
model with G = 3.7 108 . When G is this large, the wavefunction of equation (21) works
extremely well. It will be very useful as an approximate wavefunction for the particle in
examining atomic ionization caused by decay.
The reason why we took small values of G, which are unrealistic for decay, in our
numerical illustrations is the following. If G is very large and accordingly the decay is very
slow, we have to integrate the Schrodinger equation to a very large value of t before we can
see anything interesting. This is technically difficult. The quantity 0 which is proportional
to the reciprocal of the half-life, however, is a power series in 1/G starting with 1/G2 . Thus
for very large G, the half-life scales as G2 . We therefore believe that G = 100 is sufficiently
large to illustrate the essential features of cases of very strong G.
Acknowledgments
We would like to thank Dr David Kiang for many inspiring discussions on the subject over the
years. We also thank Dr Fred Kus for helpful suggestions regarding some numerical aspects
of the paper. This work was supported by the Natural Sciences and Engineering Research
Council of Canada.
Appendix
In this appendix we illustrate a situation such that, even if the expectation value of the
momentum is positive, the current density can be locally negative. This phenomenon was
first observed in connection with decay problem by Winter [9] and is corroborated by the
results of this paper.
Let us start with a free Gaussian wavepacket with its centre at rest at the origin. Its
wavefunction is [19]




x 2
it 1
1/4
(A.1)
exp
+
0 (x, t) = (2)
2m
42 + 2it/m
where (1x)t=0 = . The current density is given by
xt
j0 (x, t) = 2 4 2 |0 (x, t)|2 .
(A.2)
m +t
The expectation value of j0 (x, t) is zero.
Next we let the wavepacket move with a constant speed v > 0. The wavefunction and
current density become
1

(x, t) = 0 (x vt, t)ei(mvx 2 mv t)

j (x, t) = v|(x, t)|2 + j0 (x vt, t).

(A.3)
(A.4)

In (A.4) the second term can be negative. The first term, however, normally dominates so that
j (x, t) is positive.
A more interesting situation obtains if we assume that (x, t) is a superposition of two
wavepackets,
(x, t) = c1 1 (x, t) + c2 2 (x, t)
1 (x, t) = (x b, t)

2 (x, t) = (x + b, t).

(A.5)
(A.6)

6360

W van Dijk et al

The two wavepackets are separated by distance 2b. They both move with speed v. For
simplicity let us assume that c1 and c2 are both real and
|c1 + c2 |  1.

(A.7)

It then follows that (x, t) vanishes in the vicinity of the midpoint (x = vt) of the two
wavepackets. The current density at x = vt turns out to be


(c12 c22 )bt
2
(A.8)
j (x = vt, t) = (c1 + c2 ) v
|(x, t)|2 .
4m2 4 + t 2
If we put c1 = c +  and c2 = c +  then (c1 + c2 )2 = 4 2 and c12 c22 = 4c. If |/c|  1,
then the second term in the square brackets of (A.8) can become more important than the
first, and j (x, t) can become negative. Note that the negative current density occurs when the
probability density almost vanishes.
We have chosen c1 and c2 such that they are approximately equal in magnitude. If we
choose c1 larger than c2 in magnitude, the point at which (x, t) vanishes is shifted towards
the second wavepacket. In the vicinity of that point, j (x, t) can become negative.
References
[1] Gamow G 1928 Z. Phys. 51 204
Gamow G 1928 Z. Phys. 52 510
Gamow G and Critchfield C L 1949 Theory of Atomic Nucleus and Nuclear Energy-Sources (Oxford: Clarendon)
ch 6
[2] Condon E U and Gurney R W 1928 Nature 112 439
Condon E U and Gurney R W 1929 Phys. Rev. 33 127
[3] Weinberg A M 1952 Phys. Rev. 20 401
Hokkyo N 1965 Prog. Theor. Phys. 33 1116
Garca-Calderon G and Peierls R 1976 Nucl. Phys. A 265 443
[4] Bohm A, Gadella M and Mainland G B 1989 Am. J. Phys. 57 1103
[5] Migdal A 1941 J. Phys. (USSR) 4 449
[6] Levinger J S 1953 Phys. Rev. 90 11
[7] Revai J and Nogami Y 1992 Few-Body Syst. 13 75
[8] Petzold J 1959 Z. Phys. 155 422
[9] Winter R G 1961 Phys. Rev. 123 1503
[10] Garca-Calderon G 1992 Symmetries in Physics ed A Frank and K B Wolf (New York: Springer) p 252
[11] Garca-Calderon G, Loyola G and Moshinsky M 1992 Symmetries in Physics ed A Frank and K B Wolf (New
York: Springer) p 273
[12] Garca-Calderon G, Mateos J L and Moshinsky M 1993 Rev. Mex. Fis. (Suppl 2) 39 76
Garca-Calderon G, Mateos J L and Moshinsky M 1995 Phys. Rev. Lett. 74 337
Garca-Calderon G, Mateos J L and Moshinsky M 1996 Ann. Phys., NY 249 430
[13] Massmann H 1985 Am. J. Phys 53 679
[14] Brownstein K R 1975 Am. J. Phys. 43 173
[15] Onley D and Kumar A 1991 Am. J. Phys. 59 562
[16] Goldberg A, Schey H M and Schwartz J L 1967 Am. J. Phys. 35 177
[17] Bonche P, Koonin S and Negele J W 1976 Phys. Rev. C 13 1226
Press W H, Flannery B P, Teukolsky S A and Vetterling W T 1988 Numerical Recipes in C (Cambridge:
Cambridge University Press) pp 6613
[18] French A P and Taylor E F 1978 An Introduction to Quantum Mechanics (New York: W W Norton and Co Inc)
p 407
[19] Schiff L I 1968 Quantum Mechanics 3rd edn (New York: McGraw-Hill) ch 3